PoliGov
Choose your state to open its official voter-registration page.
PoliGov combines public election, government, polling, and map datasets into one interface. The sections below describe the current sources and how they are used.
Presidential, Senate, and gubernatorial result records are parsed from the FiveThirtyEight election-results repository. The presidential-election archive uses official Electoral College totals from the National Archives. State popular-vote returns for 1976 through 2008 come from the MIT Election Data and Science Lab. The 1964, 1968, and 1972 state tables come from the American Presidency Project, while 2012 uses the Federal Election Commission. For elections from 1824 through 1908, state popular-vote figures are supplemented by the Election History data compilation. Elections before 1824 emphasize Electoral College allocations because popular-vote records are incomplete or not nationally comparable.
The 2024 local-results layer uses the US County Level Election Results dataset. Result rows are matched to map areas using geographic FIPS or district identifiers.
Most states are shown by county or county-equivalent. Connecticut uses its nine planning regions. Alaska uses its 40 State House districts because the available presidential results are reported at that level.
Base county geometry comes from us-atlas. Connecticut’s current planning-region geometry is based on 2024 Census county-equivalent boundaries, and Alaska House district geometry comes from U.S. Census TIGERweb.
Current senator contact information comes from Senate.gov. Congressional identifiers and term metadata are supplemented by the public Congress Legislators dataset, with portraits loaded from the Congressional Biographical Directory.
Governor profiles and portraits come from the National Governors Association, while recent gubernatorial election results are derived from the FiveThirtyEight results dataset.
The Senate map is built from the Senate.gov contact roster, which supplies each senator's party, class, and office details. Next-election years are derived from each class's six-year cycle, except seats filled by special election: the Florida and Ohio Class III seats (vacated by Marco Rubio and JD Vance in January 2025) are on the November 2026 ballot. Portraits come from the public-domain congressional photo set.
Current representatives, office details, and vacancies come from the Office of the Clerk member roster, which also supplies member portraits. Congressional district geometry comes from U.S. Census TIGERweb.
PoliGov maintains a structured roster of Supreme Court justices and appointment information. Official membership changes should be confirmed through the Supreme Court’s official biographies.
Polling tables are collected on a best-effort basis from 270toWin. Only general-election polls are shown for a race; a race with primary polling only is reported as having no general-election polling. Polls are snapshots, not endorsements.
The forecast estimates each race's margin (Democratic or independent candidate minus Republican) from three inputs and then simulates the election many times to turn margins into probabilities.
1. Polling average. General-election polls from 270toWin, weighted by recency (a poll's weight halves every 14 days), sample size, and population (likely-voter polls count more than registered-voter or adult samples). Undecided voters are not allocated; only the gap between the candidates is used.
2. Expert ratings. The published ratings of up to 12 forecasters (Cook Political Report, Inside Elections, Sabato's Crystal Ball, Decision Desk HQ, The Economist, Split Ticket, Silver Bulletin, FiftyPlusOne, Fox News, NBC News, MS NOW and RealClearPolitics), as compiled by Wikipedia, converted to a margin (Safe 22, Likely 12, Lean 6, Tilt 3, Toss-up 0 points) and averaged. Ratings carry information polls do not: candidate quality, fundraising, and private polling.
3. Fundamentals. The state's or district's partisan lean (2026 Cook PVI, which accounts for this year's redrawn House maps), plus the national environment taken from the generic congressional ballot (about D+8 in late-October polling; the model centers on D+7.5), plus about three points for an incumbent seeking re-election. Governors' races are given less partisan weight because voters split tickets for governor more often.
Blending. A race with plenty of recent polling is about 62% polls, 23% ratings and 15% fundamentals. Races with little or no polling lean on ratings and fundamentals instead. House districts, which are rarely polled, are a blend of ratings and fundamentals.
Simulation. The model runs 40,000 simulated elections. Each one draws a single national error (standard deviation 3.3 points) that shifts every Senate, House and governor race in the same direction, as real polling misses do, plus a separate race-level error (about 4 points for well-polled races, up to 7 for unpolled ones) from a fat-tailed distribution. A party's win probability is the share of simulations it wins; chamber control is counted the same way. In the Senate, Republicans keep control at 50 seats because the vice president breaks ties; independents Dan Osborn (NE) and Seth Bodnar (MT) are counted separately because neither has said he will caucus with Democrats.
Limitations. Alaska uses ranked-choice voting and head-to-head polling; Louisiana's new closed primaries, candidate scandals, and late events after the model run are not modeled beyond what the polls and ratings already reflect. The House view shows each state's projected delegation rather than district shapes, because Texas, California, Missouri, North Carolina, Ohio, Utah and other states redrew their maps for 2026. After Election Day, PoliGov will score each projection against the result.
Official government portrait sources are used when available. Presidential candidate portraits are retrieved through the Wikipedia image API. PoliGov caches external data for performance, so a recent upstream change may not appear immediately.