childcare_expenses has no calibration target and sits at the top of the plausible range

Summary

childcare_expenses is an uprated FRS input with no calibration target. On the only slice for which a published benchmark exists — England, under-5s — the Enhanced FRS runs about 1.7× above it. The variable drives Tax-Free Childcare, the UC childcare element, the WTC childcare element and housing benefit, so any cost-share childcare reform scales linearly with it.

Comparing like with like

This is the part worth care, and I got it wrong on the first pass. The available benchmark covers England and the under-5s only. The model's headline aggregate is UK and all ages, and about a third of it is school-age wraparound and holiday childcare — a separate market the benchmark excludes.

enhanced_frs_2024_25 at 2027, weighted:

Against a benchmark of ~£3.75bn that is 1.69×, not the 3.05× a naive UK/all-ages comparison suggests.

Where the benchmark comes from

CMA, Early Years Education and Childcare Services Market Study — Statement of Scope (1 July 2026) estimates England's early years sector income at ~£14bn in 2025/26, derived as £8.9bn of funded entitlements ÷ 63% (the estimated share of provider income from entitlements). Netting off the ~£5bn residual's TFC (£0.6bn) and UC childcare element (£0.6bn) leaves roughly £3.5–4bn of out-of-pocket parent spend.

Two caveats, stated plainly:

• The £3.5–4bn residual is arithmetic on the CMA's figure, not a published number. No official aggregate of parent-paid childcare fees exists for England or the UK.
• The CMA flags substantial uncertainty in the £14bn, and its direction of error widens this gap: the 63% is a self-reported unweighted survey average that the CMA thinks is probably too low, and a higher true share pulls sector income and the residual down.

Corroboration from a different variable

The gap is independently visible in Tax-Free Childcare. The model's average TFC award is £1,353 against HMRC's ~£998 — 1.36× — on a caseload that is separately 1.52× too high (filed as #470). Since the award is a function of the spend it subsidises, that 1.36× is this issue seen through a different variable. Two independent routes to the same overstatement.

Suggested fix

Add a calibration target for the England under-5 slice.

Do not rebase the whole variable on it. No published aggregate exists for school-age childcare spend, so that third of the base cannot be calibrated in either direction and should be left as-is — rebasing everything on an under-5 England benchmark is the same mistake in reverse, and it cost me about £1.7bn on a reform costing before I caught it.

No files or tests are named. Start by searching the repository for childcare_expenses and existing calibration targets, then trace how slices are configured and validated. Done means adding an England under-5 calibration target without rebasing the UK all-ages variable or altering the uncalibrated school-age component.