Programmes read the raw household head flag, so a malformed input makes them disagree about whose household it is #2059

is_household_head is an input, so it can flag several members of a household, or none. Programmes that read it directly then disagree about whose household it is.

How the input goes wrong

• A user situation flags two members, for example both members of a couple, or the householder and a lodger.
• A user situation flags no one in a household. This also happens without anyone writing false: once any person in a simulation has the input, every person without it defaults to false, not to the formula.

The Family Resources Survey data is not affected. policyengine-uk-data sets the flag from the survey's household reference person (hrpid == 1).

What happens today

Council Tax Reduction settles on one head (#2009: the eldest flagged member, or the eldest member where none is flagged). These variables read the raw flag instead:

• rent: is_liable_for_household_rent, share_of_household_rent, personal_rent, pays_rent_to_householder;
• non-dependants and size criteria: is_non_dependant_of_household_head, uc_non_dep_deductions, LHA_allowed_bedrooms, housing_benefit_LHA_allowed_bedrooms, lha_renter_has_non_dependant;
• the maintenance loan proxies: maintenance_loan_living_arrangement, maintenance_loan_household_income.

Household calculations on #2009's head (4b84ea069), for a home with a £12,000 rent:

The first row is finding 8 of the review of #2009.

Proposed fix

• Add one person-level household head, exactly one per household, with Give Council Tax Reduction to the household head's family, not the oldest adult's #2009's rule: the eldest flagged member; where none is flagged, the eldest member.
• Define benunit_contains_household_head from it, and point every reader of the raw flag at one of the two.
• Test several-flagged and none-flagged households across rent, non-dependants and Council Tax Reduction, with a property test that every programme agrees on the head's family for any flag input.