Repository navigation
UK Carer's Allowance: would_claim follows reported receipt, so earnings-limit reforms reach 4,600 new carers against an official 61,000 #1319
Description
Summary
On the UK national file, raising the Carer's Allowance (CA) earnings limit adds almost no new claimants. Moving the limit from £151 to £196 a week in 2025 adds 4,600 entitled carers; the independent review of CA overpayments puts the same change at an estimated 61,000. The likely cause: the spine sets would_claim_carers_allowance only for people who report CA receipt in the FRS, so a carer above today's limit (who cannot report receipt) never claims under a reform.
File: microcosm_uk_2024_25.h5, HF policyengine/populace-uk-private, release microcosm-uk-2024-25-national-20261019T083323Z-6cba5d71 (SHA-256 810ba980…), run under policyengine-uk 2.130.1. Found during an internal Budget-day rehearsal on 19 October 2026.
Measurements
Weighted with person_weight, static, policyengine-uk 2.130.1.
| Test | File result | Official comparator |
|---|---|---|
| Limit £151 → £196 a week, 2025 (the April 2025 change) | +4,600 entitled carers, +£20m a year | "bringing an estimated 61,000 more people into eligibility" (Independent Review of Carer's Allowance Overpayments, July 2025, ch. 7) |
| Limit £204 → £230 a week, 2026 (rehearsal) | +2,600 recipients, +£11.6m a year | none published |
The file does carry carers in the band the change reaches. In 2025, 70,000 people outside Scotland report 35 or more hours of care a week and have CA earnings (carers_allowance_earnings / 52) between £151 and £196. Of these, 1,200 have would_claim_carers_allowance set. Across the whole file, would_claim_carers_allowance & ~(carers_allowance_reported > 0) has zero weight.
Mechanism
packages/microcosm-build/src/microcosm/build/uk_runtime/frs_spine.py:577:
pe_person["would_claim_carers_allowance"] = (
pe_person["carers_allowance_reported"] > 0
)The comment says this keeps the UC carer element on care_hours without paying CA to every carer (#882). policyengine-uk's carers_allowance_pre_overlap requires would_claim alongside the hours or receipt condition and the earnings limit. A carer whose earnings sit above the current limit reports no CA, so would_claim is False and no earnings-limit reform can pay them. The same gate holds for the 35-hour condition and the CA rate itself: any reform that widens CA entitlement scores only people already in receipt.
Related levels, for context
| Measure | File, 2025 | DWP |
|---|---|---|
| CA entitled | 1.087m (UK, carers_allowance_pre_overlap > 0) |
1,376,400 at August 2025, excluding Scotland (DWP benefit statistics: February 2026) |
| Entitled but not paid (overlap) | ~5,000 | 380,000 (27%) |
| Recipients with earnings | 139,000 (12.8%) | 235,000 employed or self-employed in 2023-24, 15% of claimants in payment (CA claimants analysis, November 2025) |
Geography differs: DWP's CA figures exclude Scotland (Carer Support Payment). The paid level (1.08m in the file against roughly 996,000 in payment, 1,376,400 less 380,000) is close; the entitlement-only group and earners are short. Those two gaps are to investigate separately.
To investigate
- Draw
would_claim_carers_allowancefor carers who meet the hours condition but report no CA, at a take-up rate tied to a published figure (for example, the 61,000 newly eligible from April 2025 against DWP's caseload growth since), so earnings-limit and rate reforms reach newly eligible carers. The UC carer element should keep readingcare_hoursas Align UK UC claim, child and period definitions #882 intended. - Whether the 380,000 entitlement-only claimants (mostly State Pension recipients whose pension overlaps CA) are missing because
would_claimfollows reported receipt, which the FRS records only when CA is paid. - Re-run the £151 → £196 back-test above after the change and compare with the 61,000.
Activity
Say that policyengine-uk lacks Carer Support Payment's earnings limit (…
Merge pull request #1329 from PolicyEngine/uk-prebudget-1278
juaristi22 commentedon Oct 21, 2026
#1329 (merged 21 October,
211483f3a) lands item 1 of the list above. This comment records what its review settled for this issue, so it outlives that thread.Landed in #1329
carers_allowance_take_upspine stage, afterchild_benefit_take_up, setswould_claim = reported OR (eligible AND u < 0.6).care_hoursreach the engine'sgov.dwp.carers_allowance.min_hours, who reports no allowance, and whomis_gainfully_employed_for_carers_allowanceholds gainfully employed at the survey year (the 2024-25 limit).person_id.frozen_by_adjudication, since no published take-up rate exists for newly eligible carers.Before anyone costs a Carer's Allowance earnings-limit change on this build (Vahid's review of #1329, findings 2 and 3)
The back-test on the built national file (item 3 above), on the 22 October build. It reports two weighted counts, both against the Review's 61,000:
Support is banded on distinct source households, with no figure where either side rests on one to nine.
The salary-sacrifice count. The stage reads earnings before
salary_sacrifice. So a drawn carer whom the conversion takes under the survey year's limit keeps the flag, and is paid in the baseline. The same back-test counts them.Until both are read, nobody should cost a Carer's Allowance earnings-limit change on this build.
Why 0.6 is a judgement call. The pool is wider than the population behind the Review's 61,000:
So the receipt's
realized_shareis the drawn share of that pool, not a take-up rate among the newly entitled.After the Budget
realized_share(for example todrawn_share_of_over_limit_pool) and add the back-test fields, in one receipt change.salary_sacrifice, with a draw keyed on lineage so the CGT clones keep their source's draw. Draw UK QRF stages from uniforms keyed on each record's id, and tie pension contributions to pay #1332 ties pension contributions to pay and touchesfrs_spine.py, so the set of salary-sacrifice cases may move when it lands.🤖 Generated with Claude Code
Merge main into us-calibration-epoch-thinning