From 1d6c24c71de149458fc8c4dabdb9d9e291c655a5 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Thu, 29 Oct 2026 08:31:17 -0400 Subject: [PATCH 01/17] Triple Lock post: the Budget's inputs (October forecast, 2.151.0, Microcosm) Every PolicyEngine figure now comes from uk-triple-lock's scenarios-first file at 94d3bb6d (#49): the OBR's October 2026 forecast to 2030 with its March 2026 long-term determinants after (d2124), PolicyEngine UK 2.151.0 and Microcosm UK 2024-25. Scenarios only, as under ruling (a). Co-Authored-By: Claude Opus 5.5 --- .../articles/uk-triple-lock-budget-2026.md | 134 +++++++++--------- app/src/data/posts/posts.json | 2 +- 2 files changed, 67 insertions(+), 69 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 6068a329c..f943aab1a 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,18 +1,18 @@ -> **Updated 28 October 2026.** We first published this post earlier on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We have withdrawn the average and the range around it, and this version reports scenarios instead. We no longer describe any of the model's paths as typical or likely, and we are updating our dashboard to match. +> **Updated 29 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/94d3bb6d8868265e3bd28a5a95020e0922314d27/data/interim/scenarios-first.json). Our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) still shows its pre-Budget release, which used the OBR's March 2026 forecast, PolicyEngine UK 2.125.1 and the Enhanced Family Resources Survey 2024-25. > -> Every PolicyEngine figure below comes from the pre-Budget release of our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock), finished on 15 October, before September's inflation figure (21 October) and the Budget's forecast (28 October) were published. It uses the Office for Budget Responsibility's (OBR) March 2026 forecast, PolicyEngine UK 2.125.1 and the Enhanced Family Resources Survey 2024-25. Our other Budget posts use PolicyEngine UK 2.151.0 and the OBR's October 2026 forecast. We are rebuilding the dashboard with PolicyEngine UK 2.151.0, the Microcosm UK dataset, and the OBR's October 2026 forecast to 2030 with its March 2026 long-term assumptions after that, and its figures may change. +> We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. Each April, the State Pension rises by the highest of inflation, earnings growth and 2.5%. This rule is the Triple Lock. The [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026 keeps it until April 2029 and adjusts it from April 2030. Earnings growth stops being a minimum rise, and the pension instead keeps its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this adjusted Triple Lock cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These are cash figures for Great Britain and cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)).[^name] -Key findings for 2039-40, the first of the two years DWP costs (pre-Budget estimates, UK, cash terms): +Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): -- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's March 2026 central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.9 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.3 billion. -- After the lower pension reduces income tax and raises means-tested benefits, the two savings are £0.6 billion and £7.3 billion. -- For Great Britain, the area DWP covers, our two figures are £0.8 billion and £10.0 billion, against DWP's £15 billion. Our survey counts 15% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. -- Under the OBR's long-term assumption, the full new State Pension is £865 a year lower than under the Triple Lock, and 27% of UK households have less income: nearly nine in ten of the households with someone over State Pension age. -- Had the adjusted Triple Lock started in April 2017, the full new State Pension would be £653 a year lower in 2026-27, 5.2% below the amount paid. +- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.5 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.8 billion. +- After the lower pension reduces income tax and raises means-tested benefits, the two savings are £0.3 billion and £7.3 billion. +- For Great Britain, the area DWP covers, our two figures are £0.5 billion and £10.5 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. +- Under the OBR's long-term assumption, the full new State Pension is £829 a year lower than under the Triple Lock, and 30% of UK households have less income: nearly nine in ten of the households with someone over State Pension age. +- Had the adjusted Triple Lock started in April 2017, the full new State Pension would be £642 a year lower in 2026-27, 5.1% below the amount paid. [Explore the dashboard](https://www.policyengine.org/uk/triple-lock) | [Read the code](https://github.com/PolicyEngine/uk-triple-lock) @@ -47,7 +47,7 @@ The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/governmen ### Both rules over the last ten years -To see the two rules at work, we replay today's rules on the ONS figures published in September 2026 for April 2017 to April 2026. We assume the adjusted Triple Lock started in April 2017 and protected the pension's April 2016 value relative to earnings. +To see the two rules at work, we replay today's rules on the ONS figures published in October 2026 for April 2017 to April 2026. We assume the adjusted Triple Lock started in April 2017 and protected the pension's April 2016 value relative to earnings. | April | Inflation | Earnings growth | Triple Lock | Adjusted Triple Lock | | ----- | --------- | --------------- | ----------- | -------------------- | @@ -60,9 +60,9 @@ To see the two rules at work, we replay today's rules on the ONS figures publish | 2023 | 10.1% | 5.5% | 10.1% | 10.1% | | 2024 | 6.7% | 8.5% | 8.5% | 6.7% | | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | -| 2026 | 3.8% | 4.9% | 4.9% | 3.8% | +| 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the rises were set: on these figures April 2025's rise is 4.4% and April 2026's 4.9%, while the rises paid were 4.1% and 4.8%. The figures are in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/8838fd4/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the rises were set: on these figures, for example, April 2025's rise is 4.4%, while the rise paid was 4.1%. The figures are in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/94d3bb6d8868265e3bd28a5a95020e0922314d27/data/triple_lock_actual_inputs.csv). In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. @@ -93,76 +93,74 @@ In April 2020, earnings growth of 4.0% would have lifted the floor above a pensi { "year": 2023, "triple_lock": 131.2, "adjusted": 130.2, "floor": 120.6 }, { "year": 2024, "triple_lock": 142.4, "adjusted": 138.9, "floor": 130.9 }, { "year": 2025, "triple_lock": 148.6, "adjusted": 142.4, "floor": 136.6 }, - { "year": 2026, "triple_lock": 155.9, "adjusted": 147.8, "floor": 143.3 } + { "year": 2026, "triple_lock": 155.7, "adjusted": 147.8, "floor": 143.2 } ], "yIncludesZero": false, "yLabel": "Index", "format": { "decimals": 0 }, "xFormat": { "decimals": 0 }, - "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, ONS figures as published in September 2026 (PolicyEngine/uk-triple-lock, pre-Budget release). April 2022 uses inflation for earnings under both rules." + "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, ONS figures as published in October 2026 (PolicyEngine/uk-triple-lock). April 2022 uses inflation for earnings under both rules." } ``` -By April 2026, the full new State Pension would be £11,895 a year under the adjusted Triple Lock, against the £12,548 paid: £653, or 5.2%, lower. Running PolicyEngine UK with that lower pension puts the saving at £6.0 billion in 2026-27, or £4.3 billion after the lower pension reduces income tax and raises means-tested benefits. +By April 2026, the full new State Pension would be £11,906 a year under the adjusted Triple Lock, against the £12,548 paid: £642, or 5.1%, lower. Running PolicyEngine UK with that lower pension puts the saving at £6.5 billion in 2026-27, or £4.6 billion after the lower pension reduces income tax and raises means-tested benefits. ## What it saves The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We report two scenarios. Neither is a forecast of the most likely outcome. -**The OBR's central path.** This takes the OBR's March 2026 forecast to 2030 and its long-term projections after: +**The OBR's central path.** This takes the OBR's October 2026 forecast to 2030 and its March 2026 long-term projections after: | April | Inflation | Earnings growth | Triple Lock | Adjusted Triple Lock | | ----- | --------- | --------------- | ----------- | -------------------- | -| 2028 | 2.0% | 2.4% | 2.5% | 2.5% | +| 2028 | 3.4% | 3.3% | 3.4% | 3.4% | | 2029 | 2.0% | 2.1% | 2.5% | 2.5% | -| 2030 | 2.0% | 2.2% | 2.5% | 2.5% | -| 2031 | 2.0% | 2.4% | 2.5% | 2.5% | +| 2030 | 2.1% | 2.3% | 2.5% | 2.5% | +| 2031 | 2.0% | 2.5% | 2.5% | 2.5% | | 2032 | 2.0% | 2.7% | 2.7% | 2.5% | -| 2033 | 2.0% | 3.0% | 3.0% | 2.8% | +| 2033 | 2.0% | 3.0% | 3.0% | 3.0% | | 2034 | 2.0% | 3.3% | 3.3% | 3.3% | -Earnings growth stays below 2.5% in the inputs to the rises of April 2028 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the rises of April 2030 and April 2031 lift the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 2.8%, the rise that brings the pension back to its floor, against 3.0%. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension is £18,639 a year under the adjusted Triple Lock, £73 below the Triple Lock's £18,712, and the saving is £0.9 billion. +In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 3.0%, the rise that brings the pension back to its floor, and the same as the Triple Lock. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension is £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the saving is £0.5 billion. -On the OBR's October 2026 forecast, published with the Budget, the April 2028 rise is 3.4% rather than 2.5%, because forecast inflation of 3.4% comes out highest. The rises of April 2029 to April 2031 stay at 2.5%. Our rebuild uses the October forecast to 2030 and, until the OBR publishes long-term assumptions to match it, the March 2026 ones after that. - -**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. The saving in 2039-40 is £10.3 billion, and the full new State Pension is £865 a year lower than under the Triple Lock. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and is not a forecast. +**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. The saving in 2039-40 is £10.8 billion, and the full new State Pension is £829 a year lower than under the Triple Lock. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and is not a forecast. ```chart { "type": "bar", "title": "Saving from the adjusted Triple Lock in two scenarios, 2032-33 to 2039-40", - "subtitle": "Fall in State Pension spending against the Triple Lock, UK, £ billion in cash terms. Pre-Budget estimates.", + "subtitle": "Fall in State Pension spending against the Triple Lock, UK, £ billion in cash terms. OBR October 2026 forecast to 2030, March 2026 long-term projections after.", "xKey": "year", "series": [ { "key": "central", "name": "OBR central path", "color": "quinary" }, { "key": "assumption", "name": "OBR long-term assumption", "color": "primary" } ], "data": [ - { "year": "2032-33", "central": 0.3, "assumption": 0.75 }, - { "year": "2033-34", "central": 0.63, "assumption": 1.89 }, - { "year": "2034-35", "central": 0.66, "assumption": 3.17 }, - { "year": "2035-36", "central": 0.7, "assumption": 4.24 }, - { "year": "2036-37", "central": 0.74, "assumption": 5.6 }, - { "year": "2037-38", "central": 0.78, "assumption": 7.09 }, - { "year": "2038-39", "central": 0.82, "assumption": 8.52 }, - { "year": "2039-40", "central": 0.86, "assumption": 10.29 } + { "year": "2032-33", "central": 0.33, "assumption": 0.81 }, + { "year": "2033-34", "central": 0.35, "assumption": 1.72 }, + { "year": "2034-35", "central": 0.37, "assumption": 3.09 }, + { "year": "2035-36", "central": 0.39, "assumption": 4.24 }, + { "year": "2036-37", "central": 0.41, "assumption": 5.71 }, + { "year": "2037-38", "central": 0.43, "assumption": 7.32 }, + { "year": "2038-39", "central": 0.45, "assumption": 8.87 }, + { "year": "2039-40", "central": 0.48, "assumption": 10.78 } ], "valueLabels": false, "yLabel": "Saving", "format": { "decimals": 1, "prefix": "£", "suffix": "bn" }, - "source": "Source: PolicyEngine UK (policyengine.py 6.2.6, Enhanced FRS 2024-25), pre-Budget release of PolicyEngine/uk-triple-lock. Neither scenario saves anything before 2032-33." + "source": "Source: PolicyEngine UK 2.151.0 (policyengine.py 6.11.1) on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. Neither scenario saves anything before 2032-33." } ``` ### From gross to net -With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.3 billion. Lower income tax receipts offset £2.2 billion of that, higher Pension Credit £0.6 billion, and higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, £0.2 billion, leaving a net saving of £7.3 billion. +With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.8 billion. Lower income tax receipts offset £2.4 billion of that, higher Pension Credit £0.8 billion, and higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, £0.3 billion, leaving a net saving of £7.3 billion. ```chart { "type": "waterfall", "title": "From the gross to the net saving, 2039-40", - "subtitle": "OBR long-term assumption, UK, £ billion in cash terms. Pre-Budget estimates.", + "subtitle": "OBR long-term assumption, UK, £ billion in cash terms.", "panelKey": "measure", "steps": [ { "key": "gross", "name": "Pension", "kind": "level", "legend": "Saving", "color": "primary" }, @@ -171,32 +169,32 @@ With a lower State Pension, pensioners pay less income tax and receive more mean { "key": "net", "name": "Net", "kind": "level", "legend": "Saving", "color": "primary" } ], "data": [ - { "measure": "2039-40", "gross": 10.292565051193662, "income_tax": -2.229207947842042, "benefits": -0.787262869236713, "net": 7.2760942341149075 } + { "measure": "2039-40", "gross": 10.776172900359711, "income_tax": -2.355861154190279, "benefits": -1.1459951874199987, "net": 7.274316558749433 } ], "format": { "decimals": 1, "prefix": "£", "suffix": "bn" }, - "source": "Source: PolicyEngine UK, pre-Budget release of PolicyEngine/uk-triple-lock. Pension: fall in State Pension spending. Benefits: Pension Credit (£0.6bn), and Housing Benefit, council tax reduction, Winter Fuel Payment, Universal Credit and other spending and taxes (£0.2bn). Labels rounded; steps computed from unrounded figures." + "source": "Source: PolicyEngine UK 2.151.0 on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. Pension: fall in State Pension spending. Benefits: Pension Credit (£0.8bn), and Housing Benefit, council tax reduction, Winter Fuel Payment, Universal Credit and other spending and taxes (£0.3bn). Labels rounded; steps computed from unrounded figures." } ``` ## Against the official estimates -DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our closest figures are the gross savings in Great Britain: £0.8 billion on the OBR's central path and £10.0 billion under its long-term assumption. +DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our closest figures are the gross savings in Great Britain: £0.5 billion on the OBR's central path and £10.5 billion under its long-term assumption. The estimates differ in three ways, and DWP's published assumptions do not let us size them ([HL3774](https://questions-statements.parliament.uk/written-questions/detail/2026-10-05/hl3774)): -- DWP projects the population and each person's pension with Pensim3, its dynamic microsimulation model. We reweight today's Family Resources Survey households to the ONS population projection by age and sex, holding each person's age fixed. -- Our survey counts less State Pension in payment than DWP does: £125.9 billion in Great Britain in 2026-27, against DWP's £148.3 billion, both excluding payments abroad. The saving scales with the basic and new State Pension, which DWP publishes only with payments abroad included, and only to 2030-31, so we cannot match that part of the bill exactly. +- DWP projects the population and each person's pension with Pensim3, its dynamic microsimulation model. We reweight the households in Microcosm UK, our dataset built from the Family Resources Survey, to the ONS population projection by age and sex, holding each person's age fixed. +- Our data count slightly less State Pension in payment than DWP does: £144.8 billion in Great Britain in 2026-27, against DWP's £148.3 billion, both excluding payments abroad. The saving scales with the basic and new State Pension, which DWP publishes only with payments abroad included, and only to 2030-31, so we cannot match that part of the bill exactly. - DWP says it applies "a path for uprating" but publishes no rates for it, and its answer to HL3774 says the analysis used no single-year assumptions for earnings or prices growth. Our projections end in 2039-40, so we have no figure to set against the £50 billion. -The OBR's October 2026 Economic and fiscal outlook builds the adjusted Triple Lock into its forecast from 2030-31. It finds that the adjusted Triple Lock "does not currently affect state pension costs over the forecast period", because 2.5% applies under both rules in 2030-31 and 2031-32 (paragraph 3.50 and its footnote 17). Our central path, on the March forecast, also pays 2.5% under both rules in April 2030 and April 2031, and saves nothing until 2032-33. +The OBR's October 2026 Economic and fiscal outlook builds the adjusted Triple Lock into its forecast from 2030-31. It finds that the adjusted Triple Lock "does not currently affect state pension costs over the forecast period", because 2.5% applies under both rules in 2030-31 and 2031-32 (paragraph 3.50 and its footnote 17). Our central path also pays 2.5% under both rules in April 2030 and April 2031, and saves nothing until 2032-33. -The OBR gives no figure for 2039-40. Instead, it replays both rules on prices and earnings since 1990 and finds State Pension payments 1.4% lower each year on average than under the Triple Lock (paragraph 3.51). That replay protects the pension's 1990 value relative to earnings and averages over every year since, so its 1.4% does not compare directly with our 5.2% for 2026-27 from a 2017 start. The OBR also estimates, provisionally, that if the next 50 years resembled the last 35, the adjusted Triple Lock would leave debt in 2075-76 around 25% of GDP lower than under the Triple Lock. It notes that the overall effect of the adjusted Triple Lock and the government's planned expansion of social care on the long-term public finances depends on what that expansion costs (paragraphs 3.52 to 3.54). +The OBR gives no figure for 2039-40. Instead, it replays both rules on prices and earnings since 1990 and finds State Pension payments 1.4% lower each year on average than under the Triple Lock (paragraph 3.51). That replay protects the pension's 1990 value relative to earnings and averages over every year since, so its 1.4% does not compare directly with our 5.1% for 2026-27 from a 2017 start. The OBR also estimates, provisionally, that if the next 50 years resembled the last 35, the adjusted Triple Lock would leave debt in 2075-76 around 25% of GDP lower than under the Triple Lock. It notes that the overall effect of the adjusted Triple Lock and the government's planned expansion of social care on the long-term public finances depends on what that expansion costs (paragraphs 3.52 to 3.54). ## Who has less income -Under the OBR's long-term assumption in 2039-40, 27% of UK households have less income than under the Triple Lock, by £810 a year on average. Households with someone over State Pension age make up 31% of all households, so nearly nine in ten of them lose. Households with nobody over State Pension age see no change. Pensioner couples have £1,097 a year less on average, 1.4% of their income, and single pensioners £415 less, 1.1%. By income decile, households in the second and third deciles lose the largest share of their income, 0.7%. This release records how many households lose more than £1 a year, but not how many gain. On the OBR's central path, 27% of households also lose, by £69 a year on average. +Under the OBR's long-term assumption in 2039-40, 30% of UK households have less income than under the Triple Lock, by £795 a year on average. Households with someone over State Pension age make up 34% of all households, so nearly nine in ten of them lose. Households with nobody over State Pension age see no change. Pensioner couples have £1,062 a year less on average, 1.3% of their income, and single pensioners £387 less, 0.8%. By income decile, households in the third decile lose the largest share of their income, 0.6%. Fewer than 1 in 300 households have more income. On the OBR's central path, 30% of households also lose, by £35 a year on average. ```chart { @@ -206,33 +204,33 @@ Under the OBR's long-term assumption in 2039-40, 27% of UK households have less "xKey": "decile", "series": [{ "key": "change", "name": "Change in household income", "color": "primary" }], "data": [ - { "decile": "1st", "change": -0.57 }, - { "decile": "2nd", "change": -0.7 }, - { "decile": "3rd", "change": -0.71 }, - { "decile": "4th", "change": -0.47 }, - { "decile": "5th", "change": -0.37 }, + { "decile": "1st", "change": -0.35 }, + { "decile": "2nd", "change": -0.47 }, + { "decile": "3rd", "change": -0.55 }, + { "decile": "4th", "change": -0.52 }, + { "decile": "5th", "change": -0.48 }, { "decile": "6th", "change": -0.35 }, - { "decile": "7th", "change": -0.29 }, - { "decile": "8th", "change": -0.21 }, - { "decile": "9th", "change": -0.19 }, - { "decile": "10th", "change": -0.07 } + { "decile": "7th", "change": -0.32 }, + { "decile": "8th", "change": -0.22 }, + { "decile": "9th", "change": -0.15 }, + { "decile": "10th", "change": -0.06 } ], "valueLabels": false, "yLabel": "Change in income", "format": { "decimals": 2, "suffix": "%" }, - "source": "Source: PolicyEngine UK, pre-Budget release of PolicyEngine/uk-triple-lock. The 1st decile has the lowest income." + "source": "Source: PolicyEngine UK 2.151.0 on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. The 1st decile has the lowest income." } ``` -Under the same scenario, the share of pensioners in poverty after housing costs rises from 12.9% to 13.8% on the relative measure and from 7.7% to 8.8% on the absolute measure. Across the whole population, relative poverty after housing costs rises from 21.1% to 21.3% and absolute poverty from 16.7% to 17.1%. +Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.7% to 11.5% on the relative measure and from 6.8% to 7.7% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative line is 60% of median income, and lower pensions pull the median down. ## How uncertain the saving is -The two scenarios differ more than tenfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. +The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we built a statistical model of prices and earnings and simulated 50,000 paths. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete, with a rule: if the model failed, we would report scenarios only. There were six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. Scored on all seven on 28 October, the model failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete, and committed, if the model failed, to reporting scenarios only. There were six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. Scored on all seven on 28 October, the model failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. -So we report the two scenarios, and they are not a range: the saving could fall outside them. The dashboard's next release, rebuilt on the Budget's inputs, reports scenarios only. +So we report the two scenarios, and they are not a range: the saving could fall outside them.
Methodology and limitations @@ -243,34 +241,34 @@ Both rules follow the Triple Lock through April 2029. Each April, the basic and ### Scenarios -The central path takes the OBR's [March 2026 Economic and fiscal outlook](https://obr.uk/efo/economic-and-fiscal-outlook-march-2026/) to 2030 and its [long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx) after, which the OBR describes as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants. It pays 2.5% for the rises of April 2028 to April 2031, as on the central path. From April 2034, it pays average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. +The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants. From April 2034, it pays average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. -The replay of April 2017 to April 2026 uses September CPI and May to July earnings growth as published by ONS in September 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on the 2024-25 survey cover 2024-25 to 2026-27. +The replay of April 2017 to April 2026 uses September CPI and May to July earnings growth as published by ONS in October 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. -The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000, with one further run to check a path on which both rules pay identical rates. The test registered on 6 October scored it on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022), and set limits on coverage and bias. It failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). +The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test registered on 6 October scored it on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022), and set limits on coverage and bias. It failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). ### Model and data -Every fiscal and household figure is a full run of PolicyEngine UK (policyengine.py 6.2.6, policyengine-uk 2.125.1) on the Enhanced Family Resources Survey 2024-25 (policyengine-uk-data 1.58.0). Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow it. The net saving is the change in the government's balance across every tax and benefit the model covers. +Every fiscal and household figure is a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households built from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow it. The net saving is the change in the government's balance across every tax and benefit the model covers. ### Population -We reweight the survey households each year after 2025 so that each age and sex group grows as in the ONS 2024-based population projection. Each person keeps the age they have in the survey, with ages above the survey's top code of 80 spread using ONS shares. State Pension type follows birth cohort: basic for people who reached State Pension age before 6 April 2016, new after. The survey excludes pensioners living abroad. +We reweight the households each year after 2025-26 so that each age and sex group grows as in the ONS 2024-based population projection. Each person keeps the age they have in the data. State Pension type follows birth cohort: basic for people who reached State Pension age before 6 April 2016, new after. The data exclude pensioners living abroad. ### Limitations -- The 2039-40 population is today's survey households, reweighted. The model does not simulate deaths, migration or contribution records, so a pensioner's record in 2039-40 is that of someone the same age today. -- Housing Benefit and council tax reduction respond only for households already claiming them, so nobody the adjusted Triple Lock makes newly eligible starts claiming them. That understates those offsets and so overstates the net saving. Pension Credit can go to newly eligible households, when the take-up draw stored in the survey data says they would claim it. +- The 2039-40 population is today's households, reweighted. The model does not simulate deaths, migration or contribution records, so a pensioner's record in 2039-40 is that of someone the same age today. +- Housing Benefit and council tax reduction respond only for households already claiming them, so nobody the adjusted Triple Lock makes newly eligible starts claiming them. That understates those offsets and so overstates the net saving. Pension Credit can go to newly eligible households, when the take-up draw stored in the data says they would claim it. - The Savings Credit threshold and the pension-age Housing Benefit allowances, which the model also uses for council tax reduction, rise with CPI in the model, while the Pension Credit guarantee rises with earnings. In law, maximum Savings Credit is 60% of the gap between the guarantee and the threshold, and DWP describes protected Housing Benefit allowances as the guarantee plus maximum Savings Credit. The guarantee's link to earnings does not set these amounts' future paths. Projecting them from the guarantee instead of CPI is an alternative assumption that could change the Pension Credit, Housing Benefit and council tax reduction offsets under both rules. -- The model splits each reported State Pension into a flat-rate part, up to the full basic or new rate, and anything above it, which rises with CPI under both rules. Where a pension is below the full rate, the model scales all of it with the flat rate, including any Graduated Retirement Benefit or extra pension for deferring. In law those parts rise with prices: in April 2025, they rose 1.7% while the basic State Pension rose 4.1% ([DWP](https://www.gov.uk/government/publications/benefit-and-pension-rates-2025-to-2026/benefit-and-pension-rates-2025-to-2026)). They would rise the same under both rules, so scaling them overstates the saving. -- In the survey runs, rents and council tax stay at their 2030 levels after 2030, and dividend, property, savings and self-employment income do not follow the scenario. +- The OBR publishes no May to July earnings forecast, so the Pension Credit guarantee's rises from April 2028 to April 2031 follow its April to June figure, as the Triple Lock's earnings input does. +- The model splits each reported State Pension into a flat-rate part, up to the full basic or new rate, and anything above it, which rises with CPI under both rules. Where a pension falls short of the full rate, the model scales all of it with the flat rate, including any Graduated Retirement Benefit or extra pension for deferring. In law those parts rise with prices: in April 2025, they rose 1.7% while the basic State Pension rose 4.1% ([DWP](https://www.gov.uk/government/publications/benefit-and-pension-rates-2025-to-2026/benefit-and-pension-rates-2025-to-2026)). They would rise the same under both rules, so scaling them overstates the saving. +- Rents and council tax stay at their 2030 levels after 2030, and dividend, property, savings and self-employment income do not follow the scenario. - Nobody changes how much they work, save or claim in response to the adjusted Triple Lock. - Costs are for the UK in cash terms; comparisons with DWP use our figures for Great Britain. -- Household weights sit 0.44% below their calibrated levels in every year ([policyengine-uk-data#538](https://github.com/PolicyEngine/policyengine-uk-data/issues/538)).
-[^name]: Until its next release, our dashboard calls the policy by its earlier name, the Burnham plan. +[^name]: Our dashboard's pre-Budget release calls the policy by its earlier name, the Burnham plan. --- diff --git a/app/src/data/posts/posts.json b/app/src/data/posts/posts.json index 7cabcd248..68728d4b8 100644 --- a/app/src/data/posts/posts.json +++ b/app/src/data/posts/posts.json @@ -64,7 +64,7 @@ }, { "title": "The adjusted Triple Lock in the Autumn Budget 2026", - "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. We explain how the Triple Lock works and what the adjusted rule saves: on our pre-Budget estimates, \u00a30.9 billion in 2039-40 on the OBR's March 2026 central path and \u00a310.3 billion if the Triple Lock outpaces earnings as the OBR assumes.", + "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. We explain how the Triple Lock works and what the adjusted rule saves: \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.8 billion if the Triple Lock outpaces earnings as the OBR assumes.", "date": "2026-10-28 17:00:00", "tags": ["uk", "policy", "autumn-budget"], "authors": ["max-ghenis"], From afb7552a6d9234d10d003bcf02858cc00f6d9323 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Thu, 29 Oct 2026 09:29:29 -0400 Subject: [PATCH 02/17] Triple Lock post: poverty line, April 2033 rounding, Microcosm source (Sol r8) Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index f943aab1a..fb430a0fe 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -121,7 +121,7 @@ The saving depends on how often inflation or 2.5% exceeds earnings growth after | 2033 | 2.0% | 3.0% | 3.0% | 3.0% | | 2034 | 2.0% | 3.3% | 3.3% | 3.3% | -In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 3.0%, the rise that brings the pension back to its floor, and the same as the Triple Lock. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension is £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the saving is £0.5 billion. +In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 3.0%, the same as the Triple Lock: the rise that returns the pension to its floor, rounded up to 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension is £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the saving is £0.5 billion. **The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. The saving in 2039-40 is £10.8 billion, and the full new State Pension is £829 a year lower than under the Triple Lock. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and is not a forecast. @@ -182,7 +182,7 @@ DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £5 The estimates differ in three ways, and DWP's published assumptions do not let us size them ([HL3774](https://questions-statements.parliament.uk/written-questions/detail/2026-10-05/hl3774)): -- DWP projects the population and each person's pension with Pensim3, its dynamic microsimulation model. We reweight the households in Microcosm UK, our dataset built from the Family Resources Survey, to the ONS population projection by age and sex, holding each person's age fixed. +- DWP projects the population and each person's pension with Pensim3, its dynamic microsimulation model. We reweight the households in Microcosm UK, our dataset of UK households, which starts from the Family Resources Survey, to the ONS population projection by age and sex, holding each person's age fixed. - Our data count slightly less State Pension in payment than DWP does: £144.8 billion in Great Britain in 2026-27, against DWP's £148.3 billion, both excluding payments abroad. The saving scales with the basic and new State Pension, which DWP publishes only with payments abroad included, and only to 2030-31, so we cannot match that part of the bill exactly. - DWP says it applies "a path for uprating" but publishes no rates for it, and its answer to HL3774 says the analysis used no single-year assumptions for earnings or prices growth. @@ -222,7 +222,7 @@ Under the OBR's long-term assumption in 2039-40, 30% of UK households have less } ``` -Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.7% to 11.5% on the relative measure and from 6.8% to 7.7% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative line is 60% of median income, and lower pensions pull the median down. +Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.7% to 11.5% on the relative measure and from 6.8% to 7.7% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative line is 60% of the median person's household income after housing costs, adjusted for household size, and lower pensions pull that median down. ## How uncertain the saving is @@ -249,7 +249,7 @@ The withdrawn average came from a monthly statistical model of CPI and average w ### Model and data -Every fiscal and household figure is a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households built from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow it. The net saving is the change in the government's balance across every tax and benefit the model covers. +Every fiscal and household figure is a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households, which starts from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow it. The net saving is the change in the government's balance across every tax and benefit the model covers. ### Population From 883ea82b519b30bfc9d2f409899fea0c409f7187 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Thu, 29 Oct 2026 09:30:09 -0400 Subject: [PATCH 03/17] Triple Lock post: drop the unmeasured share of pensioner households losing Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index fb430a0fe..3f4bbb4c1 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -11,7 +11,7 @@ Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): - The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.5 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.8 billion. - After the lower pension reduces income tax and raises means-tested benefits, the two savings are £0.3 billion and £7.3 billion. - For Great Britain, the area DWP covers, our two figures are £0.5 billion and £10.5 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. -- Under the OBR's long-term assumption, the full new State Pension is £829 a year lower than under the Triple Lock, and 30% of UK households have less income: nearly nine in ten of the households with someone over State Pension age. +- Under the OBR's long-term assumption, the full new State Pension is £829 a year lower than under the Triple Lock, and 30% of UK households have less income. - Had the adjusted Triple Lock started in April 2017, the full new State Pension would be £642 a year lower in 2026-27, 5.1% below the amount paid. [Explore the dashboard](https://www.policyengine.org/uk/triple-lock) | [Read the code](https://github.com/PolicyEngine/uk-triple-lock) @@ -194,7 +194,7 @@ The OBR gives no figure for 2039-40. Instead, it replays both rules on prices an ## Who has less income -Under the OBR's long-term assumption in 2039-40, 30% of UK households have less income than under the Triple Lock, by £795 a year on average. Households with someone over State Pension age make up 34% of all households, so nearly nine in ten of them lose. Households with nobody over State Pension age see no change. Pensioner couples have £1,062 a year less on average, 1.3% of their income, and single pensioners £387 less, 0.8%. By income decile, households in the third decile lose the largest share of their income, 0.6%. Fewer than 1 in 300 households have more income. On the OBR's central path, 30% of households also lose, by £35 a year on average. +Under the OBR's long-term assumption in 2039-40, 30% of UK households have less income than under the Triple Lock, by £795 a year on average. Households with someone over State Pension age make up 34% of all households, and those with nobody over State Pension age see no change on average. Pensioner couples have £1,062 a year less on average, 1.3% of their income, and single pensioners £387 less, 0.8%. By income decile, households in the third decile lose the largest share of their income, 0.6%. Fewer than 1 in 300 households have more income. On the OBR's central path, 30% of households also lose, by £35 a year on average. ```chart { From d77c4f8fde7a8b51a22e43ee93e22ab4fd413805 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Thu, 29 Oct 2026 10:10:22 -0400 Subject: [PATCH 04/17] Triple Lock post: attribute 'earnings plus 0.6 points' to the OBR row Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 3f4bbb4c1..be8ec7792 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -241,7 +241,7 @@ Both rules follow the Triple Lock through April 2029. Each April, the basic and ### Scenarios -The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants. From April 2034, it pays average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. +The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants. From April 2034, it pays what that row describes as average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. The replay of April 2017 to April 2026 uses September CPI and May to July earnings growth as published by ONS in October 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. From f98a38f0cf9057a77fbbed43e9f7059adb12677c Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 13:08:34 -0400 Subject: [PATCH 05/17] Triple Lock post: final #49 figures and the rounding note MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Read every figure from the binding results file at uk-triple-lock 26fbc89. Only the OBR long-term assumption scenario moves: £10.9bn gross and £836 a year in 2039-40, with its household, decile and poverty figures. Point the file links at that commit, date the update 30 October, and say that the gross-to-net steps do not sum exactly once rounded. Co-Authored-By: Claude Opus 5.5 --- .../articles/uk-triple-lock-budget-2026.md | 46 +++++++++---------- 1 file changed, 23 insertions(+), 23 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index be8ec7792..5d12f5c44 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,4 +1,4 @@ -> **Updated 29 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/94d3bb6d8868265e3bd28a5a95020e0922314d27/data/interim/scenarios-first.json). Our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) still shows its pre-Budget release, which used the OBR's March 2026 forecast, PolicyEngine UK 2.125.1 and the Enhanced Family Resources Survey 2024-25. +> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/26fbc89f7958c09291dcd33ec7fda8dcb067737c/data/results.json). Our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) still shows its pre-Budget release, which used the OBR's March 2026 forecast, PolicyEngine UK 2.125.1 and the Enhanced Family Resources Survey 2024-25. > > We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. @@ -8,10 +8,10 @@ The Department for Work and Pensions (DWP) estimates in its [State Pension uprat Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): -- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.5 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.8 billion. +- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.5 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.9 billion. - After the lower pension reduces income tax and raises means-tested benefits, the two savings are £0.3 billion and £7.3 billion. -- For Great Britain, the area DWP covers, our two figures are £0.5 billion and £10.5 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. -- Under the OBR's long-term assumption, the full new State Pension is £829 a year lower than under the Triple Lock, and 30% of UK households have less income. +- For Great Britain, the area DWP covers, our two figures are £0.5 billion and £10.6 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. +- Under the OBR's long-term assumption, the full new State Pension is £836 a year lower than under the Triple Lock, and 30% of UK households have less income. - Had the adjusted Triple Lock started in April 2017, the full new State Pension would be £642 a year lower in 2026-27, 5.1% below the amount paid. [Explore the dashboard](https://www.policyengine.org/uk/triple-lock) | [Read the code](https://github.com/PolicyEngine/uk-triple-lock) @@ -62,7 +62,7 @@ To see the two rules at work, we replay today's rules on the ONS figures publish | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the rises were set: on these figures, for example, April 2025's rise is 4.4%, while the rise paid was 4.1%. The figures are in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/94d3bb6d8868265e3bd28a5a95020e0922314d27/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the rises were set: on these figures, for example, April 2025's rise is 4.4%, while the rise paid was 4.1%. The figures are in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/26fbc89f7958c09291dcd33ec7fda8dcb067737c/data/triple_lock_actual_inputs.csv). In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. @@ -123,7 +123,7 @@ The saving depends on how often inflation or 2.5% exceeds earnings growth after In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 3.0%, the same as the Triple Lock: the rise that returns the pension to its floor, rounded up to 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension is £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the saving is £0.5 billion. -**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. The saving in 2039-40 is £10.8 billion, and the full new State Pension is £829 a year lower than under the Triple Lock. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and is not a forecast. +**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. The saving in 2039-40 is £10.9 billion, and the full new State Pension is £836 a year lower than under the Triple Lock. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and is not a forecast. ```chart { @@ -136,14 +136,14 @@ In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neith { "key": "assumption", "name": "OBR long-term assumption", "color": "primary" } ], "data": [ - { "year": "2032-33", "central": 0.33, "assumption": 0.81 }, - { "year": "2033-34", "central": 0.35, "assumption": 1.72 }, - { "year": "2034-35", "central": 0.37, "assumption": 3.09 }, - { "year": "2035-36", "central": 0.39, "assumption": 4.24 }, - { "year": "2036-37", "central": 0.41, "assumption": 5.71 }, - { "year": "2037-38", "central": 0.43, "assumption": 7.32 }, - { "year": "2038-39", "central": 0.45, "assumption": 8.87 }, - { "year": "2039-40", "central": 0.48, "assumption": 10.78 } + { "year": "2032-33", "central": 0.33, "assumption": 0.82 }, + { "year": "2033-34", "central": 0.35, "assumption": 1.73 }, + { "year": "2034-35", "central": 0.37, "assumption": 3.12 }, + { "year": "2035-36", "central": 0.39, "assumption": 4.28 }, + { "year": "2036-37", "central": 0.41, "assumption": 5.76 }, + { "year": "2037-38", "central": 0.43, "assumption": 7.38 }, + { "year": "2038-39", "central": 0.45, "assumption": 8.95 }, + { "year": "2039-40", "central": 0.48, "assumption": 10.87 } ], "valueLabels": false, "yLabel": "Saving", @@ -154,7 +154,7 @@ In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neith ### From gross to net -With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.8 billion. Lower income tax receipts offset £2.4 billion of that, higher Pension Credit £0.8 billion, and higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, £0.3 billion, leaving a net saving of £7.3 billion. +With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.9 billion. Lower income tax receipts offset £2.4 billion of that, higher Pension Credit £0.8 billion, and higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. ```chart { @@ -169,7 +169,7 @@ With a lower State Pension, pensioners pay less income tax and receive more mean { "key": "net", "name": "Net", "kind": "level", "legend": "Saving", "color": "primary" } ], "data": [ - { "measure": "2039-40", "gross": 10.776172900359711, "income_tax": -2.355861154190279, "benefits": -1.1459951874199987, "net": 7.274316558749433 } + { "measure": "2039-40", "gross": 10.870795836943188, "income_tax": -2.3877533883894557, "benefits": -1.1437914913336442, "net": 7.339250957220088 } ], "format": { "decimals": 1, "prefix": "£", "suffix": "bn" }, "source": "Source: PolicyEngine UK 2.151.0 on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. Pension: fall in State Pension spending. Benefits: Pension Credit (£0.8bn), and Housing Benefit, council tax reduction, Winter Fuel Payment, Universal Credit and other spending and taxes (£0.3bn). Labels rounded; steps computed from unrounded figures." @@ -178,7 +178,7 @@ With a lower State Pension, pensioners pay less income tax and receive more mean ## Against the official estimates -DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our closest figures are the gross savings in Great Britain: £0.5 billion on the OBR's central path and £10.5 billion under its long-term assumption. +DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our closest figures are the gross savings in Great Britain: £0.5 billion on the OBR's central path and £10.6 billion under its long-term assumption. The estimates differ in three ways, and DWP's published assumptions do not let us size them ([HL3774](https://questions-statements.parliament.uk/written-questions/detail/2026-10-05/hl3774)): @@ -194,7 +194,7 @@ The OBR gives no figure for 2039-40. Instead, it replays both rules on prices an ## Who has less income -Under the OBR's long-term assumption in 2039-40, 30% of UK households have less income than under the Triple Lock, by £795 a year on average. Households with someone over State Pension age make up 34% of all households, and those with nobody over State Pension age see no change on average. Pensioner couples have £1,062 a year less on average, 1.3% of their income, and single pensioners £387 less, 0.8%. By income decile, households in the third decile lose the largest share of their income, 0.6%. Fewer than 1 in 300 households have more income. On the OBR's central path, 30% of households also lose, by £35 a year on average. +Under the OBR's long-term assumption in 2039-40, 30% of UK households have less income than under the Triple Lock, by £801 a year on average. Households with someone over State Pension age make up 34% of all households, and those with nobody over State Pension age see no change on average. Pensioner couples have £1,072 a year less on average, 1.3% of their income, and single pensioners £389 less, 0.8%. By income decile, households in the third decile lose the largest share of their income, 0.6%. Fewer than 1 in 300 households have more income. On the OBR's central path, 30% of households also lose, by £35 a year on average. ```chart { @@ -205,12 +205,12 @@ Under the OBR's long-term assumption in 2039-40, 30% of UK households have less "series": [{ "key": "change", "name": "Change in household income", "color": "primary" }], "data": [ { "decile": "1st", "change": -0.35 }, - { "decile": "2nd", "change": -0.47 }, - { "decile": "3rd", "change": -0.55 }, + { "decile": "2nd", "change": -0.46 }, + { "decile": "3rd", "change": -0.56 }, { "decile": "4th", "change": -0.52 }, { "decile": "5th", "change": -0.48 }, - { "decile": "6th", "change": -0.35 }, - { "decile": "7th", "change": -0.32 }, + { "decile": "6th", "change": -0.36 }, + { "decile": "7th", "change": -0.33 }, { "decile": "8th", "change": -0.22 }, { "decile": "9th", "change": -0.15 }, { "decile": "10th", "change": -0.06 } @@ -222,7 +222,7 @@ Under the OBR's long-term assumption in 2039-40, 30% of UK households have less } ``` -Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.7% to 11.5% on the relative measure and from 6.8% to 7.7% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative line is 60% of the median person's household income after housing costs, adjusted for household size, and lower pensions pull that median down. +Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.6% to 11.4% on the relative measure and from 6.7% to 7.4% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative line is 60% of the median person's household income after housing costs, adjusted for household size, and lower pensions pull that median down. ## How uncertain the saving is From 1735181bad115b1dee754e579ebf09480bb5e674 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 13:09:52 -0400 Subject: [PATCH 06/17] =?UTF-8?q?Triple=20Lock=20post:=20carry=20=C2=A310.?= =?UTF-8?q?9=20billion=20into=20the=20description?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/posts.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/app/src/data/posts/posts.json b/app/src/data/posts/posts.json index 8548c1787..5a284c122 100644 --- a/app/src/data/posts/posts.json +++ b/app/src/data/posts/posts.json @@ -64,7 +64,7 @@ }, { "title": "The adjusted Triple Lock in the Autumn Budget 2026", - "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. We explain how the Triple Lock works and what the adjusted rule saves: \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.8 billion if the Triple Lock outpaces earnings as the OBR assumes.", + "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. We explain how the Triple Lock works and what the adjusted rule saves: \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.9 billion if the Triple Lock outpaces earnings as the OBR assumes.", "date": "2026-10-28 17:00:00", "tags": ["uk", "policy", "autumn-budget"], "authors": ["max-ghenis"], From eeefc868a319efbb84a7d3f4b5cc5d778a0baa7e Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 13:28:19 -0400 Subject: [PATCH 07/17] Triple Lock post: match the dashboard's final release The dashboard shows four scenario cards once uk-triple-lock#49 merges, so the post now covers the two earnings scenarios and the three single paths, per Max's ruling on d2212. The top note says the results file feeds the dashboard, the old-name footnote goes, and the methodology dates the long-term row from April 2032. Co-Authored-By: Claude Opus 5.5 --- .../posts/articles/uk-triple-lock-budget-2026.md | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 5d12f5c44..4d8aef8f8 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,10 +1,10 @@ -> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/26fbc89f7958c09291dcd33ec7fda8dcb067737c/data/results.json). Our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) still shows its pre-Budget release, which used the OBR's March 2026 forecast, PolicyEngine UK 2.125.1 and the Enhanced Family Resources Survey 2024-25. +> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/26fbc89f7958c09291dcd33ec7fda8dcb067737c/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. > > We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. Each April, the State Pension rises by the highest of inflation, earnings growth and 2.5%. This rule is the Triple Lock. The [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026 keeps it until April 2029 and adjusts it from April 2030. Earnings growth stops being a minimum rise, and the pension instead keeps its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." -The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this adjusted Triple Lock cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These are cash figures for Great Britain and cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)).[^name] +The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this adjusted Triple Lock cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These are cash figures for Great Britain and cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): @@ -232,6 +232,8 @@ To average over the possible years ahead, we drew 50,000 paths from a statistica So we report the two scenarios, and they are not a range: the saving could fall outside them. +Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Each is the same kind of figure as the average we withdrew, from the same model, and we estimate it from full runs on 160 paths weighted to stand for all 50,000. The failed test does not show whether these averages overstate or understate the saving, and they are not forecasts. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40. +
Methodology and limitations @@ -241,11 +243,11 @@ Both rules follow the Triple Lock through April 2029. Each April, the basic and ### Scenarios -The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants. From April 2034, it pays what that row describes as average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. +The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants from April 2032. From April 2034, it pays what that row describes as average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. The replay of April 2017 to April 2026 uses September CPI and May to July earnings growth as published by ONS in October 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. -The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test registered on 6 October scored it on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022), and set limits on coverage and bias. It failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). +The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test registered on 6 October scored it on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022), and set limits on coverage and bias. It failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. ### Model and data @@ -268,8 +270,6 @@ We reweight the households each year after 2025-26 so that each age and sex grou
-[^name]: Our dashboard's pre-Budget release calls the policy by its earlier name, the Burnham plan. - --- _Analysis by PolicyEngine. The code, results and replication steps are on [GitHub](https://github.com/PolicyEngine/uk-triple-lock)._ From 8de2d24272d80ae6ddf4c8e9875ca21af20c67fb Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 16:41:20 -0400 Subject: [PATCH 08/17] Pin the Triple Lock post's links to #49's candidate final head 0b47c87 rebuilds the results file with new wording only; the post's figures all match it. Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 4d8aef8f8..2ce71c807 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,4 +1,4 @@ -> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/26fbc89f7958c09291dcd33ec7fda8dcb067737c/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. +> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. > > We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. @@ -62,7 +62,7 @@ To see the two rules at work, we replay today's rules on the ONS figures publish | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the rises were set: on these figures, for example, April 2025's rise is 4.4%, while the rise paid was 4.1%. The figures are in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/26fbc89f7958c09291dcd33ec7fda8dcb067737c/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the rises were set: on these figures, for example, April 2025's rise is 4.4%, while the rise paid was 4.1%. The figures are in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/triple_lock_actual_inputs.csv). In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. From 5ae7f0ca5b198faf3d698c930f8a93ba4350af29 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 20:45:37 -0400 Subject: [PATCH 09/17] Say the Triple Lock scenarios rest on 160 draws, and run the voice pass The binding sample holds 160 draws of 159 distinct paths, so "160 paths" overstated the paths by one. "160 draws" states the sample the estimator uses and still matches the dashboard's count. The figures do not move. The rest applies the full voice gate to the post and its description: be-verb comparisons become comparing verbs, reduced passives name their actor, and the description drops a clause about the post itself. Co-Authored-By: Claude Opus 5.5 --- .../articles/uk-triple-lock-budget-2026.md | 50 +++++++++---------- app/src/data/posts/posts.json | 2 +- 2 files changed, 26 insertions(+), 26 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 2ce71c807..701ffd729 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,18 +1,18 @@ -> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. The figures are in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. +> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. We publish the figures in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. > > We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. -Each April, the State Pension rises by the highest of inflation, earnings growth and 2.5%. This rule is the Triple Lock. The [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026 keeps it until April 2029 and adjusts it from April 2030. Earnings growth stops being a minimum rise, and the pension instead keeps its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." +Each April, the Triple Lock raises the State Pension by the highest of inflation, earnings growth and 2.5%. The [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026 keeps it until April 2029 and adjusts it from April 2030. Earnings growth stops being a minimum rise, and the pension instead keeps its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." -The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this adjusted Triple Lock cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These are cash figures for Great Britain and cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). +The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this adjusted Triple Lock cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These cash figures for Great Britain cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): - The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.5 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.9 billion. -- After the lower pension reduces income tax and raises means-tested benefits, the two savings are £0.3 billion and £7.3 billion. -- For Great Britain, the area DWP covers, our two figures are £0.5 billion and £10.6 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. -- Under the OBR's long-term assumption, the full new State Pension is £836 a year lower than under the Triple Lock, and 30% of UK households have less income. -- Had the adjusted Triple Lock started in April 2017, the full new State Pension would be £642 a year lower in 2026-27, 5.1% below the amount paid. +- After the lower pension reduces income tax and raises means-tested benefits, the two savings fall to £0.3 billion and £7.3 billion. +- For Great Britain, the area DWP covers, we estimate £0.5 billion and £10.6 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. +- Under the OBR's long-term assumption, the full new State Pension falls £836 a year short of the Triple Lock's, and 30% of UK households have less income. +- Had the adjusted Triple Lock started in April 2017, the full new State Pension would fall £642 a year short of the amount paid in 2026-27, or 5.1%. [Explore the dashboard](https://www.policyengine.org/uk/triple-lock) | [Read the code](https://github.com/PolicyEngine/uk-triple-lock) @@ -47,7 +47,7 @@ The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/governmen ### Both rules over the last ten years -To see the two rules at work, we replay today's rules on the ONS figures published in October 2026 for April 2017 to April 2026. We assume the adjusted Triple Lock started in April 2017 and protected the pension's April 2016 value relative to earnings. +To see the two rules at work, we replay today's rules on the figures ONS published in October 2026 for April 2017 to April 2026. We assume the adjusted Triple Lock started in April 2017 and protected the pension's April 2016 value relative to earnings. | April | Inflation | Earnings growth | Triple Lock | Adjusted Triple Lock | | ----- | --------- | --------------- | ----------- | -------------------- | @@ -62,7 +62,7 @@ To see the two rules at work, we replay today's rules on the ONS figures publish | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the rises were set: on these figures, for example, April 2025's rise is 4.4%, while the rise paid was 4.1%. The figures are in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the government set the rises: on these figures, for example, April 2025's rise comes to 4.4%, against the 4.1% paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/triple_lock_actual_inputs.csv). In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. @@ -103,11 +103,11 @@ In April 2020, earnings growth of 4.0% would have lifted the floor above a pensi } ``` -By April 2026, the full new State Pension would be £11,906 a year under the adjusted Triple Lock, against the £12,548 paid: £642, or 5.1%, lower. Running PolicyEngine UK with that lower pension puts the saving at £6.5 billion in 2026-27, or £4.6 billion after the lower pension reduces income tax and raises means-tested benefits. +By April 2026, the full new State Pension would reach £11,906 a year under the adjusted Triple Lock, against the £12,548 paid: £642, or 5.1%, less. Running PolicyEngine UK with that lower pension puts the saving at £6.5 billion in 2026-27, or £4.6 billion after the lower pension reduces income tax and raises means-tested benefits. ## What it saves -The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We report two scenarios. Neither is a forecast of the most likely outcome. +The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We report two scenarios. Neither forecasts the most likely outcome. **The OBR's central path.** This takes the OBR's October 2026 forecast to 2030 and its March 2026 long-term projections after: @@ -121,9 +121,9 @@ The saving depends on how often inflation or 2.5% exceeds earnings growth after | 2033 | 2.0% | 3.0% | 3.0% | 3.0% | | 2034 | 2.0% | 3.3% | 3.3% | 3.3% | -In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 3.0%, the same as the Triple Lock: the rise that returns the pension to its floor, rounded up to 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension is £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the saving is £0.5 billion. +In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 3.0%, the same as the Triple Lock: the rise that returns the pension to its floor, rounded up to 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the adjusted Triple Lock saves £0.5 billion. -**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. The saving in 2039-40 is £10.9 billion, and the full new State Pension is £836 a year lower than under the Triple Lock. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and is not a forecast. +**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. In 2039-40, the adjusted Triple Lock saves £10.9 billion, and the full new State Pension falls £836 a year short of the Triple Lock's. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and makes no forecast. ```chart { @@ -178,7 +178,7 @@ With a lower State Pension, pensioners pay less income tax and receive more mean ## Against the official estimates -DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our closest figures are the gross savings in Great Britain: £0.5 billion on the OBR's central path and £10.6 billion under its long-term assumption. +DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our gross savings in Great Britain come closest: £0.5 billion on the OBR's central path and £10.6 billion under its long-term assumption. The estimates differ in three ways, and DWP's published assumptions do not let us size them ([HL3774](https://questions-statements.parliament.uk/written-questions/detail/2026-10-05/hl3774)): @@ -222,36 +222,36 @@ Under the OBR's long-term assumption in 2039-40, 30% of UK households have less } ``` -Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.6% to 11.4% on the relative measure and from 6.7% to 7.4% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative line is 60% of the median person's household income after housing costs, adjusted for household size, and lower pensions pull that median down. +Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.6% to 11.4% on the relative measure and from 6.7% to 7.4% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative measure sets its line at 60% of the median person's household income after housing costs, adjusted for household size, and lower pensions pull that median down. ## How uncertain the saving is The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete, and committed, if the model failed, to reporting scenarios only. There were six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. Scored on all seven on 28 October, the model failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete, and committed, if the model failed, to reporting scenarios only. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. -So we report the two scenarios, and they are not a range: the saving could fall outside them. +So we report the two scenarios, and the saving could fall outside them. -Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Each is the same kind of figure as the average we withdrew, from the same model, and we estimate it from full runs on 160 paths weighted to stand for all 50,000. The failed test does not show whether these averages overstate or understate the saving, and they are not forecasts. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40. +Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Like the average we withdrew, each comes from the same model, and we estimate it from full runs on 160 draws weighted to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and neither forecasts it. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40.
Methodology and limitations ### The rules -Both rules follow the Triple Lock through April 2029. Each April, the basic and new State Pension rise by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, with each input taken to 0.1 percentage point. From April 2030, the adjusted Triple Lock raises them by the higher of September CPI and 2.5%, or by more where needed to keep the pension at its 2029-30 level grown by earnings since. For the forward scenarios, we round that top-up up to 0.1 percentage point, which is our assumption: the Budget and DWP do not specify a rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the additional State Pension rises with September CPI and the Pension Credit guarantee with May to July earnings. +Both rules follow the Triple Lock through April 2029. Each April, the basic and new State Pension rise by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, with each input taken to 0.1 percentage point. From April 2030, the adjusted Triple Lock raises them by the higher of September CPI and 2.5%, or by more where needed to keep the pension at its 2029-30 level grown by earnings since. For the forward scenarios, we assume that top-up rounds up to 0.1 percentage point, since the Budget and DWP specify no rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the additional State Pension rises with September CPI and the Pension Credit guarantee with May to July earnings. ### Scenarios The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants from April 2032. From April 2034, it pays what that row describes as average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. -The replay of April 2017 to April 2026 uses September CPI and May to July earnings growth as published by ONS in October 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. +The replay of April 2017 to April 2026 uses the September CPI and May to July earnings growth figures ONS published in October 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. -The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test registered on 6 October scored it on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022), and set limits on coverage and bias. It failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. +The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test we registered on 6 October scored it on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022), and set limits on coverage and bias. It failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. ### Model and data -Every fiscal and household figure is a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households, which starts from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow it. The net saving is the change in the government's balance across every tax and benefit the model covers. +Every fiscal and household figure comes from a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households, which starts from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow it. The net saving is the change in the government's balance across every tax and benefit the model covers. ### Population @@ -259,17 +259,17 @@ We reweight the households each year after 2025-26 so that each age and sex grou ### Limitations -- The 2039-40 population is today's households, reweighted. The model does not simulate deaths, migration or contribution records, so a pensioner's record in 2039-40 is that of someone the same age today. +- We reweight today's households to the 2039-40 population. The model does not simulate deaths, migration or contribution records, so a pensioner in 2039-40 carries the record of someone the same age today. - Housing Benefit and council tax reduction respond only for households already claiming them, so nobody the adjusted Triple Lock makes newly eligible starts claiming them. That understates those offsets and so overstates the net saving. Pension Credit can go to newly eligible households, when the take-up draw stored in the data says they would claim it. - The Savings Credit threshold and the pension-age Housing Benefit allowances, which the model also uses for council tax reduction, rise with CPI in the model, while the Pension Credit guarantee rises with earnings. In law, maximum Savings Credit is 60% of the gap between the guarantee and the threshold, and DWP describes protected Housing Benefit allowances as the guarantee plus maximum Savings Credit. The guarantee's link to earnings does not set these amounts' future paths. Projecting them from the guarantee instead of CPI is an alternative assumption that could change the Pension Credit, Housing Benefit and council tax reduction offsets under both rules. - The OBR publishes no May to July earnings forecast, so the Pension Credit guarantee's rises from April 2028 to April 2031 follow its April to June figure, as the Triple Lock's earnings input does. - The model splits each reported State Pension into a flat-rate part, up to the full basic or new rate, and anything above it, which rises with CPI under both rules. Where a pension falls short of the full rate, the model scales all of it with the flat rate, including any Graduated Retirement Benefit or extra pension for deferring. In law those parts rise with prices: in April 2025, they rose 1.7% while the basic State Pension rose 4.1% ([DWP](https://www.gov.uk/government/publications/benefit-and-pension-rates-2025-to-2026/benefit-and-pension-rates-2025-to-2026)). They would rise the same under both rules, so scaling them overstates the saving. - Rents and council tax stay at their 2030 levels after 2030, and dividend, property, savings and self-employment income do not follow the scenario. - Nobody changes how much they work, save or claim in response to the adjusted Triple Lock. -- Costs are for the UK in cash terms; comparisons with DWP use our figures for Great Britain. +- We state costs for the UK in cash terms; comparisons with DWP use our figures for Great Britain.
--- -_Analysis by PolicyEngine. The code, results and replication steps are on [GitHub](https://github.com/PolicyEngine/uk-triple-lock)._ +_Analysis by PolicyEngine. We publish the code, results and replication steps on [GitHub](https://github.com/PolicyEngine/uk-triple-lock)._ diff --git a/app/src/data/posts/posts.json b/app/src/data/posts/posts.json index 5a284c122..053f23ef6 100644 --- a/app/src/data/posts/posts.json +++ b/app/src/data/posts/posts.json @@ -64,7 +64,7 @@ }, { "title": "The adjusted Triple Lock in the Autumn Budget 2026", - "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. We explain how the Triple Lock works and what the adjusted rule saves: \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.9 billion if the Triple Lock outpaces earnings as the OBR assumes.", + "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. The adjusted Triple Lock saves \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.9 billion if the Triple Lock outpaces earnings as the OBR assumes.", "date": "2026-10-28 17:00:00", "tags": ["uk", "policy", "autumn-budget"], "authors": ["max-ghenis"], From dacb82d95a6758558abd558ab8c858eafde0da11 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 20:46:25 -0400 Subject: [PATCH 10/17] Split three long sentences in the Triple Lock post Name the model where "it" could have meant the withdrawn average, and keep the voice gate's sentence lengths. The figures do not move. Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 701ffd729..f9da9265e 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -154,7 +154,7 @@ In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neith ### From gross to net -With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.9 billion. Lower income tax receipts offset £2.4 billion of that, higher Pension Credit £0.8 billion, and higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. +With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.9 billion. Lower income tax receipts offset £2.4 billion of that and higher Pension Credit £0.8 billion. Higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, offset £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. ```chart { @@ -228,7 +228,7 @@ Under the same scenario, the share of pensioners in poverty after housing costs The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete, and committed, if the model failed, to reporting scenarios only. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. If the model failed, we committed to reporting scenarios only. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. So we report the two scenarios, and the saving could fall outside them. @@ -247,7 +247,7 @@ The central path takes the OBR's [October 2026 Economic and fiscal outlook](http The replay of April 2017 to April 2026 uses the September CPI and May to July earnings growth figures ONS published in October 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. -The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test we registered on 6 October scored it on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022), and set limits on coverage and bias. It failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. +The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test we registered on 6 October scored the model on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022). It set limits on coverage and bias, and the model failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. ### Model and data From 12578a0e713a3a27c153205c4cf47ec49cdd32b3 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 21:00:41 -0400 Subject: [PATCH 11/17] Apply the review's voice and clarity findings to the Triple Lock post Restore the "no range" point, name the model that failed the test, and fix the pronouns and reduced passives the review listed. The OBR's own row phases in the extra 0.6 points, so the post credits the OBR. The figures do not move. Co-Authored-By: Claude Opus 5.5 --- .../articles/uk-triple-lock-budget-2026.md | 52 +++++++++---------- 1 file changed, 26 insertions(+), 26 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index f9da9265e..7b7d6ec65 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -12,7 +12,7 @@ Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): - After the lower pension reduces income tax and raises means-tested benefits, the two savings fall to £0.3 billion and £7.3 billion. - For Great Britain, the area DWP covers, we estimate £0.5 billion and £10.6 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. - Under the OBR's long-term assumption, the full new State Pension falls £836 a year short of the Triple Lock's, and 30% of UK households have less income. -- Had the adjusted Triple Lock started in April 2017, the full new State Pension would fall £642 a year short of the amount paid in 2026-27, or 5.1%. +- Had the adjusted Triple Lock started in April 2017, the full new State Pension would fall £642 a year short of what the government pays in 2026-27, a gap of 5.1%. [Explore the dashboard](https://www.policyengine.org/uk/triple-lock) | [Read the code](https://github.com/PolicyEngine/uk-triple-lock) @@ -26,7 +26,7 @@ Each April, the Triple Lock raises the basic and new State Pension by the highes | Earnings growth | Average weekly earnings over May to July, against the same months a year before | 4.8% | | Floor | A fixed minimum | 2.5% | -In April 2026, earnings growth was the highest of the three, so the full new State Pension rose 4.8%, to £12,548 a year. +In April 2026, earnings growth exceeded the other two, so the full new State Pension rose 4.8%, to £12,548 a year. Because the rise always matches or exceeds earnings growth, the pension never falls behind earnings. In a year when inflation or 2.5% exceeds earnings growth, it rises faster than earnings, and every later rise builds on that higher level. The Budget calls this the "ratchet" effect, "whereby the State Pension rises ever higher relative to earnings." @@ -41,9 +41,9 @@ Because the rise always matches or exceeds earnings growth, the pension never fa The adjusted Triple Lock keeps inflation and 2.5% as minimum rises. It replaces earnings growth with a floor: each year's rise must keep the pension at or above its April 2029 value relative to earnings. -In a year when inflation or 2.5% exceeds earnings growth, both rules pay it, and the pension rises above its floor. The rules part in the years after. When earnings growth is highest, the Triple Lock pays it, while the adjusted Triple Lock pays the higher of inflation and 2.5%. If that would leave the pension below its floor, it pays just enough to reach the floor. Once the pension is back at its floor, it rises with earnings, or as the Budget puts it, "the State Pension will track earnings over time." +In a year when inflation or 2.5% exceeds earnings growth, both rules pay it, and the pension rises above its floor. The rules part in the years after. When earnings growth comes out highest, the Triple Lock pays it, while the adjusted Triple Lock pays the higher of inflation and 2.5%. If that rise would leave the pension below its floor, the adjusted Triple Lock pays just enough to reach the floor. Once the pension returns to its floor, it rises with earnings, or as the Budget puts it, "the State Pension will track earnings over time." -The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), updated 6 October, do not name the CPI month or the earnings measure. We assume the adjusted Triple Lock uses the ones the Triple Lock uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). +The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), which DWP updated on 6 October, do not name the CPI month or the earnings measure. We assume the adjusted Triple Lock uses the ones the Triple Lock uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). ### Both rules over the last ten years @@ -62,19 +62,19 @@ To see the two rules at work, we replay today's rules on the figures ONS publish | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. Some rises differ from those paid, because ONS has revised its earnings figures since the government set the rises: on these figures, for example, April 2025's rise comes to 4.4%, against the 4.1% paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so the replay produces some different rates. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/triple_lock_actual_inputs.csv). In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. -In April 2019, 2024, 2025 and 2026, earnings growth was highest. The Triple Lock paid it, on top of the earlier gains. The adjusted Triple Lock paid the higher of inflation and 2.5%, because the pension already sat above its floor. +In April 2019, 2024, 2025 and 2026, earnings growth came out highest, so the Triple Lock paid it on top of the earlier gains. The adjusted Triple Lock paid the higher of inflation and 2.5%, because the pension already sat above its floor. In April 2020, earnings growth of 4.0% would have lifted the floor above a pension rising 2.5%. The adjusted Triple Lock paid 3.4%, just enough to bring the pension back to its floor. ```chart { "type": "line", - "title": "The State Pension under each rule, replayed for April 2017 to April 2026", - "subtitle": "Index, April 2016 = 100. The floor is the pension's April 2016 value grown with earnings since.", + "title": "The State Pension under each rule, April 2017 to April 2026", + "subtitle": "Index, April 2016 = 100. We grow the pension's April 2016 value with earnings to set the floor.", "xKey": "year", "xLabel": "April", "series": [ @@ -99,15 +99,15 @@ In April 2020, earnings growth of 4.0% would have lifted the floor above a pensi "yLabel": "Index", "format": { "decimals": 0 }, "xFormat": { "decimals": 0 }, - "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, ONS figures as published in October 2026 (PolicyEngine/uk-triple-lock). April 2022 uses inflation for earnings under both rules." + "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, figures ONS published in October 2026 (PolicyEngine/uk-triple-lock). April 2022 uses inflation for earnings under both rules." } ``` -By April 2026, the full new State Pension would reach £11,906 a year under the adjusted Triple Lock, against the £12,548 paid: £642, or 5.1%, less. Running PolicyEngine UK with that lower pension puts the saving at £6.5 billion in 2026-27, or £4.6 billion after the lower pension reduces income tax and raises means-tested benefits. +By April 2026, the full new State Pension would reach £11,906 a year under the adjusted Triple Lock, against the £12,548 the government paid: £642, or 5.1%, less. Running PolicyEngine UK with that lower pension puts the saving at £6.5 billion in 2026-27, or £4.6 billion after the lower pension reduces income tax and raises means-tested benefits. ## What it saves -The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We report two scenarios. Neither forecasts the most likely outcome. +The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We report two scenarios, and neither forecasts the most likely outcome. **The OBR's central path.** This takes the OBR's October 2026 forecast to 2030 and its March 2026 long-term projections after: @@ -121,9 +121,9 @@ The saving depends on how often inflation or 2.5% exceeds earnings growth after | 2033 | 2.0% | 3.0% | 3.0% | 3.0% | | 2034 | 2.0% | 3.3% | 3.3% | 3.3% | -In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, it pays 3.0%, the same as the Triple Lock: the rise that returns the pension to its floor, rounded up to 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the adjusted Triple Lock saves £0.5 billion. +In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, the adjusted Triple Lock pays 3.0%, matching the Triple Lock. We round the rise needed to reach the floor upwards in steps of 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the adjusted Triple Lock saves £0.5 billion. -**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, phased in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. In 2039-40, the adjusted Triple Lock saves £10.9 billion, and the full new State Pension falls £836 a year short of the Triple Lock's. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and makes no forecast. +**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, which the OBR phases in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. In 2039-40, the adjusted Triple Lock saves £10.9 billion, and the full new State Pension falls £836 a year short of the Triple Lock's. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and makes no forecast. ```chart { @@ -154,7 +154,7 @@ In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neith ### From gross to net -With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.9 billion. Lower income tax receipts offset £2.4 billion of that and higher Pension Credit £0.8 billion. Higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, offset £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. +With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.9 billion, while lower income tax receipts offset £2.4 billion of that and higher Pension Credit £0.8 billion. Higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, offset £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. ```chart { @@ -172,7 +172,7 @@ With a lower State Pension, pensioners pay less income tax and receive more mean { "measure": "2039-40", "gross": 10.870795836943188, "income_tax": -2.3877533883894557, "benefits": -1.1437914913336442, "net": 7.339250957220088 } ], "format": { "decimals": 1, "prefix": "£", "suffix": "bn" }, - "source": "Source: PolicyEngine UK 2.151.0 on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. Pension: fall in State Pension spending. Benefits: Pension Credit (£0.8bn), and Housing Benefit, council tax reduction, Winter Fuel Payment, Universal Credit and other spending and taxes (£0.3bn). Labels rounded; steps computed from unrounded figures." + "source": "Source: PolicyEngine UK 2.151.0 on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. Pension: fall in State Pension spending. Benefits: Pension Credit (£0.8bn), and Housing Benefit, council tax reduction, Winter Fuel Payment, Universal Credit and other spending and taxes (£0.3bn). We round the labels and calculate the steps from unrounded figures." } ``` @@ -228,30 +228,30 @@ Under the same scenario, the share of pensioners in poverty after housing costs The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. If the model failed, we committed to reporting scenarios only. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. -So we report the two scenarios, and the saving could fall outside them. +So we report the two scenarios. They set no range, and the saving could fall outside them. -Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Like the average we withdrew, each comes from the same model, and we estimate it from full runs on 160 draws weighted to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and neither forecasts it. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40. +Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Both averages come from the model that failed the test, like the average we withdrew. We estimate each from full runs on 160 draws and weight the results to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and we do not use either average as a forecast. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40.
Methodology and limitations ### The rules -Both rules follow the Triple Lock through April 2029. Each April, the basic and new State Pension rise by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, with each input taken to 0.1 percentage point. From April 2030, the adjusted Triple Lock raises them by the higher of September CPI and 2.5%, or by more where needed to keep the pension at its 2029-30 level grown by earnings since. For the forward scenarios, we assume that top-up rounds up to 0.1 percentage point, since the Budget and DWP specify no rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the additional State Pension rises with September CPI and the Pension Credit guarantee with May to July earnings. +Both rules follow the Triple Lock through April 2029. Each April, the basic and new State Pension rise by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, and we round each input to 0.1 percentage point. From April 2030, the adjusted Triple Lock raises them by the higher of September CPI and 2.5%, and adds any top-up needed to preserve the pension's 2029-30 value relative to earnings. For the forward scenarios, we round that top-up upwards in steps of 0.1 percentage point; the Budget and DWP specify no rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the additional State Pension rises with September CPI and the Pension Credit guarantee with May to July earnings. ### Scenarios -The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants from April 2032. From April 2034, it pays what that row describes as average earnings growth plus 0.6 percentage points, phased in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. +The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants from April 2032. From April 2034, the Triple Lock pays what that row describes as average earnings growth plus 0.6 percentage points, which the row phases in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. -The replay of April 2017 to April 2026 uses the September CPI and May to July earnings growth figures ONS published in October 2026, with April 2022's earnings figure replaced by CPI under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. +The replay of April 2017 to April 2026 uses the September CPI and May to July earnings growth figures ONS published in October 2026, and substitutes CPI for April 2022's earnings figure under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. -The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths sampled to represent the 50,000. The test we registered on 6 October scored the model on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022). It set limits on coverage and bias, and the model failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. +The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths we sampled to represent the 50,000. The test we registered on 6 October scored the model on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022). The test set limits on coverage and bias, and the model failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. ### Model and data -Every fiscal and household figure comes from a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households, which starts from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow it. The net saving is the change in the government's balance across every tax and benefit the model covers. +Every fiscal and household figure comes from a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households, which starts from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow the scenario. The net saving is the change in the government's balance across every tax and benefit the model covers. ### Population @@ -260,13 +260,13 @@ We reweight the households each year after 2025-26 so that each age and sex grou ### Limitations - We reweight today's households to the 2039-40 population. The model does not simulate deaths, migration or contribution records, so a pensioner in 2039-40 carries the record of someone the same age today. -- Housing Benefit and council tax reduction respond only for households already claiming them, so nobody the adjusted Triple Lock makes newly eligible starts claiming them. That understates those offsets and so overstates the net saving. Pension Credit can go to newly eligible households, when the take-up draw stored in the data says they would claim it. -- The Savings Credit threshold and the pension-age Housing Benefit allowances, which the model also uses for council tax reduction, rise with CPI in the model, while the Pension Credit guarantee rises with earnings. In law, maximum Savings Credit is 60% of the gap between the guarantee and the threshold, and DWP describes protected Housing Benefit allowances as the guarantee plus maximum Savings Credit. The guarantee's link to earnings does not set these amounts' future paths. Projecting them from the guarantee instead of CPI is an alternative assumption that could change the Pension Credit, Housing Benefit and council tax reduction offsets under both rules. +- Housing Benefit and council tax reduction respond only for households already claiming them, so nobody the adjusted Triple Lock makes newly eligible starts claiming them. That understates those offsets and so overstates the net saving. Pension Credit can go to newly eligible households, when their take-up draw in the data indicates they would claim it. +- The Savings Credit threshold and the pension-age Housing Benefit allowances, which the model also uses for council tax reduction, rise with CPI in the model, while the Pension Credit guarantee rises with earnings. In law, maximum Savings Credit is 60% of the gap between the guarantee and the threshold, and DWP describes protected Housing Benefit allowances as the guarantee plus maximum Savings Credit. The guarantee's link to earnings does not set these amounts' future paths. Projecting these amounts from the guarantee instead of CPI could change the Pension Credit, Housing Benefit and council tax reduction offsets under both rules. - The OBR publishes no May to July earnings forecast, so the Pension Credit guarantee's rises from April 2028 to April 2031 follow its April to June figure, as the Triple Lock's earnings input does. - The model splits each reported State Pension into a flat-rate part, up to the full basic or new rate, and anything above it, which rises with CPI under both rules. Where a pension falls short of the full rate, the model scales all of it with the flat rate, including any Graduated Retirement Benefit or extra pension for deferring. In law those parts rise with prices: in April 2025, they rose 1.7% while the basic State Pension rose 4.1% ([DWP](https://www.gov.uk/government/publications/benefit-and-pension-rates-2025-to-2026/benefit-and-pension-rates-2025-to-2026)). They would rise the same under both rules, so scaling them overstates the saving. - Rents and council tax stay at their 2030 levels after 2030, and dividend, property, savings and self-employment income do not follow the scenario. - Nobody changes how much they work, save or claim in response to the adjusted Triple Lock. -- We state costs for the UK in cash terms; comparisons with DWP use our figures for Great Britain. +- We calculate costs for the UK in cash terms; comparisons with DWP use our figures for Great Britain.
From b882289d25a63f7a7c5eefdd86436ff60820be18 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 21:17:30 -0400 Subject: [PATCH 12/17] Say the model passed the test's second set of forecasts The registered test uses two sets, and a model must pass on both. The post named only the seven March forecasts, while the dashboard now names both. The second set's 13 forecasts run from June 2010 to March 2022, and the model passed every check on them. No figure moves. Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 7b7d6ec65..a98834cdf 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -228,7 +228,7 @@ Under the same scenario, the share of pensioners in poverty after housing costs The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set, the 13 OBR forecasts from June 2010 to March 2022. So we report the two scenarios. They set no range, and the saving could fall outside them. From 88c09407e771097cd321294d749e1c7912a994aa Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 21:30:57 -0400 Subject: [PATCH 13/17] Say the second set holds one OBR forecast a year The test keeps one spring forecast a year, so "the 13 OBR forecasts" read as if the OBR made only 13 in that period. No figure moves. Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index a98834cdf..b815c8b62 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -228,7 +228,7 @@ Under the same scenario, the share of pensioners in poverty after housing costs The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set, the 13 OBR forecasts from June 2010 to March 2022. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set: 13 OBR forecasts, one a year, from June 2010 to March 2022. So we report the two scenarios. They set no range, and the saving could fall outside them. From 3cbb48062b8b803f25597c41768e4c584957eb2d Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Fri, 30 Oct 2026 21:55:20 -0400 Subject: [PATCH 14/17] Pin the post's links to the dashboard's final head Both links move from 0b47c87 to 9d99213. The results and inputs files keep their blobs, so no figure or word changes. Co-Authored-By: Claude Opus 5.5 --- app/src/data/posts/articles/uk-triple-lock-budget-2026.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index b815c8b62..99dbc008e 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,4 +1,4 @@ -> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. We publish the figures in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. +> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. We publish the figures in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. > > We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. @@ -62,7 +62,7 @@ To see the two rules at work, we replay today's rules on the figures ONS publish | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so the replay produces some different rates. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/0b47c87c451cb85c4e3b97d0804e0024fcaa1737/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so the replay produces some different rates. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/triple_lock_actual_inputs.csv). In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. From f6f2ad804b47c69989eafeb69acdbfc29cd6cb87 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Sat, 31 Oct 2026 07:39:49 -0400 Subject: [PATCH 15/17] Name the actors in the Triple Lock post Make the government, Parliament and the Work and Pensions Secretary the subjects where they act, and "we" in the replay and scenario set-up. Add what section 150A requires, what the government adds, how the rise reaches law, and that the adjusted Triple Lock needs legislation. Name gross State Pension spending where the rule-as-subject wording said "saves". Tidy a loose "it" and a "So" after a pass. No figure changes. Co-Authored-By: Claude Opus 5.5 --- .../articles/uk-triple-lock-budget-2026.md | 42 ++++++++++--------- app/src/data/posts/posts.json | 2 +- 2 files changed, 24 insertions(+), 20 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 99dbc008e..72ebb75ed 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -2,13 +2,13 @@ > > We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. -Each April, the Triple Lock raises the State Pension by the highest of inflation, earnings growth and 2.5%. The [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026 keeps it until April 2029 and adjusts it from April 2030. Earnings growth stops being a minimum rise, and the pension instead keeps its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." +Each April, the government raises the State Pension by the highest of inflation, earnings growth and 2.5%, under a commitment it calls the Triple Lock. In the [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026, the government kept that commitment until April 2029 and said it would adjust it from April 2030. It would drop earnings growth as a minimum rise and instead hold the pension at its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." -The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this adjusted Triple Lock cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These cash figures for Great Britain cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). +The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this change cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These cash figures for Great Britain cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): -- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.5 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.9 billion. +- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock cuts what the government spends on the State Pension by £0.5 billion. If the pension would otherwise rise 0.6 percentage points a year faster than earnings under the Triple Lock, as the OBR assumes in its long-term projections, the cut reaches £10.9 billion. - After the lower pension reduces income tax and raises means-tested benefits, the two savings fall to £0.3 billion and £7.3 billion. - For Great Britain, the area DWP covers, we estimate £0.5 billion and £10.6 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. - Under the OBR's long-term assumption, the full new State Pension falls £836 a year short of the Triple Lock's, and 30% of UK households have less income. @@ -18,7 +18,7 @@ Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): ## How the Triple Lock works -Each April, the Triple Lock raises the basic and new State Pension by the highest of three figures: +Each April, the government raises the basic and new State Pension by the highest of three figures: | Figure | What it measures | For the April 2026 rise | | --------------- | ------------------------------------------------------------------------------- | ----------------------- | @@ -26,7 +26,9 @@ Each April, the Triple Lock raises the basic and new State Pension by the highes | Earnings growth | Average weekly earnings over May to July, against the same months a year before | 4.8% | | Floor | A fixed minimum | 2.5% | -In April 2026, earnings growth exceeded the other two, so the full new State Pension rose 4.8%, to £12,548 a year. +In April 2026, earnings growth exceeded the other two, so the government raised the full new State Pension 4.8%, to £12,548 a year. + +Parliament requires less in law. Under [section 150A of the Social Security Administration Act 1992](https://www.legislation.gov.uk/ukpga/1992/5/section/150A), the Work and Pensions Secretary must review the pension against earnings each year. If earnings have grown, the Secretary must lay a draft order before Parliament that raises it by at least as much. The government has [committed](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) to add the inflation and 2.5% minimums for this Parliament. Once both Houses approve the draft order, the Work and Pensions Secretary makes it, and the new rates start in April. In Northern Ireland, the Department for Communities makes a [corresponding order](https://www.legislation.gov.uk/nisr/2026/59/contents/made). Because the rise always matches or exceeds earnings growth, the pension never falls behind earnings. In a year when inflation or 2.5% exceeds earnings growth, it rises faster than earnings, and every later rise builds on that higher level. The Budget calls this the "ratchet" effect, "whereby the State Pension rises ever higher relative to earnings." @@ -39,11 +41,13 @@ Because the rise always matches or exceeds earnings growth, the pension never fa | Earnings growth | Minimum rise | Not a minimum rise | | Floor on the pension's level | None | Its April 2029 value relative to earnings | -The adjusted Triple Lock keeps inflation and 2.5% as minimum rises. It replaces earnings growth with a floor: each year's rise must keep the pension at or above its April 2029 value relative to earnings. +Under the adjusted Triple Lock, the government would keep inflation and 2.5% as minimum rises and replace earnings growth with a floor: each year's rise must keep the pension at or above its April 2029 value relative to earnings. + +In a year when inflation or 2.5% exceeds earnings growth, the government raises the pension by that figure under both rules, and the pension moves above its floor. The rules part in the years after. When earnings growth comes out highest, the government raises the pension by that figure under the Triple Lock, but only by the higher of inflation and 2.5% under the adjusted Triple Lock. If that rise would leave the pension below its floor, the government adds just enough to reach the floor. Once the pension returns to its floor, it rises with earnings, or as the Budget puts it, "the State Pension will track earnings over time." -In a year when inflation or 2.5% exceeds earnings growth, both rules pay it, and the pension rises above its floor. The rules part in the years after. When earnings growth comes out highest, the Triple Lock pays it, while the adjusted Triple Lock pays the higher of inflation and 2.5%. If that rise would leave the pension below its floor, the adjusted Triple Lock pays just enough to reach the floor. Once the pension returns to its floor, it rises with earnings, or as the Budget puts it, "the State Pension will track earnings over time." +Parliament would have to change the law first. In a year when earnings growth comes out highest and the pension sits above its floor, the government would raise the pension by less than section 150A requires. The government has [said it will legislate](https://www.gov.uk/government/news/prime-minister-andy-burnham-sets-out-plans-for-a-new-national-care-service) during this Parliament, but by 31 October it had not introduced a Bill. The next general election [must come before April 2030](https://fullfact.org/government-tracker/adjusted-triple-lock/), so the next Parliament would approve the first rise under the adjusted Triple Lock. -The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), which DWP updated on 6 October, do not name the CPI month or the earnings measure. We assume the adjusted Triple Lock uses the ones the Triple Lock uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). +The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), which DWP updated on 6 October, do not name the CPI month or the earnings measure. We assume the government would use the ones it uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). ### Both rules over the last ten years @@ -62,13 +66,13 @@ To see the two rules at work, we replay today's rules on the figures ONS publish | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so the replay produces some different rates. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so we use inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so the replay produces some different rates. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/triple_lock_actual_inputs.csv). -In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. +In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the government raised the pension 2.5%. The two rules give the same rise in those years, and each rise lifted the pension relative to earnings. -In April 2019, 2024, 2025 and 2026, earnings growth came out highest, so the Triple Lock paid it on top of the earlier gains. The adjusted Triple Lock paid the higher of inflation and 2.5%, because the pension already sat above its floor. +In April 2019, 2024, 2025 and 2026, earnings growth came out highest. Under the Triple Lock, we raise the pension by that figure, on top of the earlier gains. Under the adjusted Triple Lock, we raise it by the higher of inflation and 2.5%, because the pension already sat above its floor. -In April 2020, earnings growth of 4.0% would have lifted the floor above a pension rising 2.5%. The adjusted Triple Lock paid 3.4%, just enough to bring the pension back to its floor. +In April 2020, earnings growth of 4.0% would have lifted the floor above a pension rising 2.5%. Under the adjusted Triple Lock, we raise the pension 3.4%, just enough to bring it back to its floor. ```chart { @@ -121,9 +125,9 @@ The saving depends on how often inflation or 2.5% exceeds earnings growth after | 2033 | 2.0% | 3.0% | 3.0% | 3.0% | | 2034 | 2.0% | 3.3% | 3.3% | 3.3% | -In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, the adjusted Triple Lock pays 3.0%, matching the Triple Lock. We round the rise needed to reach the floor upwards in steps of 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the adjusted Triple Lock saves £0.5 billion. +In April 2028, inflation of 3.4% comes out highest, and the government raises the pension by that figure under both rules. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so the government raises it 2.5% in each of those years under both rules. The rises of April 2028 and April 2029 set the 2029-30 level that the government would protect under the adjusted Triple Lock, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the government raises the pension 2.5% under the adjusted Triple Lock, against 2.7% under the Triple Lock. In April 2033, the rise comes to 3.0% under both rules. We round the rise needed to reach the floor upwards in steps of 0.1 percentage point. From April 2034, the rise matches earnings growth under both rules. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against £18,876 under the Triple Lock, and the government spends £0.5 billion less on the State Pension. -**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, which the OBR phases in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. In 2039-40, the adjusted Triple Lock saves £10.9 billion, and the full new State Pension falls £836 a year short of the Triple Lock's. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and makes no forecast. +**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the pension rises 0.6 percentage points a year faster than earnings on average under the Triple Lock, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, we use that assumption for the Triple Lock, with the OBR's phase-in over the rises of April 2032 and April 2033, and the central path for the adjusted Triple Lock. In 2039-40, the government spends £10.9 billion less on the State Pension, and the full new State Pension falls £836 a year short of the Triple Lock's. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the government would raise the pension by that figure under the adjusted Triple Lock too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and makes no forecast. ```chart { @@ -154,7 +158,7 @@ In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neith ### From gross to net -With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.9 billion, while lower income tax receipts offset £2.4 billion of that and higher Pension Credit £0.8 billion. Higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, offset £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. +With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, the government spends £10.9 billion less on the State Pension, collects £2.4 billion less income tax and spends £0.8 billion more on Pension Credit. Higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, cost it another £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. ```chart { @@ -228,9 +232,9 @@ Under the same scenario, the share of pensioners in poverty after housing costs The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set: 13 OBR forecasts, one a year, from June 2010 to March 2022. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored the model on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set: 13 OBR forecasts, one a year, from June 2010 to March 2022. -So we report the two scenarios. They set no range, and the saving could fall outside them. +Because the model failed, we report the two scenarios. They set no range, and the saving could fall outside them. Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Both averages come from the model that failed the test, like the average we withdrew. We estimate each from full runs on 160 draws and weight the results to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and we do not use either average as a forecast. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40. @@ -239,11 +243,11 @@ Our dashboard also shows two scenarios from the model's paths, with calendar-yea ### The rules -Both rules follow the Triple Lock through April 2029. Each April, the basic and new State Pension rise by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, and we round each input to 0.1 percentage point. From April 2030, the adjusted Triple Lock raises them by the higher of September CPI and 2.5%, and adds any top-up needed to preserve the pension's 2029-30 value relative to earnings. For the forward scenarios, we round that top-up upwards in steps of 0.1 percentage point; the Budget and DWP specify no rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the additional State Pension rises with September CPI and the Pension Credit guarantee with May to July earnings. +Both rules follow the Triple Lock through April 2029. Each April, the government raises the basic and new State Pension by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, and we round each input to 0.1 percentage point. From April 2030, under the adjusted Triple Lock, it would raise them by the higher of September CPI and 2.5%, and add any top-up needed to preserve the pension's 2029-30 value relative to earnings. For the forward scenarios, we round that top-up upwards in steps of 0.1 percentage point; the Budget and DWP specify no rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the government raises the additional State Pension with September CPI and the Pension Credit guarantee with May to July earnings. ### Scenarios -The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants from April 2032. From April 2034, the Triple Lock pays what that row describes as average earnings growth plus 0.6 percentage points, which the row phases in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. +The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, we take Triple Lock rises from the "Triple lock" row of those determinants from April 2032. From April 2034, that row gives what it describes as average earnings growth plus 0.6 percentage points, which it phases in over the two rises before. We use the central path for the adjusted Triple Lock in both scenarios. The replay of April 2017 to April 2026 uses the September CPI and May to July earnings growth figures ONS published in October 2026, and substitutes CPI for April 2022's earnings figure under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. diff --git a/app/src/data/posts/posts.json b/app/src/data/posts/posts.json index 053f23ef6..b3259186d 100644 --- a/app/src/data/posts/posts.json +++ b/app/src/data/posts/posts.json @@ -64,7 +64,7 @@ }, { "title": "The adjusted Triple Lock in the Autumn Budget 2026", - "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. The adjusted Triple Lock saves \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.9 billion if the Triple Lock outpaces earnings as the OBR assumes.", + "description": "From April 2030, the government plans to raise the State Pension by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. That cuts State Pension spending by \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.9 billion if the Triple Lock outpaces earnings as the OBR assumes.", "date": "2026-10-28 17:00:00", "tags": ["uk", "policy", "autumn-budget"], "authors": ["max-ghenis"], From 149463382978ab93aed916b68b01769ac5f36373 Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Sat, 31 Oct 2026 08:02:25 -0400 Subject: [PATCH 16/17] Apply the law, meaning and voice review of the actor rewrite Cut the claim about which Parliament approves the first adjusted rise, which the election deadline does not establish. Give the replay to "we", in the present tense, because it is our hypothetical run. Move model and method sentences to "we" and "the model", introduce ONS, and keep gross figures as State Pension spending. No figure changes. Co-Authored-By: Claude Opus 5.5 --- .../articles/uk-triple-lock-budget-2026.md | 80 +++++++++---------- app/src/data/posts/posts.json | 2 +- 2 files changed, 41 insertions(+), 41 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 72ebb75ed..251353e4c 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,16 +1,16 @@ -> **Updated 30 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. We publish the figures in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. +> **Updated 30 October 2026.** We now calculate every PolicyEngine figure below from the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. We publish the figures in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. > -> We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. +> We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it and now report scenarios. We do not describe any of the model's paths as typical or likely. Each April, the government raises the State Pension by the highest of inflation, earnings growth and 2.5%, under a commitment it calls the Triple Lock. In the [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026, the government kept that commitment until April 2029 and said it would adjust it from April 2030. It would drop earnings growth as a minimum rise and instead hold the pension at its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." -The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this change cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These cash figures for Great Britain cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). +The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that the government would spend £15 billion less on the State Pension in 2039-40 and £50 billion less in 2049-50. These cash figures for Great Britain cover direct spending only, before any effect on tax or debt interest. The government says it will use the savings to help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): -- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock cuts what the government spends on the State Pension by £0.5 billion. If the pension would otherwise rise 0.6 percentage points a year faster than earnings under the Triple Lock, as the OBR assumes in its long-term projections, the cut reaches £10.9 billion. -- After the lower pension reduces income tax and raises means-tested benefits, the two savings fall to £0.3 billion and £7.3 billion. -- For Great Britain, the area DWP covers, we estimate £0.5 billion and £10.6 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. +- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the government spends £0.5 billion less on the State Pension under the adjusted Triple Lock. If the pension would otherwise rise 0.6 percentage points a year faster than earnings under the Triple Lock, as the OBR assumes in its long-term projections, the cut reaches £10.9 billion. +- With a lower pension, pensioners pay less income tax and receive more means-tested benefits, reducing the government's savings to £0.3 billion and £7.3 billion. +- For Great Britain, the area DWP covers, we estimate £0.5 billion and £10.6 billion, against DWP's £15 billion. In our data, we count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. - Under the OBR's long-term assumption, the full new State Pension falls £836 a year short of the Triple Lock's, and 30% of UK households have less income. - Had the adjusted Triple Lock started in April 2017, the full new State Pension would fall £642 a year short of what the government pays in 2026-27, a gap of 5.1%. @@ -28,9 +28,9 @@ Each April, the government raises the basic and new State Pension by the highest In April 2026, earnings growth exceeded the other two, so the government raised the full new State Pension 4.8%, to £12,548 a year. -Parliament requires less in law. Under [section 150A of the Social Security Administration Act 1992](https://www.legislation.gov.uk/ukpga/1992/5/section/150A), the Work and Pensions Secretary must review the pension against earnings each year. If earnings have grown, the Secretary must lay a draft order before Parliament that raises it by at least as much. The government has [committed](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) to add the inflation and 2.5% minimums for this Parliament. Once both Houses approve the draft order, the Work and Pensions Secretary makes it, and the new rates start in April. In Northern Ireland, the Department for Communities makes a [corresponding order](https://www.legislation.gov.uk/nisr/2026/59/contents/made). +Parliament requires less in law. Under [section 150A of the Social Security Administration Act 1992](https://www.legislation.gov.uk/ukpga/1992/5/section/150A), the Work and Pensions Secretary must review the pension against earnings each year. If earnings have grown, the Secretary must lay a draft order before Parliament to raise the pension by at least as much. The government has [committed](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) to add the inflation and 2.5% minimums for this Parliament. Once both Houses approve the draft order, the Secretary makes it, and the new rates start in April. In Northern Ireland, the Department for Communities makes a [corresponding order](https://www.legislation.gov.uk/nisr/2026/59/contents/made). -Because the rise always matches or exceeds earnings growth, the pension never falls behind earnings. In a year when inflation or 2.5% exceeds earnings growth, it rises faster than earnings, and every later rise builds on that higher level. The Budget calls this the "ratchet" effect, "whereby the State Pension rises ever higher relative to earnings." +Because the rise always matches or exceeds earnings growth, the pension never falls behind earnings. In a year when inflation or 2.5% exceeds earnings growth, the government raises the pension faster than earnings and applies every later rise to that higher level. The Budget calls this the "ratchet" effect, "whereby the State Pension rises ever higher relative to earnings." ## What changes in April 2030 @@ -45,13 +45,13 @@ Under the adjusted Triple Lock, the government would keep inflation and 2.5% as In a year when inflation or 2.5% exceeds earnings growth, the government raises the pension by that figure under both rules, and the pension moves above its floor. The rules part in the years after. When earnings growth comes out highest, the government raises the pension by that figure under the Triple Lock, but only by the higher of inflation and 2.5% under the adjusted Triple Lock. If that rise would leave the pension below its floor, the government adds just enough to reach the floor. Once the pension returns to its floor, it rises with earnings, or as the Budget puts it, "the State Pension will track earnings over time." -Parliament would have to change the law first. In a year when earnings growth comes out highest and the pension sits above its floor, the government would raise the pension by less than section 150A requires. The government has [said it will legislate](https://www.gov.uk/government/news/prime-minister-andy-burnham-sets-out-plans-for-a-new-national-care-service) during this Parliament, but by 31 October it had not introduced a Bill. The next general election [must come before April 2030](https://fullfact.org/government-tracker/adjusted-triple-lock/), so the next Parliament would approve the first rise under the adjusted Triple Lock. +Parliament would have to change the law first. In a year when earnings growth comes out highest and the pension sits above its floor, the government would raise the pension by less than section 150A requires. The government has [said it will legislate](https://www.gov.uk/government/news/prime-minister-andy-burnham-sets-out-plans-for-a-new-national-care-service) during this Parliament, but by 31 October it had not introduced a Bill. The government plans the first adjusted rise for April 2030, after the [deadline for the next general election](https://fullfact.org/government-tracker/adjusted-triple-lock/). -The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), which DWP updated on 6 October, do not name the CPI month or the earnings measure. We assume the government would use the ones it uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). +The government does not specify the CPI month or the earnings measure in the Budget or DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), which DWP updated on 6 October. We assume the government would use the ones it uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). ### Both rules over the last ten years -To see the two rules at work, we replay today's rules on the figures ONS published in October 2026 for April 2017 to April 2026. We assume the adjusted Triple Lock started in April 2017 and protected the pension's April 2016 value relative to earnings. +We replay today's rules on the figures the Office for National Statistics (ONS) published in October 2026 for April 2017 to April 2026. We assume the government used the adjusted Triple Lock from April 2017 and protected the pension's April 2016 value relative to earnings. | April | Inflation | Earnings growth | Triple Lock | Adjusted Triple Lock | | ----- | --------- | --------------- | ----------- | -------------------- | @@ -66,13 +66,13 @@ To see the two rules at work, we replay today's rules on the figures ONS publish | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so we use inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so the replay produces some different rates. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so we use inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so we calculate some different rates in the replay. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/triple_lock_actual_inputs.csv). -In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the government raised the pension 2.5%. The two rules give the same rise in those years, and each rise lifted the pension relative to earnings. +In April 2018 and April 2023, inflation exceeds earnings growth in our replay, and in April 2021 earnings fall while we raise the pension 2.5%. The two rules give the same rise in those years, and each rise lifts the pension relative to earnings. -In April 2019, 2024, 2025 and 2026, earnings growth came out highest. Under the Triple Lock, we raise the pension by that figure, on top of the earlier gains. Under the adjusted Triple Lock, we raise it by the higher of inflation and 2.5%, because the pension already sat above its floor. +In April 2019, 2024, 2025 and 2026, earnings growth comes out highest. Under the Triple Lock, we raise the pension by that figure, on top of the earlier gains. Under the adjusted Triple Lock, we raise it by the higher of inflation and 2.5%, because the pension already sits above its floor. -In April 2020, earnings growth of 4.0% would have lifted the floor above a pension rising 2.5%. Under the adjusted Triple Lock, we raise the pension 3.4%, just enough to bring it back to its floor. +In April 2020, earnings growth of 4.0% would lift the floor above a pension rising 2.5%. Under the adjusted Triple Lock, we raise the pension 3.4%, just enough to bring it back to its floor. ```chart { @@ -103,17 +103,17 @@ In April 2020, earnings growth of 4.0% would have lifted the floor above a pensi "yLabel": "Index", "format": { "decimals": 0 }, "xFormat": { "decimals": 0 }, - "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, figures ONS published in October 2026 (PolicyEngine/uk-triple-lock). April 2022 uses inflation for earnings under both rules." + "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, figures ONS published in October 2026 (PolicyEngine/uk-triple-lock). For April 2022, we use inflation for earnings under both rules." } ``` -By April 2026, the full new State Pension would reach £11,906 a year under the adjusted Triple Lock, against the £12,548 the government paid: £642, or 5.1%, less. Running PolicyEngine UK with that lower pension puts the saving at £6.5 billion in 2026-27, or £4.6 billion after the lower pension reduces income tax and raises means-tested benefits. +By April 2026, the full new State Pension would reach £11,906 a year under the adjusted Triple Lock, against the £12,548 the government paid: £642, or 5.1%, less. We run PolicyEngine UK with that lower pension and estimate a saving of £6.5 billion in 2026-27, or £4.6 billion after pensioners pay less income tax and receive more means-tested benefits. ## What it saves -The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We report two scenarios, and neither forecasts the most likely outcome. +The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We calculate savings under two scenarios without forecasting which outcome is most likely. -**The OBR's central path.** This takes the OBR's October 2026 forecast to 2030 and its March 2026 long-term projections after: +We build the **OBR's central path** from its October 2026 forecast to 2030 and its March 2026 long-term projections for later years: | April | Inflation | Earnings growth | Triple Lock | Adjusted Triple Lock | | ----- | --------- | --------------- | ----------- | -------------------- | @@ -125,9 +125,9 @@ The saving depends on how often inflation or 2.5% exceeds earnings growth after | 2033 | 2.0% | 3.0% | 3.0% | 3.0% | | 2034 | 2.0% | 3.3% | 3.3% | 3.3% | -In April 2028, inflation of 3.4% comes out highest, and the government raises the pension by that figure under both rules. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so the government raises it 2.5% in each of those years under both rules. The rises of April 2028 and April 2029 set the 2029-30 level that the government would protect under the adjusted Triple Lock, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the government raises the pension 2.5% under the adjusted Triple Lock, against 2.7% under the Triple Lock. In April 2033, the rise comes to 3.0% under both rules. We round the rise needed to reach the floor upwards in steps of 0.1 percentage point. From April 2034, the rise matches earnings growth under both rules. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against £18,876 under the Triple Lock, and the government spends £0.5 billion less on the State Pension. +In April 2028, inflation of 3.4% comes out highest, so the government applies that rise under both rules. Neither inflation nor earnings growth exceeds 2.5% in the inputs for April 2029 to April 2031, so the government raises the pension 2.5% each year under both rules. With the rises of April 2028 and April 2029, the government sets the 2029-30 level it would protect under the adjusted Triple Lock. In April 2030, the government applies a rise of 2.5% against earnings growth of 2.3%, taking the pension above its floor. In April 2032, the government increases the pension 2.5% under the adjusted Triple Lock, against 2.7% under the Triple Lock. In April 2033, the rise comes to 3.0% under both rules. We round the rise needed to reach the floor upwards in steps of 0.1 percentage point. From April 2034, the rise matches earnings growth under both rules. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against £18,876 under the Triple Lock, and the government spends £0.5 billion less on the State Pension. -**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the pension rises 0.6 percentage points a year faster than earnings on average under the Triple Lock, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, we use that assumption for the Triple Lock, with the OBR's phase-in over the rises of April 2032 and April 2033, and the central path for the adjusted Triple Lock. In 2039-40, the government spends £10.9 billion less on the State Pension, and the full new State Pension falls £836 a year short of the Triple Lock's. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the government would raise the pension by that figure under the adjusted Triple Lock too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and makes no forecast. +The OBR's **long-term assumption** has the government raising the pension 0.6 percentage points a year faster than earnings on average under the Triple Lock. The OBR makes that stylised allowance for years when inflation or 2.5% comes out highest. We use that assumption for the Triple Lock, with the OBR's phase-in over the rises of April 2032 and April 2033, and the central path for the adjusted Triple Lock. In 2039-40, the government spends £10.9 billion less on the State Pension, and the full new State Pension falls £836 a year short of the Triple Lock's. We add no allowance under the adjusted Triple Lock for years when inflation or 2.5% exceeds earnings growth, although the government would raise the pension by that figure under the adjusted Triple Lock too. We compare the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and make no forecast. ```chart { @@ -152,13 +152,13 @@ In April 2028, inflation of 3.4% comes out highest, and the government raises th "valueLabels": false, "yLabel": "Saving", "format": { "decimals": 1, "prefix": "£", "suffix": "bn" }, - "source": "Source: PolicyEngine UK 2.151.0 (policyengine.py 6.11.1) on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. Neither scenario saves anything before 2032-33." + "source": "Source: PolicyEngine UK 2.151.0 (policyengine.py 6.11.1) on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. We estimate no saving under either scenario before 2032-33." } ``` ### From gross to net -With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, the government spends £10.9 billion less on the State Pension, collects £2.4 billion less income tax and spends £0.8 billion more on Pension Credit. Higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, cost it another £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. +With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, the government spends £10.9 billion less on the State Pension, collects £2.4 billion less income tax and spends £0.8 billion more on Pension Credit. The government also pays more Housing Benefit, council tax reduction and other benefits and collects less in other taxes, reducing its saving by another £0.3 billion to a net £7.3 billion. We round each figure, so the steps do not sum exactly. ```chart { @@ -182,19 +182,19 @@ With a lower State Pension, pensioners pay less income tax and receive more mean ## Against the official estimates -DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our gross savings in Great Britain come closest: £0.5 billion on the OBR's central path and £10.6 billion under its long-term assumption. +DWP estimates savings from the adjusted Triple Lock of £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. DWP covers direct spending in Great Britain and excludes effects on tax and debt interest. It says its analysis "isolates the modelled effect of uprating policy, rather than behavioural responses". Our gross savings in Great Britain come closest: £0.5 billion on the OBR's central path and £10.6 billion under its long-term assumption. The estimates differ in three ways, and DWP's published assumptions do not let us size them ([HL3774](https://questions-statements.parliament.uk/written-questions/detail/2026-10-05/hl3774)): - DWP projects the population and each person's pension with Pensim3, its dynamic microsimulation model. We reweight the households in Microcosm UK, our dataset of UK households, which starts from the Family Resources Survey, to the ONS population projection by age and sex, holding each person's age fixed. -- Our data count slightly less State Pension in payment than DWP does: £144.8 billion in Great Britain in 2026-27, against DWP's £148.3 billion, both excluding payments abroad. The saving scales with the basic and new State Pension, which DWP publishes only with payments abroad included, and only to 2030-31, so we cannot match that part of the bill exactly. -- DWP says it applies "a path for uprating" but publishes no rates for it, and its answer to HL3774 says the analysis used no single-year assumptions for earnings or prices growth. +- In our data, we count slightly less State Pension in payment than DWP does: £144.8 billion in Great Britain in 2026-27, against DWP's £148.3 billion, both excluding payments abroad. The saving scales with the basic and new State Pension, but DWP includes payments abroad in its figures and publishes them only to 2030-31, so we cannot match that part of the bill exactly. +- DWP says it applies "a path for uprating" but publishes no rates for it, and says in its answer to HL3774 that it used no single-year assumptions for earnings or prices growth. Our projections end in 2039-40, so we have no figure to set against the £50 billion. -The OBR's October 2026 Economic and fiscal outlook builds the adjusted Triple Lock into its forecast from 2030-31. It finds that the adjusted Triple Lock "does not currently affect state pension costs over the forecast period", because 2.5% applies under both rules in 2030-31 and 2031-32 (paragraph 3.50 and its footnote 17). Our central path also pays 2.5% under both rules in April 2030 and April 2031, and saves nothing until 2032-33. +In its October 2026 Economic and fiscal outlook, the OBR includes the adjusted Triple Lock in its forecast from 2030-31. The OBR finds that the adjusted Triple Lock "does not currently affect state pension costs over the forecast period", because 2.5% applies under both rules in 2030-31 and 2031-32 (paragraph 3.50 and its footnote 17). On our central path, we apply 2.5% under both rules in April 2030 and April 2031 and estimate no saving until 2032-33. -The OBR gives no figure for 2039-40. Instead, it replays both rules on prices and earnings since 1990 and finds State Pension payments 1.4% lower each year on average than under the Triple Lock (paragraph 3.51). That replay protects the pension's 1990 value relative to earnings and averages over every year since, so its 1.4% does not compare directly with our 5.1% for 2026-27 from a 2017 start. The OBR also estimates, provisionally, that if the next 50 years resembled the last 35, the adjusted Triple Lock would leave debt in 2075-76 around 25% of GDP lower than under the Triple Lock. It notes that the overall effect of the adjusted Triple Lock and the government's planned expansion of social care on the long-term public finances depends on what that expansion costs (paragraphs 3.52 to 3.54). +The OBR gives no figure for 2039-40. Instead, it replays both rules on prices and earnings since 1990 and finds State Pension payments 1.4% lower each year on average than under the Triple Lock (paragraph 3.51). The OBR protects the pension's 1990 value relative to earnings in that replay and averages over every year since, so its 1.4% does not compare directly with our 5.1% for 2026-27 from a 2017 start. The OBR also estimates, provisionally, that if the next 50 years resembled the last 35, the adjusted Triple Lock would leave debt in 2075-76 around 25% of GDP lower than under the Triple Lock. It notes that the overall effect of the adjusted Triple Lock and the government's planned expansion of social care on the long-term public finances depends on what that expansion costs (paragraphs 3.52 to 3.54). ## Who has less income @@ -226,51 +226,51 @@ Under the OBR's long-term assumption in 2039-40, 30% of UK households have less } ``` -Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.6% to 11.4% on the relative measure and from 6.7% to 7.4% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative measure sets its line at 60% of the median person's household income after housing costs, adjusted for household size, and lower pensions pull that median down. +Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.6% to 11.4% on the relative measure and from 6.7% to 7.4% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. We deduct housing costs from household income and adjust for household size, then set the relative poverty line at 60% of the median person's household income. Lower pensions pull that median down. ## How uncertain the saving is The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored the model on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set: 13 OBR forecasts, one a year, from June 2010 to March 2022. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored the model on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years we tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set: 13 OBR forecasts, one a year, from June 2010 to March 2022. Because the model failed, we report the two scenarios. They set no range, and the saving could fall outside them. -Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Both averages come from the model that failed the test, like the average we withdrew. We estimate each from full runs on 160 draws and weight the results to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and we do not use either average as a forecast. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40. +Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. We estimate that State Pension spending falls by an average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Both averages come from the model that failed the test, like the average we withdrew. We estimate each from full runs on 160 draws and weight the results to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and we do not use either average as a forecast. The dashboard also shows three illustrative paths under which the government spends between £11.3 billion and £19.8 billion less on the State Pension in 2039-40.
Methodology and limitations ### The rules -Both rules follow the Triple Lock through April 2029. Each April, the government raises the basic and new State Pension by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, and we round each input to 0.1 percentage point. From April 2030, under the adjusted Triple Lock, it would raise them by the higher of September CPI and 2.5%, and add any top-up needed to preserve the pension's 2029-30 value relative to earnings. For the forward scenarios, we round that top-up upwards in steps of 0.1 percentage point; the Budget and DWP specify no rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the government raises the additional State Pension with September CPI and the Pension Credit guarantee with May to July earnings. +Both rules follow the Triple Lock through April 2029. Each April, the government raises the basic and new State Pension by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, and we round each input to 0.1 percentage point. From April 2030, under the adjusted Triple Lock, the government would raise them by the higher of September CPI and 2.5% and add any top-up needed to preserve the pension's 2029-30 value relative to earnings. For the forward scenarios, we round that top-up upwards in steps of 0.1 percentage point; the government specifies no rounding rule. For the replay of past years, we use the exact top-up and report rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the government raises the additional State Pension with September CPI and the Pension Credit guarantee with May to July earnings. ### Scenarios -The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, we take Triple Lock rises from the "Triple lock" row of those determinants from April 2032. From April 2034, that row gives what it describes as average earnings growth plus 0.6 percentage points, which it phases in over the two rises before. We use the central path for the adjusted Triple Lock in both scenarios. +For the central path, we use the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, we use the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; the OBR describes them as a projection rather than a forecast. In the long-term assumption scenario, we take Triple Lock rises from the "Triple lock" row of those determinants from April 2032. From April 2034, the OBR describes that row as average earnings growth plus 0.6 percentage points, and it phases in the addition over the two rises before. We use the central path for the adjusted Triple Lock in both scenarios. -The replay of April 2017 to April 2026 uses the September CPI and May to July earnings growth figures ONS published in October 2026, and substitutes CPI for April 2022's earnings figure under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. +For the replay of April 2017 to April 2026, we use the September CPI and May to July earnings growth figures ONS published in October 2026 and substitute CPI for April 2022's earnings figure under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. -The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths we sampled to represent the 50,000. The test we registered on 6 October scored the model on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022). The test set limits on coverage and bias, and the model failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. +The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths we sampled to represent the 50,000. The test we registered on 6 October scored the model on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022). The test set limits on coverage and bias, and the model failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We use the same monthly model specification for the two earnings scenarios on our dashboard. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. ### Model and data -Every fiscal and household figure comes from a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households, which starts from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow the scenario. The net saving is the change in the government's balance across every tax and benefit the model covers. +We calculate every fiscal and household figure with a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0). We replace the model's own forecasts with each scenario's CPI and earnings, and the model then moves CPI-uprated benefits, CPI-indexed thresholds and employment income with the scenario. We calculate the net saving as the change in the government's balance across every tax and benefit the model covers. ### Population -We reweight the households each year after 2025-26 so that each age and sex group grows as in the ONS 2024-based population projection. Each person keeps the age they have in the data. State Pension type follows birth cohort: basic for people who reached State Pension age before 6 April 2016, new after. The data exclude pensioners living abroad. +We reweight the households each year after 2025-26 so that each age and sex group grows as in the ONS 2024-based population projection. We hold each person's age fixed at the value in the data. State Pension type follows birth cohort: basic for people who reached State Pension age before 6 April 2016, new after. The data exclude pensioners living abroad. ### Limitations - We reweight today's households to the 2039-40 population. The model does not simulate deaths, migration or contribution records, so a pensioner in 2039-40 carries the record of someone the same age today. -- Housing Benefit and council tax reduction respond only for households already claiming them, so nobody the adjusted Triple Lock makes newly eligible starts claiming them. That understates those offsets and so overstates the net saving. Pension Credit can go to newly eligible households, when their take-up draw in the data indicates they would claim it. +- The model adjusts Housing Benefit and council tax reduction only for households that already claim them, so it assigns no new claims when a lower State Pension makes a household eligible. That understates those offsets and so overstates the net saving. The model calculates Pension Credit for newly eligible households when their take-up draw in the data indicates they would claim it. - The Savings Credit threshold and the pension-age Housing Benefit allowances, which the model also uses for council tax reduction, rise with CPI in the model, while the Pension Credit guarantee rises with earnings. In law, maximum Savings Credit is 60% of the gap between the guarantee and the threshold, and DWP describes protected Housing Benefit allowances as the guarantee plus maximum Savings Credit. The guarantee's link to earnings does not set these amounts' future paths. Projecting these amounts from the guarantee instead of CPI could change the Pension Credit, Housing Benefit and council tax reduction offsets under both rules. - The OBR publishes no May to July earnings forecast, so the Pension Credit guarantee's rises from April 2028 to April 2031 follow its April to June figure, as the Triple Lock's earnings input does. - The model splits each reported State Pension into a flat-rate part, up to the full basic or new rate, and anything above it, which rises with CPI under both rules. Where a pension falls short of the full rate, the model scales all of it with the flat rate, including any Graduated Retirement Benefit or extra pension for deferring. In law those parts rise with prices: in April 2025, they rose 1.7% while the basic State Pension rose 4.1% ([DWP](https://www.gov.uk/government/publications/benefit-and-pension-rates-2025-to-2026/benefit-and-pension-rates-2025-to-2026)). They would rise the same under both rules, so scaling them overstates the saving. -- Rents and council tax stay at their 2030 levels after 2030, and dividend, property, savings and self-employment income do not follow the scenario. +- We hold rents and council tax at their 2030 levels after 2030 and do not adjust dividend, property, savings or self-employment income to the scenario. - Nobody changes how much they work, save or claim in response to the adjusted Triple Lock. -- We calculate costs for the UK in cash terms; comparisons with DWP use our figures for Great Britain. +- We calculate costs for the UK in cash terms and use our figures for Great Britain to compare with DWP.
diff --git a/app/src/data/posts/posts.json b/app/src/data/posts/posts.json index b3259186d..3f23b815a 100644 --- a/app/src/data/posts/posts.json +++ b/app/src/data/posts/posts.json @@ -64,7 +64,7 @@ }, { "title": "The adjusted Triple Lock in the Autumn Budget 2026", - "description": "From April 2030, the government plans to raise the State Pension by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. That cuts State Pension spending by \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.9 billion if the Triple Lock outpaces earnings as the OBR assumes.", + "description": "From April 2030, the government plans to raise the State Pension by the higher of inflation and 2.5%, or by more to hold its April 2029 value relative to earnings. The government would spend \u00a30.5 billion less on the State Pension in 2039-40 on the OBR's central path and \u00a310.9 billion less if the pension outpaces earnings under the Triple Lock, as the OBR assumes.", "date": "2026-10-28 17:00:00", "tags": ["uk", "policy", "autumn-budget"], "authors": ["max-ghenis"], From 6fbf0519555d8fe7caeb45f37211c912eb12059e Mon Sep 17 00:00:00 2001 From: Max Ghenis Date: Sat, 31 Oct 2026 08:08:55 -0400 Subject: [PATCH 17/17] Restore the text approved at 3cbb480 and date the update d2223 approved the wording at 3cbb480, so this PR lands exactly that text. The actor rewrite moves to its own branch, triple-lock-post-actors, for Max's go. Sets the date line to 31 October. Co-Authored-By: Claude Opus 5.5 --- .../articles/uk-triple-lock-budget-2026.md | 92 +++++++++---------- app/src/data/posts/posts.json | 2 +- 2 files changed, 45 insertions(+), 49 deletions(-) diff --git a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md index 251353e4c..ebc2d0640 100644 --- a/app/src/data/posts/articles/uk-triple-lock-budget-2026.md +++ b/app/src/data/posts/articles/uk-triple-lock-budget-2026.md @@ -1,16 +1,16 @@ -> **Updated 30 October 2026.** We now calculate every PolicyEngine figure below from the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. We publish the figures in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. +> **Updated 31 October 2026.** Every PolicyEngine figure below now uses the Budget's inputs: September's inflation figure, the Office for Budget Responsibility's (OBR) October 2026 forecast to 2030, PolicyEngine UK 2.151.0 and the Microcosm UK dataset, as in our other Budget posts. The OBR published no long-term assumptions with its October forecast, so after 2030 we use its March 2026 ones. We publish the figures in [this results file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/results.json), which our [Triple Lock dashboard](https://www.policyengine.org/uk/triple-lock) also draws on. > -> We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it and now report scenarios. We do not describe any of the model's paths as typical or likely. +> We first published this post on 28 October with an estimated average saving across 50,000 simulated paths of prices and earnings, £11.1 billion in 2039-40, and a range around it. The same day, the model behind those paths failed a test we had registered before the Budget ([uk-triple-lock#45](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We withdrew the average and the range around it, and the post reports scenarios instead. We do not describe any of the model's paths as typical or likely. -Each April, the government raises the State Pension by the highest of inflation, earnings growth and 2.5%, under a commitment it calls the Triple Lock. In the [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026, the government kept that commitment until April 2029 and said it would adjust it from April 2030. It would drop earnings growth as a minimum rise and instead hold the pension at its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." +Each April, the Triple Lock raises the State Pension by the highest of inflation, earnings growth and 2.5%. The [Budget](https://www.gov.uk/government/publications/budget-2026-document/budget-2026-html#box-5atriple-lock-adjustment175) on 28 October 2026 keeps it until April 2029 and adjusts it from April 2030. Earnings growth stops being a minimum rise, and the pension instead keeps its April 2029 value relative to earnings. In the Budget's words, from April 2030 "the State Pension rises each year by the highest of 2.5%, inflation, or the amount required to maintain its value relative to earnings as reached in April 2029." -The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that the government would spend £15 billion less on the State Pension in 2039-40 and £50 billion less in 2049-50. These cash figures for Great Britain cover direct spending only, before any effect on tax or debt interest. The government says it will use the savings to help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). +The Department for Work and Pensions (DWP) estimates in its [State Pension uprating analysis](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) that this adjusted Triple Lock cuts State Pension spending by £15 billion in 2039-40 and £50 billion in 2049-50. These cash figures for Great Britain cover direct spending only, before any effect on tax or debt interest. The government says the savings will help build a National Care Service in England that gives older people personal care free at the point of use. Prime Minister Andy Burnham announced the change at Labour's conference on 29 September ([BBC](https://www.bbc.co.uk/news/live/c6x2zrv774gvt?post=asset%3A6edb0175-f1b7-407f-a5b7-96e02dfa03ec#post)). Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): -- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the government spends £0.5 billion less on the State Pension under the adjusted Triple Lock. If the pension would otherwise rise 0.6 percentage points a year faster than earnings under the Triple Lock, as the OBR assumes in its long-term projections, the cut reaches £10.9 billion. -- With a lower pension, pensioners pay less income tax and receive more means-tested benefits, reducing the government's savings to £0.3 billion and £7.3 billion. -- For Great Britain, the area DWP covers, we estimate £0.5 billion and £10.6 billion, against DWP's £15 billion. In our data, we count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. +- The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030. On the OBR's central path, earnings growth comes out highest every year from April 2032, and the adjusted Triple Lock saves £0.5 billion. If the Triple Lock instead raises the pension 0.6 percentage points a year faster than earnings, as the OBR assumes in its long-term projections, it saves £10.9 billion. +- After the lower pension reduces income tax and raises means-tested benefits, the two savings fall to £0.3 billion and £7.3 billion. +- For Great Britain, the area DWP covers, we estimate £0.5 billion and £10.6 billion, against DWP's £15 billion. Our data count 2% less State Pension in payment than DWP does, and DWP does not publish the path of prices and earnings behind its estimate. - Under the OBR's long-term assumption, the full new State Pension falls £836 a year short of the Triple Lock's, and 30% of UK households have less income. - Had the adjusted Triple Lock started in April 2017, the full new State Pension would fall £642 a year short of what the government pays in 2026-27, a gap of 5.1%. @@ -18,7 +18,7 @@ Key findings for 2039-40, the first of the two years DWP costs (UK, cash terms): ## How the Triple Lock works -Each April, the government raises the basic and new State Pension by the highest of three figures: +Each April, the Triple Lock raises the basic and new State Pension by the highest of three figures: | Figure | What it measures | For the April 2026 rise | | --------------- | ------------------------------------------------------------------------------- | ----------------------- | @@ -26,11 +26,9 @@ Each April, the government raises the basic and new State Pension by the highest | Earnings growth | Average weekly earnings over May to July, against the same months a year before | 4.8% | | Floor | A fixed minimum | 2.5% | -In April 2026, earnings growth exceeded the other two, so the government raised the full new State Pension 4.8%, to £12,548 a year. +In April 2026, earnings growth exceeded the other two, so the full new State Pension rose 4.8%, to £12,548 a year. -Parliament requires less in law. Under [section 150A of the Social Security Administration Act 1992](https://www.legislation.gov.uk/ukpga/1992/5/section/150A), the Work and Pensions Secretary must review the pension against earnings each year. If earnings have grown, the Secretary must lay a draft order before Parliament to raise the pension by at least as much. The government has [committed](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/state-pension-uprating) to add the inflation and 2.5% minimums for this Parliament. Once both Houses approve the draft order, the Secretary makes it, and the new rates start in April. In Northern Ireland, the Department for Communities makes a [corresponding order](https://www.legislation.gov.uk/nisr/2026/59/contents/made). - -Because the rise always matches or exceeds earnings growth, the pension never falls behind earnings. In a year when inflation or 2.5% exceeds earnings growth, the government raises the pension faster than earnings and applies every later rise to that higher level. The Budget calls this the "ratchet" effect, "whereby the State Pension rises ever higher relative to earnings." +Because the rise always matches or exceeds earnings growth, the pension never falls behind earnings. In a year when inflation or 2.5% exceeds earnings growth, it rises faster than earnings, and every later rise builds on that higher level. The Budget calls this the "ratchet" effect, "whereby the State Pension rises ever higher relative to earnings." ## What changes in April 2030 @@ -41,17 +39,15 @@ Because the rise always matches or exceeds earnings growth, the pension never fa | Earnings growth | Minimum rise | Not a minimum rise | | Floor on the pension's level | None | Its April 2029 value relative to earnings | -Under the adjusted Triple Lock, the government would keep inflation and 2.5% as minimum rises and replace earnings growth with a floor: each year's rise must keep the pension at or above its April 2029 value relative to earnings. - -In a year when inflation or 2.5% exceeds earnings growth, the government raises the pension by that figure under both rules, and the pension moves above its floor. The rules part in the years after. When earnings growth comes out highest, the government raises the pension by that figure under the Triple Lock, but only by the higher of inflation and 2.5% under the adjusted Triple Lock. If that rise would leave the pension below its floor, the government adds just enough to reach the floor. Once the pension returns to its floor, it rises with earnings, or as the Budget puts it, "the State Pension will track earnings over time." +The adjusted Triple Lock keeps inflation and 2.5% as minimum rises. It replaces earnings growth with a floor: each year's rise must keep the pension at or above its April 2029 value relative to earnings. -Parliament would have to change the law first. In a year when earnings growth comes out highest and the pension sits above its floor, the government would raise the pension by less than section 150A requires. The government has [said it will legislate](https://www.gov.uk/government/news/prime-minister-andy-burnham-sets-out-plans-for-a-new-national-care-service) during this Parliament, but by 31 October it had not introduced a Bill. The government plans the first adjusted rise for April 2030, after the [deadline for the next general election](https://fullfact.org/government-tracker/adjusted-triple-lock/). +In a year when inflation or 2.5% exceeds earnings growth, both rules pay it, and the pension rises above its floor. The rules part in the years after. When earnings growth comes out highest, the Triple Lock pays it, while the adjusted Triple Lock pays the higher of inflation and 2.5%. If that rise would leave the pension below its floor, the adjusted Triple Lock pays just enough to reach the floor. Once the pension returns to its floor, it rises with earnings, or as the Budget puts it, "the State Pension will track earnings over time." -The government does not specify the CPI month or the earnings measure in the Budget or DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), which DWP updated on 6 October. We assume the government would use the ones it uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). +The Budget and DWP's [Triple Lock reform factsheet](https://www.gov.uk/government/publications/state-pension-uprating-analysis-2026/triple-lock-reform-factsheet), which DWP updated on 6 October, do not name the CPI month or the earnings measure. We assume the adjusted Triple Lock uses the ones the Triple Lock uses now: September CPI and May to July growth in average weekly earnings. We also assume that "its value relative to earnings as reached in April 2029" means the pension's level in 2029-30, the year that starts that April. The OBR reads it the same way, as the value "relative to earnings at 2029-30 levels" (paragraph 3.49 of its [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook)). ### Both rules over the last ten years -We replay today's rules on the figures the Office for National Statistics (ONS) published in October 2026 for April 2017 to April 2026. We assume the government used the adjusted Triple Lock from April 2017 and protected the pension's April 2016 value relative to earnings. +To see the two rules at work, we replay today's rules on the figures ONS published in October 2026 for April 2017 to April 2026. We assume the adjusted Triple Lock started in April 2017 and protected the pension's April 2016 value relative to earnings. | April | Inflation | Earnings growth | Triple Lock | Adjusted Triple Lock | | ----- | --------- | --------------- | ----------- | -------------------- | @@ -66,13 +62,13 @@ We replay today's rules on the figures the Office for National Statistics (ONS) | 2025 | 1.7% | 4.4% | 4.4% | 2.5% | | 2026 | 3.8% | 4.8% | 4.8% | 3.8% | -\*Parliament suspended the earnings link for April 2022, so we use inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so we calculate some different rates in the replay. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/triple_lock_actual_inputs.csv). +\*Parliament suspended the earnings link for April 2022, so the replay uses inflation in place of earnings growth that year, for both rules and for the floor. ONS has revised its earnings figures since the government set the rises, so the replay produces some different rates. For example, April 2025's rise comes to 4.4%, against the 4.1% the government paid. We publish the figures in the [replay's input file](https://github.com/PolicyEngine/uk-triple-lock/blob/9d99213d44f4c6b88870e2599f5e8c7e349fa569/data/triple_lock_actual_inputs.csv). -In April 2018 and April 2023, inflation exceeds earnings growth in our replay, and in April 2021 earnings fall while we raise the pension 2.5%. The two rules give the same rise in those years, and each rise lifts the pension relative to earnings. +In April 2018 and April 2023, inflation exceeded earnings growth, and in April 2021 earnings fell while the pension rose 2.5%. Both rules paid the same in those years, and each lifted the pension relative to earnings. -In April 2019, 2024, 2025 and 2026, earnings growth comes out highest. Under the Triple Lock, we raise the pension by that figure, on top of the earlier gains. Under the adjusted Triple Lock, we raise it by the higher of inflation and 2.5%, because the pension already sits above its floor. +In April 2019, 2024, 2025 and 2026, earnings growth came out highest, so the Triple Lock paid it on top of the earlier gains. The adjusted Triple Lock paid the higher of inflation and 2.5%, because the pension already sat above its floor. -In April 2020, earnings growth of 4.0% would lift the floor above a pension rising 2.5%. Under the adjusted Triple Lock, we raise the pension 3.4%, just enough to bring it back to its floor. +In April 2020, earnings growth of 4.0% would have lifted the floor above a pension rising 2.5%. The adjusted Triple Lock paid 3.4%, just enough to bring the pension back to its floor. ```chart { @@ -103,17 +99,17 @@ In April 2020, earnings growth of 4.0% would lift the floor above a pension risi "yLabel": "Index", "format": { "decimals": 0 }, "xFormat": { "decimals": 0 }, - "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, figures ONS published in October 2026 (PolicyEngine/uk-triple-lock). For April 2022, we use inflation for earnings under both rules." + "source": "Source: PolicyEngine replay of today's rules on September CPI and May to July average weekly earnings, figures ONS published in October 2026 (PolicyEngine/uk-triple-lock). April 2022 uses inflation for earnings under both rules." } ``` -By April 2026, the full new State Pension would reach £11,906 a year under the adjusted Triple Lock, against the £12,548 the government paid: £642, or 5.1%, less. We run PolicyEngine UK with that lower pension and estimate a saving of £6.5 billion in 2026-27, or £4.6 billion after pensioners pay less income tax and receive more means-tested benefits. +By April 2026, the full new State Pension would reach £11,906 a year under the adjusted Triple Lock, against the £12,548 the government paid: £642, or 5.1%, less. Running PolicyEngine UK with that lower pension puts the saving at £6.5 billion in 2026-27, or £4.6 billion after the lower pension reduces income tax and raises means-tested benefits. ## What it saves -The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We calculate savings under two scenarios without forecasting which outcome is most likely. +The saving depends on how often inflation or 2.5% exceeds earnings growth after 2030, and nobody can forecast that. We report two scenarios, and neither forecasts the most likely outcome. -We build the **OBR's central path** from its October 2026 forecast to 2030 and its March 2026 long-term projections for later years: +**The OBR's central path.** This takes the OBR's October 2026 forecast to 2030 and its March 2026 long-term projections after: | April | Inflation | Earnings growth | Triple Lock | Adjusted Triple Lock | | ----- | --------- | --------------- | ----------- | -------------------- | @@ -125,9 +121,9 @@ We build the **OBR's central path** from its October 2026 forecast to 2030 and i | 2033 | 2.0% | 3.0% | 3.0% | 3.0% | | 2034 | 2.0% | 3.3% | 3.3% | 3.3% | -In April 2028, inflation of 3.4% comes out highest, so the government applies that rise under both rules. Neither inflation nor earnings growth exceeds 2.5% in the inputs for April 2029 to April 2031, so the government raises the pension 2.5% each year under both rules. With the rises of April 2028 and April 2029, the government sets the 2029-30 level it would protect under the adjusted Triple Lock. In April 2030, the government applies a rise of 2.5% against earnings growth of 2.3%, taking the pension above its floor. In April 2032, the government increases the pension 2.5% under the adjusted Triple Lock, against 2.7% under the Triple Lock. In April 2033, the rise comes to 3.0% under both rules. We round the rise needed to reach the floor upwards in steps of 0.1 percentage point. From April 2034, the rise matches earnings growth under both rules. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against £18,876 under the Triple Lock, and the government spends £0.5 billion less on the State Pension. +In April 2028, inflation of 3.4% comes out highest, and both rules pay it. Neither inflation nor earnings growth exceeds 2.5% in the inputs to the rises of April 2029 to April 2031, so both rules pay 2.5% in each of those years. The rises of April 2028 and April 2029 set the 2029-30 level that the adjusted Triple Lock protects, and the April 2030 rise, 2.5% against earnings growth of 2.3%, lifts the pension above its floor. In April 2032, the adjusted Triple Lock pays 2.5% against the Triple Lock's 2.7%. In April 2033, the adjusted Triple Lock pays 3.0%, matching the Triple Lock. We round the rise needed to reach the floor upwards in steps of 0.1 percentage point. From April 2034, both rules pay earnings growth. In 2039-40, the full new State Pension reaches £18,840 a year under the adjusted Triple Lock, against the Triple Lock's £18,876, and the adjusted Triple Lock saves £0.5 billion. -The OBR's **long-term assumption** has the government raising the pension 0.6 percentage points a year faster than earnings on average under the Triple Lock. The OBR makes that stylised allowance for years when inflation or 2.5% comes out highest. We use that assumption for the Triple Lock, with the OBR's phase-in over the rises of April 2032 and April 2033, and the central path for the adjusted Triple Lock. In 2039-40, the government spends £10.9 billion less on the State Pension, and the full new State Pension falls £836 a year short of the Triple Lock's. We add no allowance under the adjusted Triple Lock for years when inflation or 2.5% exceeds earnings growth, although the government would raise the pension by that figure under the adjusted Triple Lock too. We compare the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and make no forecast. +**The OBR's long-term assumption.** In its long-term projections, the OBR assumes that the Triple Lock raises the pension 0.6 percentage points a year faster than earnings on average, a stylised allowance for the years when inflation or 2.5% comes out highest. In this scenario, the Triple Lock follows that assumption, which the OBR phases in over the rises of April 2032 and April 2033, while the adjusted Triple Lock follows the central path. In 2039-40, the adjusted Triple Lock saves £10.9 billion, and the full new State Pension falls £836 a year short of the Triple Lock's. The scenario gives the adjusted Triple Lock no credit for those same years: whenever inflation or 2.5% exceeds earnings growth, the adjusted Triple Lock pays it too. It compares the adjusted Triple Lock with the OBR's way of costing the Triple Lock, and makes no forecast. ```chart { @@ -152,13 +148,13 @@ The OBR's **long-term assumption** has the government raising the pension 0.6 pe "valueLabels": false, "yLabel": "Saving", "format": { "decimals": 1, "prefix": "£", "suffix": "bn" }, - "source": "Source: PolicyEngine UK 2.151.0 (policyengine.py 6.11.1) on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. We estimate no saving under either scenario before 2032-33." + "source": "Source: PolicyEngine UK 2.151.0 (policyengine.py 6.11.1) on Microcosm UK 2024-25, PolicyEngine/uk-triple-lock. Neither scenario saves anything before 2032-33." } ``` ### From gross to net -With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, the government spends £10.9 billion less on the State Pension, collects £2.4 billion less income tax and spends £0.8 billion more on Pension Credit. The government also pays more Housing Benefit, council tax reduction and other benefits and collects less in other taxes, reducing its saving by another £0.3 billion to a net £7.3 billion. We round each figure, so the steps do not sum exactly. +With a lower State Pension, pensioners pay less income tax and receive more means-tested benefits. Under the OBR's long-term assumption in 2039-40, State Pension spending falls by £10.9 billion, while lower income tax receipts offset £2.4 billion of that and higher Pension Credit £0.8 billion. Higher Housing Benefit, council tax reduction and other benefits, with lower other taxes, offset £0.3 billion, leaving a net saving of £7.3 billion. We round each figure, so the steps do not sum exactly. ```chart { @@ -182,19 +178,19 @@ With a lower State Pension, pensioners pay less income tax and receive more mean ## Against the official estimates -DWP estimates savings from the adjusted Triple Lock of £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. DWP covers direct spending in Great Britain and excludes effects on tax and debt interest. It says its analysis "isolates the modelled effect of uprating policy, rather than behavioural responses". Our gross savings in Great Britain come closest: £0.5 billion on the OBR's central path and £10.6 billion under its long-term assumption. +DWP's analysis costs the adjusted Triple Lock at £15 billion in 2039-40 and £50 billion in 2049-50 in cash terms, or £11 billion and £30 billion in 2025-26 prices. It covers direct spending in Great Britain, excludes effects on tax and debt interest, and "isolates the modelled effect of uprating policy, rather than behavioural responses". Our gross savings in Great Britain come closest: £0.5 billion on the OBR's central path and £10.6 billion under its long-term assumption. The estimates differ in three ways, and DWP's published assumptions do not let us size them ([HL3774](https://questions-statements.parliament.uk/written-questions/detail/2026-10-05/hl3774)): - DWP projects the population and each person's pension with Pensim3, its dynamic microsimulation model. We reweight the households in Microcosm UK, our dataset of UK households, which starts from the Family Resources Survey, to the ONS population projection by age and sex, holding each person's age fixed. -- In our data, we count slightly less State Pension in payment than DWP does: £144.8 billion in Great Britain in 2026-27, against DWP's £148.3 billion, both excluding payments abroad. The saving scales with the basic and new State Pension, but DWP includes payments abroad in its figures and publishes them only to 2030-31, so we cannot match that part of the bill exactly. -- DWP says it applies "a path for uprating" but publishes no rates for it, and says in its answer to HL3774 that it used no single-year assumptions for earnings or prices growth. +- Our data count slightly less State Pension in payment than DWP does: £144.8 billion in Great Britain in 2026-27, against DWP's £148.3 billion, both excluding payments abroad. The saving scales with the basic and new State Pension, which DWP publishes only with payments abroad included, and only to 2030-31, so we cannot match that part of the bill exactly. +- DWP says it applies "a path for uprating" but publishes no rates for it, and its answer to HL3774 says the analysis used no single-year assumptions for earnings or prices growth. Our projections end in 2039-40, so we have no figure to set against the £50 billion. -In its October 2026 Economic and fiscal outlook, the OBR includes the adjusted Triple Lock in its forecast from 2030-31. The OBR finds that the adjusted Triple Lock "does not currently affect state pension costs over the forecast period", because 2.5% applies under both rules in 2030-31 and 2031-32 (paragraph 3.50 and its footnote 17). On our central path, we apply 2.5% under both rules in April 2030 and April 2031 and estimate no saving until 2032-33. +The OBR's October 2026 Economic and fiscal outlook builds the adjusted Triple Lock into its forecast from 2030-31. It finds that the adjusted Triple Lock "does not currently affect state pension costs over the forecast period", because 2.5% applies under both rules in 2030-31 and 2031-32 (paragraph 3.50 and its footnote 17). Our central path also pays 2.5% under both rules in April 2030 and April 2031, and saves nothing until 2032-33. -The OBR gives no figure for 2039-40. Instead, it replays both rules on prices and earnings since 1990 and finds State Pension payments 1.4% lower each year on average than under the Triple Lock (paragraph 3.51). The OBR protects the pension's 1990 value relative to earnings in that replay and averages over every year since, so its 1.4% does not compare directly with our 5.1% for 2026-27 from a 2017 start. The OBR also estimates, provisionally, that if the next 50 years resembled the last 35, the adjusted Triple Lock would leave debt in 2075-76 around 25% of GDP lower than under the Triple Lock. It notes that the overall effect of the adjusted Triple Lock and the government's planned expansion of social care on the long-term public finances depends on what that expansion costs (paragraphs 3.52 to 3.54). +The OBR gives no figure for 2039-40. Instead, it replays both rules on prices and earnings since 1990 and finds State Pension payments 1.4% lower each year on average than under the Triple Lock (paragraph 3.51). That replay protects the pension's 1990 value relative to earnings and averages over every year since, so its 1.4% does not compare directly with our 5.1% for 2026-27 from a 2017 start. The OBR also estimates, provisionally, that if the next 50 years resembled the last 35, the adjusted Triple Lock would leave debt in 2075-76 around 25% of GDP lower than under the Triple Lock. It notes that the overall effect of the adjusted Triple Lock and the government's planned expansion of social care on the long-term public finances depends on what that expansion costs (paragraphs 3.52 to 3.54). ## Who has less income @@ -226,51 +222,51 @@ Under the OBR's long-term assumption in 2039-40, 30% of UK households have less } ``` -Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.6% to 11.4% on the relative measure and from 6.7% to 7.4% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. We deduct housing costs from household income and adjust for household size, then set the relative poverty line at 60% of the median person's household income. Lower pensions pull that median down. +Under the same scenario, the share of pensioners in poverty after housing costs rises from 10.6% to 11.4% on the relative measure and from 6.7% to 7.4% on the absolute measure. Across the whole population, absolute poverty after housing costs rises from 15.9% to 16.1%, while relative poverty falls from 19.3% to 19.1%. The relative measure sets its line at 60% of the median person's household income after housing costs, adjusted for household size, and lower pensions pull that median down. ## How uncertain the saving is The two scenarios differ more than twentyfold because the saving comes from the years in which inflation or 2.5% exceeds earnings growth. The OBR's central path has none after April 2031. The last ten years had three, in April 2018, 2021 and 2023, leaving aside April 2022, for which Parliament suspended the earnings link. -To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored the model on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years we tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set: 13 OBR forecasts, one a year, from June 2010 to March 2022. +To average over the possible years ahead, we drew 50,000 paths from a statistical model of prices and earnings. Before the Budget, we registered a test of that model on every March forecast of the OBR from 2016 whose four years of outcomes were complete. We committed in advance to reporting only scenarios if the model failed. We had six such forecasts when we registered the test, and September's inflation figure completed a seventh, from March 2022. We scored it on all seven on 28 October, and it failed one check. It expected earnings growth to come in below 2.5% in 45% of the years tested, against 18% in practice, beyond the limit we had set. Its other checks passed: its central 80% range contained the outcome in 5 of the 7 final years and 83% of the annual checks. It also passed every check on the test's second set: 13 OBR forecasts, one a year, from June 2010 to March 2022. -Because the model failed, we report the two scenarios. They set no range, and the saving could fall outside them. +So we report the two scenarios. They set no range, and the saving could fall outside them. -Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. We estimate that State Pension spending falls by an average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Both averages come from the model that failed the test, like the average we withdrew. We estimate each from full runs on 160 draws and weight the results to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and we do not use either average as a forecast. The dashboard also shows three illustrative paths under which the government spends between £11.3 billion and £19.8 billion less on the State Pension in 2039-40. +Our dashboard also shows two scenarios from the model's paths, with calendar-year earnings growth targets 0.5 percentage points lower or higher from 2031. State Pension spending falls by an estimated average of £11.4 billion in 2039-40 with lower earnings growth and £11.9 billion with higher, each with a 95% Monte Carlo margin of about £0.5 billion. Both averages come from the model that failed the test, like the average we withdrew. We estimate each from full runs on 160 draws and weight the results to stand for all 50,000 paths. The failed test does not show whether these averages overstate or understate the saving, and we do not use either average as a forecast. The dashboard also shows three single paths as illustrations, which save between £11.3 billion and £19.8 billion in 2039-40.
Methodology and limitations ### The rules -Both rules follow the Triple Lock through April 2029. Each April, the government raises the basic and new State Pension by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, and we round each input to 0.1 percentage point. From April 2030, under the adjusted Triple Lock, the government would raise them by the higher of September CPI and 2.5% and add any top-up needed to preserve the pension's 2029-30 value relative to earnings. For the forward scenarios, we round that top-up upwards in steps of 0.1 percentage point; the government specifies no rounding rule. For the replay of past years, we use the exact top-up and report rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the government raises the additional State Pension with September CPI and the Pension Credit guarantee with May to July earnings. +Both rules follow the Triple Lock through April 2029. Each April, the basic and new State Pension rise by the highest of September CPI, May to July growth in average weekly earnings (total pay) and 2.5%, and we round each input to 0.1 percentage point. From April 2030, the adjusted Triple Lock raises them by the higher of September CPI and 2.5%, and adds any top-up needed to preserve the pension's 2029-30 value relative to earnings. For the forward scenarios, we round that top-up upwards in steps of 0.1 percentage point; the Budget and DWP specify no rounding rule. The replay of past years uses the exact top-up, and its table shows rates to 0.1 percentage point. Only the basic and new State Pension differ between the rules. Under both, the additional State Pension rises with September CPI and the Pension Credit guarantee with May to July earnings. ### Scenarios -For the central path, we use the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, we use the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; the OBR describes them as a projection rather than a forecast. In the long-term assumption scenario, we take Triple Lock rises from the "Triple lock" row of those determinants from April 2032. From April 2034, the OBR describes that row as average earnings growth plus 0.6 percentage points, and it phases in the addition over the two rises before. We use the central path for the adjusted Triple Lock in both scenarios. +The central path takes the OBR's [October 2026 Economic and fiscal outlook](https://www.gov.uk/government/publications/october-2026-economic-and-fiscal-outlook) for 2026 to 2030 and for the inputs to every rise to April 2031. The OBR publishes no forecast of May to July earnings, so for the rises of April 2028 to April 2031 we use its April to June figure, and its September-quarter CPI. After 2030, the path follows the OBR's [March 2026 long-term economic determinants](https://obr.uk/docs/dlm_uploads/Long-term-economic-determinants-March-2026-EFO.xlsx), because the OBR published no new set with its October forecast; it describes them as a projection rather than a forecast. In the long-term assumption scenario, the Triple Lock follows the "Triple lock" row of those determinants from April 2032. From April 2034, the Triple Lock pays what that row describes as average earnings growth plus 0.6 percentage points, which the row phases in over the two rises before. The adjusted Triple Lock follows the central path in both scenarios. -For the replay of April 2017 to April 2026, we use the September CPI and May to July earnings growth figures ONS published in October 2026 and substitute CPI for April 2022's earnings figure under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. +The replay of April 2017 to April 2026 uses the September CPI and May to July earnings growth figures ONS published in October 2026, and substitutes CPI for April 2022's earnings figure under both rules. We scale each person's actual flat-rate pension by the ratio of the two rules' pension levels, and full runs of PolicyEngine UK on Microcosm UK 2024-25 cover 2024-25 to 2026-27. -The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths we sampled to represent the 50,000. The test we registered on 6 October scored the model on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022). The test set limits on coverage and bias, and the model failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). We use the same monthly model specification for the two earnings scenarios on our dashboard. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. +The withdrawn average came from a monthly statistical model of CPI and average weekly earnings (a vector autoregression), which we fitted to 2000 to 2026 without the furlough months. We resampled its shocks from its past errors and shifted every path so that the average calendar-year CPI and earnings growth across the 50,000 paths matched the OBR's March 2026 central path. We estimated the average saving from full runs of PolicyEngine UK on 160 paths we sampled to represent the 50,000. The test we registered on 6 October scored the model on the OBR's March forecasts from 2016 whose four years of outcomes were complete, seven by 28 October (March 2016 to March 2022). The test set limits on coverage and bias, and the model failed the limit on how often earnings growth falls below 2.5%, with a bias of 0.27 against a limit of 0.25 ([results](https://github.com/PolicyEngine/uk-triple-lock/pull/45)). The two earnings scenarios on our dashboard use the same monthly model specification. We centre its paths on the central path above, then lower or raise the calendar-year earnings targets by 0.5 percentage points a year from 2031. ### Model and data -We calculate every fiscal and household figure with a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0). We replace the model's own forecasts with each scenario's CPI and earnings, and the model then moves CPI-uprated benefits, CPI-indexed thresholds and employment income with the scenario. We calculate the net saving as the change in the government's balance across every tax and benefit the model covers. +Every fiscal and household figure comes from a full run of PolicyEngine UK (policyengine.py 6.11.1, policyengine-uk 2.151.0) on Microcosm UK 2024-25 (microcosm-data 0.1.0), PolicyEngine's dataset of UK households, which starts from the Family Resources Survey. Each scenario's CPI and earnings replace the model's own forecasts, so CPI-uprated benefits, CPI-indexed thresholds and employment income follow the scenario. The net saving is the change in the government's balance across every tax and benefit the model covers. ### Population -We reweight the households each year after 2025-26 so that each age and sex group grows as in the ONS 2024-based population projection. We hold each person's age fixed at the value in the data. State Pension type follows birth cohort: basic for people who reached State Pension age before 6 April 2016, new after. The data exclude pensioners living abroad. +We reweight the households each year after 2025-26 so that each age and sex group grows as in the ONS 2024-based population projection. Each person keeps the age they have in the data. State Pension type follows birth cohort: basic for people who reached State Pension age before 6 April 2016, new after. The data exclude pensioners living abroad. ### Limitations - We reweight today's households to the 2039-40 population. The model does not simulate deaths, migration or contribution records, so a pensioner in 2039-40 carries the record of someone the same age today. -- The model adjusts Housing Benefit and council tax reduction only for households that already claim them, so it assigns no new claims when a lower State Pension makes a household eligible. That understates those offsets and so overstates the net saving. The model calculates Pension Credit for newly eligible households when their take-up draw in the data indicates they would claim it. +- Housing Benefit and council tax reduction respond only for households already claiming them, so nobody the adjusted Triple Lock makes newly eligible starts claiming them. That understates those offsets and so overstates the net saving. Pension Credit can go to newly eligible households, when their take-up draw in the data indicates they would claim it. - The Savings Credit threshold and the pension-age Housing Benefit allowances, which the model also uses for council tax reduction, rise with CPI in the model, while the Pension Credit guarantee rises with earnings. In law, maximum Savings Credit is 60% of the gap between the guarantee and the threshold, and DWP describes protected Housing Benefit allowances as the guarantee plus maximum Savings Credit. The guarantee's link to earnings does not set these amounts' future paths. Projecting these amounts from the guarantee instead of CPI could change the Pension Credit, Housing Benefit and council tax reduction offsets under both rules. - The OBR publishes no May to July earnings forecast, so the Pension Credit guarantee's rises from April 2028 to April 2031 follow its April to June figure, as the Triple Lock's earnings input does. - The model splits each reported State Pension into a flat-rate part, up to the full basic or new rate, and anything above it, which rises with CPI under both rules. Where a pension falls short of the full rate, the model scales all of it with the flat rate, including any Graduated Retirement Benefit or extra pension for deferring. In law those parts rise with prices: in April 2025, they rose 1.7% while the basic State Pension rose 4.1% ([DWP](https://www.gov.uk/government/publications/benefit-and-pension-rates-2025-to-2026/benefit-and-pension-rates-2025-to-2026)). They would rise the same under both rules, so scaling them overstates the saving. -- We hold rents and council tax at their 2030 levels after 2030 and do not adjust dividend, property, savings or self-employment income to the scenario. +- Rents and council tax stay at their 2030 levels after 2030, and dividend, property, savings and self-employment income do not follow the scenario. - Nobody changes how much they work, save or claim in response to the adjusted Triple Lock. -- We calculate costs for the UK in cash terms and use our figures for Great Britain to compare with DWP. +- We calculate costs for the UK in cash terms; comparisons with DWP use our figures for Great Britain.
diff --git a/app/src/data/posts/posts.json b/app/src/data/posts/posts.json index 3f23b815a..053f23ef6 100644 --- a/app/src/data/posts/posts.json +++ b/app/src/data/posts/posts.json @@ -64,7 +64,7 @@ }, { "title": "The adjusted Triple Lock in the Autumn Budget 2026", - "description": "From April 2030, the government plans to raise the State Pension by the higher of inflation and 2.5%, or by more to hold its April 2029 value relative to earnings. The government would spend \u00a30.5 billion less on the State Pension in 2039-40 on the OBR's central path and \u00a310.9 billion less if the pension outpaces earnings under the Triple Lock, as the OBR assumes.", + "description": "From April 2030, the State Pension rises by the higher of inflation and 2.5%, or by more where needed to hold its April 2029 value relative to earnings. The adjusted Triple Lock saves \u00a30.5 billion in 2039-40 on the OBR's central path and \u00a310.9 billion if the Triple Lock outpaces earnings as the OBR assumes.", "date": "2026-10-28 17:00:00", "tags": ["uk", "policy", "autumn-budget"], "authors": ["max-ghenis"],