[juaristi22]

Summary

The UK lock moves from policyengine-uk 2.174.1 to 2.177.4, the newest release on PyPI (microcosm#1095). From 2.177.4 policyengine-uk requires policyengine-core 3.34.9 or later, so the UK side's core moves from 3.33.3 to 3.34.10, the newest; the US side stays on 3.32.5, and uv.lock moves those two packages alone. What each release since the lock needs is in #1095's "The next policyengine-uk lock" (the read of 10 November).

This PR stacks on #1550 (uk-event-data-pr2, at b4250ebe9), as María asked, so its CI runs once it targets main. Until #1550 merges, the pin moves to each later policyengine-uk release: PolicyEngine/policyengine-uk#2807 and PolicyEngine/policyengine-uk#2808 have merged since 2.177.4, and more are in pe-uk's merge queue.

The three new inputs, signed at their defaults

Each is in UK_ENGINE_DEFAULT_INPUTS with the error it can make, for María's signature:

• legacy_prescribed_accommodation_capital_exception (2.174.8), at False: the FRS samples private households, so it holds almost no claimant living permanently in prescribed accommodation. Such a claimant's Income Support, income-based JSA or income-related ESA is understated by the tariff income on capital between £6,000 and £10,000.
• esa_income_appointee_saving (2.175.3), at False: the FRS does not identify appointees, so every reported income-related ESA award ends at the abolition of 1 July 2026. DWP exempted around 500 such claimants. 2025, the calibration year, does not move.
• cares_for_severely_disabled_person (2.177.3), at True: the FRS records caring hours but not whom the carer cares for. At True, caring hours pay as before, so no figure moves.

An input that gained a formula

From 2.176.1, is_before_universal_credit_qualifying_young_person_terminal_date follows the date of birth, so it leaves the engine's input surface (376 to 378 inputs). frs_education still writes it, and the loader takes a stored value over the formula, as it does for any engine variable. Universal Credit reads it only for a 19-year-old, together with the education condition, which is the same test frs_education sets it from. So the stored value and the formula give the same result.

What the build's engine reads return now

• 2.174.2 to 2.175.3: as UK: every open data item: uk-data parity, engine locks, the local surface, consumption and high-value homes #1095 lists them. The qualifying young person rule for 16-year-olds; the State Pension age addition; the benefit cap reduction; the Universal Credit deduction flag; reg 60A's Working Tax Credit test; Maternity Allowance in the cap; and the income-related ESA abolition, which moves nothing before July 2026.

• 2.175.5, 2.175.8 and 2.177.2 (Council Tax Reduction):

  • the pensioner schemes in England, Scotland and Wales add child amounts and family premiums;
  • all of a Scottish pensioner's capital over £10,000 counts in tariff income;
  • Scotland's and Wales's working-age schemes get their own rules;
  • Scotland's counts Carer Support Payment after the overlapping-benefit reduction.

  The council tax rows read council tax net of the reduction.

• 2.176.0 (Model rent from boarders, lodgers and sub-tenants for the householder policyengine-uk#2002): rent from boarders, lodgers and sub-tenants for the householder.

  • Income tax gives rent-a-room relief on the boarders' and lodgers' rent, which the engine sums for the household head from the rent_paid_as_boarder and rent_paid_as_lodger the build writes.
  • The legacy means tests apply the £20 and board-and-lodging disregards.
  • A boarder or lodger draws no non-dependant deduction.
  • The build keeps the FRS sub-letting rent in property_income.

• 2.176.1: the qualifying young person's terminal date, above. It now applies to Pension Credit too.

• 2.176.2:

  • Housing Benefit's and national Council Tax Reduction's non-dependant deductions follow each scheme's rules.
  • Universal Credit changes its LHA bedrooms and its renter exemption.
  • Universal Credit's benefit cap counts welfare benefits without Housing Benefit. A dataset that supplied benefit_cap_welfare_benefits would now also need Housing Benefit before the cap; the build supplies neither.

• 2.177.0: VAT's survey coverage moves from 0.67 to 0.69. obr.vat stays held.

• 2.177.2: Marriage Allowance becomes the recipient's tax reduction, and Scottish taxpayers' tax bands apply the Scottish limits to income other than savings and dividends.

• 2.177.3: Plan 5 borrowers repay nothing before 6 April 2026, the date the student-loans stage already uses. slc.borrowers.plan_5_above_threshold then matches SLC's nil, and slc.repayments.england_plan_5, held, reads nil.

• policyengine-core 3.34.10: fixes to cached periods and to inputs on reform branches. The UK side neutralises no variable and uses none of the APIs 3.34.0 removed.

• The other releases change nothing the build reads at 2024 or 2025, or nothing at all.

What does not move

• The vendored uprating pins and the bonus-sacrifice anchor change only their version stamps. The CGT projection path keeps its values, within the pinned drift (test_uk_cgt_projection.py).
• UK_HBAI_GROSS_INCOME_DEDUCTIONS keeps its parity. 2.176.0 adds sublet_income to HBAI income, not to its subtractions, and the build keeps that income in property_income.

Texts and derived files

Every text that described the locked engine was checked against the diff from 2.174.1 to 2.177.4 and now cites 2.177.4. That covers the stage notes, code comments, test docstrings, the ISA and LISA docs, and #1550's PIP points, limited capability and trading allowance texts.

The take-up stage's note had said the build flags nobody in full-time education. Since #1535 it does, and the note now says so, as the code comment does. The coverage manifest tool's PIP notes and the eFRS known gaps follow.

The concept coverage report, the release input coverage manifest, the H2 fixture, the cited versions and APPROVED_UV_LOCK_SHA256 are regenerated.

Measurement

Deferred until #1550 merges (María's call). Then two arms run on this machine, one at a time: a control on main after #1550, and this branch's head. Each is a spine and a calibration, at a review date of 10 November 2026 or later on #1550's feed. Their results follow in a comment: both arms' gates, the fit, the first stage where the spines differ, and the rows each release moves, with rows of the SPI channel as classes.

Tests

• Engine tier on policyengine-uk 2.177.4: CI's engine-uk list (72 files), run in full locally before the last rebase onto UK November event data, PR2: feed re-pin, trading allowance claim status, Scottish council tax and NSND, young people and PIP rows, UC take-up age fit, PIP points, UC limited capability #1550, because CI stops at the first failure. UK November event data, PR2: feed re-pin, trading allowance claim status, Scottish council tax and NSND, young people and PIP rows, UC take-up age fit, PIP points, UC limited capability #1550's commits since then change the PIP points receipt, notes and tests, and their engine files pass below. 315 tests: 313 passed and 2 failed, and 5d6a3e7 fixes both. test_uk_tax_band_measure.py had pinned a known gap "so a fix shows", and 2.177.2 is that fix. The frozen eFRS candidate receipt measures the terminal-date column against its engine default, which 2.176.1 flips; the receipt keeps its original engine provenance. Those files and UK November event data, PR2: feed re-pin, trading allowance claim status, Scottish council tax and NSND, young people and PIP rows, UC take-up age fit, PIP points, UC limited capability #1550's touched engine files then pass (22 tests). The bump's checklist tests pass: the CGT projection, bus parity, UK_HBAI_GROSS_INCOME_DEDUCTIONS, the uprating pins, the bonus anchor and all 13 concept-mapping tests.
• Engine-free files on the touched surfaces: 81 files, run in three processes and counted by --junitxml. 3,904 tests: 3,877 passed, 25 skipped and 2 failed.
  • The first failure was UK November event data, PR2: feed re-pin, trading allowance claim status, Scottish council tax and NSND, young people and PIP rows, UC take-up age fit, PIP points, UC limited capability #1550's own stale stage-order pin, which 8780aa3 fixes; the file passes on the rebased head (24 tests).
  • The second was test_us_release_head_to_head_scorer, whose guard refuses this machine's Python 3.14.6.
• Checks:
  • tools/ci_test_plan.py verify;
  • tools/spec_engine_coverage.py --check (42,394 of 42,394 fields, 41 of 41 inventory checks);
  • tools/refresh_concept_coverage.py --engine policyengine-uk --check;
  • both pin tools;
  • the coverage manifest with --check;
  • the H2 fixture regenerated on the rebased tree and reproduced byte for byte by --check; the H parity test (1 test) passes.

Signatures

Awaiting María's signature: the three signed defaults, with the error each can make.

Not in this PR

• The SUBRENT move to sublet_income, which 2.176.0 now allows (Move SUBRENT from property_income to sublet_income, after policyengine-uk reads sublet_income policyengine-uk-data#508's order).
• The obr.vat rebind, whose condition in María's ruling of 24 October 2.177.0 meets.
• The restated reason for the Plan 5 repayments row.
• policyengine-uk releases after 2.177.4: the pin moves to them in this PR before UK November event data, PR2: feed re-pin, trading allowance claim status, Scottish council tax and NSND, young people and PIP rows, UC take-up age fit, PIP points, UC limited capability #1550 merges.

Disclosure

No figure in this PR, its commits or its changelog rests on survey records. The numbers are policyengine-uk's parameters and versions, and published figures cited as their publishers'.

🤖 Generated with Claude Code