UK CGT distribution: age/region conditioning, banded and residential targets, asset types (#725, #467)

changelog.d/467-cgt-band-targets.added.md

@@ -0,0 +1 @@
+Bind the HMRC CGT 2024-25 individual observations by size of gain (Table 2.1a, twelve bands from GBP 3,000 under the incumbent `hmrc/cgt_taxpayers_band_<lower>` and `hmrc/capital_gains_band_<lower>` names, closing #467), by age band (Table 6: taxpayers, gains and liability for the eight adult bands) and by country and region (Table 5 over the twelve-area region tier, restated on the individuals basis by the Table 1 share through `scaled_by_ratio`) as UK calibration targets, alongside the three national totals. The 0-2,999 gain band (below the 2024 annual exempt amount) and the 0-15 age band (no under-16 gainers on the frame) are signed out row by row; the Table 3 fence stays in place and records the promotion.

changelog.d/725-cgt-age-region-conditioning.changed.md

@@ -0,0 +1 @@
+The UK CGT amounts redraw (`hmrc_cgt_gains` and its spine projection) now conditions on age and household region as well as the taxable-income proxy: the allocation targets are raked to the Table 6 age and Table 5 region margins (the latter on the individuals basis via the Table 1 share) from the vendored `hmrc_cgt_conditioning_facts.json`, and gainers are walked inside income x age x region cells with a pooled per-income-band fallback where a cell lacks support. The stage summary reports achieved against published at the Table 6 and Table 5 grain and the age x gain-band joint microcosm#725 measured, plus the rake and fallback receipts. The rake also holds Table 6 gains by age as one linear constraint per age group (an exponential tilt of the group's cells that keeps its count), so each age group carries its published gains as well as its taxpayers, and the youngest group is 16-24 on its own (seven age groups) so its rows are no longer ranked under the 25-34 carriers. Each cell walk's signed rounding error is carried into the next cell's boundaries for the same gain band, so a band's walked total within an income band stays within one person's weight of its raked total instead of accumulating the overshoot of many cell walks.

changelog.d/725-cgt-asset-type-stage.added.md

@@ -0,0 +1 @@
+Add the `hmrc_cgt_asset_type_spine` UK spine stage after the CGT amounts redraw. Every liable gainer (net gains above the annual exempt amount) carries a main asset type in the new person column `capital_gains_asset_type` and, where flagged as a residential property disposer, its net gain in `capital_gains_residential_property`. The residential flag is a logistic in log gains solved so the household-weighted count and gains of flagged persons equal HMRC Table 8a 2024-25 restated on the individuals basis by the Table 8b individuals/all share, realised by seeded systematic sampling; the non-residential remainder draws one of the five Table 7 asset types from a size-tilted categorical whose type weights reproduce the fenced Table 7 2023-24 gains shares (diagnostic only). The rows come from the new vendored `hmrc_cgt_asset_type_facts.json`; the stage declares its operations in the manifests, a stage-health gate holds the realised residential totals within 5 percent, an enum-domain gate holds the column to the declared values, and the graph parity fixture carries synthetic facts sized to its frame.

changelog.d/725-cgt-donor-vintage.changed.md

@@ -0,0 +1 @@
+The CGT size-band donor stage (`cgt_band_donors`) now takes its nine retained HMRC Table 2.1a bands from the vendored 2024-25 conditioning facts (392,000 individuals, GBP 118.2bn from GBP 12,300 upward) instead of the retired hand-extracted 2023-24 copy `hmrc_cgt_size_bands.json`, so donor starting weights sit on the same vintage as the size-of-gain calibration targets; the reviewed retained-mass pins are asserted at stage time.

changelog.d/725-cgt-residential-property-targets.added.md

@@ -0,0 +1 @@
+Bind HMRC Table 8a 2024-25 residential property taxpayers and gains (UK residents, all reporting channels, restated on the individuals basis by the Table 8b share through `scaled_by_ratio`) as UK calibration targets `hmrc.cgt.residential_property_taxpayers` and `hmrc.cgt.residential_property_gains`, measured over the liable gainers the asset-type stage flags residential. Table 7's asset-type rows stay fenced.

changelog.d/725-cgt-table3-joint-2024-25.changed.md

@@ -0,0 +1 @@
+The UK CGT amounts redraw now draws from HMRC Table 3 for 2024-25 (the 2026 release's joint of individuals and gains by size of gain and taxable income, with the taxable-income bands moved onto the GBP 125,140 Personal Allowance taper edge) copied verbatim from the pinned Chronicle feed into `hmrc_cgt_conditioning_facts.json`, instead of the 2025 release's 2023-24 ODS rescaled onto the 2024-25 band levels. The pinned `Table_3_2025_Size_of_gain_by_income.ods` input, its `verify_pinned_cgt_ods` operation, the `cgt_published_fact_surface` artifact and the `--cgt-ods` build argument are retired; both CGT manifests declare a `verify_vendored_fact_resource` operation that checks the resource's recorded feed against `chronicle_feed.json` before a row is read, and the loader refuses a Table 3 row set whose band totals disagree with the Table 2.1a bands folded onto them beyond publication rounding.

changelog.d/725-chronicle-feed-repin-c5e5bf8.changed.md

@@ -0,0 +1 @@
+Re-pin the UK Chronicle consumer feed to `c5e5bf8` (PolicyEngine/chronicle#271 and #273: DESNZ subnational energy and NEED 2024 packages; HMRC CGT Tables 7, 8 and 9), 276,205 rows. Every national and local reference, membership register, census, validation-level block, vendored resource and compile-parity receipt regenerates from the new artifact; no compiled target value moved on either surface, and the `need_energy_facts.json` selection now carries both NEED vintages (456 rows).

changelog.d/725-filter-aware-gated-provider.changed.md

@@ -0,0 +1 @@
+`parameter_gated_threshold` bindings now honour their `filters`, `household_conditions` and a declared `groupby_variable` band after the gate, and accept a geography fan-out predicate, so a gated measure (the AEA-gated CGT taxpayer proxy) can be sliced into published age, region and size-of-gain rows instead of publishing the national population on every row. Numeric universe constraints on a Chronicle fact (`cgt_gain >= 3000`, `age >= 16`) are stamped as `ledger_filter_<variable>_lower_bound` / `_upper_bound` metadata so banded measures can read edges that publishers state as constraints rather than dimensions.

changelog.d/725-scaled-by-ratio-value-operation.added.md

@@ -0,0 +1 @@
+Add the `scaled_by_ratio` value operation for Ledger target references: a published cell is translated by the quotient of two other facts (`base * numerator / denominator`, ratio in (0, 1]), with the base, numerator, denominator, ratio and member fact keys recorded on the compiled spec. It restates HMRC CGT Table 5 country/region cells, which include trusts, on the individuals basis of the national targets using the Table 1 individuals/total share. The runtime candidate prefilter and the authoring compile now keep the facts an operand selector reaches outside the cell's own selector.