The Washington PostDemocracy Dies in Darkness

Rebutting Trump’s claims about fraud in Arizona, point by point

A Trump supporter waves a flag outside the State Capitol Executive Tower on Dec. 14 in Phoenix. (Courtney Pedroza/for The Washington Post)

This article has been updated.

As predicted, Donald Trump and his allies have seized on the preliminary report from the partisan and dubious review of voting in Maricopa County, Ariz., as validation of his long-standing claims that the election was tainted by fraud. Candidates for office (hoping to appeal to Trump’s voting base) and former staffers have lined up to elevate not the top-line recognition that President Biden won more votes in the county but, instead, the report’s assertions that certain chunks of cast votes are somehow dubious.

The former president himself weighed in Friday morning.

“[T]here were enough fraudulent votes, mystery votes, and fake votes to change the outcome of the election 4 or 5 times over,” he said in a statement. After all, the margin in the state was just under 11,000 votes and, according to the audit, some 50,000 votes in Maricopa County were shunted into the “we’ve got some questions” category. This is the vindication Trump sought.

Except that, as we reported, the process itself, the report that resulted and the specific claims it makes should be treated with enormous skepticism. Not only are the figures Trump cites unproven, it’s generally true that they should not be taken at face value. What’s more, past claims made by Trump and his allies about voting in Arizona were quietly disproved — something that itself deserves elevation given what it reinforces about Trump’s indifference to accuracy in the arguments he makes.

The new claims — and why they deserve skepticism

We’ll use Trump’s assertions, made in his statement Friday, as our baseline for evaluation, since they’ll probably serve as guideposts for his supporters.

“The number includes 23,344 mail-in ballots, despite the person no longer living at that address.” One of the themes that you’ll see in Trump’s claims is his unwavering insistence that 1) the estimates provided by the team conducting the review are, in fact, ironclad and that 2) they necessarily represent fraud to President Biden’s benefit. Neither of those is either demonstrably true or even likely to be true. Trump lost the county by only about 2 percentage points; we might therefore assume that about half of any suspect votes were cast on his behalf.

This specific claim can be found at the start of Volume III of the draft report. (That’s the volume that actually delineates the claims. The preceding two “volumes” are fairly short.) In short, the final list of voters who cast mail-in ballots in 2020 was compared to a commercial database of people who moved between households. It found that about 15,000 people had moved elsewhere in Maricopa County before the registration deadline, an additional 1,700 who moved elsewhere in the state and about 6,600 who moved out of state.

What the review did not do is verify that the database results were correct, pointing out that “a small percentage of error is expected within these results.” It did not determine that those who might have moved were ineligible to vote. It acknowledged that “it may be possible for [the voter] to receive a ballot, for example, if they know the present occupant, or if the ballot is improperly forwarded” to them. In its report, the team leading the review, called Cyber Ninjas, recommended that the county do more research.

Here it’s useful to again go big picture. A pattern that often emerges with fraud claims is that big numbers are cited as dubious only for further investigation to reveal that the numbers are inaccurate or wildly overstated. What the Arizona review does here, in other words, is ask, not answer, a question. Which, of course, was the point of the review, to make sure that as many questions as possible were raised. But there is no reason for anyone, Trump included, to assume based on this report that there were 23,344 illegally cast votes.

“The official canvass does not even match who voted, off by 11,592—more than the entire Presidential Election margin.” Here we have Trump layering an inaccuracy on top of what’s probably an innocuous claim. The Cyber Ninjas’s report asserts that the file they were given documenting cast votes had a different total than the official results. The discrepancy was about 3,400 votes, nearly all of them cast before Election Day.

There’s no explanation of why that might be the case and, again, no suggestion that anything illegal occurred. Just: There’s this discrepancy.

You probably noticed that 3,400 is not 11,592. Trump is using a figure that includes “both the counted and uncounted ballots” — meaning that he’s identifying as suspicious an unexplained discrepancy of 8,000 uncast votes. This is not strong evidence of an election tainted by fraud.

“Voters who voted in multiple counties totaled 10,342, and 2,382 ballots came from people who no longer lived in Maricopa County.” Here Trump considers two separate claims.

In the first, he’s elevating a section of the review that found more than 10,000 instances in which people with the same first, middle and last names and the same birth years had cast ballots in more than one county. On its surface, this seems significant. But when I opened up Arizona’s voter file Friday morning (data collected by the firm L2), I quickly found that there were four John Smiths living in Arizona who are reported as being 43 years old. Two of them live in Maricopa County. This took me about 30 seconds.

Again we see the same pattern: elevation of something that could be considered questionable without actually resolving it. And, again, Trump seizing upon it as proof of his point.

As for the 2,382 ballots, this was essentially the in-person-voter version of the effort to match cast ballots against that commercial database of people who had moved. It’s not clear that these voters even moved, that they cast votes improperly, that they cast votes illegally or that they cast votes for Biden. Just another bit of doubt elevated in the review that Trump tacks onto his case.

“There were also 2,592 ‘more duplicate ballots than original ballots.’ ” This cryptic assertion, unexplained by Trump, is another discrepancy in metrics. In cases where a ballot cannot be read by an electronic voting machine, it is duplicated to be scanned. The county said it duplicated about 28,000 ballots, but the Cyber Ninjas counted about 27,000 ballots that were supposed to be duplicated and about 29,500 duplicates.

What does this mean? Good question. Maybe the Cyber Ninjas miscounted. Maybe the county did. Maybe something got lost in the shuffle. Maybe something else happened. There is no evidence presented that this indicates that too many votes were tallied or even that too many votes were duplicated. It’s just a number, offered among many other numbers, as part of a review initiated to raise questions about voting in the 2020 election.

What happened to the old claims, anyway?

Meanwhile, there were a lot of old claims about fraud (including by Trump) that were quietly set aside as part of the Cyber Ninjas’s review.

You may remember, for example, that in the immediate aftermath of the election, there were allegations that certain votes cast using markers were being disqualified by authorities. The idea was that bleed-through from the markers would make it impossible to scan the ballot correctly. This was quickly debunked, but the Cyber Ninjas review nonetheless suggested that “this could cause a contest to be voted that had not been voted, a vote could be canceled, or it could cause an overvote situation.” Oh, but: “No images that were reviewed met these conditions.”

In July, Trump elevated another claim that had been made by Cyber Ninjas’s team. There were some 74,000 votes, he said, that were received but never mailed. Except that there weren’t. That claim, made at a hearing by the head of Cyber Ninjas, Doug Logan, was quickly debunked: The file being used to make that determination was known not to be a tally of all returned votes. In the draft report obtained by The Washington Post, this gets shrugged away as “unintentionally misleading.”

It’s worth remembering that Cyber Ninjas had never done this sort of work before. Its expertise, instead, was in digital investigations; hence Cyber Ninjas. A large chunk of the report is centered on that sort of forensic work, resulting in a lot of the same sort of “here are a lot of questions we had” analyses. Robert Graham, a technologist and author who has been tracking such claims for some time, offered an analysis of the digital findings on his own website.

Perhaps the most succinct description?

“They are overstretching themselves to find dirt,” Graham writes, “claiming the things they don’t understand are evidence of something bad.”

That is certainly a fair assessment of much of the rest of the report. But for Trump and his allies, that they found things that even resemble dirt, even if only when held in the right light and while squinting, is good enough.

Update: On Friday afternoon, the county weighed in on the Cyber Ninjas’ report. Here is the rebuttal on the 23,344 mail-in ballots and on the votes in multiple counties.

Rosalind Helderman contributed to this report.