, 19 tweets, 5 min read Read on Twitter
1/Lots of heat/traffic setting up opscan vs hand-count debate. I claim hand counts are impractical in US elections. They're error-prone & like all tabulations must be checked with audits. They offer few benefits at great cost. Literature is sparse/scattered. Here’s my rationale
2/Hand counts are carried out by people, so it is fair game to ask how often they get it right. The answer appears to be “not very often.” Like many aspects of voting that at first blush seem simple, this is a cognitively complex task.
3/This has nothing to do with how much we want them to do it right, or whether we are too impatient on election night. Young, trained, alert humans under ideal conditions are not good at counting ballots because of inherent cognitive limitations. Older, tired humans....well...
4/I want to focus on the accuracy of hand counting. There is a wealth of anecdotal evidence (“Jurisdiction X did hand counting and it turned out ok.”) in these twitter streams but anecdotes are not data. We don’t know that the hand counts were accurate and that things were ok.
5/In a study by Goggin, Byrne, and Gilbert, ballots were counted by small teams well-versed in disciplined team methodologies for conducting hand counts. Each team counted two two-person races out of 27 ballot contests.
6/Teams were younger and better educated than average. They were asked to count relatively few ballots. 36% had a college degree. Many had graduate degrees. The average age was 59 yrs. The total number of ballots counted was 120.
7/By comparison, 56% of poll workers in 2016 were 61 years old or older. Education not recorded nationally but 64% of jurisdictions report difficulty in recruiting poll workers. Average precinct size is around 1,100 voters.
8/40% of the groups incorrectly counted the number of valid ballots. 46% incorrectly counted votes for at least one of the four candidates. Hard to measure performance directly so one approach is to look for simpler tasks that necessarily must be completed to hand count ballots.
9/How do people perform at simpler tasks? The average precinct counts about 2 reams of paper. Simply counting the number of sheets in DIY experiments leads to varying hand counts but rarely to 1,000.
You can do this experiment at home. Try it.
10/Even if the page count is right, human baseline error rates for hand counting is about 5 per 1,000. Consistent with Goggin et al study.
11/The age of the counters and time of day both affect performance. Gather your favorite aging family members (that'd be me) and do the Minnesota DIY experiment three times: morning, noon, and night. Watch to see when boredom creeps in (for my group around 400 sheets of paper).
12/ There’s an extensive literature on how quickly cognitive capacity gets used up. Best survey is Daniel Kahneman’s “Thinking Fast and Slow” (especially the “add three card game” experiment) us.macmillan.com/books/97803745…
13/Errors rise at the end of the day (when votes are counted) as cognitive capacity gets used up.

and books.google.com/books/about/An…
14/In the Goggin et al experiment, the time (ideal conditions) was 10-20 minutes per group, so maybe 270 total person-hours per precinct. There are 3,000 precincts in Georgia, so a total effort of 800,000 person hours, an incremental cost of elections that is not affordable.
15/Not counting training and group management. Hand counting only works when the groups execute the same process in lockstep. chil.rice.edu/research/pdf/G…
Uniform execution of written procedures not a strong point (ref. 2007 Georgia Tech study)
16/Is any of this conclusive proof that hand counts are impossible? No, but that’s not the standard. I have not seen a precise hypothesis that would support the practicality and accuracy of hand counts testable by simple experiments like these.
17/If you know of any, let me know and we’ll invite you to @TheBradBlog party. Maybe explain 2 me the fascination with hand counts. If you believe hand counts provide some benefit, fine but they're no silver bullet.
18/ In the meantime, the best approach is to require RLAs to check the accuracy of whatever your tabulation approach is. BTW -- nothing in this discussion affects the superiority of hand MARKING paper ballot.
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Richard DeMillo
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!