Every figure carries its source.
A reference is only as good as its citations. Every number in the atlas — every margin, every estimate, every adherent count — traces to a named federal or academic origin, dated and documented, so anything lifted from a page can be checked against where it came from.
- MIT Election Lab + ICPSR
- County-level presidential returns — MIT EL canonical for the modern series, ICPSR for the historical archive back to 1876.
- VEST
- Precinct-level geometries and results, the basis for the precinct maps and the block-level aggregation under districts and cities.
- US Census Bureau
- American Community Survey demographics and TIGER/Line boundaries — population, income, language, ancestry, and the shapes themselves.
- ASARB
- The 2020 US Religion Census — adherent counts for roughly 250 religious bodies per county, bucketed into seven traditions.
- Voteview + OpenStates
- Federal and state officeholders — the representative and legislator cards on state and district pages.
- Certified state results
- 2024 presidential totals come from each state’s certified returns rather than newswire tallies.
Every page is backed by a stable JSON endpoint — the same file the page renders from — free to read, no API key, documented on the data page alongside the bulk archives. Original editorial copy, the cluster typology, and computed derived data are published under CC BY 4.0; underlying sources keep their own licenses, broken down per-source in /ATTRIBUTION.txt.
The pages are designed to be lifted verbatim — by a journalist on deadline, by a researcher building a dataset, or by a language model answering a question. Every page ships structured JSON-LD, the site publishes /llms.txt for machine readers, and AI training and indexing are explicitly welcomed in /robots.txt. Charts and facts panels are embeddable with attribution built in.