In the latest round of election fights, the most important fact is not the size of the numbers being thrown around, but the gulf between speculative federal claims and what careful state audits and decades of research actually show about non‑citizen voting.
Key Points
- DHS under Secretary Markwayne Mullin is touting a quarter‑million “potential non‑citizen” voter registrations in four Democratic‑led states, based on unverified data matching.
- Those figures are preliminary, represent well under 1% of the electorate, and DHS has released no methodology or proof that these individuals ever cast ballots.
- Independent audits and academic studies repeatedly find non‑citizen voting to be vanishingly rare, with incidence rates in the hundred‑thousandths of a percent and no impact on election outcomes.
- Minnesota AG Keith Ellison and other state officials argue that new federal proposals like the SAVE Act would criminalize routine election errors and erect substantial barriers to lawful voting.
- The core dispute is not whether voter rolls are imperfect—they are—but whether those imperfections justify sweeping new criminal penalties and federal control over state‑run elections.
The Trump–Mullin Numbers: What DHS Is Claiming
When President Trump and Homeland Security Secretary Markwayne Mullin talk about “over 250,000 non‑citizens” on voter rolls, they are drawing on two separate, largely undisclosed exercises in data matching. One sheet released alongside Trump’s nationally televised election‑security address lists “over 250,000” supposed non‑citizen registrants in California, New Jersey, Nevada, and Pennsylvania—states that, by DHS’s own description, had not been using the federal SAVE database for voter checks. A second figure, “over 28,000” non‑citizens, is said to come from 25 “proactive SAVE user” states that ran some 68 million records through an “enhanced” federal system.
The public numbers are detailed enough to sound precise: 190,832 in California, 35,152 in New Jersey, 15,903 in Nevada, 14,576 in Pennsylvania. They are large enough, in isolation, to alarm a lay audience. What they are not is verified. DHS’s letters themselves hedge, stating that there “may be as many as” those numbers of potential non‑citizens and acknowledging the figures are based on preliminary review of public voter files and commercial databases. A White House official later confirmed that the four‑state estimate arose from commercial database matching, a technique election‑administration specialists stress is highly prone to false positives.
The same pattern appears in the SAVE‑based claims. Mullin has told audiences that 23 states working with DHS found an additional 28,000 non‑citizens and 400,000 deceased individuals on their rolls, but in the underlying documentation only about 10,716 “non‑citizens” can be tied to any named states, and those are explicitly labelled “sample data.” Fifteen of the 25 states are not identified at all, and DHS has not provided state officials with person‑level records they can test against their own, more current data.
How Strong Is the Evidence Behind the DHS Tallies?
Once you look past the rhetoric, the evidentiary foundation for the DHS numbers is thin. Neither Trump nor Mullin has claimed—let alone demonstrated—that all or even most of the flagged registrants actually voted. CBS News notes bluntly that “neither Mullin nor Mr. Trump said that the 250,000 noncitizens who are allegedly registered to vote in the four states actually cast ballots,” and that the administration has not made public its methodology. A CNN review of the Pennsylvania letter highlights that DHS itself distinguishes between the larger speculative figure and a smaller set of “matches” in its own immigration files; even there, the agency concedes the list is unverified and under investigation.
State officials who have seen similar federal lists describe what happens when they test them. In Georgia, running a DHS‑supplied list of 2,549 “possible non‑citizens” through state systems produced only about 120 individuals who had ever voted at all. Minnesota’s Secretary of State, in formal comments on the SAVE program, reported an incidence rate of suspected non‑citizen voters of just 0.00028% after audits—barely a handful in a statewide electorate. These are not numbers of systemic fraud; they are the kind of fringe anomalies you see in any large data system.
Experts who study registration data and database matching are unsurprised. David Becker, who leads the Center for Election Innovation and Research and has worked directly with states on roll maintenance, told CBS that using commercial databases to infer citizenship is “likely to lead to false positives, significantly overestimating the number of potential noncitizens on voter rolls.” A broader review by CEIR concludes that sweeping allegations of large numbers of non‑citizen registrants almost always arise from “misunderstandings, mischaracterizations, or outright fabrications about complex voter data.” That pattern is remarkably consistent over time.
What Decades of Audits Actually Show About Non‑Citizen Voting
To understand why many election lawyers and social scientists view the current DHS claims skeptically, you have to look at the record of past investigations. In jurisdictions with high non‑citizen populations, the Brennan Center found just 30 suspected non‑citizen votes across 23.5 million ballots cast in the 2016 election—a rate of roughly 0.0001%. A multi‑state audit cited by the California Attorney General similarly found 20 non‑citizens out of 8.2 million voters in Georgia’s 2024 election. These are vanishingly small fractions.
Independent reviews across administrations tell the same story. A report tracking state inquiries into non‑citizen voting, released in 2025, concluded that “noncitizen voting occasionally happens but in minuscule numbers, and not in any coordinated way,” and that “the vast majority of allegations of noncitizen registration or voting” dissolve once states examine the individual records behind the headline numbers. The Bipartisan Policy Center, after surveying available studies and enforcement actions, found “no evidence that noncitizen voting has ever been significant enough to impact an election’s outcome.”
This is why the phrase “vanishingly rare” appears repeatedly in the literature. Federal law has prohibited non‑citizens from voting in federal elections since 1996; registration forms require applicants to attest under penalty of perjury that they are citizens, and state officials use driver records, vital statistics, and targeted checks to keep rolls accurate. Imperfections remain—people move, die, adjust their status—but when states conduct forensic audits of alleged non‑citizen voting, the result is consistently the same: negligible incidence, no organized scheme, and no measurable effect on outcomes.
Keith Ellison, the SAVE Act, and the Stakes for Election Administration
Minnesota Attorney General Keith Ellison enters this debate not as a neutral observer but as a leading critic of the Trump administration’s proposed response. In his analysis of the SAVE Act and related executive orders, Ellison argues that these measures would “create substantial barriers to voting” by transforming ordinary administrative mistakes into criminal offenses subject to five‑year prison terms. Where Trump and Mullin talk about “getting ’em all”—their phrase for purging alleged non‑citizens from the rolls and prosecuting officials who fail to do so—Ellison’s concern is who gets swept up in that dragnet.
To see why, consider how the SAVE program works. SAVE is an immigration status database designed primarily for benefit adjudications, not election administration. It does not contain every citizen, it can lag behind naturalization events, and it can misclassify lawful residents or citizens whose records are incomplete. When states have used SAVE or similar federal data to “clean” their rolls, they have at times flagged citizens as non‑citizens simply because their status was not reflected in the federal system. Ellison and Minnesota’s elections officials warn that elevating SAVE to a mandatory gatekeeper—and attaching criminal liability to any mismatch—would inevitably deny eligible voters their rights while chilling the work of already under‑resourced local officials.
From that vantage point, DHS’s current posture looks less like neutral law enforcement and more like a pressure campaign. Mullin has paired his non‑citizen figures with threats to withhold election‑related grants from states that do not adopt federal verification measures and warnings that officials who fail to act on DHS information “can be held accountable by fines, penalties, and even … prison time.” Ellison’s opposition brief frames these moves as a federal attempt to centralize control over voting systems that the Constitution assigns primarily to the states, justified by numbers that collapse upon audit.
State Officials Push Back: Methodology, Federal Overreach, and Public Trust
Ellison is not alone. Secretaries of state in both Democratic and Republican‑leaning states have criticized the current campaign as evidence‑light and corrosive to election administration. Nevada’s Francisco Aguilar dismissed DHS’s four‑state numbers as “wildly speculative at best” and noted that his office had received no underlying data to evaluate. Colorado’s Janet Griswold has highlighted that DOJ is 0‑for‑16 in lawsuits seeking sensitive voter data from states, suggesting courts see little legal basis for the federal fishing expeditions.[Face the Nation summary]
Election‑security veterans worry about the practical effects. Former CISA director Chris Krebs, who led federal–state coordination in 2020, has repeatedly emphasized that U.S. elections were secure that year because of robust partnerships and layered safeguards, not because of punitive federal threats.[Face the Nation summary] When top federal officials publicly talk about jailing local clerks and county staff on the basis of unverified lists, it becomes harder to recruit and retain the very people who administer elections; some, like Griswold, report thousands of personal threats already.[Face the Nation and MS NOW summaries]
There is also a public‑trust dimension. Claims that “hundreds of thousands” of non‑citizens and deceased individuals are “listed and active on the voter rolls” play into a narrative of rampant fraud, even when the underlying data are explicitly preliminary and likely overinclusive. Once amplified in speeches, social media, and sympathetic outlets, such claims are rarely revised downward in the public imagination, even if later audits find only a handful of questionable cases. Ellison’s warnings about “competitive authoritarianism” aim at this dynamic: use inflated, unproven numbers to justify tightening control over who can vote, then defend those restrictions as necessary to prevent the very fraud the numbers were used to conjure.[MS NOW summary]
‘They are desperate’: Stacey Abrams slams Mullin over threats against state election officials – MS NOW https://t.co/XLkGyQleEr
— Geezus1.0 (@0Geezus1) July 19, 2026
How to Read Future Claims About “Non‑Citizen Voters”
For citizens trying to make sense of these fights, a few heuristics help. First, distinguish between registrations and votes. Registrations are data records, subject to clerical error, timing lags, and imperfect database matches; votes are acts that generate a paper trail and can be tracked. Many of the most dramatic federal claims concern the former, not the latter, and even then they rest on speculative matching exercises rather than person‑level verification.
Second, look for the denominator. A quarter‑million flagged records sounds enormous until you note it represents roughly 0.16% of the 154 million ballots cast in the last presidential election. When independent audits find incidence rates on the order of 0.0001%–0.00028%, the debate is no longer about “stolen” elections but about how much friction we are prepared to impose on tens of millions of lawful voters to chase anomalies that, so far, have never altered an outcome.
Third, follow the methodology. Are the numbers tied to a transparent process, with published criteria and an opportunity for states to test the lists against their own records? Or do they emerge from black‑box commercial databases and partial tables labelled “sample data”? The latter demand more skepticism, not less, especially when they arrive packaged with threats of prosecution.
Ellison’s core contention, supported by a long line of empirical studies, is not that non‑citizen voting never happens. It is that it happens so rarely, and is already illegal and prosecutable when discovered, that it does not warrant remaking election law in ways that criminalize honest mistakes and deter eligible citizens from participating. On the evidence we have—state audits, scholarly reviews, and the federal government’s own enforcement record—that remains the more grounded reading of the numbers.
Sources:
redstate.com, wral.com, justthenews.com, youtube.com, crystalclearnews.com, facebook.com, abcnews.com, ag.state.mn.us, oag.ca.gov, nytimes.com, americanimmigrationcouncil.org, bbc.com, votingrightslab.org
