Assign official Census metro areas to congressional districts for GEOADJ

[MaxGhenis]

Summary

Currently, policyengine-us-data calculates GEOADJ for congressional districts using ACS median rent data and the GEOADJ formula. This produces unofficial estimates that don't match official Census Bureau SPM thresholds.

Proposed Solution: Census Block Assignment

Recommended approach: For each congressional district, assign a population-weighted random census block, then look up everything from that block.

Why census blocks?

1. One crosswalk, infinite lookups - Build CD → block once, then derive metro, county, tract, PUMA, etc. from block. No need for separate CD→metro, CD→county crosswalks.

2. Matches CPS methodology - Census assigns people to geographies, then looks up GEOADJ. This replicates that for simulated CD populations.

3. Captures within-CD heterogeneity - A CD spanning both SF metro (GEOADJ 1.95) and California Nonmetro (0.96) would naturally get a population-weighted mix.

4. Reusable infrastructure - Same block assignment could feed other geography-dependent calculations beyond SPM.

5. More accurate for edge cases - CDs that span metro boundaries, rural/urban splits, etc.

Implementation

1. Get census block population data (Decennial Census PL 94-171)
2. Build CD → census block crosswalk with population weights
3. For each simulation, randomly sample a block within CD (population-weighted)
4. Look up metro area (or state metro/nonmetro) for that block
5. Use official GEOADJ from Census SPM threshold data

Data Sources

• Census block populations: Decennial Census Redistricting Data
• Block to CBSA crosswalk: Census Geographic Relationship Files
• Official SPM thresholds by metro: https://www2.census.gov/programs-surveys/demo/tables/p60/287/SPM-pov-threshold-2024.xlsx

Context

• The spm-calculator web app now only uses official Census metro areas
• The unofficial CD GEOADJ estimates (computed from ACS rents with arbitrary 0.7-1.5 clamp) have been removed
• This change would allow proper CD-level thresholds using authoritative data

Related

• spm-calculator repo: https://github.com/PolicyEngine/spm-calculator
• Census SPM methodology: https://www.census.gov/topics/income-poverty/supplemental-poverty-measure.html