Data in audit report shows Cyber Ninjas' hand count is 'fiction,' election analysts say
Three top election analysts say the hand count of ballots by Cyber Ninjas is "fiction" and the Arizona Senate's own records prove it.
The hand count of 2.1 million ballots cast in Maricopa County's 2020 election is wildly inaccurate and is not supported by the data made public during a three-hour Senate hearing last week, their new report finds.
The conclusions are based on a review of 40 ballot boxes, which the Senate for the first time broke down by number, batches and various types of tallies.
The analysts found the hand count was short by nearly 16,000 ballots, or roughly one-third of the total ballots contained in the boxes.
They say the numbers contradict the main finding of the Senate's monthslong election review: That the hand count nearly matched the county's certified election results.
It did not, and the result cannot be trusted, they say.
Nor did the hand count match a machine count of ballots ordered by the Senate as a way to double check the Cyber Ninjas' work.
"The discrepancies are so large that they cannot just be brushed aside," said Larry Moore, the founder of Boston-based Clear Ballot Group. "The reality is they just made up the numbers."
Moore, along with Benny White, a prominent Pima County Republican data analyst, and Tim Halvorsen, Clear Ballot's retired chief technology officer, laid out their conclusions in a 27-page report.
They posted the findings Friday in a blog titled "Maricopa Hoax: The Ninjas made up the numbers."
"It's the Senate's own data," Moore told The Arizona Republic. "This blows apart everything that's happened in the past nine months. Anybody with a hand calculator and five minutes could figure it out."
HOW WE GOT HERE : An Arizona audit timeline
Fann: Not interested in report's findings
Senate President Karen Fann discounted those claims as "a lie that borders on inflammatory." She said the audit was conducted by good, credible people and should not be second-guessed by those not involved in it.
Fann said she wasn't interested in seeing or reading the report.
She asked instead what it meant if the hand count were wrong: "Are they saying Trump won?" she said.
Cyber Ninjas' CEO Doug Logan, who was provided a copy of the report by email, did not respond to multiple interview requests.
The hand count was supposed to be the backbone of the Senate's election review, which Fann called the "the most detailed, demanding, and uncompromising election audit that has ever been conducted."
Beginning in April, hundreds of volunteers converged daily at Veterans Memorial Coliseum to scan, count and record ballots to recount the races for president and U.S. Senate.
Ballots were photographed. They were put on color-coded lazy Susan-type tables for counters to spin through reviews. They were examined under ultraviolet light to determine if counterfeit ballots with bamboo fibers were introduced into the system.
Once the hand count ended in late June, Senate Republicans commissioned a machine count of the ballots to check the accuracy of the hand count. Then they commissioned a digital review of signatures on ballot envelopes.
Fann maintains the election review was not an attempt to overturn the 2020 election but a review to ensure the integrity of future elections. However, she hired companies with no experience in election audits and people to run it who espoused election conspiracy theories, including Logan.
The audit was funded primarily by "Stop the Steal" nonprofits with leaders determined to restore Donald Trump to the White House.
But when the Arizona Senate laid out its findings in five final reports on Sept. 24, the top line was that the hand count and the machine count matched the county's certified election results.
President Joe Biden won the election in Maricopa County by 45,000 votes. The counts also confirmed that U.S. Sen. Mark Kelly, a Democrat, won.
The Cyber Ninjas' reported there was less than a 1,000-vote difference between the county's certified ballot count and the Cyber Ninjas' count.
The findings suggested that Cyber Ninjas — a cybersecurity company that never had done an election review and funded by Trump loyalists — got the numbers right. Its count even found Trump got fewer votes than what the county reported.
Moore called it "political theater," a fiction designed to legitimize the Cyber Ninjas' work and to undermine democracy by encouraging partisan reviews in other states. He said the Arizona Senate was either victim or party to a "hoax."
He said Fann's attempt to reduce the issue to whether Trump won misses the larger concern about the audit's efficacy, which took months, cost several million dollars and sowed doubt about the security of Arizona elections.
What did the analysts' report find?
The analysts dissected the 48,371 total ballots contained in the 40 boxes with a methodology they've used to audit election results around the country.
It essentially uses data contained in election systems called "cast-vote records" to verify results.
Voting machines capture images of every ballot cast and tabulate information about every available choice in an election. Those records can be accessed to quickly and accurately confirm counts and identify anomalies in ballots and vote counts.
The Senate obtained the cast-vote records from Maricopa County under subpoena along with other election-related material and equipment. The cast vote record is a public record, but the Cyber Ninjas and the Senate have refused to make their analysis of it public record.
As a result, no independent analysis could be made of ballot boxes and batches being counted by the Cyber Ninjas — until last week, when the Senate included a series of spreadsheets in its audit reports.
The report, authored by audit spokesperson and former Arizona GOP Chair Randy Pullen, includes 17 pages excerpted from a 695-page report and uses data tables to detail the machine count results.
The report is limited to 40 numbered ballot boxes contained on one pallet. There are 46 pallets total containing 1,634 countable boxes of ballots.
Moore, White and Halvorsen, who are known in the industry as The Audit Guys, seized on the data.
The Audit Guys said the tables show the 40 boxes of ballots matched the results of the county's count results by almost 100%; 37 boxes were identical, and three boxes showed a difference of five ballots.
But the Audit Guys said the hand count numbers don't match at all. They found that the Cyber Ninjas' count of the 40 ballot boxes is off by 15,692 ballots.
Audit spokesperson: Analysts made a bad assumption
Pullen on Thursday called the Audit Guys' report inaccurate. He said it was based on an assumption that the machine count of ballots was completed after the hand count.
"The Cyber Ninja hand count was not completed before we did the machine count," he said in an email to The Republic.
The Cyber Ninjas were still in the process of checking their hand counts when the machine count launched, he said. When variations were spotted on count sheets of each ballot box, they did recounts.
Pullen's explanation raises additional questions about the reliability of data included in the Senate's own reports on the hand count and the machine count that was supposed to verify it.
It strongly suggests the machine count was not independent of the Cyber Ninjas' hand count, nor was it completed in a linear fashion, one following the other as a way to ensure accuracy.
"The purpose of the machine count was to confirm the total ballot count, which it did," Pullen said. "This does not mean it counted every ballot perfectly."
That explains why some fields in the spreadsheet were left blank, Pullen said.
"This is why there were boxes on our tabulation sheets that did not have hand counts included," Pullen said. "So the conclusion that the hand count was off (15,692) from the machine count is incorrect."
Pullen, who is a certified public accountant, appears to have included the tables as examples of the accuracy of the machine counts, not to cast doubts on the hand count numbers.
In his report to the Senate, he noted several problems with batches of ballots in the boxes. But he said the final machine count matched the county's cast-vote record with a difference of 121, an 0.0058% difference that he described as "insignificant."
Moore said Pullen is playing games with the numbers and casting doubts on the counts reported Sept. 24.
"Mr. Pullen may be attempting to say that the accounting and aggregation of the hand count was not complete on September 24, but that raises further questions — as we have done — to the reliability of the announced vote counts," Moore said. "You cannot have it both ways: Either you had a hand count or you didn’t."
Even giving Pullen the benefit of the doubt, Moore said the Audit Guys found an error rate on the hand count of 10% based on the data tables.
"That is 200,000 missing ballots," Moore said. "But these ballots aren't missing. The Cyber Ninjas' count is wrong."
Audit Guys keep pinging Senate, with no luck
The Audit Guys three times offered to perform ballot count analysis for the Senate, and Fann rejected each of their offers.
They challenged the Senate in June to test the results of the Cyber Ninjas' hand count based on two ballot boxes.
Moore said the Senate could select any unopened ballot boxes and the team could within minutes provide an accurate count of each race on the ballots inside — without ever opening them.
It would take about five minutes for the team to produce a report detailing the vote tallies inside the box, which contain an average of 1,274 ballots, Moore said.
In July, the Audit Guys expanded their challenge from two boxes to all 1,634 boxes from Maricopa County's 2020 election. Moore said they tested their methods on the machine counts of 24 boxes provided by Ken Bennett, the Senate audit liaison.
Their analysis and the box counts matched box for box with a few discrepancies, he said.
The Audit Guys upped the ante again in August by analyzing vote counts in a single race. They posted the number of votes on the ballots in those boxes won by Libertarian presidential candidate Jo Jorgenson in November. Moore said their results matched perfectly.
In a post on their website, real-audits.org, the trio asked the Senate to provide the Cyber Ninjas' findings to see how the counts compare. Again, Fann turned them down.
Maricopa County Recorder Stephen Richer said he has high regard for the Audit Guys and said their report supports the county's election results.
“White, Halvorsen, and Moore know what they’re doing," he said. "They have over 60 years of election experience together. Their efforts and studies validate the most important factor in the election administration: that Maricopa County’s tabulation of the paper ballots was fair and accurate."
Richer said he and White share similar stories.
"Neither of us were in office during the 2020 election," Richer said. "We both ran as Republicans in the 2020 election. We both voted for Trump. But we both are putting aside our disappointment in the outcome to acknowledge the truth: that the 2020 election wasn’t stolen in Maricopa County.”
Analyst: More election data should be made public
Moore said the data Pullen provided in his report on the machine counts has been closely guarded by the Senate.
He said the Senate has refused to provide ballot and vote counts by box because it could be used to discredit public statements made by Fann, Pullen and Logan.
The Republic is seeking the same records from the Senate, which has refused to turn them over. Senate lawyer Chris Kleminich said Thursday that Cyber Ninjas controls the information and has yet to turn the records over to the Senate.
Even so, Kleminich said the Senate still might claim legislative privilege and withhold the records from the public.
The Republic in July filed a lawsuit under the Arizona Public Records Law naming the Senate and the Cyber Ninjas as defendants. A Maricopa County Superior Court judge in the case ruled that records related to the audit are public.
Moore said it is clear from the audit reports that Pullen is in possession of the 695-page report of the Cyber Ninjas' hand count and the Senate's machine count.
He said the release of these records would peel back the layers of secrecy in the audit and further expose it as a sham.
The Senate needs to explain itself to the public, he said.
"Why don’t you comply with our public record request and send us the 695-page report along with whatever you think passes for vote counts?" Moore said.