Years ago, fresh out of a statistics course, I made some friends laugh by insisting that an election shouldn't be considered valid unless the difference between the vote counts is statistically significant. I was viewing an election as an experiment with the hypothesis that more of the voters wanted candidate A than candidate B. I didn't pursue the idea because a vote count seemed to me to be a fixed result. How could you measure variability without holding the election repeatedly day after day? Was I naive! I knew nothing then of chad or dimpled ballots, and I hadn't considered what it would really be like to count millions of ballots.
Nevertheless, when statisticians attracted the spotlight in the presidential election this year, I felt vindicated. While some calculated that many people had made mistakes in marking a butterfly ballot and others estimated how various recounts might change the tallies, a few made just my point. For example, mathematician John Allen Paulos