EXCLUSIVE: The ‘farcical’ stats Republicans use to claim millions voted illegally

The studies were based on surveys of as few as 14 people.

President-elect Donald Trump pauses pose for photographs as he greets Kansas Secretary of State, Kris Kobach. CREDIT: AP Photo/Carolyn Kaster
President-elect Donald Trump pauses pose for photographs as he greets Kansas Secretary of State, Kris Kobach. CREDIT: AP Photo/Carolyn Kaster

A previously unpublished document obtained by ThinkProgress further undermines unsubstantiated claims by the Trump administration and Kansas Secretary of State Kris Kobach that millions of non-citizens illegally vote in U.S. elections.

The documents reveal that claims of widespread illegal voting by noncitizens are based on samples of as few as 14 people.

In response to interviews about the newly revealed documents, its author has admitted his work does not support Trump’s claim that millions of people voted illegally.

Now, in an interview with ThinkProgress, even Kobach himself is wavering on the claim.

President Donald Trump claimed he lost the popular vote to former Secretary of State Hillary Clinton by nearly three million votes because of “the millions of people who voted illegally.” He even reportedly brought the issue up in a White House reception for congressional leaders just days after his inauguration.


“I suggest you invite Kris Kobach onto your show and he can walk you through some of the evidence of voter fraud in greater detail,” White House senior policy advisor Stephen Miller told ABC’s George Stephanopoulos in February.

The evidence Miller, Kobach, and other Trump surrogates refer to is two reports on noncitizen voting: A 2017 statistical analysis that focuses on Kansas — obtained by ThinkProgress and published here for the first time — and a 2014 national study published in the journal Electoral Studies. Both are by Jesse Richman, a professor of political science at Old Dominion University in Norfolk, Va.

But this evidence is deeply flawed. It’s likely to show accidental registrations, misunderstandings, response errors, or situations where a person registers to vote shortly before becoming naturalized, experts and advocates say.

‘[B]ordering on the absurd’

The statistical analysis in particular is riddled with small and unrepresentative samples, response errors, and other issues that undermine its legitimacy, according to Brian Schaffner, a professor of political science at the University of Massachusetts Amherst.


“Frankly, the so-called analysis is bordering on the absurd and Richman is doubling down on the same mistakes he made in his now discredited 2014 study of noncitizen voting,” Schaffner said in an email. “His estimates of noncitizen registrants in Kansas are based on literally a handful of respondents in a survey.”

The analysis gives six separate estimates, from zero noncitizen voters up to 32,000, that are based on different methodologies and data sources. One estimate is based on four potential noncitizen voters found in a sample of 14 possible noncitizens — a tiny sample from which Richman extrapolated 32,000 noncitizen voters in Kansas. Another estimate, which Kobach references often, is based on six potential noncitizen voters found in a sample of 37 noncitizens, from which Richman estimates that there are “more than 18,000” noncitizen voters across the state.

Kobach, a Trump surrogate who was considered a frontrunner for Department of Homeland Security secretary, commissioned the analysis as part of his defense in two court cases challenging a 2011 Kansas law that requires residents to provide proof of citizenship in order to complete their voter registration. Since then, he has used it to justify his campaign against voter fraud.

“A statistical analysis of the voter rolls by an academic expert at Old Dominion University determined that the total number of noncitizens on the Kansas voter rolls could be more than 18,000,” Kobach told Kansas legislators in February. This refers to an estimate from Richman’s analysis that was based on six potential noncitizen voters out of a sample of 37 noncitizens.

The next week, Kobach used Richman’s 2014 study to make an even more implausible statistical leap. “That is a projection, but if you took that projection and applied it to the numbers today, you’d be talking about in excess of 3 million aliens voting,” he told Fox Business Network’s Neil Cavuto.

Kobach himself starts to have doubts

In an interview with ThinkProgress on Wednesday, Kobach clarified that his 3 million claim was based solely on Richman’s 2014 study, not the statistical analysis. That paper, based on data from 2008 and 2010, estimated that between 38,000 and 2.8 million noncitizens could have voted in 2008.


But that paper has come under heavy fire from many political scientists. Last January, Schaffner and 190 other political scientists signed an open letter decrying Richman’s research. The numbers Richman found, they said, were largely due to reporting error — a relatively small number of survey respondents who incorrectly listed themselves as noncitizens but correctly reported their voting habits.

“The scholarly political science community has generally rejected the findings in the Richman et al. study,” the letter concluded, “and we believe it should not be cited or used in any debate over fraudulent voting.”

Richman stands by his 2014 study, but even he questions Kobach’s reading of it.

“I don’t think there’s much support at all for the notion that we’re talking about several million,” he said. “But whether we’re talking about a few thousand or hundreds of thousands, I don’t know. And that’s important, in terms of what end of that range we’re on.”

Kobach defended his use of the 3 million figure, saying it depended on what percentages and numbers one uses from Richman’s 2014 study. Pressed on why he uses the highest range of Richman’s confidence interval when making public statements, Kobach appeared to equivocate.

“I think I said it could have been as high as” 3 million, Kobach said, “so I try to be careful.”

That statement contradicts what Kobach told Neil Cavuto on Fox Business Network last February. “I think it’s probably in excess of a million,” Kobach said of noncitizen voting. “I think it’s in excess of a million if you take the entire country, for sure.”

While admitting that many of his estimates have “weak statistical power,” Richman defends his overall conclusions. “If the false narrative that Professor Shaffner has tried to peddle — that no non-citizens register or vote in U.S. elections — was true, then more of these numbers would be zero,” he said.

‘[A]lmost farcical’

Almost any evidence of noncitizens voting is significant, Richman’s analysis claims, because of the small margins of victory that can sometimes occur in state and local elections.

The analysis only provides a margin of error for one of its estimates, which Richman now says “wasn’t a good decision.”

“I didn’t feel like having someone say, ‘Richman said the rate of non-citizen registration was 54 percent’ by ripping the meaning of the confidence interval completely out of context,” he explained.

In an email, Richman provided a table of confidence intervals for five of the six estimates in his statistical analysis:

Kobach declined to discuss the statistical analysis in detail, citing the ongoing litigation. But he stood by Richman and his work. “Richman’s analysis, I think, is very good, both in our case and in his 2014 article,” he said.

Kobach also cites as evidence a list of 115 noncitizen Kansas voters he says his office has compiled. Thirty-two of those names come from a list of noncitizens who either registered or attempted to register in Sedgwick County, Kansas, according to Kobach’s Kansas Senate testimony.

Richman based one of his estimates on the Sedgwick County list. Kobach released a version of it as part of his Senate testimony. Election officials in Sedgwick County also provided ThinkProgress with a more up-to-date copy in February.

Of the 32 people on that list, 14 successfully registered before they were citizens. It’s unclear how many registered by accident while signing up for a driver’s license, for example, or mistakenly believed they could vote as legal permanent residents.

Five actually voted in at least one federal, state, or local election between 2004 and 2014, and one was prosecuted.

“[S]everal self-reported,” Sedgwick County Election Commissioner Tabitha Lehman said in an email. “Some responded to calls or notices from our staff, but most were discovered through the registration process at local naturalization ceremonies.”

Micah Kubic, executive director of the American Civil Liberties Union of Kansas, called the list “almost farcical” and said it appears most of the people on it registered by accident.

“[Kobach] had to go back years and years and years in order to assemble the list,” Kubic said in an email. “No one else has verified that the folks he identified in that list were actually non-citizens, and even then his own descriptions of them appear to be largely folks who made mistakes (checked a box by accident, etc.) and never actually tried to vote.”

Asked for the full list of 115 names, a spokesperson for Kobach told ThinkProgress to file a records request. On Friday, a federal judge in one of the two Kansas voting rights cases ordered Kobach to hand over a draft of a possible future amendment to the National Voter Registration Act, or NVRA, and a document Kobach was photographed with while entering a meeting with then President-elect Trump in November that references a possible NVRA amendment.

ThinkProgress has requested all of these documents from the Kansas Secretary of State’s Office under the state’s Open Records Act.

Read the full statistical analysis here.