Tomorrow is Election Day, so get out there and vote. Barring any major polling malfunctions, by the end of the day we’ll finally have an answer to the question of who will reside in The White House for the next four years. Almost as interestingly, tomorrow could also mark a definitive change in the way we look at Presidential campaigns, potentially for decades. In particular: if Nate Silver’s ongoing, deeply statistical analysis of the race at Fivethirtyeight turns out to be an accurate predictor of the final outcome, it may alter political punditry for a long, long time.
If you’re not familiar with Silver’s work, it’s probably a reasonable if gross characterization to say that he is a kind of ‘meta-pollster.’ Each day, he surveys the most recent state and national polls, aggregating their results using a sophisticated — but proprietary — statistical model that accounts for such factors as polling methodology, past accuracy and tendency to favor one party or another. The result is what some believe to be an exceedingly accurate picture of who is ‘winning’ at any given stage of the campaign — and, of course, a prediction of who will actually win at the close of Election Day.
Silver began doing this work in the lead-up to November 2008, and produced eye-popping results. His model correctly predicted the winner of forty-nine of the fifty states in the presidential election, and all thirty-five of the senate races held that year.
Whether that was pure luck or not is the question that will be answered when the results of tomorrow’s election are in. If his predictions are largely accurate, it will go a long way towards validating Silver’s approach. It’s my feeling too that if that happens there’s no going back; in at least the next few election cycles, you can expect to see much more attention paid to this sort of statistical evaluation of a campaign’s progress.
Continue Reading +