Reweighting acs microdata in stata

[annette jacoby]

Hello all,

We are using acs pums microdata to estimate the number of potential Temporary Protected Status (TPS) holders and their tax contributions. Since ACS does not directly identify TPS holders, we rely on reweighting methods using CRS (Congressional Research Service) reports, which provide TPS data by state OR by birthplace, but not both. However, the problem is that

• If we reweight ACS by birthplace, we match CRS birthplace totals, but our state-level distributions are off.
• If we reweight by state, we match CRS state totals, but birthplace distributions are distorted.
• We are looking for a method to balance accuracy at both levels in Stata...

Does anyone know of a reweighting approach or some kind of constrained optimization or post-stratification method to force totals to match both levels as closely as possible?

[Leonardo Guizzetti]

This is outside my area of expertise but I wonder if what you describe exactly what should happen based on the different weighting schemes? If you weight by place of birth, state totals would be different because of migration, and vice versa. Couldn’t you report the analysis both ways?