US county population-component intelligence (annual)

Census Bureau Vintage 2025 county population estimates (keyless official CSV, U.S. public domain): the wide county file reshaped to a county x estimate-year panel (3,144 county-equivalents x 2020-2025) with births, deaths, natural change, domestic and international migration, net migration and residual components, all identity-gated exactly (natural == births - deaths, net == domestic + international, residual == total - natural - net, county sums == state rows) and per-1,000 midyear-population rates cross-checked against the file's own R columns. A documented 0-100 growth-momentum score (50% total rate + 30% net-migration rate + 20% natural rate, winsorized, min-maxed within each estimate year) is ranked nationally (growth_rank) and within each state, tiered g1-g4, with a closed migration-driver taxonomy (migration-led/natural-increase-led/mixed growth and decline, stable) and top-100, bottom-100, migration-reversal, record-high/low-population and growth-streak flags. Primary key: (county_fips, estimate_year). Cadence: yearly; the upstream file is a static annual vintage, so as_of is the vintage year and identical input produces an identical content hash. Caveats: estimates are modeled and revised each vintage (single-vintage consistent series); Puerto Rico municipios are absent upstream; rates use midyear population. Sample use: order by growth_rank within an estimate year for the fastest-growing counties, or filter state_code for within-state migration geography.