juaristi22 · Nov 2, 2026 · Oct 30, 2026 · Oct 30, 2026 · Oct 30, 2026 · Oct 30, 2026
diff --git a/changelog.d/uk-company-cars-316.added.md b/changelog.d/uk-company-cars-316.added.md
@@ -0,0 +1 @@
+The UK spine writes company cars (PolicyEngine/uk-autumn-budget-dashboard-2026#316). A new stage, `company_cars`, sets six of policyengine-uk 2.165.0's company car inputs for each employee the FRS job record gives a company car (`EXPBEN01`), through four carriers the root stage now writes from the job tab: the list price, drawn within the respondent's CARVAL band from WAS round-8 company cars' prices grown by the engine's CPI index; the fuel (FUELTYP, and HMRC Table 2C's shares among the unknown fuels, without electricity where the holder reports private fuel); the CO2 figure at HMRC Table 2A's shares among cars that emit CO2, hybrids first into 1 to 50 g/km, which takes a 50-mile electric range; private fuel (EXPBEN03); and capital contributions (CARAMT). The stage runs right after the SPI income stage, so every later stage reads the tape rows' pay without the car benefits (María's ruling of 31 October 2026). A row drawn from the SPI tape keeps its twin's car only where its drawn EPB is positive, and its `employment_income`, the tape's PAY + EPB + TAXTERM, gives up the engine's car and fuel benefits, capped at EPB (María's ruling of 30 October 2026) within the pay its employee pension contribution leaves, so `salary_sacrifice` never meets a contribution above pay; where the two do not fit together the fuel goes first, and a carved row's UC gainful self-employment follows its carved pay. The channel's other rows keep their twin's pay and car. `obr.ni_employer` binds Class 1 employer NICs (`ni_class_1_employer`), since OBR's line leaves Class 1A out, and the employment income count rows count pay plus the car benefits, as the amount rows do. Three company car inputs stay signed at the engine's default: availability, payments for private use and the 2020 registration flag. No HMRC company car fact is a calibration target, and comparing the calibrated file with HMRC's tables is a manual step for now; `uk_stage_company_cars_invariants` blocks a release on any invariant break. The stage fits no model and declares no fit-weight records, so the release-cut weights audit does not read it as a fitting stage that recorded none.
diff --git a/changelog.d/uk-feed-repin-08962f1.changed.md b/changelog.d/uk-feed-repin-08962f1.changed.md
@@ -0,0 +1,5 @@
+The UK Chronicle feed is re-pinned to 08962f1, the merge of PolicyEngine/chronicle#399 on top of PolicyEngine/chronicle#396's merge (1,512,256 rows; the consumer-fact schema is unchanged). It brings HMRC's benefits in kind statistics of July 2026 (544 facts) and the UK facts of PolicyEngine/chronicle#395 (1,076). It also keeps one OBR EFO number per line and year (PolicyEngine/chronicle#397): the March 2026 edition drops the 427 record sets the October 2026 edition also publishes. A new vendored resource, `hmrc_benefits_in_kind_facts.json`, carries the 96 2024-25 rows the company cars stage reads. The other vendored resources, and the DESNZ constituency resource, carry the same rows under the new feed identity. `obr.winter_fuel_allowance` now reads the October edition's FY2025-26 outturn, 0.02% above the March projection it read before.
+
+No EFO edition now carries both years of a calendar-2025 window, so the 19 OBR targets that window FY2024-25 and FY2025-26 declare a new selector key, `source_table_by_opening_year`, as do the three council tax nation rows of PolicyEngine/microcosm#1422 and the signed-out fuel duties row (23 in all). It takes three twelfths of the March edition's FY2024-25 outturn and nine twelfths of the October edition's FY2025-26 outturn (María's ruling of 31 October 2026). Fifteen of the 19 move by under 1%. The four larger moves are domestic rates (+3.2%), Pension Credit (+2.3%), Jobseeker's Allowance (−2.5%) and SDLT (−1.2%); income tax moves −0.5% and Universal Credit +0.7%. Three report-only diagnostics pinned the March edition's FY2025-26 forecasts, which the feed no longer carries: the OBR's capital gains tax receipts, State Pension spending and total NICs. They now pin the October edition's FY2025-26 outturns, and none of them is in the fit.
+
+The national membership also catches up with policyengine-uk 2.164.1's economic assumptions (the October 2026 EFO), which PolicyEngine/microcosm#1421 locked after PolicyEngine/microcosm#1420 had generated the membership. The eight SPI Table 3.10 self-employment bands move −0.55% (OBR mixed income per head), and the thirteen SPI dividend bands move by under 0.001% (GDP per head). The national compile-parity receipt, which builds check against, already carried their new values.
diff --git a/docs/concept-coverage/policyengine-uk.md b/docs/concept-coverage/policyengine-uk.md
@@ -44,7 +44,7 @@ Engine version `2.165.1`; concept schema `5a55669d0447`. Generated by `tools/ref
 | `fact:person.legal_marital_status` | policyengine-uk derives marital status from benefit-unit composition; marital_status and is_married are formula-owned. |
 | `fact:person.partner_person_id` | Partnerships reach policyengine-uk only through benefit-unit membership, a structure the bundle must already carry (the adapter builds no units); is_benunit_head and is_married are formula-owned. |
 | `fact:household.reference_person_id` | is_household_head is formula-owned (the oldest member). The UK build overrides it with the FRS household reference person (frs_spine.py), an override the loader accepts, but it is not an input, so the mapping does not target it. Every housing input is household-level, so nothing is allocated to the reference person. |
-| `fact:person.liquid_financial_assets` | policyengine-uk holds financial wealth on the household: savings ('Household liquid savings'), corporate_wealth ('Wealth held in corporations directly or through investment funds'), gross_financial_wealth and net_financial_wealth are household inputs, and no transform puts a person amount on a household input. Its person-level GBP stock inputs are student_loan_balance, a debt; car_list_price, the price vehicle excise duty's expensive-car supplement tests; and lifetime_isa_balance ('Current value of the person's Lifetime ISAs, cash and stocks and shares together'), one savings wrapper, which the engine documents as already inside gross_financial_wealth and net_financial_wealth. Since policyengine-uk 2.130.0 three more describe a company the person owns, for Universal Credit's regulation 77 (owned_company_capital, owned_company_holding_value and owned_company_trade_assets); every one defaults to no such company, and the build sets none. Since policyengine-uk 2.164.0 cash_isa_balance and stocks_and_shares_isa_balance hold the person's ISAs, and the household cash_isa and stocks_and_shares_isa sum them; the isa_subscriptions stage splits the WAS household totals into them (microcosm#1415), a split of one wrapper's balance, not of liquid financial assets. savings and share_like_capital are two of Universal Credit's capital sources (gov.dwp.universal_credit.means_test.capital.sources); since policyengine-uk 2.147.0 share_like_capital is corporate_wealth unless its components, the household inputs directly_held_shares and unit_and_investment_trusts, add up to more; the WAS stage writes both on its corporate_wealth less the stocks-and-shares ISA, so they and the ISA sum to it (microcosm#1095). Universal Credit's one person-level source is lifetime_isa_countable_capital, the Lifetime ISA at its surrender value (capital.person_sources). uc_reported_capital (benefit unit, default -1) is Universal Credit capital; since policyengine-uk 2.136.1 a value of zero or more set directly is the unit's complete countable capital, in place of every source, property included. benunit_reported_capital, a benefit-unit input ('Savings and investments of the benefit unit's adults (the claimant and any partner) as recorded'), replaces the financial sources only: the household pool it otherwise sums from the household sources, net of what other benefit units report or override, and apportions by claimant and partner count, and the person-level sources of the unit's claimant and partner. Property is valued separately (uc_property_capital) and shared across every claimant and partner in the household. Feeding either would be a modelling choice for a UK build, not a mapping of this concept. |
+| `fact:person.liquid_financial_assets` | policyengine-uk holds financial wealth on the household: savings ('Household liquid savings'), corporate_wealth ('Wealth held in corporations directly or through investment funds'), gross_financial_wealth and net_financial_wealth are household inputs, and no transform puts a person amount on a household input. Its person-level GBP stock inputs are student_loan_balance, a debt; car_list_price, the price vehicle excise duty's expensive-car supplement tests; and lifetime_isa_balance ('Current value of the person's Lifetime ISAs, cash and stocks and shares together'), one savings wrapper, which the engine documents as already inside gross_financial_wealth and net_financial_wealth. Since policyengine-uk 2.130.0 three more describe a company the person owns, for Universal Credit's regulation 77 (owned_company_capital, owned_company_holding_value and owned_company_trade_assets); every one defaults to no such company, and the build sets none. Since policyengine-uk 2.164.0 cash_isa_balance and stocks_and_shares_isa_balance hold the person's ISAs, and the household cash_isa and stocks_and_shares_isa sum them; the isa_subscriptions stage splits the WAS household totals into them (microcosm#1415), a split of one wrapper's balance, not of liquid financial assets. Since policyengine-uk 2.165.0 company_car_list_price and company_car_capital_contributions describe a car the person's employer makes available, whose benefit is taxed as employment income (ITEPA 2003 ss. 121 and 132); the company_cars stage writes them from the FRS job table, and neither is the person's asset. savings and share_like_capital are two of Universal Credit's capital sources (gov.dwp.universal_credit.means_test.capital.sources); since policyengine-uk 2.147.0 share_like_capital is corporate_wealth unless its components, the household inputs directly_held_shares and unit_and_investment_trusts, add up to more; the WAS stage writes both on its corporate_wealth less the stocks-and-shares ISA, so they and the ISA sum to it (microcosm#1095). Universal Credit's one person-level source is lifetime_isa_countable_capital, the Lifetime ISA at its surrender value (capital.person_sources). uc_reported_capital (benefit unit, default -1) is Universal Credit capital; since policyengine-uk 2.136.1 a value of zero or more set directly is the unit's complete countable capital, in place of every source, property included. benunit_reported_capital, a benefit-unit input ('Savings and investments of the benefit unit's adults (the claimant and any partner) as recorded'), replaces the financial sources only: the household pool it otherwise sums from the household sources, net of what other benefit units report or override, and apportions by claimant and partner count, and the person-level sources of the unit's claimant and partner. Property is valued separately (uc_property_capital) and shared across every claimant and partner in the household. Feeding either would be a modelling choice for a UK build, not a mapping of this concept. |
 | `fact:person.public_pension_income` | The State Pension is computed from state_pension_reported, which is formula-owned: the loader accepts it as an override, but it is not an input. |
 | `fact:household.source_household_id` | Original donor household lineage is build provenance and weight diagnostic support, not a policyengine-uk rules input. |
 

diff --git a/docs/uk-chronicle-feed-repin.md b/docs/uk-chronicle-feed-repin.md
@@ -351,3 +351,52 @@ from the same `53815b2` artifact: `hmrc_isa_statistics_facts.json`, the 1,074 ro
 statistics the ISA subscriptions stage and the Lifetime ISA anchor read. None is a calibration
 target, so no reference moves for it. That pull request first pinned #392's head (`e56af9a`) and
 then its merge (`0b721a4`), and took `53815b2` when PR C merged first.
+
+The `08962f1` re-pin (PolicyEngine/microcosm#1430, the merge of PolicyEngine/chronicle#399 on top
+of the `7d5d627` merge of PolicyEngine/chronicle#396) brought HMRC's benefits in kind statistics of
+July 2026 (544 facts) and the UK facts of PolicyEngine/chronicle#395 (1,076). It also carries
+PolicyEngine/chronicle#397, María's rule of 30 October that of two EFO editions' numbers for one
+line and year Chronicle keeps the later, so the March 2026 edition drops the 427 record sets the
+October 2026 edition also publishes (1,512,256 rows; the schema digest is unchanged). The merge
+commit's tree is the pull request head's (`ac9d7bb`), so the artifact was built from that head's
+bundle and carries the merge commit's identity.
+
+A new vendored resource, `hmrc_benefits_in_kind_facts.json`, carries the 96 2024-25 rows the
+company cars stage reads. The other vendored resources, and the constituency energy resource
+rebuilt beside them, carry the same rows under the new feed identity.
+
+After #397 no EFO edition carries both years of a calendar-2025 window: March keeps each line's
+FY2024-25 outturn and October carries FY2025-26 on. The 19 OBR rows that window the two years
+declare `source_table_by_opening_year`, which takes each year from the table that carries it
+(María's ruling of 2026-10-31; `docs/uk-income-anchors-280.md`), as do the three council tax nation
+rows of PolicyEngine/microcosm#1422 and the fuel duties row, which is signed out of the references:
+23 in all. The 19's values move with October's outturns: fifteen by under 1%, with domestic rates +3.2%, Pension Credit +2.3%, Jobseeker's
+Allowance −2.5% and SDLT −1.2%. `obr.winter_fuel_allowance` drops its pin to March's expenditure
+table and reads October's FY2025-26 outturn (+0.02%). Three report-only diagnostics pinned March's
+FY2025-26 forecasts by fact key: `obr.capital_gains_tax`, `obr.state_pension` and the `obr.ni`
+total. They now pin October's outturns (`required_assertion: observation`). The test feed
+fixture's three captured March FY2025-26 rows (income tax, incapacity benefits and capital gains
+tax) give way to October's.
+
+The national membership also catches up with policyengine-uk 2.164.1's October-EFO economic
+assumptions. PolicyEngine/microcosm#1421 locked them after PolicyEngine/microcosm#1420 had
+generated the membership, so the 21 SPI rows grown by an OBR per-head index still carried the
+old factors there: the eight Table 3.10 self-employment bands move −0.55% (mixed income per
+head), and the thirteen dividend bands by under 0.001% (GDP per head). The national compile-parity
+receipt already carried their new values. No other national value moves. The receipt moves the 19
+windowed values and gains two drift rows, domestic rates and Winter Fuel, which had equalled the
+frozen incumbent. The OBR rationales of the parity tool name the two editions, and Winter Fuel has
+a rationale of its own.
+
+On the local surface the reference file and the compile-parity receipt are byte-identical. The
+membership moves only its feed label and the 1,263 DESNZ keys, and the census restates the pin.
+Both surfaces were regenerated from cuts of the pinned artifact, as for `200bf6d` (national
+940,989 rows, local 885,543).
+
+The national register version moves with the OBR values (the SPI rows already compiled at their
+new values), so the uk-candidate-eval scoring register is re-frozen before the next national
+solve. The 23 maps name
+EFO editions, so a later EFO moves them: once Chronicle keeps a newer edition's FY2025-26,
+October's number leaves the feed and the windows refuse until their maps name the new table. The
+high-value home stock still names `735820c`: its inputs are the same on `08962f1` (the feed-gated
+fingerprint check passes), so it was not rebuilt.
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		The UK spine writes company cars (PolicyEngine/uk-autumn-budget-dashboard-2026#316). A new stage, `company_cars`, sets six of policyengine-uk 2.165.0's company car inputs for each employee the FRS job record gives a company car (`EXPBEN01`), through four carriers the root stage now writes from the job tab: the list price, drawn within the respondent's CARVAL band from WAS round-8 company cars' prices grown by the engine's CPI index; the fuel (FUELTYP, and HMRC Table 2C's shares among the unknown fuels, without electricity where the holder reports private fuel); the CO2 figure at HMRC Table 2A's shares among cars that emit CO2, hybrids first into 1 to 50 g/km, which takes a 50-mile electric range; private fuel (EXPBEN03); and capital contributions (CARAMT). The stage runs right after the SPI income stage, so every later stage reads the tape rows' pay without the car benefits (María's ruling of 31 October 2026). A row drawn from the SPI tape keeps its twin's car only where its drawn EPB is positive, and its `employment_income`, the tape's PAY + EPB + TAXTERM, gives up the engine's car and fuel benefits, capped at EPB (María's ruling of 30 October 2026) within the pay its employee pension contribution leaves, so `salary_sacrifice` never meets a contribution above pay; where the two do not fit together the fuel goes first, and a carved row's UC gainful self-employment follows its carved pay. The channel's other rows keep their twin's pay and car. `obr.ni_employer` binds Class 1 employer NICs (`ni_class_1_employer`), since OBR's line leaves Class 1A out, and the employment income count rows count pay plus the car benefits, as the amount rows do. Three company car inputs stay signed at the engine's default: availability, payments for private use and the 2020 registration flag. No HMRC company car fact is a calibration target, and comparing the calibrated file with HMRC's tables is a manual step for now; `uk_stage_company_cars_invariants` blocks a release on any invariant break. The stage fits no model and declares no fit-weight records, so the release-cut weights audit does not read it as a fitting stage that recorded none.