Health Insurance by Citizenship Status (ACS)

Limitaciones

citizenship_status is a hierarchy, not a partition: all = native + foreign_born and foreign_born = naturalized + noncitizen, so never sum across it. Private and public coverage overlap (a person can hold both), so never add num_private_coverage and num_public_coverage or derive a 'private only' count; only insured + uninsured partition a group. Universe: the civilian noninstitutionalized population, excluding active-duty military and institutional group quarters. The private + public sum double-counts people holding both (e.g. employer insurance plus Medicare) and so is at least, and almost always more than, num_insured. The 'all' universe is smaller than the total population (10,746,448 vs 10,940,407 in population_by_race for 2024) because prisons, nursing homes, and military personnel are outside it; the gap is material in counties with large institutional or military populations. Coverage is measured at the time of the interview, not over the year; people whose only coverage is the Indian Health Service count as uninsured. The source publishes coverage cells only for native, naturalized, and noncitizen residents: the 'all' and 'foreign_born' rows' uninsured, insured, private, and public counts are derived as exact sums of those cells, with margins from the Census sqrt-sum-of-squares approximation under the zero-estimate rule (non-integer, served as floats), while num_insurance_universe on every row is the published group total with its published margin. uninsured_rate_moe is derived from the count margins with the Census ratio formula (addition-form fallback whenever the subtraction form's radicand is not positive, so a served margin is never zero; a 0 or 100 percent rate takes the Census's own model margin for such estimates); a degenerate small-sample margin above 1.0 is set to NULL. Published zeros are real estimates, not suppression: small foreign-born cells in rural counties and many tracts are 0 with a positive margin, and a NULL uninsured_rate means the group has no civilian noninstitutionalized members in the geography (a zero universe - a true absence). The ACS asks citizenship, not legal status: noncitizen covers lawful permanent residents, temporary visa holders, humanitarian migrants, and unauthorized migrants alike. No annotation sentinel appears in vintages 2020-2024, so no count is masked; a NULL count margin beside a present estimate would be a controlled estimate, never suppression. ACS 5-year period estimates: the year column is the vintage END year and each estimate pools five years of survey responses (2024 = 2020-2024). Consecutive vintages share about 80% of their sample, so the series is heavily smoothed - never difference overlapping vintages as if they were annual change. Estimates are survey-based and carry 90% margins of error (the *_moe columns); tract-level margins can be large relative to the estimate. All levels serve vintages 2020-2024 only, on 2020-vintage tract boundaries. State rows have NULL county_fips, district_code, and tract_geoid; county rows carry only county_fips, district rows only district_code, tract rows only tract_geoid. District rows cover the ~180 standard (city/county) districts only: the Census publishes school-district estimates at unified school district geography, which has no codes for charter schools, RESAs, or state agencies, and its few Census-only districts (e.g. the Fort Stewart military district) have no GCD district code and are dropped.