The truth about noncitizen voting in federal elections

— Former president Donald Trump, remarks at Eagle Pass, Tex., Feb. 29

“That’s why they are allowing these people to come in — people that don’t speak our language — they are signing them up to vote.”

— Trump, remarks during a rally in Sioux Center, Iowa, Jan. 5

Decrying the surge of undocumented immigrants at the southern border, Trump has suggested a nefarious reason for it — a desire by President Biden to tip the 2024 vote by enlisting migrants to cast ballots. Trump’s musings are reflected on social media, with the phrase “they are importing voters” spread across X, formerly known as Twitter, by many users (including by the site’s owner, Elon Musk).

Trump’s fearmongering about votes allegedly cast by noncitizens is not new. But fresh research by a professor often cited by Trump’s supporters further undercuts his claims.

After he was elected president, with a narrow margin in the electoral college, he insisted (without evidence) that the reason Hillary Clinton won 2.8 million more votes was that at least 3 million votes were cast by undocumented immigrants.

When Trump lost the 2020 election — where, by the popular vote count, Biden beat him by 7 million votes — Trump claimed that Biden’s margin of victory in key states, such as in Arizona, was provided by noncitizens. Biden won Arizona by just over 10,000 votes, and Trump falsely said that 36,000 votes in the state were cast by noncitizens.

Trump’s reiteration of that claim in his speech on Jan. 6, 2021, to supporters who later attacked the Capitol is referenced in the federal indictment accusing him of conspiring to overturn the election results.

Trump’s supporters echoed the falsehood. Rudy Giuliani, Trump’s personal attorney, lost his law license after a committee of judges from New York’s First Department Appellate Division cited his widely divergent “false and misleading” claims about noncitizen voting in Arizona, including “way more than 10,000,” “the bare minimum is 40 or 50,000,” and “the reality is probably about 250,000.”

The claim that undocumented immigrants vote in such numbers that they can swing a presidential election has weakened considerably since 2016. A researcher long cited by Trump and his allies has now dug into Arizona’s voter and driver’s license files and, in documents filed under seal in court and obtained by The Fact Checker, dramatically recast his findings. He still believes that data suggest that noncitizens cast some votes that are not easily detected by authorities — but the numbers did not change the result in the state.

Let’s try to separate fiction from fact.

Federal law bans noncitizens from voting in federal elections, including races for president, vice president, Senate or House of Representatives. Under a law adopted in 1996, noncitizens who vote can face a fine or a prison term as long as a year, or both — not to mention deportation.

Until the 1920s, when there was a wave of anti-immigrant sentiment, 22 states and federal territories allowed noncitizens to vote in state elections. Now only a few jurisdictions in California, Illinois and Maryland — notably Chicago and San Francisco — allow some form of noncitizen voting. Some states, such as Ohio and Louisiana, in recent years have enacted constitutional bans on noncitizen voting.

A majority of states — 36 — request or require voters to show some sort of identification when they vote, according to the National Conference of State Legislatures. Increasingly, states also are seeking ways to require proof of citizenship to vote or are considering ways to update voter lists to eliminate any noncitizens. Voting rights advocates have argued that such requirements are burdensome, as few Americans carry around birth certificates, naturalization certificates or passports — even if they have such documents.

The Supreme Court, in a 7-2 vote, in 2013 invalidated an Arizona law that prospective voters in Arizona provide proof of citizenship to be able to register to vote in national elections. The majority opinion, written by Justice Antonin Scalia, said the state could not reject a federal voter registration form, which does not require such documentation; the form requires only that an applicant, under penalty of perjury, must affirm he or she is a citizen. About 19,000 registered voters in Arizona are “federal-only,” meaning they filled out a federal form and vote only in federal elections.

After the 2020 election, the Arizona legislature passed bills that required proof of citizenship to participate in state elections and verification of the status of registered voters who had not provided proof of their citizenship. U.S. District Judge Susan Bolton last week upheld some provisions, saying that “the court concludes that Arizona’s interests in preventing noncitizens from voting and promoting public confidence in Arizona’s elections outweighs the limited burden voters might encounter when required to provide” proof of citizenship.

There is scattered evidence of noncitizens voting in federal elections — sometime by mistake (such as erroneously thinking they were eligible while getting a driver’s license) but also with nefarious intent.

The conservative Heritage Foundation maintains a database, dating to 1979, that it says includes a “sampling” of election-fraud cases brought by prosecutors. In that period, about 2 billion votes have been cast in federal elections, according to a calculation for The Fact Checker by the Brennan Center for Justice. A recent search of the Heritage database found 85 cases involving allegations of noncitizen voting from 2002 to 2023. A large percentage of the cases took place in North Carolina, where authorities have been aggressive in targeting noncitizen voting.

In 2022, Georgia announced that it had completed a citizenship review of the state’s voter rolls and discovered that 1,634 people over 25 years had attempted to register to vote even though they were not U.S. citizens. But none had been permitted to register to vote and, thus, had not cast ballots.

Sometimes state efforts to track down noncitizen voting backfire. Texas in 2019 flagged 100,000 voters as possible noncitizens, but at least 25,000 were erroneously flagged because of data mix-ups. Fourteen voters were removed from voting rolls and then had to be reinstated. A federal judge halted the effort, saying it unfairly targeted naturalized citizens, and, eventually, Texas abandoned the effort. “This is a solution looking for a problem,” U.S. District Judge Fred Biery said.

Still, Texas’s difficulties illustrate how difficult it might be uncover voting by noncitizens without running afoul of constitutional protections. In the Arizona case, Bolton nullified a provision that would require voters to list their place of birth, saying it violated the Civil Rights Act and the National Voter Registration Act and would result in investigations of only naturalized citizens.

Given the paucity of evidence of noncitizen voting, many election researchers have long said that there was little to support the idea that noncitizen voting had ever affected the outcome of a major election. But that does not necessarily prove that the phenomenon does not happen. Unlike many crimes, if a noncitizen casts a ballot, there is no obvious victim to make a complaint and little public documentation to prove that a voter is not a citizen.

Trump and his supporters often point to a 2014 study by a team led by Jesse Richman of Old Dominion University. Using 2008 and 2010 data from the Cooperative Election Study, which requires respondents to opt in, Richman estimated that 6.4 percent of noncitizens voted in 2008 and 2.2 percent voted in 2010. Of those voters, 4 in 5 voted for Barack Obama, he said.

“Participation was large enough to plausibly account for Democratic victories in a few close elections,” Richman wrote in a summary of his report in The Washington Post in 2014. “Noncitizen votes could have given Senate Democrats the pivotal 60th vote needed to overcome filibusters in order to pass health-care reform and other Obama administration priorities in the 111th Congress.”

Richman’s findings immediately came under attack from other researchers, who said that the data set was too small to be useful and that some of it appeared to be misclassified. Richman stood by his report, but in 2017 he also said the Trump team misinterpreted his findings — “I can’t quite account for the math being so badly wrong in their analyses” — and that noncitizens could not have provided the popular-vote edge to Clinton.

Richman’s 2014 study continues to be cited. One researcher relied on its findings of noncitizen voting to calculate that more than 51,000 noncitizens voted in Arizona in 2020, theoretically giving Biden almost an additional 18,000 votes that tipped the state in his favor.

As an expert defense witness in the case overseen by Bolton, Richman was permitted to examine state voter files and a file from the Department of Motor Vehicles, cross-referencing the information. In her ruling, Bolton said, “the Court found Dr. Richman’s testimony credible and affords his opinions considerable weight.” He produced a detailed 123-page report and a short supplemental report, which Bolton placed under seal given the sensitivity of the data. The Fact Checker obtained a redacted version after submitting a public records request.

Richman said he found that 1,934 voters (0.43 percent of Arizona voting-age noncitizens) had records that indicated they were not citizens at the time they registered or after registering to vote. There are more than 4 million registered voters in Arizona, so the noncitizen segment would be about 0.04 percent.

Assuming a 50 percent turnout, which is typical in U.S. elections, that means fewer than 1,000 noncitizen votes out of 3.4 million cast in the 2020 election in Arizona. At the upper range — from less precise records also indicating noncitizenship — Richman said he located 6,480 voters, or 1.44 percent of voting-age noncitizens. With 50 percent turnout, that would mean about three-quarters of a percentage may have voted.

Both figures are substantially lower than the 6.4 percent of noncitizens in 2008 and 2 percent in 2010 that Richman estimated had voted in his widely cited 2014 paper.

“The records were matched on the driver’s license identification number which was a unique identifier present in both databases,” Richman told The Fact Checker in an email. “It’s possible, of course, that in some instances this number might have been entered wrong, so there is some possibility of a false match.”

“The evidence very clearly doesn’t document or demonstrate anything close to the claim that the Arizona Presidential election was decided by noncitizens in 2020,” he added.

As part of his report, Richman also examined 2022 data from the Cooperative Election Study, which now has a more precise way of asking a question about a voter’s citizenship and a larger sample size than the surveys he studied for his 2014 paper. He found that just under 1 percent of noncitizens indicated they had registered to vote — again, much smaller than his previous findings.

“These sources all align quite well in terms of suggesting that the likely rate of noncitizen registration or attempted registration nationwide is slightly less than one percent” and a voting participation of half a percent, Richman wrote. “I conclude that the incidence of noncitizen participation or attempted participation in U.S. and Arizona elections (through registration or voting) is low, but nonzero.”

Justin Levitt, a law professor at Loyola Marymount University who was skeptical of Richman’s earlier research, said in an email that “while the CES data here does look to me to be more reliable than Prof. Richman’s prior forays, I’d need some more information before I believed it were reliable.” He also said he would be curious to know how many of the noncitizens who registered in Arizona cast ballots, as turnout could be smaller than average.

The issue, Levitt said, is whether rooting out a relatively small number of noncitizen voters is worth the potential cost.

“For purposes of the policy question, for me, it’s always been a cost-benefit analysis,” he said. “I don’t think anyone’s against security measures that don’t harm any legitimate voters in the process. But if your security measure has an impact on legitimate voters far higher than any security effect on ineligible voters, it’s creating more problems than it’s solving.”

Bolton, in her ruling, said she did not find that the new Arizona laws were enacted with discriminatory intent. But she noted: “The Court finds that though it may occur, noncitizens voting in Arizona is quite rare, and noncitizen voter fraud in Arizona is rarer still.”

(About our rating scale)

Send us facts to check by filling out this form

Sign up for The Fact Checker weekly newsletter

The Fact Checker is a verified signatory to the International Fact-Checking Network code of principles