“Extensive audits and independent reviews of United States elections show that voter fraud of any type occurs in less than 0.003 percent of all ballots cast.”

• The Brennan Center's landmark review 'The Truth About Voter Fraud' (Justin Levitt, 2007) examined multi-state fraud allegations and found that credible, documented incident rates fell between 0.0003% and 0.0025% of ballots cast — an upper bound below the 0.003% figure in the claim.
• Justin Levitt's 2014 comprehensive nationwide investigation identified 31 credible incidents of voter impersonation out of roughly one billion ballots cast in general, primary, special and municipal elections from 2000–2014 — a rate of about 0.000003%.
• The News21 investigative project (Arizona State University) compiled a database of 2,068 alleged election-fraud cases nationwide from 2000–2012, including only 10 cases of in-person voter impersonation, out of hundreds of millions of votes cast.
• An Associated Press review of every potential case of voter fraud in the six battleground states disputed by Donald Trump after the 2020 election found fewer than 475 potentially fraudulent ballots out of 25.5 million cast — roughly 0.0019%, below the claimed threshold, and far too few to affect any outcome.
• State-level audits reach similar magnitudes: Ohio's secretary of state referred roughly 135 possible non-citizen or double-voting cases out of 5.6 million 2012 ballots (~0.002%); Georgia's full hand recount and multiple 2020 audits, Arizona's and Michigan's reviews, and dozens of federal and state court rulings produced no evidence of fraud at a scale remotely approaching 0.01%.
• Even the Heritage Foundation's deliberately maximalist Election Fraud Database — assembled by an organization that advocates stricter voting laws — contains only roughly 1,500–1,600 proven criminal cases accumulated over several decades, against billions of ballots cast, which works out to a vanishingly small percentage.
• Fact-checking organizations (FactCheck.org, PolitiFact, Snopes) and the MIT Election Data and Science Lab have repeatedly concluded that documented voter fraud in the U.S. is extremely rare and statistically negligible.
• Trump's own Presidential Advisory Commission on Election Integrity (2017) disbanded in 2018 without producing evidence of widespread fraud.

• The specific figure '0.003 percent' is not a single published national statistic; it is a rounded ceiling derived from the Brennan Center's 0.0003%–0.0025% range, which was based on a handful of state studies (Ohio, Washington, Missouri, New Jersey) and largely addressed impersonation and double-voting, not literally 'fraud of any type.'
• All such figures measure *detected, credible, or prosecuted* fraud, not actual fraud. Undetected fraud is by definition unmeasured, so the claim overstates what audits can prove; it should be phrased as 'confirmed fraud occurs in fewer than X% of ballots.'
• Some categories of irregularity are found at higher rates than 0.003% in specific audits — e.g., reviews of voter rolls periodically identify hundreds or thousands of ineligible registrants (deceased, non-citizen, or cross-state duplicate registrations) in a single state, though the vast majority never cast ballots and many are clerical errors rather than fraud.
• Absentee/mail-ballot fraud, the category experts consider most vulnerable, has produced real conspiracies (e.g., the 2018 NC-9 congressional race, which was overturned; various municipal-level absentee schemes in Mississippi, Texas and Pennsylvania), showing that rates can be locally much higher than the national average even if the national rate is negligible.
• Critics (notably the Heritage Foundation and some election-integrity researchers) argue that low detection rates partly reflect low investigative capacity and infrequent prosecution rather than low incidence — a contested but non-frivolous methodological objection.
• The word 'Extensive audits' slightly overstates matters: most post-election audits verify tabulation accuracy, not voter eligibility, so they are not direct measurements of fraud rates.

86/100 — the lab’s own confidence in the score above, not a second reading of the claim.

Brennan Center for Justice, NYU School of Lawedu · 85 · Brennan Center for Justiceedu · 82 · The Washington Post (Justin Levitt, Loyola Law School)news · 88 · News21 / Walter Cronkite School of Journalism, Arizona State Universityedu · 85 · Associated Pressnews · 92 · The Heritage Foundation (Election Fraud Database)other · 55 · FactCheck.org (Annenberg Public Policy Center)factcheck · 90 · National Conference of State Legislaturesgov · 85

• The strongest support is that multiple large reviews find known or credible voter-fraud incidents far below 0.003% of ballots. The Brennan Center’s review reported incident rates generally between 0.0003% and 0.0025%, below the claimed threshold.
• An Associated Press review of the 2020 election in six battleground states found fewer than 475 potential voter-fraud cases among more than 25 million ballots, roughly 0.0019%, also below 0.003%.
• Justin Levitt’s review, reported by The Washington Post, found only 31 credible in-person impersonation incidents out of more than 1 billion ballots cast from 2000-2014; that is far below 0.003%, though it covers only one kind of voter fraud.
• Government audits and reviews in states such as Georgia and Michigan after 2020 confirmed vote-count accuracy and did not find widespread fraud capable of changing outcomes; federal election-security officials likewise said there was no evidence voting systems changed, deleted, or lost votes.
• Even databases maintained by organizations concerned about voter fraud, such as the Heritage Foundation’s election-fraud database, list a small number of proven cases relative to the hundreds of millions or billions of ballots cast over the same broad period.

• The wording 'voter fraud of any type' is broader than much of the best evidence. Many often-cited studies focus on in-person impersonation, prosecutions, proven cases, or potential cases found by reviews; they cannot prove the true rate of every undetected fraud type nationwide.
• There is no single comprehensive national audit of all U.S. elections and all fraud categories. U.S. elections are decentralized, and routine post-election audits are usually designed to verify vote-count accuracy, not to detect every illegal or fraudulent ballot.
• Some official reviews identify categories such as possible double voting, felon voting, noncitizen voting, or other ineligible ballots at rates that can exceed 0.003%; however, those figures often include administrative errors, eligibility misunderstandings, or unresolved cases, not necessarily intentional voter fraud. North Carolina’s 2016 post-election audit, for example, flagged 508 ineligible votes out of about 4.8 million ballots, about 0.01%, but did not treat all as intentional fraud.
• Because the claim gives a precise numerical ceiling for 'any type' of voter fraud, it is somewhat stronger than the evidence strictly supports. The evidence supports 'documented voter fraud is extremely rare and usually well below 0.003% in major reviews,' more than it proves a universal nationwide rate for all fraud types.

82/100 — the lab’s own confidence in the score above, not a second reading of the claim.

Brennan Center for Justiceother · 88 · The Washington Postnews · 88 · Associated Pressnews · 92 · Cybersecurity and Infrastructure Security Agencygov · 92 · Georgia Secretary of Stategov · 90 · Michigan Secretary of Stategov · 90 · Heritage Foundationother · 65 · FactCheck.orgfactcheck · 90

• Extensive research by nonpartisan organizations and academic institutions consistently demonstrates that voter fraud is exceedingly rare in the United States. A comprehensive study by the Brennan Center for Justice reviewed elections that had been meticulously studied for voter fraud and found incident rates ranging between 0.0003 percent and 0.0025 percent, which directly supports the claim's 'less than 0.003 percent' statistic.
• A widely cited investigation tracking over 1 billion ballots cast in general and primary elections between 2000 and 2014 found only 31 credible instances of in-person voter impersonation.
• Even databases compiled by groups advocating for stricter voting laws to combat fraud, such as The Heritage Foundation's Election Fraud Database, document fewer than 1,550 proven instances of fraud over several decades. When compared to the billions of votes cast nationwide during that same timeframe, the rate of proven fraud remains mathematically infinitesimal and well below the 0.003 percent threshold.
• Audits of recent national elections, including a thorough review of the 2020 election by the Associated Press, found fewer than 475 potential voter fraud cases across six battleground states, an amount far too small to affect the outcome and consistent with historically negligible fraud rates.

• Critics of these statistical findings argue that the reported percentages only represent the fraud that is successfully detected and prosecuted. They suggest that because election systems often lack perfect verification mechanisms (such as strict photo ID requirements in some jurisdictions), the actual rate of fraud could be higher than what documented incidents show.
• While the national average is exceptionally low, specific local elections have occasionally seen higher concentrations of fraud. For example, the 2018 North Carolina 9th Congressional district election was overturned due to a coordinated, illegal absentee ballot harvesting scheme, demonstrating that organized election fraud can occasionally reach disruptive levels locally.

95/100 — the lab’s own confidence in the score above, not a second reading of the claim.

Brennan Center for Justiceother · 90 · The Washington Postnews · 95 · Associated Pressnews · 95 · The Heritage Foundationother · 70

• The claim is directionally consistent with extensive election research showing that widespread voter fraud is not supported by evidence. MIT’s Election Data and Science Lab describes false beliefs about the prevalence of voter and election fraud and reports that corrective information reduces those beliefs.[1]
• The Heritage Foundation’s database documents thousands of proven or alleged election-fraud cases over many years, but it explicitly says its list is only a sample and is not exhaustive; this supports the conclusion that documented fraud exists while not establishing that it exceeds 0.003 percent of ballots.[2]

• The precise figure of 0.003 percent is not established by the cited audits or reviews. The claim says voter fraud of any type, but fraud categories, definitions, detection methods, time periods, and denominators differ substantially across studies.
• The Heritage Foundation reports almost 1,600 proven cases in its database and describes them as instances of election fraud, demonstrating that fraud does occur; however, the database does not provide a national, all-ballots rate that can validate or refute 0.003 percent.[3]
• The available MIT source is a survey about perceptions of fraud, not an audit or a count of fraudulent ballots, so it cannot substantiate the numerical threshold.[1]
• Because undetected fraud cannot be directly counted, a universal statement that all types occur below an exact threshold is stronger than the available evidence supports.

84/100 — the lab’s own confidence in the score above, not a second reading of the claim.

MIT Election Data and Science Labedu · 90 · The Heritage Foundationother · 72 · The Heritage Foundationother · 70

• Brennan Center and similar reviews of U.S. elections since 2000 have found only dozens of confirmed in-person impersonation cases against more than a billion ballots cast.
• State risk-limiting audits, hand recounts, and forensic reviews after 2020 (Georgia, Wisconsin, Michigan, Arizona official canvass) produced discrepancies of hundreds of votes at most, far below 0.003% of ballots.
• Even the Heritage Foundation compilation of ~1,500 proven fraud convictions over multiple decades remains well under 0.003% when divided by cumulative federal, state, and local ballots.
• CISA, state election officials of both parties, and multiple fact-checks concluded that 2020 (and prior) fraud did not occur at outcome-changing scale.

• Convictions and detected cases undercount true incidence because many potential irregularities (duplicate voting, ineligible registrants) are never fully investigated or prosecuted.
• Voter-file matching against death, immigration, and multi-state records flags thousands of possible ineligible appearances per cycle, though most resolve as data errors.
• Organized absentee or proxy fraud has been documented in isolated local races (e.g., 2018 NC-09), showing rates can exceed the national average in low-turnout settings.
• The precise 0.003% figure is an approximate upper-bound synthesis rather than the direct published result of any single nationwide audit.

72/100 — the lab’s own confidence in the score above, not a second reading of the claim.

Brennan Center for Justiceother · 75 · The Heritage Foundationother · 64 · CISAgov · 92