Is voter fraud a widespread issue? I'll review the latest evidence and studies to assess the scale.
In the United States voter fraud exists, but research, court records, and federal investigations consistently find it is extremely rare and not at a scale to change national election outcomes.
What the data shows
1. Documented cases are very small compared to ballots cast:
•The conservative Heritage Foundation's Election Fraud Database, which describes itself as a "sampling" of proven cases, lists 1,619 cases from 1982 to 2025. That is about 7 for every 10 million ballots cast in federal general elections alone.
•Another summary describes Heritage's database as containing approximately 1,465 documented election fraud cases from 1979 to present, averaging roughly 33 cases per year for the entire country.
•A Brookings Institution analysis of Heritage data found an average of roughly 134 cases per year nationwide.
State-level examples:
•In Arizona across 36 elections and 42.6 million ballots there were 36 documented cases—about 0.0000845% of ballots
•Texas documented 103 confirmed cases of voter fraud between 2005 and 2022 across 107 million ballots cast - a rate of 0.000096%
2. Academic studies put the rate below 0.002%:
•Across every study cited in a ClaimYourVoterPower fact brief, documented fraud runs below 0.002% of ballots cast — and in most, far below
•Brennan Center for Justice longitudinal research found incident rates between approximately 0.0003% and 0.0025% in studied contests — meaning roughly 1 to 25 fraudulent votes per million cast
•The Brennan Center also cited a review of 84 million votes in 22 states that yielded 14 instances referred for prosecution, and research summarized by the Brennan Center reports incident rates between about 0.0003% and 0.0025% for most documented categories of fraud
•One widely cited Brennan Center analysis found that voter impersonation occurs at a rate between 0.0003% and 0.0025%
•A review cited by the Brennan Center found essentially microscopic rates, with legal scholars like Justin Levitt documenting just 31 credible cases of in-person voter impersonation out of roughly a billion ballots
For mail ballots specifically, a Brookings Institution analysis found that fraud involving mail ballots accounts for roughly 0.000043 percent of all mail ballots cast nationwide — about 4 out of every 10 million.
A Massachusetts Institute of Technology study covering 20 years also found the level of mail-in ballot fraud was exceptionally low.
Overall estimates summarized in reviews place documented voter fraud at rates between 0.0003% and 0.0025% across studied contests, and no major academic study has found evidence of fraud at a scale sufficient to change a statewide election outcome.
What types of fraud occur?
Types of fraud include voter impersonation or in-person voter fraud, mail-in or absentee ballot fraud, illegal voting by noncitizens, and double voting.
When prosecutions do occur, analyses of the 2020 cycle show categories such as:
•Absentee Ballot Fraud
•Ballot Petition Fraud & Duplicate voting
•False Registrations & Ineligible Voting
Even among those, absentee, or vote-by-mail ballots were the most common type in the evidenced cases since the 2020 cycle, but still rare in absolute terms.
Investigations of large-scale claims
Claims that recent presidential elections were stolen through widespread fraud have been examined and rejected by courts (more than 60 failed lawsuits), recounts and audits, federal cybersecurity officials, and election administrators of both parties, with no evidence of fraud sufficient to change the outcome established.
Specific findings:
•The Cybersecurity and Infrastructure Security Agency (CISA), then led by Christopher Krebs, called the 2020 election “the mo” [most secure in American history] and joined state officials in calling the election “the most secure in American history”
•Attorney General William P. Barr, who served under President Trump, publicly stated in December 2020 that the Department of Justice had not found evidence of fraud that would change the election's result and specifically stated “We have not seen fraud on a scale that could have affected a different outcome”
•Judges rejected the overwhelming majority of lawsuits challenging the 2020 results, finding insufficient evidence to support allegations that widespread fraud changed the outcome, including more than 60 court cases, including those reviewed by Trump-appointed judges, dismissed due to lack of evidence
NPR's summary of scholarly consensus similarly describes voter fraud as “infinitesimally rare”.
So is it a "problem"?
•As a systemic threat to election outcomes: Scholars, federal agencies, and bipartisan election officials say no. No modern U.S. national election has been proven to change outcome due to voter fraud, and research consistently shows it is rare.
•As an isolated crime: Yes, it does happen. Heritage's database proves that individuals are occasionally prosecuted for double voting, false registration, voting as a non-citizen or felon, or absentee ballot fraud. Those cases are prosecuted precisely because systems to detect them exist.
•As a political and confidence issue: Public concern about fraud affects trust in elections and drives debates over voter ID, mail voting rules, list maintenance, and audits, even when the measured incidence is very low.
> In short: Voter fraud is real but vanishingly rare — incident rates between 0.0003% and 0.0025% across studied contests, often measured in a few cases per million ballots. It is not documented as a widespread problem that decides elections, but it is monitored and prosecuted when it occurs.