Play Live Radio
Next Up:
0:00 0:00
Available On Air Stations

Are humans or machines better at counting ballots? One Arizona lawmaker wants to find out.


By Howard Fischer
Capitol Media Services

PHOENIX -- A veteran Arizona lawmaker thinks he's found a way to finally end the debate over whether humans are better or worse than machines at counting ballots.

Put them both to the test.

Sen. John Kavanagh, R-Fountain Hills, told Capitol Media Services he believes that a head-to-head contest will prove that hand counting can both be accurate and produce timely results.

And if that's the case, he said, the machines that some say can be hacked and can produce inaccurate results -- allegations never proven -- can be scrapped to help restore voter confidence.

But he conceded it is, at this point, only a theory.

The only example Arizona has was the "audit'' of two of the contests in the 2020 election in Maricopa County. And that took months and failed to show any significant different between the hand count and the official tally.

So Kavanagh has crafted SB 1471 to set up the procedures to test it out.

The measure already has gotten the endorsement of Maricopa County Recorder Stephen Richer, a Republican like Kavanagh, who dubbed it a "man-versus-machine test.'' But Richer already thinks he knows how at least part of this will go.

"This legislation will build confidence in our election system by showing that machine tabulation is highly accurate, free of bias, and fast,'' he said.

SB 1471 comes amid claims by some, including failed Republican gubernatorial hopeful Kari Lake, that machine counting is inherently suspect and susceptible to fraud and hacking. In fact, she and Mark Finchem, who ran unsuccessfully for secretary of state, are still trying to have the procedure declared the procedure illegal despite a federal court ruling against them.

And there are separate proposals at the Capitol to outlaw hand counts in one form or another.

That includes HB 2307 crafted by freshman Rep. Cory McGarr, R-Marana, to mandate hand counts in all future city, town and county elections. It was approved this past week on a 6-4 party-line vote by the House Committee on Municipal Oversight and Elections.

What Kavanagh is proposing is selecting four precincts at random and taking 100 ballots from each.

The actual ballots would be put through a tabulator, with the results of all the races -- there can be more than 80 given statewide, congressional, legislative, judicial and local races, plus special districts and ballot measures -- initially kept confidential.

Duplicate ballots would be made and given to volunteers. They would have to conduct the hand count the same way counties are now required to conduct hand counts of a few races in a few precincts: teams of volunteers composed of at least one person from two of the state's three recognized political parties.

Under the terms of SB 1417, if the numbers come out within 0.1 percent of each other, that ends that part of the test. If not, both halves are rerun using different tabulators and different hand counters.

But that's just half the issue.

The local recorder would be required then to figure out the average number of ballots counted for each counting team.

That number would be extrapolated out to determine how many people working 16-hour days would be required to hand count the nearly 2.6 million ballots cast in the last general election.

"The main argument against doing hand counts is that they're inaccurate and would take too long,'' Kavanagh said.

"So the purpose of this is to have a very controlled experiment,'' he said, to measure not just accuracy but also the time it would take to implement this on a statewide basis.

As crafted, the test mandated in SB 1471 would occur in Maricopa County. But that now is uncertain.

"I may need to relocate this,'' Kavanagh said. And the reason, he said, is Richer's comments that he believes the test will show that hand counting is inferior.

"So, he's immediately biased his county for counting,'' Kavanagh said. "The results can be challenged because his counters were biased because he had a position.''

Richer, for his part, said he's not hoping for a specific outcome. But he conceded to Capitol Media Services he thinks he knows what the test will show, regardless of in which county it is conducted.

"Unless 50 years of social science is wrong, then it will show that machines are far more accurate, far faster, far less biased when it comes to repetitive tasks,'' he said.

Arizona does have at least one example of how hand counting works.

After the 2020 election, Karen Fann, then the Senate president, contracted with Cyber Ninjas to "audit'' the 2.1 million returns only from Maricopa County, including conducting a hand count.

That ended up with two six-hour shifts, with anywhere from 50 to 100 people working at any one time, for more than two months on just two races.

In its report, Cyber Ninjas said there were more than 1,500 individuals involved at one point or another, with a total of 100,000 "hours contributed.'' But not all that was involved in counting, with time also spent examining ballots for bamboo to see if they were fake and doing microscopic reviews to see if it looked like some ovals next to candidate' names had been filled in by machine rather than by hand.

The upshot of all that?

It confirmed that Joe Biden won the vote in the state's largest county. In fact, it showed that Biden tallied 99 more votes than the county tabulators had recorded -- and Donald Trump had 261 fewer votes than the official record.

And in the race for U.S. Senate, which also was reviewed, there also was no major change, with a difference of just 1,167 votes between the two tallies. This, too, confirmed that Democrat Mark Kelly handily defeated incumbent Republican Martha McSally.

Despite that there are people who still question the results as the report did not specifically say that Biden or Kelly won. That's because Cyber Ninjas, which had no previous experience with elections and was largely funded by donations from Trump supporters, also claimed to have found other problems which, at least theoretically, could have affected the outcome.

None of those ever were proven in court challenges.

Jack Sellers, who was chairman of the county board of supervisors at the tie, seized on the report to say it exonerated the county's handling of the election.

"This means the tabulation equipment counted the ballot as they were designed to do, and the results reflect the will of the voters,'' he said. "That should be the end of the story. Everything else is just noise.''

Richer said the results of the review sought by SB 1471 should rebut arguments by those who say the experience in other countries proves that hand counting works.

Take, he said, the presidential election in France.

"You just have the presidential election on the ballot,'' Richer said. And early voting is not permitted.

Counting is done at each polling place, with initial results posted that night.

What's lost in that comparison, said Richer, is that ballots in Arizona contain not just the gubernatorial contest but a multitude of races. And each one of those races needs to be counted, not only by at least two workers from different parties, but also transcribed.