A 0.3% stratified random sample of individual-level American Community Survey records for 2008–2021 (about 69,800 people ages 18–60), carrying the same Medicaid-expansion adoption years as the aggregated medicaid panel. Because it holds individual survey weights and real covariates, it is the dataset for demonstrating repeated cross-section analysis, survey weighting, and propensity-score weighting with the stacked estimator — none of which the collapsed state panel can support.
Codebook
Variable
Type
Description
state
character
Two-letter state abbreviation
year
integer
Calendar year (2008–2021)
perwt
numeric
ACS person weight (survey sampling weight)
statefip
integer
State FIPS code
adopt_year
integer
Year the state adopted Medicaid expansion (NA if never adopted)
uninsured
integer
Health insurance status: 1 = uninsured, 0 = insured
female
integer
Sex: 1 = female, 0 = male
age
integer
Age in years (18–60)
The sample preserves the state-by-year distribution of the full ACS microdata while staying small enough to ship with the package.
Source and citation
American Community Survey via IPUMS USA (https://usa.ipums.org/); sample drawn from the full ACS microdata for adults ages 18–60.
Steven Ruggles, Sarah Flood, Matthew Sobek, Daniel Backman, Annie Chen, Grace Cooper, Stephanie Richards, Renae Rodgers, and Megan Schouweiler. IPUMS USA: Version 15.0. Minneapolis, MN: IPUMS, 2024. doi:10.18128/D010.V15.0
Method reference: Wing, C., Freedman, S., & Hollingsworth, A. (2024). Stacked Difference-in-Differences. Working paper.
state year perwt statefip adopt_year uninsured female age
1: AL 2008 116 1 NA 0 1 19
2: AL 2008 121 1 NA 1 0 46
3: AL 2008 93 1 NA 1 0 49
4: AL 2008 135 1 NA 0 1 20
5: AL 2008 94 1 NA 0 0 23
6: AL 2008 242 1 NA 0 1 51