What research tells us about voter fraud

A guide to help journalists define voter fraud, understand how rarely it occurs, and spot potential errors in government datasets

By Rachel Layne

In the lead-up to the 2026 midterm elections, the Trump administration has repeatedly used claims of widespread voter fraud to justify an unprecedented shift toward federal control over elections and an array of attempted restrictions on voting access. These include limiting the right to vote by mail, requiring documentary proof of citizenship to vote, and requiring states to share sensitive and personally identifiable voter data with the federal government.

Journalists covering these claims can turn to a significant body of evidence-based, independent, nonpartisan research into how often and under what circumstances fraud occurs in United States elections.

Evidence compiled by social scientists, independent observers and government officials, and detailed in court rulings, shows documented instances of voter fraud are exceedingly rare.

This guide explores research related to claims of noncitizen voting alleged by the current White House; what constitutes voter fraud; how uncommon it is, and why; why the U.S. doesn’t have a national voter list; why it’s important to consider statistical errors when assessing voter fraud claims; what research shows about voter fraud beliefs and how to counter them; what reporting advice researchers who study the topic offer; and where to find reliable data sources on the topic.

Voter fraud happens when someone knowingly casts a ballot illegally. This can include violation of individual state or federal laws that prohibit voting as a non-citizen, in the name of a dead person, while pretending to be a registered voter under someone else’s name, more than once or in multiple places, on behalf of someone else, after tampering with ballots or voting systems, or in a state that is not the voter’s legal domicile. Voting while incarcerated or after a felony conviction in states where that’s illegal is also considered fraud, but the particulars of what’s allowed vary widely from state to state.

For journalists reporting on claims of voter fraud, it’s important to clarify what fraud actually is, as social media posts and public officials sometimes use the term loosely or incorrectly.

That means distinguishing fraud from administrative errors that result in someone voting when they’re not authorized to do so; from other types of crime or corruption; and from campaign tactics that are aggressive or manipulative, but still legal.

Political fundraising by federal employees is an example of an act that’s illegal, but not fraud. Gerrymandering is an oft-criticized technique to swing elections, but there are no federal laws prohibiting it, and it’s not fraud. Submitting a mail ballot without a signature will render that ballot ineligible, but it’s an error, not fraud, whereas forging someone else’s signature on a mail ballot is fraud.

“There is a specific legal definition to fraud,” says Sean Morales-Doyle, senior director of voting rights and elections at the Brennan Center for Justice at New York University, which analyzes election procedures and laws and advocates for fewer barriers in access to voting.

He recommends “being very careful to distinguish between the different types of crimes and forms of corruption, and the way that those relate to the kinds of policies and responses that people are putting on the table.”

Research dating back for decades shows that documented instances of voter fraud are exceedingly rare, including in three of the categories the Trump administration has been targeting: noncitizen voting, potential interference with voting machines, and fraud related to voting by mail.

• As noted in SciLine’s reporting guide on voting by mail, “‘Studies consistently have found negligible evidence of fraud in state elections,’ researchers write in [a] 2023 paper from the MIT Election Data and Science Lab. Researchers at the Brookings Institution, a nonpartisan think tank, examined cases of fraud reported during elections in 2016, 2018, 2020, and 2022 and estimated there were about four cases of fraud for every 10 million mail ballots cast, on average.”
• Voting technology became central to conspiracy theories and disproven allegations about the results of the 2020 presidential election. President Trump and some of his supporters have revived some of those assertions during the 2026 election season. But as noted in SciLine’s reporting guide on voting machines, “Multiple studies examine and find no evidence for the claim that voting machines ‘flipped’ votes from Donald Trump to Joe Biden in the 2020 presidential election.”
• The Center for Election Innovation and Research (CEIR), a nonprofit organization that tracks public claims of possible noncitizen voter registration or voting across the U.S., reports that “sweeping allegations about possible noncitizen registration or voting appear to arise from misunderstandings, mischaracterizations, or outright fabrications about complex voter data.”

CEIR’s frequently-updated interactive tracker is a helpful resource for local journalists. In addition to documenting initial claims, it tracks official investigations and court cases about noncitizen voter fraud, noting that “even the largest claims represent tiny fractions of registered voters in any state,” and that “regardless of the state, the numbers announced after a proper investigation are invariably far smaller than any numbers announced before review.”

For example:

• In October 2025, the Texas secretary of state sent county officials names from a list of 2,724 voters flagged as potential non-citizens; for context, Texas had about 18.6 million registered voters during the 2024 election. Officials investigated those names. In September 2026, the state announced that 578 of the people on the list had “demonstrated” they were actually U.S. citizens, and that 117 cases of “potential noncitizen voting” had been referred to the state attorney general’s office for further investigation. Separately, federal and state officials said they had indicted or charged seven people for acts tied to noncitizenship and voting in Texas.

• In Washington state, 15 cases of noncitizen voting have been prosecuted from 1982 through 2025, or about 0.000006% ballots cast, according to the secretary of state’s office.
• In July, a Kentucky state legislator said he identified 159 noncitizens with voting histories, citing the past five years based on the state’s board of elections. There are some 3.2 million active registered voters in the state, making this claim, if proven valid, about 0.012% of active voters. That state board reported that 379 noncitizens had been removed from its voter list since October 2024, CEIR’s tracker notes.

Voter fraud is rare in part because states and local jurisdictions have systems in place to prevent it. The National Conference of State Legislatures provides a guide to different kinds of potential fraud and applicable state laws, checks, and mechanisms each jurisdiction employs to make sure registered voters are allowed to legally cast their ballots.

The potential payoff for committing individual voting fraud as a noncitizen is negligible compared with the penalties, which can include fines and prison time and even deportation.

“It’s really the payoff that people don’t think about very often. You get to cast one ballot. Is it worth it to someone to put themselves in that kind of risk of those kinds of consequences for the payoff of being able to cast one ballot?” Morales-Doyle says. “That’s what we’re really talking about here, and that’s why it doesn’t happen very often.”

Security protocols and registration checks by local officials often catch errors and deliberate malfeasance when they do occur, including accidental voter registrations.

In July, for example, New Jersey Governor Mikie Sherrill announced that about 6,600 people who identified as noncitizens were erroneously registered to vote in 2023 and 2024, through no fault of their own, due to a software error in the state’s Motor Vehicle Commission automatic voter registration system.

She added that a preliminary analysis showed that fewer than 400 of those people actually voted.

CEIR notes that there are about 6.1 million active registered voters in New Jersey, and, “Taken at face value, the largest claim documented [6,600 erroneous registrations] would be roughly 0.1082% of active registered voters.”

An eligible voter can also be erroneously flagged as not belonging on a state’s voter rolls for a multitude of reasons, some of which involve lag in updating a voter’s status – including address changes, prison release, and becoming a naturalized citizen.

Following an executive order in March 2025, the Justice Department began demanding full voter records from the states, including sensitive and personally identifiable information such as partial Social Security numbers. The data is meant to be folded into the U.S. Citizenship and Immigration Services’ Systematic Alien Verification for Entitlements  program. (Created in 1986, SAVE was originally designed to check immigrants’ eligibility for public benefits. The Trump administration expanded it for voter-related citizenship checks.)

Most states declined to produce unredacted records, as there are various federal and state laws requiring election officials to mask and protect certain sensitive information when sharing voting lists, but did offer to release lists with that information redacted. The DOJ subsequently filed 31 lawsuits against 30 states and Washington, D.C., for failing to provide it with the unredacted lists it demanded. As of Sept. 22, none had been successful in forcing release of the data, and federal courts had dismissed 25 of those cases.

In June, a federal judge blocked the SAVE program for use by any state for elections, citing errors in data from the program when it was merged with Texas data that erroneously removed citizens from the voter rolls, including identifying some naturalized citizens as noncitizens.

Merging big datasets tends to produce errors for individual voters, and may lead to disenfranchisement.

Hypothetically, an 0.1% error rate result from a merger of voter datasets, sometimes also called record linkage, (state data with the SAVE system, for example) of 10 million voter records would mean 10,000 citizens could be erroneously flagged and not be allowed to vote. Any large claim of fraud drawn from such a database should be approached with “a great deal of skepticism,” Morales-Doyle says, noting,“That tiny error rate can end up seeming like a really large number.”

On August 18, the Census Bureau released a “beginning analysis” of non-citizen voting.

The report said the Census Bureau “readily determined with high confidence” that more than 128 million voters in the 2020 election were citizens. It also said more than 24,000 records represented “noncitizens at the time of the election” – or less than .019% of the 128-million voter sample.

As with the SAVE program issue, the Census analysis also has a potential error rate problem, wrote Justin Levitt, a Loyola Law School professor and elections expert in an entry to his election law blog.

The report does not contain the caveats detailing potential error sources in the work typically found in Census reports. But one of the five sources that are cited at the end of the report details concerns about the method used to link voter records, indicating the likelihood of errors.

“Put differently: it’s entirely possible that every single purported noncitizen voter could instead be explained by errors in the matching process alone,” Levitt wrote.  “We have no sense of whether the errors in the private vendor’s file, or the errors in the federal government data, are larger still.”

Trump often cites perceived voter fraud as a reason to establish a national list of voting-age citizens. This may make news audiences wonder why the U.S. doesn’t have a national voter list already.

Article 1 of the U.S. Constitution places responsibility for running elections with individual states, which oversee elections in more than 10,000 local jurisdictions. This includes voter registration, administration leading up to, during, and after an election, and determination of whether a voter is eligible to cast a ballot.

Registration procedures vary by state. Most states require prospective voters to fill out a form with name, address and birth date, and, in 32 states, political party affiliation. The 49 states requiring voter registration mandate potential voters to declare they are U.S. citizens upon penalty of perjury. While some states require documentary proof of citizenship, others mandate it only if a potential voter’s status can’t be determined through other means, such as driver’s license data.

North Dakota doesn’t require voter registration in part “because of its rural character,” according to an explainer from the secretary of state’s office. Voters present “acceptable” identification at the polls, and smaller precincts are “intended to ensure that election boards know the voters who come to the polling places to vote on Election Day and can easily detect those who should not be voting in the precinct.”

State oversight of elections, a pillar of the system of federalism prescribed in the Constitution, is a big reason there isn’t a national voter list—and Congress has never required one, according to the nonpartisan Bipartisan Policy Center, a group that advises local election officials and parses these disparate policies for the public.

Scientific studies can help journalists cover claims and cases about voting fraud without fueling false beliefs. Research to date suggests:

Giving voters information on election security measures and procedures before they encounter false information—a process called “prebunking”— increases confidence in voting and lowers fraud concern, the authors of an August 2025 peer-reviewed paper in the journal Science Advances found.

Their series of studies queried thousands of voters before and after the 2022 midterm elections in the U.S., when then-former President Donald Trump was amplifying claims of voter fraud in the 2020 presidential election, and, separately, after the 2022 presidential election in Brazil, which followed years of President Jair Bolsonaro claiming that his nation’s electronic voting system was vulnerable to fraud. Bolsonaro was subsequently sentenced to 27 years in prison for attempting a coup following his loss that year.

Researchers also found that voter confidence in the election process also increased, though to a lesser degree, when the information came from “credible sources,” people who expressed confidence in the system even though it didn’t always benefit them, such as well-known political figures who supported Trump in the 2020 election, which he lost.

A study in the April 2026 issue of Social Science Quarterly, drawing on responses to the 2024 Cooperative Election Survey, found that while both Democrats and Republicans agreed that illegal acts such as casting multiple ballots are fraudulent, use of the term “election fraud” has broader context for both groups.

Democrats tended to associate the phrase “election fraud” with “the bogus nature of Trump’s narrative. As a result, they hear ‘fraud,’ and think ‘the Big Lie,’” the authors write.

“For Republicans, it was predominantly that people casting votes by mail was ‘fraudulent,’ as were any elections that didn’t feature voter identification rules,” says Soren Jordan, an associate professor at Texas A&M’s Bush School of Government and Public Service and a study co-author. “Those are the kinds of assertions that give you more pause about how the narrative is constructed about fraud, and also how we talk about either resolving fraud or the preponderance of fraud to the American public.”

The differing rules from state to state, while a testament to the federalist system outlined in the Constitution, can lead to a loss in confidence in the electoral process, writes Joshua D. Clinton, a professor and co-director of the Center for the Study of Democratic Institutions at Vanderbilt University in a May 2026 study on voter confidence published in the the American Political Science Review.

Confidence diminishes more when voters disagree with election procedures in their state, whereas confidence rises when they agree with how an election was conducted. The loss is “most pronounced among Republicans and unaffiliated voters,” the study finds.

Because the rules are different in each state, suspicion about national elections also rises when results are close, Clinton notes.

“Across states, the problem can intensify when closely contested national outcomes hinge on results produced under divergent and increasingly polarized procedures, inviting skepticism about the results of elections conducted under different rules.” he writes.

We asked several researchers to share reporting tips for local journalists in advance of the 2026 midterm elections. Here’s what they said:

“My favorite pieces of journalistic coverage all open with some broad disclaimer that says something to the effect of ‘election laws vary greatly across states,’” Jordan says. “Everyone agrees that ‘cheating’—destroying ballots, stuffing ballot boxes—is ‘fraud.’ But some groups go further: [suggesting] ‘fraud’ means voting without showing picture identification. Depending on the rules of the state, that’s not fraudulent, as it’s allowed to happen in the election.”

Interviewing an election official who can walk journalists through how they follow the rules is also important, says Thessalia Merivaki, the director of the Thomas S. Foley Institute of Public Policy and Public Service and a professor of government at Washington State University. She urges journalists to develop relationships with local election officials as soon as possible and well before an election.

“Trying to get a quote from the officials is also important,” Merivaki says. “Avoiding sensationalizing these stories is critical.”

Bright Line Watch’s surveys have asked people about a wide array of fraud types, according to John M. Carey, a professor of government at Dartmouth College who says, “the beliefs that people subscribe to about how common they are are just orders of magnitude out of whack with what any reasonable evidence would suggest.”

That means journalists and other messengers “really need to stay at it. People need to get the message more than once from more than one source. To the extent that journalists can encourage political elites and opinion makers to be reiterating these messages, so much the better,” Carey says.

If a public official makes claims about voting fraud, ask for the source of information used to identify alleged improper voting. Databases available to the public often don’t have enough information to definitively verify that a particular person has incorrectly voted. Reporters “should be wary” of these claims, says Matthew Weil, vice president of the governance program at the Bipartisan Policy Center, a nonprofit thinktank.

Some states re-examine voter rolls after an election by using their full database to check and make sure people who shouldn’t have voted did not participate in the election. Journalists can follow up when officials conduct such investigations.

The Bipartisan Policy Center is a Washington-based think tank founded in 2007 by former U.S. Senate majority leaders Howard Baker, Tom Daschle, Bob Dole and George Mitchell to seek and create bipartisan policy recommendations.

The Brennan Center for Justice at New York University studies voting rights and advocates for fewer barriers and more equitable access to voting. They follow elections and related issues closely.

The Center for Election Innovation and Research is a nonprofit organization founded in 2016 that works with local election officials to build public confidence in elections using a nonpartisan approach.

The Congressional Research Service: run through the Library of Congress, researches and creates nonpartisan reports for members of Congress on a wide variety of issues. For this report, we used those related to voting machines, election laws and regulations on the federal level.

The United States Election Assistance Commission: the federal agency created by HAVA to assist states with elections.

The MIT Election Data Science Lab: This research center at the Massachusetts Institute of Technology, founded by political scientist Charles Stewart III, collects, analyzes, and shares data on U.S. elections.

National Conference of State Legislatures: the NCSL was created by state legislatures in 1975. It tracks legislation and policy at the state level. They offer detailed explainers about elections, state laws and proposed legislation, voter registration and verification, voting systems, and other election-related topics.

National Association of Secretaries of State: This bipartisan organization meets to discuss issues that generally fall under the secretary of state role. Their media campaign #Trustedinfo can be found here, along with resources for each state.

The Ohio State University’s Election Cases Tracker is a sortable database that makes it easier to find court cases related to elections.

The University of Wisconsin’s Lawsuits Seeking States Sensitive Voter Data Tracker is a database tracking the lawsuits filed by 30 states and Washington D.C. to block the Trump administration’s demand for voter data. It’s run by the law school’s State Democracy Research Initiative.

The United States Vote Foundation, a nonpartisan nonprofit, maintains a directory of local election officials.

 Rachel Layne is a Boston-based freelance journalist whose work has appeared in outlets including CBS News, CQ Researcher, The Journalist’s Resource, HBS Working Knowledge and USA Today. Previously she spent 20 years at Bloomberg News, where she covered multinational corporations, among other roles. She is also an adjunct instructor in Emerson College’s journalism department.