DHS Says 250,000 Noncitizens Are on Four States' Voter Rolls. Its Own Letters to Those States Say 118,003.

The Department of Homeland Security told the country on July 17 that it had found more than 250,000 noncitizens illegally registered to vote in just four states. Two weeks later, the four states named in that announcement — California, Nevada, New Jersey and Pennsylvania — say the federal government still has not told them how it arrived at the figure, and the letters DHS actually sent to election officials describe something considerably smaller and far more tentative than the public fact sheet did.

The gap is not a rounding error. The DHS fact sheet stated flatly that "over 250,000 noncitizens are illegally registered to vote" across the four states. The letters that Homeland Security Secretary Markwayne Mullin sent to the same states characterized the work as a "preliminary review," described the per-state counts as potential rather than confirmed, and identified a much smaller subset — 118,003 people across all four states combined — whose name, date of birth, address and Social Security number matched a noncitizen record held by the federal government. A match on those four fields is not proof that a registered voter is a noncitizen, and DHS has not said what it did to rule out false positives.

State election officials want to know one thing above all: where the Social Security numbers came from. Publicly available voter rolls do not contain them. Nevada Secretary of State Cisco Aguilar reduced the problem to two questions — "How did you get it? And is it reliable?" California Secretary of State Shirley Weber said she had serious doubts about the reliability of the data. Pennsylvania and Nevada both formally demanded that DHS disclose its methodology. The department set a response deadline of Thursday, July 30, without ever answering that question.

Election experts suspect two possible sources, and neither is reassuring. One is commercial data brokers, whose records are assembled for marketing rather than for adjudicating citizenship. The other is the SAVE database, a federal immigration-status system that a federal judge halted in June after it produced inaccurate results. David Becker, who runs the Center for Election Innovation & Research, was blunt about what the public should do with the number in the meantime: "These numbers, until backed up by clear, transparent methodology, should be considered made up."

Mullin has paired the claim with pressure. He has promised "maximum pressure" on the states and has raised the prospect of prison time for election officials who do not cooperate with his department's requests. That is landing at a delicate moment on the calendar. Federal law imposes a 90-day quiet period before an election during which systematic voter-roll purges are barred, and this cycle that window opens on August 5 — meaning any removals prompted by the DHS list would have to happen in the next several days or wait until after the midterms.

The underlying question of how often noncitizens actually vote has been studied repeatedly, in red states and blue ones, by academics and by state officials. Those reviews have consistently found the practice to be extremely rare — typically a handful of cases out of millions of ballots, most of them traceable to registration errors rather than intent. What worries the officials on the receiving end of Mullin's letters is not the prospect of a purge before November so much as what a disputed, unexplained federal number could be used to argue after the votes are counted.