The Department of Justice told California to set aside ballots from 20,930 people it labeled likely noncitizens. A court-ordered review of that same list, filed Saturday, found that 8,880 of those names, 42 percent of the total, correspond to no one in California’s current voter file at all. Applying the government’s own matching rules to the rest of the list, the reviewers found that nearly 7,500 of the remaining names were naturalized citizens who had been flagged as ineligible. At least 22 people on the list are confirmed US citizens, and the filing says many more probably are too.
The review was done by two data scientists from UCLA’s Voting Rights Project, submitted to the US District Court for the District of Columbia by the Coalition for Humane Immigrant Rights (CHIRLA) as part of its lawsuit against the Department of Homeland Security, as Al Jazeera reported. The lawsuit, CHIRLA and Trujillo v. DHS, was filed September 2 on behalf of CHIRLA and Pedro Trujillo, a naturalized California voter, and argues the government’s matching process is built on databases too unreliable to support removing anyone from a voter roll. Chief Judge James Boasberg ordered the DOJ on October 30 to hand the 20,930-name list to CHIRLA’s lawyers by 5pm that day; the analysis is what came out within 48 hours of getting it, according to US News.
What the DOJ actually sent
California mailed ballots to registered voters on October 1. On October 13 the department delivered its list on a password-protected thumb drive and gave the state 48 hours to act on it, demanding officials withhold or segregate the named ballots so they would not be counted. California refused, and the DOJ sued the state and Los Angeles County on October 15 to force the issue.
The UCLA researchers’ central complaint is about the list’s construction, not just its conclusions: the DOJ stripped voter-ID numbers from the file before handing it over, which the filing calls the single most important field for confirming a match between a name on a federal immigration database and a specific voter registration. Without it, matching falls back on names, and names collide. A county can share a John Garcia with a thousand other Californians; a federal match on name and rough birth year will catch plenty of the wrong ones.
That 42 percent error rate, on its own, means the list cannot be read as a count of ineligible voters without a second check against the rolls it was supposedly built from. The filing argues the mandated process endangers eligible California voters: “This review proves, at least as a preponderance of the evidence, that demands of election officials under criminal and civil threats, endangers the voting rights of plaintiff CHIRLA’s members and clients,” the filing says. It was co-signed by Xavier Becerra, the Democratic candidate for California governor and a senior adviser to the UCLA Voting Rights Project.
Bill Essayli, the US Attorney for the Central District of California who has led the department’s case there, rejected the findings on X: “This is a bogus report by the activist illegal immigrant rights group CHIRLA.” A DOJ spokesperson, Kiersten Pels, had earlier defended the broader effort on different grounds: “Every ballot cast by an ineligible voter cancels out the vote of an American citizen.” The department had resisted turning over the list in the first place, calling the request a fishing expedition before Judge Boasberg ordered disclosure.
The same pattern, state by state
California’s 190,832 figure is one slice of a number Homeland Security Secretary Markwayne Mullin announced in July: roughly 250,000 people the department said its preliminary review had identified as noncitizens registered to vote across California, New Jersey, Pennsylvania and Nevada, a total Mullin posted on X. The state breakdown was California 190,832, New Jersey 35,152, Nevada 15,903 and Pennsylvania 14,576. Neither Mullin nor DHS has said how many of the 250,000 actually cast a ballot, or over what period, a gap PolitiFact flagged when fact-checking the claim, noting the total is about 0.1 percent of the roughly 250 million registered voters nationwide.
Nevada’s experience with its slice of that list is the clearest stress test so far. DHS told Nevada it had identified 15,903 noncitizen registrants. A manual review by DHS officials itself, described on a recording of an August 13 meeting and in emails obtained by CNN, cut that to 185 confirmed matches. Nevada’s deputy attorney general then checked those 185 individually. The state’s count went to zero. Deputy Attorney General Greg Ott told the department it had “provided nothing of substance” to support its claim, as The Nevada Independent reported. A Homeland Security spokesperson disputed that the matter was closed, saying immigration officials had since manually confirmed roughly 2,000 more matches from the same underlying list, without detailing which list or how those matches were verified.
California’s own counties ran faster, rougher versions of the same check once pieces of the list leaked out, and reported, in the Al Jazeera account, “numerous errors.” Washington state, which is outside Mullin’s four-state tally but received a related demand to remove 400 names, told the department its list was no more reliable, citing at least one flagged voter who had been registered since the 1990s.
Where the law stands
The DOJ’s attempt to force California to segregate the 20,930 ballots has now lost twice at the Ninth Circuit. On October 30, a 2-1 panel rejected the department’s emergency bid, with Circuit Judge Patrick Bumatay dissenting. The majority held that the federal criminal ban on noncitizen voting does not give the executive branch a civil cause of action to rewrite a state’s ballot-challenge procedure, and that Congress had not used its Elections Clause authority to replace California’s process with the one the department wanted courts to impose by injunction four days before polls closed: “We decline the United States’s invitation to contravene the Elections Clause by undermining the Constitution’s plain text,” the court wrote, a line also carried by Al Jazeera’s report on the ruling. The panel weighed the competing interest directly: “There is a public interest in maintaining the integrity of elections. But... the risks that noncitizens are voting are, at best, uncertain.” That finding tracks a federal judge’s earlier ruling, covered in the decision striking down enforcement of the federal noncitizen-voting ban, and the scrutiny a separate court applied to the government’s underlying database in an earlier review of the purge list’s sourcing.
The department has signaled it will keep pressing the case, having lost, by one tally cited in reporting on the litigation, every one of its bids this cycle for unredacted state voter files and a separate Supreme Court attempt to curtail mail balloting nationwide. California’s results from the November 3 vote will post with the 20,930 disputed ballots segregated pending further court order, unless a higher court intervenes before counting finishes. The open question is not whether the DOJ’s list has an error rate; UCLA’s data scientists put a number on that, 42 percent for a plain non-match and a further 7,500 for a method the government chose itself. The open question is whether any court will let a list with that error rate touch a single counted ballot before the department restores the voter-ID field it stripped out.