Anyone looking at a political dashboard should be able to ask “where does this number come from?” and get a real answer. This page is the real answer. Everything you see in the map and the county spotlights is documented below — every source, every variable, every join. The county FIPS code is the join key throughout.
The choropleth shows two-party Dem margin per county per presidential year. Margin is computed as (Dem votes − GOP votes) ÷ total votes, expressed as a signed fraction. Negative = the county went Republican; positive = Democratic. The diverging Red-Blue palette is a 9-stop scale from −30 pp (darkest red) to +30 pp (darkest blue).
Boundaries come from the U.S. Census Bureau's TIGER/Line 2020 county shapefile, served as a TopoJSON via us-atlas.
Two-party Democratic share, turnout-weighted. For each year, we sum the implied Dem votes across counties (county Dem share × county total votes) and divide by total votes summed. Metro = USDA Rural-Urban Continuum Codes 1–3; nonmetro = 4–9.
Turnout-weighted means a county with 1,000,000 voters counts a thousand times more than a county with 1,000 voters. This is the right way to summarize a population — the alternative (county average) treats Cook County, Illinois the same as Loving County, Texas.
“Rural” is a fuzzy word that the federal government defines two ways, and they don't agree:
They tell very different stories. The Greater LA MSA covers 4,857 sq mi (both LA and Orange counties); the LA Urban Area covers 1,637 sq mi — about a third — because the northern half of LA County is sparsely-populated desert. Same name, totally different geography.
We use the USDA Rural-Urban Continuum Code (RUCC), which sits on top of MSAs (county-level). RUCC 1–3 are metro counties graded by metro-area population; RUCC 4–9 are nonmetro counties graded by their largest town and adjacency to a metro. This is the right call for political analysis because election data is reported at the county level, county boundaries don't shift between cycles, and it joins cleanly to ACS, USDA typology, and Religion Census data — all county-keyed.
The trade-off: a few RUCC-1 metro counties contain large sparsely-populated areas (LA County desert, the parts of King County, WA outside Seattle), and a few RUCC-4+ nonmetro counties contain small but real urban clusters too small to anchor an MSA. The aggregate effect is small, but worth knowing when you're looking at a single outlier county.
Background: Jeff Davis, “The 2020 Census and ‘Urban Areas’” (Eno Center for Transportation, Sept 2023).
Every source listed below is public.
This is a public showcase of The Rural Desk's analytical depth. The actual county-committee directory we maintain — chair names, contact info, organizational addresses — is intentionally not on this site. That data is a product we deliver to clients in private engagements, not a public dataset.
What IS public: the infrastructure to ask any politically-meaningful question about any of the 3,200+ U.S. counties from a single page.
The 2008 election file has a small number of plausibly-corrupted county rows (margins outside ±100% of total votes). We filter those out at build time. If a county shows “no data” for 2008 but has data for later years, this is the most likely reason.
ACS estimates are 5-year rolling averages, which means the 2022 release reflects 2018–2022 averages. For fast-changing counties (in-migration counties especially), the demographic snapshot here lags reality by 2–3 years.
The Rural Desk. (2026). National county-level insights. Retrieved from theruraldesk.com/insights.