diff --git a/.gitattributes b/.gitattributes index 45e9324bd..b76430b71 100644 --- a/.gitattributes +++ b/.gitattributes @@ -8,3 +8,5 @@ # Methodology contracts, receipts and the decision record are bound by SHA-256. docs/uk_replay/contracts/** -text docs/uk_replay/HISTORICAL_DECISIONS.md -text +# Page extracts vendored beside a source PDF are SHA-256 pinned too. +sources/*/raw/*.txt -text diff --git a/app/public/data/lanes.json b/app/public/data/lanes.json index cdafe9685..cb4e6cf39 100644 --- a/app/public/data/lanes.json +++ b/app/public/data/lanes.json @@ -4,7 +4,7 @@ "country": "US | UK | BE | NZ \u2014 the model instance the lane feeds (issue #42)", "note": "Repo-committed status feed (issue #7); graduate to Supabase when lane count makes commits noisy. Every lane writes its transition here." }, - "updated": "2026-10-22", + "updated": "2026-10-25", "lanes": [ { "id": "urban-sotsn", @@ -599,6 +599,17 @@ "running": false, "updated": "2026-10-22", "note": "2 claims (ONS September 2026 CPI, OBR's March 2026 forecast of it) and 17 PE results, 15 of them on the HM Treasury and Resolution Foundation uprating claims; parameter-tree read-back on policyengine-uk 2.147.2" + }, + { + "id": "uk-childcare-100k", + "source": "CenTax + Centre for British Progress", + "area": "\u00a3100,000 childcare limit (30 funded hours, Tax-Free Childcare)", + "mode": 2, + "country": "UK", + "stage": "computed", + "running": false, + "updated": "2026-10-25", + "note": "22 of 36 claims computed on the certified pair (populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z, policyengine-uk 2.89.2) \u2014 CenTax Table 4.2 (24 cells: static cost comparable, net cost concept_mismatch) and CBP's technical appendix (12 abolition cells, concept_mismatch); PolicyEngine's published dashboard figure kept as history; pe_gap on the central static rows (microcosm#1126)" } ] } diff --git a/app/public/data/populations.json b/app/public/data/populations.json index 64b22cb3a..c3e16f1ff 100644 --- a/app/public/data/populations.json +++ b/app/public/data/populations.json @@ -1,9 +1,9 @@ { - "built": "2026-10-22", + "built": "2026-10-25", "note": "Non-Urban populations exported from scorecard.db: the populace reform-validation registry (issue #20) plus the compute campaign's attached comparisons (US: TPC/CPSP/PWBM/CBO/JCT; UK: the HMRC ready-reckoner family and the complete OBR Policy measures database, including historical HM Treasury scores; BE: JRC EUROMOD-BE claims with demo-grade Axiom concept-mismatch attachments and the 2026 PIT-reform scorekeeper claims; NZ: official budget scores). Claims explicitly opted into pre-result display remain in the feed with latest status not_computed and empty history. Recorded result statuses and calibration relationships are verbatim; nothing here is a pass/fail grade.", "summary": { - "claims": 29735, - "multi_release_claims": 270, + "claims": 29771, + "multi_release_claims": 274, "by_source": { "aspe_welfare_indicators": 14, "az_admin": 1, @@ -16,7 +16,8 @@ "census": 104, "census_pep": 1, "census_stc": 88, - "centax": 80, + "centax": 104, + "centre_for_british_progress": 12, "co_admin": 3, "co_fiscal_note": 1, "cour_des_comptes": 1, @@ -108,17 +109,17 @@ "wi_admin": 1 }, "by_latest_status": { - "comparable": 57, - "concept_mismatch": 17, + "comparable": 63, + "concept_mismatch": 33, "constructed": 1669, - "not_computed": 27103, + "not_computed": 27117, "pe_gap": 889 }, "by_sign_relative_to_pe": { "magnitude": 14, "opposite": 489, - "same": 28963, - "unknown": 269 + "same": 28987, + "unknown": 281 }, "by_country": { "BE": { @@ -136,11 +137,11 @@ "multi_release_claims": 0 }, "UK": { - "claims": 29174, - "with_result": 2179, - "with_value": 1194, - "without_result": 26995, - "multi_release_claims": 0 + "claims": 29210, + "with_result": 2201, + "with_value": 1216, + "without_result": 27009, + "multi_release_claims": 4 }, "US": { "claims": 540, @@ -25924,44 +25925,43 @@ } }, { - "claim_id": "2a18eefc3c44336f3a17", + "claim_id": "68a9d11c877d33191133", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "extensive_margin", "country": "UK", - "denominator": "additional tax from the reform", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "farm estates with an APR/BPR claim covering less than 20% of the estate (likely passive investors)", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "2%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.02, + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.02, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -25983,45 +25983,43 @@ "diagnosis": null }, { - "claim_id": "392f8284d4c31f166150", + "claim_id": "ae9f698e34e48978505d", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "range_low", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "denominator": "additional tax from farm estates", - "estimate_kind": "range_low", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "impacted estates worth over \u00a35 million", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Over 80%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.8, + "reform_key": "db11caf3e07ba4d1", + "external_value": 150000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.8, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 150000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26043,103 +26041,309 @@ "diagnosis": null }, { - "claim_id": "74553c98c56cca6455f2", + "claim_id": "5c7bdc603a26c62b390c", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "net_fiscal_cost", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "impacted estates over \u00a310 million", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "55%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.55, + "reform_key": "db11caf3e07ba4d1", + "external_value": -520000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.55, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -520000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "not_computed", - "engine_version": "", - "data_bundle": "", - "release": "", - "construction": "", - "policyengine_variables": [], - "computed_at": "", - "annotations": [], - "baseline": null, - "status_effective": "not_computed", - "ratio": null, - "delta": null + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.327759407536785, + "delta": -170434891.91912818 }, - "results": [], + "results": [ + { + "value": -434000000.0, + "status": "concept_mismatch", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2027 (FY 2027-28); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.434bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a30.902bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + }, + { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], "diagnosis": null }, { - "claim_id": "878f5fd8884fa99dd180", + "claim_id": "80927270ac5c346db43a", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "static_cost", "country": "UK", - "denominator": "additional tax from the reform", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "\u2018small family farms\u2019 (APR/BPR share > 60% of estate and estate value < \u00a35m)", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "15%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.15, + "reform_key": "db11caf3e07ba4d1", + "external_value": -800000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.15, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -800000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 0.8630436148989102, + "delta": 109565108.08087182 + }, + "results": [ + { + "value": -434000000.0, + "status": "constructed", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2027 (FY 2027-28); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "A static cost on both sides, but the published PolicyEngine figure is the engine plus the dashboard's six local corrections to policyengine-uk 2.102.3, so it is a construction rather than the engine as served.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.434bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a30.902bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "constructed" + }, + { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "PolicyEngine's static cost on the certified pair is \u00a30.690bn in 2027-28, 14% below CenTax's \u00a30.80bn. Our side first: on this pair the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours (measured), and with every family at the full 30 hours the same reform costs \u00a31.404bn (diagnostic run). microcosm#1126 identifies this draw as the main reason PolicyEngine's free-hours spending is low and derives DfE's mean as about 28.5 hours. The fix merged in microcosm on 16 October 2026 (microcosm#1299); this certified pair was built on 19 June 2026, before it. policyengine.py#685 (merged 21 October 2026) certified a Microcosm pair with the fixed draw, national cut 6cba5d71 with policyengine-uk 2.132.6, which this lane has not run yet (policyengine-scorecard#228). PolicyEngine's published dashboard figure, on policyengine.org from 12 to 21 October 2026, ran on the 2 October Microcosm release that microcosm#1126 examines, with the same draw. Two smaller PolicyEngine-side factors are recorded on the result: the universal-hours switch-off (policyengine-uk#1930, changed after the pin by policyengine-uk#2177) and no hours for children under 1.", + "action_link": "https://github.com/PolicyEngine/microcosm/issues/1126" + } + }, + { + "claim_id": "ee1a11d985fbb0514bed", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26161,44 +26365,43 @@ "diagnosis": null }, { - "claim_id": "8d00404bfd0eecdebbdd", + "claim_id": "ec77f8f7a193a4ec2fc1", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "deceased with income over \u00a3100,000", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Almost half (47%)" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.47, + "reform_key": "db11caf3e07ba4d1", + "external_value": 100000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.47, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 100000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26220,138 +26423,259 @@ "diagnosis": null }, { - "claim_id": "adfa5c7278bede4c71c6", + "claim_id": "f8c6e59097968afab7c5", "source": "centax", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", "benchmark_class": "different_model", + "component": "net_fiscal_cost", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "incidence": "full incidence on partners, zero behavioural response", - "income_concept": "total income (all taxpayers)", - "income_group": "decile_10", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "98%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -610000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -610000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.1318604785559478, + "delta": -80434891.91912818 + }, + "results": [ + { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "06bdb7943962912e4b32", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.98, + "reform_key": "db11caf3e07ba4d1", + "external_value": -820000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.98, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -820000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "pe_gap", + "value": -690434891.9191282, + "status": "comparable", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." ], "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null + "status_effective": "comparable", + "ratio": 0.8419937706330831, + "delta": 129565108.08087182 }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -690434891.9191282, + "status": "comparable", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "comparable" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "b546dfcd9e2a9adfafc9", + "claim_id": "45836aa1bcf5a8bfe397", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "extensive_margin", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "deceased with income over \u00a3500,000", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "23%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.23, + "reform_key": "db11caf3e07ba4d1", + "external_value": 90000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.23, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 90000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26373,139 +26697,317 @@ "diagnosis": null }, { - "claim_id": "b9572e40f5a923d1494a", + "claim_id": "ed9d406aaafbf04d32cb", "source": "centax", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "range_high", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "estimate_kind": "range_high", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "incidence": "full incidence on partners, zero behavioural response", - "income_concept": "total income (all taxpayers)", - "income_group": "bottom_50pct", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "less than 0.4%" - }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.004, + "reform_key": "db11caf3e07ba4d1", + "external_value": 630000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.004, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 630000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { "value": null, - "status": "pe_gap", - "engine_version": "2.89.2", - "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", - "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" - ], - "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null - }, + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "4ba02ab047bc8d8f8b8c", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 140000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 140000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -4.931677799422344, + "delta": -830434891.9191282 + }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -690434891.9191282, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "concept_mismatch" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "b97a79c5409ab13c7372", + "claim_id": "6f05b84a4298c5d21883", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "static_cost", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "deceased with total income under \u00a325,000 per year (five-year average before death)", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Around 10%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.1, + "reform_key": "db11caf3e07ba4d1", + "external_value": -580000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.1, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -580000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 1.1904049860674624, + "delta": -110434891.91912818 + }, + "results": [ + { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": null + }, + { + "claim_id": "5dd545130bb41b108b55", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 130000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 130000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26527,286 +27029,3090 @@ "diagnosis": null }, { - "claim_id": "ccd04b8d9b2351d508ba", + "claim_id": "2f83ae44603194b65ccb", "source": "centax", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "London", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "London", - "geography_level": "ITL1", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "46%" - }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.46, + "reform_key": "db11caf3e07ba4d1", + "external_value": 210000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.46, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 210000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { "value": null, - "status": "pe_gap", + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "fc87562c6825d37097b6", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -640000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -640000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.30996690907305, + "delta": -198378821.80675197 + }, + "results": [ + { + "value": -505000000.0, + "status": "concept_mismatch", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2029 (FY 2029-30); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.505bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a31.037bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + }, + { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "ed292fa51b2c4a5aab13", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -980000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -980000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 0.8554885936803591, + "delta": 141621178.19324803 + }, + "results": [ + { + "value": -505000000.0, + "status": "constructed", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2029 (FY 2029-30); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "A static cost on both sides, but the published PolicyEngine figure is the engine plus the dashboard's six local corrections to policyengine-uk 2.102.3, so it is a construction rather than the engine as served.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.505bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a31.037bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "constructed" + }, + { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "PolicyEngine's static cost on the certified pair is \u00a30.838bn in 2029-30, 14% below CenTax's \u00a30.98bn. Our side first: on this pair the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours (measured), and with every family at the full 30 hours the same reform costs \u00a31.715bn (diagnostic run). microcosm#1126 identifies this draw as the main reason PolicyEngine's free-hours spending is low and derives DfE's mean as about 28.5 hours. The fix merged in microcosm on 16 October 2026 (microcosm#1299); this certified pair was built on 19 June 2026, before it. policyengine.py#685 (merged 21 October 2026) certified a Microcosm pair with the fixed draw, national cut 6cba5d71 with policyengine-uk 2.132.6, which this lane has not run yet (policyengine-scorecard#228). PolicyEngine's published dashboard figure, on policyengine.org from 12 to 21 October 2026, ran on the 2 October Microcosm release that microcosm#1126 examines, with the same draw. Two smaller PolicyEngine-side factors are recorded on the result: the universal-hours switch-off (policyengine-uk#1930, changed after the pin by policyengine-uk#2177) and no hours for children under 1.", + "action_link": "https://github.com/PolicyEngine/microcosm/issues/1126" + } + }, + { + "claim_id": "9d5a7abbae3f7723c566", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "0fbec555b7bbc88c956d", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "intensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "ae95388df478c4d32b54", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -770000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -770000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.0888036646840935, + "delta": -68378821.80675197 + }, + "results": [ + { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "c517308e20f471a1b547", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -1010000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -1010000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 0.8300780413928237, + "delta": 171621178.19324803 + }, + "results": [ + { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": null + }, + { + "claim_id": "4c9b74cff2348b60b79e", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 90000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 90000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "3fcf35303682330ee20c", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "intensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 770000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 770000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "f74e6038249cb2f6f76d", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 160000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 160000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -5.2398676362922, + "delta": -998378821.806752 + }, + "results": [ + { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "27f00fcd76e9326d6c57", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -700000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -700000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 1.1976840311525028, + "delta": -138378821.80675197 + }, + "results": [ + { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": null + }, + { + "claim_id": "2a18eefc3c44336f3a17", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from the reform", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "farm estates with an APR/BPR claim covering less than 20% of the estate (likely passive investors)", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "2%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.02, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.02, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "392f8284d4c31f166150", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "range_low", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "estimate_kind": "range_low", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "impacted estates worth over \u00a35 million", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Over 80%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.8, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.8, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "74553c98c56cca6455f2", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "impacted estates over \u00a310 million", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "55%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.55, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.55, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "878f5fd8884fa99dd180", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from the reform", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "\u2018small family farms\u2019 (APR/BPR share > 60% of estate and estate value < \u00a35m)", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "15%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.15, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.15, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "8d00404bfd0eecdebbdd", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "deceased with income over \u00a3100,000", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Almost half (47%)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.47, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.47, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "adfa5c7278bede4c71c6", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "incidence": "full incidence on partners, zero behavioural response", + "income_concept": "total income (all taxpayers)", + "income_group": "decile_10", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "98%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.98, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.98, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "b546dfcd9e2a9adfafc9", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "deceased with income over \u00a3500,000", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "23%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.23, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.23, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "b9572e40f5a923d1494a", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "range_high", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "estimate_kind": "range_high", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "incidence": "full incidence on partners, zero behavioural response", + "income_concept": "total income (all taxpayers)", + "income_group": "bottom_50pct", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "less than 0.4%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.004, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.004, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "b97a79c5409ab13c7372", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "deceased with total income under \u00a325,000 per year (five-year average before death)", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Around 10%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.1, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.1, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "ccd04b8d9b2351d508ba", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "London", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "London", + "geography_level": "ITL1", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "46%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.46, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.46, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "d207cb71ef81a7f6d14e", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "incidence": "full incidence on partners, zero behavioural response", + "income_concept": "total income (all taxpayers)", + "income_group": "top_0.1pct", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "58%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.58, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.58, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "dd2cde755a07c0ed9ba7", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "range_high", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "estimate_kind": "range_high", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "impacted farm estates valued at less than \u00a32 million", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Less than 1%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.01, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.01, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "3de3f320ce2d052aec1c", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "window": "", + "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", + "metric": "taxpayer_count", + "unit_concept": "persons", + "value_kind": "point", + "period": 2020, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "Kensington", + "program": null, + "conditions": { + "benchmark_class": "administrative_fact", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2019-20", + "geography": "Kensington", + "geography_level": "westminster_constituency", + "subgroup": "partners", + "value_verbatim": "1800 partners" + }, + "measure": null, + "reform_framework": "baseline", + "reform_key": "03395a10966bb849", + "external_value": 1800.0, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 1800.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "55e1fd35afbb78cc5b82", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "window": "", + "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", + "metric": "taxpayer_count", + "unit_concept": "persons", + "value_kind": "point", + "period": 2020, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "Wales and Northern Ireland", + "program": null, + "conditions": { + "benchmark_class": "administrative_fact", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2019-20", + "geography": "Wales and Northern Ireland", + "subgroup": "partners", + "value_verbatim": "65,000 partners" + }, + "measure": null, + "reform_framework": "baseline", + "reform_key": "03395a10966bb849", + "external_value": 65000.0, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 65000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "1661f467b23f54be83e4", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2026/27", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2026-27", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 70000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -12.667872072679184, + "delta": -956751045.0875429 + }, + "results": [ + { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "b4009e63d3d8f0bdf659", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2026/27", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2026-27", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 330000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.6871243790531603, + "delta": -1216751045.087543 + }, + "results": [ + { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "ba95ee167f32ce7268e6", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2027/28", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 77000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -12.562997492208506, + "delta": -1044350806.9000549 + }, + "results": [ + { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "16bf746b2bd5bd96c22b", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2027/28", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 349000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.7717788163325356, + "delta": -1316350806.900055 + }, + "results": [ + { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "b7ae540056a5cc8df027", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2028/29", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2029, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2028-29", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 91000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -11.51635975540591, + "delta": -1138988737.7419379 + }, + "results": [ + { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "d91bc28b78ea053677de", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2028/29", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2029, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2028-29", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 399000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.6265381898294184, + "delta": -1446988737.7419379 + }, + "results": [ + { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "d62224a3f05da8864a7a", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2029/30", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 132000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1153495218.9554157, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -8.738600143601634, + "delta": -1285495218.9554157 + }, + "results": [ + { + "value": -1153495218.9554157, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "62f2fa619e6b7620ef64", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2029/30", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 427000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1153495218.9554157, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CenTax quotes this cell as \u00a3425 million for 2030; the appendix prints 0.427 (\u00a3427 million) for 2029/30.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null + "status_effective": "concept_mismatch", + "ratio": -2.7013939553991, + "delta": -1580495218.9554157 }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -1153495218.9554157, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CenTax quotes this cell as \u00a3425 million for 2030; the appendix prints 0.427 (\u00a3427 million) for 2029/30.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "concept_mismatch" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "d207cb71ef81a7f6d14e", - "source": "centax", + "claim_id": "1440ce517b94707280ec", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2030/31", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2031, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", + "behavioral_response": "endogenous_pension_exogenous_labour", "benchmark_class": "different_model", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "incidence": "full incidence on partners, zero behavioural response", - "income_concept": "total income (all taxpayers)", - "income_group": "top_0.1pct", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "58%" - }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2030-31", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.58, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.58, + "reform_key": "6afa9ff6f3c2b955", + "external_value": 147000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "pe_gap", + "value": -1344575699.5687501, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null + "status_effective": "concept_mismatch", + "ratio": -9.146773466454082, + "delta": -1491575699.5687501 }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -1344575699.5687501, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "concept_mismatch" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "dd2cde755a07c0ed9ba7", - "source": "centax", + "claim_id": "41bd121814939659b513", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2030/31", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "range_high", - "period": 2027, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2031, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "basis": "static", + "behavioral_response": "endogenous_labour_no_pension", "benchmark_class": "different_model", "country": "UK", - "denominator": "additional tax from farm estates", - "estimate_kind": "range_high", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "impacted farm estates valued at less than \u00a32 million", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Less than 1%" + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2030-31", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.01, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.01, + "reform_key": "6afa9ff6f3c2b955", + "external_value": 462000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "not_computed", - "engine_version": "", - "data_bundle": "", - "release": "", - "construction": "", - "policyengine_variables": [], - "computed_at": "", - "annotations": [], - "baseline": null, - "status_effective": "not_computed", - "ratio": null, - "delta": null + "value": -1344575699.5687501, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.9103370120535716, + "delta": -1806575699.5687501 }, - "results": [], + "results": [ + { + "value": -1344575699.5687501, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], "diagnosis": null }, { - "claim_id": "3de3f320ce2d052aec1c", - "source": "centax", + "claim_id": "2a34d8beac7637566ee1", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2031/32", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", - "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", - "metric": "taxpayer_count", - "unit_concept": "persons", - "value_kind": "point", - "period": 2020, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2032, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "Kensington", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "benchmark_class": "administrative_fact", + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2019-20", - "geography": "Kensington", - "geography_level": "westminster_constituency", - "subgroup": "partners", - "value_verbatim": "1800 partners" + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2031-32", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural", + "pe_missing": "fiscal year 2031-32 is outside the data years of every certified pair this lane has run (pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z 2023-2030; data/uk/replay_variants.json)" }, "measure": null, - "reform_framework": "baseline", - "reform_key": "03395a10966bb849", - "external_value": 1800.0, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 1800.0, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 172000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26828,39 +30134,42 @@ "diagnosis": null }, { - "claim_id": "55e1fd35afbb78cc5b82", - "source": "centax", + "claim_id": "f66050f5b64017fd56a4", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2031/32", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", - "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", - "metric": "taxpayer_count", - "unit_concept": "persons", - "value_kind": "point", - "period": 2020, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2032, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "Wales and Northern Ireland", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "benchmark_class": "administrative_fact", + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2019-20", - "geography": "Wales and Northern Ireland", - "subgroup": "partners", - "value_verbatim": "65,000 partners" + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2031-32", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural", + "pe_missing": "fiscal year 2031-32 is outside the data years of every certified pair this lane has run (pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z 2023-2030; data/uk/replay_variants.json)" }, "measure": null, - "reform_framework": "baseline", - "reform_key": "03395a10966bb849", - "external_value": 65000.0, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 65000.0, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 507000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { diff --git a/app/src/sourceLabels.ts b/app/src/sourceLabels.ts index 4d64dd7a5..3787972bc 100644 --- a/app/src/sourceLabels.ts +++ b/app/src/sourceLabels.ts @@ -42,6 +42,7 @@ const SOURCE_SPECIAL: Record = { scottish_fiscal_commission: "Scottish Fiscal Commission", wbg: "Women's Budget Group", centax: "CenTax", + centre_for_british_progress: "Centre for British Progress", tax_policy_associates: "Tax Policy Associates", policy_in_practice: "Policy in Practice", entitledto: "entitledto", diff --git a/data/externals/cbp-rates-and-wrongs.json b/data/externals/cbp-rates-and-wrongs.json new file mode 100644 index 000000000..404c4fa2c --- /dev/null +++ b/data/externals/cbp-rates-and-wrongs.json @@ -0,0 +1,1442 @@ +[ + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "abolish_cliff", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 70000000, + "value_verbatim": "0.070", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "abolish_cliff", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 77000000, + "value_verbatim": "0.077", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "abolish_cliff", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 91000000, + "value_verbatim": "0.091", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "abolish_cliff", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 132000000, + "value_verbatim": "0.132", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "abolish_cliff", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 147000000, + "value_verbatim": "0.147", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "abolish_cliff", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 172000000, + "value_verbatim": "0.172", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_1pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 106000000, + "value_verbatim": "0.106", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 1%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_1pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 120000000, + "value_verbatim": "0.120", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 1%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_1pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 136000000, + "value_verbatim": "0.136", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 1%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_1pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 224000000, + "value_verbatim": "0.224", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 1%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_1pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 247000000, + "value_verbatim": "0.247", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 1%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_1pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 282000000, + "value_verbatim": "0.282", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 1%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_2pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 150000000, + "value_verbatim": "0.150", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 2%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_2pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 169000000, + "value_verbatim": "0.169", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 2%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_2pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 187000000, + "value_verbatim": "0.187", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 2%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_2pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 282000000, + "value_verbatim": "0.282", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 2%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_2pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 310000000, + "value_verbatim": "0.310", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 2%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_2pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 352000000, + "value_verbatim": "0.352", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 2%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_3pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 189000000, + "value_verbatim": "0.189", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 3%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_3pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 212000000, + "value_verbatim": "0.212", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 3%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_3pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 232000000, + "value_verbatim": "0.232", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 3%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_3pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 333000000, + "value_verbatim": "0.333", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 3%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_3pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 365000000, + "value_verbatim": "0.365", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 3%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_3pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 412000000, + "value_verbatim": "0.412", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 3%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_4pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 223000000, + "value_verbatim": "0.223", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 4%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_4pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 247000000, + "value_verbatim": "0.247", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 4%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_4pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 271000000, + "value_verbatim": "0.271", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 4%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_4pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 375000000, + "value_verbatim": "0.375", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 4%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_4pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 413000000, + "value_verbatim": "0.413", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 4%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_4pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 465000000, + "value_verbatim": "0.465", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 4%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_5pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 253000000, + "value_verbatim": "0.253", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 5%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_5pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 279000000, + "value_verbatim": "0.279", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 5%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_5pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 305000000, + "value_verbatim": "0.305", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 5%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_5pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 413000000, + "value_verbatim": "0.413", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 5%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_5pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 454000000, + "value_verbatim": "0.454", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 5%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_pension_exogenous_labour", + "geography": "England", + "country": "UK", + "table": "2", + "pdf_page": 8, + "policy": "taper_5pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 512000000, + "value_verbatim": "0.512", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "status": "ok", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), 5%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "abolish_cliff", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 330000000, + "value_verbatim": "0.330", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "abolish_cliff", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 349000000, + "value_verbatim": "0.349", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "abolish_cliff", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 399000000, + "value_verbatim": "0.399", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "abolish_cliff", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 427000000, + "value_verbatim": "0.427", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "abolish_cliff", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 462000000, + "value_verbatim": "0.462", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "abolish_cliff", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 507000000, + "value_verbatim": "0.507", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_1pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 398000000, + "value_verbatim": "0.398", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 1%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_1pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 427000000, + "value_verbatim": "0.427", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 1%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_1pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 483000000, + "value_verbatim": "0.483", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 1%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_1pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 518000000, + "value_verbatim": "0.518", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 1%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_1pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 560000000, + "value_verbatim": "0.560", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 1%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_1pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 615000000, + "value_verbatim": "0.615", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 1%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_2pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 472000000, + "value_verbatim": "0.472", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 2%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_2pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 508000000, + "value_verbatim": "0.508", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 2%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_2pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 569000000, + "value_verbatim": "0.569", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 2%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_2pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 611000000, + "value_verbatim": "0.611", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 2%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_2pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 662000000, + "value_verbatim": "0.662", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 2%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_2pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 726000000, + "value_verbatim": "0.726", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 2%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_3pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 537000000, + "value_verbatim": "0.537", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 3%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_3pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 580000000, + "value_verbatim": "0.580", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 3%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_3pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 646000000, + "value_verbatim": "0.646", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 3%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_3pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 694000000, + "value_verbatim": "0.694", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 3%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_3pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 752000000, + "value_verbatim": "0.752", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 3%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_3pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 825000000, + "value_verbatim": "0.825", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 3%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_4pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 595000000, + "value_verbatim": "0.595", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 4%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_4pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 643000000, + "value_verbatim": "0.643", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 4%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_4pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 715000000, + "value_verbatim": "0.715", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 4%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_4pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 767000000, + "value_verbatim": "0.767", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 4%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_4pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 832000000, + "value_verbatim": "0.832", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 4%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_4pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 912000000, + "value_verbatim": "0.912", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 4%, 2031/32" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_5pct", + "fy": "2026-27", + "period": 2027, + "unit_concept": "gbp", + "value": 647000000, + "value_verbatim": "0.647", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 5%, 2026/27" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_5pct", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 700000000, + "value_verbatim": "0.700", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 5%, 2027/28" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_5pct", + "fy": "2028-29", + "period": 2029, + "unit_concept": "gbp", + "value": 775000000, + "value_verbatim": "0.775", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 5%, 2028/29" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_5pct", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 833000000, + "value_verbatim": "0.833", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 5%, 2029/30" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_5pct", + "fy": "2030-31", + "period": 2031, + "unit_concept": "gbp", + "value": 903000000, + "value_verbatim": "0.903", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 5%, 2030/31" + }, + { + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": null, + "variant": "endogenous_labour_no_pension", + "geography": "England", + "country": "UK", + "table": "3", + "pdf_page": 10, + "policy": "taper_5pct", + "fy": "2031-32", + "period": 2032, + "unit_concept": "gbp", + "value": 990000000, + "value_verbatim": "0.99", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "status": "ok", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), 5%, 2031/32" + } +] diff --git a/data/externals/centax-childcare-100k.json b/data/externals/centax-childcare-100k.json new file mode 100644 index 000000000..67b1dd278 --- /dev/null +++ b/data/externals/centax-childcare-100k.json @@ -0,0 +1,722 @@ +[ + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_before_removal", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "persons", + "value": 13360, + "value_verbatim": "13,360", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Number Bunching, Before Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_after_removal", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "persons", + "value": 3850, + "value_verbatim": "3,850", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Number Bunching, After Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "static_cost", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": -800000000, + "value_verbatim": "-800m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Static Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "intensive_margin", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 150000000, + "value_verbatim": "+150m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Behavioural Gain (£), Intensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "extensive_margin", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 120000000, + "value_verbatim": "+120m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Behavioural Gain (£), Extensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "net_fiscal_cost", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": -520000000, + "value_verbatim": "-520m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Net Fiscal Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_before_removal", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "persons", + "value": 16280, + "value_verbatim": "16,280", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Number Bunching, Before Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_after_removal", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "persons", + "value": 4300, + "value_verbatim": "4,300", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Number Bunching, After Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "static_cost", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": -980000000, + "value_verbatim": "-980m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Static Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "intensive_margin", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 210000000, + "value_verbatim": "+210m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Behavioural Gain (£), Intensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "extensive_margin", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 130000000, + "value_verbatim": "+130m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Behavioural Gain (£), Extensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "central", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "net_fiscal_cost", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": -640000000, + "value_verbatim": "-640m", + "status": "ok", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Net Fiscal Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_before_removal", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "persons", + "value": 9710, + "value_verbatim": "9,710", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Number Bunching, Before Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_after_removal", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "persons", + "value": 3850, + "value_verbatim": "3,850", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Number Bunching, After Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "static_cost", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": -820000000, + "value_verbatim": "-820m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Static Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "intensive_margin", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 100000000, + "value_verbatim": "+100m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Behavioural Gain (£), Intensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "extensive_margin", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 120000000, + "value_verbatim": "+120m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Behavioural Gain (£), Extensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "net_fiscal_cost", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": -610000000, + "value_verbatim": "-610m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Net Fiscal Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_before_removal", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "persons", + "value": 10860, + "value_verbatim": "10,860", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Number Bunching, Before Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_after_removal", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "persons", + "value": 4300, + "value_verbatim": "4,300", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Number Bunching, After Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "static_cost", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": -1010000000, + "value_verbatim": "-1,010m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Static Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "intensive_margin", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 120000000, + "value_verbatim": "+120m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Behavioural Gain (£), Intensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "extensive_margin", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 120000000, + "value_verbatim": "+120m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Behavioural Gain (£), Extensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "high_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "net_fiscal_cost", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": -770000000, + "value_verbatim": "-770m", + "status": "ok", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Net Fiscal Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_before_removal", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "persons", + "value": 58510, + "value_verbatim": "58,510", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Number Bunching, Before Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_after_removal", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "persons", + "value": 3850, + "value_verbatim": "3,850", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Number Bunching, After Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "static_cost", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": -580000000, + "value_verbatim": "-580m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Static Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "intensive_margin", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 630000000, + "value_verbatim": "+630m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Behavioural Gain (£), Intensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "extensive_margin", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 90000000, + "value_verbatim": "+90m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Behavioural Gain (£), Extensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "net_fiscal_cost", + "tax_year_label": "2028", + "fy": "2027-28", + "period": 2028, + "unit_concept": "gbp", + "value": 140000000, + "value_verbatim": "+140m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Net Fiscal Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_before_removal", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "persons", + "value": 67650, + "value_verbatim": "67,650", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Number Bunching, Before Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "parents_bunching", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "bunching_after_removal", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "persons", + "value": 4300, + "value_verbatim": "4,300", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Number Bunching, After Removal" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "static_cost", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": -700000000, + "value_verbatim": "-700m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Static Cost (£)" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "intensive_margin", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 770000000, + "value_verbatim": "+770m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Behavioural Gain (£), Intensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "extensive_margin", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 90000000, + "value_verbatim": "+90m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Behavioural Gain (£), Extensive Margin" + }, + { + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change", + "subgroup": null, + "variant": "low_cost", + "geography": "England", + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": "net_fiscal_cost", + "tax_year_label": "2030", + "fy": "2029-30", + "period": 2030, + "unit_concept": "gbp", + "value": 160000000, + "value_verbatim": "+160m", + "status": "ok", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Net Fiscal Cost (£)" + } +] diff --git a/data/lanes.json b/data/lanes.json index cdafe9685..cb4e6cf39 100644 --- a/data/lanes.json +++ b/data/lanes.json @@ -4,7 +4,7 @@ "country": "US | UK | BE | NZ \u2014 the model instance the lane feeds (issue #42)", "note": "Repo-committed status feed (issue #7); graduate to Supabase when lane count makes commits noisy. Every lane writes its transition here." }, - "updated": "2026-10-22", + "updated": "2026-10-25", "lanes": [ { "id": "urban-sotsn", @@ -599,6 +599,17 @@ "running": false, "updated": "2026-10-22", "note": "2 claims (ONS September 2026 CPI, OBR's March 2026 forecast of it) and 17 PE results, 15 of them on the HM Treasury and Resolution Foundation uprating claims; parameter-tree read-back on policyengine-uk 2.147.2" + }, + { + "id": "uk-childcare-100k", + "source": "CenTax + Centre for British Progress", + "area": "\u00a3100,000 childcare limit (30 funded hours, Tax-Free Childcare)", + "mode": 2, + "country": "UK", + "stage": "computed", + "running": false, + "updated": "2026-10-25", + "note": "22 of 36 claims computed on the certified pair (populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z, policyengine-uk 2.89.2) \u2014 CenTax Table 4.2 (24 cells: static cost comparable, net cost concept_mismatch) and CBP's technical appendix (12 abolition cells, concept_mismatch); PolicyEngine's published dashboard figure kept as history; pe_gap on the central static rows (microcosm#1126)" } ] } diff --git a/data/populations.json b/data/populations.json index 64b22cb3a..c3e16f1ff 100644 --- a/data/populations.json +++ b/data/populations.json @@ -1,9 +1,9 @@ { - "built": "2026-10-22", + "built": "2026-10-25", "note": "Non-Urban populations exported from scorecard.db: the populace reform-validation registry (issue #20) plus the compute campaign's attached comparisons (US: TPC/CPSP/PWBM/CBO/JCT; UK: the HMRC ready-reckoner family and the complete OBR Policy measures database, including historical HM Treasury scores; BE: JRC EUROMOD-BE claims with demo-grade Axiom concept-mismatch attachments and the 2026 PIT-reform scorekeeper claims; NZ: official budget scores). Claims explicitly opted into pre-result display remain in the feed with latest status not_computed and empty history. Recorded result statuses and calibration relationships are verbatim; nothing here is a pass/fail grade.", "summary": { - "claims": 29735, - "multi_release_claims": 270, + "claims": 29771, + "multi_release_claims": 274, "by_source": { "aspe_welfare_indicators": 14, "az_admin": 1, @@ -16,7 +16,8 @@ "census": 104, "census_pep": 1, "census_stc": 88, - "centax": 80, + "centax": 104, + "centre_for_british_progress": 12, "co_admin": 3, "co_fiscal_note": 1, "cour_des_comptes": 1, @@ -108,17 +109,17 @@ "wi_admin": 1 }, "by_latest_status": { - "comparable": 57, - "concept_mismatch": 17, + "comparable": 63, + "concept_mismatch": 33, "constructed": 1669, - "not_computed": 27103, + "not_computed": 27117, "pe_gap": 889 }, "by_sign_relative_to_pe": { "magnitude": 14, "opposite": 489, - "same": 28963, - "unknown": 269 + "same": 28987, + "unknown": 281 }, "by_country": { "BE": { @@ -136,11 +137,11 @@ "multi_release_claims": 0 }, "UK": { - "claims": 29174, - "with_result": 2179, - "with_value": 1194, - "without_result": 26995, - "multi_release_claims": 0 + "claims": 29210, + "with_result": 2201, + "with_value": 1216, + "without_result": 27009, + "multi_release_claims": 4 }, "US": { "claims": 540, @@ -25924,44 +25925,43 @@ } }, { - "claim_id": "2a18eefc3c44336f3a17", + "claim_id": "68a9d11c877d33191133", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "extensive_margin", "country": "UK", - "denominator": "additional tax from the reform", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "farm estates with an APR/BPR claim covering less than 20% of the estate (likely passive investors)", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "2%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.02, + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.02, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -25983,45 +25983,43 @@ "diagnosis": null }, { - "claim_id": "392f8284d4c31f166150", + "claim_id": "ae9f698e34e48978505d", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "range_low", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "denominator": "additional tax from farm estates", - "estimate_kind": "range_low", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "impacted estates worth over \u00a35 million", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Over 80%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.8, + "reform_key": "db11caf3e07ba4d1", + "external_value": 150000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.8, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 150000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26043,103 +26041,309 @@ "diagnosis": null }, { - "claim_id": "74553c98c56cca6455f2", + "claim_id": "5c7bdc603a26c62b390c", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "net_fiscal_cost", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "impacted estates over \u00a310 million", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "55%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.55, + "reform_key": "db11caf3e07ba4d1", + "external_value": -520000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.55, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -520000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "not_computed", - "engine_version": "", - "data_bundle": "", - "release": "", - "construction": "", - "policyengine_variables": [], - "computed_at": "", - "annotations": [], - "baseline": null, - "status_effective": "not_computed", - "ratio": null, - "delta": null + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.327759407536785, + "delta": -170434891.91912818 }, - "results": [], + "results": [ + { + "value": -434000000.0, + "status": "concept_mismatch", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2027 (FY 2027-28); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.434bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a30.902bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + }, + { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], "diagnosis": null }, { - "claim_id": "878f5fd8884fa99dd180", + "claim_id": "80927270ac5c346db43a", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Central Scenario, tax year 2028: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "static_cost", "country": "UK", - "denominator": "additional tax from the reform", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "\u2018small family farms\u2019 (APR/BPR share > 60% of estate and estate value < \u00a35m)", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "15%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.15, + "reform_key": "db11caf3e07ba4d1", + "external_value": -800000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.15, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -800000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 0.8630436148989102, + "delta": 109565108.08087182 + }, + "results": [ + { + "value": -434000000.0, + "status": "constructed", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2027 (FY 2027-28); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "A static cost on both sides, but the published PolicyEngine figure is the engine plus the dashboard's six local corrections to policyengine-uk 2.102.3, so it is a construction rather than the engine as served.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.434bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a30.902bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "constructed" + }, + { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "PolicyEngine's static cost on the certified pair is \u00a30.690bn in 2027-28, 14% below CenTax's \u00a30.80bn. Our side first: on this pair the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours (measured), and with every family at the full 30 hours the same reform costs \u00a31.404bn (diagnostic run). microcosm#1126 identifies this draw as the main reason PolicyEngine's free-hours spending is low and derives DfE's mean as about 28.5 hours. The fix merged in microcosm on 16 October 2026 (microcosm#1299); this certified pair was built on 19 June 2026, before it. policyengine.py#685 (merged 21 October 2026) certified a Microcosm pair with the fixed draw, national cut 6cba5d71 with policyengine-uk 2.132.6, which this lane has not run yet (policyengine-scorecard#228). PolicyEngine's published dashboard figure, on policyengine.org from 12 to 21 October 2026, ran on the 2 October Microcosm release that microcosm#1126 examines, with the same draw. Two smaller PolicyEngine-side factors are recorded on the result: the universal-hours switch-off (policyengine-uk#1930, changed after the pin by policyengine-uk#2177) and no hours for children under 1.", + "action_link": "https://github.com/PolicyEngine/microcosm/issues/1126" + } + }, + { + "claim_id": "ee1a11d985fbb0514bed", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26161,44 +26365,43 @@ "diagnosis": null }, { - "claim_id": "8d00404bfd0eecdebbdd", + "claim_id": "ec77f8f7a193a4ec2fc1", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "deceased with income over \u00a3100,000", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Almost half (47%)" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.47, + "reform_key": "db11caf3e07ba4d1", + "external_value": 100000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.47, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 100000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26220,138 +26423,259 @@ "diagnosis": null }, { - "claim_id": "adfa5c7278bede4c71c6", + "claim_id": "f8c6e59097968afab7c5", "source": "centax", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", "benchmark_class": "different_model", + "component": "net_fiscal_cost", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "incidence": "full incidence on partners, zero behavioural response", - "income_concept": "total income (all taxpayers)", - "income_group": "decile_10", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "98%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -610000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -610000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.1318604785559478, + "delta": -80434891.91912818 + }, + "results": [ + { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "06bdb7943962912e4b32", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2028: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.98, + "reform_key": "db11caf3e07ba4d1", + "external_value": -820000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.98, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -820000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "pe_gap", + "value": -690434891.9191282, + "status": "comparable", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." ], "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null + "status_effective": "comparable", + "ratio": 0.8419937706330831, + "delta": 129565108.08087182 }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -690434891.9191282, + "status": "comparable", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "comparable" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "b546dfcd9e2a9adfafc9", + "claim_id": "45836aa1bcf5a8bfe397", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "extensive_margin", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "deceased with income over \u00a3500,000", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "23%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.23, + "reform_key": "db11caf3e07ba4d1", + "external_value": 90000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.23, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 90000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26373,139 +26697,317 @@ "diagnosis": null }, { - "claim_id": "b9572e40f5a923d1494a", + "claim_id": "ed9d406aaafbf04d32cb", "source": "centax", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "range_high", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "estimate_kind": "range_high", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "incidence": "full incidence on partners, zero behavioural response", - "income_concept": "total income (all taxpayers)", - "income_group": "bottom_50pct", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "less than 0.4%" - }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.004, + "reform_key": "db11caf3e07ba4d1", + "external_value": 630000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.004, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 630000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { "value": null, - "status": "pe_gap", - "engine_version": "2.89.2", - "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", - "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" - ], - "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null - }, + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "4ba02ab047bc8d8f8b8c", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 140000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 140000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -4.931677799422344, + "delta": -830434891.9191282 + }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -690434891.9191282, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "concept_mismatch" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "b97a79c5409ab13c7372", + "claim_id": "6f05b84a4298c5d21883", "source": "centax", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2028: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "basis": "static", "benchmark_class": "different_model", + "component": "static_cost", "country": "UK", - "denominator": "additional tax from farm estates", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "deceased with total income under \u00a325,000 per year (five-year average before death)", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Around 10%" + "data_scope": "hmrc_administrative", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.1, + "reform_key": "db11caf3e07ba4d1", + "external_value": -580000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.1, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -580000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 1.1904049860674624, + "delta": -110434891.91912818 + }, + "results": [ + { + "value": -690434891.9191282, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.690bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.404bn in 2027-28 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": null + }, + { + "claim_id": "5dd545130bb41b108b55", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 130000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 130000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26527,286 +27029,3090 @@ "diagnosis": null }, { - "claim_id": "ccd04b8d9b2351d508ba", + "claim_id": "2f83ae44603194b65ccb", "source": "centax", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "London", - "program": null, + "geography": "England", + "program": "free_childcare_hours", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", "benchmark_class": "different_model", + "component": "intensive_margin", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "London", - "geography_level": "ITL1", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "46%" - }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.46, + "reform_key": "db11caf3e07ba4d1", + "external_value": 210000000.0, "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.46, + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 210000000.0, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { "value": null, - "status": "pe_gap", + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "fc87562c6825d37097b6", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -640000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -640000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.30996690907305, + "delta": -198378821.80675197 + }, + "results": [ + { + "value": -505000000.0, + "status": "concept_mismatch", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2029 (FY 2029-30); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.505bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a31.037bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + }, + { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "ed292fa51b2c4a5aab13", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Central Scenario, tax year 2030: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "central", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -980000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -980000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 0.8554885936803591, + "delta": 141621178.19324803 + }, + "results": [ + { + "value": -505000000.0, + "status": "constructed", + "engine_version": "2.102.3", + "data_bundle": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "release": "1002T230158Z", + "construction": "published: the 30-hours leg of PolicyEngine's dashboard (https://policyengine.org/uk/childcare-100k-threshold), budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended), year 2029 (FY 2029-30); PolicyEngine/uk-childcare-100k-threshold@f25f0db data/results.json (sha256 4193f13b2a1f); policyengine 6.2.1, dataset passed unmanaged (pinned_by_repo_unmanaged_by_bundle)", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-09T11:50:21+00:00", + "annotations": [ + "A static cost on both sides, but the published PolicyEngine figure is the engine plus the dashboard's six local corrections to policyengine-uk 2.102.3, so it is a construction rather than the engine as served.", + "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at \u00a30.159bn in 2027-28 and \u00a30.173bn in 2029-30).", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "Published figure: \u00a30.505bn. Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly \u00a3100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126.", + "On https://policyengine.org/uk/childcare-100k-threshold from 2026-10-12 to 2026-10-21: policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'.", + "Since then https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8 merged (2026-10-24): the dashboard's own deployment, https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25), serves \u00a31.037bn in this year from PolicyEngine/uk-childcare-100k-threshold@1b83dd4 on microcosm-uk-2024-25-national-20261023T075039Z-a133f37a with policyengine-uk 2.147.2. Not attached: policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + ], + "baseline": "current_law", + "status_effective": "constructed" + }, + { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "PolicyEngine's static cost on the certified pair is \u00a30.838bn in 2029-30, 14% below CenTax's \u00a30.98bn. Our side first: on this pair the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours (measured), and with every family at the full 30 hours the same reform costs \u00a31.715bn (diagnostic run). microcosm#1126 identifies this draw as the main reason PolicyEngine's free-hours spending is low and derives DfE's mean as about 28.5 hours. The fix merged in microcosm on 16 October 2026 (microcosm#1299); this certified pair was built on 19 June 2026, before it. policyengine.py#685 (merged 21 October 2026) certified a Microcosm pair with the fixed draw, national cut 6cba5d71 with policyengine-uk 2.132.6, which this lane has not run yet (policyengine-scorecard#228). PolicyEngine's published dashboard figure, on policyengine.org from 12 to 21 October 2026, ran on the 2 October Microcosm release that microcosm#1126 examines, with the same draw. Two smaller PolicyEngine-side factors are recorded on the result: the universal-hours switch-off (policyengine-uk#1930, changed after the pin by policyengine-uk#2177) and no hours for children under 1.", + "action_link": "https://github.com/PolicyEngine/microcosm/issues/1126" + } + }, + { + "claim_id": "9d5a7abbae3f7723c566", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "0fbec555b7bbc88c956d", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "intensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 120000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 120000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "ae95388df478c4d32b54", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -770000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -770000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": 1.0888036646840935, + "delta": -68378821.80675197 + }, + "results": [ + { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "c517308e20f471a1b547", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, High-Cost Scenario, tax year 2030: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "high_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -1010000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -1010000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 0.8300780413928237, + "delta": 171621178.19324803 + }, + "results": [ + { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": null + }, + { + "claim_id": "4c9b74cff2348b60b79e", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Extensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "extensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 90000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 90000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "3fcf35303682330ee20c", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Behavioural Gain (\u00a3), Intensive Margin", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "intensive_margin", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')", + "pe_missing": "behavioural leg: this lane's PolicyEngine counterparts are static runs (policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 770000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 770000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "f74e6038249cb2f6f76d", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Net Fiscal Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "net_fiscal_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "post_behavioural", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": 160000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": 160000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -5.2398676362922, + "delta": -998378821.806752 + }, + "results": [ + { + "value": -838378821.806752, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below \u00a3100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "27f00fcd76e9326d6c57", + "source": "centax", + "country": "UK", + "source_column": "Table 4.2, Low-Cost Scenario, tax year 2030: Static Cost (\u00a3)", + "name": "Removing the childcare cliff-edge: impacts and cost of reform", + "window": "", + "publication_title": "Removing the childcare cliff-edge: impacts and cost of reform", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours", + "conditions": { + "benchmark_class": "different_model", + "component": "static_cost", + "country": "UK", + "data_scope": "hmrc_administrative", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_100k_limit_removed", + "program": "free_childcare_hours", + "scenario": "low_cost", + "scoring_method": "static", + "sign_convention": "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "db11caf3e07ba4d1", + "external_value": -700000000.0, + "sign_relative_to_pe": "same", + "sign_basis": "stated:exchequer:same", + "external_value_pe_sign": -700000000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable", + "ratio": 1.1976840311525028, + "delta": -138378821.80675197 + }, + "results": [ + { + "value": -838378821.806752, + "status": "comparable", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above \u00a3100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above \u00a3100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "CenTax costs the free childcare hours only ('removing the \u00a3100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own \u00a3100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The \u00a3500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.000bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.838bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.000bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 100000 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).", + "Diagnostic, not attached: with every family at the full 30 hours (the input set to 30 in both worlds before any calculation) the same reform costs \u00a31.715bn in 2029-30 on this pair." + ], + "baseline": "current_law", + "status_effective": "comparable" + } + ], + "diagnosis": null + }, + { + "claim_id": "2a18eefc3c44336f3a17", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from the reform", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "farm estates with an APR/BPR claim covering less than 20% of the estate (likely passive investors)", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "2%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.02, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.02, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "392f8284d4c31f166150", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "range_low", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "estimate_kind": "range_low", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "impacted estates worth over \u00a35 million", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Over 80%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.8, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.8, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "74553c98c56cca6455f2", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "impacted estates over \u00a310 million", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "55%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.55, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.55, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "878f5fd8884fa99dd180", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from the reform", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "\u2018small family farms\u2019 (APR/BPR share > 60% of estate and estate value < \u00a35m)", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "15%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.15, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.15, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "8d00404bfd0eecdebbdd", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "deceased with income over \u00a3100,000", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Almost half (47%)" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.47, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.47, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "adfa5c7278bede4c71c6", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "incidence": "full incidence on partners, zero behavioural response", + "income_concept": "total income (all taxpayers)", + "income_group": "decile_10", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "98%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.98, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.98, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "b546dfcd9e2a9adfafc9", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "deceased with income over \u00a3500,000", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "23%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.23, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.23, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "b9572e40f5a923d1494a", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "range_high", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "estimate_kind": "range_high", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "incidence": "full incidence on partners, zero behavioural response", + "income_concept": "total income (all taxpayers)", + "income_group": "bottom_50pct", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "less than 0.4%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.004, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.004, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "b97a79c5409ab13c7372", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "deceased with total income under \u00a325,000 per year (five-year average before death)", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Around 10%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.1, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.1, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "ccd04b8d9b2351d508ba", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "London", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "London", + "geography_level": "ITL1", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "46%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.46, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.46, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "d207cb71ef81a7f6d14e", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "window": "", + "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "point", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "incidence": "full incidence on partners, zero behavioural response", + "income_concept": "total income (all taxpayers)", + "income_group": "top_0.1pct", + "measure_key": "ab2025_option__nics_on_partnership_income", + "pe_expressibility": "not_expressible", + "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", + "value_verbatim": "58%" + }, + "measure": { + "key": "ab2025_option__nics_on_partnership_income", + "program": "national_insurance", + "computability": "not_expressible", + "head_variables": [], + "parameter_paths": [] + }, + "reform_framework": "policy_ref", + "reform_key": "7718692432a4c4f2", + "external_value": 0.58, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.58, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap", + "ratio": null, + "delta": null + }, + "results": [ + { + "value": null, + "status": "pe_gap", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", + "policyengine_variables": [], + "computed_at": "2026-09-24T00:00:00+00:00", + "annotations": [ + "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", + "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + ], + "baseline": "current_law", + "status_effective": "pe_gap" + } + ], + "diagnosis": { + "class": "pe_gap", + "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", + "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" + } + }, + { + "claim_id": "dd2cde755a07c0ed9ba7", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "window": "", + "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", + "metric": "revenue_share", + "unit_concept": "share", + "value_kind": "range_high", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "UK", + "program": null, + "conditions": { + "basis": "static", + "benchmark_class": "different_model", + "country": "UK", + "denominator": "additional tax from farm estates", + "estimate_kind": "range_high", + "fiscal_event": "autumn_budget_2025", + "fy": "2026-27", + "geography": "UK", + "measure_type": "per year of deaths, uprated to 2027", + "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", + "subgroup": "impacted farm estates valued at less than \u00a32 million", + "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", + "value_verbatim": "Less than 1%" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "78c68f7e47327f86", + "external_value": 0.01, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 0.01, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "3de3f320ce2d052aec1c", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "window": "", + "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", + "metric": "taxpayer_count", + "unit_concept": "persons", + "value_kind": "point", + "period": 2020, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "Kensington", + "program": null, + "conditions": { + "benchmark_class": "administrative_fact", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2019-20", + "geography": "Kensington", + "geography_level": "westminster_constituency", + "subgroup": "partners", + "value_verbatim": "1800 partners" + }, + "measure": null, + "reform_framework": "baseline", + "reform_key": "03395a10966bb849", + "external_value": 1800.0, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 1800.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "55e1fd35afbb78cc5b82", + "source": "centax", + "country": "UK", + "source_column": "", + "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "window": "", + "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", + "metric": "taxpayer_count", + "unit_concept": "persons", + "value_kind": "point", + "period": 2020, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "Wales and Northern Ireland", + "program": null, + "conditions": { + "benchmark_class": "administrative_fact", + "country": "UK", + "fiscal_event": "autumn_budget_2025", + "fy": "2019-20", + "geography": "Wales and Northern Ireland", + "subgroup": "partners", + "value_verbatim": "65,000 partners" + }, + "measure": null, + "reform_framework": "baseline", + "reform_key": "03395a10966bb849", + "external_value": 65000.0, + "sign_relative_to_pe": "same", + "sign_basis": "level", + "external_value_pe_sign": 65000.0, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": null, + "status": "not_computed", + "engine_version": "", + "data_bundle": "", + "release": "", + "construction": "", + "policyengine_variables": [], + "computed_at": "", + "annotations": [], + "baseline": null, + "status_effective": "not_computed", + "ratio": null, + "delta": null + }, + "results": [], + "diagnosis": null + }, + { + "claim_id": "1661f467b23f54be83e4", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2026/27", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2026-27", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 70000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -12.667872072679184, + "delta": -956751045.0875429 + }, + "results": [ + { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "b4009e63d3d8f0bdf659", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2026/27", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2027, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2026-27", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 330000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.6871243790531603, + "delta": -1216751045.087543 + }, + "results": [ + { + "value": -886751045.0875429, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2026 (FY 2026-27); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2026-27 (engine period 2026, the start year, for claim period 2027), people in English households: extended hours \u00a3+0.714bn, universal \u00a3-0.085bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.258bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.891bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.004bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.85m under current law, 0.98m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2026: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "ba95ee167f32ce7268e6", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2027/28", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 77000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -12.562997492208506, + "delta": -1044350806.9000549 + }, + "results": [ + { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "16bf746b2bd5bd96c22b", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2027/28", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2028, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2027-28", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 349000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.7717788163325356, + "delta": -1316350806.900055 + }, + "results": [ + { + "value": -967350806.9000549, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2027 (FY 2027-28); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2027-28 (engine period 2027, the start year, for claim period 2028), people in English households: extended hours \u00a3+0.792bn, universal \u00a3-0.101bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.277bn (reform minus current law). UK-wide, gov_spending moves \u00a3+0.973bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.006bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 0.99m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2027: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "b7ae540056a5cc8df027", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2028/29", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2029, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2028-29", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 91000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -11.51635975540591, + "delta": -1138988737.7419379 + }, + "results": [ + { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "d91bc28b78ea053677de", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2028/29", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2029, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2028-29", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 399000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.6265381898294184, + "delta": -1446988737.7419379 + }, + "results": [ + { + "value": -1047988737.7419379, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2028 (FY 2028-29); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2028-29 (engine period 2028, the start year, for claim period 2029), people in English households: extended hours \u00a3+0.868bn, universal \u00a3-0.118bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.298bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.058bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.00m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2028: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "d62224a3f05da8864a7a", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2029/30", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 132000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1153495218.9554157, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -8.738600143601634, + "delta": -1285495218.9554157 + }, + "results": [ + { + "value": -1153495218.9554157, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], + "diagnosis": null + }, + { + "claim_id": "62f2fa619e6b7620ef64", + "source": "centre_for_british_progress", + "country": "UK", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2029/30", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "window": "", + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2030, + "time_basis": "fiscal_year", + "period_start": null, + "period_end": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", + "conditions": { + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", + "country": "UK", + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2029-30", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" + }, + "measure": null, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 427000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, + "calibration_relationship": "held_out", + "claim_baseline": "current_law", + "latest": { + "value": -1153495218.9554157, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CenTax quotes this cell as \u00a3425 million for 2030; the appendix prints 0.427 (\u00a3427 million) for 2029/30.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null + "status_effective": "concept_mismatch", + "ratio": -2.7013939553991, + "delta": -1580495218.9554157 }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -1153495218.9554157, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2029 (FY 2029-30); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CenTax quotes this cell as \u00a3425 million for 2030; the appendix prints 0.427 (\u00a3427 million) for 2029/30.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2029-30 (engine period 2029, the start year, for claim period 2030), people in English households: extended hours \u00a3+0.959bn, universal \u00a3-0.121bn, targeted \u00a3+0.000bn, Tax-Free Childcare \u00a3+0.315bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.164bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.84m under current law, 1.01m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.3 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2029: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "concept_mismatch" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "d207cb71ef81a7f6d14e", - "source": "centax", + "claim_id": "1440ce517b94707280ec", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2030/31", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Equalising National Insurance on Partnership Income: Revenue and Distributional Effects", - "url": "https://centax.org.uk/wp-content/uploads/2025/09/AdvaniGazmuribarkerLonsdaleSummers2025_PartnershipNICs.pdf", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "point", - "period": 2027, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2031, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136", - "basis": "static", + "behavioral_response": "endogenous_pension_exogenous_labour", "benchmark_class": "different_model", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "incidence": "full incidence on partners, zero behavioural response", - "income_concept": "total income (all taxpayers)", - "income_group": "top_0.1pct", - "measure_key": "ab2025_option__nics_on_partnership_income", - "pe_expressibility": "not_expressible", - "pe_missing": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "reform_hint": "\u2018Partnership NICs\u2019: an equivalent to Class 1 Secondary (Employer) NICs at 13.04% on partnership profits above a \u00a35,000 Partners\u2019 Exempt Amount, with a \u00a310,500 Partnership Allowance applied in proportion to each partner's profit share", - "value_verbatim": "58%" - }, - "measure": { - "key": "ab2025_option__nics_on_partnership_income", - "program": "national_insurance", - "computability": "not_expressible", - "head_variables": [], - "parameter_paths": [] + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2030-31", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" }, + "measure": null, "reform_framework": "policy_ref", - "reform_key": "7718692432a4c4f2", - "external_value": 0.58, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.58, + "reform_key": "6afa9ff6f3c2b955", + "external_value": 147000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "pe_gap", + "value": -1344575699.5687501, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap", - "ratio": null, - "delta": null + "status_effective": "concept_mismatch", + "ratio": -9.146773466454082, + "delta": -1491575699.5687501 }, "results": [ { - "value": null, - "status": "pe_gap", + "value": -1344575699.5687501, + "status": "concept_mismatch", "engine_version": "2.89.2", "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", "release": "8c73-4aa4b14", - "construction": "pe_gap:not_expressible:ab2025_option__nics_on_partnership_income", - "policyengine_variables": [], - "computed_at": "2026-09-24T00:00:00+00:00", - "annotations": [ - "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "Searched BOTH the parameter tree and the VARIABLE list for /partner|llp|self_employment_income|class_4/. Parameters: 5 node(s) ['gov.hmrc.income_tax.bases.taxable_self_employment_income_deductions', 'gov.hmrc.national_insurance.class_4.rates.additional', 'gov.hmrc.national_insurance.class_4.rates.main', 'gov.hmrc.national_insurance.class_4.thresholds.lower_profits_limit', 'gov.hmrc.national_insurance.class_4.thresholds.upper_profits_limit']. Variables: 6 ['self_employment_income', 'ni_class_4_maximum', 'ni_class_4', 'ni_class_4_main', 'partners_unused_personal_allowance', 'taxable_self_employment_income']. Both are searched because a parameter-only search misses a variable (#106). Re-run at policyengine-uk 2.89.2.", - "no run executed: the verdict is the registry's, stated against the certified engine 2.89.2" + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." ], "baseline": "current_law", - "status_effective": "pe_gap" + "status_effective": "concept_mismatch" } ], - "diagnosis": { - "class": "pe_gap", - "rationale": "Partnership and LLP profits are not identified within self_employment_income on the bundle, and Class 4 NICs (the nearest lever, ni_class_4 at 6% above the LPL) applies to all self-employed; a charge on partners only cannot be targeted.", - "action_link": "https://github.com/PolicyEngine/policyengine-scorecard/issues/136" - } + "diagnosis": null }, { - "claim_id": "dd2cde755a07c0ed9ba7", - "source": "centax", + "claim_id": "41bd121814939659b513", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2030/31", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Policy Report: The Impact of Changes to Inheritance Tax on Farm Estates (landing page, executive summary)", - "url": "https://centax.org.uk/policy-brief-the-impact-of-changes-to-inheritance-tax-on-farm-estates/", - "metric": "revenue_share", - "unit_concept": "share", - "value_kind": "range_high", - "period": 2027, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2031, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "UK", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "basis": "static", + "behavioral_response": "endogenous_labour_no_pension", "benchmark_class": "different_model", "country": "UK", - "denominator": "additional tax from farm estates", - "estimate_kind": "range_high", - "fiscal_event": "autumn_budget_2025", - "fy": "2026-27", - "geography": "UK", - "measure_type": "per year of deaths, uprated to 2027", - "reform_hint": "the planned reform: from April 2026 the first \u00a31 million of combined qualifying agricultural and business property keeps 100% APR/BPR relief and relief above that is reduced to 50% (Autumn Budget 2024); estates uprated to 2027, per year of deaths", - "subgroup": "impacted farm estates valued at less than \u00a32 million", - "unit_population": "farm estates (estates of individuals who died owning farm assets on which relief was claimed)", - "value_verbatim": "Less than 1%" + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2030-31", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural" }, "measure": null, "reform_framework": "policy_ref", - "reform_key": "78c68f7e47327f86", - "external_value": 0.01, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 0.01, + "reform_key": "6afa9ff6f3c2b955", + "external_value": 462000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { - "value": null, - "status": "not_computed", - "engine_version": "", - "data_bundle": "", - "release": "", - "construction": "", - "policyengine_variables": [], - "computed_at": "", - "annotations": [], - "baseline": null, - "status_effective": "not_computed", - "ratio": null, - "delta": null + "value": -1344575699.5687501, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch", + "ratio": -2.9103370120535716, + "delta": -1806575699.5687501 }, - "results": [], + "results": [ + { + "value": -1344575699.5687501, + "status": "concept_mismatch", + "engine_version": "2.89.2", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "release": "8c73-4aa4b14", + "construction": "static: the certified world with gov.dfe.extended_childcare_entitlement.income.limit, gov.hmrc.tax_free_childcare.income.income_limit set to 1e+100 minus the certified world as served, engine period 2030 (FY 2030-31); minus the change in extended_childcare_entitlement + universal_childcare_entitlement + targeted_childcare_entitlement + tax_free_childcare for people in English households (microdf weights); artifact results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json", + "policyengine_variables": [ + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", + "tax_free_childcare" + ], + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "annotations": [ + "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at \u00a3100,000 only when that gains them more than a \u00a31,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "CBP's subsidy includes Tax-Free Childcare: its tables assume \u00a3500 of Tax-Free Childcare per child (the model description says up to \u00a32,000). The PolicyEngine counterpart therefore removes both \u00a3100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "Measured on the certified pair, 2030-31 (engine period 2030, the start year, for claim period 2031), people in English households: extended hours \u00a3+1.160bn, universal \u00a3-0.140bn, targeted \u00a3-0.008bn, Tax-Free Childcare \u00a3+0.333bn (reform minus current law). UK-wide, gov_spending moves \u00a3+1.355bn and gov_tax \u00a3+0.000bn; Tax-Free Childcare outside England moves \u00a3+0.010bn, left out because both offers are England's.", + "Families receiving the extended hours: 0.82m under current law, 1.03m with the limit removed.", + "Hours-usage draw: the families receiving the extended hours use a weighted mean of 15.4 of 30 weekly hours under current law (maximum_extended_childcare_hours_usage). microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "Take-up and usage the run read: the weighted share of benefit units whose take-up input is true under current law is would_claim_extended_childcare 0.807, would_claim_targeted_childcare 0.601, would_claim_tfc 0.573, would_claim_universal_childcare 0.570; maximum_extended_childcare_hours_usage is the dataset's own input.", + "Executed limits on 30 April 2030: 30 hours 1e+100, Tax-Free Childcare 1e+100 (1e+100 is the repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT)." + ], + "baseline": "current_law", + "status_effective": "concept_mismatch" + } + ], "diagnosis": null }, { - "claim_id": "3de3f320ce2d052aec1c", - "source": "centax", + "claim_id": "2a34d8beac7637566ee1", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "source_column": "Technical appendix Table 2 (Endogenous Pension, Exogenous Labour), Abolish Cliff, 2031/32", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", - "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", - "metric": "taxpayer_count", - "unit_concept": "persons", - "value_kind": "point", - "period": 2020, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2032, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "Kensington", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "benchmark_class": "administrative_fact", + "behavioral_response": "endogenous_pension_exogenous_labour", + "benchmark_class": "different_model", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2019-20", - "geography": "Kensington", - "geography_level": "westminster_constituency", - "subgroup": "partners", - "value_verbatim": "1800 partners" + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2031-32", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural", + "pe_missing": "fiscal year 2031-32 is outside the data years of every certified pair this lane has run (pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z 2023-2030; data/uk/replay_variants.json)" }, "measure": null, - "reform_framework": "baseline", - "reform_key": "03395a10966bb849", - "external_value": 1800.0, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 1800.0, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 172000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { @@ -26828,39 +30134,42 @@ "diagnosis": null }, { - "claim_id": "55e1fd35afbb78cc5b82", - "source": "centax", + "claim_id": "f66050f5b64017fd56a4", + "source": "centre_for_british_progress", "country": "UK", - "source_column": "", - "name": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", + "source_column": "Technical appendix Table 3 (Endogenous Labour, No Pension), Abolish Cliff, 2031/32", + "name": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", "window": "", - "publication_title": "Policy report: Equalising National Insurance on Partnership Income: Revenue and Distributional Effects (landing page, key findings)", - "url": "https://centax.org.uk/policy-report-equalising-national-insurance-on-partnership-income-revenue-and-distributional-effects/", - "metric": "taxpayer_count", - "unit_concept": "persons", - "value_kind": "point", - "period": 2020, + "publication_title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "metric": "revenue_change", + "unit_concept": "gbp", + "value_kind": "gbp", + "period": 2032, "time_basis": "fiscal_year", "period_start": null, "period_end": null, - "geography": "Wales and Northern Ireland", - "program": null, + "geography": "England", + "program": "free_childcare_hours_and_tax_free_childcare", "conditions": { - "benchmark_class": "administrative_fact", + "behavioral_response": "endogenous_labour_no_pension", + "benchmark_class": "different_model", "country": "UK", - "fiscal_event": "autumn_budget_2025", - "fy": "2019-20", - "geography": "Wales and Northern Ireland", - "subgroup": "partners", - "value_verbatim": "65,000 partners" + "data_scope": "synthetic_ons_births_spi_2022_23", + "fy": "2031-32", + "geography": "England", + "measure_key": "uk_free_childcare_hours_and_tfc_100k_limit_removed", + "program": "free_childcare_hours_and_tax_free_childcare", + "scoring_method": "post_behavioural", + "pe_missing": "fiscal year 2031-32 is outside the data years of every certified pair this lane has run (pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z 2023-2030; data/uk/replay_variants.json)" }, "measure": null, - "reform_framework": "baseline", - "reform_key": "03395a10966bb849", - "external_value": 65000.0, - "sign_relative_to_pe": "same", - "sign_basis": "level", - "external_value_pe_sign": 65000.0, + "reform_framework": "policy_ref", + "reform_key": "6afa9ff6f3c2b955", + "external_value": 507000000.0, + "sign_relative_to_pe": "unknown", + "sign_basis": "no_stated_convention", + "external_value_pe_sign": null, "calibration_relationship": "held_out", "claim_baseline": "current_law", "latest": { diff --git a/docs/ARCHITECTURE.md b/docs/ARCHITECTURE.md index 48bef8b8e..db0d3e2b1 100644 --- a/docs/ARCHITECTURE.md +++ b/docs/ARCHITECTURE.md @@ -34,8 +34,8 @@ document is the contract that makes instance N cheap. ## The tidy row schema (adapter contract) -Every adapter emits rows of exactly this shape; nothing downstream knows -anything source-specific except the counterpart mapping: +Every adapter emits rows with at least these fields; nothing downstream +knows anything source-specific except the counterpart mapping: ```json { @@ -56,7 +56,11 @@ anything source-specific except the counterpart mapping: Conventions: counts in raw units (adapter undoes source scaling), rates as fractions, `subgroup` slugs shared across sources where concepts genuinely match (age bands, race, disability), `variant` for source-side methodological -forks (e.g. TANF with/without solely-state-funded programs). +forks (e.g. TANF with/without solely-state-funded programs, a costing's +behavioural scenario). An adapter may carry its source's own facets beside +these fields (a table and column, a fiscal-year label, the value as printed); +the lane's ingester must name every field it accepts and reject any other +(`harvest.require_fields`). ## The PE side @@ -227,6 +231,25 @@ Operating rules for the agent stage: `consumed_as_target`: PolicyEngine-UK takes ONS September rates and OBR determinants as inputs, so a match checks the input, not model skill. +- **The £100,000 childcare limit (UK)** (`sources/centax-childcare-100k/`, + `sources/cbp-rates-and-wrongs/`, lane `uk-childcare-100k`): CenTax's + Table 4.2 (static cost, two behavioural margins and net cost; three + scenarios; 2027-28 and 2029-30) and the Centre for British Progress + appendix (net of behaviour, 2026-27 to 2031-32). PDF adapters read a + vendored page extract that a test re-extracts from the PDF. Counterparts + come from `pipeline/compute_uk_childcare_100k.py` on the certified + populace-uk pair: static, the three free-hours entitlements (plus + Tax-Free Childcare for CBP) summed for people in English households. + Each certified variant's artifact keeps its run in the history, and a + claim's latest result is the newest run that computed its year. Ingest + takes an artifact only for a pair with a `verified_pairs` entry, and an + engine or run note attaches only to the pairs its `variants` names. + Static against static is `comparable`; net costs are `concept_mismatch`. + PolicyEngine's published dashboard figure (on policyengine.org from 12 to + 21 October 2026) stays as an older history row, so the published miss + stays on the page, and the central static rows carry a `pe_gap` diagnosis + linked to microcosm#1126 (the hours-usage draw). + ### Staged for Chronicle, not claims Administrative outturn tables (caseloads, units, spending) are calibration diff --git a/pipeline/compute_uk_childcare_100k.py b/pipeline/compute_uk_childcare_100k.py new file mode 100644 index 000000000..10fd65e20 --- /dev/null +++ b/pipeline/compute_uk_childcare_100k.py @@ -0,0 +1,530 @@ +"""PolicyEngine counterparts for the £100,000 childcare limit costings. + +Two external scores cost removing the £100,000 adjusted-net-income limit on +the working-parent childcare offers in England: + + sources/centax-childcare-100k CenTax (Advani, Flew, Pepin-Hall and + Summers, September 2026), Table 4.2: + the free hours only + sources/cbp-rates-and-wrongs Centre for British Progress (25 March + 2026), technical appendix Tables 2-3: + the free hours and Tax-Free Childcare + +This script runs the managed policyengine-uk simulation on the certified +populace-uk bundle (the default replay variant, data/uk/certified_bundle.json) +through the same preflight as the variant's event replays +(pipeline/compute_uk_event.py), and writes one artifact with weighted SUMS +and COUNTS per world and year, UK-wide and for people in English households. +Every ratio and delta is derived downstream +(scorecard_db/ingest_uk_childcare_100k.py). + +Worlds (static: no behavioural response is applied): + + baseline the certified world as served (current law) + free_hours the 30-hours limit removed + (gov.dfe.extended_childcare_entitlement.income.limit) + free_hours_and_tfc both limits removed (also + gov.hmrc.tax_free_childcare.income.income_limit) + +A removed limit executes as compute_uk_ab2025.NO_LIMIT (1e100), the repo's +sentinel for "no limit" in a policyengine-core reform dict; the artifact +records the executed value of each limit in each world and year. + +Diagnostic worlds (2027 and 2029 by default): baseline and free_hours with +the input ``maximum_extended_childcare_hours_usage`` set to 30 for every +benefit unit in every simulated year, before any calculation (the engine's +own default for the variable, i.e. full use of the 30 hours). They measure +how much of the static cost turns on the dataset's hours-usage draw +(microcosm#1126); they are never attached to a claim. + +Offline and pinned, through compute_uk_event's preflight for the variant +(the legacy one or a materialised one): the installed wheels checked +against the variant record before the engine is imported (versions, RECORD +hashes, the packaged bundle manifest, Python), each engine package imported +from its own distribution, Hugging Face offline mode, and the runtime +release bundle equal to the record's. Each simulation must report as its +runtime_dataset_source the exact file hashed against the certified bundle +immediately before construction, and that file is hashed again after the +world's reads. One artifact per variant; ingest keeps every certified +variant's run in the history. + + PYTHONPATH=. .venv-pe289m/bin/python pipeline/compute_uk_childcare_100k.py + PYTHONPATH=. /bin/python pipeline/compute_uk_childcare_100k.py --variant +""" + +from __future__ import annotations + +import argparse +import datetime as dt +import gc +import inspect +import json +import string +import sys +import time +from pathlib import Path + +import numpy as np + +ROOT = Path(__file__).resolve().parent.parent +sys.path.insert(0, str(ROOT)) + +from pipeline import compute_uk_ab2025 as ab +from pipeline import compute_uk_event as event +from pipeline import compute_uk_obr_costings as fiscal +from pipeline import uk_replay_variants as variants + +OUTPUT_DIR = ROOT / "results" / "uk" / "childcare_100k" +YEARS = (2026, 2027, 2028, 2029, 2030) +DIAGNOSTIC_YEARS = (2027, 2029) + +THIRTY_HOURS_LIMIT = "gov.dfe.extended_childcare_entitlement.income.limit" +TFC_LIMIT = "gov.hmrc.tax_free_childcare.income.income_limit" +LIMITS = (THIRTY_HOURS_LIMIT, TFC_LIMIT) +REFORMS = { + "free_hours": (THIRTY_HOURS_LIMIT,), + "free_hours_and_tfc": (THIRTY_HOURS_LIMIT, TFC_LIMIT), +} +HOURS_USAGE = "maximum_extended_childcare_hours_usage" +FULL_USE_HOURS = 30.0 +# Within a data bundle the scorecard's comparisons view prefers these runs +# (scorecard_db/db.py, run_id LIKE 'platform-grid-%', which ignores ASCII +# case), and ingest refuses an artifact that takes it. +RESERVED_RUN_PREFIX = "platform-grid-" +_ASCII_LOWER = str.maketrans(string.ascii_uppercase, string.ascii_lowercase) +# Benefit-unit take-up inputs (each offer's eligibility is defined_for one; +# the engine's default is True). +TAKE_UP = ( + "would_claim_extended_childcare", + "would_claim_universal_childcare", + "would_claim_targeted_childcare", + "would_claim_tfc", +) +# Inputs whose presence in the dataset is recorded: holder periods filled at +# construction, before any calculation, are the dataset's own values. +DATASET_INPUTS = (HOURS_USAGE, *TAKE_UP, "childcare_expenses") + +# The three free early-years entitlements (policyengine-uk's gov_spending +# lists all three) and Tax-Free Childcare. +FREE_HOURS = ( + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", +) +TFC = "tax_free_childcare" +SUMMED = (*FREE_HOURS, TFC, "gov_spending", "gov_tax") +# Both offers are England's, but policyengine-uk 2.89.2 applies Tax-Free +# Childcare UK-wide (only the extended hours test country == ENGLAND), so +# the childcare heads are also summed for people in English households. +HEADS = (*FREE_HOURS, TFC) +ENGLAND = "ENGLAND" +# Variables whose engine metadata is recorded (the pe_meta contract): every +# variable year_block calculates. +METADATA = ( + *SUMMED, + HOURS_USAGE, + "is_child", + "is_child_receiving_extended_childcare", + "country", + *TAKE_UP, + "childcare_expenses", + "max_free_entitlement_hours_used", + "extended_childcare_entitlement_eligible", + "extended_childcare_entitlement_meets_income_requirements", + "universal_childcare_entitlement_eligible", + "targeted_childcare_entitlement_eligible", + "tax_free_childcare_meets_income_requirements", + "adjusted_net_income", + "age", +) + + +def worlds(years) -> dict[str, dict | None]: + """name -> executed reform dict (None: the certified world as served).""" + out: dict[str, dict | None] = {"baseline": None} + for name, paths in REFORMS.items(): + rd, sentinels = ab.reform_dict({p: None for p in paths}, years=list(years)) + if sorted(sentinels) != sorted(paths): + raise SystemExit(f"{name}: sentinel paths {sentinels} != {paths}") + out[name] = rd + return out + + +def variable_metadata(system) -> dict: + out = {} + for name in METADATA: + v = system.variables[name] + source = inspect.getsourcefile(type(v)) or "" + out[name] = { + "label": v.label, + "entity": v.entity.key, + "definition_period": str(v.definition_period), + "unit": str(getattr(v, "unit", None)), + "documentation": v.documentation, + "defined_for": str(getattr(v, "defined_for", None) or "") or None, + "default_value": v.default_value + if isinstance(v.default_value, (int, float, bool, str)) + else None, + "engine_source": "policyengine_uk/" + + source.split("/policyengine_uk/", 1)[-1] + if "/policyengine_uk/" in source + else source, + } + return out + + +def _weighted(series) -> tuple[np.ndarray, np.ndarray]: + return np.asarray(series.values, dtype=float), np.asarray( + series.weights, dtype=float + ) + + +def _share(series) -> float: + values, weights = _weighted(series) + return float((values * weights).sum() / weights.sum()) + + +def limit_values(sim, year: int) -> dict[str, float]: + params = sim.tax_benefit_system.parameters + out = {} + for path in LIMITS: + node = ab._resolve_parameter(params, path) + out[path] = float(node(f"{year}-04-30")) + return out + + +def england_masks(sim, year: int) -> dict[str, np.ndarray]: + """entity key -> True where the household is in England. + + country is a household variable; core's map_result projects it onto + each member (household -> person), and a benefit unit lies inside one + household, so any member gives its country (person -> benunit).""" + country = np.asarray(sim.calculate("country", year).values) + unknown = set(np.unique(country)) - { + "ENGLAND", + "NORTHERN_IRELAND", + "SCOTLAND", + "WALES", + } + if unknown or ENGLAND not in set(country): + raise SystemExit(f"{year}: unexpected country values {sorted(unknown)}") + person = np.asarray( + sim.map_result(country == ENGLAND, "household", "person"), dtype=bool + ) + benunit = np.asarray( + sim.map_result(person, "person", "benunit", how="any"), dtype=bool + ) + if not np.array_equal( + person, np.asarray(sim.map_result(benunit, "benunit", "person"), dtype=bool) + ): + raise SystemExit(f"{year}: a benefit unit spans two countries") + return {"person": person, "benunit": benunit} + + +def england_sums(sim, year: int, masks: dict[str, np.ndarray]) -> dict[str, float]: + out = {} + for v in HEADS: + values, weights = _weighted(sim.calculate(v, year)) + mask = masks[sim.tax_benefit_system.variables[v].entity.key] + out[v] = float((values * weights)[mask].sum()) + return out + + +def year_block(sim, year: int) -> dict: + """Weighted sums and counts for one world and year (microdf weights).""" + sums = {v: float(sim.calculate(v, year).sum()) for v in SUMMED} + masks = england_masks(sim, year) + ext, w_bu = _weighted(sim.calculate("extended_childcare_entitlement", year)) + hours, _ = _weighted(sim.calculate(HOURS_USAGE, year)) + receiving = ext > 0 + age, w_p = _weighted(sim.calculate("age", year)) + is_child = np.asarray(sim.calculate("is_child", year).values, dtype=bool) + universal, _ = _weighted(sim.calculate("universal_childcare_entitlement", year)) + targeted, _ = _weighted(sim.calculate("targeted_childcare_entitlement", year)) + tfc, _ = _weighted(sim.calculate(TFC, year)) + extended_child = np.asarray( + sim.calculate("is_child_receiving_extended_childcare", year).values, + dtype=bool, + ) + under_one = is_child & (age < 1) + fractional_age = np.abs(age - np.round(age)) > 1e-9 + return { + "sums_gbp": sums, + "sums_gbp_england": england_sums(sim, year, masks), + "counts": { + "benunits_receiving_extended": float(w_bu[receiving].sum()), + "children_receiving_extended": float(w_p[extended_child].sum()), + "children_receiving_universal": float(w_p[universal > 0].sum()), + "children_receiving_targeted": float(w_p[targeted > 0].sum()), + "people_receiving_tfc": float(w_p[tfc > 0].sum()), + "people_receiving_tfc_england": float( + w_p[(tfc > 0) & masks["person"]].sum() + ), + "children_under_one": float(w_p[under_one].sum()), + "children_under_one_receiving_extended": float( + w_p[under_one & extended_child].sum() + ), + "people_with_fractional_age": float(w_p[fractional_age].sum()), + }, + # weighted share of benefit units with each take-up input True + "take_up": {v: _share(sim.calculate(v, year)) for v in TAKE_UP}, + "hours_usage": { + "mean_all_benunits": float((hours * w_bu).sum() / w_bu.sum()), + # no mean (None, not 0/0) when no weight receives the offer + "mean_benunits_receiving_extended": float( + (hours[receiving] * w_bu[receiving]).sum() / w_bu[receiving].sum() + ) + if w_bu[receiving].sum() > 0 + else None, + }, + "limits_executed": limit_values(sim, year), + "records": { + "benunits_receiving_extended": int(receiving.sum()), + }, + } + + +def set_full_use(sim, years) -> dict[int, float]: + """Set the hours-usage input to full use in every simulated year before + anything that reads it is calculated; return the post-set weighted mean + per requested year and the (empty) dependent holders checked first.""" + # Nothing that reads the input may have been calculated yet, or the + # override would not reach it: every dependent holder must be empty. + computed = { + v: [str(p) for p in sim.get_holder(v).get_known_periods()] + for v in (*FREE_HOURS, "extended_childcare_entitlement_eligible") + } + if any(computed.values()): + raise SystemExit(f"calculated before the override: {computed}") + dataset_years = sorted(int(y) for y in sim.dataset.years) + # the entity count, read without any calculation + n = sim.populations["benunit"].count + for y in dataset_years: + sim.set_input(HOURS_USAGE, y, np.full(n, FULL_USE_HOURS)) + out = {"dependents_calculated_before_override": computed} + for y in years: + h, w = _weighted(sim.calculate(HOURS_USAGE, y)) + out[y] = float((h * w).sum() / w.sum()) + if not np.allclose(h, FULL_USE_HOURS): + raise SystemExit(f"{HOURS_USAGE} override did not take effect in {y}") + return out + + +def dataset_inputs(sim) -> dict[str, list[str]]: + """input -> the periods the dataset filled, read before any calculation.""" + return { + v: sorted(str(p) for p in sim.get_holder(v).get_known_periods()) + for v in DATASET_INPUTS + } + + +def check_runtime(sim, name: str, pre: dict) -> list[int]: + """The simulation read the hashed file, under the certified bundle, with + the record's data years; returns those years.""" + bundle = dict(getattr(sim, "policyengine_bundle", {}) or {}) + reported = bundle.get("runtime_dataset_source") + source = Path(pre["runtime_dataset_source"]) + if not isinstance(reported, str) or Path(reported).resolve() != source.resolve(): + raise SystemExit( + f"{name}: the managed simulation read {reported!r}, not the file " + f"hashed before construction ({source})" + ) + fiscal.assert_managed_bundle(bundle, pre["release_bundle"]) + dataset_years = sorted(int(y) for y in sim.dataset.years) + if dataset_years != pre["data_years"]: + raise SystemExit( + f"{name}: dataset years {dataset_years} != the variant's {pre['data_years']}" + ) + return dataset_years + + +def run_world(name: str, reform, years, pre: dict, *, full_use: bool = False): + t0 = time.perf_counter() + source = Path(pre["runtime_dataset_source"]) + before = fiscal.verify_runtime_dataset_sha256( + source, pre["sha256"], phase=f"{name}: immediately before construction" + ) + sim = ab.build_sim(reform) + dataset_years = check_runtime(sim, name, pre) + supplied = dataset_inputs(sim) + check = set_full_use(sim, years) if full_use else None + blocks = {str(y): year_block(sim, y) for y in years} + metadata = variable_metadata(sim.tax_benefit_system) + after = fiscal.verify_runtime_dataset_sha256( + source, + pre["sha256"], + phase=f"{name}: after the world's reads", + before_sha256=before, + ) + del sim + gc.collect() + print(f"{name}: {len(years)} years in {time.perf_counter() - t0:.0f}s", flush=True) + out = { + "reform_dict": reform, + "years": blocks, + "runtime_dataset_source": fiscal.relative_to_root(source), + "dataset_sha256_before": before, + "dataset_sha256_after": after, + "dataset_years": dataset_years, + "dataset_inputs": supplied, + } + if full_use: + out["input_override"] = { + "variable": HOURS_USAGE, + "value": FULL_USE_HOURS, + "set_for_years": dataset_years, + "dependents_calculated_before_override": check.pop( + "dependents_calculated_before_override" + ), + "post_set_weighted_mean": {str(k): v for k, v in check.items()}, + } + return out, metadata + + +def reserved_run_id(run_id: str) -> bool: + """The view's LIKE match: the prefix, ASCII letters in any case.""" + head = run_id[: len(RESERVED_RUN_PREFIX)] + return head.translate(_ASCII_LOWER) == RESERVED_RUN_PREFIX + + +def default_run_id(variant_id: str, started: dt.datetime) -> str: + """Distinct across variants, and across one variant's runs started in + different seconds. Ingest refuses two artifacts that share a run_id.""" + return f"childcare-100k-{variant_id}-{started:%Y%m%dT%H%M%SZ}" + + +def main(argv: list[str] | None = None) -> int: + started = dt.datetime.now(dt.timezone.utc) + ap = argparse.ArgumentParser(description=__doc__) + ap.add_argument("--years", nargs="*", type=int, default=list(YEARS)) + ap.add_argument( + "--diagnostic-years", nargs="*", type=int, default=list(DIAGNOSTIC_YEARS) + ) + ap.add_argument( + "--run-id", help="default: childcare-100k--" + ) + ap.add_argument("--output-dir", type=Path, default=OUTPUT_DIR) + ap.add_argument( + "--variant", + help="a certified variant in data/uk/replay_variants.json (default: its default)", + ) + ap.add_argument( + "--dataset-file", + type=Path, + help="a local file holding a materialised variant's certified dataset bytes", + ) + args = ap.parse_args(argv) + # absolute before a variant preflight moves into its runtime directory + args.output_dir = args.output_dir.resolve() + + variant = variants.get_variant(args.variant) + run_id = args.run_id or default_run_id(variant["variant_id"], started) + if reserved_run_id(run_id): + raise SystemExit( + f"--run-id may not start with {RESERVED_RUN_PREFIX!r} in any case" + ) + missing = sorted( + (set(args.years) | set(args.diagnostic_years)) - set(variant["data_years"]) + ) + if missing: + # a year outside the dataset calculates defaults without raising + raise SystemExit(f"{missing} are outside {variant['variant_id']}'s data years") + if variants.is_legacy(variant): + if args.dataset_file is not None: + raise SystemExit("the legacy variant reads its pinned Hugging Face cache") + # Engine-free and before any engine import: the record is certified + # and the installed wheels are its own. Then the event runner's + # legacy preflight: offline, isolated, the managed release equal to + # certified_bundle.json and to the record, import origins checked, + # and a verified writable mirror that POLICYENGINE_UK_DATA_REPO + # points at. + installed = event.check_variant_environment(variant) + pre = { + **event.preflight(variant), + "data_years": list(variant["data_years"]), + } + else: + # The same checks for a variant whose dataset is materialised into + # its runtime directory (the event runner's variant preflight). + pre = event.variant_preflight( + variant, + dataset_source=args.dataset_file.resolve() if args.dataset_file else None, + ) + installed = pre["installed"] + revision = fiscal.data_bundle_id(pre["release_bundle"]) + identity = variant["identity"] + if (revision, pre["sha256"]) != ( + identity["dataset"]["revision"], + identity["dataset"]["sha256"], + ): + raise SystemExit("the managed release and the variant's record disagree") + print( + f"certified artifact {revision} sha256 verified at " + f"{fiscal.relative_to_root(Path(pre['runtime_dataset_source']))}", + flush=True, + ) + executed = worlds(args.years) + out_worlds = {} + metadata = None + for name, reform in executed.items(): + out_worlds[name], metadata = run_world(name, reform, args.years, pre) + diag = {} + if args.diagnostic_years: + dworlds = worlds(args.diagnostic_years) + for name in ("baseline", "free_hours"): + diag[f"{name}__full_use_hours"], _ = run_world( + f"{name}__full_use_hours", + dworlds[name], + args.diagnostic_years, + pre, + full_use=True, + ) + variants.verify_import_origins() + packages = { + p: { + "version": installed["packages"][p]["version"], + "record_sha256": installed["packages"][p]["record_sha256"], + } + for p in variants.PACKAGES + } + artifact = { + "schema_version": 2, + "run_id": run_id, + "computed_at": dt.datetime.now(dt.timezone.utc).isoformat(), + "variant_id": variant["variant_id"], + "variant_identity_sha256": variant["identity_sha256"], + "engine_version": packages["policyengine-uk"]["version"], + "packages": packages, + "python": installed["python"], + "packaged_bundle_manifest_sha256": installed["packaged_bundle_manifest_sha256"], + "release_bundle": pre["release_bundle"], + "data_bundle": revision, + "dataset": { + "repo_id": identity["dataset"]["repo_id"], + "revision": revision, + "sha256": pre["sha256"], + "hf_commit": variants_hf_commit() if variants.is_legacy(variant) else None, + }, + "years": args.years, + "fy_rule": "PE year Y is fiscal year Y-(Y+1): policyengine_uk/utils/parameters.py convert_to_fiscal_year_parameters sets every gov parameter for year Y to its 30 April Y value", + "limit_sentinel": ab.NO_LIMIT, + "behavioural_response": "none (static)", + "worlds": out_worlds, + "diagnostics": diag, + "variable_metadata": metadata, + } + args.output_dir.mkdir(parents=True, exist_ok=True) + path = args.output_dir / f"{variant['variant_id']}.json" + path.write_text(json.dumps(artifact, indent=1, sort_keys=True) + "\n") + print(f"wrote {path.relative_to(ROOT)}") + return 0 + + +def variants_hf_commit() -> str: + from pipeline import modal_uk_event + + return modal_uk_event.HF_COMMIT + + +if __name__ == "__main__": + sys.exit(main()) diff --git a/results/uk/childcare_100k/construction_annotations.json b/results/uk/childcare_100k/construction_annotations.json new file mode 100644 index 000000000..022caaed4 --- /dev/null +++ b/results/uk/childcare_100k/construction_annotations.json @@ -0,0 +1,38 @@ +{ + "_what": "Notes on how the certified engine builds the PolicyEngine counterpart in lane uk-childcare-100k (CenTax and CBP claims alike), and what was verified about each certified pair the lane has run. Each note cites the engine file it was read from and names, in variants, the pairs it was checked on; it attaches only to those pairs' results. verified_pairs holds each pair's own statements; ingest refuses an artifact whose pair is not listed, so no note written for one pair attaches to another. The measured numbers that go with them are written onto each result by scorecard_db/ingest_uk_childcare_100k.py from the artifact beside this file, not copied here.", + "annotations": [ + { + "id": "pe-england-only", + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "text": "Both offers are England's. policyengine-uk 2.89.2 tests country == ENGLAND for the extended, universal and targeted hours but not for Tax-Free Childcare, which it applies across the UK. The counterpart sums each head for people in English households (the household's country projected onto its members); ingest checks that no free-hours head moves outside England, and states the Tax-Free Childcare change outside England.", + "basis": "policyengine_uk 2.89.2 variables/gov/dfe/extended_childcare_entitlement/extended_childcare_entitlement_eligible.py, universal_childcare_entitlement/universal_childcare_entitlement_eligible.py and targeted_childcare_entitlement/targeted_childcare_entitlement_eligible.py (in_england); variables/gov/hmrc/tax_free_childcare/ (no country test); pipeline/compute_uk_childcare_100k.py england_masks; results/uk/childcare_100k/.json sums_gbp_england" + }, + { + "id": "pe-universal-switch-off", + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "text": "policyengine-uk 2.89.2 switches a child's universal and targeted hours off once the family is eligible for the extended hours, and values the extended hours at the family's drawn weekly usage (maximum_extended_childcare_hours_usage). The counterpart sums all three entitlements, so the universal hours a newly eligible family loses are netted out rather than double-counted. policyengine-uk#2177 (merged 12 October 2026, after the pin) changed this, closing policyengine-uk#1930.", + "basis": "policyengine_uk 2.89.2 variables/gov/dfe/universal_childcare_entitlement/universal_childcare_entitlement_eligible.py and targeted_childcare_entitlement/targeted_childcare_entitlement_eligible.py (~has_extended_childcare), extended_childcare_entitlement/extended_childcare_entitlement.py (min_ with maximum_extended_childcare_hours_usage); https://github.com/PolicyEngine/policyengine-uk/issues/1930; https://github.com/PolicyEngine/policyengine-uk/pull/2177" + }, + { + "id": "pe-under-ones", + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "text": "The offer starts at 9 months. The engine's extended-hours scale starts at age 0.75, but every age on this pair is a whole number of years, so a child under 1 has age 0 and no child under 1 receives extended hours in any world or year (both measured). microcosm#1126 records the same for the Microcosm release.", + "basis": "policyengine_uk 2.89.2 parameters/gov/dfe/extended_childcare_entitlement/hours.yaml (threshold 0.75 from 2025); results/uk/childcare_100k/.json counts.people_with_fractional_age and counts.children_under_one_receiving_extended (0 in every world and year); https://github.com/PolicyEngine/microcosm/issues/1126" + } + ], + "verified_pairs": { + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z": { + "checked": "2026-10-25", + "static_because": "policyengine-uk 2.89.2's labour-supply elasticities are zero, so employment_income_behavioral_response is zero", + "hours_draw": "microcosm#1126 traces this draw and puts DfE's mean at about 28.5 hours.", + "diagnosis": "microcosm#1126 identifies this draw as the main reason PolicyEngine's free-hours spending is low and derives DfE's mean as about 28.5 hours. The fix merged in microcosm on 16 October 2026 (microcosm#1299); this certified pair was built on 19 June 2026, before it. policyengine.py#685 (merged 21 October 2026) certified a Microcosm pair with the fixed draw, national cut 6cba5d71 with policyengine-uk 2.132.6, which this lane has not run yet (policyengine-scorecard#228). PolicyEngine's published dashboard figure, on policyengine.org from 12 to 21 October 2026, ran on the 2 October Microcosm release that microcosm#1126 examines, with the same draw. Two smaller PolicyEngine-side factors are recorded on the result: the universal-hours switch-off (policyengine-uk#1930, changed after the pin by policyengine-uk#2177) and no hours for children under 1.", + "basis": "every note whose variants name this pair, read against policyengine_uk 2.89.2 at the pin and results/uk/childcare_100k/.json; static_because: the centax-static-vs-static basis; hours_draw and diagnosis: https://github.com/PolicyEngine/microcosm/issues/1126, https://github.com/PolicyEngine/microcosm/pull/1299 (merged 2026-10-16), https://github.com/PolicyEngine/policyengine.py/pull/685 (merged 2026-10-21; policyengine.py 6.11.0, released 2026-10-25, still pins that pair in src/policyengine/data/bundle/uk.trace.tro.jsonld), sources/centax-childcare-100k/pe/published_dashboard.json (live), this variant's dataset revision in data/uk/replay_variants.json (built 2026-06-19T02:37:11Z), and the pe-universal-switch-off and pe-under-ones notes" + } + } +} diff --git a/results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json b/results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json new file mode 100644 index 000000000..a47af7bf5 --- /dev/null +++ b/results/uk/childcare_100k/pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z.json @@ -0,0 +1,1594 @@ +{ + "behavioural_response": "none (static)", + "computed_at": "2026-10-25T20:13:47.333098+00:00", + "data_bundle": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "dataset": { + "hf_commit": "a75a9a831d6b07aaffbd09713f2a1124f5c0f08f", + "repo_id": "policyengine/populace-uk-private", + "revision": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "sha256": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833" + }, + "diagnostics": { + "baseline__full_use_hours": { + "dataset_inputs": { + "childcare_expenses": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "maximum_extended_childcare_hours_usage": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_extended_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_targeted_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_tfc": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_universal_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ] + }, + "dataset_sha256_after": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_sha256_before": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_years": [ + 2023, + 2024, + 2025, + 2026, + 2027, + 2028, + 2029, + 2030 + ], + "input_override": { + "dependents_calculated_before_override": { + "extended_childcare_entitlement": [], + "extended_childcare_entitlement_eligible": [], + "targeted_childcare_entitlement": [], + "universal_childcare_entitlement": [] + }, + "post_set_weighted_mean": { + "2027": 29.999999999999996, + "2029": 29.999999999999996 + }, + "set_for_years": [ + 2023, + 2024, + 2025, + 2026, + 2027, + 2028, + 2029, + 2030 + ], + "value": 30.0, + "variable": "maximum_extended_childcare_hours_usage" + }, + "reform_dict": null, + "runtime_dataset_source": "data/uk/policyengine-local/policyengine_uk_data/storage/populace_uk_2023.h5", + "years": { + "2027": { + "counts": { + "benunits_receiving_extended": 843962.9334253073, + "children_receiving_extended": 902772.134937644, + "children_receiving_targeted": 109997.17004776001, + "children_receiving_universal": 494404.7652310133, + "children_under_one": 811770.268560648, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 657574.6903080121, + "people_receiving_tfc_england": 556458.3579444885, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 29.999999999999996, + "mean_benunits_receiving_extended": 30.0 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 100000.0, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 13000 + }, + "sums_gbp": { + "extended_childcare_entitlement": 9248197846.434986, + "gov_spending": 500215116616.9632, + "gov_tax": 1087041211510.9722, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 572390232.9536195, + "universal_childcare_entitlement": 1796609944.425384 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 9248197846.434986, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 486798668.62185365, + "universal_childcare_entitlement": 1796609944.425384 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568691628, + "would_claim_targeted_childcare": 0.6007047434930658, + "would_claim_tfc": 0.572554624268406, + "would_claim_universal_childcare": 0.5704284221502278 + } + }, + "2029": { + "counts": { + "benunits_receiving_extended": 843440.6454449892, + "children_receiving_extended": 902082.5987404585, + "children_receiving_targeted": 110923.08130383492, + "children_receiving_universal": 500957.0870989561, + "children_under_one": 818603.4279367328, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 653347.6676185802, + "people_receiving_tfc_england": 552430.3794765472, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 29.999999999999996, + "mean_benunits_receiving_extended": 30.0 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 100000.0, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 13100 + }, + "sums_gbp": { + "extended_childcare_entitlement": 9608473104.49916, + "gov_spending": 518573790893.1057, + "gov_tax": 1166473554077.0928, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 557138975.6736407, + "universal_childcare_entitlement": 1893960607.279742 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 9608473104.49916, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 470303946.7636819, + "universal_childcare_entitlement": 1893960607.279742 + }, + "take_up": { + "would_claim_extended_childcare": 0.806652256925277, + "would_claim_targeted_childcare": 0.6007047439941354, + "would_claim_tfc": 0.5725546246889881, + "would_claim_universal_childcare": 0.5704284220637051 + } + } + } + }, + "free_hours__full_use_hours": { + "dataset_inputs": { + "childcare_expenses": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "maximum_extended_childcare_hours_usage": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_extended_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_targeted_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_tfc": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_universal_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ] + }, + "dataset_sha256_after": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_sha256_before": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_years": [ + 2023, + 2024, + 2025, + 2026, + 2027, + 2028, + 2029, + 2030 + ], + "input_override": { + "dependents_calculated_before_override": { + "extended_childcare_entitlement": [], + "extended_childcare_entitlement_eligible": [], + "targeted_childcare_entitlement": [], + "universal_childcare_entitlement": [] + }, + "post_set_weighted_mean": { + "2027": 29.999999999999996, + "2029": 29.999999999999996 + }, + "set_for_years": [ + 2023, + 2024, + 2025, + 2026, + 2027, + 2028, + 2029, + 2030 + ], + "value": 30.0, + "variable": "maximum_extended_childcare_hours_usage" + }, + "reform_dict": { + "gov.dfe.extended_childcare_entitlement.income.limit": { + "2027-01-01.2035-12-31": 1e+100 + } + }, + "runtime_dataset_source": "data/uk/policyengine-local/policyengine_uk_data/storage/populace_uk_2023.h5", + "years": { + "2027": { + "counts": { + "benunits_receiving_extended": 988159.5778554678, + "children_receiving_extended": 1061176.17544353, + "children_receiving_targeted": 109997.17004776001, + "children_receiving_universal": 466532.1504074335, + "children_under_one": 811770.268560648, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 657574.6903080121, + "people_receiving_tfc_england": 556458.3579444885, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 29.999999999999996, + "mean_benunits_receiving_extended": 30.0 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 14980 + }, + "sums_gbp": { + "extended_childcare_entitlement": 10752992875.988857, + "gov_spending": 501618625794.7997, + "gov_tax": 1087041211510.9722, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 572390232.9536195, + "universal_childcare_entitlement": 1695324074.0398438 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 10752992875.988857, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 486798668.62185365, + "universal_childcare_entitlement": 1695324074.0398438 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568691628, + "would_claim_targeted_childcare": 0.6007047434930658, + "would_claim_tfc": 0.572554624268406, + "would_claim_universal_childcare": 0.5704284221502278 + } + }, + "2029": { + "counts": { + "benunits_receiving_extended": 1006805.6732779741, + "children_receiving_extended": 1083586.7581790686, + "children_receiving_targeted": 110923.08130383492, + "children_receiving_universal": 469037.7084082365, + "children_under_one": 818603.4279367328, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 653347.6676185802, + "people_receiving_tfc_england": 552430.3794765472, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 29.999999999999996, + "mean_benunits_receiving_extended": 30.0 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 15280 + }, + "sums_gbp": { + "extended_childcare_entitlement": 11444630856.059612, + "gov_spending": 520289271546.09143, + "gov_tax": 1166473554077.0928, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 557138975.6736407, + "universal_childcare_entitlement": 1773283512.562594 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 11444630856.059614, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 470303946.7636819, + "universal_childcare_entitlement": 1773283512.562594 + }, + "take_up": { + "would_claim_extended_childcare": 0.806652256925277, + "would_claim_targeted_childcare": 0.6007047439941354, + "would_claim_tfc": 0.5725546246889881, + "would_claim_universal_childcare": 0.5704284220637051 + } + } + } + } + }, + "engine_version": "2.89.2", + "fy_rule": "PE year Y is fiscal year Y-(Y+1): policyengine_uk/utils/parameters.py convert_to_fiscal_year_parameters sets every gov parameter for year Y to its 30 April Y value", + "limit_sentinel": 1e+100, + "packaged_bundle_manifest_sha256": "255dbb893f455d078a6bcece87399f0a22478f56bc4c6e526b758c4fd5b4db24", + "packages": { + "policyengine": { + "record_sha256": "8d5eec548a5ce206a8fe59ef018542c701e26c0f927351f92339a7a57591b990", + "version": "5.0.2" + }, + "policyengine-core": { + "record_sha256": "103128a1c2a3397268b65e33b83edddfa8e406a4eb64fb6f4ff2310db421c770", + "version": "3.27.1" + }, + "policyengine-uk": { + "record_sha256": "4628bbb8ad80eec4624fc4c18214c186d1405098fa8cf40894a4437e071e1acc", + "version": "2.89.2" + } + }, + "python": "3.12", + "release_bundle": { + "bundle_id": "uk-5.0.2", + "certified_by": "policyengine.py bundle certification", + "certified_data_artifact_sha256": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "certified_data_build_id": "populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "compatibility_basis": "built_with_model_package", + "country_id": "uk", + "data_build_fingerprint": null, + "data_build_model_git_sha": null, + "data_build_model_version": "2.89.2", + "data_package": "populace-data", + "data_version": "0.1.0", + "default_dataset": "populace_uk_2023", + "default_dataset_uri": "hf://policyengine/populace-uk-private/populace_uk_2023.h5@populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "model_package": "policyengine-uk", + "model_version": "2.89.2", + "policyengine_version": "5.0.2" + }, + "run_id": "childcare-100k-20261025", + "schema_version": 2, + "variable_metadata": { + "adjusted_net_income": { + "default_value": 0, + "defined_for": null, + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/hmrc/income_tax/adjusted_net_income.py", + "entity": "person", + "label": "Taxable income after tax reliefs and before allowances", + "unit": "currency-GBP" + }, + "age": { + "default_value": 40.0, + "defined_for": null, + "definition_period": "year", + "documentation": "Age in years", + "engine_source": "policyengine_uk/variables/input/demographic.py", + "entity": "person", + "label": "age", + "unit": "year" + }, + "childcare_expenses": { + "default_value": 0, + "defined_for": null, + "definition_period": "year", + "documentation": "Total amount spent on childcare", + "engine_source": "policyengine_uk/variables/input/consumption/childcare_expenses.py", + "entity": "person", + "label": "childcare consumption", + "unit": "currency-GBP" + }, + "country": { + "default_value": null, + "defined_for": null, + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/household/demographic/country.py", + "entity": "household", + "label": "Country of the UK", + "unit": "None" + }, + "extended_childcare_entitlement": { + "default_value": 0, + "defined_for": "extended_childcare_entitlement_eligible", + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/dfe/extended_childcare_entitlement/extended_childcare_entitlement.py", + "entity": "benunit", + "label": "annual extended childcare entitlement expenses", + "unit": "currency-GBP" + }, + "extended_childcare_entitlement_eligible": { + "default_value": false, + "defined_for": "would_claim_extended_childcare", + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/dfe/extended_childcare_entitlement/extended_childcare_entitlement_eligible.py", + "entity": "benunit", + "label": "eligibility for extended childcare entitlement", + "unit": "None" + }, + "extended_childcare_entitlement_meets_income_requirements": { + "default_value": false, + "defined_for": null, + "definition_period": "year", + "documentation": "Whether this person meets the income requirements for extended childcare entitlement based on age and income thresholds", + "engine_source": "policyengine_uk/variables/gov/dfe/extended_childcare_entitlement/conditions/extended_childcare_entitlement_meets_income_requirements.py", + "entity": "person", + "label": "Income eligible for the extended childcare entitlement", + "unit": "None" + }, + "gov_spending": { + "default_value": 0, + "defined_for": null, + "definition_period": "year", + "documentation": "Government spending impact in respect of this household.", + "engine_source": "policyengine_uk/variables/gov/gov_spending.py", + "entity": "household", + "label": "government spending", + "unit": "currency-GBP" + }, + "gov_tax": { + "default_value": 0, + "defined_for": null, + "definition_period": "year", + "documentation": "Government tax revenue impact in respect of this household.", + "engine_source": "policyengine_uk/variables/gov/gov_tax.py", + "entity": "household", + "label": "government tax revenue", + "unit": "currency-GBP" + }, + "is_child": { + "default_value": false, + "defined_for": null, + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/household/demographic/is_child.py", + "entity": "person", + "label": "Is a child", + "unit": "None" + }, + "is_child_receiving_extended_childcare": { + "default_value": false, + "defined_for": null, + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/dfe/extended_childcare_entitlement/is_child_receiving_extended_childcare.py", + "entity": "person", + "label": "child is eligible for extended childcare entitlement", + "unit": "None" + }, + "max_free_entitlement_hours_used": { + "default_value": 30.0, + "defined_for": null, + "definition_period": "year", + "documentation": "The maximum weekly hours of free childcare entitlement used by the person", + "engine_source": "policyengine_uk/variables/gov/dfe/max_free_entitlement_hours_used.py", + "entity": "person", + "label": "maximum hours of free childcare entitlement used", + "unit": "None" + }, + "maximum_extended_childcare_hours_usage": { + "default_value": 30.0, + "defined_for": null, + "definition_period": "year", + "documentation": "The maximum number of weekly extended childcare hours that this family uses", + "engine_source": "policyengine_uk/variables/gov/dfe/extended_childcare_entitlement/maximum_extended_childcare_hours_usage.py", + "entity": "benunit", + "label": "maximum extended childcare hours used", + "unit": "None" + }, + "targeted_childcare_entitlement": { + "default_value": 0, + "defined_for": "targeted_childcare_entitlement_eligible", + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/dfe/targeted_childcare_entitlement/targeted_childcare_entitlement.py", + "entity": "person", + "label": "targeted childcare entitlement amount", + "unit": "currency-GBP" + }, + "targeted_childcare_entitlement_eligible": { + "default_value": false, + "defined_for": "would_claim_targeted_childcare", + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/dfe/targeted_childcare_entitlement/targeted_childcare_entitlement_eligible.py", + "entity": "benunit", + "label": "eligibility for targeted childcare entitlement", + "unit": "None" + }, + "tax_free_childcare": { + "default_value": 0, + "defined_for": "tax_free_childcare_eligible", + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/hmrc/tax_free_childcare/tax_free_childcare.py", + "entity": "person", + "label": "government contribution through tax-free childcare", + "unit": "currency-GBP" + }, + "tax_free_childcare_meets_income_requirements": { + "default_value": false, + "defined_for": null, + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/hmrc/tax_free_childcare/conditions/tax_free_childcare_meets_income_requirements.py", + "entity": "person", + "label": "income eligible for the tax-free childcare", + "unit": "None" + }, + "universal_childcare_entitlement": { + "default_value": 0, + "defined_for": "universal_childcare_entitlement_eligible", + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/dfe/universal_childcare_entitlement/universal_childcare_entitlement.py", + "entity": "person", + "label": "universal childcare entitlement amount per year", + "unit": "currency-GBP" + }, + "universal_childcare_entitlement_eligible": { + "default_value": false, + "defined_for": "would_claim_universal_childcare", + "definition_period": "year", + "documentation": null, + "engine_source": "policyengine_uk/variables/gov/dfe/universal_childcare_entitlement/universal_childcare_entitlement_eligible.py", + "entity": "person", + "label": "eligible for universal childcare entitlement", + "unit": "None" + }, + "would_claim_extended_childcare": { + "default_value": true, + "defined_for": null, + "definition_period": "year", + "documentation": "Whether this family would claim extended childcare entitlement if eligible", + "engine_source": "policyengine_uk/variables/gov/dfe/extended_childcare_entitlement/would_claim_extended_childcare.py", + "entity": "benunit", + "label": "would claim extended childcare entitlement", + "unit": "None" + }, + "would_claim_targeted_childcare": { + "default_value": true, + "defined_for": null, + "definition_period": "year", + "documentation": "Whether this family would claim targeted childcare entitlement if eligible", + "engine_source": "policyengine_uk/variables/gov/dfe/targeted_childcare_entitlement/would_claim_targeted_childcare.py", + "entity": "benunit", + "label": "would claim targeted childcare entitlement", + "unit": "None" + }, + "would_claim_tfc": { + "default_value": true, + "defined_for": null, + "definition_period": "year", + "documentation": "Whether this family would claim Tax-Free Childcare if eligible", + "engine_source": "policyengine_uk/variables/gov/hmrc/tax_free_childcare/would_claim_tfc.py", + "entity": "benunit", + "label": "would claim Tax-Free Childcare", + "unit": "None" + }, + "would_claim_universal_childcare": { + "default_value": true, + "defined_for": null, + "definition_period": "year", + "documentation": "Whether this BenUnit would claim universal childcare entitlement if eligible", + "engine_source": "policyengine_uk/variables/gov/dfe/universal_childcare_entitlement/would_claim_universal_childcare.py", + "entity": "benunit", + "label": "would claim universal childcare entitlement", + "unit": "None" + } + }, + "variant_id": "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z", + "variant_identity_sha256": "3f1a33f31ac70a8f0beac2373b6408c8be3366128ec0d0e3c1b716de3880003c", + "worlds": { + "baseline": { + "dataset_inputs": { + "childcare_expenses": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "maximum_extended_childcare_hours_usage": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_extended_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_targeted_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_tfc": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_universal_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ] + }, + "dataset_sha256_after": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_sha256_before": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_years": [ + 2023, + 2024, + 2025, + 2026, + 2027, + 2028, + 2029, + 2030 + ], + "reform_dict": null, + "runtime_dataset_source": "data/uk/policyengine-local/policyengine_uk_data/storage/populace_uk_2023.h5", + "years": { + "2026": { + "counts": { + "benunits_receiving_extended": 850789.8653590679, + "children_receiving_extended": 909072.516887188, + "children_receiving_targeted": 109591.68084144592, + "children_receiving_universal": 491224.9334603548, + "children_under_one": 808777.7939686179, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 671531.6491290182, + "people_receiving_tfc_england": 570314.5973646641, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390542723075, + "mean_benunits_receiving_extended": 15.34339149077211 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 100000.0, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 13020 + }, + "sums_gbp": { + "extended_childcare_entitlement": 4618830588.478893, + "gov_spending": 485663250574.56665, + "gov_tax": 1046803930669.7205, + "targeted_childcare_entitlement": 548184865.1414684, + "tax_free_childcare": 581610665.1900003, + "universal_childcare_entitlement": 1744925346.957487 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 4618830588.478893, + "targeted_childcare_entitlement": 548184865.1414684, + "tax_free_childcare": 496179476.63279283, + "universal_childcare_entitlement": 1744925346.957487 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522569653839, + "would_claim_targeted_childcare": 0.6007047440587696, + "would_claim_tfc": 0.5725546245577701, + "would_claim_universal_childcare": 0.5704284220657957 + } + }, + "2027": { + "counts": { + "benunits_receiving_extended": 843962.9334253073, + "children_receiving_extended": 902772.134937644, + "children_receiving_targeted": 109997.17004776001, + "children_receiving_universal": 494404.7652310133, + "children_under_one": 811770.268560648, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 657574.6903080121, + "people_receiving_tfc_england": 556458.3579444885, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390541809284, + "mean_benunits_receiving_extended": 15.324696395426356 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 100000.0, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 13000 + }, + "sums_gbp": { + "extended_childcare_entitlement": 4698921793.176687, + "gov_spending": 495665840652.964, + "gov_tax": 1087041211510.9722, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 572390232.9536195, + "universal_childcare_entitlement": 1796609944.425384 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 4698921793.176687, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 486798668.62185365, + "universal_childcare_entitlement": 1796609944.425384 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568691628, + "would_claim_targeted_childcare": 0.6007047434930658, + "would_claim_tfc": 0.572554624268406, + "would_claim_universal_childcare": 0.5704284221502278 + } + }, + "2028": { + "counts": { + "benunits_receiving_extended": 841466.2223488092, + "children_receiving_extended": 899851.2818461657, + "children_receiving_targeted": 110437.15839982033, + "children_receiving_universal": 500177.8273534775, + "children_under_one": 815017.3566737771, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 657434.4996615499, + "people_receiving_tfc_england": 557298.5232937336, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390541930215, + "mean_benunits_receiving_extended": 15.324756486898808 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 100000.0, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 13000 + }, + "sums_gbp": { + "extended_childcare_entitlement": 4773423463.507309, + "gov_spending": 504606216699.1529, + "gov_tax": 1125627399884.0078, + "targeted_childcare_entitlement": 576419792.0210685, + "tax_free_childcare": 562335816.4203691, + "universal_childcare_entitlement": 1853937569.584144 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 4773423463.507308, + "targeted_childcare_entitlement": 576419792.0210685, + "tax_free_childcare": 477422242.5407297, + "universal_childcare_entitlement": 1853937569.584144 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522566035516, + "would_claim_targeted_childcare": 0.6007047438456758, + "would_claim_tfc": 0.5725546247129096, + "would_claim_universal_childcare": 0.5704284223016919 + } + }, + "2029": { + "counts": { + "benunits_receiving_extended": 843440.6454449892, + "children_receiving_extended": 902082.5987404585, + "children_receiving_targeted": 110923.08130383492, + "children_receiving_universal": 500957.0870989561, + "children_under_one": 818603.4279367328, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 653347.6676185802, + "people_receiving_tfc_england": 552430.3794765472, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390542601741, + "mean_benunits_receiving_extended": 15.25951363302705 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 100000.0, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 13100 + }, + "sums_gbp": { + "extended_childcare_entitlement": 4850984036.267317, + "gov_spending": 513816301823.54456, + "gov_tax": 1166473554077.0928, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 557138975.6736407, + "universal_childcare_entitlement": 1893960607.279742 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 4850984036.267317, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 470303946.7636819, + "universal_childcare_entitlement": 1893960607.279742 + }, + "take_up": { + "would_claim_extended_childcare": 0.806652256925277, + "would_claim_targeted_childcare": 0.6007047439941354, + "would_claim_tfc": 0.5725546246889881, + "would_claim_universal_childcare": 0.5704284220637051 + } + }, + "2030": { + "counts": { + "benunits_receiving_extended": 819202.4852102995, + "children_receiving_extended": 878108.328037858, + "children_receiving_targeted": 110668.70008111, + "children_receiving_universal": 498650.2043390274, + "children_under_one": 822287.1459171176, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 652124.2098141834, + "people_receiving_tfc_england": 550752.7940165997, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.20439053965408, + "mean_benunits_receiving_extended": 15.430488814263628 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 100000.0, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 13020 + }, + "sums_gbp": { + "extended_childcare_entitlement": 4865927077.297628, + "gov_spending": 521430338965.97986, + "gov_tax": 1209433399914.0254, + "targeted_childcare_entitlement": 600962653.349386, + "tax_free_childcare": 560539512.2327334, + "universal_childcare_entitlement": 1922939694.2282991 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 4865927077.297628, + "targeted_childcare_entitlement": 600962653.349386, + "tax_free_childcare": 472147602.9521218, + "universal_childcare_entitlement": 1922939694.2282991 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568008421, + "would_claim_targeted_childcare": 0.6007047441439074, + "would_claim_tfc": 0.5725546245110287, + "would_claim_universal_childcare": 0.5704284224929835 + } + } + } + }, + "free_hours": { + "dataset_inputs": { + "childcare_expenses": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "maximum_extended_childcare_hours_usage": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_extended_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_targeted_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_tfc": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_universal_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ] + }, + "dataset_sha256_after": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_sha256_before": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_years": [ + 2023, + 2024, + 2025, + 2026, + 2027, + 2028, + 2029, + 2030 + ], + "reform_dict": { + "gov.dfe.extended_childcare_entitlement.income.limit": { + "2026-01-01.2035-12-31": 1e+100 + } + }, + "runtime_dataset_source": "data/uk/policyengine-local/policyengine_uk_data/storage/populace_uk_2023.h5", + "years": { + "2026": { + "counts": { + "benunits_receiving_extended": 981372.1325457096, + "children_receiving_extended": 1053449.3776535988, + "children_receiving_targeted": 109591.68084144592, + "children_receiving_universal": 467272.0593005419, + "children_under_one": 808777.7939686179, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 671531.6491290182, + "people_receiving_tfc_england": 570314.5973646641, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390542723075, + "mean_benunits_receiving_extended": 15.339836252951857 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 14840 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5332959142.19636, + "gov_spending": 486292293907.79565, + "gov_tax": 1046803930669.7205, + "targeted_childcare_entitlement": 548184865.1414684, + "tax_free_childcare": 581610665.1900003, + "universal_childcare_entitlement": 1659840135.668403 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5332959142.19636, + "targeted_childcare_entitlement": 548184865.1414684, + "tax_free_childcare": 496179476.63279283, + "universal_childcare_entitlement": 1659840135.668403 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522569653839, + "would_claim_targeted_childcare": 0.6007047440587696, + "would_claim_tfc": 0.5725546245577701, + "would_claim_universal_childcare": 0.5704284220657957 + } + }, + "2027": { + "counts": { + "benunits_receiving_extended": 988159.5778554678, + "children_receiving_extended": 1061176.17544353, + "children_receiving_targeted": 109997.17004776001, + "children_receiving_universal": 466532.1504074335, + "children_under_one": 811770.268560648, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 657574.6903080121, + "people_receiving_tfc_england": 556458.3579444885, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390541809284, + "mean_benunits_receiving_extended": 15.345486788026847 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 14980 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5490642555.481356, + "gov_spending": 496356275526.1062, + "gov_tax": 1087041211510.9722, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 572390232.9536195, + "universal_childcare_entitlement": 1695324074.0398438 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5490642555.481356, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 486798668.62185365, + "universal_childcare_entitlement": 1695324074.0398438 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568691628, + "would_claim_targeted_childcare": 0.6007047434930658, + "would_claim_tfc": 0.572554624268406, + "would_claim_universal_childcare": 0.5704284221502278 + } + }, + "2028": { + "counts": { + "benunits_receiving_extended": 995248.2811945677, + "children_receiving_extended": 1071693.0096548796, + "children_receiving_targeted": 110437.15839982033, + "children_receiving_universal": 468398.27882409096, + "children_under_one": 815017.3566737771, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 657434.4996615499, + "people_receiving_tfc_england": 557298.5232937336, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390541930215, + "mean_benunits_receiving_extended": 15.323384733633484 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 15060 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5641624166.776461, + "gov_spending": 505356624726.9332, + "gov_tax": 1125627399884.0078, + "targeted_childcare_entitlement": 576419792.0210685, + "tax_free_childcare": 562335816.4203691, + "universal_childcare_entitlement": 1736144865.1878035 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5641624166.776461, + "targeted_childcare_entitlement": 576419792.0210685, + "tax_free_childcare": 477422242.5407297, + "universal_childcare_entitlement": 1736144865.1878033 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522566035516, + "would_claim_targeted_childcare": 0.6007047438456758, + "would_claim_tfc": 0.5725546247129096, + "would_claim_universal_childcare": 0.5704284223016919 + } + }, + "2029": { + "counts": { + "benunits_receiving_extended": 1006805.6732779741, + "children_receiving_extended": 1083586.7581790686, + "children_receiving_targeted": 110923.08130383492, + "children_receiving_universal": 469037.7084082365, + "children_under_one": 818603.4279367328, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 653347.6676185802, + "people_receiving_tfc_england": 552430.3794765472, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390542601741, + "mean_benunits_receiving_extended": 15.289917643821507 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 15280 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5810039952.791217, + "gov_spending": 514654680627.84424, + "gov_tax": 1166473554077.0928, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 557138975.6736407, + "universal_childcare_entitlement": 1773283512.562594 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5810039952.791217, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 470303946.7636819, + "universal_childcare_entitlement": 1773283512.562594 + }, + "take_up": { + "would_claim_extended_childcare": 0.806652256925277, + "would_claim_targeted_childcare": 0.6007047439941354, + "would_claim_tfc": 0.5725546246889881, + "would_claim_universal_childcare": 0.5704284220637051 + } + }, + "2030": { + "counts": { + "benunits_receiving_extended": 1025872.7970594168, + "children_receiving_extended": 1102999.3975883722, + "children_receiving_targeted": 109182.81524300575, + "children_receiving_universal": 462291.46033763885, + "children_under_one": 822287.1459171176, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 652124.2098141834, + "people_receiving_tfc_england": 550752.7940165997, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.20439053965408, + "mean_benunits_receiving_extended": 15.306084159042578 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 100000.0 + }, + "records": { + "benunits_receiving_extended": 15420 + }, + "sums_gbp": { + "extended_childcare_entitlement": 6025585886.4349, + "gov_spending": 522441719149.1526, + "gov_tax": 1209433399914.0254, + "targeted_childcare_entitlement": 592893874.2436018, + "tax_free_childcare": 560539512.2327334, + "universal_childcare_entitlement": 1782729840.7795672 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 6025585886.4349, + "targeted_childcare_entitlement": 592893874.2436018, + "tax_free_childcare": 472147602.9521218, + "universal_childcare_entitlement": 1782729840.7795672 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568008421, + "would_claim_targeted_childcare": 0.6007047441439074, + "would_claim_tfc": 0.5725546245110287, + "would_claim_universal_childcare": 0.5704284224929835 + } + } + } + }, + "free_hours_and_tfc": { + "dataset_inputs": { + "childcare_expenses": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "maximum_extended_childcare_hours_usage": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_extended_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_targeted_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_tfc": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ], + "would_claim_universal_childcare": [ + "2023", + "2024", + "2025", + "2026", + "2027", + "2028", + "2029", + "2030" + ] + }, + "dataset_sha256_after": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_sha256_before": "f17306ccb2aad7ff0130be3589b560afb2e2a12a943570911cd0c77f07934833", + "dataset_years": [ + 2023, + 2024, + 2025, + 2026, + 2027, + 2028, + 2029, + 2030 + ], + "reform_dict": { + "gov.dfe.extended_childcare_entitlement.income.limit": { + "2026-01-01.2035-12-31": 1e+100 + }, + "gov.hmrc.tax_free_childcare.income.income_limit": { + "2026-01-01.2035-12-31": 1e+100 + } + }, + "runtime_dataset_source": "data/uk/policyengine-local/policyengine_uk_data/storage/populace_uk_2023.h5", + "years": { + "2026": { + "counts": { + "benunits_receiving_extended": 981372.1325457096, + "children_receiving_extended": 1053449.3776535988, + "children_receiving_targeted": 109591.68084144592, + "children_receiving_universal": 467272.0593005419, + "children_under_one": 808777.7939686179, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 919736.191150099, + "people_receiving_tfc_england": 813812.5363624096, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390542723075, + "mean_benunits_receiving_extended": 15.339836252951857 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 1e+100 + }, + "records": { + "benunits_receiving_extended": 14840 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5332959142.19636, + "gov_spending": 486554068721.5693, + "gov_tax": 1046803930669.7205, + "targeted_childcare_entitlement": 548184865.1414684, + "tax_free_childcare": 843385496.3274453, + "universal_childcare_entitlement": 1659840135.668403 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5332959142.19636, + "targeted_childcare_entitlement": 548184865.1414684, + "tax_free_childcare": 753887179.2919536, + "universal_childcare_entitlement": 1659840135.668403 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522569653839, + "would_claim_targeted_childcare": 0.6007047440587696, + "would_claim_tfc": 0.5725546245577701, + "would_claim_universal_childcare": 0.5704284220657957 + } + }, + "2027": { + "counts": { + "benunits_receiving_extended": 988159.5778554678, + "children_receiving_extended": 1061176.17544353, + "children_receiving_targeted": 109997.17004776001, + "children_receiving_universal": 466532.1504074335, + "children_under_one": 811770.268560648, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 925148.3011144027, + "people_receiving_tfc_england": 817602.842798233, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390541809284, + "mean_benunits_receiving_extended": 15.345486788026847 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 1e+100 + }, + "records": { + "benunits_receiving_extended": 14980 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5490642555.481356, + "gov_spending": 496639075753.5206, + "gov_tax": 1087041211510.9722, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 855190456.9509137, + "universal_childcare_entitlement": 1695324074.0398438 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5490642555.481356, + "targeted_childcare_entitlement": 562866851.850638, + "tax_free_childcare": 763714583.6027803, + "universal_childcare_entitlement": 1695324074.0398438 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568691628, + "would_claim_targeted_childcare": 0.6007047434930658, + "would_claim_tfc": 0.572554624268406, + "would_claim_universal_childcare": 0.5704284221502278 + } + }, + "2028": { + "counts": { + "benunits_receiving_extended": 995248.2811945677, + "children_receiving_extended": 1071693.0096548796, + "children_receiving_targeted": 110437.15839982033, + "children_receiving_universal": 468398.27882409096, + "children_under_one": 815017.3566737771, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 937194.4040367007, + "people_receiving_tfc_england": 828310.1297223568, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390541930215, + "mean_benunits_receiving_extended": 15.323384733633484 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 1e+100 + }, + "records": { + "benunits_receiving_extended": 15060 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5641624166.776461, + "gov_spending": 505664185132.1551, + "gov_tax": 1125627399884.0078, + "targeted_childcare_entitlement": 576419792.0210685, + "tax_free_childcare": 869896259.0417359, + "universal_childcare_entitlement": 1736144865.1878035 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5641624166.776461, + "targeted_childcare_entitlement": 576419792.0210685, + "tax_free_childcare": 775002981.4098558, + "universal_childcare_entitlement": 1736144865.1878033 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522566035516, + "would_claim_targeted_childcare": 0.6007047438456758, + "would_claim_tfc": 0.5725546247129096, + "would_claim_universal_childcare": 0.5704284223016919 + } + }, + "2029": { + "counts": { + "benunits_receiving_extended": 1006805.6732779741, + "children_receiving_extended": 1083586.7581790686, + "children_receiving_targeted": 110923.08130383492, + "children_receiving_universal": 469037.7084082365, + "children_under_one": 818603.4279367328, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 941658.7718813494, + "people_receiving_tfc_england": 831954.693198204, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.204390542601741, + "mean_benunits_receiving_extended": 15.289917643821507 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 1e+100 + }, + "records": { + "benunits_receiving_extended": 15280 + }, + "sums_gbp": { + "extended_childcare_entitlement": 5810039952.791217, + "gov_spending": 514979942266.5021, + "gov_tax": 1166473554077.0928, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 882400604.3274827, + "universal_childcare_entitlement": 1773283512.562594 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 5810039952.791217, + "targeted_childcare_entitlement": 590534607.5321062, + "tax_free_childcare": 785420343.9123458, + "universal_childcare_entitlement": 1773283512.562594 + }, + "take_up": { + "would_claim_extended_childcare": 0.806652256925277, + "would_claim_targeted_childcare": 0.6007047439941354, + "would_claim_tfc": 0.5725546246889881, + "would_claim_universal_childcare": 0.5704284220637051 + } + }, + "2030": { + "counts": { + "benunits_receiving_extended": 1025872.7970594168, + "children_receiving_extended": 1102999.3975883722, + "children_receiving_targeted": 109182.81524300575, + "children_receiving_universal": 462291.46033763885, + "children_under_one": 822287.1459171176, + "children_under_one_receiving_extended": 0.0, + "people_receiving_tfc": 953766.260276176, + "people_receiving_tfc_england": 843568.513585329, + "people_with_fractional_age": 0.0 + }, + "hours_usage": { + "mean_all_benunits": 15.20439053965408, + "mean_benunits_receiving_extended": 15.306084159042578 + }, + "limits_executed": { + "gov.dfe.extended_childcare_entitlement.income.limit": 1e+100, + "gov.hmrc.tax_free_childcare.income.income_limit": 1e+100 + }, + "records": { + "benunits_receiving_extended": 15420 + }, + "sums_gbp": { + "extended_childcare_entitlement": 6025585886.4349, + "gov_spending": 522785230154.9248, + "gov_tax": 1209433399914.0254, + "targeted_childcare_entitlement": 592893874.2436018, + "tax_free_childcare": 904050526.948855, + "universal_childcare_entitlement": 1782729840.7795672 + }, + "sums_gbp_england": { + "extended_childcare_entitlement": 6025585886.4349, + "targeted_childcare_entitlement": 592893874.2436018, + "tax_free_childcare": 805343125.9381161, + "universal_childcare_entitlement": 1782729840.7795672 + }, + "take_up": { + "would_claim_extended_childcare": 0.8066522568008421, + "would_claim_targeted_childcare": 0.6007047441439074, + "would_claim_tfc": 0.5725546245110287, + "would_claim_universal_childcare": 0.5704284224929835 + } + } + } + } + }, + "years": [ + 2026, + 2027, + 2028, + 2029, + 2030 + ] +} diff --git a/scorecard_db/README.md b/scorecard_db/README.md index 1152e8845..10cd2e2eb 100644 --- a/scorecard_db/README.md +++ b/scorecard_db/README.md @@ -507,3 +507,48 @@ this lane's PMD scores by fiscal event from compact, lossless shards. All scores remain accessible. Coverage distinguishes scores with a numerical PE value, recorded findings and scores without a recorded result, so adding unanswered scores does not imply that a model comparison has been completed. + +## 2026-10-21 population: the £100,000 childcare limit (UK) + +Two costings of removing the £100,000 adjusted-net-income limit on the +working-parent childcare offers in England, with PolicyEngine's static +counterparts, from `ingest_uk_childcare_100k` (lane `uk-childcare-100k`): + +```bash +PYTHONPATH=. python -m scorecard_db.ingest_uk_childcare_100k data/scorecard.db +``` + +- **CenTax**, Table 4.2 of "Removing the childcare cliff-edge" (September + 2026, `sources/centax-childcare-100k/`): 24 claims, the static cost, both + behavioural margins and the net cost for three scenarios in 2027-28 and + 2029-30. The free hours only. The 12 bunching counts stay in + `data/externals` with a tallied reason. +- **Centre for British Progress**, Tables 2 and 3 of the "Rates and wrongs" + technical appendix (25 March 2026, `sources/cbp-rates-and-wrongs/`): 12 + claims, the net impact of abolishing the cliff on the free hours and + Tax-Free Childcare, after behaviour, 2026-27 to 2031-32. The 60 taper + cells stay in `data/externals`. + +All claims are `revenue_change` in pounds, held out, keyed to two +`policy_ref` worlds (`measure_key` carries the same slug for retrieval +across producers), and published without a result where PolicyEngine has +none. The reason is in `publication.pe_missing`, read off the years the +certified runs computed; it sits outside the claim id, so a new run never +changes an id, and the feed shows it as `conditions.pe_missing`. The new +`scenario` condition carries CenTax's scenario. + +Results come from `pipeline/compute_uk_childcare_100k.py` on the certified +pair (the default replay variant): 22 rows, `comparable` for CenTax's static +cost and `concept_mismatch` for every net figure. Each status is the one +status-setting entry of the source's `annotations.json` that matches the +row, so no status is a default. A note whose basis is the engine or a run +names the certified pairs it was checked on (`variants`) and attaches only +to their results; ingest refuses an artifact whose pair has no +`verified_pairs` entry in `results/uk/childcare_100k/construction_annotations.json`. +PolicyEngine's published dashboard figure (£0.434bn in 2027-28, on +policyengine.org from 12 to 21 October 2026) is four further result rows, +older than the certified run: the history keeps the published miss beside +the current figure. The central static rows carry a `pe_gap` diagnosis +linked to microcosm#1126 (the hours-usage draw). policyengine.py's current +certified pair, which has the fixed draw, is not yet run in this lane +(#228). See `sources/centax-childcare-100k/README.md`. diff --git a/scorecard_db/build_db.py b/scorecard_db/build_db.py index 198a14389..5bc047c18 100644 --- a/scorecard_db/build_db.py +++ b/scorecard_db/build_db.py @@ -41,6 +41,8 @@ produce_uk + campaign_uk archive-resolved UK reckoner attaches uk_benefit_uprating September 2026 CPI claims + pe-uk benefit-rate read-backs (also on the HMT and RF uprating claims) + uk_childcare_100k CenTax + CBP costings of removing the £100,000 + childcare limit, certified-pair counterparts nz_budget_scores New Zealand Treasury Budget score tables (official-only) be_pit_reform Belgian PIT-reform claims + Axiom result attachments be_jrc JRC EUROMOD-BE model claims + honest demo attachments; @@ -86,6 +88,7 @@ ingest_lpc_minimum_wage, ingest_uk_ab2025, ingest_uk_benefit_uprating, + ingest_uk_childcare_100k, ingest_uk_thinktanks, ingest_us_admin_outturns, ingest_ons_etb, @@ -297,6 +300,10 @@ def build(db_path: Path) -> dict: "uk_benefit_uprating", lambda: ingest_uk_benefit_uprating.ingest(db_path), ), + # The £100,000 childcare limit costings: CenTax (Table 4.2) and the + # Centre for British Progress appendix, with the certified pair's + # static counterparts and the published dashboard figure as history. + ("uk_childcare_100k", lambda: ingest_uk_childcare_100k.ingest(db_path)), ("nz_budget_scores", lambda: ingest_official_budget_scores.ingest(db_path)), ("be_pit_reform", lambda: ingest_pit_reform_2026.ingest(db_path)), ("be_jrc", lambda: ingest_jrc_country_report.ingest(db_path)), diff --git a/scorecard_db/export_populations.py b/scorecard_db/export_populations.py index 3908c1733..3e8eda7e3 100644 --- a/scorecard_db/export_populations.py +++ b/scorecard_db/export_populations.py @@ -278,6 +278,11 @@ def export( publication = json.loads(c["publication"]) reform = json.loads(c["reform_json"]) conditions = json.loads(c["conditions"]) + # A reason that depends on which PolicyEngine runs exist rides in + # publication, outside the claim id (lane uk-childcare-100k); the + # feed states it where every other lane's reason sits. + if "pe_missing" in publication and "pe_missing" not in conditions: + conditions["pe_missing"] = publication["pe_missing"] if publication.get("registry") in PMD_REGISTRIES: event = conditions["fiscal_event"] archive_events[c["claim_id"]] = ( diff --git a/scorecard_db/ingest_uk_childcare_100k.py b/scorecard_db/ingest_uk_childcare_100k.py new file mode 100644 index 000000000..21a3e6750 --- /dev/null +++ b/scorecard_db/ingest_uk_childcare_100k.py @@ -0,0 +1,1330 @@ +"""Ingest the £100,000 childcare limit costings and their PolicyEngine +counterparts (lane uk-childcare-100k). + +Two external scores of removing the £100,000 adjusted-net-income limit on +the working-parent childcare offers in England: + + sources/centax-childcare-100k CenTax, "Removing the childcare + cliff-edge" (September 2026), Table 4.2: the free hours only, on + HMRC administrative data; static cost, two behavioural margins and + the net cost, for three scenarios in 2027-28 and 2029-30 + sources/cbp-rates-and-wrongs Centre for British Progress, "Rates and + wrongs" technical appendix (25 March 2026), Tables 2-3: the free + hours and Tax-Free Childcare on a synthetic population; net of + behaviour only, two behavioural models, 2026-27 to 2031-32 + +Inputs: the adapters' tidy rows (data/externals/.json, QC'd there), +each source's annotations.json, every certified-pair artifact +results/uk/childcare_100k/.json (pipeline/compute_uk_childcare_100k.py), +each checked against its own record in data/uk/replay_variants.json, with +their construction_annotations.json, and the published dashboard figure in +sources/centax-childcare-100k/pe/published_dashboard.json. + +Claims: 24 CenTax money cells and 12 CBP "Abolish Cliff" cells, all +revenue_change in pounds, the value as published; the printed form of each +value is publication provenance, outside the claim id. Tallied drops (still in +data/externals): CenTax's 12 bunching counts (a behavioural-model count of +parents with no registered metric; minting one waits for a counterpart) +and CBP's 60 taper cells (a per-child taper replacing the cliff, which no +PolicyEngine construction in this lane executes). + +Results: + - certified pair (the default replay variant, data/uk/certified_bundle.json): + CenTax static rows (6) are `comparable`, CenTax net rows (6) and CBP + rows (10) `concept_mismatch`; the status comes from the source's + annotations.json, and exactly one status-setting annotation must match. + Both offers are England's, so each counterpart sums the childcare heads + for people in English households (the engine applies Tax-Free + Childcare UK-wide). + - earlier certified pairs: an artifact for another certified variant stays + in results/uk/childcare_100k/ beside a newer one, and its run stays in + the history; each claim's latest is the newest run that computed its + year. A note + whose basis is the engine or a run names the pairs it was checked on + (`variants`) and attaches only to their results; an artifact whose pair + has no verified_pairs entry in construction_annotations.json is refused. + - published dashboard history (4): PolicyEngine's published 30-hours cost + (£0.434bn in 2027-28, £0.505bn in 2029-30; on policyengine.org from + 12 to 21 October 2026, then taken down while the dashboard's costing + is revised; what the dashboard has served since is recorded on the + row, not attached) on the CenTax central + static and net rows, the comparison the dashboard itself drew: + `constructed` (static; the dashboard's local corrections) and + `concept_mismatch` (net). It is older than the certified run, so the + comparisons view and the app's latest point at the certified run and + the published miss stays in the history beside it. +CenTax's behavioural margins (12) and CBP's 2031-32 cells (2) carry no +result; they are published without one and say why in ``pe_missing``, read +off the runs that exist (the years each computed). It rides in publication, +outside the claim id, so a new run never changes a claim's id; the feed +shows it as conditions.pe_missing. + +Diagnosis: the CenTax central static rows carry a pe_gap diagnosis linked +to microcosm#1126 (the hours-usage draw), with the evidence measured on the +claim's latest run and the history that run's pair's verified_pairs entry +states. + +calibration_relationship comes from relationships.uk_relationship: held out +for both sources. Each is a reform estimate, and no calibration target is a +reform-world quantity (tests/test_uk_childcare_100k.py checks the built DB). + + PYTHONPATH=. python -m scorecard_db.ingest_uk_childcare_100k data/scorecard.db +""" + +from __future__ import annotations + +import datetime as dt +import json +import math +import re +import string +from collections import Counter +from pathlib import Path + +from .db import DIAGNOSES_SQL, LANE_SQL, RESULTS_SQL, SCORES_SQL, ScorecardDB +from .harvest import REPO, finish, policy_ref, require_fields +from .models import ( + BASELINE, + STANDARD_CONDITIONS, + ComparisonStatus, + DiagnosisClass, + ExternalScore, + Metric, + PEResult, + TimeBasis, + UnitConcept, +) +from .relationships import uk_relationship + +EXTERNALS = REPO / "data" / "externals" +CENTAX_DIR = REPO / "sources" / "centax-childcare-100k" +CBP_DIR = REPO / "sources" / "cbp-rates-and-wrongs" +RESULTS_DIR = REPO / "results" / "uk" / "childcare_100k" +VARIANTS = REPO / "data" / "uk" / "replay_variants.json" +CERTIFIED = REPO / "data" / "uk" / "certified_bundle.json" +CONSTRUCTION = RESULTS_DIR / "construction_annotations.json" +# The artifact schema this ingest reads (pipeline/compute_uk_childcare_100k.py). +SCHEMA_VERSION = 2 + +CENTAX = "centax" +CBP = "centre_for_british_progress" +SOURCES = (CENTAX, CBP) +SOURCE_MODELS = { + # the spelling the Autumn Budget 2025 harvest already uses for CenTax + CENTAX: "hmrc_administrative_microdata", + CBP: "cbp_synthetic_microsimulation", +} +MEASURES = { + CENTAX: "uk_free_childcare_hours_100k_limit_removed", + CBP: "uk_free_childcare_hours_and_tfc_100k_limit_removed", +} +PROGRAMS = { + CENTAX: "free_childcare_hours", + CBP: "free_childcare_hours_and_tax_free_childcare", +} +# which certified-pair world answers which source +WORLDS = {CENTAX: "free_hours", CBP: "free_hours_and_tfc"} +FREE_HOURS = ( + "extended_childcare_entitlement", + "universal_childcare_entitlement", + "targeted_childcare_entitlement", +) +TFC = "tax_free_childcare" +VARIABLES = {CENTAX: list(FREE_HOURS), CBP: [*FREE_HOURS, TFC]} +HOURS_USAGE_INPUT = "maximum_extended_childcare_hours_usage" +# the extended offer's weekly hours: the full-use diagnostic sets every +# family's usage to them +FULL_USE_HOURS = 30.0 +FULL_USE = { + "baseline": "baseline__full_use_hours", + "free_hours": "free_hours__full_use_hours", +} +# What set_full_use reads before it sets the input: the holders of every +# variable that reads it, each still empty. +FULL_USE_DEPENDENTS = (*FREE_HOURS, "extended_childcare_entitlement_eligible") +# The limits each world records as executed on 30 April, and the value a +# removed limit executes as (compute_uk_ab2025.NO_LIMIT: policyengine-core +# has no infinity in a reform dict). +LIMITS = ( + "gov.dfe.extended_childcare_entitlement.income.limit", + "gov.hmrc.tax_free_childcare.income.income_limit", +) +NO_LIMIT = 1e100 +_PERIOD = re.compile(r"\d{4}-\d{2}-\d{2}\.\d{4}-\d{2}-\d{2}") +# Within a data bundle the comparisons view prefers these runs (db.py: the +# canonical Urban grid), so a run of this lane may not take the prefix: its +# latest result must be its newest run. +RESERVED_RUN_PREFIX = "platform-grid-" +_ASCII_LOWER = str.maketrans(string.ascii_uppercase, string.ascii_lowercase) +# the artifact's sums for people in English households +ENGLAND = "sums_gbp_england" + +LANE_ID = "uk-childcare-100k" +LANE_UPDATED = "2026-10-25" +# The UK family's shared top-level feed literal (sync_lane_feed's contract). +FEED_UPDATED = "2026-08-19" +LANE_FEED_META = { + "source": "CenTax + Centre for British Progress", + "area": "£100,000 childcare limit (30 funded hours, Tax-Free Childcare)", + "mode": 2, + "country": "UK", +} +REGISTRY_MARK = "uk_childcare_100k" + +CENTAX_PUBLICATION = { + "title": "Removing the childcare cliff-edge: impacts and cost of reform", + "authors": "Advani, Flew, Pepin-Hall and Summers", + "publisher": "CenTax", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "date": "2026-09", + "table": "4.2", + "printed_page": "41", + "vintage": "as published September 2026 (fetched 2026-10-21)", +} +CBP_PUBLICATION = { + "title": "Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper", + "publisher": "Centre for British Progress", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "date": "2026-03-25", + "vintage": "as published 25 March 2026 (fetched 2026-10-21)", +} + +CENTAX_MONEY = ( + "static_cost", + "intensive_margin", + "extensive_margin", + "net_fiscal_cost", +) +CENTAX_ATTACHED = ("static_cost", "net_fiscal_cost") +CENTAX_SIGN = ( + "positive = gain to the Exchequer (published: 'negative = cost to the Exchequer')" +) +SCORING = { + "static_cost": "static", + "net_fiscal_cost": "post_behavioural", +} +# A basis citing either of these rests on the engine or a run, so its note +# names the certified pairs it was checked on (`variants`). +_RUN_BASIS = ("policyengine_uk ", "results/uk/childcare_100k/") + +# the tidy-row core (docs/ARCHITECTURE.md), then each adapter's facets +_CORE_FIELDS = frozenset( + { + "source", + "program", + "metric", + "subgroup", + "variant", + "geography", + "unit_concept", + "period", + "value", + "status", + "source_column", + } +) +_CENTAX_FIELDS = _CORE_FIELDS | { + "column", + "country", + "fy", + "printed_page", + "table", + "tax_year_label", + "value_verbatim", +} +_CBP_FIELDS = _CORE_FIELDS | { + "assumptions", + "country", + "fy", + "pdf_page", + "policy", + "table", + "value_verbatim", +} +_DROPS = { + "centax_bunching_counts": ( + "CenTax's Table 4.2 bunching counts (parents holding income at " + "£100,000, before and after removal) are outputs of its behavioural " + "model with no registered metric here; minting one waits for a " + "counterpart or a second source. The rows stay in data/externals." + ), + "cbp_taper_rows": ( + "CBP's 1%-5% rows replace the cliff with a per-child opt-in taper " + "(section 1.3.6), a reform no PolicyEngine construction in this lane " + "executes. The rows stay in data/externals." + ), +} + + +def _load(adapter_id: str, known: frozenset, source: str) -> list[dict]: + path = EXTERNALS / f"{adapter_id}.json" + if not path.exists(): + raise FileNotFoundError( + f"{path} missing — run python sources/{adapter_id}/adapter.py" + ) + rows = json.loads(path.read_text()) + for row in rows: + require_fields(row, known, source) + missing = _CORE_FIELDS - set(row) + if missing: + raise ValueError(f"{source}: rows lack the tidy core {sorted(missing)}") + identity = (row["source"], row["country"], row["geography"], row["program"]) + if identity != (adapter_id, "UK", "England", PROGRAMS[source]): + raise ValueError(f"{source}: unexpected row identity {row}") + return rows + + +def _check_conditions(cond: dict) -> dict: + unknown = set(cond) - STANDARD_CONDITIONS + if unknown: + raise ValueError(f"conditions outside STANDARD_CONDITIONS: {sorted(unknown)}") + return cond + + +def _revenue_row(row: dict, source: str) -> dict: + if row["metric"] != Metric.REVENUE_CHANGE.value or row["subgroup"] is not None: + raise ValueError(f"{source}: not a whole-population revenue row {row}") + return row + + +def _centax_claim(row: dict, pe_missing: str | None = None) -> ExternalScore: + _revenue_row(row, CENTAX) + component = row["column"] + cond = { + "country": "UK", + "geography": "England", + "fy": row["fy"], + "program": PROGRAMS[CENTAX], + "measure_key": MEASURES[CENTAX], + "component": component, + "scenario": row["variant"], + "benchmark_class": "different_model", + "data_scope": "hmrc_administrative", + "sign_convention": CENTAX_SIGN, + } + if component in SCORING: + cond["scoring_method"] = SCORING[component] + relationship, _ = uk_relationship(CENTAX, Metric.REVENUE_CHANGE) + return ExternalScore( + source=CENTAX, + source_model=SOURCE_MODELS[CENTAX], + metric=Metric.REVENUE_CHANGE, + unit_concept=UnitConcept.GBP, + period=int(row["period"]), + time_basis=TimeBasis.FISCAL_YEAR, + value=float(row["value"]), + conditions=_check_conditions(cond), + reform=policy_ref(MEASURES[CENTAX]), + calibration_relationship=relationship, + source_column=row["source_column"], + publication=_without_result( + { + **CENTAX_PUBLICATION, + "value_verbatim": row["value_verbatim"], + "registry": REGISTRY_MARK, + "publish_without_result": True, + }, + pe_missing, + ), + value_kind="gbp", + status=row["status"], + ) + + +def _without_result(publication: dict, pe_missing: str | None) -> dict: + # Why a claim has no PolicyEngine value depends on which certified runs + # exist (the years they computed, the pair's engine), so the reason rides + # in publication, outside the claim id; export_populations shows it as + # conditions.pe_missing, where every other lane's reason sits. + if pe_missing is not None: + publication["pe_missing"] = pe_missing + return publication + + +def _cbp_claim(row: dict, pe_missing: str | None = None) -> ExternalScore: + _revenue_row(row, CBP) + cond = { + "country": "UK", + "geography": "England", + "fy": row["fy"], + "program": PROGRAMS[CBP], + "measure_key": MEASURES[CBP], + "scoring_method": "post_behavioural", + "behavioral_response": row["variant"], + "benchmark_class": "different_model", + "data_scope": "synthetic_ons_births_spi_2022_23", + # no sign_convention: the appendix states none, so the exporter's + # signs.classify reads "no_stated_convention" (unknown); CenTax's + # reading of the sign is in the cbp-sign-unstated annotation + } + relationship, _ = uk_relationship(CBP, Metric.REVENUE_CHANGE) + return ExternalScore( + source=CBP, + source_model=SOURCE_MODELS[CBP], + metric=Metric.REVENUE_CHANGE, + unit_concept=UnitConcept.GBP, + period=int(row["period"]), + time_basis=TimeBasis.FISCAL_YEAR, + value=float(row["value"]), + conditions=_check_conditions(cond), + reform=policy_ref(MEASURES[CBP]), + calibration_relationship=relationship, + source_column=row["source_column"], + publication=_without_result( + { + **CBP_PUBLICATION, + "table": row["table"], + "pdf_page": str(row["pdf_page"]), + "assumptions": row["assumptions"], + "value_verbatim": row["value_verbatim"], + "registry": REGISTRY_MARK, + "publish_without_result": True, + }, + pe_missing, + ), + value_kind="gbp", + status=row["status"], + ) + + +def check_artifact(art: dict, path: Path, variant: dict) -> None: + """The artifact ran on exactly its variant's certified pairing: the + installed wheels, Python and packaged manifest the run verified, the + runtime release bundle, the dataset each world read, and its years.""" + vid = variant["variant_id"] + identity = variant["identity"] + dataset = identity["dataset"] + if art.get("schema_version") != SCHEMA_VERSION: + raise ValueError( + f"{path.name}: schema_version {art.get('schema_version')} != " + f"{SCHEMA_VERSION}; rerun pipeline/compute_uk_childcare_100k.py" + ) + if (art["variant_id"], path.stem) != (vid, vid): + raise ValueError(f"{path.name}: artifact variant {art['variant_id']} != {vid}") + packages = { + name: {f: pkg[f] for f in ("version", "record_sha256")} + for name, pkg in identity["packages"].items() + } + got = { + "variant_identity_sha256": art["variant_identity_sha256"], + "packages": art["packages"], + "engine_version": art["engine_version"], + "python": art["python"], + "packaged_bundle_manifest_sha256": art["packaged_bundle_manifest_sha256"], + "release_bundle": art["release_bundle"], + "dataset": ( + art["data_bundle"], + art["dataset"]["revision"], + art["dataset"]["sha256"], + ), + } + want = { + "variant_identity_sha256": variant["identity_sha256"], + "packages": packages, + "engine_version": packages["policyengine-uk"]["version"], + "python": identity["python"], + "packaged_bundle_manifest_sha256": identity["packaged_bundle_manifest_sha256"], + "release_bundle": identity["release_bundle"], + "dataset": (dataset["revision"], dataset["revision"], dataset["sha256"]), + } + wrong = sorted(k for k in want if got[k] != want[k]) + if wrong: + raise ValueError(f"{path.name}: not {vid}'s certified pairing: {wrong}") + if variant["layout"] == "legacy_events": + cert = json.loads(CERTIFIED.read_text()) + if (dataset["revision"], dataset["sha256"]) != ( + cert["revision"], + cert["sha256"], + ): + raise ValueError(f"{path.name}: data bundle is not the certified bundle") + for name, world in (*art["worlds"].items(), *art["diagnostics"].items()): + if not ( + world["dataset_sha256_before"] + == world["dataset_sha256_after"] + == dataset["sha256"] + ): + raise ValueError( + f"{path.name} {name}: dataset digest is not the certified one" + ) + if not world.get("runtime_dataset_source"): + raise ValueError(f"{path.name} {name}: no runtime dataset source recorded") + if world["dataset_years"] != variant["data_years"]: + raise ValueError(f"{path.name} {name}: dataset years are not the variant's") + if art["behavioural_response"] != "none (static)": + raise ValueError(f"{path.name}: artifact is not a static run") + + +def load_artifacts( + results_dir: Path = RESULTS_DIR, variants_path: Path = VARIANTS +) -> list[tuple[dict, Path]]: + """Every certified-pair artifact, each checked against its own certified + variant record, oldest first. Each claim's latest result is the newest + run that computed its year, so run ids and UTC computed_at times must be + distinct.""" + index = json.loads(variants_path.read_text()) + records = {v["variant_id"]: v for v in index["variants"]} + out = [] + for path in sorted(results_dir.glob("*.json")): + if path.name == CONSTRUCTION.name: + continue + art = json.loads(path.read_text()) + variant = records.get(art.get("variant_id")) + if variant is None or variant["status"] != "certified": + raise ValueError(f"{path.name}: names no certified variant") + check_artifact(art, path, variant) + out.append((art, path)) + out.sort(key=lambda a: a[0]["computed_at"]) + if not out: + raise FileNotFoundError( + f"no artifact in {results_dir}; run pipeline/compute_uk_childcare_100k.py" + ) + check_history(out) + return out + + +def _utc(stamp: str, where: str) -> None: + """A UTC time in the one form datetime.isoformat() writes for it, so + string order is time order (the database orders computed_at as text, + and '+' sorts before the '.' of a fraction) and two strings differ only + when the times do.""" + when = dt.datetime.fromisoformat(stamp) + if when.utcoffset() != dt.timedelta(0) or when.isoformat() != stamp: + raise ValueError( + f"{where}: {stamp!r} is not a UTC time as datetime.isoformat() writes it" + ) + + +def reserved_run_id(run_id: str) -> bool: + """Does the comparisons view's run_id LIKE 'platform-grid-%' match this + run id? SQLite's LIKE ignores case for ASCII letters only, so only those + are folded here.""" + head = run_id[: len(RESERVED_RUN_PREFIX)] + return head.translate(_ASCII_LOWER) == RESERVED_RUN_PREFIX + + +def check_history(artifacts: list[tuple[dict, Path]]) -> None: + """The run history each stage reads: distinct run ids outside the + reserved prefix, and distinct canonical UTC computed_at times, so each + claim's newest run, here and in the database, is one run.""" + if not artifacts: + raise ValueError("no certified artifact") + for art, path in artifacts: + if reserved_run_id(art["run_id"]): + raise ValueError( + f"{path.name}: run_id {art['run_id']!r} takes the reserved prefix " + f"{RESERVED_RUN_PREFIX!r}" + ) + _utc(art["computed_at"], f"{path.name} computed_at") + if len({a["run_id"] for a, _ in artifacts}) != len(artifacts): + raise ValueError("two artifacts share a run_id") + if len({a["computed_at"] for a, _ in artifacts}) != len(artifacts): + raise ValueError("two artifacts share a computed_at") + + +def load_artifact() -> tuple[dict, Path]: + """The newest certified artifact.""" + return load_artifacts()[-1] + + +def _pe_missing( + source: str, + component: str | None, + period: int, + fy: str, + artifacts: list[tuple[dict, Path]], + notes: dict, +) -> str | None: + """Why a claim has no certified PolicyEngine value, read off the runs + that exist (the years each computed, not the years its dataset holds), + or None when one of them computed it.""" + if source == CENTAX and component not in CENTAX_ATTACHED: + # the pair of the newest run that computed this year, as for a result + runs = [a for a, _ in artifacts] + covering = [ + a for a in runs if str(period - 1) in a["worlds"][WORLDS[CENTAX]]["years"] + ] + newest = max(covering or runs, key=lambda a: a["computed_at"]) + return ( + "behavioural leg: this lane's PolicyEngine counterparts are static " + f"runs ({verified_pair(notes, newest)['static_because']})" + ) + year = period - 1 + if any(str(year) in a["worlds"][WORLDS[source]]["years"] for a, _ in artifacts): + return None + holding = sorted( + a["variant_id"] + for a, _ in artifacts + if year in a["worlds"]["baseline"]["dataset_years"] + ) + if holding: + return ( + f"fiscal year {fy}: no certified run in this lane computed it, though it is " + f"among the data years of {', '.join(holding)} (rerun " + f"pipeline/compute_uk_childcare_100k.py --variant with {year} in --years)" + ) + spans = "; ".join( + f"{a['variant_id']} {ys[0]}-{ys[-1]}" + for a, _ in artifacts + for ys in [a["worlds"]["baseline"]["dataset_years"]] + ) + return ( + f"fiscal year {fy} is outside the data years of every certified pair this " + f"lane has run ({spans}; data/uk/replay_variants.json)" + ) + + +def stage( + notes: dict | None = None, + results_dir: Path = RESULTS_DIR, + variants_path: Path = VARIANTS, +) -> tuple[list[ExternalScore], dict, list[tuple[dict, Path]]]: + """(claims, drop tally, every certified artifact oldest first). Each + artifact's pair must be a verified one; no claim id depends on which + runs exist.""" + notes = notes or load_notes() + artifacts = load_artifacts(results_dir, variants_path) + for art, _ in artifacts: + verified_pair(notes, art) + scores: list[ExternalScore] = [] + drops: Counter = Counter() + for row in _load("centax-childcare-100k", _CENTAX_FIELDS, CENTAX): + if row["column"] in CENTAX_MONEY: + missing = _pe_missing( + CENTAX, row["column"], int(row["period"]), row["fy"], artifacts, notes + ) + scores.append(_centax_claim(row, missing)) + else: + drops["centax_bunching_counts"] += 1 + for row in _load("cbp-rates-and-wrongs", _CBP_FIELDS, CBP): + if row["policy"] == "abolish_cliff": + missing = _pe_missing( + CBP, None, int(row["period"]), row["fy"], artifacts, notes + ) + scores.append(_cbp_claim(row, missing)) + else: + drops["cbp_taper_rows"] += 1 + return finish(scores, LANE_ID), dict(sorted(drops.items())), artifacts + + +# --- annotations ------------------------------------------------------------- + + +def _spec(path: Path) -> list[dict]: + spec = json.loads(path.read_text()) + for a in spec["annotations"]: + if not a.get("basis"): + raise ValueError(f"annotation {a['id']} has no basis") + # a note resting on the engine or a run holds only where it was checked + if any(m in a["basis"] for m in _RUN_BASIS) != bool(a.get("variants")): + raise ValueError( + f"annotation {a['id']}: a note whose basis is the engine or a run " + "names the certified pairs it was checked on in `variants`, and " + "only such a note does" + ) + return spec["annotations"] + + +def load_notes() -> dict: + """Every annotation in the lane, and what was verified about each pair: + each source's annotations.json, the construction notes, and their + verified_pairs (results/uk/childcare_100k/construction_annotations.json).""" + notes = { + CENTAX: _spec(CENTAX_DIR / "annotations.json"), + CBP: _spec(CBP_DIR / "annotations.json"), + "construction": _spec(CONSTRUCTION), + "pairs": json.loads(CONSTRUCTION.read_text())["verified_pairs"], + } + named = { + v + for kind in (CENTAX, CBP, "construction") + for a in notes[kind] + for v in a.get("variants", []) + } + if named - set(notes["pairs"]): + raise ValueError( + f"annotations name pairs with no verified_pairs entry: " + f"{sorted(named - set(notes['pairs']))}" + ) + return notes + + +def verified_pair(notes: dict, art: dict) -> dict: + """The pair's verified statements. An artifact for any other pair is + refused rather than given notes that were checked on another engine.""" + pair = notes["pairs"].get(art["variant_id"]) + if pair is None: + raise ValueError( + f"{art['variant_id']}: no note in this lane was verified on this pair. " + "Read each note whose basis is the engine or a run at this pair's pin, " + "add the variant to the variants of those that hold, write new notes " + "where they don't, and add its verified_pairs entry to " + f"{CONSTRUCTION.relative_to(REPO)}" + ) + return pair + + +def _scoped(spec: list[dict], variant: str | None) -> list[dict]: + """The notes that hold for a result: report-only notes always, and a + note resting on the engine or a run only on a pair it was checked on + (never on the published figure, which is no certified pair).""" + return [a for a in spec if "variants" not in a or variant in a["variants"]] + + +_MATCHERS = { + CENTAX: {"components", "scenarios", "periods", "results"}, + CBP: {"models", "periods", "results"}, +} + + +def _facets(score: ExternalScore) -> dict: + c = score.conditions + if score.source == CENTAX: + return { + "components": c["component"], + "scenarios": c["scenario"], + "periods": score.period, + } + return {"models": c["behavioral_response"], "periods": score.period} + + +def match_annotations( + score: ExternalScore, + result_kind: str, + spec: list[dict], + variant: str | None = None, +) -> tuple[list[str], ComparisonStatus]: + """(texts, status) for one result on `variant`'s pair (None for the + published figure); exactly one status-setting annotation must match, so + a status is always a stated decision.""" + facets = {**_facets(score), "results": result_kind} + texts, statuses = [], [] + for a in _scoped(spec, variant): + match = a["applies_to"] + if set(match) != _MATCHERS[score.source]: + raise ValueError(f"annotation {a['id']}: matcher keys {sorted(match)}") + if any(v is not None and facets[k] not in v for k, v in match.items()): + continue + texts.append(a["text"]) + if "status" in a: + statuses.append((a["id"], a["status"])) + if len(statuses) != 1: + raise ValueError( + f"{score.source} {score.conditions.get('component', '')} {score.period} " + f"{result_kind}: status-setting annotations {statuses} (want exactly one)" + ) + return texts, ComparisonStatus(statuses[0][1]) + + +# --- results ----------------------------------------------------------------- + + +def _bn(x: float) -> str: + return f"£{x / 1e9:+.3f}bn" + + +def _delta( + art: dict, world: str, year: int, variable: str, sums: str = "sums_gbp" +) -> float: + ref = art["worlds"][world]["years"][str(year)][sums][variable] + base = art["worlds"]["baseline"]["years"][str(year)][sums][variable] + return ref - base + + +def counterpart(art: dict, source: str, year: int) -> float: + """PolicyEngine's static revenue_change, oriented to the Exchequer + (negative = cost): minus the change in the counted spending for people + in English households.""" + return -sum( + _delta(art, WORLDS[source], year, v, ENGLAND) for v in VARIABLES[source] + ) + + +def _receiving_mean(block: dict) -> tuple[bool, float | None]: + """(sound, mean): a block's weighted mean hours usage of the families + receiving the extended hours. year_block writes no mean exactly when no + weight receives them, so the record is sound only with a finite mean and + a positive weighted count, or no mean and a zero count.""" + mean = (block.get("hours_usage") or {}).get("mean_benunits_receiving_extended") + weight = (block.get("counts") or {}).get("benunits_receiving_extended") + sound = _finite(weight) and ( + (mean is None and weight == 0) or (_finite(mean) and weight > 0) + ) + return sound, mean + + +def _finite(x) -> bool: + """A finite JSON number (a bool is not one).""" + return type(x) in (int, float) and math.isfinite(x) + + +def receiving_hours(art: dict, year: int) -> float | None: + """The current-law weighted mean hours usage of the families receiving + the extended hours, or None when no weight receives them; an unsound + record (_receiving_mean) is refused.""" + block = art["worlds"]["baseline"]["years"][str(year)] + sound, mean = _receiving_mean(block) + if not sound: + weight = (block.get("counts") or {}).get("benunits_receiving_extended") + raise ValueError( + f"{art['run_id']} {year}: receiving-hours mean {mean!r} with " + f"receiving weight {weight!r}" + ) + return mean + + +def _no_recipients(fy: str) -> str: + return ( + f"no family receives the extended hours under current law in {fy}, so " + "the run has no usage mean for them" + ) + + +def measured_notes(art: dict, source: str, year: int, fy: str, pair: dict) -> list[str]: + world = WORLDS[source] + w = art["worlds"][world]["years"][str(year)] + b = art["worlds"]["baseline"]["years"][str(year)] + d = { + v: _delta(art, world, year, v) + for v in (*FREE_HOURS, TFC, "gov_spending", "gov_tax") + } + eng = {v: _delta(art, world, year, v, ENGLAND) for v in (*FREE_HOURS, TFC)} + # Nothing outside the four childcare heads may move (£1m tolerance on + # sums of order £1e11), and the free-hours-only world leaves + # Tax-Free Childcare where it was; the annotations say both. + heads = sum(d[v] for v in (*FREE_HOURS, TFC)) + if abs(d["gov_spending"] - heads) > 1e6 or abs(d["gov_tax"]) > 1e6: + raise ValueError( + f"{source} {year}: the reform moved spending or tax outside the childcare heads" + ) + if source == CENTAX and abs(d[TFC]) > 1e6: + raise ValueError( + f"{year}: removing the 30-hours limit moved Tax-Free Childcare" + ) + # The engine tests country == ENGLAND for all three free-hours offers, + # so none may move outside England; only Tax-Free Childcare can. + outside = {v: d[v] - eng[v] for v in FREE_HOURS} + if any(abs(x) > 1e6 for x in outside.values()): + raise ValueError(f"{source} {year}: free hours moved outside England {outside}") + # an input counts as the dataset's own only in a year the dataset holds it + supplied = { + v: str(year) in periods + for v, periods in art["worlds"]["baseline"]["dataset_inputs"].items() + } + take_up = b["take_up"] + receiving = receiving_hours(art, year) + notes = [ + ( + f"Measured on the certified pair, {fy} (engine period {year}, the start " + f"year, for claim period {year + 1}), people in English households: " + f"extended hours {_bn(eng['extended_childcare_entitlement'])}, universal " + f"{_bn(eng['universal_childcare_entitlement'])}, targeted " + f"{_bn(eng['targeted_childcare_entitlement'])}, Tax-Free Childcare " + f"{_bn(eng[TFC])} (reform minus current law). UK-wide, gov_spending " + f"moves {_bn(d['gov_spending'])} and gov_tax {_bn(d['gov_tax'])}; " + f"Tax-Free Childcare outside England moves {_bn(d[TFC] - eng[TFC])}, " + "left out because both offers are England's." + ), + ( + "Families receiving the extended hours: " + f"{b['counts']['benunits_receiving_extended'] / 1e6:.2f}m under current " + f"law, {w['counts']['benunits_receiving_extended'] / 1e6:.2f}m with the " + "limit removed." + ), + " ".join( + filter( + None, + ( + ( + "Hours-usage draw: the families receiving the extended hours " + f"use a weighted mean of {receiving:.1f} of " + f"{FULL_USE_HOURS:g} weekly hours under current law " + "(maximum_extended_childcare_hours_usage)." + ) + if receiving is not None + else f"Hours-usage draw: on this pair, {_no_recipients(fy)}.", + pair.get("hours_draw"), + ), + ) + ), + ( + "Take-up and usage the run read: the weighted share of benefit units " + "whose take-up input is true under current law is " + + ", ".join( + f"{v} {take_up[v]:.3f}" + + ("" if supplied.get(v) else " (engine default; not in the dataset)") + for v in sorted(take_up) + ) + + f"; {HOURS_USAGE_INPUT} is " + + ( + "the dataset's own input." + if supplied.get(HOURS_USAGE_INPUT) + else "the engine's default (not in the dataset for this year)." + ) + ), + "Executed limits on 30 April {y}: 30 hours {a:g}, Tax-Free Childcare {t:g} (1e+100 is the " + "repo's no-limit sentinel, compute_uk_ab2025.NO_LIMIT).".format( + y=year, + a=w["limits_executed"][ + "gov.dfe.extended_childcare_entitlement.income.limit" + ], + t=w["limits_executed"]["gov.hmrc.tax_free_childcare.income.income_limit"], + ), + ] + full = full_use_cost(art, year) if source == CENTAX else None + if full is not None: + notes.append( + f"Diagnostic, not attached: with every family at the full {FULL_USE_HOURS:g} hours (the " + f"input set to {FULL_USE_HOURS:g} in both worlds before any calculation) the same reform " + f"costs £{full / 1e9:.3f}bn in {fy} on this pair." + ) + return notes + + +def _full(x) -> bool: + return x is not None and abs(x - FULL_USE_HOURS) < 1e-6 + + +def _in_force(periods, year: int): + """The value a reform dict's period map sets for engine year `year`: on + 30 April of that year (fy_rule), each period read as policyengine-core's + Reform.from_dict reads start.stop, both days included. None unless the + map is well formed and the periods covering the date set one value (the + artifact sorts its keys, so it does not show the order overlapping + periods were applied in).""" + if not isinstance(periods, dict): + return None + on = dt.date(year, 4, 30) + hits = [] + for period, value in periods.items(): + if not (isinstance(period, str) and _PERIOD.fullmatch(period)): + return None + try: + start, stop = (dt.date.fromisoformat(d) for d in period.split(".")) + except ValueError: + return None + if start > stop: + return None + if start <= on <= stop: + hits.append(value) + return hits[0] if hits and all(h == hits[0] for h in hits) else None + + +def _same_reform(reform, main, year: int) -> bool: + """The diagnostic world ran the main world's reform: both none, or the + same limits, each removed (NO_LIMIT) in both on the year's parameter + date.""" + if reform is None or main is None: + return reform is None and main is None + return ( + isinstance(reform, dict) + and isinstance(main, dict) + and bool(main) + and sorted(reform) == sorted(main) + and all( + _in_force(reform[p], year) == NO_LIMIT == _in_force(main[p], year) + for p in main + ) + ) + + +def full_use_cost(art: dict, year: int) -> float | None: + """The reform's free-hours cost with every family at the full 30 hours, + from the run's two full-use worlds, or None when it ran none for this + year. Each world must record what the text states: + - the hours input set to 30 in this year, with every holder that reads + it (FULL_USE_DEPENDENTS) still empty, and read back at 30; + - the main world's reform: the same limits removed on 30 April of the + year, under the repo's NO_LIMIT sentinel; + - the main world's executed limits, read back from the engine on that + date, each at NO_LIMIT exactly when the reform removes it. + Anything else raises, so no note or diagnosis claims full use without + the evidence.""" + diag, y = art["diagnostics"], str(year) + held = {w: y in diag.get(name, {}).get("years", {}) for w, name in FULL_USE.items()} + if not any(held.values()): + return None + for world, name in FULL_USE.items(): + if not held[world]: + raise ValueError( + f"{art['run_id']} {year}: no {name} world beside the other" + ) + d, main = diag[name], art["worlds"][world] + o = d.get("input_override") or {} + deps = o.get("dependents_calculated_before_override") + block = d["years"][y] + hours = block.get("hours_usage") or {} + reform = d.get("reform_dict") + executed = block.get("limits_executed") + removed = set(reform or {}) + receiving = hours.get("mean_benunits_receiving_extended") + evidence = { + "variable": o.get("variable") == HOURS_USAGE_INPUT, + "value": o.get("value") == FULL_USE_HOURS, + "year set": year in o.get("set_for_years", []), + "set first": isinstance(deps, dict) + and sorted(deps) == sorted(FULL_USE_DEPENDENTS) + and all(deps[v] == [] for v in deps), + "read back": _full(o.get("post_set_weighted_mean", {}).get(y)), + "mean, all": _full(hours.get("mean_all_benunits")), + # year_block writes no mean when no weight receives the offer + "mean, receiving": _full(receiving) + or (receiving is None and _receiving_mean(block)[0]), + "sentinel": art.get("limit_sentinel") == NO_LIMIT, + "reform": _same_reform(reform, main.get("reform_dict"), year), + "executed limits": isinstance(executed, dict) + and sorted(executed) == sorted(LIMITS) + and executed == main["years"].get(y, {}).get("limits_executed") + and all((executed[p] == NO_LIMIT) == (p in removed) for p in LIMITS), + } + if not all(evidence.values()): + wrong = sorted(k for k, ok in evidence.items() if not ok) + raise ValueError( + f"{art['run_id']} {name} {year}: the record does not show full use: {wrong}" + ) + return sum( + diag[FULL_USE["free_hours"]]["years"][y]["sums_gbp"][v] + - diag[FULL_USE["baseline"]]["years"][y]["sums_gbp"][v] + for v in FREE_HOURS + ) + + +def _construction(art: dict, path: Path, source: str, year: int) -> str: + world = WORLDS[source] + limits = sorted(art["worlds"][world]["reform_dict"]) + return ( + f"static: the certified world with {', '.join(limits)} set to " + f"{art['limit_sentinel']:g} minus the certified world as served, engine period {year} " + f"(FY {year}-{(year + 1) % 100:02d}); minus the change in " + f"{' + '.join(VARIABLES[source])} for people in English households (microdf " + f"weights); artifact {path.relative_to(REPO)}" + ) + + +def published_result( + score: ExternalScore, spec: list[dict], pub: dict +) -> PEResult | None: + """The published figure, on the central rows: the comparison the + dashboard itself drew (its benchmark text names CenTax's central + static and net costs).""" + if score.conditions["scenario"] != "central": + return None + year = score.period - 1 + value = pub["values_gbp_bn"]["thirty_hours"].get(str(year)) + if value is None: + return None + texts, status = match_annotations(score, "published_dashboard", spec) + after = pub["after"] + return PEResult( + claim_id=score.claim_id(), + computed_value=-value * 1e9, + status=status, + engine_version=pub["engine"]["policyengine_uk"], + data_bundle=pub["dataset"]["release"], + pe_construction=( + f"published: the 30-hours leg of PolicyEngine's dashboard ({pub['page']}), " + f"{pub['values_gbp_bn']['field']}, year {year} (FY {year}-{(year + 1) % 100:02d}); " + f"{pub['repo']}@{pub['commit'][:7]} {pub['results_file']} (sha256 " + f"{pub['results_file_sha256'][:12]}); policyengine {pub['engine']['policyengine']}, " + f"dataset passed unmanaged ({pub['dataset']['management']})" + ), + run_id=pub["run_id"], + computed_at=pub["generated_at"], + annotations=[ + *texts, + f"Published figure: £{value:.3f}bn. {pub['corrections']}", + f"{pub['hours_draw']}.", + ( + f"On {pub['live']['on']} from {pub['live']['from']} to " + f"{pub['live']['until']}: {pub['live']['until_basis']}." + ), + ( + f"Since then {after['pr']} merged ({after['merged']}): the dashboard's " + f"own deployment, {after['served_at']}, serves " + f"£{after['thirty_hours_gbp_bn'][str(year)]:.3f}bn in this year from " + f"{pub['repo']}@{after['commit'][:7]} on {after['release']} with " + f"policyengine-uk {after['engine']['policyengine_uk']}. Not attached: " + f"{after['why_not_attached']}" + ), + ], + baseline_key=BASELINE.baseline_key(), + policyengine_variables=list(FREE_HOURS), + ) + + +def certified_result( + s: ExternalScore, notes: dict, art: dict, path: Path +) -> PEResult | None: + check_history([(art, path)]) + pair = verified_pair(notes, art) + year = s.period - 1 + if str(year) not in art["worlds"][WORLDS[s.source]]["years"]: + return None + vid = art["variant_id"] + texts, status = match_annotations(s, "certified", notes[s.source], vid) + construction = [a["text"] for a in _scoped(notes["construction"], vid)] + return PEResult( + claim_id=s.claim_id(), + computed_value=counterpart(art, s.source, year), + status=status, + engine_version=art["engine_version"], + data_bundle=art["data_bundle"], + pe_construction=_construction(art, path, s.source, year), + run_id=art["run_id"], + computed_at=art["computed_at"], + annotations=[ + *texts, + *construction, + *measured_notes(art, s.source, year, s.conditions["fy"], pair), + ], + baseline_key=BASELINE.baseline_key(), + policyengine_variables=VARIABLES[s.source], + ) + + +def _published() -> dict: + path = CENTAX_DIR / "pe" / "published_dashboard.json" + pub = json.loads(path.read_text()) + _utc(pub["generated_at"], f"{path.name} generated_at") + if reserved_run_id(pub["run_id"]): + raise ValueError(f"{path.name}: run_id takes the reserved prefix") + return pub + + +def results( + scores: list[ExternalScore], + artifacts: list[tuple[dict, Path]], + notes: dict | None = None, +) -> list[PEResult]: + """Every certified artifact's run on each attached claim, each with the + notes verified on its own pair, and the published dashboard figure on + CenTax's central rows, older than all.""" + notes = notes or load_notes() + check_history(artifacts) + for art, _ in artifacts: + verified_pair(notes, art) + pub = _published() + oldest = min(art["computed_at"] for art, _ in artifacts) + out = [] + for s in scores: + if s.source == CENTAX and s.conditions["component"] not in CENTAX_ATTACHED: + continue + for art, path in artifacts: + r = certified_result(s, notes, art, path) + if r is not None: + out.append(r) + if s.source == CENTAX: + hist = published_result(s, notes[CENTAX], pub) + if hist is not None: + if hist.computed_at >= oldest: + raise ValueError( + "the published figure must be older than every certified run" + ) + out.append(hist) + return out + + +def _covering(s: ExternalScore, artifacts: list[tuple[dict, Path]]) -> dict | None: + """The newest certified run that computed this claim's year: its latest + certified result, so the run its diagnosis must come from.""" + year = str(s.period - 1) + runs = [ + art for art, _ in artifacts if year in art["worlds"][WORLDS[s.source]]["years"] + ] + return max(runs, key=lambda a: a["computed_at"]) if runs else None + + +def _diagnosed(s: ExternalScore, art: dict | None, pair: dict | None) -> bool: + """A CenTax central static row whose latest certified result comes from + a run on a pair whose verified_pairs entry states a diagnosis.""" + c = s.conditions + return ( + art is not None + and pair.get("diagnosis") is not None + and (s.source, c.get("component"), c.get("scenario")) + == (CENTAX, "static_cost", "central") + ) + + +def diagnoses( + scores: list[ExternalScore], + artifacts: list[tuple[dict, Path]], + notes: dict | None = None, +) -> list[tuple]: + """pe_gap on the CenTax central static rows, linked to microcosm#1126, + each as the pair of its own latest certified result states it: the + measured lead from that run, then the pair's verified diagnosis (none for + a pair that states none). A newer run that skips a claim's year leaves + that claim's diagnosis with the run that computed it.""" + notes = notes or load_notes() + check_history(artifacts) + pairs = {art["run_id"]: verified_pair(notes, art) for art, _ in artifacts} + rows = [] + for s in scores: + art = _covering(s, artifacts) + pair = pairs[art["run_id"]] if art is not None else None + if not _diagnosed(s, art, pair): + continue + c = s.conditions + year = s.period - 1 + pe = counterpart(art, CENTAX, year) + hours = receiving_hours(art, year) + usage = ( + "the families receiving the extended hours use a weighted mean of " + f"{hours:.1f} of {FULL_USE_HOURS:g} weekly hours (measured)" + if hours is not None + else _no_recipients(c["fy"]) + ) + full = full_use_cost(art, year) + if full is None: + raise ValueError( + f"{art['variant_id']}: the diagnosis needs the full-use diagnostic for " + f"{year}; rerun with it in --diagnostic-years" + ) + rationale = ( + f"PolicyEngine's static cost on the certified pair is £{-pe / 1e9:.3f}bn in {c['fy']}, " + f"{abs(pe / s.value - 1) * 100:.0f}% {'below' if pe > s.value else 'above'} CenTax's " + f"£{-s.value / 1e9:.2f}bn. Our side first: on this pair {usage}, and with every " + f"family at the full {FULL_USE_HOURS:g} hours the same reform costs " + f"£{full / 1e9:.3f}bn (diagnostic run). {pair['diagnosis']}" + ) + rows.append( + ScorecardDB.diagnosis_row( + s.claim_id(), + DiagnosisClass.PE_GAP.value, + rationale, + "https://github.com/PolicyEngine/microcosm/issues/1126", + ) + ) + return rows + + +def expected( + scores: list[ExternalScore], + artifacts: list[tuple[dict, Path]], + notes: dict | None = None, +) -> dict: + """The tally the claims, each run's computed years and the published + file imply, derived without the annotations: a certified run attaches + to each CenTax static or net row and each CBP row whose year it + computed; CenTax's static rows are comparable and every net figure a + concept_mismatch; the published figure attaches to the central static + (constructed) and net (concept_mismatch) rows in its years.""" + notes = notes or load_notes() + check_history(artifacts) + pairs = {art["run_id"]: verified_pair(notes, art) for art, _ in artifacts} + published_years = _published()["values_gbp_bn"]["thirty_hours"] + kinds: Counter = Counter() + statuses: Counter = Counter() + for s in scores: + component = s.conditions.get("component") + if s.source == CENTAX and component not in CENTAX_ATTACHED: + continue + year = str(s.period - 1) + static = component == "static_cost" + for art, _ in artifacts: + if year in art["worlds"][WORLDS[s.source]]["years"]: + kinds["certified"] += 1 + statuses["comparable" if static else "concept_mismatch"] += 1 + if ( + s.source == CENTAX + and s.conditions["scenario"] == "central" + and year in published_years + ): + kinds["published_dashboard"] += 1 + statuses["constructed" if static else "concept_mismatch"] += 1 + return { + "claims": 36, + "claims_by_source": {CENTAX: 24, CBP: 12}, + "results": sum(kinds.values()), + "results_by_kind": dict(kinds), + "statuses": dict(statuses), + "diagnoses": sum( + _diagnosed(s, art, pairs[art["run_id"]] if art else None) + for s in scores + for art in [_covering(s, artifacts)] + ), + "drops": {"cbp_taper_rows": 60, "centax_bunching_counts": 12}, + } + + +def accounting(scores, pe, diag, drops, artifacts) -> dict: + certified = {art["run_id"] for art, _ in artifacts} + return { + "claims": len(scores), + "claims_by_source": dict(Counter(s.source for s in scores)), + "results": len(pe), + "results_by_kind": dict( + Counter( + "certified" if r.run_id in certified else "published_dashboard" + for r in pe + ) + ), + "statuses": dict(Counter(r.status.value for r in pe)), + "diagnoses": len(diag), + "drops": drops, + } + + +def ingest(db_path: Path) -> dict: + notes = load_notes() + scores, drops, artifacts = stage(notes) + art = artifacts[-1][0] + pe = results(scores, artifacts, notes) + diag = diagnoses(scores, artifacts, notes) + counts = accounting(scores, pe, diag, drops, artifacts) + want = expected(scores, artifacts, notes) + if counts != want: + raise ValueError(f"claim accounting drifted: {counts} != {want}") + db = ScorecardDB(db_path) + + from .baselines import register_baselines_txn + from .ingest_harvest import sync_lane_feed + + with db.conn: + old = [ + r[0] + for r in db.conn.execute( + "SELECT claim_id FROM external_scores " + "WHERE json_extract(publication, '$.registry') = ?", + (REGISTRY_MARK,), + ) + ] + for table in ("diagnoses", "pe_results", "external_scores"): + db.conn.executemany( + f"DELETE FROM {table} WHERE claim_id = ?", [(c,) for c in old] + ) + db.conn.executemany(SCORES_SQL, [ScorecardDB.score_row(s) for s in scores]) + db.conn.executemany(RESULTS_SQL, [ScorecardDB.result_row(r) for r in pe]) + db.conn.executemany(DIAGNOSES_SQL, diag) + register_baselines_txn(db) + runs = {a["run_id"] for a, _ in artifacts} + computed = {r.claim_id for r in pe if r.run_id in runs} + pairs = ( + "the certified pair" + if len(artifacts) == 1 + else f"{len(artifacts)} certified pairs, the newest" + ) + detail = ( + f"{len(computed)} of {len(scores)} claims computed on {pairs} " + f"({art['data_bundle']}, policyengine-uk {art['engine_version']}) — " + "CenTax Table 4.2 (24 cells: static cost comparable, net cost concept_mismatch) and CBP's " + "technical appendix (12 abolition cells, concept_mismatch); PolicyEngine's published " + "dashboard figure kept as history" + + ("; pe_gap on the central static rows (microcosm#1126)" if diag else "") + ) + db.conn.execute(LANE_SQL, (LANE_ID, "computed", detail, LANE_UPDATED)) + sync_lane_feed( + db, + REPO / "data" / "lanes.json", + FEED_UPDATED, + lanes={LANE_ID: LANE_FEED_META}, + ) + db.close() + return counts + + +if __name__ == "__main__": + import sys + + out = Path(sys.argv[1] if len(sys.argv) > 1 else "data/scorecard.db") + print(json.dumps(ingest(out), indent=1)) diff --git a/scorecard_db/models.py b/scorecard_db/models.py index 274c3236e..2ddf3557b 100644 --- a/scorecard_db/models.py +++ b/scorecard_db/models.py @@ -528,6 +528,11 @@ class UnitConcept(str, Enum): # publication's own words "horizon", "horizon_note", + # scenario the source's named scenario where one costing is + # published under several assumption sets (CenTax's + # childcare cliff-edge Table 4.2: "central" | + # "high_cost" | "low_cost"); identity-bearing + "scenario", } ) diff --git a/scorecard_db/relationships.py b/scorecard_db/relationships.py index 41453f560..ee6e49f08 100644 --- a/scorecard_db/relationships.py +++ b/scorecard_db/relationships.py @@ -619,6 +619,22 @@ def us_source_relationship(source, program, metric, period=None): } +# Centre for British Progress, "Rates and wrongs" technical appendix (lane +# uk-childcare-100k), read 2026-10-21: the certified populace-uk build's 149 +# calibration targets (microcosm fixture packages/microcosm-data/tests/ +# fixtures/uk_june_2023/populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z/ +# calibration_diagnostics.json) include no childcare entitlement, Tax-Free +# Childcare or CBP quantity, and policyengine-uk 2.89.2 names neither CBP +# nor CenTax anywhere in its package. CenTax's own entry is _AB2025_HELD. +_CBP_HELD = ( + CR.HELD_OUT, + "CBP's synthetic-microsimulation reform estimates are scored, never " + "consumed: none of the certified populace-uk build's 149 targets is a " + "childcare or CBP quantity, and policyengine-uk 2.89.2 reads nothing " + "from CBP (read 2026-10-21)", +) + + def uk_relationship(source, metric, program=None, kind=None): """(CalibrationRelationship, basis) for a UK claim, keyed exactly. @@ -713,6 +729,8 @@ def uk_relationship(source, metric, program=None, kind=None): ) if source in _AB2025_HELD: return _AB2025_HELD[source] + if source == "centre_for_british_progress": + return _CBP_HELD if source == "dwp": return ( CR.HELD_OUT, diff --git a/sources/cbp-rates-and-wrongs/adapter.py b/sources/cbp-rates-and-wrongs/adapter.py new file mode 100644 index 000000000..7d315af29 --- /dev/null +++ b/sources/cbp-rates-and-wrongs/adapter.py @@ -0,0 +1,216 @@ +"""Adapter: Centre for British Progress, "Rates and wrongs" technical +appendix (25 March 2026), Tables 2 and 3 -> tidy rows (lane +uk-childcare-100k). + +Input (raw/): + rates_and_wrongs_technical_appendix.pdf + the technical appendix as published (19 pages), pinned by sha256 + pages_layout.txt + ``pdftotext -layout`` (poppler 26.09.0) of PDF pages 1, 3, 4, 5, 8 + and 10, each under a ``=== PDF page N ===`` header; pinned by sha256 +Output: data/externals/cbp-rates-and-wrongs.json + +Rows: section 1 (free childcare hours and Tax-Free Childcare), Table 2 +(endogenous pension, exogenous labour; PDF page 8) and Table 3 (endogenous +labour, no pension; PDF page 10): "Projected Net Fiscal Impact of Childcare +Entitlements (£bn)" for abolishing the cliff and for tapers of 1% to 5%, +financial years 2026/27 to 2031/32. 2 tables x 6 policies x 6 years = 72 +rows, values in pounds (£bn x 1e9) as printed. The appendix states no sign +rule; the values are carried as printed. + +Years: the column labels are financial years ("2026/27"); ``period`` is the +end year (the repo's UK convention) and ``fy`` the label ("2026-27"). + +QC (any failure raises): + 1. the PDF and the extract match their source.json digests + 2. each table parses to exactly the six policies, in order, each with + six values, under the six year columns 2026/27 .. 2031/32 + 3. each table carries its own model title and the note "full utilisation + of FCWP and £500 utilisation of TFC" + 4. within each year, the published values rise with the taper rate + (abolish < 1% < ... < 5%), the ordering the appendix's model implies + for a surcharge added on top of an abolished cliff (section 1.3.6); + a parse that shifted a column would break it + +Run (stdlib only): + python sources/cbp-rates-and-wrongs/adapter.py +""" + +from __future__ import annotations + +import hashlib +import json +import re +from pathlib import Path + +HERE = Path(__file__).resolve().parent +RAW = HERE / "raw" +OUT = HERE.parent.parent / "data" / "externals" / "cbp-rates-and-wrongs.json" +PDF = RAW / "rates_and_wrongs_technical_appendix.pdf" +PAGES = RAW / "pages_layout.txt" + +TABLES = { + "2": { + "page": 8, + "title": "Table 2: Projected Net Fiscal Impact of Childcare Entitlements (£bn): Endoge-", + "model": "endogenous_pension_exogenous_labour", + "model_label": "Endogenous Pension, Exogenous Labour", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000", + "note": "It also assumes a friction parameter of £1000.", + }, + "3": { + "page": 10, + "title": "Table 3: Projected Net Fiscal Impact of Childcare Entitlements (£bn): Endoge-", + "model": "endogenous_labour_no_pension", + "model_label": "Endogenous Labour, No Pension", + "assumptions": "full utilisation of FCWP and £500 utilisation of TFC; friction parameter £1000; Frisch elasticity 0.3", + "note": "It also assumes a friction parameter of £1000 and Frisch elasticity of 0.3.", + }, +} +POLICIES = ( + ("Abolish Cliff", "abolish_cliff"), + ("1%", "taper_1pct"), + ("2%", "taper_2pct"), + ("3%", "taper_3pct"), + ("4%", "taper_4pct"), + ("5%", "taper_5pct"), +) +YEAR_LABELS = ("2026/27", "2027/28", "2028/29", "2029/30", "2030/31", "2031/32") +NOTE = "This assumes full utilisation of FCWP and £500 utilisation of TFC." +EXPECTED_ROWS = len(TABLES) * len(POLICIES) * len(YEAR_LABELS) +_VALUE = r"\d+\.\d+" + + +def sha256(path: Path) -> str: + return hashlib.sha256(path.read_bytes()).hexdigest() + + +def check_digests() -> None: + """QC 1.""" + spec = json.loads((HERE / "source.json").read_text())["raw_sha256"] + for path in (PDF, PAGES): + got = sha256(path) + if got != spec[path.name]: + raise ValueError( + f"QC failed: {path.name} sha256 {got} != {spec[path.name]}" + ) + + +def pages() -> dict[int, str]: + out: dict[int, str] = {} + current = None + for line in PAGES.read_text().splitlines(keepends=True): + m = re.match(r"=== PDF page (\d+) ===$", line.rstrip("\n")) + if m: + current = int(m.group(1)) + out[current] = "" + continue + if current is None: + raise ValueError("QC failed: extract text before the first page header") + out[current] += line + return out + + +def parse_table(number: str, text: str) -> dict[str, list[tuple[str, float]]]: + """QC 2-3: {policy: [(verbatim, value) per year]}.""" + spec = TABLES[number] + start = text.index(spec["title"]) + end = text.index(NOTE, start) + block = text[start:end] + title = " ".join(text[start : start + 200].split()).replace( + "Endoge- nous", "Endogenous" + ) + if spec["model_label"] not in title: + raise ValueError( + f"QC failed: Table {number} title lacks {spec['model_label']!r}" + ) + header = next(line for line in block.splitlines() if "Taper Rate" in line) + if header.split()[2:] != list(YEAR_LABELS): + raise ValueError(f"QC failed: Table {number} year columns {header.split()[2:]}") + out = {} + body = [line for line in block.splitlines() if re.search(_VALUE, line)] + if len(body) != len(POLICIES): + raise ValueError(f"QC failed: Table {number} has {len(body)} value rows") + for line, (label, key) in zip(body, POLICIES): + stripped = line.strip() + if not stripped.startswith(label): + raise ValueError( + f"QC failed: Table {number} row {stripped!r} is not {label!r}" + ) + cells = re.findall(_VALUE, stripped.removeprefix(label)) + if len(cells) != len(YEAR_LABELS): + raise ValueError( + f"QC failed: Table {number} {label} has {len(cells)} values" + ) + out[key] = [(c, float(c)) for c in cells] + note = " ".join(text[end : end + 300].split()) + if not note.startswith(f"{NOTE} {spec['note']}"): + raise ValueError(f"QC failed: Table {number} note changed: {note[:160]!r}") + return out + + +def check_order(number: str, table: dict) -> None: + """QC 4.""" + for i, label in enumerate(YEAR_LABELS): + column = [table[key][i][1] for _, key in POLICIES] + if column != sorted(column) or len(set(column)) != len(column): + raise ValueError( + f"QC failed: Table {number} {label} not rising with the taper" + ) + + +def tidy(number: str, table: dict) -> list[dict]: + spec = TABLES[number] + rows = [] + for label, key in POLICIES: + for (verbatim, value), year_label in zip(table[key], YEAR_LABELS): + start, end = year_label.split("/") + rows.append( + { + # the tidy-row core (docs/ARCHITECTURE.md); variant is + # CBP's behavioural model + "source": "cbp-rates-and-wrongs", + "program": "free_childcare_hours_and_tax_free_childcare", + "metric": "revenue_change", + "subgroup": None, + "variant": spec["model"], + "geography": "England", + # the costing facets beside it + "country": "UK", + "table": number, + "pdf_page": spec["page"], + "policy": key, + "fy": f"{start}-{end}", + "period": int(start[:2] + end), + "unit_concept": "gbp", + "value": round(value * 1e9), + "value_verbatim": verbatim, + "assumptions": spec["assumptions"], + "status": "ok", + "source_column": ( + f"Technical appendix Table {number} ({spec['model_label']}), " + f"{label}, {year_label}" + ), + } + ) + return rows + + +def run() -> list[dict]: + check_digests() + text = pages() + rows = [] + for number, spec in TABLES.items(): + table = parse_table(number, text[spec["page"]]) + check_order(number, table) + rows += tidy(number, table) + if len(rows) != EXPECTED_ROWS: + raise SystemExit(f"row contract broken: {len(rows)} != {EXPECTED_ROWS}") + OUT.parent.mkdir(parents=True, exist_ok=True) + OUT.write_text(json.dumps(rows, indent=1, ensure_ascii=False) + "\n") + return rows + + +if __name__ == "__main__": + out = run() + print(f"wrote {len(out)} rows to {OUT}") diff --git a/sources/cbp-rates-and-wrongs/annotations.json b/sources/cbp-rates-and-wrongs/annotations.json new file mode 100644 index 000000000..a6b1f1ee9 --- /dev/null +++ b/sources/cbp-rates-and-wrongs/annotations.json @@ -0,0 +1,105 @@ +{ + "_schema": { + "applies_to": "matcher on {models, periods, results}; null = any. models: endogenous_pension_exogenous_labour (Table 2), endogenous_labour_no_pension (Table 3); periods: claim periods (financial-year end years); results: certified", + "variants": "the certified variants (data/uk/replay_variants.json) on which a note resting on the engine or a run was checked; it attaches only to those pairs' results. A note resting on the report alone names none and applies to every run", + "severity": "period | definitional | construction | caution | gap", + "status": "the comparison status this annotation sets, where it sets one; at most one status-setting annotation matches a result", + "basis": "traceability: raw/pages_layout.txt (PDF page), engine files at the certified pin, or the measured artifact results/uk/childcare_100k/.json" + }, + "annotations": [ + { + "id": "cbp-net-after-behaviour", + "applies_to": { + "models": null, + "periods": null, + "results": null + }, + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "severity": "definitional", + "status": "concept_mismatch", + "text": "CBP's figures are net fiscal impacts after a behavioural response: parents choose pension contributions with labour income fixed (Table 2), or choose labour supply with no pension (Table 3), and bunch at £100,000 only when that gains them more than a £1,000 friction. PolicyEngine's counterpart is a static cost (the engine's labour-supply elasticities are zero at the pin), so the two measure different things. CBP publishes no static leg.", + "basis": "raw/pages_layout.txt PDF page 7 (section 1.3.4, the bunching rule with a £1k friction), PDF page 8 (section 1.3.7, Table 2 and note), PDF page 9 (section 1.4.5, the same rule) and PDF page 10 (section 1.4.6, Table 3 and note); policyengine_uk 2.89.2 variables/input/employment_income.py (adds employment_income_behavioral_response) and variables/gov/simulation/labour_supply_response/employment_income_behavioral_response.py (returns 0 when income_elasticity and substitution_elasticity are both 0); parameters/gov/simulation/labour_supply_responses/income_elasticity.yaml and substitution_elasticity.yaml (0 from 2020-01-01), which no world's reform_dict sets; results/uk/childcare_100k/.json behavioural_response 'none (static)' and gov_tax, which moves by under £1m in every world and year (checked at ingest)" + }, + { + "id": "cbp-includes-tfc", + "applies_to": { + "models": null, + "periods": null, + "results": null + }, + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "severity": "definitional", + "text": "CBP's subsidy includes Tax-Free Childcare: its tables assume £500 of Tax-Free Childcare per child (the model description says up to £2,000). The PolicyEngine counterpart therefore removes both £100,000 limits and adds Tax-Free Childcare to the three free-hours entitlements. Both are counted for England only, as CBP's children are England's births: the engine applies Tax-Free Childcare across the UK, and the change outside England is stated on each result and left out.", + "basis": "raw/pages_layout.txt PDF page 4 (section 1.2.1: 'the forecast number of births for England'), PDF page 5 (section 1.2.3: 'Up to £2000 is then added to the subsidy to simulate the effect of Tax Free Childcare') and PDF pages 8 and 10 (table notes: '£500 utilisation of TFC'); results/uk/childcare_100k/.json worlds.free_hours_and_tfc sums_gbp and sums_gbp_england; the pe-england-only construction note" + }, + { + "id": "cbp-full-utilisation", + "applies_to": { + "models": null, + "periods": null, + "results": null + }, + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "severity": "construction", + "text": "CBP assumes every eligible family uses the full working-parent hours, and gives every child two parents and no siblings. PolicyEngine runs the dataset's own families, values the extended hours at each family's weekly usage input (maximum_extended_childcare_hours_usage), and pays each offer only where the family's take-up input is true. Each result states the usage and take-up the run read and whether the dataset or the engine default supplied them.", + "basis": "raw/pages_layout.txt PDF page 5 (section 1.2.2: 'every child has two parents and no siblings') and PDF pages 8 and 10 (table notes: 'full utilisation of FCWP'); policyengine_uk 2.89.2 variables/gov/dfe/extended_childcare_entitlement/extended_childcare_entitlement.py (min_ with maximum_extended_childcare_hours_usage), the eligibility variables' defined_for (would_claim_extended_childcare, would_claim_universal_childcare, would_claim_targeted_childcare, would_claim_tfc; each defaults to true); results/uk/childcare_100k/.json worlds.baseline.dataset_inputs and years.*.take_up and hours_usage" + }, + { + "id": "cbp-synthetic-population", + "applies_to": { + "models": null, + "periods": null, + "results": null + }, + "severity": "caution", + "text": "CBP's population is synthetic: children from ONS births for England, each given two parents whose incomes are drawn from the 2022-23 Survey of Personal Incomes (ages 25 to 44). CenTax's Appendix D says the two estimates differ principally because CenTax observes in administrative data what the synthetic model had to assume, and names the treatment of frictions as the most consequential assumption.", + "basis": "raw/pages_layout.txt PDF pages 4-5 (sections 1.2.1-1.2.2); sources/centax-childcare-100k/raw/pages_layout.txt PDF pages 100-101 (printed pages 99-100, Appendix D)" + }, + { + "id": "cbp-sign-unstated", + "applies_to": { + "models": null, + "periods": null, + "results": null + }, + "severity": "caution", + "text": "The appendix gives no sign rule for its tables, and its values are carried as printed. CenTax reads them as revenue raised. PolicyEngine's value is negative for a cost to the Exchequer.", + "basis": "raw/pages_layout.txt PDF pages 8 and 10 (titles and notes state no sign rule); sources/centax-childcare-100k/raw/pages_layout.txt PDF page 100 (printed page 99: 'CBP estimate that abolishing the threshold would raise between £130 million and £425 million in 2030')" + }, + { + "id": "cbp-centax-quote", + "applies_to": { + "models": [ + "endogenous_labour_no_pension" + ], + "periods": [ + 2030 + ], + "results": null + }, + "severity": "caution", + "text": "CenTax quotes this cell as £425 million for 2030; the appendix prints 0.427 (£427 million) for 2029/30.", + "basis": "raw/pages_layout.txt PDF page 10 (Table 3, Abolish Cliff, 2029/30); sources/centax-childcare-100k/raw/pages_layout.txt PDF page 100 (printed page 99, Appendix D)" + }, + { + "id": "cbp-financial-years", + "applies_to": { + "models": null, + "periods": null, + "results": null + }, + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "severity": "period", + "text": "CBP's columns are financial years (2026/27 is 2026-27). PolicyEngine's annual period 2026 runs the 2026-27 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "basis": "raw/pages_layout.txt PDF pages 8 and 10 (column labels); policyengine_uk 2.89.2 utils/parameters.py convert_to_fiscal_year_parameters, called in tax_benefit_system.py process_parameters" + } + ] +} diff --git a/sources/cbp-rates-and-wrongs/raw/pages_layout.txt b/sources/cbp-rates-and-wrongs/raw/pages_layout.txt new file mode 100644 index 000000000..358686076 --- /dev/null +++ b/sources/cbp-rates-and-wrongs/raw/pages_layout.txt @@ -0,0 +1,370 @@ +=== PDF page 1 === + Technical Appendix + Underlying modelling and analysis for each of the policies in the Rates and + Wrongs paper. + + Centre for British Progress + 2026-03-25 + + +Contents +1 Free Childcare Hours and Tax-Free Childcare 3 + 1.1 Policy Background . . . . . . . . . . . . . . . . . . . . . . . . . . 3 + 1.2 Creating a Synthetic Microsimulation . . . . . . . . . . . . . . . 4 + 1.2.1 The Children . . . . . . . . . . . . . . . . . . . . . . . . . 4 + 1.2.2 The Parents . . . . . . . . . . . . . . . . . . . . . . . . . . 5 + 1.2.3 Inflators and Childcare Subsidy . . . . . . . . . . . . . . . 5 + 1.3 The Exogenous labour and Endogenous Pension Model . . . . . . 5 + 1.3.1 Utility Function . . . . . . . . . . . . . . . . . . . . . . . 5 + 1.3.2 The Budget Constraint . . . . . . . . . . . . . . . . . . . 6 + 1.3.3 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . 6 + 1.3.4 Global Maximization . . . . . . . . . . . . . . . . . . . . . 7 + 1.3.5 Calibration of γ . . . . . . . . . . . . . . . . . . . . . . . . 7 + 1.3.6 Modelling a Taper . . . . . . . . . . . . . . . . . . . . . . 7 + 1.3.7 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + 1.4 Endogenous Labour Supply and No Pension Model . . . . . . . . 8 + 1.4.1 Utility Function . . . . . . . . . . . . . . . . . . . . . . . 8 + 1.4.2 The Budget Constraint . . . . . . . . . . . . . . . . . . . 9 + 1.4.3 Calibration of α . . . . . . . . . . . . . . . . . . . . . . . 9 + 1.4.4 Optimisation . . . . . . . . . . . . . . . . . . . . . . . . . 9 + 1.4.5 Global Maximization . . . . . . . . . . . . . . . . . . . . . 9 + 1.4.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 + +2 Personal Allowance Withdrawal 10 + 2.1 How much bunching? . . . . . . . . . . . . . . . . . . . . . . . . . 10 + 2.2 Policies to Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 + 2.3 Modelling the Policies . . . . . . . . . . . . . . . . . . . . . . . . 13 + 2.3.1 Static Response . . . . . . . . . . . . . . . . . . . . . . . . 13 + 2.3.2 Labour Response . . . . . . . . . . . . . . . . . . . . . . . 14 + 2.3.3 Pension Response . . . . . . . . . . . . . . . . . . . . . . . 15 + + + + + 1 + === PDF page 3 === +1 Free Childcare Hours and Tax-Free Childcare +1.1 Policy Background +Families are currently eligible for free childcare hours mainly through two gov- +ernment schemes: Free Childcare for Working Parents (FCWP) and 15 hours +free childcare for 3 and 4-year-olds. + +FCWP has, since September 2025, entitled families with children aged between +9 months and four years to 30 hours per week of free childcare. This scheme +has a number of requirements but the relevant one for this paper is that both +parents of the relevant child must not have an ”expected adjustable net income +of over £100,000 in the current tax year”. This creates a notch in the tax sys- +tem where households earning just above the threshold can miss out on all of +the free childcare hours. The value of these free childcare hours can be large, +especially for children aged 2 and under, and those living in London (where the +cost of early years is higher). For example, a child born on 1st January 2024 +that is growing up in Westminster was entitled to childcare worth approximately +£16,480.80 in the year 2025/26. + +The other scheme is 15 hours free childcare for 3 and 4-year-olds. This childcare +has no restriction based on income so households where parents have incomes +over £100k are eligible for these 15 hours. The two schemes cannot be claimed +simultaneously so parents eligible for both will only get a maximum of 30 hours +per week of childcare for 3 and 4-year-olds. + +The childcare entitlement of 30 hours per week is only available for 38 weeks +in a year (a total of 1140 hours). The annual hourly entitlement to childcare +varies with the date of birth of the child. The hours can only be claimed from +the start of three dates: 1st January, 1st April and 1st September. The child +must be at least 9 months old on the start date to be eligible for the childcare +for that term, creating discontinuities in entitlement based on date of birth. + +As can be seen in Table 1, the hourly funding rate varies considerably by the +age of the child and the local authority of the providers 1 + + + + + 1 We discuss this more later but this essentially means that there is a much greater disin- + +centive to supply labour for high earners in the most productive regions of the UK. + + + 3 + === PDF page 4 === + Table 1: Maximum Value of Childcare Subsidies by Age and Location + + Hourly Value of + Age of Child Location + Funding Rate Subsidy + Under 2 National Average £11.54 £13,156 + 2 years old National Average £8.53 £9,724 + 3 and 4 year olds National Average £6.06 £3,454 + Under 2 Inner London £15.95 £18,183 + 2 years old Inner London £11.61 £13,235 + 3 and 4 year olds Inner London £8.07 £4,600 +Uses 2025-26 funding rates. Inner London rate is the median for the region. + + + +1.2 Creating a Synthetic Microsimulation +There is not a publicly available dataset with the gross incomes (not adjustable +net incomes) of parents of nursery age children and the implicit childcare sub- +sidy they receive. HMT responded to an FOI in November 2025 confirming +they do not hold information relating to the number of “parents of children +aged between 9 months and 4 years who are impacted by the £100,000 thresh- +old when measured by gross income, rather than adjusted net income.” This is a +crucial piece of evidence which is required to fully estimate the fiscal cost of the +childcare related cliffs; many parents substantially increase their salary sacrifice +pension contributions to ensure that their ANI is below £100k. To demonstrate +the strength of this behaviour change, we have created a synthetic dataset that +recreates known facts about this group of people to see the effect of the tax cliff +and to simulate new policy effects. + + +1.2.1 The Children +The model uses the forecast number of births for England over the next 10 +years and the actual number of births in England over the previous 5 years. +The sources for this are the ONS. + +The children are then assigned a birthday in each year with the probability of +having been born in the UK on any particular day in 2024. This is to ensure +that the seasonal variation in births is picked up in the dataset - entitlement to +early years can sometimes vary by over £10k depending on the date of birth of +the child. + + + + + 4 + === PDF page 5 === +1.2.2 The Parents +The simulation assumes that every child has two parents and no siblings. This is +the most conservative way of modelling the fiscal impact of the policy change as +multiple children increase the total value of the household’s childcare subsidies, +increasing the incentive for the family to cut their incomes to under £100k. +Likewise, the requirement of two parents means that, where both earn over +£100k, it has to be worthwhile for them both to cut their ANI to under £100k, +a greater hurdle. + +Parents’ incomes are assigned using the total incomes in the 2022-23 SPI data +where only those aged 25 to 44 are considered and the sex of each of the partners +is different (both assumptions can be relaxed). The dataset ensures there is a +correlation coefficient between each of the male and female partner’s incomes of +around 0.5. + +The parents are then assigned their child at random. + + +1.2.3 Inflators and Childcare Subsidy +The maximum entitled number of childcare hours in each funding age group +(under 2, 2 year old and 3 & 4) is modelled for each financial year. The national +average funding rate is used to calculate the value of the subsidy in each finan- +cial year. Different funding rates are used as a sensitivity test. The average +rates for all of London are used as well as just Inner London. Both lead to +higher static costs but stronger behavioural change. Up to £2000 is then added +to the subsidy to simulate the effect of Tax Free Childcare. + +Incomes and the Childcare hours funding are increased in line with the OBR’s +forecast for average earnings growth. The tax system keeps in line with the +Government’s current personal taxation plans. + + +1.3 The Exogenous labour and Endogenous Pension Model +We model the household’s response to the tax system as a choice over pension +contributions. The agent treats their gross labour income Z as fixed (exogenous +labour supply) but chooses the optimal allocation between current consumption +(c) and future pension wealth (p) to maximize utility. + +1.3.1 Utility Function +We employ a Quasi-Linear utility function. This specification implies that the +demand for pension savings is driven by the relative price (tax relief) rather +than wealth effects. + + + + 5 + === PDF page 7 === +Step B: Corner Solutions (Bunching) +The agent also evaluates the specific strategy of bunching exactly at the thresh- +old K to retain the benefit B. + + pbunch = max(0, Z − K) (6) + +In this scenario, consumption is fixed at cbunch = K − T (K) + B. + +1.3.4 Global Maximization +The agent compares the indirect utility between the two strategies. They only +choose to bunch at £100k if the utility from doing so is greater than from the +optimal pension saving scenario and a friction value (in our central scenario, +that is £1k). + + ( + ∗ pbunch if U (c(pnormal ), pnormal ) + f riction < U (c(pbunch ), pbunch ) + p = + pnormal else + (7) +Agents only choose to respond if they can secure themselves £1k of extra net +income - this compensates for the planning and tax advice which may be required +to embark on Salary Sacrifice. +This discrete choice framework naturally generates: + • Smooth responses: Agents gradually increasing pensions as γ rises. + • Clumping: A mass of agents setting p = Z − 100, 000 because the utility + jump from the benefit B outweighs the liquidity cost of the extra saving. + +1.3.5 Calibration of γ +The parameter γ determines the baseline savings rate. A conservative estimate +of γ = 0.04 has been used here. Where τ = 0, workers will naturally save 4% of +their income. + +1.3.6 Modelling a Taper +Under a taper policy, the binary cliff edge is replaced by a gradual withdrawal +rate s (e.g., 3%). Agents decide whether they want to opt-in to the childcare +benefits, leading to an indefinite taper above £100k, or to avoid the additional +charge. Mechanically, this acts as a surcharge on the marginal tax rate for +Adjusted Net Income above K.The effective marginal tax rate τef f becomes: + ( + τbase + s if K < (Z − p) + τef f = (8) + τbase otherwise + +The optimization problem remains very similar to Step A (Interior Solutions). +The agent calculates the optimal pension p∗ using the First Order Condition, + + + 7 + === PDF page 8 === +but substituting τef f for τj . They then compare this scenario against one where +the original optimal pension provision is used and where no childcare benefit is +received. Simply, they compare whether it is worth paying the surtax in order +to gain the additional net income from the childcare policy (this is the case for +almost all couples where the s value is small. The higher utility alternative is +chosen. + +Under the Taper, the friction parameter is dropped. + +1.3.7 Results + +Table 2: Projected Net Fiscal Impact of Childcare Entitlements (£bn): Endoge- +nous Pension, Exogenous Labour + + Taper Rate 2026/27 2027/28 2028/29 2029/30 2030/31 2031/32 + Abolish Cliff 0.070 0.077 0.091 0.132 0.147 0.172 + 1% 0.106 0.120 0.136 0.224 0.247 0.282 + 2% 0.150 0.169 0.187 0.282 0.310 0.352 + 3% 0.189 0.212 0.232 0.333 0.365 0.412 + 4% 0.223 0.247 0.271 0.375 0.413 0.465 + 5% 0.253 0.279 0.305 0.413 0.454 0.512 +This assumes full utilisation of FCWP and £500 utilisation of TFC. It also assumes a +friction parameter of £1000. + + + + +1.4 Endogenous Labour Supply and No Pension Model +We model the household’s response to the tax system as a discrete choice over +labour. The agent no longer optimizes their pension contributions (which are +now treated as exogenous or zero); instead, they can choose to reduce their +labour supply in response to the tax cliff edge; they choose h ∈ {1, 0.9, 0.8, 0.7, ...} +to maximize utility. + +1.4.1 Utility Function +We employ a Quasi-Linear utility function with iso-elastic labour supply. This +specification eliminates income effects on labour supply, isolating the substitu- +tion effects driven by the tax notch. + α 1+ 1ε + U (c, h) = c − 1h (9) + 1+ ε +Where: + • c: Net consumption. + + + 8 + === PDF page 9 === + • h: Labour hours (normalized such that h = 1 represents status quo full- + time work). + • Zpot : Potential Earnings Capacity (Exogenous scalar). + • α: Disutility of labour parameter (calibrated individually). + + • ε: Frisch elasticity of labour supply. + +1.4.2 The Budget Constraint +The household faces a budget constraint defined by the tax function T (·) and +the conditional Child Benefit B. The benefit is withdrawn if ”Adjusted Net +Income” (AN I) exceeds the threshold K = £100, 000. + + c = w · h − T (w · h) + B · I(w · h ≤ K) (10) +This creates a non-convex budget set. To solve the optimization problem, we +evaluate two distinct regimes for each household and select the global maximum. + +1.4.3 Calibration of α +To ensure the model reproduces the status quo distribution, we calibrate αi for +each individual i such that their observed choice (hi = 1) is optimal under the +current tax system. + Using the inverse of the labour FOC: + + αi = wi (1 − τi ) (11) + +This ensures that any deviation from the status quo in our counterfactual sim- +ulations is driven purely by changes in the tax incentives τ or the benefit rules +B. + +1.4.4 Optimisation +We find the maximum h that only just brings total income below £100k. + + hbunch = max h | h · Z ≤ 100, 000 (12) + +1.4.5 Global Maximization +This is very similar to the previous model. The agent compares the indirect +utility between the scenario where h = 1 and h = hbunch . They only choose to +bunch at £100k if the utility from doing so is greater than from the normal in- +come (with its associated disutility) and a friction value (in our central scenario, +that is £1k). + +1.4.6 Results + + + + 9 + === PDF page 10 === +Table 3: Projected Net Fiscal Impact of Childcare Entitlements (£bn): Endoge- +nous Labour, No Pension + + Taper Rate 2026/27 2027/28 2028/29 2029/30 2030/31 2031/32 + Abolish Cliff 0.330 0.349 0.399 0.427 0.462 0.507 + 1% 0.398 0.427 0.483 0.518 0.560 0.615 + 2% 0.472 0.508 0.569 0.611 0.662 0.726 + 3% 0.537 0.580 0.646 0.694 0.752 0.825 + 4% 0.595 0.643 0.715 0.767 0.832 0.912 + 5% 0.647 0.700 0.775 0.833 0.903 0.99 +This assumes full utilisation of FCWP and £500 utilisation of TFC. It also assumes a +friction parameter of £1000 and Frisch elasticity of 0.3. + + +2 Personal Allowance Withdrawal +The withdrawal of the Personal Allowance causes substantial bunching at the +£100k kink in the tax system. For an employee, the marginal tax rate increases +from 42% to 62% at this point, disincentivizing work at or around this thresh- +old. There are approximately 20% more people with incomes bunched around +the threshold than would be expected if the density of income remained smooth +at the kink. This causes significant distortions in labour supply and tax revenue. + +This issue is further exacerbated for individuals with undergraduate and post- +graduate student loans, where the effective marginal tax rate can be 15 percent- +age points higher. The 62% tax rate caused by the withdrawal of the Personal +Allowance generally ends at £125,140, where the marginal rate drops to 47%. + +We simulate the fiscal effects of various policies intended to mitigate the im- +pact of this sharp kink in the tax system while also reducing the peak effective +marginal tax rate. + + +2.1 How much bunching? +We use the Survey of Personal Incomes: Public Use Tape 2022-23 to conduct a +bunching analysis of the change in the marginal tax rate at the £100k kink. + +This paper estimates the average compensated elasticity of taxable income with +respect to marginal tax rates, following the methodology of Saez (2010). Esti- +mating the average elasticity is necessary for two reasons. First, agents possess +heterogeneous elasticities, meaning some are more responsive to tax schedule +kinks than others. This heterogeneity is essential to explain why bunching +occurs at specific thresholds while some density remains just above the kink. + + + 10 + \ No newline at end of file diff --git a/sources/cbp-rates-and-wrongs/raw/rates_and_wrongs_technical_appendix.pdf b/sources/cbp-rates-and-wrongs/raw/rates_and_wrongs_technical_appendix.pdf new file mode 100644 index 000000000..abb798a29 Binary files /dev/null and b/sources/cbp-rates-and-wrongs/raw/rates_and_wrongs_technical_appendix.pdf differ diff --git a/sources/cbp-rates-and-wrongs/source.json b/sources/cbp-rates-and-wrongs/source.json new file mode 100644 index 000000000..3cd050be2 --- /dev/null +++ b/sources/cbp-rates-and-wrongs/source.json @@ -0,0 +1,44 @@ +{ + "id": "cbp-rates-and-wrongs", + "name": "Centre for British Progress — Rates and wrongs: technical appendix (25 March 2026), section 1, Tables 2 and 3", + "organization": "Centre for British Progress", + "authors": ["E. Cohen", "M. Stubbs"], + "authors_basis": "CenTax report footnote 46 (sources/centax-childcare-100k/raw/pages_layout.txt, printed page 99); the technical appendix itself names no authors", + "country": "UK", + "claim_class": "reform_score", + "url": "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf", + "data_urls": [ + "https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf" + ], + "report_url_note": "CenTax's footnote 46 cites the main report at https://britishprogress.org/reports/rates-and-wrongs-fixing-uk-tax-cliff-edges-and-tap (truncated in the PDF). On 2026-10-21 that slug, /reports/rates-and-wrongs and /reports/rates-and-wrongs-fixing-uk-tax-cliff-edges-and-tapers returned the site's 'Article Not Found' page, and britishprogress.org/reports carried no link to the report. The technical appendix, CBP's own publication of the numbers (CenTax footnote 47), is what is vendored.", + "raw_sha256": { + "rates_and_wrongs_technical_appendix.pdf": "53697ee68ecc07bf96d2ff387643523c97907fbcee5a7382a3c6d8bbe7774b79", + "pages_layout.txt": "8ae90abbf3a6211887ca29e6ad8c2e5d78328cf26465c8f94594660d8ce98fce" + }, + "pages_extracted": { + "tool": "pdftotext -layout (poppler 26.09.0), one PDF page at a time", + "pdf_pages": [1, 3, 4, 5, 7, 8, 9, 10] + }, + "citation": "Centre for British Progress (2026), Technical Appendix: underlying modelling and analysis for each of the policies in the Rates and Wrongs paper, 25 March 2026.", + "published": "2026-03-25", + "fetched": "2026-10-21", + "pdf_metadata": "19 pages; CreationDate 2026-03-25 (pdfinfo); 359,068 bytes; HTTP 200 on 2026-10-21", + "model": { + "name": "CBP synthetic microsimulation of parents of nursery-age children", + "open_source": false, + "method": "Children from ONS births for England (actual and forecast), each given a birthday and two parents with no siblings; parents' incomes drawn from the 2022-23 Survey of Personal Incomes (ages 25-44, opposite sex, partner correlation about 0.5). The subsidy is the maximum entitled hours at national average funding rates, plus Tax-Free Childcare (up to £2,000 in the model description; the tables assume £500). Two behavioural models are run separately: an endogenous pension choice with fixed labour income (Table 2) and an endogenous labour choice with no pension (Table 3, Frisch elasticity 0.3); in both an agent bunches at £100,000 only if it gains at least £1,000 (the friction).", + "method_basis": "raw/pages_layout.txt: PDF pages 4-5 (sections 1.2.1-1.2.3), 7 and 9 (sections 1.3.4 and 1.4.5, the £1k friction), 8 (section 1.3.7, Table 2) and 10 (section 1.4.6, Table 3)", + "data": "Synthetic: ONS births and the SPI 2022-23 (HMRC's sample of income tax records)" + }, + "reform": "Abolish the £100,000 cliff on Free Childcare for Working Parents and Tax-Free Childcare (the 'Abolish Cliff' row); the 1%-5% rows replace the cliff with a per-child opt-in taper on income above £100,000 (section 1.3.6).", + "baseline_key": "current law with the £100,000 cliff", + "period": "Financial years 2026/27 to 2031/32, labelled as such", + "metrics": ["revenue_change (net fiscal impact, after behaviour)"], + "units": { + "money": "gbp — raw pounds (published in £bn to three decimals), as printed" + }, + "sign_note": "The appendix states no sign rule for its tables. CenTax's Appendix D (printed page 99) reads the abolition figures as raising revenue ('would raise between £130 million and £425 million in 2030'). The values are carried as printed and the sign is left unclassified.", + "geographies": "England (children from England's births)", + "coverage_note": "Tables 2 and 3 in full: 72 cells. The 12 'Abolish Cliff' cells become claims; the 60 taper cells stay in data/externals with a tallied reason (no PolicyEngine construction in this lane).", + "diagnosis_upstream": "CBP's model is not published as code; a divergence classified as an external-model issue becomes a comparison memo." +} diff --git a/sources/centax-childcare-100k/README.md b/sources/centax-childcare-100k/README.md new file mode 100644 index 000000000..d958791c6 --- /dev/null +++ b/sources/centax-childcare-100k/README.md @@ -0,0 +1,173 @@ +# CenTax: removing the £100,000 childcare limit (lane `uk-childcare-100k`) + +CenTax's costing of removing the £100,000 adjusted-net-income limit on the +working-parent free childcare hours in England (Advani, Flew, Pepin-Hall and +Summers, [Removing the childcare cliff-edge](https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf), +September 2026, Table 4.2), set against PolicyEngine's static cost of the +same reform on the scorecard's certified UK pair. The lane also carries the +Centre for British Progress estimate the report reviews +([`../cbp-rates-and-wrongs/`](../cbp-rates-and-wrongs/)). + +## What is here + +| file | what it is | +|---|---| +| `raw/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf` | the report as published (sha256 in `source.json`) | +| `raw/pages_layout.txt` | `pdftotext -layout` of the twelve pages the rows and annotations cite; a test re-extracts them from the PDF | +| `adapter.py` | Table 4.2 → `data/externals/centax-childcare-100k.json` (36 cells), with a sum check, a cross-check against Table 4.3 and a check of the prose under the table | +| `annotations.json` | concept differences, each with its basis; the one that sets a row's status is marked | +| `pe/published_dashboard.json` | PolicyEngine's published figure for the same reform, kept as history | + +The PolicyEngine side is `pipeline/compute_uk_childcare_100k.py`, which wrote +`results/uk/childcare_100k/.json`; the ingest is +`scorecard_db/ingest_uk_childcare_100k.py`. + +## Years and scope + +- CenTax names tax years by their later year: 2028 is 2027-28 and 2030 is + 2029-30 (editorial note, printed page 2). Claims key `period` on that + later year, the repo's UK convention, and carry `fy`. +- The reform covers the free hours only. Tax-Free Childcare has its own + £100,000 limit and is not part of it. The £500 Tax-Free Childcare + addition in the report's Appendix D belongs to CBP's estimate. +- Both offers are England's. The counterparts sum each childcare head for + people in English households: policyengine-uk 2.89.2 restricts the three + free-hours offers to England but applies Tax-Free Childcare across the + UK. Ingest checks that no free-hours head moves outside England. + +## The comparison (£bn, negative = cost to the Exchequer) + +| | 2027-28 | 2029-30 | +|---|---:|---:| +| CenTax static cost, central | −0.80 | −0.98 | +| CenTax net fiscal cost, central | −0.52 | −0.64 | +| PolicyEngine, certified pair (static) | −0.690 | −0.838 | +| PolicyEngine, published dashboard (static, history; on policyengine.org 12–21 October 2026) | −0.434 | −0.505 | +| PolicyEngine dashboard since PR #8 merged (1b83dd46, 24 October 2026; its own Vercel deployment, not policyengine.org; not attached) | −0.902 | −1.037 | + +The certified pair is the default replay variant: populace-uk build +`populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z` with policyengine-uk +2.89.2 (`data/uk/certified_bundle.json`). Statuses: + +- static against static: `comparable` (all three scenarios, which differ in + CenTax's assumed bunching friction; PolicyEngine has one current-law + distribution); +- CenTax's net cost, which adds the intensive and extensive behavioural + gains: `concept_mismatch`; +- the published dashboard figure: `constructed` (the dashboard applied six + local corrections to policyengine-uk 2.102.3), and older than the + certified run, so it shows in the history rather than as the latest. + policyengine.org served it from 12 October 2026 until it took the page + down on 21 October while the dashboard's costing is revised + (policyengine-app-v2 de2b07cf); the path redirects to /uk/research. + Since [uk-childcare-100k-threshold#8](https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8) + merged on 24 October, the dashboard's own deployment serves £0.902bn and + £1.037bn on Microcosm's 23 October release (a133f37a, policyengine-uk + 2.147.2), which is in no certified pair. The row records it and does not + attach it; when policyengine.org serves the page again, its figure is a + new published row. + +CenTax's two behavioural margins are on the page without a PolicyEngine +value. Their `pe_missing` says why: this lane's counterparts are static +runs. + +## Why PolicyEngine is below CenTax + +The central static rows carry a `pe_gap` diagnosis linked to +[microcosm#1126](https://github.com/PolicyEngine/microcosm/issues/1126). +Measured on the certified pair: families receiving the extended hours use a +weighted mean of 15.3 of the 30 weekly hours (`maximum_extended_childcare_hours_usage`); +#1126 puts DfE's mean at about 28.5 and identifies the draw as the main +reason PolicyEngine's free-hours spending is low. With every family at the +full 30 hours, the same reform costs £1.40bn in 2027-28 and £1.72bn in +2029-30 on this pair (a diagnostic run, never attached). Ingest writes that +figure only when both full-use worlds record it: the input set to 30 before +any holder that reads it was filled, read back at 30, and the main world's +reform and executed limits on 30 April of the year. Smaller factors on +our side, recorded on each result: the engine switches universal and +targeted hours off once a family qualifies for the extended hours +([policyengine-uk#1930](https://github.com/PolicyEngine/policyengine-uk/issues/1930), +changed after the pin by [policyengine-uk#2177](https://github.com/PolicyEngine/policyengine-uk/pull/2177)), +and no child under 1 receives hours. + +The published dashboard figure ran on the 2 October Microcosm release that +#1126 examines, with the same draw. The fix merged in microcosm on 16 +October 2026 (microcosm#1299). The certified pair was built on 19 June 2026, +before it. [policyengine.py#685](https://github.com/PolicyEngine/policyengine.py/pull/685) +(merged 21 October 2026) certified a Microcosm pair with the fixed draw: +national cut `microcosm-uk-2024-25-national-20261019T083323Z-6cba5d71` with +policyengine-uk 2.132.6, which policyengine.py 6.11.0 (25 October) still +pins. This repository has it only as an experiment population, not a replay +variant, so this lane has not run it yet: +[#228](https://github.com/PolicyEngine/policyengine-scorecard/issues/228). +Until it does, the latest result is the 19 June pair's and the row shows +that pair's miss. + +To add a pair, run the compute script with `--variant ` in that +variant's environment. The run id defaults to +`childcare-100k--`. Two runs of one variant +started in the same second would share it, and ingest refuses two +artifacts with one run id. The artifact goes beside this one in +`results/uk/childcare_100k/`. Before ingest takes it, check each note whose +basis is the engine or a run at the new pin. +These are the entries with a `variants` list in both sources' +`annotations.json` and in `results/uk/childcare_100k/construction_annotations.json`. +Add the pair to the notes that still hold, and write replacements for those +that do not. Then add the pair's `verified_pairs` entry, which says why its +run is static, what is known about its hours draw, and the `pe_gap` +diagnosis, if one still applies. Ingest refuses an artifact on a pair with +no such entry. It attaches each note only to results on the pairs it names, +and it refuses an artifact whose package, release-bundle or dataset identity +differs from its variant record. It keeps every certified artifact's run. +Each claim's latest result is the newest run that computed the claim's +year; older runs and the published figure stay in the history. Run ids may +not start with `platform-grid-` in any letter case: the comparisons view +prefers those runs within a data bundle, and its `LIKE` ignores ASCII case. +Each `computed_at` must be a distinct UTC time, written as +`datetime.isoformat()` writes it, so that text order is time order. + +## Never a calibration target + +Both sources are held out (`relationships.uk_relationship`). A reform +estimate from another model is never a calibration target under the +2026-08-02 rulings. `never_calibrate()` is keyed on metric and covers the +poverty family, so `tests/test_uk_childcare_100k.py` checks the rule on the +built database for every source: any claim that scores a world other than +the baseline is held out. + +## CenTax moments as calibration targets + +Arun Advani suggested calibrating PolicyEngine to CenTax's moments. Under +the 2026-08-02 rulings that splits in two: + +- **Never targets:** CenTax's model outputs. That covers every Table 4.2 + cell, including the bunching counts (projections for 2028 and 2030 from + the bunching model), and estimator outputs such as the 2,610 excess + parents bunching in 2022 (key findings, printed page 3: a comparison + with a counterfactual density). +- **Possible targets, if published as administrative quantities:** genuine + HMRC administrative counts, the quadrant the rulings allow (admin + tax-benefit quantities). The report quotes some, rounded, in prose, such + as about 11,000 parents in 2022 with a live childcare account for part of + the year and an actual adjusted net income over £100,000 (printed page 42). + A target needs CenTax (or HMRC) to publish such counts as administrative + tabulations with a definition, a year and a population, not a rounded + figure in a policy report. + +Decided 25 October 2026: CenTax's administrative tabulations become +calibration targets in Microcosm, where that work happens; its model +outputs stay out. Nothing changes in this lane: both sources stay held out, +and no costing here is ever a target. + +## Reproducing + +```bash +python sources/centax-childcare-100k/adapter.py +python sources/cbp-rates-and-wrongs/adapter.py +# the counterpart: the certified environment (docs/uk_replay/requirements.txt) +# and the certified artifact in the Hugging Face cache, offline; the run +# copies it into a verified writable mirror under data/uk/policyengine-local/ +# (about 1.3 GB, git-ignored) and hashes the file each simulation reads +PYTHONPATH=. .venv-pe289m/bin/python pipeline/compute_uk_childcare_100k.py +PYTHONPATH=. python -m scorecard_db.ingest_uk_childcare_100k data/scorecard.db +``` diff --git a/sources/centax-childcare-100k/adapter.py b/sources/centax-childcare-100k/adapter.py new file mode 100644 index 000000000..d0f0033fa --- /dev/null +++ b/sources/centax-childcare-100k/adapter.py @@ -0,0 +1,285 @@ +"""Adapter: CenTax, "Removing the childcare cliff-edge" (September 2026), +Table 4.2 -> tidy rows (lane uk-childcare-100k). + +Input (raw/): + AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf + the report as published (108 pages), pinned by sha256 in source.json + pages_layout.txt + ``pdftotext -layout`` (poppler 26.09.0) of the twelve PDF pages the + rows and annotations cite, each under a ``=== PDF page N (printed + page M) ===`` header; pinned by sha256. tests/test_uk_childcare_100k.py + re-extracts the pages from the PDF and compares, where pdftotext + is installed. +Output: data/externals/centax-childcare-100k.json + +Rows: Table 4.2 (printed page 41), "Fiscal cost estimates of removing +£100,000 threshold, 2028 and 2030": three scenarios (central, high cost, +low cost) x two tax years (2028, 2030) x six columns — the number of +parents bunching before and after removal, the static cost, the intensive- +and extensive-margin behavioural gains, and the net fiscal cost. 36 rows. +Money is in pounds (the table's £m x 1e6), sign as published (negative = +cost to the Exchequer); counts are parents. + +Years: the report names tax years by their later year (editorial note, +printed page 2: "tax year 2021-22 is given as 2022"), so 2028 is 2027-28. +``period`` is that later year, the repo's UK convention (FY end year), and +``fy`` the explicit label. + +QC (any failure raises): + 1. the PDF and the extract match their source.json digests + 2. the table block parses to exactly six rows, years 2028, 2030 in each + scenario, scenario labels in the order central, high, low + 3. every money cell is a multiple of £10m (the table's rounding) + 4. static + intensive + extensive = net within £20m in every row: four + figures each rounded to the nearest £10m can disagree by at most + 3 x 5 + 5 (the notes: "Figures may not sum due to rounding") + 5. Table 4.3's "Abolish the threshold" rows (printed page 44) equal the + central rows of Table 4.2 cell for cell (static, both margins, net, + and the after-removal bunching count) + 6. the prose under the table (printed pages 41-42) restates the central + 2030 figures: £340 million between static and net cost, £210 million + from the intensive margin, £130 million from the extensive margin + +Run (stdlib only): + python sources/centax-childcare-100k/adapter.py +""" + +from __future__ import annotations + +import hashlib +import json +import re +from pathlib import Path + +HERE = Path(__file__).resolve().parent +RAW = HERE / "raw" +OUT = HERE.parent.parent / "data" / "externals" / "centax-childcare-100k.json" +PDF = RAW / "AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf" +PAGES = RAW / "pages_layout.txt" + +SCENARIOS = ("central", "high_cost", "low_cost") +SCENARIO_LABELS = { + "central": "Central Scenario", + "high_cost": "High-Cost Scenario", + "low_cost": "Low-Cost Scenario", +} +YEARS = (2028, 2030) +# Table 4.2's columns, in order, after the tax year. +COUNTS = ("bunching_before_removal", "bunching_after_removal") +MONEY = ("static_cost", "intensive_margin", "extensive_margin", "net_fiscal_cost") +COLUMN_LABELS = { + "bunching_before_removal": "Number Bunching, Before Removal", + "bunching_after_removal": "Number Bunching, After Removal", + "static_cost": "Static Cost (£)", + "intensive_margin": "Behavioural Gain (£), Intensive Margin", + "extensive_margin": "Behavioural Gain (£), Extensive Margin", + "net_fiscal_cost": "Net Fiscal Cost (£)", +} +ROUNDING_GBP_M = 10 +SUM_TOLERANCE_GBP_M = 3 * ROUNDING_GBP_M // 2 + ROUNDING_GBP_M // 2 # 20 +EXPECTED_ROWS = len(SCENARIOS) * len(YEARS) * (len(COUNTS) + len(MONEY)) + +_NUM = r"[\d,]+" +_MONEY = r"[+-][\d,]+m" +_T42_ROW = re.compile( + rf"(20\d\d)\s+({_NUM})\s+({_NUM})\s+({_MONEY})\s+({_MONEY})\s+({_MONEY})\s+({_MONEY})\s*$" +) +_T43_ROW = re.compile( + rf"(20\d\d)\s+({_NUM})\s+({_MONEY})\s+({_MONEY})\s+({_MONEY})\s+({_MONEY})\s*$" +) + + +def sha256(path: Path) -> str: + return hashlib.sha256(path.read_bytes()).hexdigest() + + +def check_digests() -> dict: + """QC 1.""" + spec = json.loads((HERE / "source.json").read_text())["raw_sha256"] + for path in (PDF, PAGES): + got = sha256(path) + if got != spec[path.name]: + raise ValueError( + f"QC failed: {path.name} sha256 {got} != {spec[path.name]}" + ) + return spec + + +def pages() -> dict[int, str]: + """{PDF page number: layout text} from the vendored extract.""" + out: dict[int, str] = {} + current = None + for line in PAGES.read_text().splitlines(keepends=True): + m = re.match( + r"=== PDF page (\d+) \(printed page (\d+)\) ===$", line.rstrip("\n") + ) + if m: + current = int(m.group(1)) + if int(m.group(2)) != current - 1: + raise ValueError(f"QC failed: page header {line!r}") + out[current] = "" + continue + if current is None: + raise ValueError("QC failed: extract text before the first page header") + out[current] += line + return out + + +def _gbp_m(cell: str) -> int: + """'-1,010m' -> -1010 (£m, sign as printed).""" + return int(cell.removesuffix("m").replace(",", "")) + + +def _count(cell: str) -> int: + return int(cell.replace(",", "")) + + +def parse_table_4_2(text: str) -> dict[tuple[str, int], dict[str, int]]: + """QC 2-3: {(scenario, year): {column: value}} from printed page 41.""" + start = text.index( + "Table 4.2: Fiscal cost estimates of removing £100,000 threshold" + ) + end = text.index("Notes: This table presents the results", start) + block = text[start:end] + order = [block.find(w) for w in ("Central", "High-", "Low-")] + if -1 in order or order != sorted(order): + raise ValueError(f"QC failed: scenario labels out of order {order}") + rows = [m.groups() for line in block.splitlines() if (m := _T42_ROW.search(line))] + if len(rows) != len(SCENARIOS) * len(YEARS): + raise ValueError(f"QC failed: Table 4.2 parsed {len(rows)} rows, want 6") + out = {} + for i, groups in enumerate(rows): + scenario = SCENARIOS[i // len(YEARS)] + year = int(groups[0]) + if year != YEARS[i % len(YEARS)]: + raise ValueError(f"QC failed: {scenario} row {i} has year {year}") + values = {c: _count(v) for c, v in zip(COUNTS, groups[1:3])} + values |= {c: _gbp_m(v) for c, v in zip(MONEY, groups[3:])} + for c in MONEY: + if values[c] % ROUNDING_GBP_M: + raise ValueError( + f"QC failed: {scenario} {year} {c} not a £10m multiple" + ) + out[(scenario, year)] = values + out[(scenario, year)]["_verbatim"] = dict(zip((*COUNTS, *MONEY), groups[1:])) + return out + + +def check_sums(table: dict) -> None: + """QC 4.""" + for (scenario, year), v in table.items(): + gap = ( + v["static_cost"] + + v["intensive_margin"] + + v["extensive_margin"] + - v["net_fiscal_cost"] + ) + if abs(gap) > SUM_TOLERANCE_GBP_M: + raise ValueError( + f"QC failed: {scenario} {year} components miss net by £{gap}m" + ) + + +def check_table_4_3(table: dict, text: str) -> None: + """QC 5: Table 4.3's abolition rows equal Table 4.2's central rows.""" + start = text.index("Table 4.3: Fiscal cost estimates of childcare policy reforms") + rows = [ + m.groups() for line in text[start:].splitlines() if (m := _T43_ROW.search(line)) + ] + for groups in rows[:2]: + year = int(groups[0]) + central = table[("central", year)] + got = { + "bunching_after_removal": _count(groups[1]), + **{c: _gbp_m(v) for c, v in zip(MONEY, groups[2:])}, + } + want = {k: central[k] for k in got} + if got != want: + raise ValueError( + f"QC failed: Table 4.3 abolition {year} {got} != Table 4.2 {want}" + ) + if [int(g[0]) for g in rows[:2]] != list(YEARS): + raise ValueError( + "QC failed: Table 4.3's first rows are not the abolition years" + ) + + +def check_prose(table: dict, text: str) -> None: + """QC 6: the sentences under the table restate the central 2030 row.""" + # the page footers ("41 © CenTax") split a sentence across pages + body = "\n".join( + line + for line in text.splitlines() + if not re.fullmatch(r"\s*\d+\s+© CenTax\s*", line) + ) + flat = " ".join(body.split()) + c = table[("central", 2030)] + claims = { + "around £340 million less than the static cost": c["net_fiscal_cost"] + - c["static_cost"], + "£210 million of this is estimated to arise": c["intensive_margin"], + "Another £130 million is expected": c["extensive_margin"], + } + for phrase, value in claims.items(): + if phrase not in flat: + raise ValueError(f"QC failed: prose phrase missing: {phrase!r}") + stated = int(re.search(r"£(\d+) million", phrase).group(1)) + if stated != value: + raise ValueError(f"QC failed: prose {phrase!r} != table {value}") + + +def tidy(table: dict) -> list[dict]: + rows = [] + for (scenario, year), values in table.items(): + fy = f"{year - 1}-{year % 100:02d}" + for column in (*COUNTS, *MONEY): + money = column in MONEY + rows.append( + { + # the tidy-row core (docs/ARCHITECTURE.md); variant is + # CenTax's friction scenario + "source": "centax-childcare-100k", + "program": "free_childcare_hours", + "metric": "revenue_change" if money else "parents_bunching", + "subgroup": None, + "variant": scenario, + "geography": "England", + # the costing facets beside it + "country": "UK", + "table": "4.2", + "printed_page": 41, + "column": column, + "tax_year_label": str(year), + "fy": fy, + "period": year, + "unit_concept": "gbp" if money else "persons", + "value": values[column] * 1_000_000 if money else values[column], + "value_verbatim": values["_verbatim"][column], + "status": "ok", + "source_column": ( + f"Table 4.2, {SCENARIO_LABELS[scenario]}, tax year {year}: " + f"{COLUMN_LABELS[column]}" + ), + } + ) + return rows + + +def run() -> list[dict]: + check_digests() + text = pages() + table = parse_table_4_2(text[42]) + check_sums(table) + check_table_4_3(table, text[45]) + check_prose(table, text[42] + text[43]) + rows = tidy(table) + if len(rows) != EXPECTED_ROWS: + raise SystemExit(f"row contract broken: {len(rows)} != {EXPECTED_ROWS}") + OUT.parent.mkdir(parents=True, exist_ok=True) + OUT.write_text(json.dumps(rows, indent=1, ensure_ascii=False) + "\n") + return rows + + +if __name__ == "__main__": + out = run() + print(f"wrote {len(out)} rows to {OUT}") diff --git a/sources/centax-childcare-100k/annotations.json b/sources/centax-childcare-100k/annotations.json new file mode 100644 index 000000000..c43a0b1dc --- /dev/null +++ b/sources/centax-childcare-100k/annotations.json @@ -0,0 +1,143 @@ +{ + "_schema": { + "applies_to": "matcher on {components, scenarios, periods, results}; null = any. components: Table 4.2 money columns (static_cost, intensive_margin, extensive_margin, net_fiscal_cost); scenarios: central, high_cost, low_cost; periods: claim periods (the tax year's later year); results: certified (the certified-pair run) or published_dashboard (the history row)", + "variants": "the certified variants (data/uk/replay_variants.json) on which a note resting on the engine or a run was checked; it attaches only to those pairs' results. A note resting on the report alone names none and applies to every run", + "severity": "period | definitional | construction | caution | gap", + "status": "the comparison status this annotation sets, where it sets one; at most one status-setting annotation matches a result", + "basis": "traceability: raw/pages_layout.txt (PDF page and printed page), engine files at the certified pin, or the measured artifact results/uk/childcare_100k/.json" + }, + "annotations": [ + { + "id": "centax-static-vs-static", + "applies_to": { + "components": [ + "static_cost" + ], + "scenarios": null, + "periods": null, + "results": [ + "certified" + ] + }, + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "severity": "construction", + "status": "comparable", + "text": "Both numbers are static costs. CenTax's static cost is the cost of funding the hours for children of parents above £100,000 before any behavioural response (Table 4.2 notes; Appendix A, 'Static cost'). PolicyEngine's is the change in spending on the three free-hours entitlements in England with the limit removed; at the pin the engine's labour-supply elasticities are zero, so no behavioural response enters.", + "basis": "raw/pages_layout.txt PDF page 42 (printed page 41, Table 4.2 notes) and PDF pages 75-76 (printed pages 74-75, Appendix A 'Static cost'); policyengine_uk 2.89.2 variables/input/employment_income.py (adds employment_income_behavioral_response) and variables/gov/simulation/labour_supply_response/employment_income_behavioral_response.py (returns 0 when income_elasticity and substitution_elasticity are both 0); parameters/gov/simulation/labour_supply_responses/income_elasticity.yaml and substitution_elasticity.yaml (0 from 2020-01-01), which no world's reform_dict sets; results/uk/childcare_100k/.json behavioural_response 'none (static)' and gov_tax, which moves by under £1m in every world and year (checked at ingest)" + }, + { + "id": "centax-static-published-dashboard", + "applies_to": { + "components": [ + "static_cost" + ], + "scenarios": null, + "periods": null, + "results": [ + "published_dashboard" + ] + }, + "severity": "construction", + "status": "constructed", + "text": "A static cost on both sides, but the published PolicyEngine figure is the engine plus the dashboard's six local corrections to policyengine-uk 2.102.3, so it is a construction rather than the engine as served.", + "basis": "pe/published_dashboard.json (corrections: src/childcare_100k/corrections.py at PolicyEngine/uk-childcare-100k-threshold f25f0db)" + }, + { + "id": "centax-net-includes-behaviour", + "applies_to": { + "components": [ + "net_fiscal_cost" + ], + "scenarios": null, + "periods": null, + "results": null + }, + "severity": "definitional", + "status": "concept_mismatch", + "text": "CenTax's net fiscal cost adds two behavioural gains to the static cost: tax from parents who stop holding their income below £100,000 (the intensive margin) and tax and National Insurance from partners who enter work (the extensive margin). PolicyEngine's counterpart is the static cost alone, so the two measure different things.", + "basis": "raw/pages_layout.txt PDF page 42 (printed page 41, Table 4.2 notes: 'The intensive margin behavioural gain ... The extensive margin behavioural gain ...') and PDF page 76 (printed page 75, Appendix A 'Behavioural revenue gain')" + }, + { + "id": "centax-scenario-friction", + "applies_to": { + "components": [ + "static_cost", + "net_fiscal_cost" + ], + "scenarios": [ + "high_cost", + "low_cost" + ], + "periods": null, + "results": null + }, + "severity": "construction", + "text": "CenTax's scenarios assume different falls in the bunching friction after 2022 (none in the high-cost scenario, a 50% fall in the low-cost one). That changes how many parents are above £100,000 under current law, so the static cost differs by scenario. PolicyEngine has one current-law income distribution, so one PolicyEngine value meets all three scenarios.", + "basis": "raw/pages_layout.txt PDF pages 40-41 (printed pages 39-40, section 4.1) and PDF page 75 (printed page 74, Appendix A 'Static cost': parents between £100,000 and the marginal buncher's income) and PDF page 43 (printed page 42: in the low-cost scenario 'there are fewer parents remaining above the threshold that would represent a new cost')" + }, + { + "id": "centax-free-hours-only", + "applies_to": { + "components": null, + "scenarios": null, + "periods": null, + "results": [ + "certified" + ] + }, + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "severity": "definitional", + "text": "CenTax costs the free childcare hours only ('removing the £100,000 threshold for eligibility for free childcare hours'). Tax-Free Childcare, which has its own £100,000 limit, is not part of the reform. PolicyEngine's counterpart matches: it removes only the 30-hours limit and sums the three free-hours entitlements, and the run measured no change in Tax-Free Childcare. The £500 Tax-Free Childcare addition in CenTax's Appendix D belongs to the Centre for British Progress estimate CenTax reviews, not to CenTax's figures.", + "basis": "raw/pages_layout.txt PDF page 42 (printed page 41, Table 4.2 notes), PDF page 9 (printed page 8, section 1.1: Tax-Free Childcare 'withdrawn in full ... once either parent has an ANI above £100,000') and PDF page 100 (printed page 99, Appendix D: CBP's 'value of the entitlement ... with £500 added for Tax-Free Childcare'); results/uk/childcare_100k/.json worlds.free_hours: limits_executed and the tax_free_childcare sum" + }, + { + "id": "centax-year-convention", + "applies_to": { + "components": null, + "scenarios": null, + "periods": null, + "results": [ + "certified" + ] + }, + "variants": [ + "pe5.0.2_uk2.89.2_populace-uk-2023-dd68c73-4aa4b14-20260619T023711Z" + ], + "severity": "period", + "text": "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. PolicyEngine's annual period 2027 runs the 2027-28 rules: policyengine-uk sets each government parameter for year Y to its value on 30 April of Y.", + "basis": "raw/pages_layout.txt PDF page 3 (printed page 2, editorial note: 'tax year 2021-22 is given as 2022'); policyengine_uk 2.89.2 utils/parameters.py convert_to_fiscal_year_parameters, called in tax_benefit_system.py process_parameters" + }, + { + "id": "centax-free-hours-only-published", + "applies_to": { + "components": null, + "scenarios": null, + "periods": null, + "results": [ + "published_dashboard" + ] + }, + "severity": "definitional", + "text": "CenTax costs the free childcare hours only. The dashboard's reform also removed the Tax-Free Childcare limit; the figure here is its 30-hours leg, which excludes Tax-Free Childcare (the dashboard puts that leg at £0.159bn in 2027-28 and £0.173bn in 2029-30).", + "basis": "raw/pages_layout.txt PDF page 42 (printed page 41, Table 4.2 notes); pe/published_dashboard.json (budget.gross_bn.thirty_hours and budget.gross_bn.tax_free_childcare of data/results.json at f25f0db)" + }, + { + "id": "centax-year-convention-published", + "applies_to": { + "components": null, + "scenarios": null, + "periods": null, + "results": [ + "published_dashboard" + ] + }, + "severity": "period", + "text": "CenTax names tax years by their later year, so 2028 is 2027-28 and 2030 is 2029-30. The dashboard names fiscal years by their start year (2027 is 2027-28) and compared its 2029-30 figure with CenTax's 2030.", + "basis": "raw/pages_layout.txt PDF page 3 (printed page 2, editorial note); pe/published_dashboard.json (src/childcare_100k/config.py at f25f0db: 'Fiscal years named by start year (\"2026\" = 2026-27)'; data/results.json benchmarks[0].like_for_like)" + } + ] +} diff --git a/sources/centax-childcare-100k/pe/published_dashboard.json b/sources/centax-childcare-100k/pe/published_dashboard.json new file mode 100644 index 000000000..c3f32a75d --- /dev/null +++ b/sources/centax-childcare-100k/pe/published_dashboard.json @@ -0,0 +1,79 @@ +{ + "_what": "PolicyEngine's PUBLISHED estimate of the same reform, kept as a history row on the CenTax claims so the published miss stays on the page. Read from the dashboard's own results file; nothing here is recomputed. live records when policyengine.org served it; after records what the dashboard has served since, which is not attached.", + "page": "https://policyengine.org/uk/childcare-100k-threshold", + "repo": "PolicyEngine/uk-childcare-100k-threshold", + "commit": "f25f0dbd46c31d336bde06898d801a1887a8cf89", + "commit_note": "main from the merge of #6 (2026-10-12) until #8 merged (2026-10-24); data/results.json and dashboard/public/data/results.json, the file the dashboard serves, are the same bytes at this commit", + "results_file": "data/results.json", + "results_file_sha256": "4193f13b2a1f40a809d5738a96f6558a3633978bd3bd5efd4da5db211b72f50a", + "generated_at": "2026-10-09T11:50:21+00:00", + "results_git_revision": "1b57118fe8dab4d82b456cbaff24d0306c74a1f6", + "run_id": "uk-childcare-100k-dashboard-1b57118", + "engine": { + "policyengine": "6.2.1", + "policyengine_uk": "2.102.3", + "policyengine_bundle": "uk-6.2.1" + }, + "dataset": { + "release": "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68", + "repo_id": "policyengine/populace-uk-private", + "revision": "f9d1922cddab6b54a0dd37794a9bac74e3780c88", + "sha256": "aa31bdf67c977927ea2b325567d1cf7a79d94381239bc79918a0a0fc9c9588af", + "management": "pinned_by_repo_unmanaged_by_bundle" + }, + "reform": { + "removed_limits": [ + "gov.dfe.extended_childcare_entitlement.income.limit", + "gov.hmrc.tax_free_childcare.income.income_limit" + ], + "basis": "src/childcare_100k/config.py at f25f0db (REFORM_PARAMETERS, parameter_changes: each limit set to infinity in every analysis year)" + }, + "corrections": "Six local corrections applied to every run, baseline and reform alike (src/childcare_100k/corrections.py at f25f0db): exactly £100,000 qualifies; a partner on a specified benefit need not pass the minimum earnings test; the universal and targeted 15 hours are a floor that eligibility for the extended hours never removes (policyengine-uk#1930); salary sacrifice returned to pay from April 2029 is not income for the limits; Tax-Free Childcare regards a partner on a caring or incapacity benefit as in qualifying paid work; both schemes test the claimant and partner.", + "fiscal_year_rule": "src/childcare_100k/config.py at f25f0db: 'Fiscal years named by start year (\"2026\" = 2026-27).'", + "values_gbp_bn": { + "field": "budget.gross_bn.thirty_hours (the 30-hours leg: extended, universal and targeted entitlements; budget.thirty_hours_components_bn puts it all in extended)", + "thirty_hours": { + "2026": 0.41, + "2027": 0.434, + "2028": 0.484, + "2029": 0.505 + }, + "tax_free_childcare": { + "2026": 0.148, + "2027": 0.159, + "2028": 0.168, + "2029": 0.173 + } + }, + "hours_draw": "microcosm#1126 (closed 2026-10-16) reports that this release draws maximum_extended_childcare_hours_usage ~ clip(N(15.019, 4.972), 0, 30), a mean of about 15 of the 30 weekly hours, and derives DfE's mean as about 28.5 hours: https://github.com/PolicyEngine/microcosm/issues/1126", + "live": { + "on": "https://policyengine.org/uk/childcare-100k-threshold", + "from": "2026-10-12", + "from_basis": "the repository's Production deployment of this commit at 2026-10-12T06:39:44Z (GitHub deployments API); policyengine.org's app listing dates the dashboard 2026-10-12, and its zone route sent the page to the production alias uk-childcare-100k-threshold.vercel.app (policyengine-app-v2 app/src/data/apps/apps.json and website/src/data/appZoneRoutes.ts before de2b07cf)", + "until": "2026-10-21", + "until_basis": "policyengine-app-v2 de2b07cf (2026-10-21T14:59:14Z), 'Take down the childcare cliff edge dashboard temporarily': it removes the listing, the featured card and the route, and redirects /uk/childcare-100k-threshold/:path* to /uk/research (permanent: false) 'while its costing is revised (uk-childcare-100k-threshold#8)'" + }, + "after": { + "pr": "https://github.com/PolicyEngine/uk-childcare-100k-threshold/pull/8", + "merged": "2026-10-24", + "commit": "1b83dd465848e12bbf28aa61e7d0cb969800dbf4", + "commit_note": "main and the Production deployment of 2026-10-24T22:51:59Z, read on 2026-10-25", + "served_at": "https://uk-childcare-100k-threshold.vercel.app/uk/childcare-100k-threshold (the repository's homepage; policyengine.org still redirects the page on 2026-10-25)", + "results_file_sha256": "a069113caa09c0ceb788893d5a41ed7ab36a6f06115b091ec8e05cde9923dc0c", + "engine": { + "policyengine": "6.10.2", + "policyengine_uk": "2.147.2", + "policyengine_bundle": "uk-6.10.2" + }, + "release": "microcosm-uk-2024-25-national-20261023T075039Z-a133f37a", + "release_note": "built from microcosm 829c1bc26, which contains microcosm#1299 (merge 866fee4e, 2026-10-16), the fix for microcosm#1126's draw; per the results file, no policyengine.py bundle includes it", + "thirty_hours_gbp_bn": { + "2026": 0.845, + "2027": 0.902, + "2028": 0.955, + "2029": 1.037 + }, + "attached": false, + "why_not_attached": "policyengine.org does not serve it, and its Microcosm release is in no certified pair. When policyengine.org serves the page again, the figure it shows is a new published row." + } +} diff --git a/sources/centax-childcare-100k/raw/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf b/sources/centax-childcare-100k/raw/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf new file mode 100644 index 000000000..a279399e9 Binary files /dev/null and b/sources/centax-childcare-100k/raw/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf differ diff --git a/sources/centax-childcare-100k/raw/pages_layout.txt b/sources/centax-childcare-100k/raw/pages_layout.txt new file mode 100644 index 000000000..51bda71cf --- /dev/null +++ b/sources/centax-childcare-100k/raw/pages_layout.txt @@ -0,0 +1,598 @@ +=== PDF page 3 (printed page 2) === +Executive summary + +Since 2010, all parents of 3- and 4-year-olds in England have been able to +claim 15 hours a week of free childcare funded by government for 38 weeks +of the year. Since 2017, working parents of 3- and 4-year-olds have been able +to claim a further 15 hours, provided that certain conditions are met. These +include the requirement that neither parent expects to have income above +£100,000. Families with a parent whose expected income is above £100,000 +– after adjustments such as for pension contributions and charitable +donations – face the prospect of losing access to some or all of their free +childcare, potentially leaving them worse off overall. This childcare ‘cliff- +edge’ is in addition to and, for those affected, more severe than the +withdrawal of the Personal Allowance for Income Tax, which also starts at +£100,000. + +The size of the childcare cliff-edge has been getting bigger, as the value of +the free childcare entitlement has increased following recent reforms and +rising childcare costs over time. Starting in April 2024, the entitlement has +also been gradually expanded to cover a wider range of ages. Working +parents of children aged 9 months to 2 years old are now eligible for 30 hours +per week of free childcare, provided they both expect their income for the +tax year to be below £100,000. Parents of under-3s therefore lose 30 hours of +free childcare by crossing the threshold, rather than the 15 hours that +parents of 3- and 4-year-olds lose. The threshold has also stayed fixed at +£100,000 in cash terms since 2017, so rising incomes pull more parents into +its range each year. The combined effect is that the median affected +parent would have had to reach income of roughly £105,000 to be no +worse off after losing free childcare in the tax year 2021/22, but this will +rise to £124,000 by the end of this parliament, in 2029/30.1 + +The childcare cliff-edge has been widely criticised for discouraging work and +creating incentives that hold back economic growth. However, quantifying +these impacts and estimating the fiscal cost of reform requires analysis of +HMRC administrative tax data covering the full population of parents +affected by the cliff-edge, which has not previously been available. In this +report, we provide the first detailed evidence on who is affected by the +withdrawal of free childcare, how they respond, and how much it would cost +to remove this cliff-edge by making the free childcare entitlement available +to parents with incomes over £100,000, or other options short of this. + +Editorial note: all references to years in this report are based on tax years, giving the later +1 + +year e.g., tax year 2021-22 is given as 2022. + + + 2 © CenTax + === PDF page 9 (printed page 8) === +In Wales, all 3- and 4-year-olds receive a universal Foundation Phase +entitlement of at least 10 hours per week with no income test. Since April +2019 working parents can add to this through the Childcare Offer for Wales, +taking combined provision up to 30 hours per week for up to 48 weeks.4 The +income condition for this is assessed on each parent's gross income being +£100,000 or less. Gross income is measured before deductions such as +pension contributions and salary sacrifice, in contrast to ANI used in +England.5 + +Alongside these devolved schemes, Tax-Free Childcare operates across the +whole UK. It creates a childcare account that both parents and the +government pay into. For every £8 a parent pays into the account, the +government will top it up by £2, up to £2,000 per child a year (higher for +disabled children). It is withdrawn in full, with no taper, once either parent +has an ANI above £100,000.6 + +1.2. Understanding the effect of the childcare entitlement threshold in +England + +The design of the childcare entitlement in England means that very small +changes in expected incomes can result in large reductions in the number +of hours per year that the government will fund for an early-years child. This +creates a strong incentive for parents who choose to make use of the +childcare entitlement to keep their income below the threshold. + +Starting in April 2024, the entitlement has also been gradually expanded to +cover a wider range of ages. This now provides 30 hours for 38 weeks per +year for children from 9-months to 2 years, in addition to the 15 hours for 3- +and 4-year-olds introduced in 2017. Significantly, the universal free 15 hours +of childcare has not also been expanded beyond 3- and 4-year-olds, which +means that parents of under-3s risk losing 1,140 hours of annual funded +childcare, rather than the 570 hours that parents of 3- and 4-year-olds would +lose.7 + +Figure 1.1 shows the range of the implicit value of childcare entitlement +across England, driven by large variation in the hourly cost of childcare +parents face when paying for it out of their own pocket. For most working +parents of a 3- or 4-year-old in 2025, losing a full year of the childcare + + +4 + https://www.gov.wales/get-30-hours-childcare-3-and-4-year-olds/nursery-education +5 + The number of parents of young children in Wales receiving this level of income is small, +meaning we are not able to perform statistical analysis on this group. +6 + https://www.gov.uk/tax-free-childcare +7 + 30 hours per week multiplied by 38 weeks per year. + + + 8 © CenTax + === PDF page 40 (printed page 39) === +Table 4.1: Relative bunching and friction parameter estimates, 2019-2022 + + Employees Non-Employees + Tax Year Relative Friction Relative Friction + Bunching Parameter Bunching Parameter + 2019 1.33 0.94 8.29 0.72 + + 2020 1.43 0.93 9.19 0.72 + + 2021 1.68 0.92 9.79 0.68 + + 2022 1.79 0.93 12.41 0.67 +Notes: This table shows the relative bunching and friction parameter estimates for +employee and non-employee parents in each tax year 2019 to 2022. Relative bunching refers +to the estimated excess number of parents with an income just below £100,000 for that tax +year divided by the number estimated to have an income of £100,000 were there no +incentive to bunch at £100,000. The friction parameter estimates are the share of parents +remaining that have an income that is in a ‘dominated region’. That is, where the value of +the childcare entitlement exceeds the additional post-tax income they receive over +£100,000. + +Source: Authors' calculations based on HMRC administrative datasets. + +Given this ongoing trend, our modelling takes as a conservative central +estimate the assumption that this trend continues in coming years. By 2028 +this would mean the friction parameter for employees would have reduced +by another 2%, and for non-employees by about 13%, which would imply an +overall reduction of around 5%, given the relative share of employees and +non-employees in the sample. Increased response by parents to the current +policy reduces the cost of reform, since relaxation of restrictions on childcare +entitlement would lead (some of) them to have higher incomes, and pay +more tax, while continuing to receive the entitlement as now. + +Since there is now a larger age range in which children are eligible for the +entitlement, as well as a greater number of free hours lost for under-3s, +some parents that previously would not have considered it worth the hassle +to respond may now do so. In this case the friction parameter might have +reduced more quickly than the previous trend indicates. As a lower bound +on the financial cost of reforms (or upper bound of the benefits), we assume +that uptake of childcare entitlements rises to match the share just below +the threshold. As shown in Figure 2.1, this is around 60%, providing a likely +‘floor’ on how much further the friction parameter can fall, as we would +expect at least 40% of parents above £100,000 to not make use of the +entitlement even if it were available. + + + + + 39 © CenTax + === PDF page 41 (printed page 40) === +Conversely, since we have no quantitative evidence that the trend in the +friction parameter has continued post-2022, as an upper bound on the +financial cost we assume that the friction parameter remains unchanged at +its 2022 level. + +We therefore present a high cost, low cost, and central cost scenario +illustrating the potential fiscal effects of reforms. In the high-cost scenario, +we assume that the level of friction remains as it was in 2022. In the central +scenario, we assume the trend of the change in responsiveness between +2019 and 2022 continues into the future. In the low-cost scenario, we assume +the friction parameter falls by 50%, which corresponds approximately to the +‘floor’ in which the only parents not responding are those that would not be +making use of the entitlement regardless. The scenarios presented in +Section 4.2 are intended as bounds on the estimated effects of any reforms, +with our central estimate representing a conservative prediction of how +responsiveness to the cliff-edge has increased since 2022. As we cannot +directly observe what the magnitude of frictions are beyond 2022, Appendix +C presents the results for a range of changes to the friction parameters in +order to show the estimated fiscal effects change with the friction +parameter, as well as the modelling results for each of the reforms under the +low-cost and high-cost scenarios. + +4.2. Reform modelling results + +4.2.1. Abolish the threshold +We take as our baseline abolishing the threshold and providing the +childcare entitlement with no limit on how high a parent’s income can be +before losing any entitlement. There are no significant administrative +considerations in doing so, as it simply removes the complexity from the +system. Childcare accounts could continue to be used to administer the +entitlement but significantly simplified given there is no longer a link to +individual incomes or the tax system. + + + + + 40 © CenTax + === PDF page 42 (printed page 41) === +Table 4.2: Fiscal cost estimates of removing £100,000 threshold, 2028 +and 2030 + + Number Behavioural + Bunching Gain (£) Net + Tax Static Fiscal + Scenario + Year Cost (£) Cost + Before After Intensive Extensive (£) + Removal Removal Margin Margin + + + + 2028 13,360 3,850 -800m +150m +120m -520m + Central + Scenario + 2030 16,280 4,300 -980m +210m +130m -640m + + + High- 2028 9,710 3,850 -820m +100m +120m -610m + Cost + Scenario 2030 10,860 4,300 -1,010m +120m +120m -770m + + + Low- 2028 58,510 3,850 -580m +630m +90m +140m + Cost + Scenario 2030 67,650 4,300 -700m +770m +90m +160m + +Notes: This table presents the results of modelling the fiscal effects of removing the +£100,000 threshold for eligibility for free childcare hours. The static cost is the estimated +cost of providing the modelled childcare policy, relative to the status quo, before accounting +for any behavioural effects. The intensive margin behavioural gain is the estimated fiscal +effect of the modelled policy due to taxpayers changing their ANI due to the change in +incentive to bunch at the threshold. The extensive margin behavioural gain is the modelled +effect of change in the share of partners above the threshold that enter the labour market +and begin paying Income Tax and National Insurance Contributions. Figures may not sum +due to rounding. Figures in £m; negative = cost to the Exchequer + +Source: Authors' calculations based on HMRC administrative datasets. + +Table 4.2 presents the estimated effects of removing the threshold. Under +our central estimate, the static cost of providing childcare to those above +the threshold is estimated to be just below £1 billion by 2030. Fiscal drag +means that by this point there will be almost 200,000 potentially eligible +parents with an income greater than £100,000. + +Before the removal of threshold, the number of parents estimated to be +bunching at £100,000 in our central scenario is about 13,000 and 16,000 in +2028 and 2030, respectively. Following the removal of the entitlement +threshold, we estimate around 4,000 parents will continue bunching due to +the Personal Allowance withdrawal. After accounting for these changes in +parents’ behaviour, the overall cost is estimated to be around £340 million + + + 41 © CenTax + === PDF page 43 (printed page 42) === +less than the static cost. £210 million of this is estimated to arise from +parents ceasing to hold their income below £100,000 and thus generating +additional tax revenue. Another £130 million is expected to come from an +increase in the participation rate of the partners of parents currently +ineligible for the entitlement, who are now able to re-enter the labour +market. + +Under the lower bound cost scenario, where the share of parents assumed +to remain above the threshold is half that observed in 2022, the removal of +the threshold is revenue positive. This is because under this scenario, around +four times as many parents are estimated to bunch due to the childcare +entitlement threshold. This reduces the cost of expanding the entitlement, +as there are fewer parents remaining above the threshold that would +represent a new cost, while greatly increasing the behavioural revenue +effect, as there are many more parents that would increase their incomes +following the removal of the threshold. The extensive margin effect is +smaller for the same reason the static cost is smaller – that is there are fewer +currently-ineligible parents whose partners are assumed to return to work +after becoming eligible. + +4.2.2. Change to an actual ANI threshold +As explained in Section 2.5, some parents with incomes above £100,000 for +the tax year have still had a live childcare account for some period of the +year. One possible reform to the entitlement would be to change the +income measure used to determine eligibility for the entitlement away from +expected ANI and use parents’ actual ANI for the year. This would mean that +parents that were deemed eligible for at least part of the year because they +had expected to have an income below £100,000 but ended up with an +actual ANI greater than £100,000 would no longer be considered eligible for +the entitlement in that year. + +In 2022 there were approximately 11,000 parents who had a live childcare +account for at least part of the year while going on to have an actual ANI +greater than £100,000. If the free childcare that was provided to these +parents was instead provided on the basis of actual ANI, the upper income +threshold could be expanded to about £111,000 at the same fiscal cost.26 + + + + +26 + This is estimated by calculating the number of additional parents with incomes just above +£100,000 that could be provided the free childcare entitlement if the total quarters of +childcare claimed by all parents above £100,000 in 2022 were instead allocated to parents +just above £100,000, assuming the average number of quarters claimed by these parents +would be the same as observed for parents below £100,000. + + + 42 © CenTax + === PDF page 45 (printed page 44) === +Table 4.3: Fiscal cost estimates of childcare policy reforms, 2028 and +2030: actual ANI, central scenario + + Net + Behavioural Gain (£) + Tax Number Static Fiscal + Reform + Year Bunching Cost (£) Intensive Extensive Cost + Margin Margin (£) + + Abolish 2028 3,850 -800m +150m +120m -520m + the + threshold 2030 4,300 -980m +210m +130m -640m + Abolish + the 2028 11,930 -300m +80m +50m -170m + threshold + for the + first 15 2030 14,690 -370m +110m +50m -210m + Hours to + Under-3s + Raise the 2028 7,880 -230m +100m +50m -90m + threshold + to + £125,000 2030 9,570 -290m +130m +50m -100m + + Raise the 2028 5,060 -400m +120m +70m -210m + threshold + to + £150,000 2030 6,360 -490m +160m +80m -250m + + Additional 2028 6,390 -440m +120m +70m -250m + 8% + Marginal + Rate 2030 8,460 -570m +170m +80m -320m + + Additional 2028 7,300 -220m +120m +40m -60m + 18% + Marginal + Rate 2030 9,450 -300m +170m +50m -80m + + Additional 2028 8,060 -120m +120m +30m +30m + 28% + Marginal + Rate 2030 10,230 -180m +160m +40m +20m +Notes: This table presents the results of modelling the fiscal effects of replacing the +£100,000 threshold with other reforms. The static cost is the estimated cost of providing the +modelled childcare policy, relative to the status quo, before accounting for any behavioural +effects. The intensive margin behavioural gain is the estimated fiscal effect of the modelled +policy due to taxpayers changing their ANI due to the change in incentive to bunch at the +threshold. The extensive margin behavioural gain is the modelled effect of change in the +share of partners above the threshold that enter the labour market and begin paying +Income Tax and National Insurance Contributions. Figures may not sum due to rounding. +Negative = cost to the Exchequer. Source: Authors' calculations based on HMRC +administrative datasets. + + + + 44 © CenTax + === PDF page 74 (printed page 73) === +including those that were already eligible for the 3- and 4-year-olds offer +within the tax year. + +Since we would expect a larger response from parents that face a larger +notch, we separate the new sample into parents that would have any +eligibility for under-3 childcare (as for these ages, the full 30 hours is lost) and +those that would only have eligibility for 3- and 4-year-olds childcare (as only +15 hours would be lost). We therefore estimate the income response +separately for four subsamples: (i) employees with any under-3 childcare +eligibility; (ii) employees with only 3- and 4-year-old eligibility; (iii) non- +employees with any under-3 childcare eligibility; (iv) non-employees with +only 3- and 4-year-old eligibility. + +To predict the income response for each of these subsamples, we first +calculate average notch size faced by the parents in the subsample, based +on the number of hours lost if passing the £100,000 threshold, and the +hourly cost in their region for a child of the relevant age. + +We then use the estimated elasticity and the relationship between the +elasticity and the marginal bunching individual to estimate the income of +the marginal bunching individual in each of the subsamples. + +Finally, to estimate the absolute number of individuals that would bunch, +we estimate the counterfactual income distribution for that subsample in +the same way as for the original sample and assume the share of missing +mass would follow the same relative shape as estimated on our original +sample. Concretely, we regress 𝑚𝑗 on 𝑧𝑗𝑟𝑒𝑙 , where 𝑚𝑗 is the share of missing +mass in income bin 𝑗 and 𝑧𝑗𝑟𝑒𝑙 = (𝑧𝑗 − 𝑧 ∗ )/Δ𝑧 ∗ is the position of income bin 𝑗 +relative to the notch threshold 𝑧 ∗ and marginal bunching income bin Δ𝑧 ∗ , +and then predict 𝑚 ̂𝑗 for each income bin in each subsample, using the +relative income bins positions calculated from the estimated marginal +bunching income. + +This leaves us with a predicted number of taxpayers in each income bin had +the expanded childcare offer been in place in the 2022 tax year, and a +counterfactual number of taxpayers in each bin if no kink or notch existed. + +Modelling reforms to the childcare policy + +Removal of the £100,000 threshold +We first model the fiscal effects of removing the £100,000 income eligibility +threshold completely. We do this by first calculating the static cost of +increasing the number of eligible children to those with a parent with an + + + + 73 © CenTax + === PDF page 75 (printed page 74) === +ANI above £100,000. We then estimate the behavioural revenue gain arising +from parents having less of an incentive to hold their income below +£100,000 and so increasing their incomes that generates additional tax +revenue. + +Static cost +We first estimate the ‘static’ (i.e. mechanical) cost of this reform, in the +following way. For those with incomes above the estimated marginal +buncher’s income, we assume none of these parents reduce their income +to below £100,000 if the child threshold was in place. We sum up the implied +cost to the government of funding hours for children of these parents, +assuming their use of free hours would be the same (on average) as use by +children of parents just below the threshold. + +For those with counterfactual incomes between £100,000 and the marginal +buncher’s income, some share of these taxpayers we assume will bunch +their income below £100,000. Our previous estimation approach allows us +to understand the distribution of where this bunching comes from, +although we cannot identify which specific individuals would respond. We +therefore calculate average cost to government per taxpayer in each income +bin and multiply this by the number of taxpayers assumed to remain in this +income bin if the expanded offer and income eligibility threshold are in +place. + +To obtain the final total static cost, we adjust the total cost to account for +the share of parents with incomes over £100,000 whose partner also has an +income over £100,000, to avoid counting the cost of the same child twice. +We then also adjust the cost to assume there is no provision of free childcare +hours for the share of parents with incomes over £100,000 whose partner is +not working in that tax year, as these children would remain ineligible under +this reform. + +We also adjust the cost of removing the £100,000 threshold to account for +parents with live accounts with ANI of over £100,000. These live accounts are +already in receipt of (some) free childcare hours and so would not represent +a full additional cost of removing the threshold. However, as shown earlier in +this appendix, the average number of quarters for which the live accounts +above £100,000 are eligible is significantly lower than the live accounts +below £100,000, so if they were to become eligible independent of income +then they would still represent some additional cost. To account for this, we +assume that post-removal of the £100,000 threshold, they would on average +utilise the free childcare hours for the same number of quarters as parents +with ANI below £100,000 + + + 74 © CenTax + === PDF page 76 (printed page 75) === +Finally, we assume that take-up rates of the benefit for the newly eligible +parents would be the same as currently estimated take-up rate. For parents +of 3- and 4-year-olds, we use the estimated take-up rate for the year in +question from the DfE’s Funded early education and childcare annual +January statistics.44 As the expanded childcare offer to under-3s has only +been in place since April 2024, there is only one year of take-up rate data for +this group. Therefore, for this group we apply this take-up rate in each year. + +Behavioural revenue gain + +To estimate the amount of additional tax revenue that would be gained +from parents who would no longer keep their income below £100,000, we +first estimate the number of parents who would no longer be bunching +based on who we model as bunchers under the expanded offer and +£100,000 threshold policy. As there would still be an incentive to bunch at +£100,000 (for example, due to the kink in the tax schedule generated by the +tapering of the Personal Allowance), we assume there would be some +residual excess mass remaining at £100,000. + +To account for this, we use the equivalent population of parents located in +Scotland, where free childcare hours are universal, as a control population +for the number of parents we estimate would continue to bunch in England +if the £100,000 threshold was removed. We implement this by estimating +the relative excess mass (that is, total observed excess mass divided by the +estimated counterfactual mass in the income bin at £100,000) for the +pooled population of Scottish parents of children of eligible age across the +tax years 2019 to 2022. + +We then divide the relative bunching estimated from our control Scottish +population by the relative bunching estimated for our treated English +population to estimate the share of taxpayers who would not increase their +incomes following the removal of the £100,000 threshold. + +For the parents we assume do increase their incomes, we assume that the +number that ‘debunch’ into each income bin between the £100,000 +threshold and the marginal bunching income is proportional to the amount +of missing mass in that income bin. We then calculate the implied increase +in tax liability from incomes increasing from £100,000 to the average +income within each income bin above £100,000 and multiply this by the + + + +44 + Up until the latest year (2025). For modelling beyond this year, we assume take up rates +remain constant at the last observed year in the data (https://explore-education- +statistics.service.gov.uk/find-statistics/funded-early-education-and-childcare/2025) + + + 75 © CenTax + === PDF page 100 (printed page 99) === +Appendix D: Comparison to previous work +The Centre for British Progress (CBP) has published the only other +quantitative estimate of the fiscal effect of reforming the £100,000 childcare +threshold.46 Their report covers four cliff-edges and tapers in the tax and +benefit system; we comment here only on the childcare component. CBP +estimate that abolishing the threshold would raise between £130 million +and £425 million in 2030, and that replacing it with a 3% per-child surtax on +income above £100,000 would raise between £330 million and £700 +million.47 +Their approach is a synthetic microsimulation. Children are generated from +Office for National Statistics (ONS) birth numbers and assigned a date of +birth reflecting seasonal variation in births, which determines their +entitlement. Each child is assigned two parents whose incomes are drawn +from the 2023 Survey of Personal Incomes (SPI), restricted to those aged 25 +to 44 and of opposite sex, with an imposed correlation of 0.5 between +partners' incomes. The value of the entitlement is calculated from national +average hourly funding rates, with £500 added for Tax-Free Childcare. Two +behavioural models are then run separately, each allowing only one margin +of response. In the first, labour income is fixed and individuals choose how +much of their income above £100,000 to divert into a salary sacrifice +pension. In the second, pension saving is set aside and individuals choose +labour supply. In both, an individual bunches their income at the threshold +if doing so leaves them at least £1,000 better off. +The synthetic approach was necessary because the data needed to answer +this question directly were not in the public domain. Estimating the cost of +reform requires identifying which taxpayers around £100,000 have a child of +qualifying age, measuring their income on the ANI basis that determines +eligibility rather than on a gross basis, observing whether they actually claim +the entitlement, and observing their partner's income and work status. +None of this is available in published statistics. HM Treasury confirmed in +response to a Freedom of Information request in November 2025 that it +holds no estimate of the number of parents affected by the threshold +measured on gross income.48 The SPI data that CBP use measure neither +ANI nor gross income, so it captures some, but not all, of the bunching. CBP +are explicit in their report that their analysis was subject to these constraints. +Our estimates differ from theirs principally because we can directly observe + +46 + Cohen, E. and Stubbs, M., Rates and wrongs: fixing UK tax cliff edges and tapers, Centre +for British Progress, 25 March 2026, https://britishprogress.org/reports/rates-and-wrongs- +fixing-uk-tax-cliff-edges-and-tap +47 + Figures are drawn from their accompanying technical appendix, +https://britishprogress.org/api/media/file/rates_and_wrongs_technical_appendix.pdf +48 + Freedom of Information response, HM Treasury, November 2025, +https://www.whatdotheyknow.com/request/impact_of_ps100000_threshold_on_2 + + + 99 © CenTax + === PDF page 101 (printed page 100) === +these quantities in administrative data, and the assumptions the synthetic +model made in place of data turn out to be highly consequential. +The most important of these is the treatment of frictions. CBP assume that +an individual reduces their income to the threshold whenever doing so +leaves them at least £1,000 better off. Set against entitlements which their +model values at between roughly £3,500 and £18,000 a year, this is a low +barrier, and the model consequently predicts that almost everyone with a +financial incentive to hold their income down will do so. That prediction +implies that the 'dominated region' above £100,000 should be nearly empty +for parents of eligible-age children. As set out in Section 4.1, we estimate that +93% of employees and 67% of non-employees whose incomes placed them +in their own dominated region in 2022 remained there. Prior work on +bunching in UK incomes by Adam et al. (2021) provided previously available +evidence of the importance of such frictions, outside the childcare context, +finding substantial non-response to tax incentives. +A related assumption is that all eligible parents take up the full entitlement. +We observe that around 60% of parents with incomes below £100,000 held +a live childcare account in 2022 (Figure 2.1). Parents who do not want the +free hours have nothing to gain by reducing their income, so the population +who would bunch even with no frictions is substantially smaller than the +overall parent population at these incomes. Unlike the previous point, this +was not evident from previously published data. The incomplete take-up by +parents means both that the behavioural effect CBP estimate is too large, +but also that their static cost is too large, so the impact on the post- +behavioural costing is smaller. + +The labour supply model raises two further issues. The first concerns the +elasticity. CBP use a Frisch elasticity of 0.3, which is a reasonable central +value in the labour supply literature. However, their utility function is quasi- +linear and isoelastic in hours, which removes income effects and so makes +the Frisch, compensated and uncompensated elasticities coincide. Since +the model contains no margin beyond labour supply, this single parameter +determines the whole response of ANI, so it is in practice functioning as an +Elasticity of Taxable Income while being interpreted entirely as a labour +supply response. The relevant comparison is therefore with our structural +estimates of 0.13 for employees and 0.19 for non-employees, rather than our +reduced-form figures of 0.31 and 0.45. On that basis their elasticity is 1.5-2.3 +times too large for the total response. As we show, 'paper' responses +(pension contributions, rather than labour supply) account for around half +of the employee response (Section 3.3), so the pure labour supply parameter +we estimate is lower still. Because the income response to a notch scales +approximately with the square root of the elasticity, this difference alone +raises the predicted response by between a quarter and a half. + + + 100 © CenTax + \ No newline at end of file diff --git a/sources/centax-childcare-100k/source.json b/sources/centax-childcare-100k/source.json new file mode 100644 index 000000000..d28e4fb8c --- /dev/null +++ b/sources/centax-childcare-100k/source.json @@ -0,0 +1,43 @@ +{ + "id": "centax-childcare-100k", + "name": "CenTax — Removing the childcare cliff-edge: impacts and cost of reform (September 2026), Table 4.2", + "organization": "Centre for the Analysis of Taxation (CenTax)", + "authors": ["Arun Advani", "Josh Flew", "Jack Pepin-Hall", "Andy Summers"], + "country": "UK", + "claim_class": "reform_score", + "url": "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf", + "data_urls": [ + "https://centax.org.uk/wp-content/uploads/2026/09/AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf" + ], + "raw_sha256": { + "AdvaniFlewPepin-HallSummers2026_Removing-the-childcare-cliff-edge.pdf": "733b7a1d48409071135303c3363482126ead83c4d6e02d9b7f049e4d3113140a", + "pages_layout.txt": "4769e0ee90902559e924a79b531cdb0134747ad66a9327464fa8d668bd488f8a" + }, + "pages_extracted": { + "tool": "pdftotext -layout (poppler 26.09.0), one PDF page at a time", + "pdf_pages": [3, 9, 40, 41, 42, 43, 45, 74, 75, 76, 100, 101], + "printed_page_rule": "printed page = PDF page - 1 (the cover is unnumbered)" + }, + "citation": "Advani, A., Flew, J., Pepin-Hall, J. and Summers, A. (2026), Removing the childcare cliff-edge: impacts and cost of reform, CenTax, September 2026.", + "published": "2026-09", + "fetched": "2026-10-21", + "pdf_metadata": "108 pages; CreationDate 2026-09-09 (pdfinfo); 2,709,003 bytes; HTTP 200 from centax.org.uk on 2026-10-21", + "model": { + "name": "CenTax bunching and static-cost model on HMRC administrative data", + "open_source": false, + "method": "Static cost: the cost of funding hours for children of parents above £100,000 adjusted net income, assuming their use of free hours equals that of children of parents just below the threshold, net of double-counted couples, parents whose partner is not working, and parents already holding live childcare accounts above £100,000; take-up from DfE's Funded early education and childcare statistics. Behavioural gains: an intensive margin (parents who stop holding income below £100,000, with Scottish parents as the control for residual Personal Allowance bunching) and an extensive margin (partners entering work). Three scenarios differ in the assumed fall in the bunching friction parameter after 2022: none (high cost), the 2019-2022 trend continued (central), a 50% fall (low cost).", + "method_basis": "raw/pages_layout.txt: printed pp. 38-40 (section 4.1, scenarios), p. 41 (Table 4.2 and notes), pp. 73-75 (Appendix A, static cost and behavioural revenue gain)", + "data": "HMRC administrative datasets (Table 4.2 source line); the most recent year observed is 2022 (tax year 2021-22)" + }, + "reform": "Remove the £100,000 adjusted-net-income limit on eligibility for the free childcare hours for working parents in England (section 4.2.1, 'Abolish the threshold'). Tax-Free Childcare is not part of the costed reform.", + "baseline_key": "current law (the status quo with the £100,000 limit)", + "period": "Tax years 2027-28 and 2029-30, which the report labels 2028 and 2030 (editorial note, printed page 2: 'tax year 2021-22 is given as 2022')", + "metrics": ["revenue_change (static cost, intensive and extensive margins, net fiscal cost)", "bunching counts (staged in data/externals; not claims)"], + "units": { + "money": "gbp — raw pounds (published in £m), sign as published: negative = cost to the Exchequer", + "bunching": "persons — parents estimated to bunch at £100,000" + }, + "geographies": "England (the free-hours offer)", + "coverage_note": "Table 4.2 in full: 36 cells (3 scenarios x 2 years x 6 columns). The 24 money cells become claims; the 12 bunching counts stay in data/externals with a tallied reason (see scorecard_db/ingest_uk_childcare_100k.py). Table 4.3's 'Abolish the threshold' rows are the central rows of Table 4.2 and are used as a QC cross-check, not staged twice; its other reforms (threshold at £125,000 or £150,000, tapers, 15 hours for under-3s) have no PolicyEngine construction in this lane.", + "diagnosis_upstream": "CenTax's model is closed and runs on HMRC microdata; a divergence classified as an external-model issue becomes a comparison memo, not an upstream PR. Divergences that are ours route to microcosm (data) or policyengine-uk (rules)." +} diff --git a/tests/test_build_chain.py b/tests/test_build_chain.py index d40210f22..c1106535d 100644 --- a/tests/test_build_chain.py +++ b/tests/test_build_chain.py @@ -59,6 +59,7 @@ "uk_ab2025_verdicts", "uk_ab2025_lanes", "uk_benefit_uprating", + "uk_childcare_100k", "nz_budget_scores", "be_pit_reform", "be_jrc", diff --git a/tests/test_export_populations.py b/tests/test_export_populations.py index 3857d336c..56bc6785b 100644 --- a/tests/test_export_populations.py +++ b/tests/test_export_populations.py @@ -411,8 +411,10 @@ def test_integration_committed_db(tmp_path): # historical/comprehensive claims and the existing main feed stay intact. # + 13 UK rows from lane uk-benefit-uprating: the 11 Resolution # Foundation uprating claims, now with parameter-tree results, and the - # ONS and OBR September 2026 CPI claims. - assert dist == {"US": 540, "UK": 5163 + 23_752 + 246 + 13, "BE": 9, "NZ": 12} + # ONS and OBR September 2026 CPI claims; + 36 from uk-childcare-100k: + # the £100,000 childcare limit costings (24 CenTax Table 4.2 cells, 12 + # CBP abolition cells), all opted into display. + assert dist == {"US": 540, "UK": 5163 + 23_752 + 246 + 13 + 36, "BE": 9, "NZ": 12} nz = [row for row in payload["rows"] if row["country"] == "NZ"] assert len(nz) == 12 assert {row["source"] for row in nz} == {"nz_treasury"} @@ -447,7 +449,7 @@ def test_integration_committed_db(tmp_path): assert committed["summary"] == payload["summary"] assert committed["event_archives"] == payload["event_archives"] # Every event is reachable in the deployed copy, with bounded requests. - assert len(payload["rows"]) == 5737 + assert len(payload["rows"]) == 5737 + 36 for archive in payload["event_archives"]: relative = archive["path"].removeprefix("data/") freshly_exported = (tmp_path / relative).read_bytes() diff --git a/tests/test_uk_childcare_100k.py b/tests/test_uk_childcare_100k.py new file mode 100644 index 000000000..4c139a10d --- /dev/null +++ b/tests/test_uk_childcare_100k.py @@ -0,0 +1,2645 @@ +"""The £100,000 childcare limit costings (lane uk-childcare-100k): CenTax's +Table 4.2 and the Centre for British Progress appendix are exactly what the +adapters derive from the vendored PDFs; the QC catches broken tables; the +certified-pair artifact ran on exactly its variant's pairing and moves only +the childcare heads, in England; every result's status is one stated +annotation; every certified run and the published dashboard figure stay in +the history; and a reform estimate is never a calibration target.""" + +import ast +import builtins +import copy +import datetime as dt +import hashlib +import importlib.util +import json +import re +import shutil +import sqlite3 +import subprocess +import sys +import tempfile +from collections import Counter +from itertools import product +from pathlib import Path +from types import SimpleNamespace + +import numpy as np +import pytest +from hypothesis import example, given, settings +from hypothesis import strategies as st + +from scorecard_db import ingest_uk_childcare_100k as lane +from scorecard_db.models import STANDARD_CONDITIONS +from scorecard_db.models import CalibrationRelationship as CR +from scorecard_db.relationships import uk_relationship + +REPO = Path(__file__).resolve().parent.parent +CENTAX_SRC = REPO / "sources" / "centax-childcare-100k" +CBP_SRC = REPO / "sources" / "cbp-rates-and-wrongs" +DB = REPO / "data" / "scorecard.db" +CERTIFIED = json.loads((REPO / "data" / "uk" / "certified_bundle.json").read_text()) +VARIANTS = json.loads((REPO / "data" / "uk" / "replay_variants.json").read_text()) + +# Table 4.2 as printed (printed page 41), read from the PDF on 2026-10-21: +# (scenario, year) -> (bunching before, after, static, intensive, extensive, net), £m +TABLE_4_2 = { + ("central", 2028): (13_360, 3_850, -800, 150, 120, -520), + ("central", 2030): (16_280, 4_300, -980, 210, 130, -640), + ("high_cost", 2028): (9_710, 3_850, -820, 100, 120, -610), + ("high_cost", 2030): (10_860, 4_300, -1_010, 120, 120, -770), + ("low_cost", 2028): (58_510, 3_850, -580, 630, 90, 140), + ("low_cost", 2030): (67_650, 4_300, -700, 770, 90, 160), +} +# CBP technical appendix, "Abolish Cliff" rows, £bn as printed +CBP_ABOLISH = { + "endogenous_pension_exogenous_labour": (0.070, 0.077, 0.091, 0.132, 0.147, 0.172), + "endogenous_labour_no_pension": (0.330, 0.349, 0.399, 0.427, 0.462, 0.507), +} + + +def _module(name: str, path: Path): + # a unique module name: several lanes ship an adapter.py; registered as + # an import registers it, so sys.modules[__name__] reads as in a real run + spec = importlib.util.spec_from_file_location(name, path) + module = importlib.util.module_from_spec(spec) + sys.modules[name] = module + spec.loader.exec_module(module) + return module + + +@pytest.fixture(scope="module") +def centax(): + return _module("centax_childcare_adapter", CENTAX_SRC / "adapter.py") + + +@pytest.fixture(scope="module") +def cbp(): + return _module("cbp_rates_and_wrongs_adapter", CBP_SRC / "adapter.py") + + +@pytest.fixture(scope="module") +def artifact(): + return lane.load_artifact()[0] + + +# --- the vendored sources --------------------------------------------------- + + +@pytest.mark.parametrize("src", [CENTAX_SRC, CBP_SRC]) +def test_raw_files_match_their_recorded_hashes(src): + recorded = json.loads((src / "source.json").read_text())["raw_sha256"] + assert set(recorded) == {p.name for p in (src / "raw").iterdir()} + for name, digest in recorded.items(): + assert hashlib.sha256((src / "raw" / name).read_bytes()).hexdigest() == digest + + +@pytest.mark.skipif(shutil.which("pdftotext") is None, reason="needs poppler") +@pytest.mark.parametrize("src", [CENTAX_SRC, CBP_SRC]) +def test_page_extract_is_a_faithful_extract_of_the_pdf(src): + """Every page of the extract re-extracts from the PDF byte for byte, so + a number in the extract is a number in the publication.""" + spec = json.loads((src / "source.json").read_text()) + pdf = next(p for p in (src / "raw").iterdir() if p.suffix == ".pdf") + text = (src / "raw" / "pages_layout.txt").read_text() + chunks = text.split("=== PDF page ")[1:] + assert [int(c.split(" ", 1)[0].split("=")[0]) for c in chunks] == spec[ + "pages_extracted" + ]["pdf_pages"] + for chunk in chunks: + header, body = chunk.split("===\n", 1) + page = int(header.split()[0]) + fresh = subprocess.run( + ["pdftotext", "-layout", "-f", str(page), "-l", str(page), str(pdf), "-"], + capture_output=True, + text=True, + check=True, + ).stdout + assert body == fresh, f"{src.name} PDF page {page}" + + +def test_centax_extract_is_the_adapters_output_and_the_published_table(centax): + rows = centax.tidy(centax.parse_table_4_2(centax.pages()[42])) + assert rows == json.loads( + (REPO / "data/externals/centax-childcare-100k.json").read_text() + ) + assert len(rows) == centax.EXPECTED_ROWS == 36 + columns = (*centax.COUNTS, *centax.MONEY) + for (scenario, year), printed in TABLE_4_2.items(): + got = { + r["column"]: r["value"] + for r in rows + if (r["variant"], r["period"]) == (scenario, year) + } + want = { + c: v * (1_000_000 if c in centax.MONEY else 1) + for c, v in zip(columns, printed) + } + assert got == want, (scenario, year) + assert {r["fy"] for r in rows if r["period"] == 2028} == {"2027-28"} + assert {r["fy"] for r in rows if r["period"] == 2030} == {"2029-30"} + + +def test_centax_run_passes_every_qc(centax): + assert len(centax.run()) == 36 + + +def test_centax_qc_catches_a_component_that_no_longer_adds_up(centax): + table = centax.parse_table_4_2(centax.pages()[42]) + table[("central", 2028)]["static_cost"] -= 30 + with pytest.raises(ValueError, match="components miss net"): + centax.check_sums(table) + + +def test_centax_qc_catches_table_4_3_disagreeing(centax): + pages = centax.pages() + table = centax.parse_table_4_2(pages[42]) + table[("central", 2030)]["net_fiscal_cost"] = -650 + with pytest.raises(ValueError, match="Table 4.3"): + centax.check_table_4_3(table, pages[45]) + + +def test_centax_qc_catches_prose_disagreeing(centax): + pages = centax.pages() + table = centax.parse_table_4_2(pages[42]) + table[("central", 2030)]["intensive_margin"] = 200 + with pytest.raises(ValueError, match="prose"): + centax.check_prose(table, pages[42] + pages[43]) + + +def test_centax_qc_catches_a_shifted_row(centax): + text = centax.pages()[42].replace("13,360", "13,360 9", 1) + with pytest.raises(ValueError): + centax.parse_table_4_2(text) + + +def test_cbp_extract_is_the_adapters_output_and_the_published_tables(cbp): + rows = cbp.run() + assert rows == json.loads( + (REPO / "data/externals/cbp-rates-and-wrongs.json").read_text() + ) + assert len(rows) == cbp.EXPECTED_ROWS == 72 + for model, printed in CBP_ABOLISH.items(): + got = [ + r["value"] + for r in rows + if (r["variant"], r["policy"]) == (model, "abolish_cliff") + ] + assert got == [round(v * 1e9) for v in printed] + assert [r["fy"] for r in rows[:6]] == [ + "2026-27", + "2027-28", + "2028-29", + "2029-30", + "2030-31", + "2031-32", + ] + assert [r["period"] for r in rows[:6]] == [2027, 2028, 2029, 2030, 2031, 2032] + + +def test_cbp_qc_catches_a_column_shift(cbp): + pages = cbp.pages() + table = cbp.parse_table("2", pages[8]) + table["taper_1pct"][3] = ("0.100", 0.100) + with pytest.raises(ValueError, match="not rising"): + cbp.check_order("2", table) + + +# --- properties of the adapters' arithmetic ---------------------------------- + + +@settings(deadline=None) +@given( + st.lists( + st.floats(min_value=-2_000, max_value=2_000, allow_nan=False), + min_size=3, + max_size=3, + ) +) +def test_rounding_tolerance_covers_any_three_components(parts): + """Invariant behind QC 4: if net is the exact sum of three components and + each of the four figures is rounded to the nearest £10m, the rounded + components miss the rounded net by at most SUM_TOLERANCE_GBP_M.""" + a = _module("centax_childcare_adapter_p", CENTAX_SRC / "adapter.py") + step = a.ROUNDING_GBP_M + + def rnd(x): + return step * round(x / step) + + net = sum(parts) + assert abs(sum(rnd(p) for p in parts) - rnd(net)) <= a.SUM_TOLERANCE_GBP_M + + +@settings(deadline=None) +@given(st.integers(min_value=-99_999, max_value=99_999)) +def test_money_cells_round_trip(value): + a = _module("centax_childcare_adapter_r", CENTAX_SRC / "adapter.py") + cell = f"{'+' if value >= 0 else '-'}{abs(value):,}m" + assert a._gbp_m(cell) == value + + +# --- the certified-pair artifact --------------------------------------------- + + +def test_artifact_ran_on_the_certified_bundle_without_behaviour(artifact): + assert artifact["data_bundle"] == CERTIFIED["revision"] + assert artifact["dataset"]["sha256"] == CERTIFIED["sha256"] + assert artifact["engine_version"] == "2.89.2" + assert artifact["behavioural_response"] == "none (static)" + for world in (*artifact["worlds"].values(), *artifact["diagnostics"].values()): + assert world["dataset_sha256_before"] == world["dataset_sha256_after"] + + +def test_each_world_removes_exactly_its_limits(artifact): + hours = "gov.dfe.extended_childcare_entitlement.income.limit" + tfc = "gov.hmrc.tax_free_childcare.income.income_limit" + for year, block in artifact["worlds"]["baseline"]["years"].items(): + assert block["limits_executed"] == {hours: 100_000.0, tfc: 100_000.0}, year + for year, block in artifact["worlds"]["free_hours"]["years"].items(): + assert block["limits_executed"] == {hours: 1e100, tfc: 100_000.0}, year + for year, block in artifact["worlds"]["free_hours_and_tfc"]["years"].items(): + assert block["limits_executed"] == {hours: 1e100, tfc: 1e100}, year + + +def _d(art, world, year, var, group="worlds", base="baseline", sums="sums_gbp"): + years = art[group] + return ( + years[world]["years"][year][sums][var] - years[base]["years"][year][sums][var] + ) + + +def test_reforms_move_only_the_childcare_heads(artifact): + heads = (*lane.FREE_HOURS, lane.TFC) + for world in ("free_hours", "free_hours_and_tfc"): + for year in artifact["worlds"][world]["years"]: + moved = sum(_d(artifact, world, year, v) for v in heads) + assert abs(_d(artifact, world, year, "gov_spending") - moved) < 1e6 + assert abs(_d(artifact, world, year, "gov_tax")) < 1e6 + # removing the 30-hours limit leaves Tax-Free Childcare alone + for year in artifact["worlds"]["free_hours"]["years"]: + assert abs(_d(artifact, "free_hours", year, lane.TFC)) < 1e6 + + +def test_free_hours_leg_is_the_same_in_both_worlds(artifact): + """Differential: the TFC limit does not touch the free hours, so the two + reform worlds agree on every free-hours head.""" + for year in artifact["years"]: + for v in lane.FREE_HOURS: + a = artifact["worlds"]["free_hours"]["years"][str(year)]["sums_gbp"][v] + b = artifact["worlds"]["free_hours_and_tfc"]["years"][str(year)][ + "sums_gbp" + ][v] + assert a == pytest.approx(b, rel=1e-12, abs=1.0), (year, v) + + +def test_removing_the_limit_costs_money_and_switches_universal_hours_off(artifact): + for year in artifact["worlds"]["free_hours"]["years"]: + assert lane.counterpart(artifact, lane.CENTAX, int(year)) < 0 + assert _d(artifact, "free_hours", year, "extended_childcare_entitlement") > 0 + assert _d(artifact, "free_hours", year, "universal_childcare_entitlement") < 0 + + +def test_full_use_diagnostic_took_effect_and_costs_more(artifact): + diag = artifact["diagnostics"] + for world in diag.values(): + assert world["input_override"]["value"] == 30.0 + for mean in world["input_override"]["post_set_weighted_mean"].values(): + assert mean == pytest.approx(30.0) + # set before anything that reads it was calculated ... + overridden = world["input_override"]["dependents_calculated_before_override"] + assert overridden and not any(overridden.values()) + # ... so it reached the entitlement: full use funds more extended hours + for year in diag["baseline__full_use_hours"]["years"]: + full = diag["baseline__full_use_hours"]["years"][year]["sums_gbp"] + drawn = artifact["worlds"]["baseline"]["years"][year]["sums_gbp"] + ext = "extended_childcare_entitlement" + assert full[ext] > drawn[ext] + for year in diag["baseline__full_use_hours"]["years"]: + full = sum( + diag["free_hours__full_use_hours"]["years"][year]["sums_gbp"][v] + - diag["baseline__full_use_hours"]["years"][year]["sums_gbp"][v] + for v in lane.FREE_HOURS + ) + assert full > -lane.counterpart(artifact, lane.CENTAX, int(year)) + + +def test_no_child_under_one_receives_extended_hours(artifact): + """The pe-under-ones construction note states both facts for every world + and year; the artifact must keep them true.""" + for group in ("worlds", "diagnostics"): + for world in artifact[group].values(): + for block in world["years"].values(): + assert block["counts"]["children_under_one_receiving_extended"] == 0 + assert block["counts"]["people_with_fractional_age"] == 0 + assert block["counts"]["children_under_one"] > 0 + + +def test_counterparts_count_england_only(artifact): + """Both offers are England's: each counterpart is minus the change in + the heads for people in English households, no free-hours head moves + outside England, and England's Tax-Free Childcare is part of the UK's.""" + for world, source in ( + ("free_hours", lane.CENTAX), + ("free_hours_and_tfc", lane.CBP), + ): + for year in artifact["worlds"][world]["years"]: + eng = { + v: _d(artifact, world, year, v, sums=lane.ENGLAND) + for v in (*lane.FREE_HOURS, lane.TFC) + } + assert lane.counterpart(artifact, source, int(year)) == pytest.approx( + -sum(eng[v] for v in lane.VARIABLES[source]), abs=1.0 + ) + for v in lane.FREE_HOURS: + assert abs(_d(artifact, world, year, v) - eng[v]) < 1e6, (year, v) + block = artifact["worlds"][world]["years"][year] + assert ( + 0 <= block["sums_gbp_england"][lane.TFC] <= block["sums_gbp"][lane.TFC] + ) + + +def test_free_hours_moving_outside_england_is_refused(artifact): + bad = copy.deepcopy(artifact) + year = next(iter(bad["worlds"]["free_hours"]["years"])) + sums = bad["worlds"]["free_hours"]["years"][year]["sums_gbp"] + # a UK-wide move with no English counterpart, kept inside gov_spending + sums["extended_childcare_entitlement"] += 5e6 + sums["gov_spending"] += 5e6 + with pytest.raises(ValueError, match="outside England"): + lane.measured_notes(bad, lane.CENTAX, int(year), "fy", {}) + + +def test_artifact_records_the_inputs_and_metadata_it_read(artifact): + compute = _module( + "compute_uk_childcare_100k", REPO / "pipeline" / "compute_uk_childcare_100k.py" + ) + assert set(artifact["variable_metadata"]) == set(compute.METADATA) + for group in ("worlds", "diagnostics"): + for world in artifact[group].values(): + assert set(world["dataset_inputs"]) == set(compute.DATASET_INPUTS) + for block in world["years"].values(): + assert set(block["take_up"]) == set(compute.TAKE_UP) + assert all(0 <= x <= 1 for x in block["take_up"].values()) + + +class Unvouched(AssertionError): + """The metadata contract cannot vouch for a name the source calculates.""" + + +_COMPS = (ast.ListComp, ast.SetComp, ast.DictComp, ast.GeneratorExp) +_SCOPES = (ast.FunctionDef, ast.AsyncFunctionDef, ast.Lambda, ast.ClassDef, *_COMPS) +# Forms that bind a name out of a scope scan's sight (an assignment +# expression in a comprehension or a default binds the enclosing function; +# global and nonlocal reach other scopes; match patterns bind bare strings), +# or reach calculate other than by a direct call. The compute script uses +# none, so the scan refuses each wherever it appears. +_HIDDEN = (ast.NamedExpr, ast.Global, ast.Nonlocal, ast.Match) +_DYNAMIC = frozenset( + { + "eval", + "exec", + "compile", + "globals", + "locals", + "vars", + "__import__", + "setattr", + "delattr", + "attrgetter", + "methodcaller", + "__builtins__", + } +) + + +# Vetted functions that walk to whatever attribute a dotted path names: +# compute_uk_ab2025._resolve_parameter is getattr in a loop, so the scan +# holds each call to a direct call whose path it resolves, as it resolves a +# calculate name, to parameter paths that never name calculate. +_GETTERS = frozenset({"_resolve_parameter"}) +_PARAMETER_PATH = re.compile(r"gov(\.[a-z][a-z0-9_]*(\[\d+\])?)+") + +# What the compute script reaches numpy, the engine and the standard library +# through: each import, builtin name and attribute name it uses (getattr's +# literal names included). Each was read for a route to calculate other than +# a direct .calculate(...) call. The one such route is a getter in _GETTERS, +# held to parameter paths; so with the forms refused here every calculate +# the script's own code makes is a direct call the scan resolves. A name not +# listed is refused until it is read and added. +_VETTED_IMPORTS = frozenset( + { + ("", "argparse", None), + ("", "datetime", "dt"), + ("", "gc", None), + ("", "inspect", None), + ("", "json", None), + ("", "numpy", "np"), + ("", "string", None), + ("", "sys", None), + ("", "time", None), + ("__future__", "annotations", None), + ("pathlib", "Path", None), + ("pipeline", "compute_uk_ab2025", "ab"), + ("pipeline", "compute_uk_event", "event"), + ("pipeline", "compute_uk_obr_costings", "fiscal"), + ("pipeline", "modal_uk_event", None), + ("pipeline", "uk_replay_variants", "variants"), + } +) +_VETTED_NAMES = frozenset( + { + "SystemExit", + "__doc__", + "__file__", + "__name__", + "any", + "bool", + "dict", + "float", + "getattr", + "int", + "isinstance", + "len", + "list", + "print", + "set", + "sorted", + "str", + "tuple", + "type", + } +) +_VETTED_ATTRIBUTES = frozenset( + { + "ArgumentParser", + "HF_COMMIT", + "NO_LIMIT", + "PACKAGES", + "_resolve_parameter", + "abs", + "add_argument", + "allclose", + "array_equal", + "asarray", + "ascii_lowercase", + "ascii_uppercase", + "assert_managed_bundle", + "build_sim", + "calculate", + "check_variant_environment", + "collect", + "count", + "data_bundle_id", + "dataset", + "dataset_file", + "datetime", + "default_value", + "defined_for", + "definition_period", + "diagnostic_years", + "documentation", + "dumps", + "entity", + "exit", + "full", + "get", + "get_holder", + "get_known_periods", + "get_variant", + "getsourcefile", + "insert", + "is_legacy", + "isoformat", + "items", + "key", + "label", + "maketrans", + "map_result", + "mkdir", + "ndarray", + "now", + "output_dir", + "parameters", + "parent", + "parse_args", + "path", + "perf_counter", + "policyengine_bundle", + "pop", + "populations", + "preflight", + "reform_dict", + "relative_to", + "relative_to_root", + "resolve", + "round", + "run_id", + "set_input", + "split", + "sum", + "tax_benefit_system", + "timezone", + "translate", + "unique", + "unit", + "utc", + "values", + "variables", + "variant", + "variant_preflight", + "verify_import_origins", + "verify_runtime_dataset_sha256", + "weights", + "write_text", + "years", + } +) + + +def _used(tree) -> tuple[frozenset, frozenset, frozenset]: + """(imports, builtin and dunder names, attribute names) a module uses; + an import is (module with any leading dots, name, alias).""" + imports, names, attributes = set(), set(), set() + for node in ast.walk(tree): + if isinstance(node, ast.Import): + imports |= {("", a.name, a.asname) for a in node.names} + elif isinstance(node, ast.ImportFrom): + module = "." * node.level + (node.module or "") + imports |= {(module, a.name, a.asname) for a in node.names} + elif isinstance(node, ast.Name) and ( + hasattr(builtins, node.id) or node.id.startswith("__") + ): + names.add(node.id) + elif isinstance(node, ast.Attribute): + attributes.add(node.attr) + if ( + isinstance(node, ast.Call) + and isinstance(node.func, ast.Name) + and node.func.id == "getattr" + and len(node.args) > 1 + and isinstance(node.args[1], ast.Constant) + ): + attributes.add(node.args[1].value) + return frozenset(imports), frozenset(names), frozenset(attributes) + + +def _binds(node, name: str) -> bool: + """Does this node bind `name` in the scope it sits in?""" + if isinstance(node, ast.Name): + return node.id == name and isinstance(node.ctx, ast.Store | ast.Del) + if isinstance(node, ast.arg): + return node.arg == name + if isinstance(node, ast.alias): + return (node.asname or node.name).split(".")[0] == name + if isinstance(node, ast.ExceptHandler): + return node.name == name + if isinstance(node, ast.FunctionDef | ast.AsyncFunctionDef | ast.ClassDef): + return node.name == name + return False + + +def _stores(scope, name: str) -> list: + """Every binding of `name` in this scope, not in scopes nested in it.""" + out = [] + + def visit(node): + for child in ast.iter_child_nodes(node): + if _binds(child, name): + out.append(child) + if not isinstance(child, _SCOPES): + visit(child) + + if isinstance(scope, ast.FunctionDef | ast.AsyncFunctionDef | ast.Lambda): + out += [a for a in ast.walk(scope.args) if _binds(a, name)] + visit(scope) + return out + + +def _refuse_hidden(tree, parents) -> None: + """Raise on any form in _HIDDEN or _DYNAMIC, on 'calculate' or a getter + in _GETTERS named as a string or reached other than by a direct call, on + getattr without a literal attribute name, and on an attribute write that + rebinds a module constant of another object (sys.modules[__name__].X).""" + constants = { + t.id + for node in tree.body + if isinstance(node, ast.Assign | ast.AnnAssign) + for t in (node.targets if isinstance(node, ast.Assign) else [node.target]) + if isinstance(t, ast.Name) + } + for node in ast.walk(tree): + line = getattr(node, "lineno", "?") + parent = parents.get(node) + called = isinstance(parent, ast.Call) and parent.func is node + if isinstance(node, _HIDDEN): + raise Unvouched(f"line {line}: {type(node).__name__} binds out of sight") + if isinstance(node, ast.Name) and node.id in _DYNAMIC: + raise Unvouched(f"line {line}: {node.id} can bind or call by name") + if isinstance(node, ast.Constant) and node.value in {"calculate", *_GETTERS}: + raise Unvouched(f"line {line}: {node.value!r} named by a string") + if isinstance(node, ast.Name) and node.id == "getattr": + attr = parent.args[1] if called and len(parent.args) > 1 else None + if not (isinstance(attr, ast.Constant) and isinstance(attr.value, str)): + raise Unvouched(f"line {line}: getattr without a literal name") + if isinstance(node, ast.Attribute): + if node.attr in {"calculate", *_GETTERS} and not called: + raise Unvouched(f"line {line}: {node.attr} used but not called") + if node.attr == "__dict__": + raise Unvouched(f"line {line}: a namespace read as a dict") + if isinstance(node.ctx, ast.Store | ast.Del) and node.attr in constants: + raise Unvouched(f"line {line}: an attribute write rebinds {node.attr}") + + +def _refuse_unvetted(tree) -> None: + """Raise on any import, builtin name or attribute name not vetted.""" + vetted = (_VETTED_IMPORTS, _VETTED_NAMES, _VETTED_ATTRIBUTES) + for kind, used, known in zip(("import", "name", "attribute"), _used(tree), vetted): + if used - known: + raise Unvouched(f"unvetted {kind}s {sorted(used - known, key=str)}") + + +def _refuse_classes(tree, parents) -> None: + """Raise on a class the script defines, by a class statement or by + type() with more than one argument. The script defines none. A method + of such a class (resolve, __fspath__, __hash__, __float__, ...) is the + script's code that library or engine code can call implicitly, inside + a stub, an array conversion or a lookup, where the stand-in run does + not follow.""" + for node in ast.walk(tree): + line = getattr(node, "lineno", "?") + if isinstance(node, ast.ClassDef): + raise Unvouched(f"line {line}: class {node.name} could carry a callback") + if isinstance(node, ast.Name) and node.id == "type": + call = parents.get(node) + if not ( + isinstance(call, ast.Call) + and call.func is node + and len(call.args) == 1 + and not call.keywords + and not isinstance(call.args[0], ast.Starred) + ): + raise Unvouched(f"line {line}: type() other than on one object") + + +def _calculated(path: Path, compute) -> set[str]: + """Every name the source passes to sim.calculate, read as Python scopes + read it. A name must be a literal, or bound exactly once on the way out + from the call: by the plain target of a loop or comprehension around the + call, over a tuple display or a module tuple; or nowhere, a module + string constant. Any other binding (an assignment, a second loop, a + rebinding generator, an argument, a tuple target) raises, and so does + any form _refuse_hidden lists, so a name cannot change unnoticed by a + route this scan reads past. Each call of a getter in _GETTERS must have + a path resolved the same way, to parameter paths only; then any name not + vetted raises, so no route other than a direct call reaches calculate, + and so does any class (_refuse_classes), so no method the script defines + is called back unread. The stand-in run backs it up (_observed).""" + tree = ast.parse(path.read_text()) + parents = { + child: node for node in ast.walk(tree) for child in ast.iter_child_nodes(node) + } + _refuse_hidden(tree, parents) + + def scopes(node): + while node in parents: + node = parents[node] + if isinstance(node, _SCOPES): + yield node + + def module_value(name: ast.Name): + for scope in scopes(name): + if _stores(scope, name.id): + raise Unvouched(f"{name.id!r} is rebound before line {name.lineno}") + found = _stores(tree, name.id) + if len(found) != 1 or not isinstance( + parents[found[0]], ast.Assign | ast.AnnAssign + ): + raise Unvouched( + f"{name.id!r} is not one plain module assignment ({len(found)} bindings)" + ) + return getattr(compute, name.id) + + def values(expr) -> list: + if isinstance(expr, ast.Constant): + return [expr.value] + if isinstance(expr, ast.Name): + value = module_value(expr) + if not isinstance(value, tuple): + raise Unvouched(f"{expr.id!r} is not a module tuple") + return list(value) + if isinstance(expr, ast.Tuple | ast.List): + out = [] + for e in expr.elts: + if isinstance(e, ast.Starred): + out += values(e.value) + elif isinstance(e, ast.Name): + value = module_value(e) + if not isinstance(value, str): + raise Unvouched(f"{e.id!r} is not a module string") + out.append(value) + else: + out += values(e) + return out + raise Unvouched(f"unresolved loop iterable {ast.dump(expr)}") + + def inside(node, block) -> bool: + while node in parents: + if any(node is b for b in block): + return True + node = parents[node] + return False + + def resolve(name: ast.Name) -> list: + for scope in scopes(name): + found = _stores(scope, name.id) + if not found: + continue + if len(found) > 1: + raise Unvouched( + f"{name.id!r} is bound {len(found)} times around line {name.lineno}" + ) + target = parents[found[0]] + comprehension = ( + isinstance(scope, _COMPS) + and isinstance(target, ast.comprehension) + and target.target is found[0] + ) + loop = ( + isinstance(target, ast.For) + and target.target is found[0] + and inside(name, target.body) + ) + if comprehension or loop: + return values(target.iter) + raise Unvouched( + f"{name.id!r} at line {name.lineno} is bound by " + f"{type(target).__name__}, not a loop over known names" + ) + value = module_value(name) + if not isinstance(value, str): + raise Unvouched(f"{name.id!r} is not a module string") + return [value] + + calculated = set() + for node in ast.walk(tree): + if not ( + isinstance(node, ast.Call) + and isinstance(node.func, ast.Attribute) + and node.func.attr == "calculate" + ): + continue + if node.keywords or not node.args or isinstance(node.args[0], ast.Starred): + raise Unvouched(f"line {node.lineno}: calculate not called (name, period)") + arg = node.args[0] + if isinstance(arg, ast.Constant): + calculated.add(arg.value) + elif isinstance(arg, ast.Name): + calculated |= set(resolve(arg)) + else: + raise Unvouched(f"unresolved calculate argument {ast.dump(arg)}") + for node in ast.walk(tree): + if not ( + isinstance(node, ast.Call) + and isinstance(node.func, ast.Attribute) + and node.func.attr in _GETTERS + ): + continue + if ( + node.keywords + or len(node.args) != 2 + or isinstance(node.args[1], ast.Starred) + ): + raise Unvouched( + f"line {node.lineno}: {node.func.attr} not called (node, path)" + ) + arg = node.args[1] + if isinstance(arg, ast.Constant): + paths = [arg.value] + elif isinstance(arg, ast.Name): + paths = resolve(arg) + else: + raise Unvouched( + f"line {node.lineno}: unresolved getter path {ast.dump(arg)}" + ) + for p in paths: + if not (isinstance(p, str) and _PARAMETER_PATH.fullmatch(p)) or ( + "calculate" in p.split(".") + ): + raise Unvouched(f"line {node.lineno}: {p!r} is not a parameter path") + _refuse_unvetted(tree) + _refuse_classes(tree, parents) + return calculated + + +_PLAIN = (str, int, float, bool, type(None), type(Path())) + + +def _plain(*values) -> None: + """Refuse any value whose type is not exactly one of _PLAIN, or a dict, + list or tuple of them. A stub drops what it is passed, so an object the + script defines could carry a callback that the real function would make + (resolve, __fspath__, __eq__, ...) and the stand-in run would not see.""" + for v in values: + if type(v) is dict: + _plain(*v.keys(), *v.values()) + elif type(v) in (list, tuple): + _plain(*v) + elif type(v) not in _PLAIN: + raise Unvouched(f"a stub was passed a {type(v).__name__}") + + +def _array(values) -> None: + """Refuse anything but a plain ndarray of numbers, bools or strings: the + engine converts an input array (astype(float)), which calls each member + of an object array.""" + if type(values) is not np.ndarray: + raise Unvouched(f"a stub was passed a {type(values).__name__}, not an array") + if values.dtype.hasobject: + raise Unvouched(f"a stub was passed an array of {values.dtype}") + + +def _stub(result): + """A function outside the script that returns `result` and refuses any + argument that is not plain (_plain).""" + + def call(*args, **kwargs): + _plain(*args, *kwargs.values()) + return result + + return call + + +class _Series: + def __init__(self, values): + self.values = np.asarray(values) + self.weights = np.ones(len(self.values)) + + def sum(self): + return float((self.values * self.weights).sum()) + + +class _Parameter: + """A parameter node: reading a child names it, and calling it on a date + gives the reform's value there (start.stop, both days in) or £100,000.""" + + def __init__(self, reform, path=""): + self._reform, self._path = reform, path + + def __getattr__(self, part): + return _Parameter(self._reform, f"{self._path}.{part}".lstrip(".")) + + def __call__(self, date): + _plain(date) + for period, value in self._reform.get(self._path, {}).items(): + start, stop = period.split(".") + if start <= date <= stop: + return float(value) + return 100_000.0 + + +class _Variable: + def __init__(self, name): + self.label = name + self.entity = SimpleNamespace(key="person") + self.definition_period = "year" + self.unit = self.documentation = self.defined_for = None + self.default_value = 0.0 + + +class _Variables(dict): + def __missing__(self, name): + return _Variable(name) + + +class _StandIn: + """Stands in for each simulation the compute script builds, with no + engine and no data, and records each (name, period) passed to + calculate. It offers only what the script reads; reading anything else + raises, so no other engine route goes unseen, and each method refuses an + argument that is not plain (_plain) or an array. The hours input is + drawn below 30 until set_input sets it, so the ordinary and full-use + worlds each take their own branch, and every household but one is in + England.""" + + DRAWN = (12.0, 20.0, 25.0) + + def __init__(self, run, reform, years, source): + self._run, self._inputs = run, {} + self.dataset = SimpleNamespace(years=list(years)) + self.populations = {"benunit": SimpleNamespace(count=len(self.DRAWN))} + self.policyengine_bundle = {"runtime_dataset_source": str(source)} + self.tax_benefit_system = SimpleNamespace( + parameters=_Parameter(reform or {}), variables=_Variables() + ) + + def __getattr__(self, name): + raise Unvouched(f"the run read sim.{name}, which the stand-in does not offer") + + def calculate(self, variable, period): + _plain(variable, period) + self._run.calls.append((variable, int(period))) + if variable == "country": + return _Series(["ENGLAND", "WALES", "ENGLAND"]) + if variable == lane.HOURS_USAGE_INPUT: + key = (variable, int(period)) + self._run.hours.append( + (int(period), "set" if key in self._inputs else "drawn") + ) + return _Series(self._inputs.get(key, self.DRAWN)) + if (variable, int(period)) in self._inputs: + return _Series(self._inputs[variable, int(period)]) + return _Series([1000.0, 0.0, 500.0]) + + def map_result(self, values, *entities, **how): + _array(values) + _plain(*entities, *how.values()) + return values + + def get_holder(self, variable): + _plain(variable) + return SimpleNamespace(get_known_periods=list) + + def set_input(self, variable, period, values): + _plain(variable, period) + _array(values) + self._inputs[variable, int(period)] = values + + +def _observed(compute, monkeypatch) -> SimpleNamespace: + """The compute script's main() run end to end on stand-in simulations: + every world in every default year, and both full-use worlds. Only what + lies outside the script and needs the engine, the data or the variant + registry is stubbed: the variant record and preflight, the dataset hash, + the bundle checks, the import-origin check and the simulation it builds + (ab.build_sim). Each stub refuses an argument that is not plain, and the + path helper (fiscal.relative_to_root) runs as it is. Returns each + (name, period) passed to calculate, each read of the hours input (drawn + or set) and each world's reform; the artifact written must record + metadata for exactly METADATA.""" + run = SimpleNamespace(calls=[], hours=[], reforms=[]) + years = list(range(2023, 2031)) + out = Path(tempfile.mkdtemp(prefix="childcare-stand-in-")).resolve() + source = out / "dataset.h5" + sha = "0" * 64 + variant = { + "variant_id": "stand-in", + "data_years": years, + "identity": { + "dataset": {"revision": "stand-in", "sha256": sha, "repo_id": "stand-in"} + }, + "identity_sha256": sha, + } + pre = { + "installed": { + "packages": { + p: {"version": "0", "record_sha256": sha} + for p in compute.variants.PACKAGES + }, + "python": "stand-in", + "packaged_bundle_manifest_sha256": sha, + }, + "runtime_dataset_source": str(source), + "sha256": sha, + "release_bundle": {}, + "data_years": years, + } + + def build(reform): + _plain(reform) + run.reforms.append(reform) + return _StandIn(run, reform, years, source) + + monkeypatch.setattr(compute, "ROOT", out) + monkeypatch.setattr(compute.variants, "get_variant", _stub(variant)) + monkeypatch.setattr(compute.variants, "is_legacy", _stub(False)) + monkeypatch.setattr(compute.variants, "verify_import_origins", _stub(None)) + monkeypatch.setattr(compute.event, "variant_preflight", _stub(pre)) + monkeypatch.setattr(compute.fiscal, "data_bundle_id", _stub("stand-in")) + monkeypatch.setattr(compute.fiscal, "verify_runtime_dataset_sha256", _stub(sha)) + monkeypatch.setattr(compute.fiscal, "assert_managed_bundle", _stub(None)) + monkeypatch.setattr(compute.ab, "build_sim", build) + try: + assert compute.main(["--output-dir", str(out)]) == 0 + written = json.loads((out / "stand-in.json").read_text()) + finally: + shutil.rmtree(out) + assert set(written["variable_metadata"]) == set(compute.METADATA) + return run + + +def _contract(path: Path, name: str, monkeypatch) -> set[str]: + """The pe_meta contract on one source file: the scan resolves every + calculate argument, the stand-in run of main() sees exactly those + names, and METADATA lists them all.""" + compute = _module(name, path) + scanned = _calculated(path, compute) + seen = {v for v, _ in _observed(compute, monkeypatch).calls} + if seen != scanned: + raise Unvouched(f"the run and the scan differ on {sorted(seen ^ scanned)}") + if not scanned <= set(compute.METADATA): + raise Unvouched(f"no metadata for {sorted(scanned - set(compute.METADATA))}") + return scanned + + +def test_every_variable_the_run_calculates_has_its_metadata_recorded(monkeypatch): + """The pe_meta contract, read off the source and observed on a + recording simulation: each sim.calculate name is in METADATA, with each + loop variable read in its own scope.""" + path = REPO / "pipeline" / "compute_uk_childcare_100k.py" + compute = _module("compute_uk_childcare_100k_src", path) + calculated = _contract(path, "compute_uk_childcare_100k_contract", monkeypatch) + assert calculated == _calculated(path, compute) + # both reused loop names resolve to their own tuples + assert set(compute.SUMMED) | set(compute.TAKE_UP) <= calculated + assert { + "is_child", + "is_child_receiving_extended_childcare", + "country", + } <= calculated + assert calculated <= set(compute.METADATA) + # each vetted name is one the script uses, so the lists are its own + assert _used(ast.parse(path.read_text())) == ( + _VETTED_IMPORTS, + _VETTED_NAMES, + _VETTED_ATTRIBUTES, + ) + + +def test_the_stand_in_run_reaches_every_world_year_and_hours_branch(monkeypatch): + """main() builds the three worlds and the two full-use worlds, the + calculations run in every default year, and the hours input is read + drawn in the ordinary worlds and set to 30 in the full-use ones.""" + path = REPO / "pipeline" / "compute_uk_childcare_100k.py" + compute = _module("compute_uk_childcare_100k_reach", path) + run = _observed(compute, monkeypatch) + main = compute.worlds(compute.YEARS) + diag = compute.worlds(compute.DIAGNOSTIC_YEARS) + assert run.reforms == [*main.values(), diag["baseline"], diag["free_hours"]] + assert {p for _, p in run.calls} == set(compute.YEARS) + assert {p for p, how in run.hours if how == "drawn"} == set(compute.YEARS) + assert {p for p, how in run.hours if how == "set"} == set(compute.DIAGNOSTIC_YEARS) + + +def test_the_metadata_scan_catches_a_summed_variable_left_out(): + """The mutation the review of #207 ran: drop the two SUMMED-only + variables from METADATA and the contract must fail.""" + path = REPO / "pipeline" / "compute_uk_childcare_100k.py" + compute = _module("compute_uk_childcare_100k_src_mut", path) + pruned = set(compute.METADATA) - {"gov_spending", "gov_tax"} + assert not _calculated(path, compute) <= pruned + + +def test_the_stand_in_run_sees_what_the_scan_cannot(monkeypatch): + """A rebinding no source form shows (here a module constant changed at + run time) reaches the stand-in run, which the contract then refuses.""" + path = REPO / "pipeline" / "compute_uk_childcare_100k.py" + compute = _module("compute_uk_childcare_100k_runtime", path) + scanned = _calculated(path, compute) + monkeypatch.setattr(compute, "HOURS_USAGE", "employment_income") + seen = {v for v, _ in _observed(compute, monkeypatch).calls} + assert "employment_income" in seen + assert seen != scanned + + +_UNREAD = '"employment_income"' +_YEAR_DOC = ( + ' """Weighted sums and counts for one world and year (microdf weights)."""\n' +) +_HEADS_LOOP = " for v in HEADS:\n" +_LIMITS_BODY = " params = sim.tax_benefit_system.parameters\n" +_BUILT = '"cal" + "culate"' +_IN_2029 = " if year == 2029:\n" +# the round-5 review's thirteen survivors: calculate reached through a name +# the scan did not vet, in a helper the old recording run replaced, in a +# year it did not run (2029) or on the drawn-hours branch it never took +_SURVIVORS = { + "import_alias_in_helper": ( + _LIMITS_BODY, + ( + " from builtins import getattr as lookup\n" + f" lookup(sim, {_BUILT})({_UNREAD}, year)\n{_LIMITS_BODY}" + ), + ), + "getattribute_in_helper": ( + _LIMITS_BODY, + f" sim.__getattribute__({_BUILT})({_UNREAD}, year)\n{_LIMITS_BODY}", + ), + "class_body_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC}{_IN_2029} class ReadExtra:\n" + f" extra = sim.__getattribute__({_BUILT})({_UNREAD}, year)\n" + ), + ), + "decorator_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} def decorate(f):\n {_IN_2029}" + f" sim.__getattribute__({_BUILT})({_UNREAD}, year)\n" + " return f\n @decorate\n def unused():\n pass\n" + ), + ), + "dict_of_functions_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC}{_IN_2029}" + f' calls = {{"read": sim.__getattribute__({_BUILT})}}\n' + f' calls["read"]({_UNREAD}, year)\n' + ), + ), + "exec_alias_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} from builtins import exec as run_code\n{_IN_2029}" + ' run_code("sim.cal" + "culate(\'employment_income\', year)")\n' + ), + ), + "getattr_alias_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} from builtins import getattr as lookup\n{_IN_2029}" + f" lookup(sim, {_BUILT})({_UNREAD}, year)\n" + ), + ), + "getattribute_in_2029": ( + _YEAR_DOC, + f"{_YEAR_DOC}{_IN_2029} sim.__getattribute__({_BUILT})({_UNREAD}, year)\n", + ), + "drawn_hours_branch": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} if (sim.calculate(HOURS_USAGE, year).values < 30).any():\n" + f" sim.__getattribute__({_BUILT})({_UNREAD}, year)\n" + ), + ), + "importlib_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} import importlib as loader\n{_IN_2029}" + ' loader.import_module("builtins").getattr(sim, ' + f"{_BUILT})({_UNREAD}, year)\n" + ), + ), + "operator_alias_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} import operator as op\n{_IN_2029}" + f" op.methodcaller({_BUILT}, {_UNREAD}, year)(sim)\n" + ), + ), + "operator_from_alias_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} from operator import methodcaller as invoke\n{_IN_2029}" + f" invoke({_BUILT}, {_UNREAD}, year)(sim)\n" + ), + ), + "partial_in_2029": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} from functools import partial\n{_IN_2029}" + f" partial(sim.__getattribute__({_BUILT}), {_UNREAD})(year)\n" + ), + ), +} +_CALLBACK = f" ab._resolve_parameter(sim, {_BUILT})({_UNREAD}, year)\n" +# the round-6 review's survivor, a callback the old run's path-helper stub +# dropped, and the getter routes to calculate it used: a vetted walker on a +# built, a literal or an aliased path +_ROUND_6 = { + "callback_through_path_helper": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} class ChildcarePath:\n def resolve(self):\n" + f"{_CALLBACK} return ROOT\n" + " fiscal.relative_to_root(ChildcarePath())\n" + ), + ), + "getter_built_path": ( + _YEAR_DOC, + f"{_YEAR_DOC} ab._resolve_parameter(sim, {_BUILT})({_UNREAD}, year)\n", + ), + "getter_literal_path": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} ab._resolve_parameter(sim, " + f'"calculate.__call__")({_UNREAD}, year)\n' + ), + ), + "getter_alias": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} walk = ab._resolve_parameter\n" + f' walk(sim, "calculate.__call__")({_UNREAD}, year)\n' + ), + ), + # the same callback handed to an engine method (a holder lookup hashes + # its name) and to a function the run stubs + "callback_through_engine_method": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} class Name(str):\n def __hash__(self):\n" + f"{_CALLBACK} return 0\n" + ' sim.get_holder(Name("age"))\n' + ), + ), + "callback_through_stub": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} class Dataset:\n def __fspath__(self):\n" + f"{_CALLBACK} return str(ROOT)\n" + " fiscal.verify_runtime_dataset_sha256(" + 'Dataset(), "0" * 64, phase="year")\n' + ), + ), +} +_INPUT = '"employment_income_behavioral_response"' +# the round-7 review's survivor: the getter read by a literal getattr, from +# the __float__ of objects in an array the engine converts; the getter read +# by a string alone; and classes the class rule alone refuses, by a class +# statement and by type(), whose callbacks are direct calls +_ROUND_7 = { + "object_array_callback": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} class ChildcareValue:\n def __float__(self):\n" + f' getattr(ab, "_resolve_parameter")(sim, {_BUILT})' + f"({_UNREAD}, year)\n return 1.0\n" + f" sim.set_input(\n {_INPUT},\n year,\n" + " np.asarray(" + '[ChildcareValue()] * len(sim.calculate("age", year).values)),\n )\n' + ), + ), + "getter_named_by_string": ( + _YEAR_DOC, + ( + f'{_YEAR_DOC} getattr(ab, "_resolve_parameter")(sim, {_BUILT})' + f"({_UNREAD}, year)\n" + ), + ), + "class_calls_back": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} class Reader:\n def __float__(self):\n" + f" return float(sim.calculate({_UNREAD}, year).sum())\n" + f" sim.set_input({_INPUT}, year, np.asarray([Reader()] * 3))\n" + ), + ), + "class_by_type": ( + _YEAR_DOC, + ( + f'{_YEAR_DOC} Reader = type("Reader", (), {{"__float__": lambda _: ' + f"float(sim.calculate({_UNREAD}, year).sum())}})\n" + f" sim.set_input({_INPUT}, year, np.asarray([Reader()] * 3))\n" + ), + ), +} +# anchor -> mutant: each calculates a name the metadata does not list, by a +# route the contract must refuse +_MUTANTS = { + # the round-3 review's two: a rebinding generator, and an assignment + # inside the loop + "rebinding_generator": ( + "for v in SUMMED}", + f"for v in SUMMED for v in ({_UNREAD},)}}", + ), + "assignment_in_loop": (_HEADS_LOOP, f"{_HEADS_LOOP} v = {_UNREAD}\n"), + "shadowed_module_tuple": (_YEAR_DOC, f"{_YEAR_DOC} SUMMED = ({_UNREAD},)\n"), + "module_constant_bound_twice": ( + 'HOURS_USAGE = "maximum_extended_childcare_hours_usage"\n', + ( + 'HOURS_USAGE = "maximum_extended_childcare_hours_usage"\n' + f"HOURS_USAGE = {_UNREAD}\n" + ), + ), + "tuple_target": ("for v in SUMMED}", "for v, _ in zip(SUMMED, SUMMED)}"), + # the round-4 review's eight survivors and its mapping capture + "getattr_call": ( + _YEAR_DOC, + f'{_YEAR_DOC} getattr(sim, "calculate")({_UNREAD}, year)\n', + ), + "alias_call": ( + _YEAR_DOC, + f"{_YEAR_DOC} calculate = sim.calculate\n calculate({_UNREAD}, year)\n", + ), + "walrus_in_comprehension": ( + _YEAR_DOC, + f"{_YEAR_DOC} [(HOURS_USAGE := {_UNREAD}) for _ in (0,)]\n", + ), + "walrus_in_generator": ( + _YEAR_DOC, + f"{_YEAR_DOC} tuple((HOURS_USAGE := {_UNREAD}) for _ in (0,))\n", + ), + "nested_nonlocal": ( + _HEADS_LOOP, + ( + f"{_HEADS_LOOP} def rebind():\n nonlocal v\n" + f" v = {_UNREAD}\n rebind()\n" + ), + ), + "nested_global": ( + _YEAR_DOC, + ( + f"{_YEAR_DOC} def rebind():\n global HOURS_USAGE\n" + f" HOURS_USAGE = {_UNREAD}\n rebind()\n" + ), + ), + "default_walrus": ( + _YEAR_DOC, + f"{_YEAR_DOC} def _unused(arg=(HOURS_USAGE := {_UNREAD})):\n pass\n", + ), + "default_getattr": ( + _YEAR_DOC, + ( + f'{_YEAR_DOC} def _unused(arg=getattr(sim, "calculate")({_UNREAD}, year)):\n' + " pass\n" + ), + ), + "match_mapping_rest": ( + _HEADS_LOOP, + f"{_HEADS_LOOP} match {{}}:\n case {{**v}}:\n pass\n", + ), + # and a module constant rebound through the module object + "module_attribute_write": ( + _YEAR_DOC, + f"{_YEAR_DOC} sys.modules[__name__].HOURS_USAGE = {_UNREAD}\n", + ), + **_SURVIVORS, + **_ROUND_6, + **_ROUND_7, +} +# What the stand-in run alone makes of each mutant (the scan refuses them +# all): it sees an unlisted name, or it does not, or a stub refuses the +# object a mutant passes it, or the mutant crashes before an artifact. +_RUN_ALONE = { + # calculates only listed names + "tuple_target": "listed", + # METADATA is built from the rebound name, so the run lists it + "module_constant_bound_twice": "listed", + # TypeError (an unhashable dict) before the run writes an artifact + "match_mapping_rest": "crash", + "callback_through_engine_method": "refused", + "callback_through_stub": "refused", + # set_input refuses an object array + "object_array_callback": "refused", + "class_calls_back": "refused", + "class_by_type": "refused", +} + + +def _mutant(tmp_path, anchor, mutant) -> Path: + path = REPO / "pipeline" / "compute_uk_childcare_100k.py" + source = path.read_text() + assert source.count(anchor) == 1, anchor + mutated = tmp_path / "compute_uk_childcare_100k.py" + mutated.write_text(source.replace(anchor, mutant)) + return mutated + + +@pytest.mark.parametrize("anchor, mutant", _MUTANTS.values(), ids=list(_MUTANTS)) +def test_the_metadata_contract_refuses_a_hidden_name( + tmp_path, monkeypatch, anchor, mutant +): + """Each mutant calculates a name the metadata does not list; the + contract must refuse it rather than read the name's old binding.""" + mutated = _mutant(tmp_path, anchor, mutant) + with pytest.raises(Unvouched): + _contract(mutated, f"compute_mutant_{tmp_path.name}", monkeypatch) + + +@pytest.mark.parametrize("anchor, mutant", _MUTANTS.values(), ids=list(_MUTANTS)) +def test_the_scan_alone_refuses_each_hidden_name(tmp_path, anchor, mutant): + """The two layers each hold: with no run, the scan refuses every mutant.""" + mutated = _mutant(tmp_path, anchor, mutant) + compute = _module(f"compute_scan_{tmp_path.name}", mutated) + with pytest.raises(Unvouched): + _calculated(mutated, compute) + + +def _alone(compute, monkeypatch) -> str: + try: + seen = {v for v, _ in _observed(compute, monkeypatch).calls} + except Unvouched: + return "refused" + except TypeError: + return "crash" + return "unlisted" if seen - set(compute.METADATA) else "listed" + + +@pytest.mark.parametrize("name", list(_MUTANTS)) +def test_what_the_stand_in_run_alone_makes_of_each_mutant(tmp_path, monkeypatch, name): + """With the scan set aside, the stand-in run of main() reaches the + calculation of every mutant but those in _RUN_ALONE: the helper each + round-5 survivor hides in runs, and so do 2029, the drawn-hours branch + and the path helper the round-6 callback goes through.""" + mutated = _mutant(tmp_path, *_MUTANTS[name]) + compute = _module(f"compute_alone_{tmp_path.name}", mutated) + assert _alone(compute, monkeypatch) == _RUN_ALONE.get(name, "unlisted") + + +def test_a_stub_refuses_any_argument_the_script_could_define(): + """Each stub, and each stand-in method, refuses any argument whose type + the script could have defined; plain values, paths and arrays pass.""" + _stub(None)(Path("x"), {"a": [1, 2.0, None]}, phase="p") + _array(np.zeros(2)) + + class Text(str): + pass + + for value in (Text("x"), SimpleNamespace(), np.int64(1), [object()], {1: Text()}): + with pytest.raises(Unvouched): + _stub(None)(value) + for values in ( + [1.0], + np.asarray([object()]), + np.zeros(2, dtype=[("a", object)]), + ): + with pytest.raises(Unvouched): + _array(values) + + +def test_default_run_ids_differ_across_variants_and_runs(): + compute = _module( + "compute_uk_childcare_100k_run_id", + REPO / "pipeline" / "compute_uk_childcare_100k.py", + ) + t = dt.datetime(2026, 10, 25, 20, 13, 47, tzinfo=dt.timezone.utc) + ids = { + compute.default_run_id(VARIANTS["default_variant"], t), + compute.default_run_id(_wide_certified()["variant_id"], t), + compute.default_run_id( + VARIANTS["default_variant"], t + dt.timedelta(seconds=1) + ), + } + assert len(ids) == 3 + + +# --- the artifact's identity and history (synthetic: no engine, no run) -------- + + +def _record(variant_id=None): + vid = variant_id or VARIANTS["default_variant"] + return copy.deepcopy( + next(v for v in VARIANTS["variants"] if v["variant_id"] == vid) + ) + + +def _synthetic(variant, *, run_id="run", computed_at="2026-10-25T00:00:00+00:00"): + identity = copy.deepcopy(variant["identity"]) + sha = identity["dataset"]["sha256"] + world = { + "dataset_sha256_before": sha, + "dataset_sha256_after": sha, + "runtime_dataset_source": "data/uk/policyengine-local/populace.h5", + "dataset_years": list(variant["data_years"]), + "years": {}, + } + return { + "schema_version": lane.SCHEMA_VERSION, + "run_id": run_id, + "computed_at": computed_at, + "variant_id": variant["variant_id"], + "variant_identity_sha256": variant["identity_sha256"], + "packages": { + name: {"version": pkg["version"], "record_sha256": pkg["record_sha256"]} + for name, pkg in identity["packages"].items() + }, + "engine_version": identity["packages"]["policyengine-uk"]["version"], + "python": identity["python"], + "packaged_bundle_manifest_sha256": identity["packaged_bundle_manifest_sha256"], + "release_bundle": identity["release_bundle"], + "data_bundle": identity["dataset"]["revision"], + "dataset": {"revision": identity["dataset"]["revision"], "sha256": sha}, + "behavioural_response": "none (static)", + "worlds": {"baseline": world}, + "diagnostics": {"baseline__full_use_hours": copy.deepcopy(world)}, + } + + +def _path(art): + return Path(f"{art['variant_id']}.json") + + +def _other_certified(): + return next( + v["variant_id"] + for v in VARIANTS["variants"] + if v["status"] == "certified" and v["variant_id"] != VARIANTS["default_variant"] + ) + + +@pytest.mark.parametrize("variant_id", [None, "other"]) +def test_check_artifact_accepts_exactly_its_variants_pairing(variant_id): + variant = _record(_other_certified() if variant_id else None) + art = _synthetic(variant) + lane.check_artifact(art, _path(art), variant) + + +_ZERO = "0" * 64 +MUTATIONS = { + "schema": lambda a: a.update(schema_version=1), + "wrapper version": lambda a: a["packages"]["policyengine"].update(version="9.9.9"), + "core RECORD": lambda a: a["packages"]["policyengine-core"].update( + record_sha256=_ZERO + ), + "engine": lambda a: a.update(engine_version="0.0.1"), + "python": lambda a: a.update(python="3.99"), + "packaged manifest": lambda a: a.update(packaged_bundle_manifest_sha256=_ZERO), + "release bundle": lambda a: a["release_bundle"].update(bundle_id="uk-other"), + "variant identity": lambda a: a.update(variant_identity_sha256=_ZERO), + "dataset": lambda a: a["dataset"].update(sha256=_ZERO), + "data bundle": lambda a: a.update(data_bundle="another-build"), + "runtime source": lambda a: a["worlds"]["baseline"].pop("runtime_dataset_source"), + "digest after": lambda a: a["diagnostics"]["baseline__full_use_hours"].update( + dataset_sha256_after=_ZERO + ), + "data years": lambda a: a["worlds"]["baseline"]["dataset_years"].pop(), + "behaviour": lambda a: a.update(behavioural_response="dynamic"), +} + + +@pytest.mark.parametrize("what", sorted(MUTATIONS)) +def test_check_artifact_refuses_any_other_pairing(what): + variant = _record() + art = _synthetic(variant) + MUTATIONS[what](art) + with pytest.raises(ValueError): + lane.check_artifact(art, _path(art), variant) + + +def test_an_artifact_must_sit_under_its_own_variant_id(): + variant = _record() + art = _synthetic(variant) + with pytest.raises(ValueError): + lane.check_artifact(art, Path(f"{_other_certified()}.json"), variant) + + +def _write(directory, *arts): + for art in arts: + (directory / _path(art)).write_text(json.dumps(art)) + # the construction notes live beside the artifacts and are not one + shutil.copy(lane.CONSTRUCTION, directory / lane.CONSTRUCTION.name) + + +def test_every_certified_artifact_stays_and_the_newest_is_latest(tmp_path): + older = _synthetic( + _record(_other_certified()), + run_id="older", + computed_at="2026-10-22T00:00:00+00:00", + ) + newer = _synthetic(_record(), run_id="newer") + _write(tmp_path, newer, older) + got = lane.load_artifacts(tmp_path) + assert [a["run_id"] for a, _ in got] == ["older", "newer"] + + +def test_artifacts_must_have_distinct_runs_and_certified_variants(tmp_path): + a = _synthetic(_record(_other_certified()), run_id="same") + b = _synthetic(_record(), run_id="same") + _write(tmp_path, a, b) + with pytest.raises(ValueError, match="run_id"): + lane.load_artifacts(tmp_path) + index = copy.deepcopy(VARIANTS) + for v in index["variants"]: + if v["variant_id"] == a["variant_id"]: + v["status"] = "retired" + (tmp_path / "variants.json").write_text(json.dumps(index)) + b["run_id"] = "other" + _write(tmp_path, b) + with pytest.raises(ValueError, match="no certified variant"): + lane.load_artifacts(tmp_path, tmp_path / "variants.json") + + +@pytest.mark.parametrize( + "run_id, when, match", + [ + ("a", "2026-10-25T00:00:00+00:00", "share a computed_at"), + ("a", "2026-10-25T01:00:00+01:00", "not a UTC time"), + # the same instant as b, and a time before it that sorts after it, + # each in a UTC spelling isoformat() does not write + ("a", "2026-10-25T00:00:00.000000+00:00", "not a UTC time"), + ("a", "2026-10-24T23:59:59Z", "not a UTC time"), + ("platform-grid-a", "2026-10-24T00:00:00+00:00", "reserved prefix"), + # the view's LIKE ignores ASCII case + ("Platform-grid-a", "2026-10-24T00:00:00+00:00", "reserved prefix"), + ("PLATFORM-GRID-a", "2026-10-24T00:00:00+00:00", "reserved prefix"), + ], +) +def test_each_claims_newest_run_is_unambiguous(tmp_path, run_id, when, match): + """Each claim's latest is its newest covering run, here and in the + database, which orders computed_at as text and prefers platform-grid + runs within a bundle; so two runs at one time, a time not in isoformat's + UTC form and a platform-grid run id are refused.""" + a = _synthetic(_record(_other_certified()), run_id=run_id, computed_at=when) + b = _synthetic(_record(), run_id="b") + _write(tmp_path, a, b) + with pytest.raises(ValueError, match=match): + lane.load_artifacts(tmp_path) + + +@settings(deadline=None) +@given( + st.datetimes(timezones=st.just(dt.timezone.utc)), + st.datetimes(timezones=st.just(dt.timezone.utc)), +) +def test_canonical_utc_strings_sort_as_the_times_do(a, b): + """The premise of ordering runs by their computed_at strings: in the + form _utc accepts, string order is time order and equal strings are + equal times.""" + sa, sb = a.isoformat(), b.isoformat() + lane._utc(sa, "a") + lane._utc(sb, "b") + assert (sa < sb) == (a < b) + assert (sa == sb) == (a == b) + + +def test_every_stage_checks_the_run_history(artifact): + """Called directly, results, diagnoses and expected refuse the histories + load_artifacts refuses, so no path reads a tie or a duplicate run id.""" + scores, _, artifacts = lane.stage() + path = artifacts[-1][1] + twin = copy.deepcopy(artifact) + twin["run_id"] = "twin" + same_run = copy.deepcopy(artifact) + same_run["computed_at"] = "2026-10-25T23:00:00+00:00" + spelled = copy.deepcopy(artifact) + spelled["computed_at"] = artifact["computed_at"].replace("+00:00", "Z") + reserved = copy.deepcopy(artifact) + reserved["run_id"] = "platform-grid-childcare" + mixed = copy.deepcopy(artifact) + mixed["run_id"] = "Platform-Grid-childcare" + for history, match in ( + ([(artifact, path), (twin, path)], "share a computed_at"), + ([(artifact, path), (same_run, path)], "share a run_id"), + ([(spelled, path)], "not a UTC time"), + ([(reserved, path)], "reserved prefix"), + ([(mixed, path)], "reserved prefix"), + ): + for stage in (lane.results, lane.diagnoses, lane.expected): + with pytest.raises(ValueError, match=match): + stage(scores, history) + if len(history) == 1: + with pytest.raises(ValueError, match=match): + lane.certified_result(scores[0], lane.load_notes(), *history[0]) + + +@pytest.mark.parametrize( + "run_id", ["platform-grid-childcare", "Platform-grid-childcare", "PLATFORM-GRID-x"] +) +def test_the_compute_script_refuses_a_reserved_run_id(run_id): + compute = _module( + "compute_uk_childcare_100k_reserved", + REPO / "pipeline" / "compute_uk_childcare_100k.py", + ) + assert compute.RESERVED_RUN_PREFIX == lane.RESERVED_RUN_PREFIX + with pytest.raises(SystemExit, match="may not start with"): + compute.main(["--run-id", run_id]) + + +def _view_prefers(conn, run_id: str) -> bool: + """The comparisons view's own platform-grid test, evaluated by SQLite.""" + sql = (REPO / "scorecard_db" / "db.py").read_text() + (test,) = set(re.findall(r"p\.run_id LIKE '[^']*'", sql)) + like = test.removeprefix("p.run_id ") + return bool(conn.execute(f"SELECT ? {like}", (run_id,)).fetchone()[0]) + + +_CASED = st.lists(st.booleans(), min_size=14, max_size=14).map( + lambda up: "".join( + c.upper() if u else c for c, u in zip(lane.RESERVED_RUN_PREFIX, up) + ) +) + + +@settings(deadline=None) +@given( + st.one_of( + st.text(), + st.tuples(_CASED, st.text()).map("".join), + st.tuples(st.text(max_size=2), _CASED, st.text()).map("".join), + ) +) +@example("Platform-grid-childcare") +@example("platform-grİd-x") # dotted capital I: SQLite folds ASCII only +@example("platform-grid") +def test_reserved_run_ids_are_those_the_database_view_prefers(run_id): + """Differential: ingest's and the compute script's reserved test agree + with SQLite evaluating the comparisons view's LIKE on every run id.""" + compute = _module( + "compute_uk_childcare_100k_prefix", + REPO / "pipeline" / "compute_uk_childcare_100k.py", + ) + conn = sqlite3.connect(":memory:") + try: + want = _view_prefers(conn, run_id) + finally: + conn.close() + assert lane.reserved_run_id(run_id) == want + assert compute.reserved_run_id(run_id) == want + + +def test_the_published_run_may_not_take_the_reserved_prefix(tmp_path, monkeypatch): + pe = tmp_path / "pe" + pe.mkdir() + pub = json.loads((lane.CENTAX_DIR / "pe" / "published_dashboard.json").read_text()) + lane._published() + pub["run_id"] = "Platform-grid-dashboard" + (pe / "published_dashboard.json").write_text(json.dumps(pub)) + monkeypatch.setattr(lane, "CENTAX_DIR", tmp_path) + with pytest.raises(ValueError, match="reserved prefix"): + lane._published() + + +def test_results_keep_an_earlier_certified_run_beside_the_latest(artifact): + """History across variants: an earlier certified run attaches to every + claim the latest does, and the published figure is older than both.""" + scores, _, artifacts = lane.stage() + latest, path = artifacts[-1] + earlier = copy.deepcopy(latest) + earlier["run_id"] = "earlier-certified-run" + earlier["computed_at"] = "2026-10-20T00:00:00+00:00" + history = [(earlier, path), (latest, path)] + out = lane.results(scores, history) + runs = Counter(r.run_id for r in out) + assert runs["earlier-certified-run"] == runs[latest["run_id"]] == 22 + # ingest's tally scales with the runs, so a second artifact builds + drops = lane.stage()[1] + diag = lane.diagnoses(scores, history) + want = lane.expected(scores, history) + assert lane.accounting(scores, out, diag, drops, history) == want + assert want["results"] == 2 * 22 + 4 + one = lane.results(scores, artifacts) + assert lane.accounting( + scores, one, lane.diagnoses(scores, artifacts), drops, artifacts + ) == lane.expected(scores, artifacts) + # the published figure must stay older than every certified run + earlier["computed_at"] = "2026-10-01T00:00:00+00:00" + with pytest.raises(ValueError, match="older than every certified run"): + lane.results(scores, [(earlier, path), (latest, path)]) + + +def _wide_certified(): + """A registered certified variant other than the default whose data + years reach 2031, the engine year of CBP's last claims (2031-32).""" + return next( + v + for v in VARIANTS["variants"] + if v["status"] == "certified" + and v["variant_id"] != VARIANTS["default_variant"] + and 2031 in v["data_years"] + ) + + +def _on_pair(art, variant, *, run_id, with_2031=False): + """The committed run's numbers relabelled as a run on `variant` (its + identity from the record, as _synthetic builds it), to exercise ingest's + control flow only: they claim nothing about that engine's costs.""" + out = copy.deepcopy(art) + label = _synthetic(variant, run_id=run_id, computed_at="2026-10-26T00:00:00+00:00") + out.update({k: v for k, v in label.items() if k not in ("worlds", "diagnostics")}) + sha = variant["identity"]["dataset"]["sha256"] + for group in ("worlds", "diagnostics"): + for world in out[group].values(): + world.update( + dataset_sha256_before=sha, + dataset_sha256_after=sha, + dataset_years=list(variant["data_years"]), + ) + if with_2031 and group == "worlds": + world["years"]["2031"] = copy.deepcopy(world["years"]["2030"]) + return out + + +def _in_repo(artifacts): + # a result's construction names its artifact's path in the repo + return [(a, lane.RESULTS_DIR / p.name) for a, p in artifacts] + + +def _verified_for(variant_id, *, except_ids=()): + """The lane's notes with `variant_id` added as a checked pair to every + pair-scoped note but `except_ids`, and a pair entry stating no diagnosis.""" + notes = copy.deepcopy(lane.load_notes()) + for kind in (lane.CENTAX, lane.CBP, "construction"): + for a in notes[kind]: + if "variants" in a and a["id"] not in except_ids: + a["variants"].append(variant_id) + notes["pairs"][variant_id] = { + "static_because": "STATED FOR THE TEST", + "hours_draw": None, + "diagnosis": None, + } + return notes + + +def test_an_artifact_on_an_unverified_pair_is_refused(artifact, tmp_path): + """The registered 2.130.1 pairs differ from 2.89.2 in mechanisms these + notes state (the review of #207 read the pinned wheel), so ingest + refuses a pair no note was checked on rather than give it 2.89.2's.""" + other = _on_pair(artifact, _wide_certified(), run_id="other") + _write(tmp_path, artifact, other) + with pytest.raises(ValueError, match="no note in this lane was verified"): + lane.stage(results_dir=tmp_path) + scores, _, artifacts = lane.stage() + with pytest.raises(ValueError, match="no note in this lane was verified"): + lane.results(scores, [*artifacts, (other, lane.RESULTS_DIR / _path(other))]) + with pytest.raises(ValueError, match="no note in this lane was verified"): + lane.diagnoses(scores, [*artifacts, (other, lane.RESULTS_DIR / _path(other))]) + with pytest.raises(ValueError, match="no note in this lane was verified"): + lane.expected(scores, [*artifacts, (other, lane.RESULTS_DIR / _path(other))]) + # refused even when the unverified run computes no claim's year + idle = copy.deepcopy(other) + for world in idle["worlds"].values(): + world["years"] = {} + with pytest.raises(ValueError, match="no note in this lane was verified"): + lane.results(scores, [*artifacts, (idle, lane.RESULTS_DIR / _path(idle))]) + + +def test_notes_reach_only_the_pairs_they_were_checked_on(artifact, tmp_path): + vid = _wide_certified()["variant_id"] + other = _on_pair(artifact, _wide_certified(), run_id="other") + notes = _verified_for(vid, except_ids={"pe-universal-switch-off"}) + _write(tmp_path, artifact, other) + scores, _, artifacts = lane.stage(notes, results_dir=tmp_path) + out = lane.results(scores, _in_repo(artifacts), notes) + runs = { + run: [r for r in out if r.run_id == run] + for run in (artifact["run_id"], "other") + } + switch_off = next( + a["text"] for a in notes["construction"] if a["id"] == "pe-universal-switch-off" + ) + draw = notes["pairs"][VARIANTS["default_variant"]]["hours_draw"] + assert len(runs["other"]) == len(runs[artifact["run_id"]]) == 22 + for r in runs["other"]: + assert switch_off not in r.annotations + assert not any(draw in t for t in r.annotations) + for r in runs[artifact["run_id"]]: + assert switch_off in r.annotations + assert any(draw in t for t in r.annotations) + # the newest pair states no diagnosis, so none is drawn from the old one + assert lane.diagnoses(scores, _in_repo(artifacts), notes) == [] + # and the behavioural legs say why from the newest pair + legs = [s for s in scores if s.conditions.get("component") == "intensive_margin"] + assert legs and all( + "STATED FOR THE TEST" in s.publication["pe_missing"] for s in legs + ) + + +def test_tally_and_pe_missing_follow_each_runs_computed_years(artifact, tmp_path): + """Coverage, not dataset support: a run that holds 2031 among its data + years but computes only the default years leaves CBP's 2031-32 claims + without a value and says so; one that computes 2031 answers them. The + tally follows, and no claim id moves across the three histories.""" + wide = _wide_certified() + notes = _verified_for(wide["variant_id"]) + histories = {"alone": lane.stage(notes)} + for name, with_2031 in (("default_years", False), ("with_2031", True)): + (tmp_path / name).mkdir() + other = _on_pair(artifact, wide, run_id="other", with_2031=with_2031) + _write(tmp_path / name, artifact, other) + histories[name] = lane.stage(notes, results_dir=tmp_path / name) + ids = {k: [s.claim_id() for s in h[0]] for k, h in histories.items()} + assert ids["alone"] == ids["default_years"] == ids["with_2031"] + + def last(name): + return [ + s for s in histories[name][0] if s.source == lane.CBP and s.period == 2032 + ] + + assert len(last("alone")) == 2 + assert all( + "outside the data years" in s.publication["pe_missing"] for s in last("alone") + ) + for s in last("default_years"): + assert ( + "no certified run in this lane computed it" in s.publication["pe_missing"] + ) + assert wide["variant_id"] in s.publication["pe_missing"] + assert all("pe_missing" not in s.publication for s in last("with_2031")) + for name, certified in (("alone", 22), ("default_years", 44), ("with_2031", 46)): + scores, drops, arts = histories[name] + out = lane.results(scores, _in_repo(arts), notes) + diag = lane.diagnoses(scores, _in_repo(arts), notes) + got = lane.accounting(scores, out, diag, drops, arts) + assert got == lane.expected(scores, arts, notes) + assert got["results_by_kind"] == { + "certified": certified, + "published_dashboard": 4, + } + + +def test_a_newer_run_that_skips_a_year_leaves_that_years_diagnosis(artifact, tmp_path): + """The round-3 review's case: a newer run on a verified pair that + computes only 2031 takes no claim's latest result in the years it skips, + so the diagnoses of those years stay with the run that computed them.""" + wide = _wide_certified() + notes = _verified_for(wide["variant_id"]) + other = _on_pair(artifact, wide, run_id="other", with_2031=True) + for world in other["worlds"].values(): + world["years"] = {y: b for y, b in world["years"].items() if y == "2031"} + _write(tmp_path, artifact, other) + scores, drops, arts = lane.stage(notes, results_dir=tmp_path) + arts = _in_repo(arts) + out = lane.results(scores, arts, notes) + diag = lane.diagnoses(scores, arts, notes) + default = notes["pairs"][VARIANTS["default_variant"]]["diagnosis"] + assert len(diag) == 2 and all(default in str(row) for row in diag) + got = lane.accounting(scores, out, diag, drops, arts) + assert got == lane.expected(scores, arts, notes) + assert got["results_by_kind"] == {"certified": 24, "published_dashboard": 4} + assert got["diagnoses"] == 2 + + +_BASE_FULL = "baseline__full_use_hours" +_REFORM_FULL = "free_hours__full_use_hours" +_LIMIT_30H = "gov.dfe.extended_childcare_entitlement.income.limit" + + +def _override(a, world): + return a["diagnostics"][world]["input_override"] + + +def _block(a, world, year="2027"): + return a["diagnostics"][world]["years"][year] + + +def _periods(period, value=None): + """Set the reform world's reform to one period (at the artifact's own + sentinel unless a value is given).""" + + def change(a): + d = a["diagnostics"][_REFORM_FULL] + d["reform_dict"] = { + k: {period: a["limit_sentinel"] if value is None else value} + for k in d["reform_dict"] + } + + return change + + +def _both(change): + def apply(a): + for world in (_BASE_FULL, _REFORM_FULL): + change(a["diagnostics"][world]) + + return apply + + +def _false_sentinel(a): + a["limit_sentinel"] = 100_000 + _periods("2026-01-01.2035-12-31", 100_000)(a) + + +def _receiving_none(a): + _block(a, _BASE_FULL)["hours_usage"]["mean_benunits_receiving_extended"] = None + + +# what the record must show before any text claims full use, each broken +_NOT_FULL_USE = { + "value": lambda a: _override(a, _REFORM_FULL).update(value=17.0), + "variable": lambda a: _override(a, _BASE_FULL).update( + variable="childcare_expenses" + ), + "year not set": lambda a: _override(a, _REFORM_FULL).update(set_for_years=[2029]), + "set after a dependent": lambda a: _override(a, _BASE_FULL)[ + "dependents_calculated_before_override" + ].update(extended_childcare_entitlement=["2027"]), + "read back": lambda a: _override(a, _REFORM_FULL)["post_set_weighted_mean"].update( + {"2027": 17.0} + ), + "mean": lambda a: _block(a, _BASE_FULL)["hours_usage"].update( + mean_benunits_receiving_extended=17.0 + ), + "no override": lambda a: a["diagnostics"][_REFORM_FULL].pop("input_override"), + "one world": lambda a: a["diagnostics"][_BASE_FULL]["years"].pop("2027"), + "other reform": lambda a: a["diagnostics"][_REFORM_FULL].update(reform_dict={}), + "reform not in force": _periods("2028-01-01.2035-12-31"), + # the round-5 review's: holders, periods, sentinel and executed limits + "an unrelated holder checked": _both( + lambda d: d["input_override"].update( + dependents_calculated_before_override={"unrelated": []} + ) + ), + "a reading holder unchecked": _both( + lambda d: d["input_override"]["dependents_calculated_before_override"].pop( + "extended_childcare_entitlement" + ) + ), + "no reform periods": lambda a: a["diagnostics"][_REFORM_FULL].update( + reform_dict={k: {} for k in a["diagnostics"][_REFORM_FULL]["reform_dict"]} + ), + "reform starts after 30 April": _periods("2027-05-01.2035-12-31"), + "reform stops before 30 April": _periods("2026-01-01.2027-04-29"), + "invalid reform dates": _periods("2027-99-99.2035-99-99"), + "reversed reform dates": _periods("2027-12-31.2027-01-01"), + "overlapping periods disagree": lambda a: a["diagnostics"][_REFORM_FULL].update( + reform_dict={ + k: {"2026-01-01.2035-12-31": 1e100, "2027-01-01.2027-12-31": 200_000} + for k in a["diagnostics"][_REFORM_FULL]["reform_dict"] + } + ), + "a false sentinel": _false_sentinel, + "another sentinel named": lambda a: a.update(limit_sentinel=1e99), + "an executed limit the main world did not run": lambda a: _block(a, _REFORM_FULL)[ + "limits_executed" + ].update({_LIMIT_30H: 100_000}), + "no executed limits": _both(lambda d: d["years"]["2027"].pop("limits_executed")), + "no receiving mean with recipients": _receiving_none, +} + + +def test_the_full_use_cost_rests_on_the_recorded_override(artifact): + """The committed run's record shows full use in both diagnostic years; + the cost is the reform world's free hours minus the baseline's.""" + for year in (2027, 2029): + d = artifact["diagnostics"] + want = sum( + d[_REFORM_FULL]["years"][str(year)]["sums_gbp"][v] + - d[_BASE_FULL]["years"][str(year)]["sums_gbp"][v] + for v in lane.FREE_HOURS + ) + assert lane.full_use_cost(artifact, year) == want > 0 + assert lane.full_use_cost(artifact, 2026) is None + + +@pytest.mark.parametrize("what", sorted(_NOT_FULL_USE)) +def test_no_text_claims_full_use_the_record_does_not_show(artifact, what): + """A note or diagnosis says every family is at the full 30 hours only + when both full-use worlds record the override, set before any holder + that reads it and read back at 30, on the main world's reform in force + on 30 April of the year, with the main world's executed limits.""" + bad = copy.deepcopy(artifact) + _NOT_FULL_USE[what](bad) + with pytest.raises(ValueError, match="full use|beside the other"): + lane.full_use_cost(bad, 2027) + with pytest.raises(ValueError, match="full use|beside the other"): + lane.measured_notes(bad, lane.CENTAX, 2027, "fy", {}) + scores, _, artifacts = lane.stage() + with pytest.raises(ValueError, match="full use|beside the other"): + lane.diagnoses(scores, [(bad, artifacts[-1][1])]) + + +def _zero_recipients(a): + for world in (_BASE_FULL, _REFORM_FULL): + block = _block(a, world) + block["hours_usage"]["mean_benunits_receiving_extended"] = None + block["counts"]["benunits_receiving_extended"] = 0 + block["records"]["benunits_receiving_extended"] = 0 + + +def _per_year_windows(a): + from pipeline import compute_uk_ab2025 as ab + + d = a["diagnostics"][_REFORM_FULL] + d["reform_dict"], _ = ab.reform_dict( + {k: {"2027": None, "2029": None} for k in d["reform_dict"]}, years=[2027, 2029] + ) + + +# records the compute script can write that show full use, each kept +_TRUE_FULL_USE = { + "per-year windows from reform_dict": _per_year_windows, + "a one-day window on 30 April": _periods("2027-04-30.2027-04-30"), + "overlapping periods that agree": lambda a: a["diagnostics"][_REFORM_FULL].update( + reform_dict={ + k: {"2026-01-01.2035-12-31": 1e100, "2027-01-01.2027-12-31": 1e100} + for k in a["diagnostics"][_REFORM_FULL]["reform_dict"] + } + ), + "no benefit unit receiving": _zero_recipients, +} + + +@pytest.mark.parametrize("what", sorted(_TRUE_FULL_USE)) +def test_the_full_use_cost_keeps_every_true_record_shape(artifact, what): + """The check refuses only what the record does not show: a reform + window each year, a window that covers just the parameter date, + overlapping periods that set one value, and no benefit unit receiving + the offer (year_block then writes no receiving mean) all show full use.""" + good = copy.deepcopy(artifact) + _TRUE_FULL_USE[what](good) + assert lane.full_use_cost(good, 2027) == lane.full_use_cost(artifact, 2027) + + +def _current_law(a, year=2027): + return a["worlds"]["baseline"]["years"][str(year)] + + +def test_no_family_receiving_under_current_law_is_stated_not_formatted(artifact): + """The round-6 review's robustness case. year_block itself, on a + stand-in where no benefit unit gets the extended hours, writes no + receiving mean and no receiving records; ingest then says so in the + note and the diagnosis instead of formatting a mean it does not have.""" + compute = _module( + "compute_uk_childcare_100k_no_recipients", + REPO / "pipeline" / "compute_uk_childcare_100k.py", + ) + run = SimpleNamespace(calls=[], hours=[], reforms=[]) + sim = _StandIn(run, None, [2027], Path("stand-in.h5")) + for v in ( + "extended_childcare_entitlement", + "is_child_receiving_extended_childcare", + ): + sim.set_input(v, 2027, np.zeros(len(_StandIn.DRAWN))) + made = compute.year_block(sim, 2027) + assert made["hours_usage"]["mean_benunits_receiving_extended"] is None + assert made["records"]["benunits_receiving_extended"] == 0 + assert made["counts"]["benunits_receiving_extended"] == 0 + none = copy.deepcopy(artifact) + for key in ("hours_usage", "counts", "records"): + _current_law(none)[key] = made[key] + assert lane.receiving_hours(none, 2027) is None + stated = "no family receives the extended hours under current law in 2027-28" + notes = lane.measured_notes(none, lane.CENTAX, 2027, "2027-28", {}) + assert sum(stated in n for n in notes) == 1 + assert not any("weekly hours under current law" in n for n in notes) + scores, _, artifacts = lane.stage() + rows = lane.diagnoses(scores, [(none, artifacts[-1][1])]) + assert sum("no family receives the extended hours" in str(r) for r in rows) == 1 + assert sum("weighted mean of" in str(r) for r in rows) == len(rows) - 1 + + +class _UnweightedReceivers(_StandIn): + """A stand-in whose benefit units given the extended hours all carry + zero weight, while one that is not given them carries weight.""" + + def calculate(self, variable, period): + series = super().calculate(variable, period) + if variable == "extended_childcare_entitlement": + series.weights = np.where(series.values > 0, 0.0, 1.0) + return series + + +def test_receiving_records_with_no_receiving_weight_are_stated_not_nan(artifact): + """The round-7 review's producer boundary. Records that receive the + offer but weigh nothing in total give year_block no mean (None, not + 0/0 = nan), and ingest, keyed on the weighted count, says no family + receives the offer.""" + compute = _module( + "compute_uk_childcare_100k_unweighted_receivers", + REPO / "pipeline" / "compute_uk_childcare_100k.py", + ) + sim = _UnweightedReceivers( + SimpleNamespace(calls=[], hours=[]), None, [2027], Path("stand-in.h5") + ) + made = compute.year_block(sim, 2027) + assert made["records"]["benunits_receiving_extended"] == 2 + assert made["counts"]["benunits_receiving_extended"] == 0 + assert made["hours_usage"]["mean_benunits_receiving_extended"] is None + none = copy.deepcopy(artifact) + for key in ("hours_usage", "counts", "records"): + _current_law(none)[key] = made[key] + assert lane.receiving_hours(none, 2027) is None + notes = lane.measured_notes(none, lane.CENTAX, 2027, "2027-28", {}) + stated = "no family receives the extended hours under current law in 2027-28" + assert sum(stated in n for n in notes) == 1 + assert not any("nan" in n for n in notes) + + +@pytest.mark.parametrize( + "mean, weight", + [ + (None, "kept"), + (15.3, 0), + (float("nan"), "kept"), + (float("inf"), "kept"), + (15.3, float("nan")), + (None, float("nan")), + (15.3, "missing"), + (None, "missing"), + (15.3, -1.0), + ("15.3", "kept"), + (True, "kept"), + (15.3, True), + ], + ids=str, +) +def test_a_receiving_mean_must_agree_with_the_receiving_weight(artifact, mean, weight): + """year_block writes no mean exactly when no weight receives the offer, + so ingest refuses a mean without receiving weight, receiving weight + without a mean, and any mean or weight that is not a finite number.""" + bad = copy.deepcopy(artifact) + block = _current_law(bad) + block["hours_usage"]["mean_benunits_receiving_extended"] = mean + if weight == "missing": + del block["counts"]["benunits_receiving_extended"] + elif weight != "kept": + block["counts"]["benunits_receiving_extended"] = weight + with pytest.raises(ValueError, match="receiving-hours mean"): + lane.receiving_hours(bad, 2027) + with pytest.raises(ValueError, match="receiving-hours mean"): + lane.measured_notes(bad, lane.CENTAX, 2027, "fy", {}) + + +def test_the_full_use_check_reads_the_compute_scripts_own_constants( + artifact, monkeypatch +): + """Differential: the sentinel, limits, hours and the holders checked + before the override are the compute script's, and the holders are read + off a stand-in run of set_full_use itself.""" + from pipeline import compute_uk_ab2025 as ab + + compute = _module( + "compute_uk_childcare_100k_constants", + REPO / "pipeline" / "compute_uk_childcare_100k.py", + ) + assert lane.NO_LIMIT == ab.NO_LIMIT == artifact["limit_sentinel"] + assert lane.LIMITS == compute.LIMITS + assert (lane.HOURS_USAGE_INPUT, lane.FULL_USE_HOURS) == ( + compute.HOURS_USAGE, + compute.FULL_USE_HOURS, + ) + sim = _StandIn(SimpleNamespace(calls=[], hours=[]), None, [2027], "x") + checked = compute.set_full_use(sim, [2027])["dependents_calculated_before_override"] + assert sorted(checked) == sorted(lane.FULL_USE_DEPENDENTS) + + +def _ingest_isolated(tmp_path, monkeypatch, notes, staged) -> sqlite3.Connection: + """The lane's real ingest into an empty database, with this history and + these notes, and its lane feed written beside it, not into data/.""" + from scorecard_db import ingest_harvest + + scores, drops, artifacts = staged + monkeypatch.setattr(lane, "stage", lambda n=None: (scores, drops, artifacts)) + monkeypatch.setattr(lane, "load_notes", lambda: notes) + feed = tmp_path / "lanes.json" + shutil.copy(REPO / "data" / "lanes.json", feed) + sync = ingest_harvest.sync_lane_feed + monkeypatch.setattr( + ingest_harvest, + "sync_lane_feed", + lambda db, path, updated, **kw: sync(db, feed, updated, **kw), + ) + db = tmp_path / "lane.db" + lane.ingest(db) + conn = sqlite3.connect(db) + conn.row_factory = sqlite3.Row + return conn + + +def test_the_database_shows_each_claims_newest_covering_run( + artifact, tmp_path, monkeypatch +): + """Differential: on a history of three runs, two of them on pairs that + share a data bundle and the newest computing only 2027, the comparisons + view's latest result for each claim is the run _covering picks, and + the diagnosis on it is that run's pair's.""" + default = VARIANTS["default_variant"] + shared = Counter( + v["identity"]["dataset"]["revision"] + for v in VARIANTS["variants"] + if v["status"] == "certified" and v["variant_id"] != default + ) + bundle = next(b for b, n in shared.items() if n >= 2) + v1, v2 = [ + v + for v in VARIANTS["variants"] + if v["status"] == "certified" + and v["variant_id"] != default + and v["identity"]["dataset"]["revision"] == bundle + ][:2] + notes = _verified_for(v1["variant_id"]) + for kind in (lane.CENTAX, lane.CBP, "construction"): + for a in notes[kind]: + if "variants" in a: + a["variants"].append(v2["variant_id"]) + notes["pairs"][v1["variant_id"]]["diagnosis"] = "FIRST SHARED-BUNDLE PAIR" + notes["pairs"][v2["variant_id"]] = { + **notes["pairs"][v1["variant_id"]], + "diagnosis": "SECOND SHARED-BUNDLE PAIR", + } + older = _on_pair(artifact, v1, run_id="shared-older") + newer = _on_pair(artifact, v2, run_id="shared-newer") + newer["computed_at"] = "2026-10-27T00:00:00+00:00" + for world in newer["worlds"].values(): + world["years"] = {"2027": world["years"]["2027"]} + (tmp_path / "in").mkdir() + _write(tmp_path / "in", artifact, older, newer) + scores, drops, arts = lane.stage(notes, results_dir=tmp_path / "in") + arts = _in_repo(arts) + conn = _ingest_isolated(tmp_path, monkeypatch, notes, (scores, drops, arts)) + shown = { + row["claim_id"]: row + for row in conn.execute( + "SELECT c.claim_id, r.run_id, c.diagnosis_rationale FROM comparisons c " + "JOIN pe_results r ON r.claim_id = c.claim_id " + "AND r.computed_at = c.computed_at" + ) + } + picked = Counter() + for s in scores: + run = lane._covering(s, arts) + attached = not ( + s.source == lane.CENTAX + and s.conditions["component"] not in lane.CENTAX_ATTACHED + ) + if run is None or not attached: + assert s.claim_id() not in shown or shown[s.claim_id()]["run_id"] not in { + a["run_id"] for a, _ in arts + } + continue + row = shown[s.claim_id()] + assert row["run_id"] == run["run_id"], s.claim_id() + picked[run["run_id"]] += 1 + pair = notes["pairs"][run["variant_id"]] + if lane._diagnosed(s, run, pair): + assert pair["diagnosis"] in row["diagnosis_rationale"] + else: + assert row["diagnosis_rationale"] is None + # the newest run is latest only in the year it computed + assert set(picked) == {"shared-older", "shared-newer"} + diagnosed = [ + r["diagnosis_rationale"] for r in shown.values() if r["diagnosis_rationale"] + ] + assert len(diagnosed) == 2 + for marker in ("FIRST SHARED-BUNDLE PAIR", "SECOND SHARED-BUNDLE PAIR"): + assert sum(marker in t for t in diagnosed) == 1 + + +def test_a_note_on_the_engine_or_a_run_names_its_pairs(tmp_path): + spec = json.loads(lane.CONSTRUCTION.read_text()) + spec["annotations"][0].pop("variants") + path = tmp_path / "construction_annotations.json" + path.write_text(json.dumps(spec)) + with pytest.raises(ValueError, match="variants"): + lane._spec(path) + for kind in (lane.CENTAX, lane.CBP, "construction"): + for a in lane.load_notes()[kind]: + if "variants" in a: + assert a["variants"] == [VARIANTS["default_variant"]], a["id"] + + +@settings(deadline=None) +@given(value=st.integers(-(10**10), 10**10), printed=st.text(min_size=1, max_size=12)) +def test_claim_ids_ignore_the_value_its_printed_form_and_pe_coverage(value, printed): + """A corrected extraction (value or formatting) keeps every claim id, and + so does any reason PolicyEngine has no value (it depends on the runs), so + links and history stay attached.""" + centax = json.loads( + (REPO / "data/externals/centax-childcare-100k.json").read_text() + ) + cbp = json.loads((REPO / "data/externals/cbp-rates-and-wrongs.json").read_text()) + edit = {"value": value, "value_verbatim": printed} + for build, rows in ( + (lane._centax_claim, [r for r in centax if r["metric"] == "revenue_change"]), + (lane._cbp_claim, [r for r in cbp if r["policy"] == "abolish_cliff"]), + ): + row = rows[0] + assert build({**row, **edit}, printed).claim_id() == build(row).claim_id() + + +def test_adapter_rows_carry_the_tidy_core(): + for adapter_id, source in ( + ("centax-childcare-100k", lane.CENTAX), + ("cbp-rates-and-wrongs", lane.CBP), + ): + rows = json.loads((REPO / f"data/externals/{adapter_id}.json").read_text()) + for r in rows: + assert lane._CORE_FIELDS <= set(r) + assert (r["geography"], r["program"], r["subgroup"]) == ( + "England", + lane.PROGRAMS[source], + None, + ) + + +# --- claims, statuses, annotations ------------------------------------------- + + +def test_claims_drops_relationships_and_conditions(): + scores, drops, _ = lane.stage() + assert Counter(s.source for s in scores) == {lane.CENTAX: 24, lane.CBP: 12} + assert drops == {"cbp_taper_rows": 60, "centax_bunching_counts": 12} + for s in scores: + assert s.calibration_relationship == CR.HELD_OUT + assert s.reform.framework == "policy_ref" + assert set(s.conditions) <= STANDARD_CONDITIONS + assert s.publication["publish_without_result"] is True + assert s.conditions["measure_key"] == lane.MEASURES[s.source] + # the printed value is provenance, outside the claim id + assert "value_verbatim" not in s.conditions + assert s.publication["value_verbatim"] + # why a claim has no value depends on the runs, so it is outside the id + assert "pe_missing" not in s.conditions + assert uk_relationship(lane.CBP, None)[0] == CR.HELD_OUT + assert len({s.claim_id() for s in scores}) == len(scores) + + +def _all_facets(source): + if source == lane.CENTAX: + return [ + {"component": c, "scenario": sc, "period": p} + for c, sc, p in product( + lane.CENTAX_ATTACHED, ("central", "high_cost", "low_cost"), (2028, 2030) + ) + ] + return [ + {"model": m, "period": p} for m, p in product(CBP_ABOLISH, range(2027, 2032)) + ] + + +def test_every_attached_result_has_exactly_one_stated_status(): + """Exhaustive over the attached facets: no result's status is a + default, and the statuses are the ones the lane promises.""" + scores, _, _ = lane.stage() + specs = { + lane.CENTAX: lane._spec(CENTAX_SRC / "annotations.json"), + lane.CBP: lane._spec(CBP_SRC / "annotations.json"), + } + seen = Counter() + for s in scores: + if ( + s.source == lane.CENTAX + and s.conditions["component"] not in lane.CENTAX_ATTACHED + ): + continue + kinds = ["certified"] + if s.source == lane.CENTAX and s.conditions["scenario"] == "central": + kinds.append("published_dashboard") + for kind in kinds: + variant = VARIANTS["default_variant"] if kind == "certified" else None + _, status = lane.match_annotations(s, kind, specs[s.source], variant) + key = (s.source, s.conditions.get("component"), kind) + seen[(key, status.value)] += 1 + assert seen == { + ((lane.CENTAX, "static_cost", "certified"), "comparable"): 6, + ((lane.CENTAX, "net_fiscal_cost", "certified"), "concept_mismatch"): 6, + ((lane.CENTAX, "static_cost", "published_dashboard"), "constructed"): 2, + ( + (lane.CENTAX, "net_fiscal_cost", "published_dashboard"), + "concept_mismatch", + ): 2, + ((lane.CBP, None, "certified"), "concept_mismatch"): 12, + } + + +def test_every_annotation_has_a_basis_and_cites_files_that_exist(): + for path in ( + CENTAX_SRC / "annotations.json", + CBP_SRC / "annotations.json", + REPO / "results/uk/childcare_100k/construction_annotations.json", + ): + for a in lane._spec(path): + assert a["basis"] + for token in a["basis"].replace(";", " ").split(): + if token.startswith(("raw/", "pe/")): + assert (path.parent / token.rstrip(",")).exists(), (a["id"], token) + if token.startswith("sources/"): + assert (REPO / token.rstrip(",")).exists(), (a["id"], token) + + +# --- the built database -------------------------------------------------------- + + +def _rows(sql, *args): + conn = sqlite3.connect(DB) + conn.row_factory = sqlite3.Row + out = conn.execute(sql, args).fetchall() + conn.close() + return out + + +def test_built_db_lane_counts_and_latest_result_is_the_certified_run(): + rows = _rows( + "SELECT * FROM comparisons WHERE json_extract(publication, '$.registry') = ?", + lane.REGISTRY_MARK, + ) + assert len(rows) == 36 + computed = [r for r in rows if r["pe_value"] is not None] + assert len(computed) == 22 + assert {r["data_bundle"] for r in computed} == {CERTIFIED["revision"]} + assert {r["engine_version"] for r in computed} == {"2.89.2"} + lanes = _rows("SELECT * FROM lanes WHERE lane = ?", lane.LANE_ID) + assert [r["stage"] for r in lanes] == ["computed"] + # the note counts claims on the latest run, not results across the history + assert lanes[0]["detail"].startswith("22 of 36 claims computed") + + +def test_published_dashboard_miss_stays_in_the_history(): + rows = _rows( + "SELECT s.period, json_extract(s.conditions, '$.component') AS comp," + " r.computed_value, r.status, r.engine_version, r.data_bundle, r.computed_at" + " FROM pe_results r JOIN external_scores s USING (claim_id)" + " WHERE s.source = 'centax' AND r.run_id = 'uk-childcare-100k-dashboard-1b57118'" + " ORDER BY s.period, comp" + ) + got = {(r["period"], r["comp"]): (r["computed_value"], r["status"]) for r in rows} + assert got == { + (2028, "net_fiscal_cost"): (-434_000_000.0, "concept_mismatch"), + (2028, "static_cost"): (-434_000_000.0, "constructed"), + (2030, "net_fiscal_cost"): (-505_000_000.0, "concept_mismatch"), + (2030, "static_cost"): (-505_000_000.0, "constructed"), + } + assert {r["engine_version"] for r in rows} == {"2.102.3"} + assert {r["data_bundle"] for r in rows} == { + "microcosm-uk-2024-25-national-20261002T230158Z-5c6b3f68" + } + # history, not the latest: the lane's certified run is newer + certified = _rows( + "SELECT MIN(computed_at) AS t FROM pe_results WHERE run_id = ?", + lane.load_artifact()[0]["run_id"], + )[0]["t"] + assert max(r["computed_at"] for r in rows) < certified + + +def test_the_published_row_says_when_it_was_live_and_what_followed(): + """policyengine.org served the figure from 12 to 21 October 2026; what + the dashboard has served since is stated on the row, never a value.""" + pub = lane._published() + assert (pub["live"]["from"], pub["live"]["until"]) == ("2026-10-12", "2026-10-21") + assert pub["after"]["attached"] is False + scores, _, artifacts = lane.stage() + published = [ + r for r in lane.results(scores, artifacts) if r.run_id == pub["run_id"] + ] + assert len(published) == 4 + for r in published: + year = str(next(s for s in scores if s.claim_id() == r.claim_id).period - 1) + after = f"£{pub['after']['thirty_hours_gbp_bn'][year]:.3f}bn in this year" + assert any(after in t for t in r.annotations) + assert any("from 2026-10-12 to 2026-10-21" in t for t in r.annotations) + after_m = sorted( + {-round(v * 1e3) for v in pub["after"]["thirty_hours_gbp_bn"].values()} + ) + marks = ",".join("?" * len(after_m)) + assert not _rows( + "SELECT 1 FROM pe_results r JOIN external_scores s USING (claim_id)" + f" WHERE s.source = 'centax' AND ROUND(r.computed_value / 1e6) IN ({marks})", + *after_m, + ) + + +def test_central_static_rows_carry_the_hours_draw_diagnosis(): + rows = _rows( + "SELECT period, diagnosis_class, action_link, diagnosis_rationale" + " FROM comparisons WHERE source = 'centax'" + " AND json_extract(conditions, '$.component') = 'static_cost'" + " AND diagnosis_class IS NOT NULL ORDER BY period" + ) + assert [(r["period"], r["diagnosis_class"]) for r in rows] == [ + (2028, "pe_gap"), + (2030, "pe_gap"), + ] + assert {r["action_link"] for r in rows} == { + "https://github.com/PolicyEngine/microcosm/issues/1126" + } + assert "15.3 of 30 weekly hours" in rows[0]["diagnosis_rationale"] + + +def test_no_reform_estimate_is_a_calibration_target(): + """The 2026-08-02 rulings: another model's reform estimate is never a + calibration target. never_calibrate() is keyed on metric (the poverty + family), so this checks the rule on the built DB for every source: + a claim that scores any world other than the baseline (a policy_ref + reform, a PolicyEngine reform or an input-override counterfactual) is + held out, never consumed_as_target or seed_source.""" + rows = _rows( + "SELECT source, calibration_relationship, COUNT(*) AS n FROM external_scores" + " WHERE json_extract(reform_json, '$.framework') != 'baseline'" + " AND calibration_relationship != 'held_out'" + " GROUP BY source, calibration_relationship" + ) + assert [tuple(r) for r in rows] == [] + lane_rows = _rows( + "SELECT DISTINCT json_extract(reform_json, '$.framework') AS fw," + " calibration_relationship AS cr FROM external_scores" + " WHERE json_extract(publication, '$.registry') = ?", + lane.REGISTRY_MARK, + ) + assert [tuple(r) for r in lane_rows] == [("policy_ref", "held_out")] + + +def test_exported_feed_rows_state_the_lane_signs_and_history(): + """The deployed feed is what a reader sees: CenTax rows point the same + way as PolicyEngine's; CBP states no sign rule, so its rows are + unknown; the central static rows carry both results, the published + figure first and the certified run as the latest.""" + feed = json.loads((REPO / "data" / "populations.json").read_text()) + rows = [ + r + for r in feed["rows"] + if r["conditions"].get("measure_key") in set(lane.MEASURES.values()) + ] + assert len(rows) == 36 + got = Counter( + (r["source"], r["sign_relative_to_pe"], r["sign_basis"]) for r in rows + ) + assert got == { + (lane.CENTAX, "same", "stated:exchequer:same"): 24, + (lane.CBP, "unknown", "no_stated_convention"): 12, + } + central = [ + r + for r in rows + if r["conditions"].get("scenario") == "central" + and r["conditions"].get("component") == "static_cost" + ] + assert len(central) == 2 + for r in central: + assert [x["engine_version"] for x in r["results"]] == ["2.102.3", "2.89.2"] + assert r["latest"]["data_bundle"] == CERTIFIED["revision"] + assert r["latest"]["status_effective"] == "comparable" + assert r["results"][0]["status_effective"] == "constructed" + unscored = [r for r in rows if not r["results"]] + assert len(unscored) == 14 # 12 CenTax margins + CBP's two 2031-32 cells + assert all(r["conditions"].get("pe_missing") for r in unscored)