Proofiness - Charles Seife [55]
Sometimes a miscount made votes materialize out of thin air. One precinct in Rice County tallied twenty-five more votes than voters who showed up on election day. Fran Windschitl, the election official in charge, suggested that a simple counting error was to blame: “Our accounting procedures dictate that you arrange them in piles of twenty-five; they must have counted one too many piles,” he said. Sometimes a miscount caused votes to disappear. In a precinct in Blue Earth County, there were twenty-five fewer ballots than voters who showed up to the polls. Worse yet, a careful look at the data implied that the missing votes had all been cast for Norm Coleman.52 (A number of the miscount errors seemed to happen in bundles of twenty-five, which can easily be explained by missing or double-counting one of the crosswise stacks of twenty-five ballots described in Minnesota law.) Sometimes the errors weren’t a simple miscount, but had a more complex explanation. All around the state, absentee ballots, particularly those that had been duplicated so they could be fed through the scanning machines, were causing troubles. Sometimes the originals disappeared, and sometimes the duplicates went missing—potentially introducing errors into the final count. All told, there were oodles of miscounted ballots. Hundreds and hundreds of votes had vanished and hundreds of others had materialized. Some of these errors were in favor of Franken, some were in favor of Coleman, and some were a wash.53 Many of these errors were corrected by the time the recount ended, but many—the number is probably in the hundreds—were not.
Counting errors can’t be prevented; they can be reduced somewhat, but they can never be eliminated completely. Even under the absolute best conditions imaginable, counting and tabulation errors still occur. The Minnesota voting data prove it. Errors exist no matter how small the number you’re counting. People can make errors counting on their own fingers. Okay, the error rate is small and they might not do it very often, but they do it, and reasonably often if they do the same task over and over again all day. (Witness how often people will hand you the wrong amount of change because they counted out pennies incorrectly.) Some of the precincts in which errors were found were quite small; one where the audit mismatched the recount only had about 260 voters.
Minnesota’s electoral law requires that there be a post-election audit. Before the election, a few hundred precincts around the state are randomly selected54 to perform a hand recount of their votes to make sure that all the counting equipment was working properly. The law dictates that these recounts must use the same procedure as a full-on recount in the case of a close election. Multiple observers go through the votes as carefully and deliberately as humanly possible, trying as hard as they can to ensure that they make no errors at all. This is the ne plus ultra of counting ballots by hand. Then, because the Minnesota Senate race was so close, the very same precincts had to participate in the recount, counting those ballots again—by hand, as carefully and as deliberately as humanly possible, again ensuring that they make no errors. In these precincts, the same ultra-precise counting procedure was repeated twice. And even under these stringent conditions, the numbers didn’t match up perfectly.
The errors were small, to be sure; everyone did a marvelous job preventing them. But errors there were. In the roughly two hundred precincts subject to the post-election audit, about fifty votes changed hands between the two counts. This meant that about forty-thousandths of a percent—0.04 percent—of the votes in the audit had changed between one ultra-careful count and another ultra-careful count.55 This level of error wouldn’t be a problem if Coleman were beating Franken by 1 percent or even half of a percent. But when the margin between the two candidates is seven-thousandths of a percent—0.007 percent—suddenly even this tiny, tiny error