One consistent UK target surface: enforced cross-grain reconciliation, declared uprating holds, nation household controls and the 942a1fa re-pin (#1123 PR A)

[juaristi22]

Part of #1123 (PR A: items 1, 2, 6 and 7; private rent is PR B). Addresses #104.

Why

The K=25 audit found that most local misses surviving the dense solve were on the target side. Local rows were never uprated, 509 national rows were held at an older period with no declared index, nothing refused an undeclared hold, and reconciliation across grains was opt-in. This PR makes the UK target surface one consistent, enforced surface.

What changes

• Cross-grain reconciliation is enforced. Every family bound at more than one grain declares how its grains agree (uk/cross_grain_declarations.json: bridges, partitions, relations) or the compile refuses. The lattice gains a nation grain, rows that span several legs, and same-geography partitions. The national register is reconciled across UK, nation and region in both loaders, and the joint local pass refuses a national factor away from one.
• Uprating doctrine. A target bound from a fact older than the calibration year carries a declared index or a declared hold (uk/uprating_holds.json, 16 entries). The compile refuses an undeclared, unsigned, expired or stale hold, nationally and locally. María signed all 16 on 2026-10-11; each expires on 2026-12-11, so every hold needs a review before then. The compile warns 14 days before an expiry, and a test on the committed register fails in the same window, so CI turns red on 2026-11-27.
• Chronicle re-pin to 942a1fa (UK local-surface facts for the 2025 calibration: household counts by nation and area, communal-establishment residents by age, mid-2025 population estimates, UC households by area for calendar 2025, and BRMA geography and rent inputs (PolicyEngine/microcosm#1123, PolicyEngine/microcosm#1090) chronicle#313 via Add UK 2025 calibration publisher inputs chronicle#314), on top of the 1ee7dfe pin that Set UK road-fuel and energy levels in the LCFS stage on a calendar-2024 basis (#1113) #1167 put on main and UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145 kept: 942a1fa holds every 1ee7dfe row. National population, local-authority ages, land, SLC repayments, ISC, DfI and public sector employment move to their 2025 editions. Spine inputs move only by publisher revisions (revised mid-2024 population, DfI rounding).
• Households: nations win. Each nation's official estimate controls its census cells: LFS 2025 for England's nine regions, NRS for Scotland, Welsh Government rolled by population for Wales, Census 2021 rolled by LPS dwelling stock for Northern Ireland. The UK composition takes the nations' sum (29,082,020; the LFS UK total is a diagnostic, 0.27% lower).
• HMRC. Income targets are on a taxpayer basis at both grains; area cells take HMRC's all-ranges growth and reconcile to the national SPI bands. Both parts of the factor they take are gated at each grain: source agreement within 2%, and the gap between the two sides' uprating indices within each bridge's measured gap plus half a point. Higher and additional rate taxpayers bind for each English region and nation (Income Tax liabilities Table 2.2, calendar 2025, on the engine's tax_band), and higher rate taxpayers bind by constituency and local authority in England, Wales and Northern Ireland (919 cells). The 13 SPI property-income amounts that UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145 binds carry the same taxpayer filter as every other SPI row here, which is Table 3.7's own universe. main measures them over every landlord in the band. On UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145's calibrated candidate (arm c9, policyengine-uk 2.123.0 at 2025), people who pay no income tax hold 2.1% of the property income in those bands (£0.74bn of £35.29bn) and 9.4% in the three bands below £30,000. They are 47,000 of 2.66m landlords; 31,000 of them have a finance-cost reduction, which takes their liability to nil. María decided on 2026-10-12 to keep the filter; the measurement by band is in this comment.
• Cells that are off the surface. A leg's control covers every area in it, but some signed cells are not bound: deferred with a published value (the City of London, the Isles of Scilly) or absent from the publisher's table (HMRC's local-authority tables carry no Torbay row). Where a control moves their leg, those cells join the reconciliation frame, so the bound cells do not take their share; they never reach the solve. A cell with no value is implied from the target's other local tier, and its share of its leg is bounded by 1.5 times its areas' share of the leg's census households.
• UC by area binds the calendar-2025 mean of DWP's monthly series, and the child bands follow each constituency's caseload. Open: the area series counts every household on UC (6.73m) and the national control counts paid claims (6.20m), so the bridge applies the GB paid share (0.921) to every area. The fix is a paid-claims area series from Chronicle (requested in UC households by area: re-extract the constituency, local-authority and by-children series with Payment Indicator as a dimension (PolicyEngine/microcosm#1198) chronicle#339) and a re-pin.
• Tenure. England's census shares drift by ONS SPREE's 2022 to 2024 share change, which carries information on social stock only; a share that moves by more than 10% holds that authority's census shares (Rushmoor). Tenure cells follow their authority's households.
• Small cells. A signed rule defers cells below a floor, with an expiry, and a deferred cell still counts in its leg's control (uk/local_small_cell_rule.json). Two rules are proposed and unsigned, so neither is in force and their cells are bound: council-tax band cells under 50 dwellings (16 cells) and higher rate area cells of 5,000 or fewer, where HMRC's rounding to the thousand can reach 10% (192 cells).
• Make zero-valued calibration targets intentionally nonzero-weighted #104. The loss-scale options carry a census of zero-valued targets. The weighting rule is unchanged (UK sparse selection: keep household mass, composition, nation shares and per-area ESS at a tractable size — test the target-weight rule and the L2 penalty first, dataset size only if they fall short #1124 owns it).
• Tooling. A rolled_forward_by_ratio value operation; both UK target kernels hash every module that decides a target value; tools/uk_target_surface_receipt.py writes and diffs receipts of the compiled surface; the national compile scans only each reference's own sources (the national parity receipts drop from about 2h25m to under 20 minutes on the 927k-fact feed, with identical candidate facts).

Value moves (receipts in docs/evidence/uk-target-surface-1123/)

The branch is rebased on main at 51a263726. Before is that commit with only the receipt tool; after is cd39d037e, run on a clean tree. Each receipt records what its working tree adds to the commit it names.

• National, bound rows: 1,177 to 1,213. 36 added (12 nation household cells, 24 marginal-rate cells), none removed, 552 moved.
  • By this PR's own reconciliation: 360 SPI regional cells (the band bridges) and 24 CGT regional cells.
  • By the re-pin and the household controls: 149 population rows, the 10 household-composition rows, land (3), SLC repayments (3), ISC, public sector employment and DfI.
• Local: 20,885 to 21,804 cells. 919 added, none removed, 18,811 moved (every age, census-household, tenure and UC cell; the four existing HMRC area targets; private rent through tenure; and 437 council-tax band cells in London, the South West and Scotland, by up to 0.5%, because a deferred authority no longer leaves its share on its neighbours).
• Reconciliation factors:
  • National band bridges onto the SPI regional cells: 1.002 to 1.057, each inside its declared bound. The gap is the taxpayers Table 3.11 places in no region, concentrated in the lowest band (2.7% of taxpayers and 5.7% of income tax at £12,570 to £15,000). The compile checks that on the values: each factor sits within half a point of Table 3.11's UK row over the regional cells' sum (21 of 24 match to five decimals; the widest gap is 0.19 points).
  • HMRC area cells to the national bands, source agreement: 0.999 to 1.006 at constituency grain and 1.003 to 1.008 at local-authority grain (gate 2%). The gap between the two sides' uprating indices is 0.997 to 1.014 (1.44% on employment counts), each inside its bridge's bound. The factors the cells take are the product, 0.996 to 1.019.
  • Local-authority legs under the constituency totals: 1.018 to 1.021 on employment counts and 1.008 to 1.011 on employment amounts. Torbay and the Isles of Scilly, which HMRC's local-authority tables do not carry, are implied at 2.3% of the South West's employees; they hold 2.6% of its households.
  • Nation household controls over constituency census cells: 0.973 (North East) to 1.038 (London); Scotland 0.989, Wales 0.994, Northern Ireland 0.998.
  • Higher rate area cells to their 2025 rows: 1.19 to 1.40.
  • UC households by area to the GB caseload: 0.921 (0.942 before this PR). This is the paid share, see above.

Verification

• tools/uk_target_surface_receipt.py --communal runs clean end to end on the pinned feed, on a clean tree at cd39d037e.
• Every new check passes on the feed: both fan-out gates at both grains, the 24 band-bridge bounds and their explained factors, all 20 subset tests (none skipped), the 44 off-surface cells, no row without a readable period, and the roll-forward base and series checks for Wales and Northern Ireland.
• Local runs at cd39d037e: the engine-free job's files, 16,693 tests with 6 failures that are local to this machine or to live state (two Orrery parser tests need a node package; one CGT test expects no pe-uk in the engine-free venv; one telemetry test is unstable under parallel workers; one head-to-head scorer test needs Python 3.14.7; one microcosm-data test reads a live Hugging Face pointer that CI skips); engine-uk 180 of 180; lint, tools/ci_test_plan.py verify and the vendored-resource check ok. CI on the previous head b8509c765 was green on every lane, engine-us included.
• The Make zero-valued calibration targets intentionally nonzero-weighted #104 edit to microcosm.calibrate.solve moves the loader golden vector, the US seed digests, the US spec digest, the US coverage evidence and the calibrate graph parity pins; all are re-pinned as main does for solver edits, and recomputed after each rebase where main moved them too.
• Not run: the dense re-solve, so the effect on fit is not yet measured. main retired the West Midlands income-tax deferral (UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145), so that cell is bound against the target this PR moves; its fit shows at the next national build.

Departures from the plan

• No closure gate in gates.json: closure fails closed at compile instead.
• No council-tax band partition, and the preflight tool is not updated.
• The A15 census-household uprating step is kept; it now targets the nations' sum and the nation bridge then rescales.

Open decisions (María)

Before merge:

• Signatures: the two small-cell rules. The 16 holds are signed and expire on 2026-12-11; keep the expiry test, or replace it with a dated tracking issue.
• UC by area: the paid-claims series is requested in UC households by area: re-extract the constituency, local-authority and by-children series with Payment Indicator as a dimension (PolicyEngine/microcosm#1198) chronicle#339; UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145 and Set UK road-fuel and energy levels in the LCFS stage on a calendar-2024 basis (#1113) #1167 are merged, so its re-pin has no open PR to sequence with.
• England households: LFS regional controls (as built) or the England total with projection shares.
• Scotland and Wales households by authority: the feed carries NRS 2025 by council and Welsh Government mid-2024 by authority; the surface holds census shares instead. Bind them here or later.
• Scope: Scottish higher rate area cells and all additional rate area cells stay unbound; Scotland's two national marginal-rate cells are bound although the engine classifies Scottish taxpayers on total income (tax_band compares a Scottish taxpayer's total income with the Scottish thresholds; the Scottish rates apply to non-savings, non-dividend income only policyengine-uk#2350). Keep them bound, or defer them until that issue is fixed.
• Two signed adjudications need their wording amended (full_frs_tei_band_unavailable, population_universe_private_households).
• Accept the departures above, or build the gate and remove A15.

Can follow as issues:

• Clip self-employment losses in the amount rows to match HMRC.
• Salary-sacrifice relief growth.
• The Northern Ireland household method.
• An edition preference in the resolver, so the land rows resolve at 2023 again.

After merge

• Re-freeze the scoring register in uk-candidate-eval (the reconciled national values change the registry version).
• PR B (rent) rebases on this branch.

🤖 Generated with Claude Code

[juaristi22]

Independent adversarial review (Claude Code) at 67fc1dd04

Static review only. I read the code and the committed registers at this head and at the merge base aca40a692, and I checked the UC source package in Chronicle at 942a1fa. I ran no tests, no compile and no receipt tool, so every factor below is control ÷ the sum of the committed cells before reconciliation, the same ratio the code takes. Engine CI was still queued when I looked.

Verdict. The core machinery holds up. Cross-grain coverage is complete for the committed contract: 15 multi-grain ids, 28 overlap pairs, none undeclared. The holds cover every lagged national row, and the #104 edit to solve.py changes no numbers. But four things should change before this leaves draft:

• the UC bridge reconciles two different concepts;
• the 2% fan-out gate never sees the local-authority grain, which needs 2.1%;
• the new national band bridges have no tolerance;
• the signatures the body asks for aren't enforced by any code.

Must fix

1. UC by area reconciles all-claims cells onto a paid-claims control.

   • National control: dwp.uc.households binds payment_indicator: "Yes" (uk_population_targets.json:14716).
   • Area cells: the selector (:28243) reads universal_credit_households with no payment dimension. The Chronicle package at 942a1fa (dwp/uc_households_by_constituency_january_2025_may_2026) has only the date and PCON24 fields, so it counts all UC households, nil-award claims included. docs/uk-uc-paid-target-contract.md is why the national rows moved to paid claims in the first place.
   • Size: the areas sum to about 6.73m against 6.20m for GB, hence the 0.921. Aligning the windows widened the gap (0.942 → 0.921), which points at concept, not timing.
   • The selector note says the bridge "reconciles like with like"; it doesn't.
   • Effects:
     • The uniform factor assumes the same nil-award share in every constituency.
     • The child bands keep their all-claims shares, so after the bridge they disagree with the national paid rows: +2.8% with no children, −1.5% with one, −3.5% with two and −5.1% with three or more. (The national no-children figure is the total less the child rows.) The solve gets conflicting targets across grains.
     • The coverage check can't see this, because the signatures differ.
   • Fix: extract the area series with Payment Indicator = Yes (UC_Households carries the field).

2. The 2% fan-out gate is never applied at local-authority grain, which needs more than 2%.

   • Factors: at constituency grain the committed values reproduce the body exactly: 1.0096, 1.0192, 1.0082 and 0.9958. At local-authority grain they are 1.0118, 1.0211 (employment counts), 1.0160 and 1.0023. Torbay and Scilly have no SPI LA cell.
   • Why nothing refuses 1.0211:
     • hmrc.employment_income.count at constituency and LA forms an exact-signature group with no control-grain row, so constituency parents LA under the single-winner rule.
     • The sort key at cross_grain.py:256-266, (precedence[lower], -precedence[winning], id), runs constituency_over_la before the bridge's country_over_la.
     • So by the time the bridge pair runs, LA already equals the reconciled constituency total. Its factor is about 1, and _assert_uk_fanout_sum_factors (uk_runtime/ledger_targets.py:2579) only reads bridge groups.
   • The declaration says "the constituency and local-authority cells reconcile to it per grain… a factor beyond 2% is refused". test_uk_fanout_sum_bridges.py only builds a constituency receipt.
   • Fix: either check the bridge against the pre-reconciliation LA total, or gate the signature group too.
   • A related point: the gate's rationale is "the tables agree within 0.5% at source", yet employment counts land at 1.92%. The likely cause is that the area cells take all-ranges growth while the bands take band-specific growth. If so, the gate mostly measures that index choice, and with 0.08 points of headroom the next ITL re-pin may trip it.

3. The 24 national band bridges move cells by up to 5.7% with no bound.

   • No bound: _apply_band_bridges (national_reconciliation.py:~229) takes factor = control / raw_total and refuses only a non-positive total.
   • Wrong reason: all 24 declarations give the same reason, "about 0.5 percent short".
   • Factors:
     • income tax £12,570–15,000: 1.0573
     • taxpayers £12,570–15,000: 1.0274
     • total income £12,570–15,000: 1.0228
     • income tax £15,000–20,000: 1.0142
     • the rest: 1.002–1.008
   • Knock-on: the £12,570 income-tax bridge includes the West Midlands cell deferred in target_fit_reviewed_exclusions.json:5. That deferral was measured against a 55.1m target, which becomes about 58.3m. The body's "Reconciliation factors" section leaves the band bridges out.
   • Fix: add a max_factor_shift as the fan-out bridges have, state the real gaps per bridge, and re-measure the deferral (UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145 holds the same deferral as stale).

4. Nothing enforces the signatures the body asks for.

   • Unsigned: all 15 holds and the small-cell rule are approved_by: null, status: "proposed".
   • What the loader checks: load_uk_uprating_holds (uprating_holds.py:62-84) checks only kind and reason. expires_on is optional (:169); every entry has one today, so that part is latent.
   • Contrast: the measure exclusions refuse an entry without approved_by, approved_on or expires_on (measure_simulation.py:719).
   • Small cells: the 16 cells are labelled signed_deferred in the membership while the rule is unsigned.
   • Fix: make proposed or a null approver fail closed, so the open decision can't be merged around.

Should fix

5. "Value moves" doesn't match the committed registers, and no receipt is committed.

   • Active national rows go from 1,231 to 1,267: 36 added, not 37, and none removed.
     • dwp.hb.amount_pension_age is already active at the base, from UK: track the defects uk-data is fixing that microcosm still has #1095. The comment at test_country_spec.py:1163 says 12 + 24.
     • The two UC bands listed as removed are the ones UK: track the defects uk-data is fixing that microcosm still has #1095 already excluded, so they aren't new removals.
   • Moved: 160 rows move at register level, against the body's 168.
   • Missing from the count: the moves this PR's own reconciliation makes (the band bridges on the regional SPI rows and the CGT regional rows). So the "before" receipt seems to postdate 7a7ddbdf0.
   • Can't be re-run: tools/uk_target_surface_receipt.py needs local .codex-work inputs.
   • Please commit the before receipt taken at aca40a692 and the after receipt, or at least their diff, with the band-bridge and both fan-out grains' legs.

6. The doctrine has escape paths.

   • Relations: declared relations (31) only silence the static overlap check (cross_grain_declarations.py:373-395). Nothing tests "subset" (parts ≤ whole) numerically, so an infeasible pair would compile.
   • Coverage path: the coverage check runs only on the joint path (uk_runtime/ledger_targets.py:2464), not in the national-only loader.
   • is_uprating_hold: it returns False for any row carrying uprating_factor (uprating_holds.py:110). tenure_drift.py:119 stamps factor 1 on every non-English tenure cell, so the Welsh, Scottish and NI tenure cells (a few hundred) are held at census shares without a declaration.
   • Missing periods: a missing or unparseable period also returns "not held" instead of refusing.

7. The 919 higher-rate area cells.

   • One factor per region: each region's factor (1.19 London to 1.40 NI LAs) keeps 2023-24 shares within the region. Across regions the growth is steeper where the base share is lower: about 1.37 for NI at roughly a 5% share, against 1.19 for London at about 12%. Within-region spreads are wider than that, so a uniform factor within a region probably overstates affluent areas and understates poor ones.
   • Rounding: every one of the 919 values is a multiple of 1,000, 192 are at or below 5,000, and the minimum is 2,000. That's the rounding the PR cites for leaving additional-rate cells unbound. NI constituencies sum to 89k against 87k for NI LAs.
   • Adjudications not amended: local_binding_adjudications.json full_frs_tei_band_unavailable still says hmrc_income_by_area binds only the employment and self-employment amount and count cells. population_universe_private_households still says mid-2024.

Minor and questions

8. Land: the production-2023 parity receipt gains ledger_absent for the three ons.land.* rows, which resolved at 2023 on the base.
9. rolled_forward_by_ratio:
   • The base sums every eligible fact at the latest period with no count or duplicate check (ledger_targets.py:~2359).
   • The operands aren't required to be the same series apart from the period; only the (0.5, 2] bound guards them.
   • The current NI and Wales values are right.
10. SPREE drift: as far as I can tell it only moves social renting against everything else; the private-rent-to-owned ratio barely moves in any authority. Social-stock factors such as Rushmoor's (about 0.76) look like reclassifications, and nothing bounds them. That matters for PR B, which builds on tenure/private_rent.
11. Question: Scottish tax_band. The engine applies Scottish thresholds to income including savings and dividends, while HMRC's Scottish rates apply to non-savings, non-dividend income. This affects only Scotland's national higher and additional rows. It's tested against fakes only, as the body says.
12. Question: Scotland and Wales. The feed at 942a1fa appears to carry NRS 2025 households by council area and Welsh Government mid-2024 households by LA. The census-share hold keeps 2022 and 2021 shares instead. Is that deliberate, given rules 1 and 2 of the vintage audit's source preference?
13. Stale text.
    • docs/uk-local-cross-grain-reconciliation-802.md:115 and :165-168 (A15 and the composition partition).
    • uk_local_target_census.json:848 and :871 (the "all cells bind" counts, which now include deferred cells).
    • d10b8b96b's rationale: those UC rows were already calendar averages at the base.
    • "eleven" DfI values: 17 changed.
    • The Make zero-valued calibration targets intentionally nonzero-weighted #104 changelog says zero targets "keep a positive loss weight", but the code only reports it.

Interaction with #1145 and #1167

• Feed pins: Set UK road-fuel and energy levels in the LCFS stage on a calendar-2024 basis (#1113) #1167 pins the feed to 1ee7dfe, which is an ancestor of this PR's 942a1fa. Whichever merges second keeps 942a1fa. If Set UK road-fuel and energy levels in the LCFS stage on a calendar-2024 basis (#1113) #1167 lands first, this PR must carry its new vendored resources (desnz_road_fuel_by_vehicle.json and the Consumer Trends rows) through the re-vendoring.
• Conflicts: 20+ files with Set UK road-fuel and energy levels in the LCFS stage on a calendar-2024 basis (#1113) #1167 and 9 with UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145.
• If UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145 lands first: its 13 property rows go active (about 1,280 references), and the parity receipts, membership, counts and uk_spine.json need regenerating. No hold or declaration is needed: they carry an uprating_index and sit at one grain. But UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145's fit inherits this PR's taxpayer filter and band bridges, so it needs re-measuring.

Checked and fine:

• The nations sum to 29,082,020, and A15 doesn't double-scale.
• The nation household factors match the body.
• Deferred small cells count in their leg's control and don't inflate neighbours.
• Spine inputs move only by publisher revisions: the revised mid-2024 population and the DfI rounding. ons_household_expenditure_facts is metadata only.
• The US and loader digest moves come from solve.py being an attested kernel, with no US value change.
• The source-indexed national compile selects the same candidates by construction.

[juaristi22]

Reply to the adversarial review (Claude Code), at 40a69a911

I checked each finding against the code and the pinned feed, and they hold. Fixes are in 68a0025d8..40a69a911; one must-fix needs a Chronicle extract before it can close. Engine CI failed on two stale pins (the US spec digest and a diesel donor factor), fixed in 68a0025d8; the engine-uk lane passes locally.

Must fix

1. UC by area: confirmed, and only partly fixable here. On the feed the area cells average 6.73m across Great Britain and the national paid caseload is 6.20m, so the 0.921 is the paid share, not a households-against-benefit-units gap. Your child-band figures reproduce exactly (+2.8%, −1.5%, −3.5%, −5.1%).
   • Done here: the selector notes, the bridge and the partition no longer say "like with like". They state the two concepts, the single GB factor and the child-band spread.
   • Still open: the paid-claims area series needs a Chronicle re-extract with Payment Indicator and a re-pin. A request to Chronicle is drafted and not yet filed. The item I added to WS-C deferred adjudications and the reconciliation layer #736 misdiagnosed the gap; it is corrected.
2. Fan-out gate at local-authority grain: fixed. Your factors reproduce (1.0118, 1.0211, 1.0160, 1.0023), and so does the cause: the single-winner rule put the local-authority cells on the constituency total before the bridge leg ran.
   • The gate now reads the cells before any reconciliation, at each grain.
   • It tests source agreement: the bands over the cells before either side's uprating index. Your related point was right: 1.44% of the employment-count factor is index choice (band-by-band growth against all-ranges growth).
   • Measured source factors: 0.999 to 1.007 at constituency grain and 1.006 to 1.014 at local-authority grain, against the 2% bound. The applied factor and the index gap are receipted beside them.
3. Band bridges: bounded, with their real reasons.
   • The gap is not "about 0.5% short". It is the taxpayers Table 3.11's UK row carries in no region: 2,850k in the lowest band against 2,774k across the twelve areas. They are concentrated in the lowest band, hence 2.7% of taxpayers and 5.7% of income tax there, and under 1% above £20,000.
   • Each bridge now declares max_factor_shift (its measured gap plus half a point) and states its own gap. A wider factor is refused.
   • Not done: re-measuring the West Midlands deferral against its moved target (55.1m to about 58.3m) needs a calibration run. It is owed at the next national build.
4. Signatures: enforced.
   • Holds: an entry without approved_by and approved_on, or still proposed, is not in force. The compile refuses its rows, and the loader requires tracking and expires_on.
   • Small-cell rule: an unsigned rule defers nothing, so the 16 cells are bound again, and the runtime refuses a membership that defers under an unsigned rule.
   • test_every_committed_hold_is_signed fails until the register is signed. That is deliberate: engine-free CI stays red on that one test until the decision is taken.

Should fix

5. Value moves and receipts: corrected and committed. You were right on each count.
   • The "before" receipt was wrong. It postdated this branch's own national reconciliation, so those moves were missing. It is now the merge base aca40a692 with only the receipt tool, on the feed it pins.
   • Bound national rows: 1,167 to 1,203. 36 added, none removed, 552 moved: 360 SPI regional and 24 CGT regional cells by reconciliation, and 168 by the re-pin and the household controls.
   • Local cells: 20,885 to 21,804. 919 added, none removed, 18,374 moved.
   • Committed: docs/evidence/uk-target-surface-1123/ holds the row-level national diff, the local diff by target and grain, and the reconciliation (band bridges with bounds, both fan-out grains, every group's factors, the subset relations, the holds), plus the extraction script. The full receipts are 6.5 and 7.4 MB and stay outside the tree.
   • One caveat: unsigned holds refuse the compile, so the after receipt ran over a register signed for the run only. The receipt records the file under code_modified_paths.
   • Re-running still needs the Chronicle consumer artifact and the ladder; the tool's docstring and the evidence README now say where they come from.
6. Escape paths: closed.
   • Subset relations name their part and whole and are tested on the compiled values in both loaders. All 17 hold on the feed; the six CGT cross-table pairs carry a 1% rounding allowance.
   • The national-only loader runs the coverage check.
   • uprating_factor of one: an applier that moved nothing now stamps uprating_hold and stays a hold. The 264 held tenure cells (Wales, Scotland, Northern Ireland and Rushmoor) are declared under a new, unsigned entry.
   • A row with no readable period is refused. None exists on the feed at either grain.
7. Higher rate area cells.
   • Rounding: confirmed. 192 of the 919 cells are at or below 5,000 (155 constituencies, 37 local authorities), and the median constituency cell is 8,000. I added a proposed rule that would defer those 192. It is unsigned, so it is not in force: sign it or strike it.
   • One factor per region: not changed. It is the stated limit of a control_rescaled hold until the 2024-25 area tables arrive.
   • Adjudications: both entries are yours, signed 2026-09-02, so I have not edited them. Proposed wording is below for your approval.

Minor and questions

8. Land at 2023. The three rows vanish at period 2023 because the selector now pins publication_status: preliminary, which only the end-2025 rows carry. Without the pin the 2025 compile sees two series and refuses. I left it: the 2023 surface is the legacy production fixture. A proper fix is an edition preference in the resolver; say if you want it here.
9. Roll-forward: fixed. The base must resolve exactly one cell, or one for each geography it names, and the two operands must be one series at two periods. Both hold for Wales and Northern Ireland on the feed.
10. SPREE: confirmed and bounded. The private-rent-to-owned ratio moves by under 0.1% in every authority, so the drift carries information on social stock only; the docstring now says so. A share that moves by more than 10% in two years holds that authority's census shares. Rushmoor (social stock −22%) is the only one.
11. Scottish tax_band: you are right, and it is an engine question. The engine classifies a Scottish taxpayer on total income against the Scottish thresholds; HMRC applies Scottish rates to non-savings, non-dividend income only. A new engine test pins the case: £40,000 of earnings plus £5,000 of dividends is HIGHER in Scotland on the installed engine and BASIC in England, where HMRC would count no higher rate taxpayer. The same test pins the other classifications the targets rely on, so the measure is no longer tested against fakes only. I have not measured the size of the over-count. Your call: keep Scotland's two national cells bound and file the engine issue, or defer them until the engine is fixed.
12. Scotland and Wales by authority: not deliberate. The feed carries NRS 2025 households for all 32 councils (summing to the Scotland control exactly) and Welsh Government mid-2024 households for all 22 authorities, plus LPS stock by district. By rules 1 and 2 those should set the authority cells, with census shares kept only where no estimate exists. I have not built it: it changes the household design (three nation-scoped authority targets, their bridges and the tenure partition's parents). Say if you want it in this PR.
13. Stale text: fixed, apart from the commit message of d10b8b96b, which I can only correct by rewriting pushed history. You are right that those UC rows were calendar averages at the base, so the hold I had declared for them was never needed.

Verification at 40a69a911

• microcosm-build UK and shared: 7,133 of 7,135. One failure is the signature test, by design; the other is local (pe-uk installed in the engine-free venv).
• engine-uk: 164 of 164, the eight new tax_band tests among them.
• microcosm-calibrate 361 of 361; microcosm-graph 755 of 757 (two Orrery parser tests need a node package not installed here).
• Every new check passes on the pinned feed.

#1145 and #1167

Noted. Whichever merges second carries 942a1fa, and a paid-claims UC re-pin would move the pin again, so the three should be sequenced before any of them leaves draft.

Proposed adjudication wording (for your signature)

• full_frs_tei_band_unavailable: "…the hmrc_income_by_area family binds the unbanded per-area employment and self-employment amount and count references and the higher rate taxpayer count (England, Wales and Northern Ireland)…"
• population_universe_private_households: "…bind the ONS total-usual-resident universe as published: mid-2025 for local authorities, mid-2024 for constituencies until their mid-2025 edition, the same universe the national mid-2025 controls bind…"

[juaristi22]

Re-review (Claude Code) at 40a69a911

Static review of 67fc1dd04..40a69a911 against the previous review and the reply. I checked the code, docs/evidence/uk-target-surface-1123/ and the registers, and ran nothing. The CI run for this head was still queued.

Verdict.

• Fixed: items 3, 4, 6, 9, 10, 11 and 13.
• Reproduced exactly: the committed receipts. National goes 1,167 → 1,203 (+36, 0 removed, 552 moved = 360 SPI regional + 24 CGT regional + 168). Local goes 20,885 → 21,804 (+919, 18,374 moved).
• Needs another pass: must-fix 2 was closed by changing what the gate measures, so the factor the cells actually take is now unbounded. The deliberately red signature test also stops the whole engine-free job.

Previous items

#	Status	Note
1 UC by area	Text fixed; substance open	No "like with like" remains. The paid-claims extract is pending in Chronicle.
2 Fan-out gate	Partly fixed; see A	It now reads the pre-reconciliation cells at each grain (uk_runtime/ledger_targets.py:2621-2694). The source factors reproduce: 0.9992–1.0066 constituency, 1.0058–1.0144 LA. Only the source factor is gated.
3 Band bridges	Fixed	Bounds are required and enforced (national_reconciliation.py:82, :247), and all 24 factors sit inside them. Reasons match the receipt to 0.1 pt. The West Midlands re-measure is still owed.
4 Signatures	Fixed	The loader requires tracking and expires_on. An unsigned hold or small-cell rule is out of force, and the 16 cells are bound again. The head can't compile a UK surface until the holds are signed, which is intended. See C and D.
5 Receipts	Fixed	The numbers above verify, and the extraction script is correct. See G for provenance.
6 Escape paths	Fixed; see F	Subset relations are checked in both loaders, and the national loader runs coverage. The tenure applier now stamps uprating_hold (264 = 65 × 4 + Rushmoor × 4), and a period that can't be read is refused.
7 Higher-rate small cells	Inert, as stated	192 cells (155 + 37); the rule is unsigned so it defers nothing.
8 Land	Open, acknowledged
9 Roll-forward	Fixed	One base cell per geography; operands one series at two periods.
10 SPREE	Fixed	Rushmoor (0.763) is the only share past 10% in two years; the next widest is 6.9%.
11 Scottish tax_band	Pinned	test_uk_tax_band_measure.py pins the Scotland-vs-England case. The decision is still yours.
12 Scotland and Wales households	Open (question to María)
13 Stale text	Fixed, with new spots	See H.

68a0025d8 is legitimate:

• The diesel factor: it follows this PR's own re-pin of UK mid-2024 population (69,281,437 → 69,256,274). The tolerance is unchanged; only the centre moved.
• The US digest: it comes from this PR's solve.py edit, which us-f0-coverage.json already records.

New

A. The fan-out gate no longer bounds what the cells take.

• _assert_uk_fanout_sum_factors refuses only source_factor. applied_factor and the index gap are receipt-only.
• So LA employment counts take ×1.0211 with nothing checking it, and the 1.44-point index gap passes.
• The test that refused an applied 1.03 is gone. test_uk_fanout_sum_bridges.py:151 now asserts that a 5.4% applied factor passes.
• Gating source agreement is the right addition, but it replaces the bound instead of joining it. A future ITL re-pin that widens the band-vs-all-ranges gap would move the area targets silently.
• Fix: also bound the index (or applied) factor per grain, measured gap + 0.5 pt as the band bridges do.

B. LA cells take per-region factors up to 1.043 that nothing gates.

• Factors (reconciliation.json, constituency_over_la groups):
  • employment count, South West: 1.0430 (every other region 1.018–1.021);
  • employment amount, South West: 1.0294;
  • self-employment amount, North East: 1.0294.
• Cause: the single-winner rule scales LA cells to the constituency total region by region. Torbay and the Isles of Scilly have no SPI LA cell, so their share of the South West lands on the other South West authorities.
• Consequence: South West LA targets ask about 2.4% more than the constituency targets for the same households.
• The body's "0.996 to 1.021" and the gate's applied_factor are UK-level, so this doesn't show. It predates the fix commits.
• Fix: drop the absent authorities' constituencies from that leg's parent, or gate per-leg factors.

C. The deliberately failing test stops the engine-free job.

• tools/run_engine_free_tests.sh runs --maxfail=1 -n 2. When test_every_committed_hold_is_signed fails, pytest stops scheduling.
• So "red on that one test" means the rest of the engine-free suite goes unverified in CI on every push until the holds are signed.
• Fix: mark it xfail(strict=True) with the reason, so CI goes red only once it unexpectedly passes. Or move the check into a merge gate.

Minor

• D. "Signed" means "not proposed". uk_hold_is_signed (uprating_holds.py:162-172) and uk_small_cell_rule_is_signed (ledger_targets.py:882) accept rejected, withdrawn or draft once an approver and date are set. Use an allowlist (status == "approved").
• E. The source factor falls back silently to the uprated value when ledger_value_before_alignment is missing (_spec_value_before_alignment, ledger_targets.py:2617; also national_reconciliation.py:407). That would mix periods on one side of the gate. It should refuse instead.
• F. The subset checks are thin.
  • They compare aggregate totals only.
  • age_0_9_by_region ≤ uk_total is trivially true (ratio 0.106), while the stated constraint (all bands ≤ total less 90+) isn't tested.
  • max_excess is unbounded at runtime; only the test caps it.
  • A relation with an unbound side is skipped silently (national_reconciliation.py:460).
• G. Receipt provenance.
  • code_modified_paths records paths only, with no hash or diff, and ignores untracked files (tools/uk_target_surface_receipt.py:75). So "the signed-for-the-run overlay moved no value" can't be checked from the receipt.
  • The "before" receipt names code_head: aca40a692, but compile_uk_full_target_surface doesn't exist there. The README says so in words; the receipt doesn't.
• H. Stale text. docs/uk-local-cross-grain-reconciliation-802.md:98 ("a factor more than 2% from one is refused") and changelog.d/1123-hmrc-area-taxpayer-basis.changed.md ("within 2%") still describe the applied-factor gate.
• I. The band-bridge bounds are drift tripwires (measured + 0.5 pt by construction), not plausibility bounds. If the feed carries Table 3.11's UK row, bounding each bridge by UK row ÷ Σ regions would check the stated explanation in code.
• Nits.
  • Eight taxpayer band-bridge reasons read "the taxpayers of taxpayers Table 3.11's UK row carries".
  • The higher-rate rule says "rounding error of at least 10%"; it means up to 10% (25% at 2,000).
  • assert_uk_uprating_holds_declared(holds=...) skips the loader's validation, and a missing expires_on now raises a KeyError.
  • A deferral under a rule id later deleted from the register still passes assert_uk_local_deferrals_in_force.

Still open, as the reply says:

• the paid-claims UC extract;
• the West Midlands re-measure against its moved target;
• land at 2023;
• the Scotland and Wales household estimates;
• the Scottish tax_band decision;
• signing the holds, the tenure entry and the two small-cell rules.

No tolerance was loosened. The count edits (21,788 → 21,804, 1,598 → 1,582) follow from the 16 cells being bound again. The only test weakened is the fan-out applied-factor refusal (A).

[juaristi22]

Reply to the re-review at b8509c765 (Claude Code), now at a4df36092

All three minor items and the nits are addressed, with two exceptions that I state below. The fixes are in a864c842a..a4df36092.

The branch is rebased again, onto main at 51a263726, because #1145, #1188 and #1211 merged. Every commit id changed again: the head you reviewed, b8509c765, is now a864c842a.

CI on b8509c765 finished green on every lane, engine-us on both Python versions included. That closes the "unverified pins" nit for that head: the US spec digest, the seed digests and the coverage evidence passed where they are checked.

Minor items

1. The expiry cliff: fixed, with a clock in CI.

• The compile warns when a held target's declaration is within 14 days of its expiry (UK_HOLD_EXPIRY_NOTICE_DAYS), and the holds receipt carries the earliest expiry.
• A test fails in the same window: test_no_committed_hold_is_about_to_lapse. From 2026-11-27 the engine-free lane is red on every PR until the holds are reviewed and signed again, or replaced by a newer fact or an index. The builds still pass for those 14 days.
• That is a deliberate clock. María, say if you prefer a dated tracking issue and no test; it is a one-test removal.
• You are right that most of these holds will need a new signature, not retirement: the constituency mid-2025 ages and the SPI 2024-25 area tables land after 11 December.

2. Implied cells: bounded. Per-leg tier factors: not bounded, on purpose.

• An implied row's share of its leg is now at most 1.5 times its areas' share of the leg's census households (UK_IMPLIED_SHARE_MAX_RATIO). Past that the compile refuses: the leg's bound cells are short by more than the absent areas can explain.
• On the feed the ratios are 0.32 to 0.89:
  • Torbay and the Isles of Scilly hold 2.61% of the South West's households. They are implied at 2.33% of its employees (0.89), 1.97% of employment income (0.75), 1.75% and 1.71% of self-employment counts and income (0.67 and 0.66) and 0.85% of higher rate taxpayers (0.32).
  • The North West constituency holds 1.31% of the region's households and is implied at 0.48% of its self-employment income (0.37).
• Two refusals beside the bound: an implied area with no census households to set it against is refused, and a surface that binds no census households says in the receipt that the row is unbounded.
• I did not bound the leg factors per target.
  • The local-authority census-household legs at 0.973 to 1.038 are the nation and region controls themselves (decision "nations win"): the constituency cells take them, and the authority cells follow. A bound there would be a bound on the LFS and NRS estimates.
  • The North East self-employment amount at 1.029 is the two tables' gap at source, and the standing rule lets the constituency tier win.
  • Both stay in the receipt by leg.

3. The ladder rule: tightened, with a different guard from the one you proposed.

• An off-surface area outside a run's ladder is left out only when the membership binds some cell for that area. If the membership binds nothing for the area, the compile refuses.
• Why that closes the hole. A whole run binds that cell, and the frame refuses a bound cell that its ladder cannot place. So a production ladder that lacks Torbay refuses on Torbay's bound age and household cells before any off-surface cell is read. A run that reaches the skip is a narrowed one, which is what the fixture is.
• A test pins the premise: every committed off-surface area is one a whole run binds.
• I did not add a dev or sample mode flag. The guard above is derived from the data, so a production caller cannot switch it off.

Nits

• full_problem.py: fixed. Its own surface passes the membership's deferred and off-surface cells.
• Deferred cells in non-controlled legs: fixed. A deferred cell with a value counts in its leg's total in every leg, and adds a frame row where a control moves either tier of its leg.
• G, the before overlay: the extraction now writes the before receipt's overlay as before-overlay.diff in the evidence folder, with its digest in surface-diff.json.
• Signature dates: fixed. A hold is in force from approved_on through expires_on, as a measure exclusion is.
• The obr.fuel_duties_cars hold: not edited. The entry is signed, so its text is María's. Proposed sentence for her next review of the register: "Since microcosm#1113 this row is a measured diagnostic and not a calibration target; the hold stays because holds are asserted on the full register."
• Bisect: unchanged, and one span longer. This rebase adds a second reconciliation commit (21da310f7). The branch's commits between main and that commit carry branch-side conflict picks. A squash at merge removes the problem; that choice is María's.
• F, local parts on their grain total: noted, not changed.

Rebase onto main at 51a263726

• Nine files conflicted. 21da310f7 records the reconciliation.
• One real merge decision: the SPI property-income amounts. UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145 binds the 13 Table 3.7 amounts unscaled, on the band index. This branch puts every SPI row on a taxpayer basis. The merged target has both: main's note and index, and this branch's income_tax > 0 filter. So those 13 rows are now measured over taxpayers only, which differs from what UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145 fitted. The filter is the SPI's own universe. The 13 target values are the same on main and here. I have not measured how much the filter changes their fit; that shows at the next national build.
• National references: 1,281 active, the branch's 1,268 plus the 13 property amounts.
• The West Midlands deferral is retired on main, so nothing is deferred against the cell this branch moves. Its fit is measured with every other row at the next national build.
• No branch value moved in the rebase or in this round. The compiled surface has 1,213 bound national rows, the 1,200 of the last head plus the 13 property amounts, and the same 21,804 local cells. Every value equals its value at b8509c765.

Left over from #1145

Vahid's review of 63943a76b had three items that #1145 recorded in its description only. They are now in experiments/1106-uk-property-income-receipts.md (section C11, cd39d037e): the HBAI size of the FRS add-back at 2025 with PolicyEngine/policyengine-uk#2390, furnished holiday lettings from 2025-26, and capital netting in the finance-cost draw. The figures are the description's. The control arm on the current base is still owed.

Value moves

• Against the new main: 1,177 to 1,213 bound national rows. 36 added, none removed, 552 moved, the same rows as before. 20,885 to 21,804 local cells: 919 added, none removed, 18,811 moved.
• This round moves no value. The share bound, the ladder guard and the holds notice are checks.

Verification at cd39d037e

cd39d037e is the last code commit; a4df36092 adds only the evidence files.

• engine-uk: 180 of 180.
• The engine-free job's files: 16,693 tests; 141 skip and 6 fail. The six are the same as in the last round and none is from this branch: two Orrery parser tests (a node package), one CGT test (pe-uk installed in the engine-free environment), one telemetry test (unstable under parallel workers), one head-to-head scorer test (needs Python 3.14.7), and one microcosm-data test that reads a live Hugging Face pointer.
• Receipts: the after receipt ran clean at cd39d037e on a clean tree; the before receipt is main at 51a263726. On the feed: 44 off-surface frame rows, an empty areas_outside_the_ladder, all 20 subset tests, and an earliest hold expiry of 2026-12-11 at both grains.
• Lint, the CI test plan, the burndown check and the vendored-resource check: ok.
• Not run here: the engine-us lane (no US engine on this machine). The US spec digest is unchanged by this rebase, and the coverage tool passes its 41 inventory checks on the merged code.
• Not run: the dense re-solve.

Still open

• UC by area: UC households by area: re-extract the constituency, local-authority and by-children series with Payment Indicator as a dimension (PolicyEngine/microcosm#1198) chronicle#339.
• Scottish tax_band: tax_band compares a Scottish taxpayer's total income with the Scottish thresholds; the Scottish rates apply to non-savings, non-dividend income only policyengine-uk#2350; the decision on Scotland's two national cells is María's.
• The two small-cell rules are unsigned.
• The holds expire on 2026-12-11. The new test turns CI red from 2026-11-27.
• Land at 2023 and the Scotland and Wales household estimates stand as before.

[juaristi22]

The taxpayer filter on the 13 SPI property-income rows: measurement and decision (Claude Code)

Decision (María, 2026-10-12): keep the filter. The 13 SPI Table 3.7 property-income amounts that #1145 binds are measured over taxpayers only on this branch (income_tax > 0), as every other SPI row is. main measures them over every person in the band. The 13 target values are the same on both sides.

Why the two measures differ

• Table 3.7 counts taxpayers only. The feed records these facts with the entity role taxpayer.
• Some landlords in the bands pay no income tax. The main cause is the s. 274A finance-cost reduction, which UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145's finance costs bring into play. On policyengine-uk 2.123.0, a landlord with £20,000 of profit, £10,000 of finance costs and no other income has a total income of £20,000 and an income tax of £0: the reduction of £1,486 removes the whole liability.

What was measured

• Candidate: arm c9 of UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145, calibrated on main's measure (no filter).
• Engine: policyengine-uk 2.123.0, a plain Microsimulation at 2025. The run reproduces UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145's figures: £35.86bn of property_income on 2.80m landlords.
• Per band of total_income: the weighted property_income of all people and of people with income_tax > 0; the weighted landlords (property_income > 0); and the landlords who pay no tax, with those who have property_finance_cost_relief > 0.

Total income band	All people, £bn	Taxpayers, £bn	Gap, £bn	Gap	Landlords, thousands	Of them, pay no tax	Of them, with finance-cost relief	Records that pay no tax
£12,570 to £15,000	0.74	0.67	0.065	8.8%	106.1	7.9	0.2	103
£15,000 to £20,000	1.73	1.58	0.158	9.1%	228.5	13.2	7.0	30
£20,000 to £30,000	4.02	3.63	0.384	9.6%	434.2	21.6	19.5	42
£30,000 to £40,000	4.47	4.37	0.093	2.1%	411.7	3.6	3.2	16
£40,000 to £50,000	4.44	4.40	0.040	0.9%	367.0	1.0	1.0	2
£50,000 to £70,000	6.90	6.90	0.000	0.0%	486.6	0.0	0.0	0
£70,000 to £100,000	4.81	4.81	0.000	0.0%	277.0	0.0	0.0	0
£100,000 to £150,000	3.30	3.30	0.000	0.0%	171.5	0.0	0.0	0
£150,000 to £200,000	1.40	1.40	0.000	0.0%	63.4	0.0	0.0	0
£200,000 to £300,000	1.34	1.34	0.000	0.0%	49.4	0.0	0.0	0
£300,000 to £500,000	0.98	0.98	0.000	0.0%	27.7	0.0	0.0	0
£500,000 to £1,000,000	0.61	0.61	0.000	0.0%	13.0	0.0	0.0	0
£1,000,000 and over	0.55	0.55	0.000	0.0%	23.3	0.0	0.0	0
All 13 bands	35.29	34.55	0.740	2.1%	2,659.4	47.3	30.8	193

• Below £12,570 (outside every band, on both measures): £0.58bn on 138 thousand landlords.
• The engine run is at 2025 values, so the banded total is 5.0% above the sum of the 13 targets (£33.60bn), which the calibration measures on its own path. The shares are the result; the £ gap is about £0.7bn on the targets' basis.

What it means

• People who pay no income tax hold 2.1% of the property income in the bands (£0.74bn of £35.29bn).
• The gap is in the low bands: 9.4% in the three bands below £30,000, 1% to 2% from £30,000 to £50,000, and nil above £50,000.
• 47,281 of 2.66m landlords in the bands pay no income tax. 30,823 of them have a finance-cost reduction, which takes their liability to nil. In the lowest band most have another cause, which I did not decompose.
• At the next calibration the three low bands start about 9% short on this branch's measure, so the solver raises the weight of taxpaying landlords there. UK landlords: bind the SPI net property amounts, leave property_wealth to the engine, add receipts and finance costs #1145's fit of 0.3% on these rows was measured on main's measure and does not carry over.
• Total property income ends about 2% higher than on main's measure, because the landlords who pay no tax are then outside the target and not inside it. The PRIS diagnostics move up with it. Neither is measured: no calibration was run.
• The rows now depend on the modelled finance costs, which decide who pays no tax. The gap rests on 193 records.

The script

"""Property income by SPI total-income band on the #1145 c9 candidate: all people against taxpayers."""
import json, os, sys, warnings
os.environ.setdefault("HDF5_USE_FILE_LOCKING", "FALSE")
warnings.simplefilter("ignore")
import numpy as np
import importlib.metadata as metadata
from policyengine_uk import Microsimulation
YEAR = 2025
dataset = sys.argv[3]  # the calibrated c9 candidate, microcosm_uk_2024_25.h5
sim = Microsimulation(dataset=dataset)
def col(name):
    v = sim.calculate(name, YEAR)
    return np.asarray(v.values, dtype=float), np.asarray(v.weights, dtype=float)
prop, w = col("property_income")
total, _ = col("total_income")
tax, _ = col("income_tax")
relief, _ = col("property_finance_cost_relief")
costs, _ = col("property_finance_costs")
edges = [12570, 15000, 20000, 30000, 40000, 50000, 70000, 100000, 150000, 200000, 300000, 500000, 1000000, np.inf]
receipt = json.load(open(sys.argv[1]))  # a target-surface receipt, for the 13 target values
targets = {r["name"]: r["value"] for r in receipt["national"] if r["name"].startswith("hmrc/property_income_income_band_")}
landlord = prop > 0
payer = tax > 0
rows = []
for lo, hi in zip(edges[:-1], edges[1:]):
    band = (total >= lo) & (total < hi)
    all_amt = float((prop * w)[band].sum()); pay_amt = float((prop * w)[band & payer].sum())
    all_n = float(w[band & landlord].sum()); pay_n = float(w[band & landlord & payer].sum())
    nonpay = band & landlord & ~payer
    rows.append({"band": lo, "all_gbp": all_amt, "taxpayers_gbp": pay_amt, "all_landlords": all_n, "taxpayer_landlords": pay_n,
                 "nonpayer_with_relief": float(w[nonpay & (relief > 0)].sum()), "nonpayer_records": int(nonpay.sum())})
out = {"engine": metadata.version("policyengine-uk"), "year": YEAR, "dataset": dataset, "bands": rows,
       "below_12570": {"all_gbp": float((prop*w)[total < 12570].sum()), "landlords": float(w[(total < 12570) & landlord].sum())},
       "all_property_income_gbp": float((prop*w).sum()), "landlords": float(w[landlord].sum()),
       "targets": targets}
json.dump(out, open(sys.argv[2], "w"), indent=1)
ta = sum(r["all_gbp"] for r in rows); tp = sum(r["taxpayers_gbp"] for r in rows)
print("engine", out["engine"], "| all property income £bn", round(out["all_property_income_gbp"]/1e9,2), "landlords m", round(out["landlords"]/1e6,3))
print("in the 13 bands: all people £bn", round(ta/1e9,2), "| taxpayers £bn", round(tp/1e9,2), "| non-taxpayers £bn", round((ta-tp)/1e9,2), f"({(ta-tp)/ta:.1%})", "| targets £bn", round(sum(targets.values())/1e9,2), len(targets))
print("landlords in bands: all m", round(sum(r['all_landlords'] for r in rows)/1e6,3), "taxpayers m", round(sum(r['taxpayer_landlords'] for r in rows)/1e6,3), "non-taxpayers with relief m", round(sum(r['nonpayer_with_relief'] for r in rows)/1e6,3), "records", sum(r['nonpayer_records'] for r in rows))
for r in rows:
    print(f"  {r['band']:>8}: all £{r['all_gbp']/1e9:6.2f}bn taxpayers £{r['taxpayers_gbp']/1e9:6.2f}bn gap {1-r['taxpayers_gbp']/r['all_gbp'] if r['all_gbp'] else 0:6.1%} | landlords {r['all_landlords']/1e3:7.1f}k taxpayers {r['taxpayer_landlords']/1e3:7.1f}k | non-payers with relief {r['nonpayer_with_relief']/1e3:6.1f}k ({r['nonpayer_records']} records)")
print("below 12,570: £bn", round(out["below_12570"]["all_gbp"]/1e9,2), "landlords m", round(out["below_12570"]["landlords"]/1e6,3))

The output

{
 "engine": "2.123.0",
 "year": 2025,
 "bands": [
  {
   "band": 12570,
   "all_gbp": 739209114.2499585,
   "taxpayers_gbp": 673939177.727847,
   "all_landlords": 106072.8392278701,
   "taxpayer_landlords": 98167.14690719545,
   "nonpayer_with_relief": 164.04813385009766,
   "nonpayer_records": 103
  },
  {
   "band": 15000,
   "all_gbp": 1734569916.506525,
   "taxpayers_gbp": 1576771007.521718,
   "all_landlords": 228460.7329572402,
   "taxpayer_landlords": 215265.92346588895,
   "nonpayer_with_relief": 7034.7943078279495,
   "nonpayer_records": 30
  },
  {
   "band": 20000,
   "all_gbp": 4016857379.8852468,
   "taxpayers_gbp": 3632881784.709773,
   "all_landlords": 434235.018026039,
   "taxpayer_landlords": 412633.3857693523,
   "nonpayer_with_relief": 19504.546014398336,
   "nonpayer_records": 42
  },
  {
   "band": 30000,
   "all_gbp": 4467844392.178325,
   "taxpayers_gbp": 4374841463.2569685,
   "all_landlords": 411677.84457619116,
   "taxpayer_landlords": 408056.6843499057,
   "nonpayer_with_relief": 3161.4855183959007,
   "nonpayer_records": 16
  },
  {
   "band": 40000,
   "all_gbp": 4440908596.628444,
   "taxpayers_gbp": 4401200624.90477,
   "all_landlords": 367039.04541874677,
   "taxpayer_landlords": 366080.90053547174,
   "nonpayer_with_relief": 958.144883275032,
   "nonpayer_records": 2
  },
  {
   "band": 50000,
   "all_gbp": 6896780804.749105,
   "taxpayers_gbp": 6896780804.749105,
   "all_landlords": 486628.99004568905,
   "taxpayer_landlords": 486628.99004568905,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  },
  {
   "band": 70000,
   "all_gbp": 4811365242.8906,
   "taxpayers_gbp": 4811365242.8906,
   "all_landlords": 276958.4430429302,
   "taxpayer_landlords": 276958.4430429302,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  },
  {
   "band": 100000,
   "all_gbp": 3301603916.086019,
   "taxpayers_gbp": 3301603916.086019,
   "all_landlords": 171501.82781397924,
   "taxpayer_landlords": 171501.82781397924,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  },
  {
   "band": 150000,
   "all_gbp": 1401580825.345178,
   "taxpayers_gbp": 1401580825.345178,
   "all_landlords": 63391.11676461995,
   "taxpayer_landlords": 63391.11676461995,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  },
  {
   "band": 200000,
   "all_gbp": 1337064761.5412693,
   "taxpayers_gbp": 1337064761.5412693,
   "all_landlords": 49441.01511074789,
   "taxpayer_landlords": 49441.01511074789,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  },
  {
   "band": 300000,
   "all_gbp": 976727442.7887601,
   "taxpayers_gbp": 976727442.7887601,
   "all_landlords": 27743.166271453723,
   "taxpayer_landlords": 27743.166271453723,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  },
  {
   "band": 500000,
   "all_gbp": 610383423.6390421,
   "taxpayers_gbp": 610383423.6390421,
   "all_landlords": 12970.121472515166,
   "taxpayer_landlords": 12970.121472515166,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  },
  {
   "band": 1000000,
   "all_gbp": 552367083.8276165,
   "taxpayers_gbp": 552367083.8276165,
   "all_landlords": 23303.13035662286,
   "taxpayer_landlords": 23303.13035662286,
   "nonpayer_with_relief": 0.0,
   "nonpayer_records": 0
  }
 ],
 "below_12570": {
  "all_gbp": 575829853.843927,
  "landlords": 137855.0947300978
 },
 "all_property_income_gbp": 35863092754.16002,
 "landlords": 2797278.385814743,
 "targets": {
  "hmrc/property_income_income_band_100_000_to_150_000": 3151071428.571429,
  "hmrc/property_income_income_band_12_570_to_15_000": 703500000.0,
  "hmrc/property_income_income_band_150_000_to_200_000": 1332747524.7524753,
  "hmrc/property_income_income_band_15_000_to_20_000": 1651431535.2697096,
  "hmrc/property_income_income_band_1_000_000_to_inf": 525041487.8397711,
  "hmrc/property_income_income_band_200_000_to_300_000": 1272816349.3840985,
  "hmrc/property_income_income_band_20_000_to_30_000": 3831324110.671937,
  "hmrc/property_income_income_band_300_000_to_500_000": 927337625.9798431,
  "hmrc/property_income_income_band_30_000_to_40_000": 4252163461.538461,
  "hmrc/property_income_income_band_40_000_to_50_000": 4229783653.8461537,
  "hmrc/property_income_income_band_500_000_to_1_000_000": 580715469.6132597,
  "hmrc/property_income_income_band_50_000_to_70_000": 6559199999.999999,
  "hmrc/property_income_income_band_70_000_to_100_000": 4579293333.333333
 },
 "dataset": "arm c9 of PolicyEngine/microcosm#1145 (microcosm_uk_2024_25.h5)"
}