Letter on Hand Counted Ballots

Submitted to the Des Moines Register on Mar. 13, 2022

Part of the Voting and Elections web pages, http://homepage.cs.uiowa.edu/~dwjones/voting/
by Douglas W. Jones
THE UNIVERSITY OF IOWA Department of Computer Science

The proposal to abandon voting machinery in favor of hand-counted paper ballots reported in Sunday's Des Moines Register (Mar. 13, page 5A) poses significant problems.

Hand counting works well where there is only one race on the ballot. In many parliamentary democracies, you vote for MP, the parliament elects the prime minster, and that's it. Hand counting such ballots is very fast, you just divide the ballots into stacks according to how they are voted and then count the number of pieces of paper in each stack.

In contrast, hand counts of US-style ballots are messy. A small rural county might have 20 races on the ballot, and larger urban counties usually have many more.

By around 1930, the majority of urban voters in the United States were using mechanical voting machines, so you have to look before that to see how well, or poorly, hand counting works.

Clerical errors were common when hand counting was the norm. We still see these errors when votes are processed by hand. The wild swings in Iowa's second congressional district in 2020 that led to a recount were all attributed to clerical errors. Machines were used to count the ballots in that race, but combining the results from multiple precincts and counties was not fully automated.

Outright fraud attracts the most attention. In the days of hand counting, tally clerks sometimes parked pieces of pencil lead under their fingernails so they could mark on ballots they disliked. The mark didn't even need to look like a vote because most states have laws that disqualify ballots having "identifying marks."

An even more subtle approach was for biased election workers to demand strict enforcement of disqualifying rules for ballots they disliked while being lenient about accepting ballots they liked. This leads to a spiraling arms race where partisan vote counters disqualify an ever increasing fraction of the ballots until the election is determined by who can disqualify the most ballots instead of by how the voters feel.

It takes armies of biased ballot counting clerks to corrupt a hand-counted election, but our political parties have such a armies. When these armies are devoted to campaigning, get-out-the-vote work and poll watching, it is good for democracy. When they are put to work on manipulating the vote count, it destroys democracy.

Problems like those described above led election reformers in the US to demand mechanical voting machines a century ago. They hoped that voting machines would break the back of political machines.

They were wrong. It turns out that moving vote counting from the hands of precinct election workers to voting machines moved the integrity of the process from the army of precinct workers to the voting machine technicians. With the move to computerized electronic ballot counting, the technicians in the county building are no longer as important as the programmers working for the voting machine vendors.

We know, from numerous studies, that the electronic ballot counting machines currently on the market are vulnerable to various attacks, but when they are operated properly, they count very accurately.

Back in 1965, the state of California addressed this problem by requiring a hand-counted audit after each election. The California model was to do a limited hand count to check the honesty of the machine count. They selected precincts at random to cover one percent of the ballots cast and then audited those. No other state did this until the controversy surrounding the 2000 Bush versus Gore election. Today, a majority of states, including Iowa, require some kind of post election auditing.

In the decades after 2000, we've developed more sophisticated audit models. When the margin in an election is large, you don't need to hand count many ballots to check the result. When the margin in the election is small, though, one percent is not nearly enough. The risk-limiting audits pioneered in Colorado and Rhode Island addresse this by hand-counting more ballots in close elections and even more when discrepancies are found.

There is plenty of room to improve how most states audit elections, but I believe that the combination of machine counted paper ballots with rigorous post-election audits is far more likely to give us an accurate and trustworthy vote count than the complete abandonment of machine counted ballots.

Douglas W. Jones is a retired professor of computer science at the University of Iowa. He was chair of the Iowa Board of Examiners for Voting Machines and Electronic Voting Systems in 2000 and is co-author of the book "Broken Ballots" (CSLI Press, 2012).