How Elections Simplified works
Elections Simplified is an independent, nonpartisan guide to what is on your ballot: every race for your address, who is running, what they have actually done, and what each ballot measure means. This page explains who runs the site, where the data comes from, how it is checked, how the ratings are calculated, and how to get a mistake fixed.
Who runs Elections Simplified?
Elections Simplified is published by Elections Simplified Inc. and funded by reader contributions and by its founders. It takes no money from any candidate, campaign, committee, or party. It endorses no one. The source code is public at GitHub, and you can reach the team at contact@electionssimplified.com. Read more about why the site exists on the Mission page.
Where does the data come from?
Races and candidate lists come from state and county election offices: certified candidate lists, sample ballots, and filing records. Candidate records come from official sources: recorded votes, bills sponsored, court and agency records, and a candidate’s own official statements. Campaign finance comes from the Federal Election Commission and state and city disclosure agencies. Ballot measure explanations come from the official measure text and the state’s own voter guide. Every record on the site shows its source link and the date it was researched, so you can check it yourself.
How is the research done?
AI models search for verifiable public records and write plain-language summaries of them. Nothing goes live on that alone: every payload passes automated validators that reject missing sources, dead links, and claims without a source; the sources themselves are checked to be official or primary; and researchers review the results and rerun any search that fails. AI calls are never automatic: each research run is started by a person. Summaries are limited to two sentences, use plain words, and describe what a candidate did, not how the race is going.
What do we leave out?
No endorsements, no predictions, and no opinion about who should win. We do not publish unresolved accusations: a lawsuit filed against a candidate appears only once a court has ruled or an official body has acted. Trivia, biography filler, and campaign talking points are not records. If a candidate has no verifiable public record, the page says so instead of filling the space.
How is “vote power” calculated?
Vote power answers one question: how much can your single vote change the outcome here, compared with other races on your ballot? It combines two measures, representation and decisiveness, into one of six ratings: very low, low, average, above average, high, very high.
Representation
Representation compares the size of the electorate to a statewide vote in the same state. The score is 50 + 50 × ln(state population ÷ district population) ÷ ln(50,000), clamped between 50 and 100. A statewide race scores exactly 50 (the “average” baseline); a district 50,000 times smaller than its state scores 100. Grades: 66 and up is high, 55 and up is above average, 33 and up is average, below that is low. Populations come from the U.S. Census.
Decisiveness
Decisiveness estimates how likely the race is to be close. If every candidate on the ballot wins a seat (one candidate for one seat, or three for three), the race is uncontested and decisiveness is none. Otherwise the grade follows the competitiveness label below: a toss-up or very competitive race is high, competitive or somewhat competitive is medium, and safe is low.
Combining the two
The rating is read off a fixed table of the two grades. High representation with high decisiveness is very high; high representation with an uncontested race is low; average representation needs a high-decisiveness race to reach high; and an uncontested race in a low-representation district is very low. When one measure is unknown, the rating rests on the other and the page says which. Ballot measures are rated on representation alone, because you vote directly on the policy. Judicial retention questions get no rating.
Every election page shows this calculation with the real numbers under “How do we calculate my vote power?”.
How is competitiveness decided?
When a current analyst rating exists for the race, it drives the label. We use the published ratings of Inside Elections and Sabato’s Crystal Ball, averaged when both rate the race, and show the date the ratings were read. Otherwise the label comes from past results for the same office: the winner’s margin over the runner-up as a share of all votes, blended over recent contests when several exist. A margin of 2 points or less is a toss-up, up to 5 is very competitive, up to 10 is competitive, up to 15 is somewhat competitive, and wider than 15 is safe. Past margins are flagged as stale after redistricting.
How are candidate summaries written?
A candidate summary is at most two sentences and 300 characters. It says what the person has done in and around public office, in plain words, with no campaign language and no horse-race framing. Positions on issues come from the candidate’s records: each record is tagged with the issue it touches and whether the action was for or against it, and the page’s stance summary counts those records. A ballot measure gets a two-sentence summary plus a plain statement of what a yes vote and a no vote each mean.
How are issue labels assigned?
Each issue area has a one-sentence goal, such as “ensure elections are secure, accurate, auditable, and trusted by the public” for Election Integrity. A record is labeled “for” an area when the action it describes directly and materially advances that goal, and “against” when it directly cuts against it. The label follows the direction of the action, not the word yes or no: a vote against a bill that weakens audits is for election integrity. A record that does not take a clear position on any area, or whose effect is indirect or speculative, is filed under General with no stance. Changes to how elections are run, such as adopting or banning ranked-choice voting, changing a primary format, or moving an election date, are filed under General whichever side the candidate took: they are positions on a voting method, not on whether elections are secure or accurate. Records on affirmative action, race- or sex-based preferences, and diversity, equity and inclusion programs are also filed under General whichever side the candidate took, not under Civil Rights, unless the record itself acts on a ban on discrimination, such as by adding or removing a protected class or enforcing or weakening an anti-discrimination law. The same written rule is given to the AI labeler and to the people who review its output, and the full prompt is in the public source code.
How is ranked-choice voting shown?
When a race is counted by ranked-choice voting, its card and its page say so and link to a plain explanation of how to fill out the ballot and how the rounds are counted. Where the voting method is not recorded, the race is shown without a note. See ranked-choice voting.
How do I report an error?
Every election page, candidate page, and individual record has a “Report an issue” button. Tell us what is wrong and, if you can, where the correct information is. Reports go to a person, who checks the source, fixes the entry, and reruns the research when a source has changed. You can also email contact@electionssimplified.com. Candidates and campaigns are welcome to send corrections the same way; we correct facts, not framing.
What are the limits?
AI-assisted research. Verify voting information with official election authorities. Elections Simplified is not an official election source: confirm registration, deadlines, and polling places with your state or county election office. Coverage grows race by race, so an empty ballot means “not researched yet”, not “no elections”. Research is dated on every page; a record can be out of date until its next research pass. The full scope and limits are in the Disclaimer.