Skip to content

Bind England's PIP cases at 16–19 and 20–24 in place of 16–24 - #1671

Closed
juaristi22 wants to merge 1 commit into
mainfrom
uk-milburn-pip-split
Closed

juaristi22 wants to merge 1 commit into
mainfrom
uk-milburn-pip-split

Conversation

@juaristi22

Copy link
Copy Markdown
Collaborator

Why

The benchmarks check of the Milburn review dashboard (PolicyEngine/uk-milburn-review-dashboard#1) found that on build fcfac709, England's PIP band at 16–24 is calibrated and fits. Within it, though, 16–19 ran 17.6% high and 20–24 ran 15.9% low. The review's measures cut PIP at 18.

This was the first half of #1663.

Changes

uk/calibration_measure_exclusions.json:

  • dwp.pip.cases_england_aged_16_to_19 and dwp.pip.cases_england_aged_20_to_24 now bind.
  • dwp.pip.cases_england_aged_16_to_24 is held out as their sum, as dwp.uc.people_aged_16_to_24 is for its two bound parts.
  • England's eleven other published bands stay held out as checks. Their restated exclusion carries her call, dated 2026-11-17 and expiring with the family on 2027-02-15.

The register's count pin in test_uk_measure_simulation.py moves from 143 to 142, and a changelog fragment records the change.

The commit is #1663's first, b1fe150b1, cherry-picked onto main at 549938277.

Testing

Disclosure

Every figure here is national and rests on 30 or more survey households.

🤖 Generated with Claude Code

María's call of 17 November 2026 ("do the pip age-split and lcwra in a
single pr"), after the Milburn review's benchmarks check
(PolicyEngine/uk-milburn-review-dashboard#1). The review's measures cut
PIP at 18, inside the 16 to 24 band, and on build fcfac709 the split
within it ran 17.6% high at 16 to 19 and 15.9% low at 20 to 24 while the
band itself fitted.

The two published bands bind; the 16 to 24 band is held out as their sum,
as dwp.uc.people_aged_16_to_24 is for its two bound parts. England's
eleven other published bands stay held out as checks on the split within
their broad bands; their restated exclusion carries her call, dated
2026-11-17 and expiring with its family on 2027-02-15.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit b1fe150)
juaristi22 added a commit that referenced this pull request Nov 18, 2026
… a count

Part 2 of the UC health fix, on María's go of 18 November 2026. The LCWRA
fit's one odds ratio holds the element's households at every age, so nothing
holds young people with it, and binding England's PIP at 20 to 24 (#1671)
pushed them up through the units the proxy flags.

Given DWP's count of young people with the element, the drawn units in which
a claimant or partner aged 16 to 24 would hold it claim at a third odds ratio
in place of the LCWRA fit's, solved in turn with the age fit's and the LCWRA
fit's ratios. A young unit whose element an older partner would hold keeps
the LCWRA fit's ratio.

The stage passes no such count until Chronicle carries the fact
(PolicyEngine/chronicle#442), so the build is unchanged until then: without
the count the redraw takes the two fits' path as before. The function's tests
exercise the three-ratio solve, its refusals and the stage's count of young
people holding the flag.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@juaristi22

Copy link
Copy Markdown
Collaborator Author

Closing: #1663, merged in 6d32375b3, ships this change unchanged.

The branch stays: its commit 48bfbc5be is the source of the listed build uk-frs-calibration-attempt-20261117T212629Z-55dae354.

Drafted with AI assistance.

@juaristi22 juaristi22 closed this Nov 19, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant