250,000? DHS Math Doesn’t Add Up

The fight over alleged non‑citizen voters is less about proven illegal ballots and more about how unverified, often inflated numbers are being weaponized to justify criminal penalties and new barriers to voting.

At a Glance

DHS has publicized a headline figure of roughly 250,000 “potential non‑citizens” on voter rolls in four Democratic‑led states, but admits the numbers are preliminary and unverified.
Independent audits and multi‑state reviews consistently find actual non‑citizen voting to be vanishingly rare, with incidence rates measured in ten‑thousandths of a percent.
Minnesota AG Keith Ellison and other state officials argue the SAVE‑based approach criminalizes honest mistakes, intimidates election workers, and misreads imperfect voter data as mass fraud.
The pattern repeats past “ghost voter” scares: large, speculative claims collapse under scrutiny, while the political pressure they generate lingers and shapes policy debates.

From Big Numbers to Thin Evidence: What DHS Actually Found

The current controversy traces back to Homeland Security Secretary Markwayne Mullin’s assertion that a Department of Homeland Security review found “over 250,000” non‑citizens registered to vote in California, New Jersey, Nevada, and Pennsylvania. In letters to state officials and public briefings, Mullin broke that figure into specific state estimates—about 190,832 in California, 35,152 in New Jersey, 15,903 in Nevada, and 14,576 in Pennsylvania—framed as potential illegal registrations requiring urgent attention. The administration coupled these numbers with threats: election officials who fail to act, Mullin warned, could face fines, penalties, and even prison.

Those numbers appear dramatic until you examine what DHS has and has not shown. The department describes its work as a “preliminary review” of public voter files and commercial databases, not a confirmed count of non‑citizen voters. Its own letters to states use hedged language—“may be as many as”—and in at least one case distinguish between a larger speculative figure and a much smaller set of people who simply matched DHS records. A White House official told reporters the 250,000 figure was built from commercial data and broad matching, a method election experts say is prone to false positives and exaggeration. Even taken at face value, the 250,000 alleged registrations amount to around 0.16 percent of the roughly 154 million voters in the 2024 presidential election.

Crucially, neither Mullin nor President Trump has claimed—let alone demonstrated—that these individuals actually cast ballots. No methodology has been released that would allow independent analysts or the affected states to verify the names, citizenship status, or voting history behind the numbers. This gap between headline figures and verifiable facts is the fulcrum of the debate.

Ellison’s Critique: Criminalizing Error, Misreading Data

Into this vacuum stepped Minnesota Attorney General Keith Ellison, who has emerged as a leading critic of the administration’s approach. Ellison’s objections operate on two levels: the legal structure of the SAVE‑based regime, and the empirical record of how rarely non‑citizen voting actually occurs.

First, he argues that the Trump‑backed SAVE America Act and related DHS directives would expose election officials to felony liability for what are, in practice, inevitable data imperfections. When voter rolls are compared against large federal databases, mismatches happen for mundane reasons—name changes, naturalization not yet reflected in federal records, clerical errors. Under Mullin’s threats, Ellison contends, local officials could face up to five years in prison for failing to purge people whom flawed federal data wrongly labels non‑citizens. That, he says, flips the presumption of good‑faith administration into a presumption of criminality and risks chilling the recruitment of competent election workers.

Second, Ellison points to the numbers from actual state audits rather than speculative national tallies. In formal comments on DHS’s SAVE program, Minnesota’s Secretary of State reported an incidence rate of suspected non‑citizen voters of just 0.00028 percent—less than three in a million registrations. A multi‑state letter from California’s attorney general cited Georgia’s 2024 audit, which found only 20 non‑citizens among 8.2 million voters, and a 2016 analysis that identified about 30 suspected non‑citizen voting cases out of 23.5 million votes, a rate of roughly 0.0001 percent. These are not numbers associated with systemic fraud; they are statistical rounding errors in a sprawling electoral system.

On that evidence, Ellison’s core claim is straightforward: non‑citizen voting exists, but at a level too small to justify criminalizing administrative mistakes or erecting new documentation hurdles that would burden lawful voters more than they deter rare misconduct.

The Broader Record: Non‑Citizen Voting Is Real but Vanishingly Rare

Ellison’s argument sits atop a decade of independent research that converges on the same conclusion. Bipartisan and nonpartisan organizations—from the Center for Election Innovation and Research to the Brennan Center for Justice—have repeatedly reviewed allegations of large‑scale non‑citizen voting and found that they collapse under scrutiny. CEIR’s updated analysis notes that sweeping claims about “millions” or “hundreds of thousands” of illegal non‑citizen votes almost always stem from misunderstandings or mischaracterizations of complex voter data. When states actually audit their rolls, they uncover a small number of registrations by non‑citizens—often tied to confusion at DMVs or mistaken completion of forms—and even fewer actual ballots cast.

A national review released in 2025, tracking state investigations across the country, reaffirmed that non‑citizen voting “occasionally happens but in minuscule numbers, and not in any coordinated way.” Reuters’ examination of DHS enforcement activity likewise found investigations into more than 325 people across at least six states, but concluded that these cases did not support claims of rampant illegal voting. Congressional testimony summarizing multiple studies put the point bluntly: every credible study has found non‑citizen voting in federal elections to be “practically nil.”

In parallel, organizations like the American Immigration Council and Voting Rights Lab have dissected datasets often cited by fraud alarmists, showing how small absolute case counts are inflated through loose definitions or double‑counting. Their consistent finding mirrors Ellison’s: you can find instances of non‑citizen registration and even voting, but they are rare, typically inadvertent, and nowhere near the scale needed to alter election outcomes.

Why the Numbers Inflate: Imperfect Rolls and Over‑Aggressive Matching

To understand why the DHS figures diverge so sharply from audit results, you have to understand how voter rolls and federal databases actually work. Voter registration lists are large, dynamic, and inevitably imperfect; people move, change names, die, naturalize, and interact with different state agencies at different times. Federal citizenship databases are likewise incomplete snapshots, lagging real‑world changes and organized around identifiers that do not always align cleanly with state records.

When you perform broad, automated matches between these systems—especially using commercial data or partial identifiers—you generate substantial numbers of “possible” non‑citizens that, on closer inspection, turn out to be citizens. A DHS document dissected by independent analysts conceded that large portions of its own 28,000 non‑citizen figure for SAVE‑participating states were labeled “sample data” or not assigned to any specific jurisdiction. Election officials in Georgia and North Carolina told reporters that DHS tallies represented individuals flagged as possible non‑citizens, not confirmed ineligible voters; after manual review, Georgia found that only about 120 of 2,549 flagged individuals had ever voted.

This is the recurring “ghost voter” pattern: big, alarming numbers generated by rough matching, followed by careful audits that whittle them down to tiny fractions. The driving force is not malice so much as the mismatch between complex data and political appetite for simple, dramatic narratives.

From Data Disputes to Power Struggles: Federal Pressure on States

Where Ellison and many of his counterparts sharpen their critique is in the use of these inflated, unverified numbers to justify federal threats. Mullin has paired the 250,000 claim with warnings that election officials who fail to “scrub” their rolls using federal data could face prosecution and with proposals to condition federal election grants on adopting DHS‑preferred security measures. DOJ letters have demanded sensitive voter data from all 50 states, backed by lawsuits that federal courts have rejected 16 times out of 16.

State officials across the political map have bristled. Arizona’s Republican‑turned‑Democratic Secretary of State Adrian Fontes dismissed Mullin’s threat—colourfully—and detailed the extensive testing and auditing already in place in his state. Colorado Secretary of State Janet Griswold has publicly vowed not to “bend to intimidation,” pointing to thousands of personal threats she has received and the chilling effect of constant federal accusations on election workers’ safety and morale. These responses frame the conflict not just as a debate over data quality, but as a constitutional clash over who controls elections and how far the federal executive can go in pressuring state officials.

Ellison’s position aligns with that broader resistance. In his analysis of the SAVE Act and related directives, he argues they are less about fixing genuine vulnerabilities than about centralizing control and creating leverage over state election systems. By making continued funding contingent on embracing a flawed federal matching regime and by attaching criminal penalties to disagreements over data interpretation, he contends, the administration is attempting to nationalize election administration under the banner of fighting non‑citizen voting.

What the Evidence Supports—and What It Doesn’t

When you weigh both sides—the DHS assertions, Ellison’s critique, and the neutral research record—certain points are solid while others remain speculative or unsubstantiated. There is credible evidence that cross‑referencing voter rolls with federal and state databases occasionally uncovers non‑citizens who are registered and, more rarely, who vote. There is also clear statutory authority to punish intentional illegal voting, with criminal penalties and, for non‑citizens, immigration consequences.

What the evidence does not support is the leap from those modest case counts to claims of widespread, outcome‑shifting fraud. Multi‑state audits, academic studies, and bipartisan reviews all land on the same conclusion: non‑citizen voting is statistically negligible and has never been shown to alter the result of major U.S. elections. The DHS 250,000 figure, built on commercial data and described by the agency itself as preliminary and unverified, has not yet been backed by public documentation sufficient to overturn that consensus.

In that evidentiary landscape, Ellison’s warning looks proportionate. He is not denying that non‑citizen registration can occur; he is arguing that the remedy should be targeted cleanup and better data-sharing, not criminalization of honest mistakes or sweeping new documentation demands. For voters and election officials alike, the stakes lie less in the abstract number of “potential non‑citizens” than in how those numbers are used—as a reasoned basis for improving systems, or as a blunt instrument to intimidate and restrict.

Step four: use threats and intimidation to try to get your way. Friday, Secretary of Homeland Security Mullin threatened state election officials with possible prison time if they don’t comply with Trump administration efforts to determine whether noncitizens are on voting lists.

— Jane Doe’s Addiction (@Jane_Plain_Doe) July 19, 2026