About 75,000 Republican-leaning voters in Arizona's two most populous counties did not vote to re-elect President Donald Trump in the 2020 election, according to an analysis of every vote cast by a longtime Arizona Republican Party election observer and election technologists familiar with vote-counting data.
The analysis from Maricopa and Pima Counties underscored that the Arizona state Senate's ongoing audit of 2.1 million ballots from Maricopa County's November 2020 election was based on a false premise—that Democrats stole Arizona's election where Trump lost statewide to Joe Biden by 10,457 votes.
"I am continuing my analysis of why Trump lost in Arizona," Benny White, a former military and commercial pilot who has been a Republican election observer for years in Pima County and was part of the research team, said in a May 10 Facebook post. "Bottom line: Republicans and non-partisans who voted for other Republicans on the ballot did not vote for Trump, some voted for Biden and some simply did not cast an effective vote for President."
The analysis, whose methodology is similar to academic research by political scientists, offers a counternarrative to Trump's continuing claims that he lost a rigged election. It also underscores that election experts can extract records from voting systems to affirm and explain the results, such as showing that at least 75,000 Arizonans voted for many other GOP candidates but not for Trump.
"The data is all there to form a justified belief that there wasn't anything amiss, and you should be looking at that [data] before you turn ballots over to partisan third parties," said Larry Moore, who founded Clear Ballot, a federally certified firm that helps local and state governments to count and verify election results, and helped White analyze to fall 2020's vote patterns from the two counties.
"This needs to be treated like a giant accounting problem where everything has to add up," Moore said. "We have been working on this nonstop for days. The [state] Senate's auditors don't know what they are doing… The county election officials and their attorneys also don't realize the power of the [data] tools that they have."
The analysis was based on the "cast-vote record" of every vote on every ballot in the two counties, which White obtained in a public records request and analyzed. The state Senate's auditors, led by the pro-Trump contractor Cyber Ninjas, were given the same data in February, but have not used it to cross-reference the subtotals in their hand count of Maricopa County's presidential and U.S. Senate votes, audit officials told Voting Booth. Cyber Ninjas has not yet issued any findings about several audits it is supervising.
"I want voters to decide the results of the election, not lawyers and judges, which is what is occurring in Maricopa County with this [Senate-led] audit that is extremely disruptive," said White in an interview. "It is really undermining the public's confidence in our election systems, and it's completely unnecessary."
Bryan Blehm, an Arizona attorney representing Cyber Ninjas, replied to White's post on Facebook—without identifying that relationship—by saying that White was not working with reliable data and was angling for a job with Arizona's Secretary of State Katie Hobbs, a Democrat.
"Of course data facts matter," Blehm wrote on May 10. "That is why Mr. White relies on data supplied to him by government buearocrats [sic] rather than the actual real data. Hence, he questions anyone actually working with the underlying real data. I think Mr. White is pushing for a job with the Secretary of State."
However, a handful of political scientists who study voter turnout confirmed that using cast-vote records to analyze voting patterns, including voters who split their votes between major party candidates, was a standard research methodology.
"Yes, political scientists have done research using cast-vote records," said Charles Stewart III, who directs the Caltech/MIT Voting Technology Project. "Last year, I published a co-authored article that looked at the 2016 election, and we concluded that Republicans were much more likely to abstain in that election than Democrats—and the Republicans who did abstain had been anti-Trump in the primary."
"This approach is similar to research we have done," said Matthew Thornburg, a University of South Carolina Aiken assistant professor of political science. "What we find in political science research is that voters are more likely to defect in races they know more about. In presidential races, everyone knows the candidates well by November and can be persuaded by factors other than partisanship."
"Given that you have the data, you can make all kinds of analyses," said Duncan Buell, chair emeritus of the Computer Science and Engineering Department at the University of South Carolina, who has analyzed public election records in a half-dozen states. "This is not rocket science, and it is not partisan."
A Political Science Analysis
The approach that White and Moore used echoed what political scientists do when analyzing split-ticket voting patterns (when voters diverge between different parties' candidates as they fill out a ballot) or partisan voting patterns based on a precinct's demographics.
To start, White obtained the cast-vote record from the counties, both of which saw more voters support Biden than Trump in the 2020 election. This public document is a series of elaborate computer files that contain every vote cast in every race. Those records are organized as individual folders each containing batches of several hundred ballots. Maricopa's data was in 10,300 folders, White said.
White reached out to Moore, who enlisted Tim Halvorsen, Clear Ballot's former CTO. The Boston-based firm's expertise is based on analyzing digital images of every scanned paper ballot to double-check election results. After obtaining the cast-vote records, the researchers had to identify Republican-leaning voters.
General election ballots don't identify voters or list their party affiliation. White's team noted that there were 15 contests with Republican candidates for county or higher offices in Maricopa County in the 2020 election. There were 13 such contests in Pima County.
To identify Republican-inclined voters, Halvorsen created a search tool to identify the ballots where half or more of the votes in these contests were for Republicans. That meant at least eight votes for Republican candidates in Maricopa County and seven Republicans in Pima County.
The search tool also identified how many ballots contained a majority of votes for Republicans—but not for Trump. It found about 60,000 such ballots in Maricopa County and slightly more than 15,000 ballots in Pima County. White said that he needed an experienced voting system programmer to help process the data.
"I don't want to trivialize this analysis because it is very difficult," White said. "You have to have knowledge of the election administration process. You have to have knowledge of the way voting machines work. You have to have knowledge of what might be available to you in all of the public records… It takes actual expertise to be able to do that."
The finding that some number of Republican voters were turned off by Trump and did not vote for him in 2020's general election is not unique.
Michael McDonald, a University of Florida political science professor who tracks voter turnout patterns nationally, said "it seems consistent with what we've seen elsewhere concerning the suburban shift toward Biden."
"The upshot is that voters defect more the higher up the ballot the race is and the more information they have," said Thornburg, citing published research about this pattern. "This result does not surprise me."
"Fifty-nine thousand votes in Maricopa County amounts to only approximately 2.8 percent of the votes that were cast there," he continued. "Assuming (generously) that loyal Republicans made up just 45 percent of Maricopa's 2020 voters, that's only about 6.3 percent of loyal Republicans (which is in line with national exit poll results that show approximately 6 percent of Republicans and 5 percent of Democrats voted for the other party's presidential candidate)."
There was also a drop-off in Republican voter turnout in Georgia between its November 2020 election, which Trump also lost, and the turnout in early January's U.S. Senate runoffs, which Trump repeatedly said would be fraudulent and where the Democrats prevailed—returning the majority to Democrats. Trump's rhetoric has been seen as suppressing his party's turnout in the Senate runoffs.
Arizona Investigation Continues
The investigation by White and Moore was also using other public records to debunk another conspiratorial claim about Maricopa County's election: that 40,000 ballots were smuggled into vote-counting centers after midnight on November 4.
Using Arizona's public voter history file and eligible voter file, White found there were no unusual spikes in precinct-level turnout patterns (Maricopa County's turnout was 80.5 percent), Moore said. White also verified the identities of all but 720 voters out of the 2.1 million people who voted in the 2020 election in Maricopa, Moore said, adding the exceptions were people whose identities were protected as crime victims, law enforcement officers or public officials like judges.
"It completely checked out—all 2.1 million voters," Moore said. "There were no unknown names except for those 720… And the tool we used to explain all this [assertion] was mapping. We show by precinct the percentage turnout and the actual numbers of turnout. There's no [conspiratorial] there, there."
On May 17, Blehm, Cyber Ninjas' attorney, also commented on Facebook, again criticizing White for this line of inquiry. "Pretty map," he wrote. "And it shows you are up on the data they give you. So much for reality because you apparently only need what they feed you."
Ken Bennett, a Republican and former Arizona secretary of state who is serving as a liaison for the Senate's audit of Maricopa County's 2020 election, declined to comment on the research by White and Moore. Previously, Bennett has said that he hopes to oversee several audit procedures to address the persistent belief among Trump supporters that Arizona's 2020 presidential election was dishonest.
"The power of this [cast-vote record] analysis is dealing with the complete record of all votes and not just statistical estimates," Moore said. "This is not based on estimates. There are no confidence intervals. These numbers are based on 100 percent of all the voters voting."
Steven Rosenfeld is the editor and chief correspondent of Voting Booth, a project of the Independent Media Institute. He has reported for National Public Radio, Marketplace, and Christian Science Monitor Radio, as well as a wide range of progressive publications including Salon, AlterNet, the American Prospect, and many others.