NI local UC targets count claimants but are used as households, and Ards and North Down has no LA target #590

Found by the independent review of #589 (#468's fix). The UK hub verified it on main at daad71c. Neither problem is caused by #589.

1. NI local UC targets count claimants, but are used as households

utils/uc_data.py (constituencies) and L211-L236 (LAs) read the NI rows from dfc-ni-uc-stats-supp-tables-may-2025.ods, sheets 5b and 5c:

• 5b is "Table 5b - Claimants by Parliamentary Constituencies, October 2017 to May 2025";
• 5c is "Table 5c - Claimants by Local Government Districts, October 2017 to May 2025".

The GB rows come from Stat-Xplore "Households on Universal Credit". All of them go into the same household_count column, which the loss builders fit against modelled UC households (uc_households, plus the four uc_hh_*_children columns).

A joint claim has two claimants and one household, so the NI targets are on a different basis from GB's and overstate NI UC households. Then the combined GB + NI rows are rescaled to one national total, which shifts weight between NI and GB.

2. Ards and North Down has no local-authority target

local_authorities_2021.csv lists 10 NI districts (N09000001 to N09000010), so Ards and North Down (N09000011) is missing. uc_data.py:226-227 skips it "to match target datasets (they have 10 NI LGDs, not 11)". The review puts that district at 15,320 claimants in May 2025.

NI has had 11 local government districts since 2015, so one NI district's households get no LA-level UC (or age or income) target.

Note: the "No payment" exclusion

The rescale total (_parse_uc_national_payment_dist) leaves out Stat-Xplore's "No payment" households (477,409, per the review). That is consistent with the model indicator the builders use, universal_credit > 0, so it is recorded here as a definition and not as a bug. If the target is ever meant to mean all UC households, both sides need to change together.

Fix

• Convert the NI claimant counts to households: use a DfC household table if one exists, otherwise apply a documented claimants-per-household factor with its source. Otherwise, keep NI out of the household-count targets and say so.
• Add N09000011 Ards and North Down to the LA code list and its targets (age, income, UC). Check every source that was trimmed to 10 NI districts.
• Add tests that each NI source's unit matches its target's unit, and that the LA code list has all 11 NI districts.