Report each pull request's effect on national aggregates on the certified dataset by MaxGhenis · Pull Request #2335 · PolicyEngine/policyengine-uk

MaxGhenis · 2026-10-11T17:46:21Z

This is workstream W3 of the Autumn Budget 2026 CI plan. It stays a draft until the independent GPT-6.1 Sol review is back.

The workflow has already run on this PR: see the "Aggregate impact on certified data" comment below. Both sides ran on uk-data 1.58.0 in identical environments (the same environment hash), at about 5 minutes a side. That comment was removed with the data-protection fix below and the next run replaces it. The result was pass, with no moves, as expected for a PR that changes no formula.

What it adds

An aggregate-impact report. The module is policyengine_uk.utils.aggregate_impact, and the workflow is .github/workflows/aggregate-impact.yaml. On every PR, the workflow measures the base and the head on the same certified dataset, on the same runner image. Each side syncs its own uv.lock and runs the PR head's copy of the tool. The report covers every year from 2025-26 to 2031-32:
- receipts by tax head;
- spending and caseloads by benefit;
- poverty: absolute and relative, BHC and AHC, for everyone, children, working-age adults and pensioners;
- the Gini coefficient;
- mean income in each side's own deciles, and household-matched decile changes within the base deciles;
- ten standard reform costings: the eight fixtures, plus a 10% rise in the Universal Credit standard allowance and in the State Pension.
Base and head are measured in one job. The dataset is licensed survey microdata, so only aggregates leave that job; a test enforces it (see "Data protection" below).

It posts one comment and updates it on later runs. It fails on:
- a crash;
- a different dataset on the two sides;
- non-finite figures;
- any check that the base passes and the head breaks.
Moves above the flag thresholds are listed, not failed.
A fiscal partition, policyengine_uk/data/fiscal_partition.yaml. The tax heads are the specification from the Corporate SDLT is counted twice in household and government tax totals #2159 work (Count corporate SDLT once and let SDLT abolition remove all of it #2235), which was written from the OBR receipts lines and each variable's formula. They are encoded here for main before that fix: corporate_sdlt is a declared double count inside expected_sdlt. The spending heads cover gov_spending and household_benefits.
Immutable test data and a mandatory-data mode, in tests/conftest.py.
- tests/certified_dataset.yaml pins uk-data 1.58.0's enhanced_frs_2024_25.h5 by Hugging Face commit 47bbb58d, with sha256 5ba399ea…. That is the artifact policyengine.py#555 certified with model 2.125.1.
- Data tests verify those bytes once per process before they run.
- With POLICYENGINE_UK_REQUIRE_TEST_DATA=1, every data test fails in setup when the dataset is unavailable, and runs only on the pinned file.
- Without the flag, tests skip as before. Credential-free unit tests are unaffected.
- W2 wires the flag into the existing workflows; this PR does not touch them.
Reform fixtures rebased onto 1.58.0. reforms_config.yaml keeps every earlier value as a receipt, with the model and data that produced it. It also records current main's figure on 1.40.3, so the data's share of each move is visible. The earlier values were computed on whatever Enhanced FRS was latest until the 1.40.3 pin landed on 2026-04-05.

Invariants the report enforces

Closure. For each of gov_tax, household_tax, pre_budget_change_household_tax, gov_spending and household_benefits, the model's own component list is exactly the partition's heads plus its declared double counts, each listed once. Each head's declared entity is its variable's entity.
Identities, household by household.
- Heads plus double counts equal each model total, respecting dropped_when (council tax abolition).
- gov_balance equals gov_tax − gov_spending.
- Household net income equals market income plus benefits, less taxes and pension contributions.
Counted once. Adding £1 to any component of a tax total moves that total by exactly £1 per unit. On a spending total it moves by at most £1 per unit: means tests can claw some of it back, but no other benefit may rise.
- A declared double count must still be contained in its head, or the check fails as stale.
- The check reuses one simulation, restores every input it overrode, and requires the recomputed baseline to match the first run exactly.
Bounds and finiteness. Weights are finite and non-negative. Poverty rates lie in [0, 1]. Every figure is finite. Heads declared sign: nonnegative are never negative.
Determinism. A rebuilt unreformed simulation reproduces the baseline gov_balance exactly.
Reform responsiveness. Each standard reform names the variable it must move, the direction and a minimum (standard_reforms.yaml). A reform that moves its target by less, by nothing or the wrong way in any year fails responsive:<id>, so a reform the model silently ignores cannot pass as a small number.
- Relative reforms are built from each side's own baseline parameters: that model year's value times the factor.
- On 1.58.0 all ten respond in every year. The UC standard allowance rise adds £3.0bn to £3.4bn of Universal Credit; the State Pension rise adds £12.4bn to £14.4bn.
Comparison.
- Comparing a side with itself changes nothing and flags nothing.
- Changes are antisymmetric.
- A flag fires exactly when |head − base| > max(absolute, relative × |base|).
- The comment always carries its marker and fits GitHub's limit.

Tests

Property-based (Hypothesis):
- the data plan over random environments: mandatory mode never skips and runs only on the pin, and unrelated variables never change the plan;
- pin parsing: only a 40-hex commit passes;
- byte verification against random tampering;
- partition mutations: a duplicate head, a missing head, an extra component, a mis-scoped double count;
- head-order invariance;
- counted-once on a fake simulation with known containment: the coefficient is 2 exactly when a head contains another;
- responsiveness: a reform passes exactly when every year moves enough in the right direction, and the verdict is symmetric under negation;
- comparison identity, antisymmetry and thresholds;
- comment size.
Differential:
- the partition against the model's own lists;
- decile impacts against a pandas groupby;
- Gini against microdf;
- head totals against Microsimulation.calculate(...).sum(), comparing person and benefit-unit weights with household-mapped sums.
On certified data (microsimulation marker): test_aggregate_impact_certified.py runs 2026-27 with the counted-once check on every tax component and four benefits, and the two relative reforms.

Commands run locally:

pytest policyengine_uk/tests/aggregate_impact: 79 passed without credentials (7 data tests skipped); the 7 data tests pass with POLICYENGINE_UK_REQUIRE_TEST_DATA=1 and a token.
pytest policyengine_uk/tests -m microsimulation on the 1.58.0 pin: every test outside the rebased fixtures passed before the rebase; the rebased fixtures were then recomputed.
A base-against-head rehearsal of the workflow (main against this branch): pass, no moves.

The full Python suite was not run locally; CI runs it.

What the report already finds on main

These figures are from certified 1.58.0, 2026-27, main at 6d89338. The first two are filed as #2337 and #2338, and the negative VAT as policyengine-uk-data#565.

gov_tax, household_tax and pre_budget_change_household_tax count corporate_sdlt twice: coefficient 2.000, £3.13bn a year (Corporate SDLT is counted twice in household and government tax totals #2159, fixed in Count corporate SDLT once and let SDLT abolition remove all of it #2235).
Universal Credit rises about £0.15 for each £1 of Statutory Sick Pay or Statutory Maternity Pay (Universal Credit rises with SSP and SMP: tax on them is deducted but they aren't counted as earnings #2337).
- taxable_employment_income adds employment_benefits, which is SSP plus SMP, so UC deducts their income tax and NI.
- But uc_individual_earned_income_before_mif counts only employment, miscellaneous and self-employment income as earnings.
- This shows as failing on both sides, so it doesn't block PRs.
Scottish Child Payment (£0.52bn) and Healthy Start (£0.05bn) are in household_benefits but missing from gov_spending, so gov_balance leaves out their cost (Scottish Child Payment and Healthy Start are household benefits but missing from government spending #2338).
76 households have negative VAT. This also fails on both sides.
Alcohol duty, tobacco duty and VED are £0 for every household on 1.58.0.
The dataset is extended explicitly only to 2030: extend_single_year_dataset(end_year=2030).
- 2031-32 is filled by policyengine-core instead (simulations/simulation.py ~1121–1215): an input with an uprating index is uprated from 2030, and other inputs are carried over unchanged. Council tax per household, for example, is flat 2030→2031.
- So 2031-32 figures are provisional.
VAT on 1.58.0 is £326bn in 2026-27, against £201bn with the same model on 1.40.3, because recorded consumption is higher and the 0.38 coverage factor is unchanged (Re-derive microdata_vat_coverage: the value has been 0.383 since 2010 and overshoots OBR VAT receipts on current datasets #1996, Split VAT grossing into household share and survey coverage #2187). This is why the VAT 2pp fixture moves from 21.3 to 34.5.

Data protection

An earlier version of this workflow (commits 073e808 to 8d3e482) uploaded each side's measurement as an Actions artifact, including the household-level arrays the decile comparison uses, and cached the dataset in the Actions cache. Those artifacts and the cache were deleted on 11 October, and fd94415 removes both:

both sides are measured in one job, and only the report directory is uploaded (two result.json files of aggregates, the comment and the summary);
the dataset is never cached;
household detail is written under private/;
check details give counts and weighted totals, never one household's value.

test_aggregate_impact_workflow.py fails if the workflow uploads anything else, uses the cache, or splits the measurement across jobs.

Thresholds and methodology

These are proposals pending cos d1344:

the flag thresholds in report_spec.yaml: max(£100m, 0.1%) for £ totals, max(5,000, 0.5%) for caseloads, 0.1pp for rates and 0.001 for Gini. Flags never fail the job;
the responsiveness minimums in standard_reforms.yaml, each well below the reform's smallest yearly effect on 1.58.0.

No model semantics change in this PR.

Not in this PR

W2 sets POLICYENGINE_UK_REQUIRE_TEST_DATA in the existing workflows.
The uk-data gates (model-fixed comparison, final-fit and target-coverage gates, single-age and HBAI checks, OBR ingestion) come in separate uk-data PRs for the d833 batch.

axiom: n/a: CI, test-data and reporting infrastructure; no policy change

🤖 Generated with Claude Code

The report measures one model checkout on the pinned certified dataset over 2025-26 to 2031-32: receipts by tax head, spending and caseloads by benefit, poverty by measure and group, inequality, own-decile incomes and the standard reform costings. It checks them against a fiscal partition of the model's totals (policyengine_uk/data/fiscal_partition.yaml): - closure: each total's components are exactly the partition's heads and declared double counts; - accounting identities, household by household; - a counted-once check, which perturbs each component by £1 and requires each tax total to move by exactly £1 per unit, and each spending total by at most £1. compare then sets a base measurement against a head one. It fails on a crash, a different dataset, non-finite figures or a check the base passes and the head breaks. It flags moves above the proposed thresholds, and renders a pull request comment. The partition's tax heads follow the #2159 work's specification, encoded for main before that fix: corporate_sdlt is a declared double count inside expected_sdlt. Spending heads cover gov_spending and household_benefits. Test data: conftest pins enhanced_frs_2024_25 from uk-data 1.58.0 by Hugging Face commit and SHA-256 (tests/certified_dataset.yaml), verifies the bytes before data tests run, and adds a mandatory-data mode (POLICYENGINE_UK_REQUIRE_TEST_DATA). In that mode, data tests fail rather than skip without data, and run only on the pinned dataset. The reform-cost fixtures are rebased onto 1.58.0, keeping each earlier value as a receipt with the model and data that produced it. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

.github/workflows/aggregate-impact.yaml measures the pull request's base and head on the same runner image. Each side syncs its own uv.lock and runs the head's copy of the tool against its own model, on the pinned dataset whose bytes the tool verifies. The report job then compares the two, writes the job summary and one updated comment, and fails on structural failures. Fork and Dependabot runs, which cannot read the private dataset, record a skip instead. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Sections with no change collapse into one line, an unchanged decile table into one sentence, and the counted-once summary lists only real violations (a tax total moving by other than £1 per unit, or a spending total by more than £1), not the means-test offsets every benefit shows. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The aggregate impact report (#2335) adds gov_spending, household_benefits and spending_heads to fiscal_partition.yaml. The tax tests now check the three tax totals only, require that spending heads borrow no tax head id or variable and that no spending head names a tax total, and apply dropped_when to the tax totals alone. Checked against #2335's file with these tax sections merged in: this file's tests and #2335's partition and guard tests pass. The vat head records that 76 households have negative VAT on policyengine-uk-data 1.58.0 (consumption or the full-rate expenditure share is negative in the data; policyengine-uk-data issue 565). Refs #2159 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…lookup - Set POLICYENGINE_UK_REQUIRE_TEST_DATA=1 (W3's conftest switch, #2335) wherever require-data is on: main pushes, merge queue groups and releases. A test that needs the dataset then fails with its own message instead of skipping. Unused until #2335 lands. - A manual dispatch of the main workflow always tests; the merge queue lookup applies only to pushes, and a failed lookup parses to zero. - Document the re-run path's residual and where docs-only pull requests build the documentation; the workflow-shape test covers the release workflow's require-data and the lookup. From the GPT-6.1 Sol delta review of 96e9cec (APPROVE, non-blocking). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The standard reforms move into the tool (standard_reforms.yaml), so a vendored copy carries them: the eight reform-cost fixtures, plus a 10% rise in the Universal Credit standard allowance and in the State Pension. Each names the variable it must move, the direction and a minimum. A reform that moves its target by less, by nothing or the wrong way in any year fails responsive:<id>, so a reform the model silently ignores cannot pass as a small number. A relative reform is built from each side's own baseline parameters: that model year's value times the factor, for every year of the horizon. On uk-data 1.58.0 all ten respond in every year, 2025-26 to 2031-32. The UC standard allowance rise adds £3.0bn to £3.4bn of Universal Credit, and the State Pension rise £12.4bn to £14.4bn. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The certified dataset is licensed survey microdata, and only aggregates may be published. The workflow uploaded each side's measurement as an artifact, including the household-level arrays the decile comparison uses, and cached the dataset in the Actions cache. Both are removed: - base and head are now measured in one job, so the household detail never travels between jobs; only the report directory (the two result.json files, the comment and the summary) is uploaded; - the dataset is downloaded on each run and never cached; - measure writes household detail under private/, apart from result.json; - check details give counts and weighted totals, not one household's value. test_aggregate_impact_workflow.py fails if the workflow uploads anything but the report, uses the Actions cache, copies anything private into the report, or splits the measurement across jobs. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The aggregate impact report (#2335) adds gov_spending, household_benefits and spending_heads to fiscal_partition.yaml. The tax tests now check the three tax totals only, require that spending heads borrow no tax head id or variable and that no spending head names a tax total, and apply dropped_when to the tax totals alone. Checked against #2335's file with these tax sections merged in: this file's tests and #2335's partition and guard tests pass. The vat head records that 76 households have negative VAT on policyengine-uk-data 1.58.0 (consumption or the full-rate expenditure share is negative in the data; policyengine-uk-data issue 565). Refs #2159 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

github-actions · 2026-10-11T23:43:46Z

Aggregate impact on certified data: pass

Base 3f174d07ec against head c6d99a2944, both on enhanced_frs_2024_25.h5 (policyengine-uk-data 1.58.0, sha256 5ba399ea5ab9), fiscal years 2025-26 to 2031-32. Each year is a full microsimulation run on each side.

Checks: 0 broken by this change, 4 failing on both sides, 0 fixed. Moves flagged for review: 0 (thresholds proposed in report_spec.yaml, pending cos d1344; a flag does not fail the job).

Workflow run and full report

Moves flagged for review

None.

Household net income by decile

No household's net income changes in any year.

No change in any year: receipts by tax head; spending by benefit and service; caseloads; government balance, each component counted once; model totals; components the model counts twice; household income totals; population; poverty rates; people in poverty; inequality; mean household net income in each side's own deciles; standard reform costings (change in gov_balance); standard reforms: change in each reform's target.

Checks failing on both base and head (4)

VAT is never negative (sign:vat): 2025-26: 76 units below zero, a weighted £-69,996,281 in total; 2026-27: 76 units below zero, a weighted £-71,878,301 in total; 2027-28: 76 units below zero, a weighted £-73,587,136 in total
At most one of expected_sdlt, expected_ltt, expected_lbtt is non-zero for any household (overlap_guard:property_transaction_tax_geography): 2025-26: 86 households have more than one, a weighted 39,273; 2026-27: 86 households have more than one, a weighted 39,422; 2027-28: 88 households have more than one, a weighted 39,718
Raising statutory_maternity_pay by £1 raises its own head by £1, moves no other tax head and raises no other benefit (counted_once:statutory_maternity_pay): gov_spending: universal_credit rises by 0.0068 per unit of statutory_maternity_pay; household_benefits: universal_credit rises by 0.0068 per unit of statutory_maternity_pay
Raising statutory_sick_pay by £1 raises its own head by £1, moves no other tax head and raises no other benefit (counted_once:statutory_sick_pay): gov_spending: universal_credit rises by 0.0099 per unit of statutory_sick_pay; household_benefits: universal_credit rises by 0.0099 per unit of statutory_sick_pay

Counted-once check

Each of 82 components was raised by £1 per unit in 2026-27: tax totals must rise by exactly £1 per unit and spending totals by at most £1. 2 failed; 0 were skipped (zero for every unit, or not money).

statutory_maternity_pay: gov_spending 1.0068, household_benefits 1.0068
statutory_sick_pay: gov_spending 1.0099, household_benefits 1.0099

In the model's own totals:

jsa_income moves gov_spending by 45.6619 per unit
jsa_income moves household_benefits by 65.8035 per unit
income_support moves gov_spending by 45.6619 per unit
income_support moves household_benefits by 65.8035 per unit
working_tax_credit moves gov_spending by 17.9161 per unit
working_tax_credit moves household_benefits by 17.9161 per unit
incapacity_benefit moves gov_spending by 27.2846 per unit
incapacity_benefit moves household_benefits by 27.2846 per unit
maternity_allowance moves gov_spending by 2.7303 per unit
maternity_allowance moves household_benefits by 2.7303 per unit
statutory_maternity_pay moves gov_spending by 1.0068 per unit
statutory_maternity_pay moves household_benefits by 1.0068 per unit
statutory_sick_pay moves gov_spending by 1.0099 per unit
statutory_sick_pay moves household_benefits by 1.0099 per unit
corporate_sdlt moves gov_tax by 2.0000 per unit
corporate_sdlt moves household_tax by 2.0000 per unit
corporate_sdlt moves pre_budget_change_household_tax by 2.0000 per unit
declared double count corporate_sdlt is still contained in expected_sdlt (coefficient 1.000; Corporate SDLT is counted twice in household and government tax totals #2159)

Provenance

Side	Commit	Runtime	Environment hash	Partition hash	Time
base	main `3f174d07ec`	uk 2.126.10, core 3.32.27, py 3.13.16 aarch64	`82ac7ba16ba2`	`779be6fce587`	384.08s
head	#2335 merged into main `c6d99a2944`	uk 2.126.10, core 3.32.27, py 3.13.16 aarch64	`82ac7ba16ba2`	`779be6fce587`	389.38s

Dataset: hf://policyengine/policyengine-uk-data-private/enhanced_frs_2024_25.h5@47bbb58d3f6467aa4a7059bc6110d2bdcf699b29, sha256 5ba399ea5ab9a6c7d4e9aaf6c623080beb34d9ad5f330f6244dbad72b8d960b6 (verified on both sides), policyengine-uk-data 1.58.0.

Receipts by tax head: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Income tax	£300.49bn	£320.54bn	£338.02bn	£353.23bn	£369.12bn	£387.23bn	£411.43bn
National Insurance (employee and self-employed)	£53.13bn	£55.50bn	£57.32bn	£59.02bn	£60.93bn	£62.87bn	£65.32bn
National Insurance (employer)	£154.54bn	£161.15bn	£166.17bn	£170.81bn	£177.11bn	£182.68bn	£189.88bn
Employer NI change borne by capital	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Employer NI change borne by consumers	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Capital gains tax	£12.72bn	£13.16bn	£13.65bn	£14.15bn	£14.65bn	£15.20bn	£15.77bn
Student loan repayments	£7.83bn	£8.13bn	£8.35bn	£8.56bn	£8.79bn	£9.04bn	£8.06bn
VAT	£317.14bn	£325.66bn	£333.41bn	£341.43bn	£349.79bn	£358.40bn	£367.14bn
Change in VAT from its baseline	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
VAT on private school fees	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Fuel duties	£24.04bn	£24.14bn	£26.10bn	£26.16bn	£25.89bn	£25.24bn	£25.82bn
Alcohol duties	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Tobacco duties	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Vehicle excise duties	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
TV licence fee	£4.31bn	£4.46bn	£4.48bn	£4.50bn	£4.52bn	£4.54bn	£4.61bn
Council tax	£49.77bn	£52.67bn	£56.01bn	£59.64bn	£63.25bn	£67.02bn	£67.31bn
High value council tax surcharge	£0.00bn	£0.00bn	£0.00bn	£0.49bn	£0.51bn	£0.52bn	£0.53bn
Northern Ireland domestic rates	£0.58bn	£0.61bn	£0.64bn	£0.67bn	£0.70bn	£0.74bn	£0.74bn
Business rates	£31.73bn	£31.73bn	£31.73bn	£31.73bn	£31.73bn	£31.73bn	£31.73bn
Stamp duty land tax (own purchases, and corporate until the	£19.45bn	£20.44bn	£21.41bn	£22.39bn	£23.39bn	£24.49bn	£25.65bn
Land transaction tax (Wales)	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Land and buildings transaction tax (Scotland)	£0.10bn	£0.10bn	£0.11bn	£0.11bn	£0.12bn	£0.13bn	£0.13bn
Unspecified tax change borne by capital	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Unspecified tax change borne by consumers	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Wealth tax	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Wealth tax excluding the main residence	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Land value tax	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Carbon tax	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn

Spending by benefit and service: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
State Pension	£123.61bn	£128.31bn	£130.13bn	£132.24bn	£136.14bn	£140.17bn	£144.30bn
Pension Credit	£6.69bn	£7.05bn	£7.03bn	£7.11bn	£7.23bn	£7.36bn	£6.71bn
Winter Fuel Payment	£1.70bn	£1.69bn	£1.62bn	£1.60bn	£1.59bn	£1.57bn	£1.56bn
Pension Age Winter Heating Payment (Scotland)	£0.20bn	£0.20bn	£0.21bn	£0.21bn	£0.22bn	£0.22bn	£0.23bn
Universal Credit	£79.30bn	£83.59bn	£85.21bn	£86.75bn	£87.59bn	£84.52bn	£84.96bn
Housing Benefit	£8.16bn	£8.48bn	£8.73bn	£8.86bn	£9.13bn	£9.36bn	£7.91bn
Council Tax Reduction	£2.11bn	£2.16bn	£2.21bn	£2.26bn	£2.32bn	£2.38bn	£2.14bn
Employment and Support Allowance (income-related)	£2.16bn	£2.22bn	£2.27bn	£2.32bn	£2.37bn	£2.43bn	£2.49bn
Employment and Support Allowance (contribution-based)	£5.19bn	£5.32bn	£5.45bn	£5.58bn	£5.72bn	£5.86bn	£6.00bn
Jobseeker's Allowance (income-based)	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Jobseeker's Allowance (contribution-based)	£0.30bn	£0.31bn	£0.31bn	£0.32bn	£0.33bn	£0.34bn	£0.34bn
Income Support	£0.17bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Working Tax Credit	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Child Tax Credit	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Incapacity Benefit	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Personal Independence Payment	£26.93bn	£28.06bn	£28.81bn	£29.51bn	£30.23bn	£30.97bn	£31.73bn
Disability Living Allowance	£7.48bn	£7.80bn	£8.01bn	£8.20bn	£8.40bn	£8.61bn	£8.82bn
Attendance Allowance	£8.74bn	£9.11bn	£9.35bn	£9.57bn	£9.81bn	£10.05bn	£10.30bn
Carer's Allowance	£4.65bn	£4.84bn	£4.97bn	£5.09bn	£5.22bn	£5.35bn	£5.48bn
Carer Support Payment (Scotland)	£0.38bn	£0.39bn	£0.40bn	£0.41bn	£0.42bn	£0.43bn	£0.44bn
Carer's Allowance Supplement (Scotland)	£0.00bn	£0.05bn	£0.05bn	£0.05bn	£0.05bn	£0.05bn	£0.05bn
Severe Disablement Allowance	£0.02bn	£0.02bn	£0.02bn	£0.03bn	£0.03bn	£0.03bn	£0.03bn
Industrial Injuries Disablement Benefit	£0.51bn	£0.52bn	£0.53bn	£0.55bn	£0.56bn	£0.57bn	£0.59bn
Armed Forces Compensation Scheme	£0.57bn	£0.58bn	£0.60bn	£0.61bn	£0.63bn	£0.64bn	£0.66bn
Child Benefit	£16.23bn	£16.90bn	£17.34bn	£17.76bn	£18.18bn	£18.62bn	£19.05bn
Scottish Child Payment	£0.50bn	£0.52bn	£0.53bn	£0.55bn	£0.56bn	£0.57bn	£0.58bn
Two Child Limit Payment (Scotland, counterfactual)	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Healthy Start vouchers	£0.05bn	£0.05bn	£0.05bn	£0.05bn	£0.05bn	£0.06bn	£0.06bn
Maternity Allowance	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Statutory Maternity Pay	£3.34bn	£3.43bn	£3.51bn	£3.59bn	£3.68bn	£3.77bn	£3.86bn
Statutory Sick Pay	£1.18bn	£1.21bn	£1.24bn	£1.27bn	£1.30bn	£1.33bn	£1.37bn
Sure Start Maternity Grant	£0.01bn	£0.01bn	£0.01bn	£0.01bn	£0.01bn	£0.01bn	£0.01bn
Bereavement Support Payment	£0.65bn	£0.67bn	£0.69bn	£0.70bn	£0.72bn	£0.74bn	£0.76bn
Tax-Free Childcare	£0.56bn	£0.57bn	£0.57bn	£0.59bn	£0.56bn	£0.58bn	£0.56bn
Extended childcare entitlement	£4.77bn	£5.31bn	£5.43bn	£5.58bn	£5.68bn	£5.86bn	£5.93bn
Universal childcare entitlement	£1.30bn	£1.37bn	£1.41bn	£1.45bn	£1.53bn	£1.57bn	£1.61bn
Targeted childcare entitlement	£0.44bn	£0.46bn	£0.47bn	£0.48bn	£0.50bn	£0.50bn	£0.50bn
Care to Learn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Childcare Grant	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Parents' Learning Allowance	£0.20bn	£0.20bn	£0.20bn	£0.20bn	£0.20bn	£0.17bn	£0.17bn
Adult Dependants' Grant	£0.04bn	£0.04bn	£0.04bn	£0.04bn	£0.04bn	£0.04bn	£0.04bn
Travel Grant	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Disabled Students' Allowance	£0.03bn	£0.03bn	£0.03bn	£0.03bn	£0.03bn	£0.03bn	£0.03bn
16 to 19 Bursary Fund	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
NHS spending	£205.23bn	£206.01bn	£206.77bn	£207.60bn	£208.52bn	£209.45bn	£210.35bn
State education spending	£0.03bn	£0.03bn	£0.03bn	£0.03bn	£0.03bn	£0.03bn	£0.03bn
Transport subsidy spending	£25.15bn	£25.25bn	£25.34bn	£25.44bn	£25.55bn	£25.67bn	£30.94bn
Cost-of-living support payment	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Energy Bills Rebate	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Energy Price Guarantee subsidy	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Basic income	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Disability basic income	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn
Other public spending budget change	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn	£0.00bn

Caseloads: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
State Pension	11,410k	11,315k	11,062k	10,978k	11,026k	11,076k	11,124k
Pension Credit	1,483k	1,481k	1,468k	1,447k	1,448k	1,445k	1,072k
Winter Fuel Payment	7,113k	7,035k	6,701k	6,596k	6,562k	6,475k	6,396k
Pension Age Winter Heating Payment (Scotland)	829k	830k	829k	827k	831k	835k	838k
Universal Credit	6,348k	6,361k	6,372k	6,388k	6,368k	6,307k	6,291k
Housing Benefit	1,232k	1,232k	1,234k	1,219k	1,224k	1,226k	1,041k
Council Tax Reduction	3,390k	3,414k	3,432k	3,455k	3,471k	3,504k	3,284k
Employment and Support Allowance (income-related)	189k	190k	190k	191k	192k	192k	193k
Employment and Support Allowance (contribution-based)	717k	719k	722k	725k	728k	731k	735k
Jobseeker's Allowance (income-based)	0k	0k	0k	0k	0k	0k	0k
Jobseeker's Allowance (contribution-based)	69k	70k	70k	70k	70k	71k	71k
Income Support	34k	0k	0k	0k	0k	0k	0k
Working Tax Credit	0k	0k	0k	0k	0k	0k	0k
Child Tax Credit	0k	0k	0k	0k	0k	0k	0k
Incapacity Benefit	0k	0k	0k	0k	0k	0k	0k
Personal Independence Payment	3,561k	3,575k	3,588k	3,602k	3,618k	3,634k	3,650k
Disability Living Allowance	1,335k	1,340k	1,345k	1,350k	1,356k	1,362k	1,368k
Attendance Allowance	1,717k	1,723k	1,729k	1,736k	1,744k	1,752k	1,759k
Carer's Allowance	1,073k	1,077k	1,081k	1,086k	1,090k	1,095k	1,100k
Carer Support Payment (Scotland)	87k	87k	87k	88k	88k	88k	89k
Carer's Allowance Supplement (Scotland)	0k	87k	87k	88k	88k	88k	89k
Severe Disablement Allowance	4k	4k	4k	4k	4k	4k	4k
Industrial Injuries Disablement Benefit	136k	137k	137k	138k	138k	139k	140k
Armed Forces Compensation Scheme	87k	88k	88k	88k	89k	89k	89k
Child Benefit	8,145k	8,170k	8,197k	8,229k	8,262k	8,296k	8,318k
Scottish Child Payment	353k	354k	356k	357k	357k	359k	357k
Two Child Limit Payment (Scotland, counterfactual)	0k	0k	0k	0k	0k	0k	0k
Healthy Start vouchers	209k	209k	210k	211k	212k	213k	214k
Maternity Allowance	0k	0k	0k	0k	0k	0k	0k
Statutory Maternity Pay	321k	323k	324k	325k	326k	328k	329k
Statutory Sick Pay	186k	187k	187k	188k	189k	190k	190k
Sure Start Maternity Grant	28k	28k	28k	28k	29k	29k	29k
Bereavement Support Payment	92k	93k	93k	93k	94k	94k	95k
Tax-Free Childcare	974k	980k	971k	977k	958k	967k	915k
Extended childcare entitlement	962k	958k	959k	964k	956k	963k	954k
Universal childcare entitlement	379k	386k	389k	391k	404k	407k	409k
Targeted childcare entitlement	92k	92k	92k	93k	93k	92k	91k
Care to Learn	0k	0k	0k	0k	0k	0k	0k
Childcare Grant	0k	0k	0k	0k	0k	0k	0k
Parents' Learning Allowance	99k	99k	100k	100k	100k	83k	83k
Adult Dependants' Grant	10k	11k	11k	11k	11k	11k	11k
Travel Grant	0k	0k	0k	0k	0k	0k	0k
Disabled Students' Allowance	2k	2k	2k	2k	2k	2k	2k
16 to 19 Bursary Fund	0k	0k	0k	0k	0k	0k	0k
Cost-of-living support payment	0k	0k	0k	0k	0k	0k	0k
Energy Bills Rebate	0k	0k	0k	0k	0k	0k	0k
Energy Price Guarantee subsidy	0k	0k	0k	0k	0k	0k	0k
Basic income	0k	0k	0k	0k	0k	0k	0k
Disability basic income	0k	0k	0k	0k	0k	0k	0k

Government balance, each component counted once: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Government receipts, each component once	£975.84bn	£1,018.31bn	£1,057.41bn	£1,092.91bn	£1,130.52bn	£1,169.83bn	£1,214.13bn
Government spending	£538.03bn	£552.21bn	£559.03bn	£566.05bn	£574.52bn	£579.31bn	£589.95bn
Government balance, each receipt once	£437.80bn	£466.11bn	£498.38bn	£526.85bn	£555.99bn	£590.53bn	£624.18bn
Benefits in household_benefits but not in gov_spending	£0.55bn	£0.57bn	£0.59bn	£0.60bn	£0.61bn	£0.63bn	£0.64bn

Model totals: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
gov_tax	£978.97bn	£1,021.44bn	£1,060.54bn	£1,096.04bn	£1,133.65bn	£1,172.96bn	£1,217.26bn
household_tax	£507.29bn	£534.63bn	£560.97bn	£583.79bn	£606.74bn	£631.88bn	£660.24bn
pre_budget_change_household_tax	£499.46bn	£526.50bn	£552.61bn	£574.74bn	£597.44bn	£622.32bn	£651.65bn
gov_spending	£538.03bn	£552.21bn	£559.03bn	£566.05bn	£574.52bn	£579.31bn	£589.95bn
household_benefits	£538.39bn	£552.58bn	£559.41bn	£566.44bn	£574.92bn	£579.71bn	£590.35bn
gov_balance	£440.93bn	£469.24bn	£501.51bn	£529.98bn	£559.12bn	£593.66bn	£627.31bn

Components the model counts twice: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
corporate_sdlt, counted again inside expected_sdlt	£3.13bn	£3.13bn	£3.13bn	£3.13bn	£3.13bn	£3.13bn	£3.13bn

Household income totals: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Household market income	£1,671.23bn	£1,733.93bn	£1,786.79bn	£1,837.49bn	£1,899.44bn	£1,958.07bn	£2,030.31bn
Household net income	£1,662.85bn	£1,710.91bn	£1,743.12bn	£1,776.97bn	£1,812.65bn	£1,849.60bn	£1,901.81bn
Household net income (HBAI definition)	£1,477.50bn	£1,524.37bn	£1,557.42bn	£1,590.63bn	£1,624.77bn	£1,659.80bn	£1,706.27bn

Population: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
People	69,713k	69,978k	70,237k	70,518k	70,828k	71,147k	71,453k
Benefit units	35,437k	35,572k	35,703k	35,846k	36,004k	36,166k	36,321k
Households	31,245k	31,364k	31,480k	31,606k	31,745k	31,888k	32,025k

Poverty rates: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Absolute poverty, before housing costs: everyone	15.8%	15.0%	15.5%	15.6%	15.7%	16.0%	15.9%
Absolute poverty, before housing costs: children	23.0%	20.6%	21.3%	21.1%	21.3%	21.7%	21.7%
Absolute poverty, before housing costs: working-age adults	13.7%	13.4%	14.0%	14.1%	14.2%	14.5%	14.2%
Absolute poverty, before housing costs: pensioners	14.1%	13.6%	13.7%	13.9%	13.9%	14.0%	14.6%
Absolute poverty, after housing costs: everyone	20.8%	20.3%	20.9%	21.1%	21.4%	22.1%	22.0%
Absolute poverty, after housing costs: children	31.7%	30.0%	30.7%	30.9%	31.3%	32.8%	32.6%
Absolute poverty, after housing costs: working-age adults	19.4%	19.2%	19.8%	20.0%	20.3%	21.0%	20.5%
Absolute poverty, after housing costs: pensioners	12.5%	12.3%	12.1%	12.3%	12.4%	12.4%	13.4%
Relative poverty, before housing costs: everyone	17.1%	16.5%	16.4%	16.7%	16.8%	16.3%	17.0%
Relative poverty, before housing costs: children	25.6%	23.3%	22.9%	23.1%	23.2%	22.2%	23.4%
Relative poverty, before housing costs: working-age adults	14.6%	14.4%	14.7%	14.9%	15.2%	14.8%	15.1%
Relative poverty, before housing costs: pensioners	15.2%	15.2%	14.7%	14.8%	14.8%	14.4%	15.8%
Relative poverty, after housing costs: everyone	21.3%	21.1%	21.1%	21.2%	21.4%	22.0%	21.9%
Relative poverty, after housing costs: children	32.5%	31.3%	31.0%	31.0%	31.3%	32.5%	32.6%
Relative poverty, after housing costs: working-age adults	19.7%	19.8%	20.0%	20.1%	20.3%	21.0%	20.5%
Relative poverty, after housing costs: pensioners	13.1%	12.8%	12.6%	12.7%	12.4%	12.4%	13.4%

People in poverty: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Absolute poverty, before housing costs: everyone	11,008k	10,507k	10,884k	10,991k	11,106k	11,381k	11,373k
Absolute poverty, before housing costs: children	3,459k	3,104k	3,225k	3,214k	3,250k	3,329k	3,339k
Absolute poverty, before housing costs: working-age adults	5,861k	5,780k	6,068k	6,170k	6,241k	6,419k	6,320k
Absolute poverty, before housing costs: pensioners	1,688k	1,623k	1,591k	1,608k	1,614k	1,633k	1,714k
Absolute poverty, after housing costs: everyone	14,535k	14,233k	14,646k	14,868k	15,139k	15,746k	15,689k
Absolute poverty, after housing costs: children	4,765k	4,529k	4,640k	4,694k	4,779k	5,023k	5,017k
Absolute poverty, after housing costs: working-age adults	8,271k	8,234k	8,594k	8,748k	8,924k	9,278k	9,102k
Absolute poverty, after housing costs: pensioners	1,499k	1,470k	1,412k	1,426k	1,435k	1,445k	1,569k
Relative poverty, before housing costs: everyone	11,894k	11,526k	11,553k	11,765k	11,930k	11,608k	12,143k
Relative poverty, before housing costs: children	3,842k	3,520k	3,468k	3,508k	3,540k	3,405k	3,610k
Relative poverty, before housing costs: working-age adults	6,228k	6,198k	6,367k	6,543k	6,676k	6,529k	6,684k
Relative poverty, before housing costs: pensioners	1,824k	1,808k	1,718k	1,714k	1,713k	1,674k	1,850k
Relative poverty, after housing costs: everyone	14,870k	14,734k	14,834k	14,981k	15,157k	15,681k	15,680k
Relative poverty, after housing costs: children	4,880k	4,722k	4,689k	4,715k	4,784k	4,986k	5,016k
Relative poverty, after housing costs: working-age adults	8,415k	8,492k	8,682k	8,795k	8,935k	9,254k	9,098k
Relative poverty, after housing costs: pensioners	1,575k	1,520k	1,463k	1,472k	1,438k	1,441k	1,566k

Inequality: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Gini coefficient, equivalised HBAI income before housing costs	0.3117	0.3094	0.3106	0.3113	0.3119	0.3137	0.3127

Mean household net income in each side's own deciles: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Decile 1: mean household_net_income per household	£18,263	£18,566	£18,588	£18,712	£19,079	£19,169	£19,488
Decile 2: mean household_net_income per household	£28,213	£29,006	£28,623	£28,878	£29,211	£29,528	£30,103
Decile 3: mean household_net_income per household	£34,413	£35,673	£36,287	£36,892	£37,204	£37,192	£37,444
Decile 4: mean household_net_income per household	£37,643	£39,340	£40,284	£41,010	£41,559	£42,170	£43,753
Decile 5: mean household_net_income per household	£45,172	£46,059	£46,924	£47,338	£48,018	£48,015	£49,373
Decile 6: mean household_net_income per household	£48,821	£50,247	£51,421	£52,277	£52,862	£54,632	£56,252
Decile 7: mean household_net_income per household	£55,388	£56,756	£57,200	£58,135	£58,986	£60,252	£62,409
Decile 8: mean household_net_income per household	£70,806	£72,028	£73,226	£74,342	£75,651	£76,729	£77,557
Decile 9: mean household_net_income per household	£78,614	£80,277	£81,557	£83,394	£84,834	£86,473	£90,592
Decile 10: mean household_net_income per household	£121,629	£124,669	£125,904	£128,225	£130,773	£133,647	£135,670

Standard reform costings (change in gov_balance): head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Raise basic rate by 1pp	£6.23bn	£6.51bn	£6.72bn	£6.92bn	£7.13bn	£7.37bn	£7.65bn
Raise higher rate by 1pp	£3.67bn	£3.98bn	£4.22bn	£4.37bn	£4.61bn	£4.89bn	£5.11bn
Raise personal allowance by ~800GBP/year	£-3.82bn	£-3.94bn	£-4.03bn	£-4.07bn	£-4.13bn	£-4.23bn	£-1.71bn
Raise child benefit by 25GBP/week per additional child	£-1.98bn	£-1.74bn	£-1.63bn	£-1.53bn	£-1.43bn	£-1.32bn	£-1.22bn
Reduce Universal Credit taper rate to 20%	£-19.99bn	£-20.59bn	£-20.87bn	£-21.21bn	£-21.27bn	£-21.32bn	£-21.42bn
Raise Class 1 main employee NICs rate to 10%	£11.19bn	£11.65bn	£11.97bn	£12.26bn	£12.57bn	£12.91bn	£13.39bn
Raise VAT standard rate by 2pp	£31.26bn	£32.10bn	£32.86bn	£33.65bn	£34.48bn	£35.33bn	£36.19bn
Raise additional rate by 3pp	£2.83bn	£3.05bn	£3.24bn	£3.55bn	£3.76bn	£4.00bn	£4.33bn
Raise the Universal Credit standard allowance by 10%	£-3.02bn	£-3.16bn	£-3.21bn	£-3.27bn	£-3.30bn	£-3.33bn	£-3.36bn
Raise the State Pension by 10%	£-9.04bn	£-9.26bn	£-9.30bn	£-9.37bn	£-9.56bn	£-9.81bn	£-10.31bn

Standard reforms: change in each reform's target: head levels

Measure (head)	2025-26	2026-27	2027-28	2028-29	2029-30	2030-31	2031-32
Raise basic rate by 1pp: change in income_tax	£6.38bn	£6.67bn	£6.89bn	£7.09bn	£7.31bn	£7.55bn	£7.82bn
Raise higher rate by 1pp: change in income_tax	£3.67bn	£3.98bn	£4.23bn	£4.38bn	£4.62bn	£4.90bn	£5.12bn
Raise personal allowance by ~800GBP/year: change in income_tax	£-3.96bn	£-4.09bn	£-4.18bn	£-4.23bn	£-4.30bn	£-4.39bn	£-1.77bn
Raise child benefit by 25GBP/week per additional child: change in child_benefit	£2.33bn	£2.14bn	£2.03bn	£1.92bn	£1.82bn	£1.71bn	£1.59bn
Reduce Universal Credit taper rate to 20%: change in universal_credit	£20.03bn	£20.62bn	£20.91bn	£21.25bn	£21.31bn	£21.36bn	£21.46bn
Raise Class 1 main employee NICs rate to 10%: change in national_insurance	£11.46bn	£11.92bn	£12.26bn	£12.57bn	£12.88bn	£13.23bn	£13.71bn
Raise VAT standard rate by 2pp: change in vat	£31.26bn	£32.10bn	£32.86bn	£33.65bn	£34.48bn	£35.33bn	£36.19bn
Raise additional rate by 3pp: change in income_tax	£2.83bn	£3.05bn	£3.24bn	£3.55bn	£3.76bn	£4.00bn	£4.33bn
Raise the Universal Credit standard allowance by 10%: change in universal_credit	£3.03bn	£3.16bn	£3.22bn	£3.28bn	£3.31bn	£3.34bn	£3.37bn
Raise the State Pension by 10%: change in state_pension	£12.36bn	£12.83bn	£13.01bn	£13.22bn	£13.61bn	£14.02bn	£14.43bn

Counted-once check. It judged each total's net movement, which a means-test fall, or the eligibility-jump exclusion, could use to hide a double count. It now reads each head's response household by household: - a component's own head must rise by exactly £1; - no other tax head may move; - no other benefit may rise in any household (rises and falls are not netted); - the model's own total must move by what its heads moved, plus one for a declared double count. That last rule catches a total whose formula adds a zero component twice. A head that declares a read may respond, but not one for one. The snapshot of supplied inputs is taken after formula results are discarded, so an earlier result is never frozen as an input. Comparison. - A missing or crashed base now fails. - A check failing in a year where the base passes is a regression, even if it already failed in another year. - A check the base ran that is missing from the head fails (or is noted, when the partition changed), and the head must have run the core checks. - A figure on one side only is shown as added or removed, never as no change. - A non-finite head figure fails directly. - Households must match one to one. Measurement. - Finiteness is checked after the reform costings. - A rebuilt baseline must match in every year. - A skipped check is recorded as skipped. - metrics_version covers the effective spec, horizon and reforms. - Geographic exclusivity guards are checked on the data. Test data. - Data tests are identified by their marker, so the data-free tests under microsimulation/ run without credentials. - The setup hook runs before pytest's skip handling, so a skip marker cannot turn missing data into a skip. - The latest-data smoke test is a named exception to the pin. The partition's notes now describe main before the #2159 fix correctly. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Qualification compares a candidate's report with the previous pair's. With only a status per check, a check that fails in more years than before looks unchanged; failing_scopes lets it be told apart. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The tool digest was read from disk when a measurement finished, so editing the tool during a run stamped the result with code that had not produced it. It is now taken when the module loads. Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

MaxGhenis · 2026-10-12T02:24:26Z

Two fixes from this PR's W3 report change the lists fiscal_partition.yaml closes against. Here is what the partition needs once each lands (checked against this PR's head 9b33fe1):

Count Scottish Child Payment, Healthy Start and the Two Child Limit Payment in government spending #2366 (fixes Scottish Child Payment and Healthy Start are household benefits but missing from government spending #2338) adds scottish_child_payment, healthy_start_vouchers and two_child_limit_payment to GOV_SPENDING_VARIABLES.
- Those three heads (around lines 744-770) change from totals: [household_benefits] to totals: [gov_spending, household_benefits].
- Their "missing from gov_spending" model notes go.
- The household_benefits total's description changes. After Count Scottish Child Payment, Healthy Start and the Two Child Limit Payment in government spending #2366 it differs from gov_spending only by pawhp (until Count the Pension Age Winter Heating Payment in household and HBAI income #2047 lands) and other_public_spending_budget_change.
Count statutory sick, maternity and paternity pay as Universal Credit earned income #2365 (fixes Universal Credit rises with SSP and SMP: tax on them is deducted but they aren't counted as earnings #2337) adds statutory_paternity_pay next to statutory_maternity_pay in four lists: HOUSEHOLD_BENEFIT_VARIABLES, GOV_SPENDING_VARIABLES, HBAI_HOUSEHOLD_NET_INCOME_ADDS and hbai_benefits.adds.
- The partition needs a statutory_paternity_pay spending head with totals: [gov_spending, household_benefits], like the SSP and SMP heads.
- uk-data never fills SPP, so no survey aggregate moves.

Neither PR touches tests/conftest.py, aggregate-impact.yaml or reforms_config.yaml.

MaxGhenis and others added 4 commits October 11, 2026 13:43

Document the aggregate-impact report and mandatory test data

cbd2fed

Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

MaxGhenis mentioned this pull request Oct 11, 2026

Stop the test shards' memory growing with every Simulation under sysmon coverage #2331

Merged

MaxGhenis and others added 2 commits October 11, 2026 19:14

Name the measured refs readably in the aggregate-impact report

afa1c2a

Refs #2334. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Merge remote-tracking branch 'origin/main' into aggregate-impact-report

accf433

MaxGhenis and others added 2 commits October 11, 2026 19:24

PolicyEngine deleted a comment from github-actions Bot Oct 11, 2026

MaxGhenis mentioned this pull request Oct 11, 2026

Count corporate SDLT once and let SDLT abolition remove all of it #2235

Draft

MaxGhenis and others added 3 commits October 11, 2026 19:44

Navigation Menu

Repository navigation

Report each pull request's effect on national aggregates on the certified dataset - #2335

Report each pull request's effect on national aggregates on the certified dataset#2335
MaxGhenis wants to merge 11 commits into
mainfrom
aggregate-impact-report

MaxGhenis commented Oct 11, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 11, 2026 •

edited

Loading

Uh oh!

MaxGhenis commented Oct 12, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

MaxGhenis commented Oct 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What it adds

Invariants the report enforces

Tests

What the report already finds on main

Data protection

Thresholds and methodology

Not in this PR

Uh oh!

github-actions Bot commented Oct 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Aggregate impact on certified data: pass

Moves flagged for review

Household net income by decile

Uh oh!

MaxGhenis commented Oct 12, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

MaxGhenis commented Oct 11, 2026 •

edited

Loading

github-actions Bot commented Oct 11, 2026 •

edited

Loading