Pin the scotgov council-tax stock targets to Scotland (activates the nine held CTAXBASE references)

[MaxGhenis]

Summary

Nine UK calibration targets — scotgov.council_tax_stock.band_a … band_h and .total — have been held as no_fact_at_or_before_period since the target surface was regenerated, although Chronicle has carried their facts (the CTAXBASE September 2025 chargeable dwellings, package scotgov-council-tax-bands-2025) all along. The blocker is the geography pin derivation in tools/generate_uk_target_references.py::_geography_id_for_target: the substring rule sees no "scotland" in "scotgov", falls through to the UK pin (K02000001), and the pin is merged into the selector (target_reference_authoring.py _target_selector) and matched against fact.geography.id — which Chronicle correctly stamps S92000003 for the Scotland row. Zero matches ever.

The fix is scoped to the stock family, as the review of the chronicle-side packages recommended: a scotgov.council_tax_stock. prefix rule returning Scotland, deliberately not a blanket scotgov.* rule (which would also re-pin scotgov.scottish_child_payment_spending, a source_projection target whose activation is a separate decision). The family accounting now counts both stock publishers (VOA for England and Wales, CTAXBASE for Scotland).

Regenerated uk/target_references.json and uk/target_reference_membership.json with the committed generator against the same feed vintage the surface was built from (rebuilt byte-identically from chronicle 1cab809 — main at the surface's regeneration date — with build-bundle --suite uk + build-consumer-artifact; 107,550 facts, sha256 226358e73e7c449e71a3f6dc91a72e8e3941e0f14265943d81c990edb21c2a6c, matching the 226358e7… the feed commit recorded). Membership movement is exactly the nine stock targets flipping to active with their CTAXBASE September 2025 values (band A 498,707 … band H 14,481, total 2,623,149) and month-to-year uprating holds (2025-09 → 2025):

• active: 388 → 397
• no_fact_at_or_before_period: 27 → 18
• council_tax_stock family: 9 → 18 active declared rows

No other target's status, pin, candidate counts or resolved value changes.

A regression test pins the scoped rule: the nine stock ids resolve to S92000003, scotgov.scottish_child_payment_spending keeps its UK pin, and the full set of Scotland-pinned targets equals the stock family plus the ids the substring rule already matched.

Context and remaining holds

Chronicle PR PolicyEngine/chronicle#189 adds the nation-level council tax billed/collectable totals and CTR costs and documents this activation question. The four obr.council_tax_*/obr.domestic_rates holds are deliberately untouched here: they fail on the same pin-vs-stamp mechanics (nation pins vs the OBR rows' K02000001 stamp), but activating them interacts with the already-active obr.council_tax UK total, which is Table 4.1 row 19 — the three nation rows (£49.569bn FY2025) plus NI domestic rates (£0.494bn) plus a council tax accruals adjustment (£0.862bn) that has no counterpart in the bound household variable — so the UK binding should be redesigned before the decomposition activates (chronicle's stacked PR PolicyEngine/chronicle#192 adds the row-15/17/18 facts that make that redesign expressible). The remaining 14 holds (dwp.pip.*, slc.*, scotgov.scottish_child_payment_spending) show the same pin-class signature (matched_fact_count_overall: 0 with GB/UK pins) and deserve the same case-by-case treatment.

Tests

• test_scotgov_council_tax_stock_pins_are_scotland_and_the_rule_is_scoped (new).
• test_uk_target_reference_membership_report_is_packaged, test_country_spec.py and test_uk_national_calibration.py reference-count pins updated for the regenerated surface (397 active, 18 no-fact holds, council_tax_stock family 18).
• uv run pytest and uv run ruff check . green.

🤖 Generated with Claude Code