On Hand-Counting of Ballots
It's a really bad method
I took the above photo in March of 2010 in Bogotá, Colombia as an election observer. It shows the poll workers (who are citizens basically serving a kind of jury duty) hand-counting the ballots in congressional elections. Each table has a ballot box (indeed, multiple boxes for different offices) and voters are assigned to a specific table. I forget the max number of voters per table, but there are a lot of tables.
For example, here’s a blurry picture of a mall parking garage set up with dozens of tables:
Here’s what a given table would look like (from the 2014 presidential elections, which I also observed):
I would note that the process described above is a pretty old one, and hardly represents the state of the art in running elections. I am not certain, but think that this is still the process in Colombia. It was fascinating to observe, but it is hardly what one would call best practices,.
All of this comes to mind because of this NBC news piece: Arizona Republicans wanted to hand-count ballots. Then they saw the price tag — and the errors.
In June, the Mohave County’s Board of Supervisors asked the county elections office to develop a plan for tabulating 2024 results by hand. Secretary of State Adrian Fontes, a Democrat, warned Mohave’s county supervisors in a June letter that they risked breaking the law if they chose to opt for hand-counting in a future election. A lawyer for the county told supervisors before they voted that the county’s legal team wasn’t sure it was legal, either.
The test run took place in late June, when elections workers spent three days hand-counting a batch of 850 test ballots from the 2022 election, bringing in seven part-time staffers eight-hour days of counting and four full-time staffers who monitored the process.
Elections Director Allen Tempert told the Board of Supervisors at a Tuesday meeting that the group was a “dream team” of experienced staffers, but the feasibility study nonetheless went poorly.
There were counting errors in 46 of 30,600 races on the ballots, as the team tallying the results of the election made mistakes. According to a report prepared for the Board of Supervisors, some of the observed errors included: bored and tired staffers who stopped watching the process, messy handwriting in tallies, fast talkers, or staffers who heard or said the wrong candidate’s name.
Each ballot took three minutes to count, Tempert said. At that pace, it would take a group of seven staffers at least 657 eight-hour days to count 105,000 ballots, the number of ballots cast in 2020. Mohave County would need to hire at least 245 people to tally results and have counting take place seven days a week, including holidays, for nearly three weeks. That estimate doesn’t include the time needed for reconciling mistakes, or counting write-in ballots, Tempert’s report added.
[…]
The total cost for the staffing, renting for a large venue for the counting, security cameras, and other associated costs was staggering: $1,108,486.
All of this should have been obvious: machines count faster and more accurately than humans. And unless you are going to go the Colombian route and basically draft hundreds, if not thousands, of poll workers, the machines are cheaper, too.
Every time I hear someone wax poetic about the wonders of hand counts, I think back to dropping a deck of playing cards all over the floor and then counting to make sure you got them all. 53? That’s not right. 51? Aarrgh!! Or just kowning the difference between hand-grading a scantron form versus letting the machine do it (spoiler: the machine is better and faster).
Back in 2010 when I was watching the Colombian poll workers sit on the floor counting I observed a few errors, all of which were about applying the rules for when a ballot should be counted as spoiled or not, although overall the process went relatively smoothly and the overall tally took less time than you might think. But, again, thousands of workers across the country were involved. Plus, it should be noted, that in this election each ballot was for one office and one office only and the number of offices being contested was somewhere in the mid-single digits. Compare that to a US ballot that might have a dozen or more offices that need counting.
So, yes, hand-counting is a bad idea, and machine counting is far superior, in case anyone was wondering. And the fact that so many Republicans now have it in their heads that it is a good idea is just one other reason that Trump has been bad for the country and our politics.
It’s one of the hallmarks of a cult;
What’s odd, though, is that, in very close elections, we “recount” the work done by the superior machines via an inferior hand count. ¯\_(ツ)_/¯
@James Joyner: IIRC, recount procedures vary based on the type of voting and jurisdiction. I think often they are machine recounts.
But yes, doing a hand recount is problematic for the reason described (although I would note that a recount would focus on only one of the elections on the ballot, not all of them).
Can someone explain this line: “There were counting errors in 46 of 30,600 races on the ballots,” It makes no sense at all.
I keep saying that for A-Republic-Not-A-Democracy, America has waaaaaaaaay too many elected offices, all but one based on a 50%+1 majority vote (and that only where a plurality won’t do). Add in local races, and those ballots are simply insanely complicated for a hand count. I don’t know how people managed it in the old days (I do know tabulating machines go farther back than most people seem to think).
It’s painfully obvious to any of us who have ever worked elections that hand-counting is a dumb-as-sh!t idea and introduces far more inaccuracy into the mix than machine counts.
So much of our current election mess has to do with people who are too dang stubborn to realize how utterly wrong they are. It’s maddening.
@Steven L. Taylor: Hand recounts can also be limited to testing the quality of certain machines by spot auditing groups of ballots.
The point about it recounting only on one election is what makes it viable. What is ignored with this is that while hand recounts have sometimes made very small differences they have not been found, anywhere I can find, in large enough numbers to make up for the thousands needed to change the votes so Trump would have won.
Steve
My 2022 ballot had more than 60 individual items once federal, state, and local offices were considered, plus state and local ballot initiatives, and a couple of referendums from the legislature(s). I don’t want to think about what condition the ballots would be in on the 60th pass through someone’s hands.
In my county there were more than a dozen types of ballot reflecting different overlapping districts.
@steve:
One has to take into account the number of votes, and the fact that counts tend to be very accurate to begin with. Other than races with a smallish number of voters, or ones that turn out to be incredibly close, no recounts stands a chance of reversing the original count’s results.
I wonder if there’s a mathematical formula for a number of votes threshold.
By incredibly close, I’m talking dozens of votes, at most, out of at least tens of thousands. the odds of that are slim.
Mohave County has a population a bit over 200,000. Serendipitously Political Wire today quotes Larry Sabato,
And adds, as one would expect,
I can’t see Bronx County, or LA County, or Cook County even thinking about hand counting.
But at least hand counting can’t be reprogrammed by Italian military satellites controlled by the late Hugo Chavez. It’s fascinating psychology to watch the deeply held, carved in stone, conservative verities shift around. FL instituted large scale mail in voting years ago. Partly in the fallout from 2000, but mostly because they noticed absentee ballots ran heavily Republican. Now for a few years mail-in ballots are a tool of Antifa, or BLM, or something worse. In a few years, FOX/GOP, having recognized they shot themselves in the foot, will re-normalize mail-in. And at every turn good GOPs will have always believed whatever it is they believe at the moment and the recent anti-mail in push will have been Dem fake news they never believed.
@Jen: Literally some of these people never voted before 2020 and had (and still have) no idea how voting works. Period. It was clear to see in the whining and false alarms they did over perfectly normal election-night practices.
And as usual for the Trump Cult, they want two completely opposite things: a hand count accomplished within 24 hours of the polls closing.
Good luck with that…
I have a friend that is irritated by the pattern he observes that liberal ideas often show up five or ten years down the road as conservative ones, though maybe with a twist.
In the early Oughts, I was very worried about computerized touch-screen voting machines because they have no auditability. Paper ballots that are machine-readable seem a much better choice. Conservatives at the time were all “what are you even worried about?” They finally figured it out, but it’s almost a parody of my concerns.
@James Joyner:
True. WAPO and NYT like to launch deep dives into the squeaky clean Clinton Foundation or counting Trump’s lies. ( I bet they regretted starting that last. How many interns do they have?) I wish one of them would do a deep dive on vote counting. People seem to have such ridiculous ideas about how it’s done. My understanding is that, post the Diebold machine issues, almost nobody does paperless voting. The Dominion and Smartmatic machines are mostly optical scanners counting hand marked paper ballots. Which are retained for recount and audit purposes. And that post-election at least sampling audits are routine. Checks on number of ballot reported cast are compared to numbers counted. The machines are kept tightly controlled, such that they have to be replaced once the chain of custody has been broken by the GOP bozos who took some. I keep seeing experts saying voting in this country is now the most honest it ever has been. The Italian satellite vote flipping by machines and garbage bags of ballots stories are ridiculous.
@gVOR10:
Who needs any of that in this era of Cuban sonic microwave weapons manipulated by the late Hugo Chavez?
Really, they need a replacement odious foreign socialist dictator, now that Kim is the Cheeto’s love interest.
@Michael Cain:
Same with my ballot.
That’s why I so strongly support vote by mail like we have in California. I can sit with my ballot for as long as I need to research candidate positions when I don’t know, to read the voter’s guide, to look deeper into ballot initiatives, or even talk something out with my sons. Then it is very simple to securely mail the ballot in or drop it off at polling place on Election Day (or sooner).
Without question, vote by mail ensures that I am voting my intent more thoroughly and accurately than I could possible achieve in a voting booth. I don’t see the arguments against that approach.
From the NBCnews article
Yet they’ll still want a hand count, with the reason being that it will be easier to cheat.
@Scott F.:
It’s become a regional thing. Next year, in the 13-state American West, >90% of all ballots will be distributed by mail. CA, CO, HI, NV, OR, UT, and WA mail ballots to all registered voters. AZ has a permanent no-excuse mail ballot list used by, IIRC, >80% of registered voters. The MT permanent no-excuse list is used by >75% of voters.
At one time there was some sentiment to bring back telephone operators to place calls “number please” and do away with dial phones and automatic switches. There was an idea that this would create jobs.
Telephony was a trade magazine that I subscribed to in the ‘80s. An article they ran then stated that there weren’t enough humans on the face of the Earth to manually switch all the telephone calls made in 24 hours.
“I’m sorry. That line is busy. Please try again later.”
We need to introduce the concept of statistical significance to vote tabulation such that a repeat vote is required when the margin between candidates is less than the bounds of error for the tabulation mechanism. I’d be happy with 95% confidence but would be Ok with a 3-sigma limit.