Post-Election Auditing Effects of Procedure and Ballot Type on Manual Counting Accuracy, Efficiency, and Auditor Satisfaction and Confidence

Election Law Journal (2012-03-05) Stephen N. Goggin, Michael D. Byrne, and Juan E. Gilbert

This Page: https://copswiki.org/Common/M1725
Media Link: https://copswiki.org/w/pub/Common/M1725/GogginByrneG_12.pdf
Remote Link: http://chil.rice.edu/research/pdf/GogginByrneG_12.pdf
More Info: Election Integrity

They conclude that:

"Overall, this study provides valuable quantitative and qualitative evidence that manual post-election auditing is not an error-free process. Depending on the procedure used, as well as the type of ballot counted, manual audits can vary in their accuracy and efficiency, as well as their appearance of validity to the auditors and outside observers. While many argue manual audits are the ‘‘gold standard’’ by which we must evaluate computerized ballot totals due to the insecure nature of such machines, we must be careful to remember that even the most basic tasks performed by humans can and do introduce error into the process."

Table 1 of the paper provides the "Total Candidate Error Rate," which ranges from 0.48% to 2.13%.

Accuracy

Overall, 40.0% (SE=6.4%) of groups provided an incorrect total number of valid ballots, and 46.7% (SE=6.5%) of groups provided an incorrect count for at least one of the four candidates. The average error percentage for the total number of valid ballots is 1.2% (SE=0.28%), and the average error percentage for candidate counts is 1.4%(SE=0.30%).

Timing

They tallied only two races on the ballots, 120 ballots. The first race took additional time than the second race. Times are net only for this activity and do not include rest breaks and other overhead of actually doing the work entailing an actual audit. Nevertheless, we can consider the time for the second race tallied (14 minutes) to be somewhat of an upper bound for 120 ballots. Thus, we have 14 min / 120 ballots = 7 seconds per ballot-race.

• GogginByrneG_12.pdf: Goggin Byrne G_12.pdf

Comments