Where our US results come from

Each race's result, the figure on its state page, comes from the most authoritative source we could find, in this order: the state's own certified results and archives; the Clerk of the House of Representatives' Statistics of the Congressional Election; the MIT Election Data and Science Lab's returns (1976 on); Patrick Heideman's county results (2026); Algara and Amlani's county returns for president, Senate and governor (2021, Harvard Dataverse); and The American Vote's state results for president.

In November 2026 every one of the roughly 14,000 races we hold was compared with an independent record: Wikipedia's tables for each year and state, the Clerk's statistics for the House from 1990, and MEDSL's returns. Any race whose winner, candidates or votes differed by more than half a percentage point of the vote was corrected, about 5,000 in all. Most corrections were names (misspellings, placeholders, a missing accent). Others were real errors: a party's votes filed under the wrong party, the votes a candidate won on a second ballot line (New York's American Labor and Liberal lines, for example) left out, a runoff added to the first round, or a year's count inflated by a county counted twice.

A few results need a note, and the race's page gives it. Where nobody won a majority and a legislature chose the governor (Georgia in 1966, several New England states in the 1800s), the winner shown is the person who took office, with the vote as cast. Where a seat was unopposed and no vote was printed (Florida and Louisiana often did not print one), the race shows no vote. North Carolina's 9th district in 2018 was never certified, so nobody is shown as elected. Candidates outside the two main parties are named wherever they won or took at least 5% of the vote: La Follette's Progressives, Minnesota's Farmer-Labor senators, Jesse Ventura, Angus King, Bernie Sanders and others.

The House seat totals for every year from 1990 to 2024 match the official totals. The electoral votes for every presidential year match those actually cast, including split states, unpledged and faithless electors.

County figures before 2000 come mostly from Algara and Amlani's county returns; from 2000 they come from MEDSL, Heideman and the states. Every race's county figures were then compared with its corrected state result. Three kinds of difference were fixed: errors in the county files (Baltimore County holding Baltimore City's votes as well, a county whose total counted its votes twice, Kansas City missing from Jackson County, Missouri); columns that meant another party (a "Democratic" column that held a third party's votes, or "other" votes that were the Republican vote); and the Democratic-Farmer-Labor and Democratic-NPL House votes filed as "other".

Where a race's county figures still differed from its certified state total by more than 3% of the vote, usually because a candidate's second ballot line was not in the county returns, each party's county votes were scaled to the state's figure, keeping each county's share of the state's vote. Those county figures are estimates, and the county pages show the source as "scaled to the certified state result". In 41 races a whole party's county votes were missing from the county returns; those races show the state result only, without a county map.

Counties change. Before 1990 our maps can draw each county as it stood on election day, using the Newberry Library's Atlas of Historical County Boundaries (CC BY-NC-SA 2.5), or share the old votes among today's counties by area. The state pages list counties as they stood; the county pages, which follow today's counties, show the shared estimates for years before 1990.

Hawaii's House results for 1990 to 1994 exist only as scanned state reports and are not yet loaded. Where a state held two special Senate elections in one year, only one is shown. A few early governor races (Tennessee 1872 to 1878, Connecticut 1876) could not be checked against a second source. Certified results for November 2026 will be added when the states certify them in December.

If you find a figure that looks wrong, tell us through the about page and we will check it against the official record.