Noncitizen Voting: What Republicans Claim vs. What the Data Shows

01What the audits actually found

The claim of millions of illegal votes runs into the audits. When states have systematically checked their rolls, the rate of noncitizen voting comes back at a tiny fraction of one percent, in the range of one in ten thousand to a few in a hundred thousand ballots.^[1] Georgia's audit found a few dozen potential noncitizen registrations out of millions. Even the Heritage Foundation, which maintains a database specifically to document voter fraud, lists only a small number of noncitizen-voting cases across decades and all 50 states.^[2] The numbers are real, they are prosecuted when found, and they are nowhere near the scale that could swing an election.

02Trump's own commission found nothing

The strongest test came from the proponents themselves. In 2017, President Trump created the Presidential Advisory Commission on Election Integrity, chaired in practice by Kris Kobach, with a mandate to document the fraud the president alleged had cost him the popular vote. It disbanded in January 2018 without producing evidence of widespread fraud.^[3] A commission built to find the fraud, given federal resources and a friendly mandate, came up empty. That is not silence from critics; it is a null result from the people most motivated to find a positive one.

03The day it went to court

The claim also had its day in a courtroom, and it lost. In Fish v. Kobach (2018), Kansas's proof-of-citizenship voter law went to trial, and Kobach, defending it personally, was asked to produce evidence of the noncitizen voting the law was meant to stop. The court found the evidence negligible, struck down the law, and the judge ordered Kobach to take remedial legal-education classes over his conduct of the case.^[4] Under the rules of evidence, with the burden on the proponent, the case for widespread noncitizen voting collapsed.

04The one study, and why it failed

Nearly every large claim traces back to a single 2014 study by Richman and colleagues that estimated 6.4 percent of noncitizens voted. Political scientists across the field, including the team that owned the survey data the study used, showed the result was a statistical artifact: a small number of citizens misclassifying themselves as noncitizens on a survey, which at the scale of a national sample produces a falsely high estimate.^[5] Corrected for that measurement error, the estimated noncitizen voting rate is statistically indistinguishable from zero. The headline number that launched a thousand claims does not survive its own data.

05The SAVE tool flags citizens too

The newest version of the claim points to the SAVE database as a way to catch noncitizen voters at scale. The tool was built to verify immigration status for federal benefits, not to vet voters, and using it for voter rolls produces false positives: naturalized citizens and others whose records are not current get flagged as noncitizens.^[6] A flag is not a finding. Purging voters on the basis of a tool with a known error rate removes eligible citizens, which is the opposite of the integrity the claim invokes.

06The weight of the evidence

Five independent lines all point one way. State audits find rates near zero. The proponents' own commission found nothing. A federal court rejected the case on the evidence. The one supporting study was a measurement error. And the database now offered as proof flags citizens as noncitizens. Noncitizen voting exists at the margins and is prosecuted when found. The claim that it happens by the millions and decides elections is false by every measure built to test it, including the ones built by the people making the claim.