Exempt non-dependants responsible for children under five from UC housing contributions

[juaristi22]

Fixes #2071.

Apply the Universal Credit housing-cost contribution exemption for a non-dependant responsible for a child under five, under Schedule 4 paragraph 16(2)(i). Use the existing own-family child-responsibility and claimant/partner roles, with an age-limit parameter dated to the GB regulations’ 29 April 2013 commencement. The corresponding Northern Ireland rule appears in Schedule 4 paragraph 15(2)(i).

Reproducer: policyengine_uk/tests/policy/integration/uc_child_responsibility_housing_contribution.yaml::UC housing contribution - non-claiming parent of a four-year-old is exempt

axiom: needed

Local validation: 67 focused checks passed, including 12 new cases and 55 existing cases. Before the change, six new cases failed and four controls passed. Independent policy/source and code reviews completed with no findings; the code reviewer also passed six held-out checks. Hosted CI will be checked after publication.

The model’s existing annual child-age convention and shared historical UC availability remain limitations. The GB parameter date does not establish Northern Ireland’s UC commencement or certify pre-rollout NI awards.

[vahid-ahmadi]

Looks right, fine to approve.

@juaristi22, the new limb matches UC Regs 2013 Sch 4 para 16(2)(i) and reg 4. I found nothing that blocks merge. Three small points.

1. nit: policyengine_uk/variables/gov/dwp/universal_credit/housing_costs_element/non_dep_deduction/uc_non_dep_deduction_exempt.py:30. Responsibility comes from the child being in the non-dependant's own benefit unit. If a dataset puts a grandchild in the renter's benefit unit instead, the parent gets no exemption, and the renter would be treated as responsible. That follows reg 4(2) and (4) on the model's reading of "normally lives with", and the FRS usually puts a lone parent and child in their own unit. A sentence in the docstring saying so would help later readers.
2. nit: child_responsibility_age_limit.yaml:3. The value is dated to the GB start, 29 April 2013, but the rule is applied UK-wide. The NI rule (SR 2016/216 Sch 4 para 15(2)(i)) is the same, and NI UC started later, so this changes no results. The PR body already says so. You could add the NI regulation as a second reference.
3. nit: draft PR #2017 also edits the return expression in uc_non_dep_deduction_exempt.py. The hunks are next to each other, so whichever merges second will need a small rebase.

What I checked

• Read issue #2071, the PR body and the full diff.
• Checked Sch 4 para 16(2)(i) ("a person who is responsible for a child under 5 years old") and reg 4(2), (4) and (6)(a) on legislation.gov.uk. Both members of a couple count as responsible, and a looked-after child confers no responsibility. uc_youngest_child_age already excludes looked-after children and returns infinity when there is no child, so the strict < 5 test is right. Age 5 is not exempt and age 0 is.
• Hand-computed case 1: 12 x £424.90 + £6,000 rent = £11,098.80 UC, no contribution. Hand-computed the unrelated-family case: one contribution of 12 x £96.55 = £1,158.60, so housing element £4,841.40 and UC £9,940.20. Both match the YAML.
• Ran policyengine-core test on the new file locally: 12 passed. Ran the existing UC housing costs element tests: 20 passed.
• Reverted the formula to the base commit: 7 of the 12 new cases fail and 5 controls pass.
• ruff check and ruff format --check on the changed Python file: clean.
• Changelog fragment is fixed. The Reproducer: and axiom: lines are present, and the Fix evidence check passed.
• CI on the current head: all checks passed. GitHub shows the branch as mergeable with main.
• Overlap: no clash with #1820, #1855, #1864, #2184, #2187, #2189, #2193, #2382, #2386 or #2433. #2184 touches HB and CTR non-dependant deductions, not UC.
• Not checked: microsimulation impact on enhanced FRS data, as I don't have the private data.

🤖 Generated with Claude Code

[vahid-ahmadi]

Approving, per my review above. The shoulds and nits can follow.