Skip to content

No calibration targets for UC elements: childcare runs 2.3x official, housing/LCWRA/carer unbenchmarked #466

Description

@vahid-ahmadi

Most of UC entitlement is calibrated against nothing. sources/dwp.py targets caseload, family type, payment bands, deductions and the two-child limit — but no element counts. Only the childcare element has a published comparator, and the model is 2.3x it.

Model vs official, enhanced FRS 2023-24 v1.56.16, 2025:

Element Model recipients Official Err
Childcare 370k 160k +131%
Child 3.077m (48.1% of hh) 47% of hh ok
Housing 5.193m no target
LCWRA / health 2.041m no target
Carer 0.560m no target

Childcare official: UC statistics to 9 Oct 2025 (160k Aug 2025, 170k Aug 2024). The rest are Stat-Xplore UC_Households only — not in any bulletin.

Related: family-type targets sum to 6.096m against a dwp/uc/households target of 6.7m (the vintage tension noted in #458), and the 3+ children rows run 16-22% low:

Family type Model DWP Err
Single, no children 3.210m 2.868m +11.9%
Couple, no children 0.280m 0.231m +21.1%
3 children 0.369m 0.473m -22.0%
4 children 0.136m 0.167m -18.3%
5+ children 0.064m 0.076m -16.2%

Next steps

  1. Register a Stat-Xplore API key and add element-count targets for housing, LCWRA, carer, childcare from UC_Households.
  2. Add the childcare 160k target now — it is published and needs no key.
  3. Refresh the family-type values to a current Stat-Xplore month so they sum consistently with dwp/uc/households (the Fix Universal Credit caseload targets #458 follow-up).
  4. Once these exist, microcosm#729's national_calibration.py picks them up through the reference registry; no microcosm change needed until then.

cc @MaxGhenis @vahid-ahmadi

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions