Voter Discrepancies Found In The Arizona 2022 General Election

Introduction

In recent years, several state officials have raised concerns that some ballots could have been counted multiple times in their elections, resulting in more ballots counted than registered voters who voted. Others have raised the question of whether ballots could have been destroyed, resulting in more registered voters who voted than ballots counted.[1] Following these concerns, the America First Policy Institute (AFPI) endeavored to conduct a study with a simple goal: to match the total number of registered voters listed as voting with the total number of ballots counted in the Arizona 2022 general election. Doing so would enable us to see if the vote totals were consistent or if they revealed discrepancies.

The races for Arizona governor and Arizona attorney general were extremely tight, so we recognized that any discrepancies could have played a role in the outcome. In the attorney general race, just 280 votes divided the Democrat and Republican attorney general candidates (Snow (2023)).

We made no attempt to determine if any discrepancies were intentional or accidental. How and for whom voters voted also were not at issue. We were strictly looking to determine whether voters and vote totals were equal.

Background and Methodology

This study was conducted with a similar methodology employed in an AFPI analysis of the national 2020 general election that found vast discrepancies. AFPI repeated the study for the November 2022 general election. To answer our specific study question about potential voter discrepancies in Arizona, we focused only on the data for the four most populous Arizona counties—Maricopa, Pima, Pinal, and Yavapai—as well as on smaller populations in tribal communities, such as Apache and Coconino, where previous concerns about voting irregularities had been raised (Davidson (2020) and Smith (2020)).

To obtain records of voter data from the general election, AFPI asked county election officers in 100 counties across the U.S., including six in Arizona, for their official tabulations of total ballots counted in the election. We submitted public records requests to these counties to provide us with a list of all voters who voted in the November 8, 2022, general election, including the following fields:

• Voter ID
• Voter name
• Registration status (i.e., Active/Canceled/Other)
• Registered address (City, Zip Code, County)
• Precinct
• Voted in 2022 GE (Yes/No)

The goal of these public records requests was to match up the total number of registered voters listed as casting ballots (RVBC) and the total number of ballots counted (TBC) by precinct in each county (the national report will be published in the summer of 2023). Total ballots counted include all ballots: absentee, mail-in, and in-person.

In theory, data should have been easy to retrieve upon request because the federal Civil Rights Act of 1960 requires the retention and preservation of all election records and papers for 22 months, unaltered (Appendix B). However, in our 2020 national election report, only 6% of county election officials and two secretaries of state for the states and counties provided data. In the counties that provided data, we found an average 2.89% discrepancy rate—where either there were more registered voters listed as voting than there were ballots counted or, in other precincts, the reverse.

Our national report showed that within days after an election, the counties tended to tell us that they had only “fluid” or “current” data files. They explained that this was due to the county updating its voter list for people who had moved or died and due to saving over the original computer file. Presumably, these counties were just sloppy. But data storage is trivially inexpensive, and saving a file time-stamped on election day would be easy. Yet, amazingly, these election bureaus frequently claim they do not archive their data.

Due to the previous report’s findings on the 2020 general election, we knew that most counties would not keep their data time-stamped or keep their files current on everyone who voted in the 2022 election. Accordingly, we alerted the counties we had studied before the November 8, 2022, election to ensure they would keep the election-day data this time. After all the counties were notified, we followed up and started contacting the Arizona counties on November 17, 2022, requesting the list of voters who voted in the November 8, 2022, general election.

Data Findings

After six months of persistence with these Arizona counties (full correspondence in Appendix A), the precinct-level data for the six Arizona counties was received in full. Analysis showed some precincts where there were more ballots being counted than there were registered voters listed as casting ballots and some precincts where the reverse occurred.[2] Across these counties, 6,057 more ballots were recorded as cast than there were registered voters listed as voting. In precincts where the reverse was true, 2,184 more registered voters were listed as voting than ballots shown as counted (Table 1). That results in an 8,241-vote discrepancy, or 0.36% of the total ballots counted. Some may say this is just a small discrepancy, but 0.36% amounts to 29.4 times the 280-vote difference in the Arizona attorney general race.

The finding of discrepancy alone isn’t the end of the story. There are three possible explanations for the discrepancy.

First, some voters, due to safety concerns (i.e., policemen, firemen, public servants, or individuals in witness protection programs), need to conceal their personal information due to threats or other concerns. As a result, their identity, including name and address, is hidden. These individuals are considered “secured” or “protected” voters; states and localities often use different terminology for these voters, so for the sake of this report, we refer to them as “secured voters.” Across these six counties, according to the county election officials, there were 4,078 “secured voters that voted in the 2022 general election.” By subtracting the secured voters (we excluded Maricopa County), this still leaves a total discrepancy of 4,187 (Table 2). That is a conservative estimate and does not include the numbers for Maricopa County due to their number of 2,888 secured voters, which is 37 votes larger than their discrepancy gap of 2,864 (See Table 2).

Second, we also learned from Yavapai County’s Registrar of Voters Office that the discrepancy might arise for another reason. They explained:

“If a voter checks in on Election Day but walks out with their ballot (does not put it in the ballot box), they will be listed as voted because they signed the register but there will not be a ballot to tabulate. I do not think this is common, but we do get reports of this happening from poll workers.”

Even if every instance of a precinct having more registered voters than ballots counted is a result of this occurrence, the reverse scenario (of more ballots counted than registered voters) still leaves 2,242 unexplained discrepancies outside of Maricopa County after taking into account the secured voters whose identities were masked; that is still eight times the 280-vote margin in the attorney general race.[3]

Third, the data we received on who voted in the November 8, 2022, election may not be accurate. Even though we requested the list of who voted on that day and the federal Civil Rights Act of 1960 requires counties to preserve those records for 22 months, many counties immediately start updating their list of voters after the election and save over their original files. Only two counties, Pima and Pinal, warned us that this might account for part of their discrepancies.

But even if the entire discrepancies in Pima and Pinal were due to not having an accurate list of who voted in the election, it would still leave a discrepancy of 688 votes (see Table 2, column 5). For column 6 in Table 2, that would mean a gap of 320. Again, these numbers are based on several conservative assumptions, but both are still larger than the 280-vote difference in the attorney general race.

There are nine other Arizona counties for which data was not requested or reviewed. If these counties had similar problems, it is another reason to believe that our most conservative estimates of 688- or 320-vote discrepancies between votes counted and registered voters listed as voting underestimate the possibility for fraud to be occurring in Arizona.

Finally, even if these six counties did not accurately record who voted in the general election last year, that itself would be a concern and a violation of federal law. There were 2.24 million votes for the attorney general race in the six counties for which we have data and 306,217 in the other nine counties. Assuming that discrepancies occur at the same rate in the nine counties we did not review, it would raise our most conservative estimated discrepancies from 688 to 782 and from 320 to 364.

Conclusion

Two possibilities remain. Either there are more votes than registered voters who voted, and that gap is larger than the vote difference in the attorney general’s race, or the counties have failed to keep accurate records of who voted in the election. Of course, both may be true.

A technologically advanced and democratic society in the digital age should never face discrepancies in its election results or be in situations where more votes are counted than voters; votes counted should always equal the number of registered voters who voted.

So, how do we fix this moving forward? The Civil Rights Act of 1960 underscored the importance of retaining Election Day voting records. Election officials in Arizona rightly changed course during the 2022 election after AFPI’s reporting brief following the 2020 cycle raised the issue. Prefiling a discrepancy analysis that highlights the ratio of votes counted to registered voters who voted — following the template provided in this report — should occur before certifying election results. This process is a way to rebuild trust in elections and provide a “receipt” to voters.   

Election officials and state legislators should follow Ohio’s example for modernizing its election laws. This year, the Ohio legislature introduced the landmark DATA Act, Senate Bill 71, which will update legal definitions of how it stores election records and add procedures for how the public can inspect records. These election result records will be open sourced and available for public inspection, helping restore confidence and trust in elections and a practice that was once only available by public records request. Open sourcing election records will streamline public records requests and help to alleviate local election officials from the burden of high-profile legal demands witnessed during the 2020 cycle. The bill creates a new Office of Data Analytics and Archives within the Office of the Secretary of State to serve as a clearinghouse for the new process and to conduct a discrepancy analysis following their elections, just like this report.

Creating an open-sourced process to verify if the number of votes counted is equal to the number of registered voters who voted in that election ensures a transparent election process. Identifying discrepancies early will give election officials a clear roadmap on how and where to address wide-ranging issues from potential fraud to inaccurate recordkeeping.

It’s time to modernize our election laws and bring how we count votes into the 21st century to make it easy to vote but hard to cheat. Doing so will ensure that all elections, especially close races like those experienced in the Arizona 2022 general election, are clear of discrepancies before election certification.

Biographies

Dr. John Lott, Ph.D., is from Detroit, Michigan. His academic journey began at the University of California, Los Angeles (UCLA), where he received his bachelor’s degree in economics in 1980. He obtained his master’s and Ph.D. in the same field at UCLA in 1982 and 1984, respectively. From there, Lott has held research and teaching positions at the Wharton Business School, Yale Law School, University of Chicago, Stanford University, UCLA, and Rice University. Outside of employment at these universities, Lott also worked as Chief Economist at the United States Sentencing Commission and senior advisor for research and statistics in the U.S. Department of Justice. He currently serves as the president of the Crime Prevention Research Center. He has published over a hundred peer-reviewed articles in academic journals and 10 books.

Steven M. Smith is from Dayton, Ohio, and serves as AFPI’s Interim Chief Operating Officer and Chief of Staff. Smith recently served as a Special Assistant to the President and Chief of Staff of the White House Office of American Innovation and the Domestic Policy Council. Smith served on the team that passed the First Step Act, a historic bipartisan criminal justice reform bill, and was heavily involved in the successes of the White House Opportunity and Revitalization Council and the COVID-19 rapid response operations team. Steve previously spent 12 years in campaign management, leading and supporting political races from city, state, federal, and presidential levels, including service on Senator Mitch McConnell’s (R-KY) 2008 campaign and as the Special Assistant to the Republican National Committee’s Co-Chairman during the 2016 cycle.

Contributions

On behalf of AFPI, we would like to extend a special thank you to Local Labs, who conducted all public record requests and dedicated thousands of hours of research and follow-up to gather the data. Also, a special thank you to Kristen Ziccarelli, who helped write this report.