Proofiness - Charles Seife [52]
By the afternoon after the election, the 725-vote margin had dropped to 477 and then to 465. Some changes were attributed to data entry errors. At one precinct, for example, a clerk had accidentally typed “24” votes for Franken instead of “124.” (Even if clerks were somehow able to count ballots with perfect accuracy, it means nothing if they can’t enter data into a computer correctly.) Absentee ballots were also changing the numbers; some had been included in the initial counts, while others took longer to tally. A week after the election, the margin between the two candidates stabilized at a 206-vote lead for Republican Norm Coleman.
It’s hard to fathom just how unbelievably close a contest that number represents. An election where the two candidates are separated by a percentage point is considered very close. If Coleman were ahead of Franken by 1 percent, that would have meant a 28,000-vote lead. To be ahead by only 206 votes means that Coleman had a lead of about seven-thousandths of a percent—less than 1 percent of 1 percent of the total votes cast. This is a mind-bogglingly tiny margin. If we elected people based upon their height instead of the number of votes they received, this election would be equivalent to having one candidate taller than the other by no more than the width of a human hair.
When “close” means 28,000 votes, hundred-vote, twenty-five-vote, or ten-vote errors (which are surprisingly common in an election this big) don’t make a real difference. A few hundred votes here or there won’t swing the election, so they are ignored—they don’t make headlines. We’re blissfully unaware that they happen regularly. However, when an election is really close—when the margin is a fraction of a percent—those mistakes suddenly make a huge difference. This is why election officials pray for electoral blowouts; big margins cover mistakes, while close ones bring every little glitch into sharp relief.
That is what happened in Minnesota. Because the margin between Franken and Coleman was so small, Minnesota’s electoral errors seemed to be huge, even though they were nothing unusual. When those errors were corrected—and when, inevitably, new mistakes were made—the margin between the two candidates changed by a tiny fraction of a percent. With an election this close, even a few hundred votes is a distressingly large amount.
As Minnesota officials began correcting errors, the margin between Coleman and Franken got narrower and narrower. Though their candidate still held the lead, the Wall Street Journal cried foul, accusing Democrats of hoping to “[steal] a Senate seat for left-wing joker Al Franken.” The beadiest-eyed of Coleman’s lawyers, Fritz Knaak, immediately told the press that the changes in vote tallies were “statistically dubious.” (When lawyers appeal to statistics, you can be certain that there’s proofiness about.)48 The reasoning was that correcting errors shouldn’t change the vote tally by much—mistakes in favor of one candidate should cancel out mistakes in favor of the other. However, this is wrong. Errors can favor one candidate or another—the errors need not behave like coin flips coming up heads 50 percent of the time and tails 50 percent of the time. Because of this, correcting those errors will not necessarily “cut both ways” as the Journal put it.49 In this particular election, there were a few reasons why the errors favored Coleman, so correcting them favored Franken.
One reason was that absentee ballots, which are used by a slightly different population than the people who go to the voting booth, tended to favor Al Franken more than the rest of the state’s voters. Thus, problems specific to counting absentee ballots (and there were many