UK: take up #1095's actionable items on policyengine-uk 2.165.1

changelog.d/1095-ledger-observation-period.added.md

@@ -0,0 +1 @@
+A Ledger target reference can declare an `observation_period`, the period its facts are selected at when that differs from the target period the model is measured at. Selection, the latest-period choice and the resolved-fact guard read it, the target keeps its own period, and the compiled metadata records it. Operations that compose periods of their own refuse it (microcosm#1095 D5).

changelog.d/1095-ledger-window-ratio-complement.added.md

@@ -0,0 +1 @@
+A `calendar_year_window_scaled_by_ratio` numerator can declare `complement`: its cells come off the denominator, so the share is the rest of the published universe, for example chargeable dwellings less second homes and empty dwellings (microcosm#1095 D14f). The compiled metadata records the rule and the formula.

changelog.d/1095-uk-council-tax-main-residence-share.changed.md

@@ -0,0 +1 @@
+`obr.council_tax` binds the main-residence share of OBR's council tax receipts: the calendar-2025 window is scaled by England's chargeable dwellings less its second homes and empty dwellings, over its chargeable dwellings (MHCLG Council Taxbase, October 2025). The frame carries only main-residence bills, so binding the whole line pushed them up (María's ruling of 2026-10-29, microcosm#1095 D14f).

changelog.d/1095-uk-council-tax-nation-rows-pin-and-hold.fixed.md

@@ -0,0 +1 @@
+`obr.council_tax_england`, `obr.council_tax_scotland` and `obr.council_tax_wales` compile their facts: OBR's table 4.1 is UK-wide, so Chronicle stamps the three nation rows K02000001, and they now pin to the UK, as `obr.council_tax` and `obr.domestic_rates` do, instead of the nation codes the substring rule gave them, where no fact sits (microcosm#1095 D14e). Each binding still scopes the model column to its nation. `obr.council_tax` binds their sum scaled to main residences (D14f), and each nation row is scaled by the same share, so the three sum to it; measure exclusions hold the three rows as diagnostics, measured on every evaluation as evidence for the CT-C comparison (María's ruling of 2026-10-29).

changelog.d/1095-uk-declared-selector-geography.changed.md

@@ -0,0 +1 @@
+A target's Chronicle selector can declare the geography its fact sits at, and the declaration wins over the generator's inferred pin: a wrong inference otherwise reads as "no fact" (microcosm#1095 R8, María's ruling of 2026-10-29). The authoring refuses a pin that disagrees with a declared geography, so the two cannot drift apart.

changelog.d/1095-uk-dfe-january-2026-census.changed.md

@@ -0,0 +1 @@
+DfE's eight funded-childcare rows bind the January 2026 census, the spring term of model year 2025 and the month and age basis the spring-term measures use, in place of the January 2025 census a year out (María's ruling of 2026-10-29, microcosm#1095 D5). Each declares `observation_period` 2026-01; the fact follows the 2025 target period, so the three in-year snapshot holds retire.

changelog.d/1095-uk-disclosure-floor-three.changed.md

@@ -0,0 +1 @@
+UK stage receipts, stage-health gates and report tools withhold a figure only when it rests on one or two distinct source households (CD171's absolute floor), not on fewer than ten (María's ruling of 2026-10-29, [recorded on microcosm#1095](https://github.com/PolicyEngine/microcosm/issues/1095#issuecomment-6541929216)). One constant, `receipt_counts.UK_DISCLOSURE_FLOOR`, sets the floor. A withheld figure reads `"<3"`. Support, extreme and rounded record counts read `"3-9"` from three to nine, so a count just above the floor never narrows a small group by difference. The statistical floors stay: the zero-support floor, donor-cell minimums, the private-rent tool's ten-row group rule, and the SPI coherence tool's cell floor for correlations, medians and quantiles, since a median of a few records is one record's own value. The SPI Public Use Tape's own two conditions (SN 9422) are unchanged: no attempt may be made to attribute a record to a named individual, and material that includes analyses of the tape goes to HMRC at least ten working days before it is distributed or published.

changelog.d/1095-uk-gate-battery-current-grain.changed.md

@@ -0,0 +1 @@
+The UK gate battery describes the current build: the zero-weight strata are capped at 0 (the June caps of 100,000 counted the retired ten-clone grain, and the national seam ships no zero-weight row), the export-surface gate drops its all-zero incapacity benefit exclusion, which never matched the pinned reference, and the battery records why it declares no macro_realism band (microcosm#1095, from #688, #796 and #757).

changelog.d/1095-uk-signed-exclusion-rulings.changed.md

@@ -0,0 +1 @@
+Scotland's households with three or more children stay signed out with a corrected reason, Chronicle's Scotland cut counting couple households only (microcosm#1095 D3); ONS's small-area income rows are kept out as circular, ONS modelling them from the FRS (D18a); and Northern Ireland's UC area cells stay out, DfC counting claimants rather than households (D18c). María's rulings of 2026-10-29.

changelog.d/1095-uk-slc-devolved-repayments-by-nation.changed.md

@@ -0,0 +1 @@
+SLC's devolved student loan repayments are one row per nation, `slc.repayments.wales_total`, `slc.repayments.scotland_total` and `slc.repayments.northern_ireland_total`, each declaring its nation's geography in its selector, in place of one devolved row that compiled no fact at the Great Britain pin (microcosm#1095 R4). Each carries SLC's 2024-25 outturn to 2025 by policyengine-uk's OBR average earnings index, since repayments are a share of earnings above the plan thresholds (María's ruling of 2026-10-30). Wales binds; Scotland and Northern Ireland are held as diagnostics until their borrowers' own plan thresholds are modelled (María's ruling of 2026-10-29).

changelog.d/1095-uk-stale-resource-and-source-text.changed.md

@@ -0,0 +1 @@
+Three vendored bus resources now say that the bus fare and bus support pricing read them, not the retired LCFS and ETB rakes, and `dft_bus_value_anchors.json` lists `bus_support_pricing` among its readers. The Incapacity Benefit SPI leaf's note now says the leaf is a structural zero in the FRS 2023-24 and 2024-25. The charter-H2 graph fixture and the release input coverage manifest follow the stage contract digests (microcosm#1095).

changelog.d/1095-uk-support-floor-deferrals-follow-scope.fixed.md

@@ -0,0 +1 @@
+The local generator's support-floor deferrals for the Isles of Scilly and the City of London follow each target's area scope: the nine Welsh and eight Scottish local-authority families have no cell at those English authorities and no longer list them (microcosm#796 item, via microcosm#1095).

changelog.d/1095-uk-take-up-stage-signal-gates.added.md

@@ -0,0 +1 @@
+The twelve take-up flags the terminal uk_take_up_signal gate measures are checked at their drawing stages too: frs_take_up, frs_person_draws and frs_household_draws each record the drawn flags' shares at the stage's weights, childcare_copy_take_up records the five childcare flags it re-draws for the copies, and four assembled-phase stage-health gates hold them to their contract targets within the terminal gate's bounds, so a draw defect fails at its stage apart from calibration's drift (María's ruling of 2026-10-29, microcosm#1095). A flag whose population rests on one or two source households is flagged unchecked at its stage, never failed (María's ruling of 2026-10-26); the terminal gate still measures it.

changelog.d/1095-uk-window-rule-for-projected-rows.changed.md

@@ -0,0 +1 @@
+Scottish Child Payment spending, HMRC's Tax-Free Childcare top-up, Northern Ireland's Translink bus passenger receipts and SLC's England repayment rows are measured at the calendar-2025 window of their fiscal years, three twelfths of 2024-25 and nine twelfths of 2025-26, like every other fiscal-year fact (María's ruling of 2026-10-29, microcosm#1095 R7). They took the 2025-26 fact alone before.

changelog.d/721-stage-record-rewrites.changed.md

@@ -0,0 +1 @@
+Stage records list the columns each stage rewrote, with their nonzero share after it, so build provenance names the last writer of a column; the UK spine manifest carries them, and the final nonzero-share list reads them instead of a hard-coded carve-out (microcosm#721, via microcosm#1095).

changelog.d/uk-policyengine-uk-2-165-1.changed.md

@@ -0,0 +1 @@
+The UK lock moves from policyengine-uk 2.164.2 to 2.165.1 (microcosm#1095). 2.164.3 (PolicyEngine/policyengine-uk#2736) counts other investment income in market income and HBAI household net income, and 2.164.4 (PolicyEngine/policyengine-uk#2746) in `capital_income`. So the stages that band households on HBAI income with its deductions added back now count it, as the LCFS, NEED and NTS gross incomes they match do, and the WAS stage's `capital_income` predictor now matches the donor's gross income from all financial investments (`DVGIINVR8`). 2.165.0 (PolicyEngine/policyengine-uk#2750) adds the company car and car fuel benefits in kind and the Class 1A NICs on them; the build writes none of their nine person inputs, so both benefits stay zero, as before, and the inputs are signed at their default. 2.165.1 (PolicyEngine/policyengine-uk#2754) lets a family keep Tax-Free Childcare in place of a smaller Universal Credit award, and takes no debt deductions from an award only a reform creates. No economic assumption moves, so the vendored engine pins, the bonus-sacrifice anchor and the CGT projection path keep their values; only their version stamps move.