[juaristi22]

What and why

Budget 2026 raises every company car appropriate percentage by one point in 2030-31: Table 6.1 row 86, +£310m in 2030-31 and +£345m in 2031-32 (OBR split: income tax 226.2, NICs 92.4). The dashboard can't score it (PolicyEngine/uk-autumn-budget-dashboard-2026#316), because neither the engine nor the data had company cars.

• The engine. Model the company car and car fuel benefits, with Class 1A NICs policyengine-uk#2750 shipped in 2.165.0. It taxes the company car and car fuel benefits, and the Class 1A NICs on them, from nine person inputs that default to no car.
• The data (this PR). It adds a stage, company_cars, that fills six of those inputs, on UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's policyengine-uk 2.165.1 lock.

Without the stage, a file on 2.165.0 still scores the rise at £0.

Status

1. The HMRC facts are in. The feed is re-pinned to 08962f1, the merge of Add HMRC benefits in kind statistics, July 2026, and keep one OBR EFO number per line and year chronicle#399 on top of UK facts for chronicle#395: Northern Ireland domestic rates, Welsh bus journeys, the Childcare Offer for Wales, VED by payer and CPI gas weights chronicle#396 (1,512,256 rows; the schema is unchanged). uk/hmrc_benefits_in_kind_facts.json vendors the 96 2024-25 rows the stage reads, and is registered in country_package.json.
   • The other 35 vendored resources, and the DESNZ constituency resource, carry the same values under the new identity.
   • The local surface is unchanged apart from its feed label and the DESNZ keys.
   • The coverage manifest and the H2 fixture are regenerated.
2. The OBR windows read both editions. Full eCPS input coverage: 158 required columns, 8 evidence-backed exclusions #399 keeps one OBR EFO number per line and year (OBR EFO March 2026: drop the years the October 2026 vintage also publishes (PolicyEngine/microcosm#1095) chronicle#397), so March 2026 now carries only FY2024-25 and October 2026 carries FY2025-26 on. The 19 OBR targets that window both years into calendar 2025 could no longer compile, since their window took both years from one series. A new selector key, source_table_by_opening_year, takes FY2024-25 from March's table and FY2025-26 from October's (María's ruling of 31 October).
   • Fifteen of the 19 move by under 1%: income tax −0.5%, UC +0.7%. The four larger moves are domestic rates (+3.2%), Pension Credit (+2.3%), Jobseeker's Allowance (−2.5%) and SDLT (−1.2%).
   • Three report-only diagnostics pinned March's FY2025-26 forecasts, which the feed no longer carries: obr.capital_gains_tax, obr.state_pension and the obr.ni total. They now pin October's outturns. None of them is in the fit.
   • obr.winter_fuel_allowance reads October's FY2025-26 (+0.02%).
   • The national membership also catches up with policyengine-uk 2.164.1's October-EFO indices, which Budget 2026 ISA data, PR B of #1418: ISA subscriptions per adult, the Lifetime ISA anchor, additive ISA income, policyengine-uk 2.164.2 #1421 locked after Budget 2026 slower data fixes, PR C of #1418: electricity at the ONS level, pension income by area #1420 had generated it: the eight SPI Table 3.10 self-employment bands move −0.55% and the thirteen dividend bands under 0.001%. The compile-parity receipt, which builds check against, already had these values.
   • The national references, membership and compile-parity receipt are regenerated from the 08962f1 national cut.
3. UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422 is in (11c1f4e), and main is merged here, so the diff is UK: company cars from the FRS job table (dashboard #316) #1430's own. This branch builds on UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's policyengine-uk 2.165.1 lock, its targets and its gates.
   • Every shared surface is regenerated on 08962f1: the vendored resources (byte-identical), the national and local references, memberships and compile-parity receipts, the coverage manifest, the H2 fixture and the concept coverage.
   • On 08962f1, UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's three OBR council tax nation rows take the same edition map as the other windowed OBR rows, and so does the signed-out obr.fuel_duties: 23 rows in all.
4. The reviews. The independent review (issuecomment-6545997069) is addressed, with María's rulings of 31 October on items 2, 3, 7 and 10, and so are the second (issuecomment-6561698051) and third (issuecomment-6564905160).
5. The build. A pair build on 2.165.1 against a control on the same base follows the merge, and its results are added below. The machine order is María's.

Merge order

• UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422 is in. Main (11c1f4e) is merged here.

• What this PR reconciles. UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422 stayed on Chronicle 53815b2 (María's call), so the feed pin and the table maps arrive with this PR. This PR also regenerates every shared surface on 08962f1.

• SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 lands after this PR and picks this code up, and builds run from there (María's call). On merging main, SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428:

  • switches its hmrc.spi.employment_income.count_by_self_employment_income_band row, its test and its Table 3.10 cross-grain reason to employment_income_with_benefits_in_kind_annual, which a test here now requires of every HMRC employment income row;
  • re-vendors its Table 3.7 savings rows on 08962f1, with its membership and parity receipt;
  • re-pins the stage count at 52;
  • regenerates the gate-battery pins, the coverage manifest and the H2 fixture.

  Its b9bcd709c removes ISA's empty fit-weight records, which the weights audit needs.

• Company car inputs. UK_ENGINE_DEFAULT_INPUTS signs only the three inputs the stage leaves at the engine's default: availability, private-use payments and the 6 April 2020 registration flag. Its reason states the full-year approximation. UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's guard test counts every stage's declared outputs and rewrites.

Changes

1. policyengine-uk 2.165.1, UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's lock. 2.165.0 (Model the company car and car fuel benefits, with Class 1A NICs policyengine-uk#2750) brings the company car inputs. 2.164.3 also counts other_investment_income in market and HBAI income, which the spine fills (UK: data fixes for the Budget 2026 household measures #1418), so HBAI-based outputs move too.

2. The FRS job tab's company car answers. The root stage carries four columns per person from their car job with the lowest JOBTYPE (one car per person, as in the engine):

   • the CARVAL band, or "unknown";
   • FUELTYP, or "unknown";
   • EXPBEN03 (private fuel);
   • CARAMT where CARCON says the employee contributed.

   They are dropped at the release boundary.

3. The company_cars stage runs right after the SPI income stage (hmrc_spi_income_spine; María's ruling of 31 October). Every later stage reads the tape rows' pay without the car benefits, as the engine's employment_income excludes them: salary_sacrifice and its bounds, the take-up draws' earnings rules and the CGT and ISA stages' income proxy. Every draw is keyed on identity. The stage fits no model and declares no fit-weight records.

   • Holders. FRS-channel people keep their car. A row drawn from the SPI tape (person_is_spi_income_recipient) keeps its twin's only where its drawn EPB is positive. The channel's undrawn copies and dependants keep their twin's pay and car, as FRS rows do.
   • Fuel. From FUELTYP: a hybrid is an alternative fuel; a diesel meets RDE2. The unknown fuels draw at HMRC Table 2C's shares together: a holder who reports private fuel draws among petrol and diesel, and the others draw electricity at the share that keeps Table 2C's. The tape rows that lose their car draw at the holders' share.
   • List price. Drawn within the CARVAL band from WAS round-8 company cars' list prices, grown by the engine's CPI index from 2021. The open and unknown bands draw by fuel class. A cell with fewer than 3 WAS households (the floor of 3) is drawn uniformly, from no lower than £5,000.
   • CO2. Zero for an electric car. For the others, HMRC Table 2A's shares among cars that emit CO2, hybrids first into 1 to 50 g/km, uniformly within each band; a car at 1 to 50 g/km takes a 50-mile electric range.
   • Defaults kept. Availability, payments for private use and the 2020 registration flag keep the engine's defaults.

4. The tape rows' pay. It is the tape's PAY + EPB + TAXTERM, so the engine's car and fuel benefits come out of employment_income, capped at EPB (María's ruling of 30 October) within the pay the row's employee pension contribution leaves. The SPI income stage capped that contribution at the drawn pay, and salary_sacrifice refuses one above the pay that remains; a row whose contribution takes all its pay loses its car. The receipt measures what the room costs beside a cap of EPB within pay alone: the cars it scales further or drops, and the list price they give up.

   • Where the car and fuel benefits don't fit together, the fuel goes first, so a car that fits without it is kept (María's ruling of 31 October).
   • A car that still doesn't fit has its list price scaled down, with its contributions dropped. A scaled price is the room over the car's appropriate percentage, so the CO2 draw, the 50-mile range and the RDE2 diesel move the 2030-31 base through these cars.
   • Below £5,000 the car is dropped.

   The stage then has the engine price its output again: on every tape row, pay plus the engine's benefits must equal the pay the row brought in. A carved row's UC gainful self-employment, which the SPI income stage set on the drawn pay, takes the root's rule again on the carved pay. Employer contributions stay as the SPI income stage set them on the drawn pay, as on every tape row that draws EPB.

5. Targets.

   • hmrc.spi.employment_income.amount_by_total_income_band and the local hmrc.employment_income.amount read employment_income_with_benefits_in_kind_annual, a new named term for employment income plus benefits_in_kind, which HMRC's employment income counts.
   • The two count rows filter on the same term, so a tape row whose pay the carve takes to nothing still counts.
   • The annual simulation makes a car available for the share of the year the pay covers.
   • obr.ni_employer binds ni_class_1_employer. OBR's line is Class 1 employer NICs (table 4.4 row 18), and Class 1A sits in row 21.

6. The facts. hmrc_company_car_statistics.py reads Full eCPS input coverage: 158 required columns, 8 evidence-backed exclusions #399's 2024-25 rows by their source record ids, in persons and pounds:

   • a cell the feed lacks is refused, never read as zero;
   • a resource whose Table 2A or 2C total, or electric count, sits from Table 1 beyond each sheet's rounding is refused too.

7. Gates. uk_stage_company_cars_invariants is release-blocking and runs on the synthetic smoke too. It fails on:

   • a non-finite or negative value;
   • a holder without a price;
   • inputs on a person without a car;
   • an unknown fuel;
   • a CO2 figure missing for a car that emits it, or set for an electric car;
   • private fuel on an electric car;
   • contributions above the price;
   • an electric range on a car outside 1 to 50 g/km;
   • an SPI benefit above EPB;
   • a tape row's employee pension contribution above its carved pay;
   • a tape row's pay plus the engine's benefits, priced again on the output, more than 5p from the pay the row brought in (single precision).

8. Not bound. No HMRC company car fact is a calibration target: the stage draws to Tables 2A and 2C, so binding them would be circular (María's ruling of 28 October on the ISA facts). Comparing the calibrated file with Tables 1, 3A and 4 is a manual step for now.

Results

PENDING the pair build, which follows the merge. Every figure is checked for disclosure before it goes in.

• Holders against HMRC's 920,000 (2024-25 provisional), and by income range against Table 3A.
• Car and fuel taxable value, income tax and Class 1A against Table 1 (£3,070m, £1,090m, £420m).
• The list price base, electric and not, measured on the calibrated file. The stage's receipt (list_price.weighted_total_gbp) is at its pre-calibration weights. Two published comparators frame it: HMRC's 2024-25 taxable value implies roughly £27bn to £37bn, and the OBR's NICs line about £60bn.
• What the base loses before calibration, from the receipt: the holders and list price the tape rows lose for want of EPB (holders.spi_dropped_without_epb; María's ruling of 31 October: measure first), and the list price the carve-out scales away or drops, including what the employee contribution's room takes.
• The static 2030-31 cost of COMPANY_CAR_TAX_RISE_2030 against Table 6.1 row 86 (+£310m). Expect it well under £310m, from the size of the base (docs/uk-company-cars-316.md, "What to expect of the 2030-31 score"). The file holds the build year's fleet: fuel and CO2 are frozen, prices grow by CPI, and the weights are not projected. A file that matches Table 1 exactly should still fall about 30 to 45% short.
• Gate and fit movement:
  • the re-pointed SPI rows and obr.income_tax. obr.ni_employer now measures Class 1 alone, but it is held out of the objective as a diagnostic until 2027-01-16, so the fit doesn't move with it;
  • HBAI income. 2.164.3 adds other_investment_income to market_income, household_market_income and hbai_household_net_income, and 2.164.4 to capital_income. Those feed stage predictors, so the draws below can move for households with other investment income:
    • was_wealth, which materializes household_net_income and capital_income;
    • was_lisa (household_net_income);
    • lcfs_consumption and etb_services (hbai_household_net_income).

Tests

• New:
  • test_uk_company_cars.py, engine-free and engine. It covers the reviews' cases:
    • fuel first, the private-fuel draw, the unknown fuels meeting Table 2C together, and the lost cars drawing at the holders' share;
    • band 1's floor, the EPB cap, and the room for the employee pension contribution with what it costs;
    • salary_sacrifice run on the stage's output;
    • the UC gainful self-employment refresh, in the helper and through the stage, and the fit-weight evidence;
    • a declaration naming the old unknown-fuel rule refused;
    • the fence count;
    • the receipts for the EPB shortfall, the scaling and the output's CO2 shares, against the draw's;
  • test_uk_company_cars_stage_health.py;
  • the carrier tests and the code-view test in test_uk_frs_spine.py;
  • the named term and the annual car availability in test_uk_measure_simulation.py;
  • the ni_class_1_employer binding and its exclusion in test_uk_national_insurance_targets.py;
  • every HMRC employment income row reading the benefit-in-kind term in test_uk_population_targets.py.
• The UK engine tests on 2.165.1 pass: company cars, coverage, the annual measures, concept mapping, the graph and the H2 parity, 75 tests. They show that:
  • the engine reads the stage's inputs;
  • a tape row's pay plus the engine's benefits is the tape's total to the penny;
  • the 2030-31 rise reaches the holders alone.
• After the third review, on main with UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422 (head de41aba):
  • the 27 touched engine-free files pass: 1,442 tests, 8 skipped;
  • the coverage manifest and the H2 fixture are regenerated for the renamed fuel rule.
• After the second review, which last touched the targets and the feed:
  • the feed-gated files, one process each against the pinned feed (CI skips them): the 15 other than bus targets, 239 passed and 2 skipped (the Scotland support and the authenticated national artifact aren't mounted here);
  • the generated surfaces are regenerated from the feed's cuts. The national compile-parity receipt, the membership's fuel duties entry, the coverage manifest and the H2 fixture changed as the fixes predict. The national references and the local receipt regenerated byte-identical.
• The full engine-free suite, run locally in chunks on the merge with UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's final head (4af602d): 14,487 tests, of which 14,332 pass and 152 skip. Three local-only checks fail: the uk-extra-absent test (this environment installs the extra), a live release-pointer test, and the orrery check, whose Node package isn't installed here. CI runs the full suite on every push.
• Before UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's third review, all 16 feed-gated files passed (252 tests, bus targets included), as did the national compile-parity regeneration on the full pinned artifact, with its release gate.
• tools/run_integration_tests.sh was not run locally; CI's integration-uk runs it.

Disclosure

The FRS and WAS are UK Data Service data (SN 9563, SN 7215). The stage's receipt carries weighted aggregates, rounded counts and support bands only. The HMRC figures quoted here are published.

Not in this PR

• Class 1A is not passed through to households (Pass Class 1A NICs through the employer NI incidence policyengine-uk#2752). The 2030-31 rise is scored statically, comparable with the OBR's costing.
• Vans, fuel provided for part of the year, optional remuneration arrangements, and the rest of the "not modelled" list in policyengine-uk#2750.
• Whether to anchor holders to Table 3A by income range, if the FRS count falls short of HMRC's. That is María's call after the results.
• A build step that compares the calibrated file with HMRC's Tables 1, 3A and 4; for now it is manual.
• Employer pension contributions on a tape row's pay without EPB: the SPI income stage sets them on PAY + EPB + TAXTERM for every tape row, car or not.
• The dashboard measure: policyengine.py certifies a build on ≥2.165.1, then the dashboard adopts the release and adds the measure.

🤖 Generated with Claude Code

[vahid-ahmadi]

Review of 21e3623 (draft, so an early look as a comment). First round; there are no earlier comments.

The stage writes what policyengine-uk 2.165.0 expects. I checked the nine inputs at the 2.165.0 tag. The six written ones have the right units, periods and enum names (GBP stocks uprated by CPI, g/km, miles, CarFuelType names, bool). The draws are keyed on identity, there is one car per person by lowest JOBTYPE, household weights pass through, and no company car fact is bound. The stage is the last spine stage, so calibration still runs after it: it adds no weights, but the solve then moves holders, which fits "measured, not bound". There are four shoulds and one nit.

Findings

1. should: on a frozen fleet, expect the static 2030-31 score well under £310m. The gap comes from the size of the base, not the EV share.

• The rise is one point in every band, zero-emission and the maximum included. So the extra taxable value is about 1% of total list price (less contributions), plus 1% of the fuel multiplier for each fuel-benefit holder. The frozen EV share barely moves the rise. It moves the baseline level instead (EVs go from 3% to 9% by 2029-30).
• The base the rise acts on is the build year's fleet, carried forward to 2030-31. Fuel type and CO2 are not uprated, list price and contributions rise by CPI, and the weights are not projected.
• Rough HMRC-implied base for 2024-25: Table 1 car taxable value £3,070m over Table 2A's band-weighted 2024-25 appropriate percentages gives about £27bn to £37bn of list price (depending on PHEV range and EV price premium). With CPI to 2030-31 and 40,000 fuel holders, that is about £330m to £440m of extra taxable value. At HMRC's 2024-25 ratios (income tax 35.5% of taxable value, Class 1A 15%), that is roughly £170m to £225m.
• The OBR's NICs line (£92.4m / 15%) implies about £616m of extra taxable value, roughly a £60bn base. Its income tax line gives the same 36.7% ratio. So even a file that hits HMRC's 920,000 holders and £3,070m exactly should come in around 30 to 45% short. Recipients rose from 840,000 to 920,000 in one year, and HMRC flags under-reporting of payrolled cars.
• Matching Table 1's taxable value does not pin this base: a file can hit £3,070m with a different EV mix and different prices. Suggest the receipt add weighted total list price (by electric and not), so Results can compare it with the HMRC-implied range and read the £310m gap as base size.

2. should: the SPI pay-identity check in the release-blocking gate can't fail.
company_cars.py:983-987 sets new_pay = pay - carve.benefit, then measures |new_pay + carve.benefit - pay|, so max_abs_pay_plus_benefit_error_gbp is zero by construction. That means uk_stage_company_cars_invariants (stage_health.py:4954-4960) never tests the claim in its notes. The real risk is the engine's benefit on the written inputs differing from the carved amount. The price scaling assumes the defaults for availability and private-use payments, and that contributions were dropped, so a later engine change would break it without the gate noticing. Suggest one more materialize of company_car_benefit + car_fuel_benefit on the output frame, comparing each SPI-channel row with carved.

3. should: one CI failure is not on the body's expected-red list.
engine-uk fails on tests/engine/uk/test_uk_annual_income_measures.py:156: set(receipt["computed_measures"]) == set(UK_ANNUAL_INCOME_MEASURES). The test never computes the new employment_income_with_benefits_in_kind_annual. This is independent of chronicle#399. Computing it there, and asserting it equals employment_income_annual when nobody has a car, would fix it. The job stops at this first failure ("1 failed" and nothing passed), so CI has not run the new engine tests in test_uk_company_cars.py yet.

4. should: the lock step pulls in more than the body lists.
2.164.2 → 2.165.0 includes 2.164.3 (other_investment_income in market_income, household_market_income and hbai_household_net_income) and 2.164.4 (also in capital_income). The body mentions only HBAI outputs. These engine values are stage predictors:

• was_wealth materializes household_net_income and capital_income (was_wealth.py:63-89);
• was_lisa reads household_net_income (was_lisa.py:164);
• lcfs_consumption and etb_services use hbai_household_net_income (lcfs_consumption.py:650, etb_services.py:62).

So the wealth, LISA, consumption and services draws can move for households with other investment income. Suggest adding these to Results' "gate and fit movement" list for the pair build. 2.165.0 also adds Class 1A to ni_employer, which the body already covers.

5. nit: unknown-fuel draws never produce a hybrid.
company_cars.py:588 draws only ELECTRIC, PETROL or diesel, and :920 flags hybrids from the FRS code alone. So "petrol/diesel including hybrids" from Table 2C never feeds the hybrids-first split into 1 to 50 g/km. The effect is small, but worth a sentence in the doc.

Checked and fine: the three defaults (availability 1, no private-use payments, the 2020 flag) only matter in 2020-22 or slightly raise each car's taxable value for part-year holders; the WAS donor (cars not vans, household weight, floor of 3); CARVAL/FUELTYP mapping and the open band; CO2 bands within Table 2A's bounds; the SPI carve-out arithmetic (fuel first, exact price scaling, £5,000 floor); the re-pointed SPI targets, since taxable_employment_income now includes benefits_in_kind, so total-income bands agree.

Verdict

Sound design. Items 2 and 3 are small code fixes. Item 1 is about setting expectations before the pair build, and item 4 is about what that build should report. Not ready until the #399 vendoring, the regenerations and the build, as the body says.

What I ran

Fresh clone at 21e3623 and uv sync --all-packages --locked (no UK engine extra), deleted afterwards; one process at a time.

• Engine-free pytest: test_uk_company_cars.py and test_uk_company_cars_stage_health.py gave 24 passed. With test_uk_frs_spine.py, test_uk_measure_simulation.py, test_uk_source_stages.py, test_uk_terminal_gates.py, the ISA, local-target, private-school and stochastic tests, test_engine_core_pins.py, test_country_spec.py and microcosm-data's test_contract.py, 848 passed and 4 failed. All 4 are KeyError: 'hmrc_benefits_in_kind_facts.json', as expected.
• I read the 2.165.0 variables, parameters and the company_car_tax_rise_2030 scenario at the tag, and the 2.164.2...2.165.0 compare.
• CI: engine-free fails on the missing facts and integration-uk on the same resource (expected), engine-uk as in item 3; lint, wheels and engine-us pass.

Not verified: the engine tests (no UK engine locally, no build or Modal), any FRS or WAS figure, and the score itself. Item 1's figures are rough arithmetic from published HMRC and OBR numbers.

🤖 Generated with Claude Code

[juaristi22]

Thanks, Vahid. All five items are addressed. Item 2 led to one more fix, and to a disclosure pass on the stage's receipt.

1. The 2030-31 expectation and the list price base.

• The receipt now records list_price.weighted_total_gbp for electric cars and for the others, each with its support band.
• docs/uk-company-cars-316.md has a new section, "What to expect of the 2030-31 score". It says why the rise acts on list price, why the frozen electric share moves the baseline but not the rise, and why the static cost should come in well under £310m: the OBR's NICs line implies a base of about £60bn, against roughly half of that from HMRC's 2024-25 taxable value. It credits your arithmetic for the 30 to 45% range.
• The body's Results now compare the receipt's base with that range, so the gap reads as base size.

2. The pay identity check couldn't fail.

• The stage now prices its output again on the engine. On every SPI-channel row, pay plus the engine's company_car_benefit and car_fuel_benefit on the written inputs must equal the pay the row brought into the stage.
• The receipt records the largest miss as spi_carve_out.max_abs_pay_identity_error_gbp, rounded up to the penny. uk_stage_company_cars_invariants allows 5p, for the engine's single precision (maximum_pay_identity_error_gbp: 0.05). The battery and certification digests are re-pinned for that gates.json change.
• A new test runs the stage with an engine stub whose car benefit ignores the list price. The check reads a miss above £1,000, so it fails as it should.
• While doing this I found one more thing to fix. salary_sacrifice keeps employee contributions and salary sacrifice within pay, and bonus sacrifice within half of it, and its comment asks any later pay rewrite to hold those bounds. The carve-out didn't. Its budget now also leaves the pay those deductions need, and the receipt counts the rows where that binds (limited_by_pension_deductions).
• The disclosure pass covers every weighted figure in the receipt:
  • each one is withheld when a cell it rests on has fewer distinct source households than the floor, so support copies count once;
  • shares that sum to one are withheld together;
  • invariant counts are zero exact and banded otherwise.

3. The engine-uk failure. test_uk_annual_income_measures.py now computes employment_income_with_benefits_in_kind_annual and checks that it equals employment_income_annual when nobody has a car. Locally on 2.165.0, that file gives 4 passed and the company cars engine tests 3 passed.

4. The 2.164.3 and 2.164.4 predictor moves. They are in Results' gate and fit movement list:

• was_wealth (household_net_income, capital_income);
• was_lisa (household_net_income);
• lcfs_consumption and etb_services (hbai_household_net_income).

5. Hybrids. The doc now says that an unknown fuel never draws a hybrid, since Table 2C counts hybrids inside petrol and diesel, and that the effect is small.

What I ran:

• engine-free company cars and stage-health tests: 28 passed;
• the engine tests above;
• the contract, gate battery pin, country spec, release certification and terminal gate tests: 523 passed.

The chronicle#399 re-pin and the BiK vendoring are pushed alongside. So is the fix for the 19 OBR windows that #397 split across the March and October EFOs: they take FY2024-25 from March and FY2025-26 from October (María's ruling of 31 October). The body's Status says what each moved.

🤖 Generated with Claude Code

[juaristi22]

Independent review of #1430 at e67e76c29

This is an independent adversarial review. Four reviewers each took one area:

• the stage;
• facts and windows;
• a cross-cutting pass;
• conventions.

A verifier then reproduced each finding on synthetic data or published tables, and a last pass looked for gaps. Vahid's round isn't repeated. Nothing here uses licensed data. Paths are relative to packages/microcosm-build/src/microcosm/build/ unless they start with tools/ or docs/.

Verdict: two things to fix before the pair build (1 and 2), and two that will skew how its Results read (3 and 4).

Fix before the build

1. obr.ni_employer now counts Class 1A; OBR's line doesn't.

• The row binds ni_employer (uk/uk_population_targets.json:2111). In policyengine-uk 2.165.0 that variable adds ni_class_1a, which is the rate × benefits_in_kind.
• OBR's line is Class 1 employer NICs only. Class 1A sits in "Other NIC" (October EFO table 4.4, row 21, footnote).
• Once this stage fills the car inputs, the measure gains Class 1A on them, about £420m in 2024-25 on HMRC Table 1, and the solve moves weight to absorb it.
• Fix: bind ni_class_1_employer. The notes at :677 and :34829 still say no household carries Class 1A.

2. The carve-out breaks the identity-stability tool.

• tools/verify_uk_identity_stability.py:2555 requires that pay plus salary sacrifice equal the pre-conversion pay exactly, on every converted record.
• company_cars lowers SPI pay after salary_sacrifice has run, and keeps no carrier for the carved amount. salary_sacrifice converts SPI rows too (salary_sacrifice.py:658).
• So a converted SPI person with a car makes --check e4, e5 and e8 raise. At full scale that overlap is near-certain.
• e8 also re-bands CGT on pay the CGT stages never saw.

What the Results will measure

3. SPI copies keep a car only on an independent EPB draw (uk_runtime/company_cars.py:991).

• A copy keeps its twin's car only where its drawn EPB is positive. EPB is drawn on identity-keyed uniforms that are never conditioned on the car (spi_rank_transport.py:123; spi_income.py:168-175).
• The SPI channel holds half of the prior working-age mass (sources.yaml:1415-1416). So before calibration, holders and list price come to about (0.5 + 0.5p) of the FRS level, where p is the share of copies that draw a positive EPB.
• No car target restores this.
• The receipt gives only a rounded row count (spi_dropped_without_epb). Meanwhile the doc's "What to expect" puts the whole shortfall against £310m down to base size.
• Suggestion: record a weighted figure for the dropped holders, or condition the keep on the twin; either would make this measurable.

4. Assumptions documented as "baseline-only" drive the 2030-31 base, through the price scaling (:918-925).

• A scaled SPI car's price is the room ÷ the appropriate percentage at 2024-25 rates, and the 2030-31 rise is 1% of price.
• On a £2,000 room and a £45k car, the scaled price is:
  • £25,000 at 8% (the fixed 50-mile range);
  • £14,286 at 14% (under 30 miles);
  • £45,000, unscaled, for an electric car at 2%;
  • £8,000 for an RDE2 diesel at 25%;
  • £6,897 at 29% for a diesel without RDE2.
• That contradicts sources.yaml:6360 ("matters only to 2027-28") and docs/uk-company-cars-316.md:40-41 and :86-87.

5. The CO2 receipt predates the carve that skews the fleet (:1009-1017, :1147).

• The carve drops or scales cars in inverse proportion to their appropriate percentage. So electric and low-CO2 cars survive, and high-CO2 cars drop first.
• On synthetic data, the kept 1-50 g/km share rises from 0.39 to 0.50, and the electric share from 0.25 to 0.38.
• The receipt reports the pre-carve shares, so "matches Table 2A" isn't true of the output.

6. The list-price base is read at pre-calibration weights (:1134).

• list_price.weighted_total_gbp is at the stage's weights, which carry the SPI channel's prior 50%. Calibration moves holders after this point, and nothing measures list price on the calibrated file.
• Yet the doc (:96-99) and the body plan to read it as the score's base.

Stage behaviour

7. Fuel-first ordering drops cars that would fit without fuel (:910-928).

• With EPB of £9,000, an appropriate percentage of 29%, a £35k car and an £8,062 fuel benefit, nothing is kept. Without the fuel flag, the car is kept at £31,034.
• This follows the 30 October ruling's wording literally, but the result isn't monotone.
• Related: an unknown-fuel holder who reported private fuel draws electric at Table 2C's share (about 51%) and so loses that fuel (:1018).

8. Count rows drop SPI rows whose pay the carve takes to zero.

• The employment-income counts filter on employment_income_annual != 0 (uk/uk_population_targets.json:3696, :31014), while the amounts read the new term.
• An SPI row whose pay is all EPB can end with pay 0 and its car kept: carve_spi_benefits(epb=2000, pay=2000, retained=0, car=6000) does this. Such a row leaves the counts but stays in the amounts.
• Suggestion: filter the counts on the new term.

9. The band-1 fallback draws prices from £0 (:197, :735).

• With no band-1 WAS donors, prices are drawn uniformly on (0, £10,000]. About half fall below £5,000, the least price the carve treats as plausible.
• A CARAMT capped at a near-zero price then zeroes the benefit.

10. Carving last leaves upstream pay rules unrestored (:188-192).

• These were all drawn on pre-carve pay:
  • the bonus-sacrifice candidacy (pay ≥ UEL, salary_sacrifice.py:286);
  • UC take-up awards;
  • the Carer's Allowance earnings test;
  • the gainful self-employment flag.
• So "nothing reads the pay it rewrites" holds only for later stages.
• Suggestion: split the car part out of EPB inside the SPI income stage. That keeps these rules consistent and also clears item 2.

11. SPI-channel rows that aren't tape rows lose the twin's car (:538-543, :992).

• Undrawn copies and dependants aged 16 to 19 keep their twin's pay with EPB 0, so no_epb drops the car.
• spi_income documents person_is_spi_income_recipient for exactly this distinction, and property_components uses it.

Coverage and texts

12. company_cars has no coverage-manifest family.

• Of the 15 stages that declare record_mass_conservation_receipt, it is the only one without a family_coverage entry. tools/build_uk_release_input_coverage_manifest.py:931 lists isa_subscriptions but not this stage.
• So release_input_coverage.py never checks its mass record, nor requires it in the plan.

13. Small items:

• the annual measure scales pay by employment_income_annual_share but not the car inputs, which puts a full-year benefit on part-year FRS pay (small);
• the BIK fact fence's count is popped but never compared;
• company_cars.py:196-216 re-declares FRS constants from frs_spine.py:957-963 (one of them unused), against docs/shared-constants.md;
• _STAGE_HELPER_MODULES["company_cars"] omits uk_engine_adapter and salary_sacrifice;
• the stage notes omit the pension-deduction cap and the re-pricing check;
• graph.py:586 still says one engine materialization.

Cross-PR with #1422

• Council tax nation rows. UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's three OBR council tax nation rows need this PR's source_table_by_opening_year map, or the merged tree refuses. This is posted on UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422.
• Company car defaults. The guard at test_uk_terminal_gates.py:512 can't see this stage's outputs. So UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's nine company car inputs signed at default would merge silently with the six this stage writes; this is also posted on UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422. The other three keep a reason that becomes false ("every list price is zero"), and the approximation they make (full-year availability, no private-use payments) isn't stated.

Checked and found true

• The OBR windows. The weights are 3/12 and 9/12, and income tax reproduces exactly. Editions match by row header for all 19 rows, the refusals work, and the value moves are as stated.
• The BIK facts. There are 96 rows, and the tolerances match Chronicle's stated rounding. The re-pin changes only feed identity on the other 35 resources, and only the keys of the DESNZ resource.
• The stage's mechanics:
  • the engine inputs' names, units and periods match 2.165.0;
  • the price scaling is exact;
  • the float32 re-pricing error stays under 2.3p;
  • the draws are deterministic;
  • the carriers are dropped at the release boundary.
• Fixtures. The H2 fixture CSVs and the test frames are synthetic.

Disclosure-control items went to the author separately.

🤖 Generated with Claude Code

[juaristi22]

Thanks for the review. Every item is addressed, on top of #1422's head, since #1422 merges first. María ruled on three of them on 31 October: the carve moves upstream (items 2 and 10), the fuel goes first (item 7), and the EPB shortfall is measured before anything corrects it (item 3).

Fix before the build

1. obr.ni_employer binds ni_class_1_employer. In 2.165.0 ni_employer adds ni_class_1_employer and ni_class_1a. Chronicle's October package labels table 4.4 row 18 "Class 1 Employer NICs", and footnote 2 puts Class 1A in row 21 ("Other NIC"). The two notes that said no household carries Class 1A now say households carry it on the company cars and fuel, and the parity tool's rationale says so too. The registry-parity reason now cites October's table 4.8 and the 2.1% gap.
2. The carve now runs upstream (items 2 and 10). company_cars runs right after hmrc_spi_income_spine. So salary_sacrifice converts the carved pay, and the identity tool's carrier holds: pay plus sacrifice is the pre-conversion pay. e8's CGT stages see the carved pay too, as do the bonus-sacrifice candidacy, the UC and Carer's Allowance earnings rules and the gainful self-employment flag. The carve no longer keeps room for pension deductions, since salary_sacrifice bounds them on the carved pay.

What the Results will measure

3. Measured first. The receipt records the weighted holders and list price that the tape rows lose for want of EPB (holders.spi_dropped_without_epb), each withheld below the floor. "What to expect" names this beside the base size.
4. Price scaling. The doc and the stage note now say a scaled price is the room over the appropriate percentage. So the CO2 draw, the 50-mile range and the RDE2 diesel move the 2030-31 base through the scaled cars, not only the baseline. Your arithmetic is in the doc. The receipt records the list price the scaling takes off, and that of the cars it drops (spi_carve_out.weighted_list_price_scaled_away_gbp and weighted_list_price_dropped_gbp).
5. CO2 after the carve. The receipt keeps the draw's shares (co2.drawn) and adds the output's (co2.output).
6. Pre-calibration weights. The doc and the body say the receipt's figures are at the stage's weights, and the results measure the base on the calibrated file.

Stage behaviour

7. Fuel first. Where the car and fuel benefits don't fit inside EPB together, the fuel goes first. Your case now keeps the car, scaled to £31,034. An unknown-fuel holder who reports private fuel draws petrol or diesel at their Table 2C shares.
8. Count rows. The two count rows filter on employment_income_with_benefits_in_kind_annual, as the amounts read it.
9. Band 1's floor. Band 1's uniform draw starts at £5,000, the carve-out's declared minimum price, read from the same declaration.
10. Tape rows only. The carve reads person_is_spi_income_recipient. The undrawn copies and the dependants aged 16 to 19 keep their twin's pay and car as FRS rows do, and the pay identity is checked on the tape rows.

Coverage and texts

12. Coverage. company_cars has a family_coverage entry, and the coverage test requires the stage.
13. Small items:
    • the annual simulation sets company_car_availability to the employment share, so the engine prorates a part-year employee's car by its own rule (ITEPA ss. 143 and 152);
    • the fence's count is compared with the vendored rows at run time;
    • CARVAL's bounds and FUELTYP's codes are defined once, in frs_spine, with a consistency test;
    • the helper modules gain uk_engine_adapter, spi_income and stochastic_assignment, and the stage no longer imports salary_sacrifice;
    • the stage notes name the re-pricing check;
    • the graph.py comment counts two materializations.

Cross-PR with #1422

• In the merged tree the three council tax nation rows carry source_table_by_opening_year.
• UK_ENGINE_DEFAULT_INPUTS signs only the three inputs the stage leaves at default, and the reason states the full-year approximation. The guard test now counts every stage's declared outputs, so a signed default on a written input fails.

What I ran, on the 08962f1 feed, on the merged tree with #1422's head 821e342:

• the engine-free suite in chunks: 14,463 tests. Three local-only checks fail: the UK extra is installed here, the live release pointer, and orrery's Node package;
• the national and local references, memberships and compile-parity receipts, and the H2 fixture, regenerated: byte-identical to the committed files;
• the feed-gated files one per process, other than bus targets: 236 passed;
• the UK engine tests for company cars, coverage, the annual measures, concept mapping, the graph and the H2 parity: 75 passed.

Before #1422's third review I also re-vendored the resources (byte-identical), regenerated the coverage manifest and the concept coverage, and ran the bus targets file and the national compile-parity regeneration on the full artifact, with its release gate. All passed. After the third review, the references and receipts those two runs check regenerate byte-identical from the feed's cuts, so I didn't repeat those 27 GB runs.

🤖 Generated with Claude Code

[juaristi22]

Second review of #1430 at 40968ff53

This re-reviews this PR's commits since e67e76c29 (eb43bbb6a, 9a656579d, 4d7356e51 and 40968ff53) against my first review. They sit on top of #1422 at 821e342d1. Three reviewers took one area each:

• re-running each round-1 finding;
• reading the stage changes for new defects;
• targets, the regenerated surfaces, the nine merge resolutions, and trial merges with UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422's current head and with SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428.

Everything ran without the engine; where engine behaviour matters, it was read from the policyengine-uk 2.165.x source. Paths are relative to packages/microcosm-build/src/microcosm/build/ unless they start with packages/, tools/ or docs/.

Verdict: twelve of the round-1 items are fixed and two are partly fixed. Two problems would each stop the pair build:

• company_cars declares empty fit-weight records, so uk_weights_audit fails every national and dense build (1);
• with the carve upstream, salary_sacrifice can abort the spine (2).

The merge with #1428 also needs steps that git won't flag (3, 4).

Fix before the build

1. Every national fails uk_weights_audit (uk_runtime/company_cars.py:1373-1375).

• Mechanism: the stage fits nothing, but declares fit_weight_records returning (). That empty list travels to the audit:
  • the stage evidence records an empty list (stage_evidence.py:69-79);
  • the sidecar keeps it (uk_runtime/graph_evidence.py:214-218);
  • rehydrate_uk_fit_weight_records turns any empty list into () (uk_runtime/release_certification.py:216-219);
  • so the release-blocking gate fails: "A production fit stage ran but emitted no FitWeightRecord evidence".
• Probe: with the real classes and binding the audit fails; without company_cars it passes.
• ISA: ISA has the same declaration at this head (isa_subscriptions.py:1646-1648). SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's b9bcd709c removes only ISA's, so the merged tree still fails on company_cars.
• Fix: drop the property and its import, as b9bcd709c does, and add the evidence test SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 added for ISA.

2. With the carve upstream, salary_sacrifice can abort the spine (uk_runtime/company_cars.py:937-959; uk_runtime/salary_sacrifice.py:745-756).

• Mechanism:
  • The SPI stage caps a tape row's employee pension contribution at its pre-carve pay (uk_runtime/spi_income.py:2553-2556).
  • company_cars now lowers that pay. eb43bbb6a removed round 1's room for pension deductions, on the grounds that salary_sacrifice bounds them.
  • It does bound salary and bonus sacrifice. For employee contributions it raises instead: "Employee pension contributions must not exceed the pay that remains". Its comment at :746-752 says a new pay rewrite must cap contributions again.
• Probes with the real classes:
  • a tape row with PAY 0, EPB £3,000, a car and a £150 contribution ends with pay 0, and raises;
  • a row with £20,000 of pay, whose contribution was capped at £20,000, is carved to £16,760, and raises;
  • the same rows without a car run.
• Also: employer contributions and the ASHE ceiling stay on pre-carve pay, so pension input is overstated on carved rows.
• Fix:
  • keep room for the employee contribution in the carve's budget, or re-cap contributions after the carve and declare that rewrite;
  • add a test that runs salary_sacrifice on the stage's output;
  • update docs/uk-company-cars-316.md:74-75, which says salary_sacrifice keeps the deductions within pay.

Merging with #1428

3. A new HMRC employment count still reads pay alone.

• SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 adds hmrc.spi.employment_income.count_by_self_employment_income_band. It filters on employment_income_annual != 0, with a note that this is "the variable Table 3.6's employment rows read". This PR moves those rows to employment_income_with_benefits_in_kind_annual.
• The files merge without conflict, so the merged tree binds one HMRC employment count on pay alone. A tape row carved to zero pay, with its car kept, then leaves that count but stays in Table 3.6's: round-1 item 8 again.
• Fix: one of the two PRs should switch the new row, and its note, to the benefit-in-kind term.

4. The merge needs steps git won't flag.

• Vendored feed: SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 vendors uk/hmrc_spi_table_3_7_savings_rows.json stamped with feed 53815b2. On this PR's 08962f1 pin, vendored_rows refuses it ("vendored from a different Chronicle feed"), so the SPI income stage stops, and the vendoring test fails. The table needs re-vendoring on 08962f1, and SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's membership and parity receipt need regenerating there.
• Stage count: both PRs move the stage count from 50 to 51 the same way, so the asserts merge silently at 51 with 52 stages. test_uk_graph.py:65 fails, and test_country_spec.py:1063 needs the same re-pin.
• Otherwise: the trial merge has 21 conflicts, all resolved by keeping both sides, and the stage order comes out right: company_cars after the SPI stage, and the receipts stage last.

Round-1 items

#	Status
1. obr.ni_employer	Fixed. It binds ni_class_1_employer, which reproduces the old measure exactly. Correction: my round-1 note said the solve would move weight. That was wrong: the row is held out of the objective until 2027-01-16, so the fix corrects a diagnostic. Leftovers are in 6 below.
2. Identity tool	Fixed. The carve runs right after the SPI stage. A probe chain (company_cars, then salary_sacrifice, then the tool's pre-sacrifice frame) holds. But see 2.
3. EPB shortfall	Fixed. The receipt measures the lost holders and list price.
4. Price scaling	Partly. The arithmetic and the receipt fields are right. But docs/uk-company-cars-316.md:46-47 and :108-109 still say the diesel supplement and the frozen electric share move only the baseline, which contradicts :86-94.
5. CO2 after the carve	Fixed (co2.drawn and co2.output).
6. Pre-calibration weights	Fixed in the texts.
7. Fuel first	Fixed, and private fuel can't draw electric. The unknown-fuel shares are in 7 below.
8. Count rows	Fixed. All four HMRC employment rows read the benefit-in-kind term; #1428's new row doesn't (3).
9. Band 1's floor	Fixed.
10. Upstream pay rules	Partly. Bonus sacrifice, UC take-up, Carer's Allowance and the CGT stages see carved pay. Exception: the UC gainful self-employment flag is set inside the SPI stage on pre-carve pay (uk_runtime/spi_income.py:3015-3047), and nothing refreshes it. A side-trade profit of £5,000, with pre-carve pay of £6,000 and carved pay of £1,000, should be flagged and isn't. The reply says this flag sees carved pay.
11. Tape rows only	Fixed.
12. Coverage	Fixed.
13. Small items	Fixed. The availability share matches ITEPA ss. 143(3) and 152(1).
Cross-PR with #1422	Fixed. The nation rows carry the edition map, three defaults stay signed, and the guard counts every stage's outputs. A forward merge with #1422's current head (4af602d81) is clean, and its tests pass.

Also new

5. The calibrated-file report the doc promises doesn't exist.

• The promise: docs/uk-company-cars-316.md:140-142 says the national reports the calibrated file against HMRC Tables 1, 3A and 4. So do the BIK facts' purpose text and the gate note (uk/gates.json:1059, "measured, not gated").
• What exists: the only reader of HMRCCompanyCarStatistics is the stage itself, before calibration.
• Fix: add the diagnostic, or say the comparison is a manual step.

6. The obr.ni_employer exclusion still describes ni_employer (uk/calibration_measure_exclusions.json:445-449).

• The texts: its reason calls ni_employer the secondary Class 1 charge. Its condition says to rebind "when a locked policyengine-uk nets the allowance out of ni_employer".
• The risk: with the row now on ni_class_1_employer, meeting that condition (pe-uk#2459) would leave the bound measure gross. Following the register would then rebind about £7.4bn high.
• Also: tests/engine_free/uk/test_uk_national_insurance_targets.py:19-21 and the body's fit-movement note are stale, and no test pins the new binding.

7. Smaller:

• Unknown-fuel shares: the unknown-fuel draw no longer meets Table 2C (uk_runtime/company_cars.py:658-672). Fuelled holders now exclude electric, while unfuelled ones keep the unconditional shares. So the electric share among unknown-fuel holders is (1 − f) × Table 2C's, where f is the share reporting private fuel.
• Income proxy: the CGT and ISA income proxy (uk_runtime/cgt_imputation.py:191-208) sums employment_income alone. With the carve upstream, tape-row holders' car benefit drops out of it, though the engine's total income includes benefits_in_kind.
• obr.fuel_duties: it is the only OBR window row without the edition map (uk/uk_population_targets.json:2214-2240). It is signed out today, but it would refuse if brought back.
• Council tax nation rows: their parity reason omits the main-residence scaling. This comes from UK: take up #1095's actionable items on policyengine-uk 2.165.1 #1422 and is carried in this PR's regenerated receipt.

8. Stale texts:

• uk/gates.json:1059 and uk_runtime/stage_health.py:5107 and :5136 still say "any SPI-channel row … less any salary sacrifice";
• uk_runtime/salary_sacrifice.py:746-752;
• docs/uk-isa-subscriptions-1415.md:5, though ISA is now last;
• three docs count 19 rows with source_table_by_opening_year, where the merged tree has 22;
• the engine test texts still say "SPI-channel".

9. Test gaps: no test does any of these:

• runs salary_sacrifice on the stage's output;
• checks the stage's fit-weight evidence;
• pins the ni_class_1_employer binding;
• compares co2.drawn with co2.output.

Checked and found true

• The scaled-price arithmetic matches policyengine-uk 2.165.1. The electric range rule in 40968ff53 matches ITEPA s. 139 and the engine.
• Draws stay keyed and uniquely salted, so moving the stage changes no other stage's draws.
• The regenerated surfaces match their generators wherever that can be checked without the feed: the coverage manifest, local census, contract digests, parity reasons, uv.lock and concept coverage.
• The nine merge resolutions drop no side.
• The 2.1% gap the parity reason cites is consistent with October's tables.

Disclosure-control items went to the author separately.

🤖 Generated with Claude Code

[juaristi22]

Thanks for the second review. Every item is addressed, on top of main now that #1422 is in (11c1f4e). Merging main brings no UK file.

Fix before the build

1. Fit-weight records. company_cars no longer declares fit_weight_records or imports FitWeightRecord, as SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's b9bcd709c does for ISA. A test reads the stage's evidence and finds none. ISA's declaration is still on main until SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 lands.

2. The carve keeps the contribution. The cap is now EPB within the pay the row's employee pension contribution leaves, so a carve never takes pay below the contribution the SPI stage capped. Your two probes are now tests, of the carve and of salary_sacrifice run on the stage's output:

   • the first keeps its car with £2,850 carved, leaving £150 of pay for its £150 contribution;
   • the second has no room, so it loses its car and keeps its £20,000.

   Both tests fail on the old cap. The stage's invariants also count a tape row whose contribution exceeds its carved pay, and the gate blocks on it. The doc, the stage note and salary_sacrifice's comment now say this. Employer contributions stay as the SPI stage set them on the drawn PAY + EPB + TAXTERM. That holds for every tape row that draws EPB, car or not, so it predates this PR; moving them onto pay alone is a change to the SPI stage. The doc says so.

Merging with #1428 (3 and 4)

These fall to whichever PR lands second. If #1430 lands first, #1428 needs to:

• switch hmrc.spi.employment_income.count_by_self_employment_income_band, and its note, to employment_income_with_benefits_in_kind_annual;
• re-vendor uk/hmrc_spi_table_3_7_savings_rows.json on 08962f1, and regenerate its membership and parity receipt there;
• re-pin the stage count at 52 in test_uk_graph.py and test_country_spec.py.

Round-1 items that were only partly fixed

• 4. The two sentences now say that the diesel supplement and the frozen electric percentage move the base through the scaled cars, as well as the baseline.
• 10. You're right: the flag is set inside the SPI stage, before the carve, so my last reply was wrong about it. A carved row now takes the root's rule (frs_spine.uc_gainful_self_employment) again on its carved pay, so your side trade of £5,000 beside pay carved from £6,000 to £1,000 is flagged. The flag is now a declared output and rewrite of the stage, and the receipt counts the rows it turns on. A test covers your case.

Also new

5. No calibrated-file report. Comparing the calibrated file with Tables 1, 3A and 4 is a manual step for now. The doc, the stage note, the fact fence, the BIK facts' purpose text and the gate note all say so.
6. The obr.ni_employer exclusion.
   • Its reason now says the row binds ni_class_1_employer.
   • Its rebind condition asks for the allowance to be netted out of that measure. It also notes that netting it out of ni_employer alone would leave the bound measure gross.
   • A new test pins the binding and the condition.
   • The NICs-total test's docstring now says households carry Class 1A on company cars and fuel.
   • The body's fit-movement note now says the row is a diagnostic.
7. Smaller:
   • Unknown fuels. Holders without private fuel now draw electricity at Table 2C's share divided by (1 − f), capped at 1, where f is the weighted share of unknown-fuel holders who report private fuel. The unknown fuels then meet Table 2C together. A test checks the shares, and the case where the private-fuel holders leave too little room.
   • Income proxy. The CGT and ISA proxy sums employment_income, so it now leaves the car benefits out on every row, as it always has on FRS rows. The doc says so. The engine computes the benefits; no column carries them for the proxy to add.
   • obr.fuel_duties carries the edition map and its note, so it reads if it comes back.
   • Council tax nation rows. Their parity reason now names the main-residence scaling, and the national receipt is regenerated.
8. Stale texts. These now say tape rows, or are restated:
   • the gate note, and the stage-health docstring and message;
   • salary_sacrifice's comment;
   • the ISA doc, which now says "after every other stage";
   • the three docs that counted 19 mapped rows, which now count 23 (adding the three nation rows and fuel duties);
   • the engine tests.
9. Tests. New tests cover:
   • salary_sacrifice on the stage's output;
   • the fit-weight evidence;
   • the ni_class_1_employer binding;
   • co2.drawn against co2.output (equal without a carve-out).

What I ran, on the 08962f1 feed, on main (11c1f4e) merged with these fixes (head 144d738):

• the 26 touched engine-free files (the stage, salary sacrifice, stage health, gates, graph, manifests, coverage, parity, contract, ISA, NI targets and others): 1,408 passed, 8 skipped;

• the generated surfaces, regenerated from the feed's cuts. Four changed as expected:

  • the national compile-parity receipt, with the three nation reasons;
  • the membership's fuel duties entry, which now matches two facts and stays signed out;
  • the coverage manifest;
  • the H2 fixture.

  The national references and the local receipt regenerated byte-identical;

• the feed-gated files one per process, other than bus targets: 239 passed and 2 skipped (the Scotland support and the authenticated national artifact aren't mounted here);

• the UK engine tests for company cars, coverage, the annual measures, concept mapping, the graph and the H2 parity: 75 passed.

CI runs the full engine-free suite on this head. Locally, the full suite passed on the previous merge with #1422's final head: 14,487 tests, and only the three checks that can't pass on this machine fail.

🤖 Generated with Claude Code

[juaristi22]

Third review of #1430 at 144d73836

This re-reviews 083ee9f7d, 51259e71b and 144d73836 against my second review at 40968ff53, along with the merges that bring in #1422 and main (14dc81547, after 11c1f4e69). Two reviewers:

• one re-ran each round-2 finding and re-applied each fix's old rule as a mutation;
• the other read the new commits for new defects and trial-merged with main and with SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's head (4dd83e999).

Everything ran without the engine; where engine behaviour matters, it was read from the policyengine-uk source at the lock (2.165.1). Paths are relative to packages/microcosm-build/src/microcosm/build/ unless they start with packages/ or docs/.

Verdict: both build-stoppers and the other round-2 items are fixed, CI is green, and the main merge is clean. Nothing new blocks a build. What's left:

• ISA still fails the audit. A national from this head, or from main, still fails uk_weights_audit, now only because of ISA's declaration (uk_runtime/isa_subscriptions.py:1646-1648). It passes once SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's b9bcd709c is in, as the reply says, so the pair needs SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 first, or the same removal for ISA here.
• The merge with SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 needs five steps beyond the reply's three, and two of them are silent.
• The stage-level UC refresh is untested (1).
• Smaller: the lost holders' fuel draw (2), what the contribution room costs (3), the obr.ni_employer texts (4), and the module docstring and fuel rule id (5).

Merging with #1428

The reply's three steps are right: switching the employment count row, re-vendoring Table 3.7 on 08962f1, and the stage count at 52. The stage-count asserts now conflict, so git flags them, and at 52 the graph and country-spec tests pass. The trial merge now has 25 conflicting files, all resolved by keeping both sides. The stage order comes out right: 52 stages, company_cars after the SPI stage, trading_allowance_receipts last.

Five more steps are needed:

• The employment count row (silent if forgotten). Switching it also means updating SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's test_the_table_3_10_employment_counts_read_annual_pay_by_profit_band (packages/microcosm-build/tests/engine_free/uk/test_uk_pension_age_benefit_targets.py), which pins employment_income_annual, and the Table 3.10 reason in uk/cross_grain_declarations.json ("adds the year's pay"). Nothing requires HMRC employment rows to read the benefit-in-kind term, so forgetting the switch merges green. A test across the hmrc.spi.employment_income.* rows would pin it.
• The gate-battery pins in packages/microcosm-data/src/microcosm/data/contract.py (6 hunks) and packages/microcosm-data/tests/engine_free/shared/test_contract.py (3). Neither side's values hold on the merged spec, and packages/microcosm-build/tests/engine_free/shared/test_gate_battery_contract_pins.py fails 3 tests until they're regenerated. SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's 4dd83e999 records a certified national that then failed the release contract on these pins.
• The coverage manifest: --check fails on the merge, so it needs regenerating.
• The H2 fixture needs regenerating.
• docs/uk-isa-subscriptions-1415.md:5 (silent). It now says ISA runs "after every other stage". SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428 runs trading_allowance_receipts after it, and the line merges without a conflict.

New

1. Nothing tests the stage-level UC refresh (packages/microcosm-build/tests/engine_free/uk/test_uk_company_cars.py:597-613; uk_runtime/company_cars.py:1241-1244).

• Why the test can't fail: the whole-stage test compares the carved rows' flags with the rule on carved pay. On its frame no carved row changes flag (seeds 1 to 5), so the "after" flag equals the "before" one, and the new receipt count is 0.
• Mutation: both reviewers made the stage's call a no-op. All 31 tests in the file stayed green, and so did the 162 in the salary_sacrifice and stage-health files. Only the helper's own unit test catches a broken helper, and the engine frame has no self-employment income.
• Fix: seed one tape holder with a side trade whose profit lies between its carved and its drawn pay, and assert both the flip and the receipt count.

2. The lost holders' fuel is now drawn at their own private-fuel share (uk_runtime/company_cars.py:1158-1174, with :688-703).

• Mechanism: the tape rows without EPB call assign_fuel_types on their own, so their unfuelled rows draw electricity at e/(1 − f) with f taken over the lost rows alone, not over the holders. The comment says their prices are "drawn as the holders' are". Before 083ee9f7d, the lost draw reproduced the holders' draw exactly.
• Synthetic probe: with Table 2C at 0.5, 0.4 and 0.1, 2,000 holders (a quarter fuelled) and 400 lost rows (none fuelled), the lost rows' electric share is 0.48 drawn alone and 0.65 drawn at the holders' share. 75 of the 400 change fuel.
• Effect: it moves holders.spi_dropped_without_epb.weighted_list_price_gbp, the measure the 31 October ruling asked for, through the electric/non-electric price class.
• Fix: pass the holders' electric probability to the lost rows' draw.
• Also: every unknown-fuel draw now depends on its population's weighted f. Reordering rows still moves none, but a person's fuel can change when other rows or weights do, for example between spines.

3. The contribution room keeps a drawn contribution at the car's expense (uk_runtime/company_cars.py:985, :1008-1010; uk_runtime/spi_income.py:2536-2555).

• Mechanism:
  • On a tape row, pay is PAY + EPB + TAXTERM. So the room lowers the budget below EPB only where the employee contribution exceeds PAY + TAXTERM, the row's cash pay.
  • That happens only because the SPI stage caps the forest-drawn net-pay contribution at a pay that includes EPB, and a net-pay contribution can't come out of benefits in kind.
  • On those rows, the carve now scales or drops the car to keep a contribution the build drew.
• The other route: re-capping the contribution at the carved pay, the second option in my round-2 list, keeps the car. I offered both without saying what the first one costs.
• Not measured: cap_below_epb_for_the_employee_contribution counts rows where the room lowers the budget, not those where it scales or drops a car. Those cars are pooled into list_price_scaled and dropped_below_minimum_price.
• Probably small. A ruling, or a count of the cars the room scales or drops, would settle it.

4. The obr.ni_employer exclusion is still partly stale (uk/calibration_measure_exclusions.json:446).

• Under-21 rate: the reason says the under-21 zero rate is one "which the engine also omits". policyengine-uk applies the under-21 and apprentice zero rates in ni_class_1_employer from 2.160.0 (utils/employer_ni.py), and the 2.165.1 lock has them. The UK local-area release: publish a K=15, 150,000-household dense dataset in at most two builds #1301 measurement it describes ran on an older engine, so the sentence needs a past tense.
• Tracker: policyengine-uk#2459 ("Model the Employment Allowance so ni_employer is net of it") and the adjudication still ask for netting out of ni_employer. The reason now says that change would leave the bound measure gross, so #2459 landing as titled would look like it meets the rebind condition when it doesn't. A note on #2459 would keep them in step.

5. Texts:

• Module docstring: step 2 (uk_runtime/company_cars.py:32-35) still says unknown fuels draw "at Table 2C's shares". Step 5 (:47-58) still caps at EPB alone, with no room for the contribution and no UC flag rewrite.
• Fuel rule id: UNKNOWN_FUEL_RULE and its error text (:227-230, :368-372) still name the old rule. So a declaration written for the old rule still validates. A new id would make the change visible.
• Pre-existing: salary_sacrifice lowers pay without re-deriving the UC flag (uk_runtime/salary_sacrifice.py:686-688), which is what this PR now does after the carve. Its edited comment (:746-752) names only contributions as what a pay rewrite must keep.

Round-2 items

#	Status
1. Fit-weight records	Fixed. The property and its import are gone. The new evidence test (packages/microcosm-build/tests/engine_free/uk/test_uk_company_cars.py:825) fails if the declaration comes back. With ISA's declaration also removed, a probe audit passes. ISA is the rest (Verdict).
2. Carve and salary_sacrifice	Fixed. The budget is EPB within pay less the employee contribution (uk_runtime/company_cars.py:985). Both of my probes now run: PAY 0, EPB £3,000 and a £150 contribution keeps its car, with £2,850 carved and £150 of pay left. £20,000 of pay with a £20,000 contribution drops the car and keeps the pay. Salary and bonus sacrifice and the private-fuel variants run on the output. The new invariant can't fire on legitimate input. See 3 for the cost.
3. #1428's employment row	Right but incomplete (Merging).
4. Merge steps	Right but incomplete (Merging).
Round-1 item 4	Fixed. The doc's two sentences agree with the scaling.
Round-1 item 10	Fixed in code. A carved row takes the root rule on its carved pay, using the SPI stage's exact inputs, and my £5,000 side-trade case is flagged. Untested at the stage (1).
5. Calibrated-file report	Fixed. It's a manual step, and every text says so.
6. obr.ni_employer	Fixed as asked. The new binding test fails on revert. Residual texts are in 4.
7. Smaller	Fixed. The unknown fuels meet Table 2C together wherever the private-fuel holders leave room (a 200,000-row probe). obr.fuel_duties carries the map, and the nation reasons name the main-residence scaling. The side effect is in 2.
8. Stale texts	Fixed, except the module docstring (5).
9. Test gaps	Fixed. Each new test fails under its mutation.

Checked and found true

• The carve: on every path the benefit is at most the budget, which is at most pay less the contribution. The SPI cap is an exact minimum on the same rows, with no stage between them. The gate blocks on every violation key.
• The UC refresh: it applies the SPI stage's rule and only turns flags on, only on carved rows. Its declaration matches across the sources, the graph, the coverage manifest, the H2 fixture and the export allow-list. Benefits in kind aren't UC earnings in 2.165, so carved pay is the right comparator.
• The fuel draw: the shares sum to 1. Where the private-fuel holders leave too little room, every unfuelled car draws electricity and petrol and diesel keep Table 2C's ratio. f is weighted, and the salts are unchanged.
• Regenerated surfaces: the nation parity reasons equal the generator's constant, and the replay passes for the national and local receipts. The fuel duties membership entry, the coverage manifest and the H2 fixture are consistent.
• Merges: the remerge-diffs of 14dc81547, 23aadd8f3 and 4d4c1a6e6 are empty, and main brings no UK file. A trial merge with main at dfa5fe399 is clean, and main's one later commit (Harden the money run-entry walk and cover its gaps #1441) touches no UK file.
• CI on 144d73836: every job passes. The touched files pass locally too: the stage, salary sacrifice, stage health, receipt counts, the UK and generic ledger targets, compile parity, the data contract, the NI targets and the graph.

Disclosure-control items went to the author separately.

🤖 Generated with Claude Code

[juaristi22]

Thanks for the third review. #1428 lands after this PR and picks this code up, and builds run from there (María's call). So the merge-time items stay with #1428, and everything that tightens this PR on its own is fixed here.

Left for #1428

• ISA's audit. SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's b9bcd709c removes ISA's empty fit-weight records, and builds run from SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428, so this PR doesn't repeat it.

• The merge. Your five steps fall to SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428, on top of the three in my last reply:

  • the employment count row, with its test and the Table 3.10 reason;
  • the gate-battery pins;
  • the coverage manifest;
  • the H2 fixture;
  • the ISA doc line.

  Two of those would have merged silently, so this PR now makes them fail loudly instead:

  • a new test pins every HMRC employment income row, amount or count, national or local, to employment_income_with_benefits_in_kind_annual, so SPI donor savings rebased to SPI's 2023-24 bands (savings follow-up to #1418) #1428's Table 3.10 row fails until it switches;
  • the ISA doc no longer names a position: the stage runs "after every stage that writes income", which stays true with trading_allowance_receipts after it.

Fixed here

1. The stage-level UC refresh is tested. The new test runs the stage once, gives a carved tape row a side trade between its carved and drawn pay, then runs the stage again. It asserts both the flip and the receipt count. Your no-op mutation now fails it.
2. The lost holders draw at the holders' share. unknown_fuel_electric_share computes the holders' electric share once, and the lost rows' draw takes that share, so they draw as the holders do. A test reproduces your probe. The doc now says that matching Table 2C at the stage's weights makes each drawn fuel depend on the other drawn cars' private fuel and weights, as the CO2 split already does.
3. The contribution room is now measured. The carve runs a second time inside EPB within pay alone. The receipt counts the cars the room scales further or drops (cars_the_employee_contribution_scales_or_drops), and the list price they give up (weighted_list_price_given_up_for_the_employee_contribution_gbp). The doc names the other route, re-capping the contribution at the carved pay, which keeps those cars. María chooses between them once the pair build has measured the cost.
4. The obr.ni_employer texts. The under-21 sentence is now in the past tense, and says the 2.165.1 lock applies the under-21 and apprentice zero rates in ni_class_1_employer. policyengine-uk#2459 now has a note that the row binds ni_class_1_employer.
5. Texts.
   • The module docstring's steps 2 and 5 now describe the joint fuel draw, the lost holders, the contribution room and the UC refresh.
   • The unknown-fuel rule is renamed table_2c_shares_together_non_electric_with_private_fuel in the schema and both manifests, and a test refuses a declaration that names the old rule.
   • salary_sacrifice's comment now says that its own conversion doesn't refresh the UC flag. That gap predates this PR, so I've left it for a follow-up.

What I ran, on head de41aba:

• the 27 touched engine-free files: 1,442 passed, 8 skipped. They cover the stage, salary sacrifice, stage health, gates, the graph, the manifests, coverage, parity, the contract, ISA, the NI and population targets, and others;
• the coverage manifest and the H2 fixture, regenerated. Only the manifest's source hash moved, plus the fixture's copy of the renamed rule and its contract digest;
• the UK engine tests for company cars, coverage, the annual measures, concept mapping, the graph and the H2 parity: 75 passed.

Nothing here touches a target or the feed, so the references, memberships and parity receipts stand as last regenerated. CI runs the full engine-free suite on this head.

🤖 Generated with Claude Code

[juaristi22]

Fourth review of #1430 at de41aba62: merge check

This checks a22e33da2 and de41aba62 against my third review at 144d73836, looking only for anything that should block the merge.

Verdict: nothing blocks the merge once CI is green. At the time of writing, lint, wheels, integration-uk and both engine-uk jobs pass, and the engine-free and engine-us jobs are still running.

What I checked

• No stage output changes:
  • The refactored fuel share gives the holders the same draws.
  • The tape rows that lose their car now draw at the holders' share, which moves only a receipt measure.
  • The second carve, inside EPB within pay alone, only measures what the contribution's room costs.
• Tests: 615 passed, none skipped, in the nine files a merge could break:
  • the gate-battery pins and the data contract, which still hold with the renamed fuel rule;
  • the stage's tests, and the new employment-row test;
  • the receipt-count guard and both coverage checks;
  • source stages and stage health.
• Main: its one commit since this PR's main merge (Harden the money run-entry walk and cover its gaps #1441) touches only transport files, and GitHub reports the PR as mergeable.

Round-3 items

#	Status
1. Stage-level UC refresh	Fixed. The new test gives a carved row a side trade between its two pays, and asserts the flip and the receipt count. It fails when the refresh is made a no-op.
2. Lost holders' fuel	Fixed. The lost rows take the holders' electric share, and the new test reproduces the probe.
3. Contribution room	Measured, with the choice of route left to a ruling after the pair build. One nit, not blocking: cars_the_employee_contribution_scales_or_drops also counts a car that keeps its price and loses only its private fuel to the room, while the name and the doc say "scales further or drops".
4. obr.ni_employer texts	Fixed. The under-21 sentence is in the past tense, and policyengine-uk#2459 has the note.
5. Texts	Fixed. The module docstring, the renamed rule (a test refuses the old name) and salary_sacrifice's comment are updated.
Merge steps	Moved to #1428, which lands after this PR. The new test makes a forgotten switch of the Table 3.10 employment row fail loudly. The ISA doc line now holds either way.

What #1428 needs after this merges

1. Switch the Table 3.10 employment count row, and its note, to employment_income_with_benefits_in_kind_annual. Update its pinning test and the Table 3.10 cross-grain reason to match.
2. Re-vendor Table 3.7 on 08962f1, and regenerate its membership and parity receipt.
3. Set the stage count to 52 in test_uk_graph.py and test_country_spec.py.
4. Regenerate the gate-battery pins in contract.py and test_contract.py.
5. Regenerate the coverage manifest and the H2 fixture.

Disclosure-control items went to the author separately.

🤖 Generated with Claude Code