Comparing the Auditability of Optical Scan, Voter Verified Paper Audit Trail (VVPAT) and Video (VVVAT) Ballot Systems

With many states beginning to require manual audits of election ballots, comparing the auditability of different types of ballot systems has become an important issue. Because the majority of counties in the United States are now using either Direct Recording Electronic (DRE) voting systems equipped with Voter Verified Paper Audit Trail (VVPAT) modules or optical scan ballot systems, we examined the usability of an audit or recount on these two systems, and compared it with the usability of a prototype Voter Verified Video Audit Trail (VVVAT) system. Error rates, time, satisfaction, and confidence in each recount were measured. For the VVPAT, Optical Scan, and Video systems, only 45.0%, 65.0% and 23.7% of participants provided the correct vote counts, respectively. VVPATs were slowest to audit. However, there were no meaningful differences in subjective satisfaction between the three methods. Furthermore, confidence in count accuracy was uncorrelated with objective accuracy. These results suggest that redundant or error-correcting count procedures are vital to ensure audit accuracy.