Election consultants: Results of Arizona ballot recount will be inaccurate at worst, incomplete at best

When Maricopa County election workers loaded 1,691 boxes of ballots onto semitrucks in April and drove them to Veterans Memorial Coliseum, they didn’t send instructions with them.

The Arizona Senate’s contractors cracked open the first box without a way of knowing how many ballots should be in each box, without a complete understanding of the complicated way the county tallies votes and stores ballots, without much to compare their results to and without a background doing this type of work. 

For the next seven weeks, six days a week, early in the morning and late into the night, a mostly volunteer crew of dozens of workers recounted the votes cast in the 2020 presidential and U.S. Senate race on nearly 2.1 million ballots.  

Four national election consultants called it an error-prone and ever-changing process.

The contractors who designed the process, led by Florida-based cybersecurity firm Cyber Ninjas, lacked the information they needed from the county as well as the knowledge of elections to do the recount correctly, according to the consultants, who have been watching the process closely since it began.

The final report on the hand count, they say, will be incomplete at best and inaccurate at worst.

Benny White, a Republican elections consultant who lives in Pima County and has helped audit elections in Arizona and other states for more than a decade, said there is a “zero percent chance" that the results will be accurate.

“They chose the least credible way to count this number of ballots and the votes on the ballots," said White, who has pressed Senate President Karen Fann to compare the contractors' results to an analysis of the election results he and a team of election analysts recently conducted.

The contractors finished recounting votes and inspecting ballots by the end of June, and initial results are expected in late July or early August, according to audit spokesperson Randy Pullen.

Cyber Ninjas disputes the claim that their process is error-prone. Company spokesman Rod Thomson said that the contractors have multiple checks and counterchecks to reduce errors and block any potential attempt to manipulate the results. Thomson said the claim that the contractors are inexperienced is unfair, considering this type of audit has not been done before.

“The multiple components, and the depth and breadth of them, is a first in American history,” Thomson said. “So no one has experience in this. So much of this long-running narrative is false.”

Yet observers with Democratic Secretary of State Katie Hobbs’ office have noted several times that workers broke their own rules for ballot security, and have raised concerns about quality control. 

The partisan nature of the effort — from the GOP Senate leaders who commissioned the effort to the people who are opaquely fundraising for it and counting the ballots — have convinced many that the results can’t be trusted.

Senate liaison Ken Bennett, a former Republican lawmaker and secretary of state, said that he has confidence in the processes and in the results to come. He said he believes that scrutiny of the results will be 10 times more than the usual scrutiny of the election results, from journalists, the Secretary of State’s Office and others.

So when this final report comes out, Bennett said, it has to be solid. The results should include results not just from the hand count but also a ballot inspection process and an examination of the county's voting machines.

Whatever the results say, they will not change the state's certified election results, which declared President Joe Biden and U.S. Sen. Mark Kelly winners.

'We humans are error-prone'

The Senate contractors’ hand count is likely to come up with different final numbers than the county’s, in part because of a process for counting ballots that left too much room for human error, according to the four national election consultants. 

They include White, and Harri Hursti, a data security expert with election auditing experience and founding partner of Nordic Innovation Labs in New York; Mark Lindeman, acting co-director of the Philadelphia-based election integrity nonprofit Verified Voting; and Jennifer Morrell, a partner at The Elections Group, a national elections consulting organization.

They say that the contractors may not be able to explain why their numbers are different from the county's, because they may lack the ability to match up the results batch by batch or even precinct by precinct. Two weeks after the recount ended, Bennett said the contractors were still trying to figure out whether it’s possible.

The contractors used an unconventional method to count ballots.

Each counting table had three counters who would look at the ballot as it came to them on a lazy Susan-style turntable, and then record three marks:

• One to acknowledge they looked at a ballot.
• One for the presidential candidate a voter selected.
• One for the U.S. Senate candidate selected.

The ballots whizzed past the workers. Observing the process on multiple occasions, it often took between 5 and 10 seconds for three workers to count two races on each ballot, each looking for only a second or two before the ballot moved on.

“Not one of the counters is guaranteed a good look at the ballot that is literally spinning by them,” Lindeman said. He and Hursti provide election auditing consulting across the country and recently were hired to help run a hand count of the 2020 election in Windham, New Hampshire.

Human error makes it practically impossible to eliminate inaccuracy when doing a ballot count by hand, Lindeman said. The contractors' process created even more room for mistakes, he said.

At first, the counters were attempting to do 100 ballots in a row without a break. Then, they moved to 50 ballots, according to Bennett.

Just counting this many ballots in a row leaves room for mistakes, Morrell said. It’s standard in ballot hand counts to only count by 20. Maricopa County counts by 10 in the post-election sample hand count of ballots that state law mandates.

Hursti said that the rotating table is “counterproductive,” in part because it doesn’t allow for contemplation, and the cavernous coliseum where workers counted ballots at as many as 40 or more tables was distracting.

“It’s not that much about being malicious,” Hursti said. “We humans are error-prone, and we do make mistakes.”

Seeking files: Republic takes Senate, Cyber Ninjas to court for audit records

Thomson with Cyber Ninjas said that people making claims that the process had errors “better have provable specifics to make such claims.” He gave an example of the way the contractors were preventing errors.

“The hand counting had three eyes on each with no talking among the group, and an overhead camera with systems to correct if all three did not agree,” he said.

After the counters looked at 100 ballots, a table manager would review the counts on each of the counter’s tally sheets to see if they matched. The process for how to reconcile differences changed at some point during the recount.

At first, if all three counters’ totals were different, they recounted the batch. If two counters’ totals matched but one was off by less than three, the table manager would simply accept the two matching counts as final.

A few weeks into the hand count, the process changed. It appeared to happen after the first company running the hand count, Wake Technology Services Inc., left and StratTech Solutions took over. 

After that point, if one of the counters came up with a different result than the other two, table managers started to review the tally sheets to try to find where the difference occurred. If they could find it, they would go back to the actual ballot in question and determine voter intent.

Hursti, Lindeman, Morrell and White said they were glad to hear that it changed and that the initial process was extremely problematic.

Hursti said that the contractors’ original plan to not reach a consensus on each vote on each ballot is “bulls---.” That leaves unresolved questions about voter intent, he said, which discredits the results.

“You just cannot do it this way,” he said. “This is 101.”

Audit in new digs: Ballots move to swamp-cooled building at fairgrounds

Bennett said he disagreed that the initial process was bad, or that it somehow discredited the results.

But the new process doesn't change that only one person is looking at a ballot at a time, which is a flawed system, Hursti, Lindeman and Morrell said. A bipartisan team of at least two people should be deciding voter intent for each ballot, they said, to come up with a definitive call on the votes on each ballot. 

Cyber Ninjas CEO Doug Logan said before the audit began that he could not guarantee that each counting table would have members of different political parties.

Morrell, who served as an observer of the audit for the Secretary of State’s Office during the first week of May, said that she hopes that, before the numbers come out, people understand that they should expect them to look different from the county’s.

This is not because there was fraud on the part of the county, she said, but because the audit was mismanaged, processes changed too frequently and there initially was no quality control.

“This was a group of inexperienced individuals, with no planning, no expertise, and changing practices,” she said. “Of course the numbers don’t match. This is why this whole endeavor has been so alarming and dangerous.”

Thomson said the Secretary of State’s Office observers are biased, noting that Hobbs is a Democrat who has opposed the audit from the beginning.

Morrell disputes the claim that she is biased, explaining that her background is in advocating for robust election audits and consulting for election officials in Republican and Democratic states.  

"If I have a bias, it is a belief in the rule of law; that we carry out elections and audits and recounts and challenges to an election in accordance with federal and state laws," she said. "My opposition to the political review happening in Arizona has nothing to do with my personal politics and everything to do with my belief in democracy and democratic norms."

How the final results were tallied

Morrell took issue not just with the process for counting ballots but with the method by which the vote counts from the tables were entered into a computer as the official count.

Morrell, while serving as an observer, said she raised a concern that people were doing this work without any oversight from others.

When she noted that, for at least a few days following, the contractors added a screen above the workers so that others could watch. Then, that practice ended.

Another observer from the Secretary of State’s Office noted that the people entering the numbers from the tally sheets onto the computer had access to red and green pens, which could have been used to alter the tally sheets because that’s the same color ink the counters used.

Bennett said a second person is comparing the final results entered on the computer to a scanned version of the original hand count tally sheets, "so, it’s not just one person’s work product."

The auditors are doing a quality control process to double check that every tally sheet has been scanned and the numbers are correct, he said.

“We want it so tight that there is no question,” Bennett said.

The Secretary of State’s Office raised — in publicly posted audit observation notes — concern about the quality control process itself, though.

When observing people recounting ballots for quality control, an observer noted that the counters had counted 24 ballots instead of 25 ballots. Instead of going back and redoing the hand count of the votes cast on those ballots, the observer said that the table manager indicated that “she thought she found the 25th ballot stuck to another ballot” and they moved on without recounting.

John Brakey, an elections activist and consultant for the audit, said he is “very concerned” about whether the hand count can be trusted.

How detailed will the results be?

Election consultants say that for a final report to be meaningful, it will need to detail any discrepancies between the hand count and the county's official results, including where and why a discrepancy occurred so that improvements can be made. Fann has maintained that improving any weaknesses in the voting system is why she ordered the audit.

However, Bennett said contractors still are attempting to figure out whether they will be able to give a comparison by precinct and even down to batches of 200 mail-in ballots.

They just recently figured out — after the hand count ended — which documents they got from the county could tell them the correct number of ballots per batch.

They were at first looking at the wrong documents for the ballot count, leading Fann to write the county a letter in May asking why ballots were missing from the batches. This was the same letter in which Fann claimed that the county deleted files, only to have the contractors say later they had the files in question.

The initial claim of missing ballots led the right-wing One America News Network to repeatedly say on air that the contractors were missing thousands of ballots.

County election officials sent back a technical letter explaining the county's system and pointing to the correct documents. The auditors got those documents not because they had asked for them in their original subpoenas but because Brakey got them through a records request. 

Brakey said the contractors weren't aware of what they needed when they began. He sometimes wonders if the secrecy on the part of the contractors “is just secrecy or it’s just stupidity,” he said.

If the contractors can’t figure out how to report details of their results, they might have just a few numbers to share on the overall election results for the presidential and U.S. Senate races, including:

• The number of total ballots voted early, either in person at vote centers before Election Day or by mail.
• The total number of ballots cast on Election Day.
• The total number of ballots cast provisionally, which happens when a voter goes to a polling place to vote but isn't on the voter rolls, so their voter registration information must be confirmed later.

Election consultants who spoke with The Arizona Republic said they aren’t sure what good those numbers would do. The final report, they say, should be able to identify specifically where the county and the contractors’ counts differed. Then, it will need to identify why that happened and how to improve it.

White said without this detail, without specifically knowing where the numbers are different, “I don’t think we should believe anything they have to say.”

“Because you simply don’t know where the differences are — where the problems are," he said.

White said this type of analysis is difficult to do. He would know, because for the past several weeks, he has worked with a team to analyze the county’s vote cast record to break down the election results to see specifically which county voters, by political affiliation, voted for each presidential candidate. He found that disaffected Republicans are what allowed Biden to take Arizona — not fraud.

White said he believes the consultants “started down the wrong path” because of a lack of experience, and couldn’t recover.

“In the meantime, the public is sitting here distrusting the election,” White said. “That’s the dangerous and destructive part of this.”

The results of the hand count also won’t reinstate former President Donald Trump, which ardent supporters say has become the new intent of this audit, whether the Arizona senators support that or not.

What it has accomplished is fundraising for the state GOP and for the nonprofit advocacy groups that say they are funding the audit. 

Lindeman said the lack of detail the Senate's consultants will be able to produce about the election results will raise additional doubts about the election. Even casting doubt that the county delivered the right number of ballots to the contractors does damage, he said.

“Tails they win,” he said. "Heads we lose.”

Reach the reporter at jen.fifield@azcentral.com or at 602-444-8763. Follow her on Twitter @JenAFifield.