Republican county officials, as well as candidates for statewide offices, suggest that counting by hand would be more reliable than machine tabulation. Research shows the opposite.
Over the past two years, Republicans have pursued an array of changes to how Americans vote. The past few weeks have drawn attention to a particularly drastic idea: counting all ballots by hand.
Officials in Cochise County, Ariz., recently pushed to do that in next month’s election, and whether or not they go through with it, the efforts may spread. Republicans in at least six states introduced bills this year that would have banned machine tabulation, and several candidates for statewide offices have expressed support, including Kari Lake and Mark Finchem, the party’s nominees for Arizona’s governor and secretary of state, and Jim Marchant, its nominee for Nevada’s secretary of state.
The New York Times spoke with six experts in election administration, and all said the same thing: While hand counting is an important tool for recounts and audits, tallying entire elections by hand in any but the smallest jurisdictions would cause chaos and make results less accurate, not more.
“People who think they would have greater confidence in this process think so because they haven’t seen it,” said Mark Lindeman, the policy and strategy director at Verified Voting, a nonpartisan organization focused on election technology. “The process in real life would not inspire confidence at all on this scale.”
The proposals often stem from false claims by former President Donald J. Trump and his allies that voting technology was somehow to blame for Mr. Trump’s loss in the 2020 presidential election. Most of those claims center on electronic voting machines, but some extend to scanners and tabulators that count paper ballots.
The right-wing arguments against tabulators rely not on evidence that they have been compromised — because there is none — but on the possibility that they could be. In a lawsuit filed in April, Ms. Lake and Mr. Finchem asked a federal court to mandate hand counting in Arizona, arguing that the state’s ballot scanners were “potentially unsecure” and denied voters “the right to have their votes counted and reported in an accurate, auditable, legal and transparent process.” The court dismissed the case, and Ms. Lake and Mr. Finchem are appealing.
Research indicates that hand counting increases errors.
A study published in 2012 looked at discrepancies between initial counts and recounts in New Hampshire and found that, on average, those discrepancies were 0.8 percentage points smaller in towns that used scanners than in towns that counted by hand. A study in 2018 analyzed two statewide races in Wisconsin and found that “vote counts originally conducted by computerized scanners were, on average, more accurate.”
Hand counting is “incredibly labor-intensive, very slow and, it turns out, subject to more error than having a machine do it,” said Stephen Ansolabehere, a professor of government at Harvard who led both studies.
The reason is simple: Humans are bad at tedious tasks.
Douglas W. Jones, a retired associate professor of computer science at the University of Iowa, former Iowa voting systems examiner and former member of the U.S. Election Assistance Commission’s Technical Guidelines Development Committee, recalled a pre-election test in Florida in 2004. He and other officials worked slowly, taking about a minute per ballot, and “despite all that care, we kept making mistakes,” he said. “You lose focus when you’re doing something very repetitive.”
Experts agree that hand recounts are valuable in close races where a small number of ballots could change the outcome; a voter’s intent may be clear to a human, for instance, when it was not to a machine. They also agree that hand audits — spot checks of a small percentage of ballots — are useful in verifying machine counts.
Audits, which are common, can confirm “whether the machine’s basically got it right,” said Charles Stewart III, a professor of political science at the Massachusetts Institute of Technology and the director of the M.I.T. Election Data and Science Lab.
But these tools are valuable because they are limited: The fewer ballots that workers have to review by hand, the more time and attention they can give each one.
This level of diligence cannot realistically be replicated for every ballot in every race, experts said.
In at least one place, Nye County, Nev., officials are trying to have it both ways — doing an initial machine count followed by a full hand count. (The American Civil Liberties Union is suing to try to stop the plan.) But that, too, has costs in time and resources.
Election staffs are often overworked, and “to layer something this time-consuming and this prone to error on top of all that other work is very concerning,” said Gowri Ramachandran, senior counsel in the Elections and Government Program at the Brennan Center for Justice. “In all but really the tiniest of jurisdictions, it just does not make sense.”
In Cochise County alone, which accounted for less than 2 percent of votes cast in Arizona in 2020, Mr. Lindeman estimated that a full hand count would require 12,000 person-hours of work. The United States District Court for the District of Arizona found in the lawsuit from Ms. Lake and Mr. Finchem that in the state’s largest county, Maricopa, a hand count would require 25,000 temporary workers.
Even if officials found this many workers, a hand count could take weeks, and many of the same right-wing figures promoting it say counting should be completed on election night. And the sheer number of participants could undermine security: How thoroughly would the workers be vetted before being entrusted with ballots?
“If you have to recruit and hire that many people and insert them into this process, you’re immediately making something less secure,” said Will Gaona, the Arizona policy advocate for Protect Democracy, a nonpartisan watchdog group.
Some countries, like France, do count ballots by hand. But their elections are nowhere near as complex as those in the United States, where dozens of races, from president down to local mosquito control boards, can be on one ballot.
In the United States, hand counting is generally used only on small scales — in towns in New Hampshire, for instance, where only a few hundred ballots are cast. But even there, Dr. Ansolabehere’s research found hand counts less accurate than machine counts.
Proponents of hand counting sometimes appeal to history, painting a rosy portrait of secure, well-administered elections before modern technology arrived. But those elections did not seem secure or well administered at the time.
“The counting of paper ballots, often lasting far into the night, and made by tired and frequently incompetent persons, is highly conducive to mistakes and frauds,” Joseph P. Harris wrote in his 1934 book “Election Administration in the United States,” adding, “It cannot be denied that the only way to avoid this tedious job of counting the ballots and to guarantee an honest and accurate count is to use voting machines.”