Skip to content

UK spine: relationship-to-head frame column, so the ONS household-composition split is measurable #791

Description

@juaristi22

Successor to the #757 `uk_target_fit` dispositions (issue comment 5427936411, adjudicated 2026-08-26).

Three ONS household-composition targets are unmeasurable on the current spine because the frame carries no relationship-to-head column: the adapter's household-type derivation cannot distinguish a second family unit from unrelated adults or non-dependent children of the head, so the categories bleed into each other. Measured on the rebind packet (403 targets, loss 0.1362):

target rel. error note
`ons.household_composition.multi_family_households` +1618% absorbs misclassified households (244k target, 4.19m estimate)
`ons.household_composition.unrelated_adult_households` +48.1% same bleed
`ons.household_composition.lone_parent_non_dependent_children_households` −0.03% fits numerically but measures the wrong concept — the split is undefined without relationship-to-head

All three carry windowed exclusions on the calibration measure-exclusion register (approved 2026-08-26, expires 2026-11-26) citing this issue as tracking. Closing this issue means: derive or impute a relationship-to-head column on the spine (the FRS household grid carries the raw relationship data), re-derive the composition categories per the ONS definitions, and retire the three exclusions.

Parent: #757 (WS-E of #145, epic #665).

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions