Do the subtraction the government has not done.
Homeland Security Secretary Markwayne Mullin told four states in July that DHS had identified over 250,000 noncitizens registered to vote across Pennsylvania, Nevada, California and New Jersey [1] [3].
Two of those states have published what DHS actually sent them.
Pennsylvania was given 14,576 potential noncitizens [1]. Nevada was given roughly 15,903 - of which, on DHS's own account, 8,576 matched immigration files on name, date of birth, address and Social Security number [2].
Add the two headline figures and you get 30,479.
For the four-state total to reach 250,000, California and New Jersey would have to supply about 219,521 between them. Neither number has been published.
Watch what happens to Nevada's figure when a stricter test is applied, because it is the only place we can see the machinery working.
15,903 suspected. 8,576 matched. Applying one filter cut the number by 46 percent [2].
Nobody outside DHS knows what a second filter would do, because DHS has not said what the first one was.
That is the actual dispute, and it is not partisan.
Nevada's Francisco Aguilar, a Democrat, asked on July 29 whether DHS used exact or approximate matching, whether it relied on artificial intelligence, and how it handled common names, shared addresses, data-entry errors and people who naturalised after the federal records were created [2].
He put the core objection plainly: 'We want to better understand your claims that thousands of alleged noncitizens on Nevada's voter rolls, again, none of which you have identified' [2].
A day later Pennsylvania's Al Schmidt - a Republican - asked for the same thing: 'please provide a document explaining your methodology, which identifies the data sources you relied on.' His letter states that '[n]otably, DHS has not identified its methodology or what data it relied on in reaching this conclusion' [1].
Schmidt also noted that SAVE has mistakenly flagged eligible voters as noncitizens in Texas and Missouri, and that the USCIS official who runs SAVE has acknowledged the system 'can't always find the most current citizenship information for people not born in the U.S.' [1]
Two things DHS has not done are worth stating precisely, because they are the difference between a finding and a headline.
It has not identified a single individual on the list. Nor has it alleged that any of them voted [1].
There is a real problem underneath this, and it deserves saying. Noncitizens on a voter roll, however few, is a genuine administrative failure, and states should fix it. New Jersey itself admitted a software glitch mistakenly added thousands.
The question is not whether accuracy matters. It is whether a number nobody can check, produced by a method nobody will describe, is the way to get it.
