UC households by area: re-extract the constituency, local-authority and by-children series with Payment Indicator as a dimension (PolicyEngine/microcosm#1198) #339

Microcosm binds Universal Credit households by constituency and local authority against the national paid-claims caseload, and the two sides count different things (PolicyEngine/microcosm#1198 review, must-fix 1).

What the feed carries at 942a1fa

• Area packages (dwp/uc_households_by_constituency_january_2025_may_2026, its local-authority counterpart and the by-children package): the Stat-Xplore UC_Households extract of 7 October 2026 has only the reference month and the area field. With no Payment Indicator in the query, each cell counts every household on UC, nil-award claims included.
• National rows: dwp.uc_households by family type and by number of children carries payment_indicator, and Microcosm binds Yes (the paid-claims contract in Microcosm's docs/uk-uc-paid-target-contract.md).

Size, calendar-2025 means for Great Britain:

• area cells: 6.73m at both grains;
• national paid caseload, as Microcosm binds it: 6.20m;
• the gap, 0.53m, matches the national series' claims not in payment (about 0.54m in the months that carry the indicator).

Request

Re-extract the three area series from UC_Households with Payment Indicator as a dimension, monthly from January 2025:

1. households by Westminster constituency (PCON24);
2. households by local authority;
3. households by constituency and number of children.

Same semantics as the current packages: publisher values as published, one record set per month. Keeping both indicator values (Yes and No) lets Microcosm select Yes and keeps the all-claims total checkable as their sum. If Stat-Xplore suppresses or perturbs small No cells, keep the publisher's cells and say so in the package notes.

Why it matters

• Microcosm currently scales every area cell by one GB factor (0.921), which assumes the same nil-award share everywhere.
• The child bands keep all-claims shares, so after scaling they sit +2.8% (no children), -1.5% (one), -3.5% (two) and -5.1% (three or more) against the national paid rows.

After it lands

Microcosm re-pins the feed, selects payment_indicator: Yes on the five area targets, and the bridge to the GB caseload reconciles one concept. PolicyEngine/microcosm#1167 pins 1ee7dfe, an ancestor of 942a1fa, so the new pin has to carry both.