As we know, modern democracies are founded on free, secret and fair electoral processes. In order to guarantee that their results reflect the will of the voters, elections are subjected to careful scrutiny by multiple actors before their opening, during their course, and after polling stations close.
Further audits usually comprise manual counting of a representative sample out of the total counted votes.
Over time, dozens of manuals have described how to carry them out. National and international observers examine these processes exhaustively in order to make sure that the official results reflect the will of the people. However, one might ask why a manual process carried out by human beings who make mistakes by nature can be taken as a reference to contrast the reliability of the results. Even worse, with the tremendously high levels of fairness and accuracy reached by electronic voting technologies such as those developed by the Smartmatic company, why grant such reliability to a manual process? Must a manual process be considered indisputable evidence?
A recent research carried out by Stephen N. Goggin, Michael D. Byrne, and Juan E. Gilbert, PhD candidates and professors from important North American universities, titled “Post-Election Auditing: Effects of Procedure and Ballot Type on Manual Counting Accuracy, Efficiency, and Auditor Satisfaction and Confidence,” offers valuable evidence, both qualitative and quantitative, of how commonly used auditing mechanisms aren’t necessarily devoid of errors. Using diverse scientific methodologies, the researchers analyzed different group counting techniques and the kinds of ballots used by certain voting systems, and they concluded that these could have a negative influence on the accuracy and efficiency of audits.
In order to analyze the level or error in manual counting, two methods of group counting called “read-and-mark” and “sort-and-stack” were used. Four auditors were used for the first—one to read the options marked in each ballot, a second one to confirm the reading, and two others marked what had been read. For the second method, all the ballots were separated into different stacks, one per candidate, and another one was created for the undervotes (ballots where not enough options were marked),plus one for the overvotes (ballots where too many options were chosen). This method used three auditors who did their own counting individually and then checked itwith the other auditors.
The study also analyzed different types of ballots: those used by some optical scanners, and those produced by voting machines that print paper trails. One problem often found with optical scanner ballots is the discrepancy between what the machine interprets as a valid vote and what the auditor interprets as a valid vote.
The study proved that inconsistencies arise even in completely controlled environments. Discrepancies were revealed between the two counting methods. Besides, when ballots are being counted, no matter how simple the task may seem, people tend to count a single ballot repeatedly, leave some ballots uncounted, or misinterpret the voter’s will.
Anyway, if further auditing is the process through which definite legitimacy is granted to an election, it becomes necessary to ask: Who audits the audits?