UK spine: the early flat UC draw moves working-age Housing Benefit reporters onto UC

Follow-up to item 7 of Vahid's second review of #1344.

Problem

frs_take_up draws would_claim_uc at one contract rate over the UC age population. spi_benefit_coherence redraws it the same way for SPI units. Neither draw looks at the legacy benefits a unit reports.
policyengine-uk continues an existing Housing Benefit award only while the unit does not claim Universal Credit. So a working-age unit that reports HB and is drawn to claim UC loses its HB, even when the engine does not move it at the legacy closure.
On the 2024-25 spines, the working-age units that keep HB are already below DWP's 2025 caseload before uc_take_up runs. UK #1334: draw UC take-up by award band, and measure the HMRC income rows over each person's year #1344 keeps these units out of its own draw and counts them in F at their stored flag, so it does not add to the gap. It does not close the gap either. The gap is older than UK national calibration moves weight away from low-income households: reconcile the HMRC low-band rows, model UC take-up by award size, impute pensioner benefit receipt (pre-Budget) #1334.

Working-age Housing Benefit claims in Great Britain (DWP benefit statistics, August 2025 and February 2026 releases):

The caseload falls fast through 2025 as DWP moves HB-only cases to UC. The build's 2025 basis is a period average, so these two points bracket it.

The model figure to compare is every working-age benefit unit with engine housing_benefit above zero in 2025, over Great Britain, at calibration weights.
The working_age_housing_benefit block in uc_take_up's receipt is not that figure. Its reporting_units leaves out the units the engine moves at the legacy closure, and it is at the stage's weights.

In frs_take_up and spi_benefit_coherence, leave would_claim_uc false for a unit that reports a legacy benefit the engine does not move, unless the unit also reports UC. This is the rule uc_take_up now applies. The legacy closure stage then decides who moves.
Draw those units onto UC at a published natural-migration rate (moves after a change of circumstances), not at the flat rate.

The model's working-age HB caseload in 2025 is compared with DWP's figures on a like-for-like basis, as a receipt line.
The receipt line shows weighted figures only where ten or more source households carry them.
The gap is closed, or its cause is written down.

🤖 Generated with Claude Code