US certified default undercounts households (122.5M) and overcounts married couples (81.5M): no household-type or marital-status targets in the registry

Summary

The certified US default (populace-us-2024-buildp-sparse-rmloss100-cae8640-20260728, HF blob 48b9d479…) matches person-level population targets but carries too few households and too many married couples. The target registry (5,659 targets, all household entity) has no household-count, family-structure, or marital-status targets, so calibration meets person-by-age targets with fewer, larger, married households.

Found while counting married one-earner families with a child under 13 for the HHS "parent-based child care" draft rule (Sept 2026); the Microcosm count came out 37% above the raw ASEC on the same definition.

Benchmark cells (same code over each file; marital unit = 2 adults; SPM-unit children)

Persons per household: Microcosm 2.78 vs ASEC 2.41.

Weight behaviour is otherwise much better than the old enhanced CPS (top-1% household weight share 11% vs 56%; realized max weight ratio 5.0), so this is a target-surface gap, not a weight-concentration problem.

Proposed targets (Chronicle-named, household entity)

• Census CPS/ASEC households by type: total households; married-couple family households; other family; nonfamily (Table HH-1 / F-1 series), national and by state where published (ACS B11001).
• BLS "Employment Characteristics of Families" Table 4: married-couple families with own children <18 (22.7M, 2025 release), by parents' employment (both employed 15.0M; one parent employed 7.1M).
• Census/ACS children by presence of parents (living with two married parents), so child-age targets are not met by concentrating children in married households.

Repro

Scripts and outputs: ~/PolicyEngine/_drafts/vance-ccdf-parent-based-care/benchmark_cells.md and newly_eligible_count_microcosm.md (numpy over the h5 entity tables, no engine run). BLS Table 4: https://www.bls.gov/news.release/famee.t04.htm. Census household type: https://www.census.gov/topics/families/families-and-households.html.

Start by reading ~/PolicyEngine/_drafts/vance-ccdf-parent-based-care/benchmark_cells.md and newly_eligible_count_microcosm.md, then inspect how the target registry represents household entities and calibration targets. Use the cited Census and BLS sources to define the proposed household-type, marital-status, and parent-employment targets. Done means the registry includes the selected targets and the benchmark cells no longer show the reported household and married-couple distortions.