Two-child limit targets declared at 2026 are structurally unreachable (limit abolished April 2026)

[vahid-ahmadi]

The nine two-child limit targets are declared at 2026 only, but the limit is abolished from 2026-04-06, so uc_is_child_limit_affected is structurally zero that year. They can never bind.

policyengine_uk_data/targets/sources/dwp.py:203-305:

Target(
    name="dwp/uc/two_child_limit/households_affected",
    variable="uc_is_child_limit_affected",
    values={2026: 453_600},   # <- never reachable
)

policyengine-uk parameters/gov/dwp/universal_credit/elements/child/limit/child_count.yaml:

2017-04-06: 2
2026-04-06: .inf   # Budget 2025 abolition

Model output confirms it (enhanced FRS 2023-24, v1.56.16): every TCL target computes to 0.0 in 2026.

These are April 2025 statistics. Moved to 2025, the model overshoots:

Target	Model 2025	Official (Apr 2025)	Err
Households affected	608k	454k	+34%
Children affected	825k	580k	+42%
Children in affected households	2,135k	1,614k	+32%

Next steps

1. Move all nine targets from 2026 to 2025.
2. Add a registry test that fails when a target is declared in a year where its variable is structurally zero — this class of dead target should not be silently possible.
3. The +34/+42% overshoot is then a live calibration signal, likely tied to the 3+ children undercount in No calibration targets for UC elements: childcare runs 2.3x official, housing/LCWRA/carer unbenchmarked #466.

cc @MaxGhenis @vahid-ahmadi

[vahid-ahmadi]

Correction to my earlier note: this is not blocked by or blocking PolicyEngine/microcosm#738. Microcosm binds the April-2025 publication facts at period 2025 (all 15 two-child-limit references, post-#735), where the limit is still in force — the 2026 abolition never touches them. The dead pinning is only here, in sources/dwp.py:209-303, still at {2026} on main. Standalone fix.