I have truly enjoyed following Nate Silver and FiveThirtyEight this political season. I’ve loved following the push and pull of the numbers against each other as the campaign went on, and Mr. Silver’s explanations of what went into polling and his particular methodology of combining polls.
The hagiography has begun in numerous articles about Nate Silver and other fellow “quants”. But Nate Silver did not predict the election. He crunched the numbers from a variety of polls, interpreted according to a particular methodology, which gave him the confidence to predict, day by day, not that Barack Obama would win the election on Nov. 6, but that if the election were held today, and the polls he aggregated were correct, Barack Obama would win.
A blurb to his book, The Signal and the Noise, contains the prediction paradox: “The more humility we have about our ability to make predictions, the more successful we can be in planning for the future.” As I followed his blogs, I also followed the comments from readers, especially from the Romney supporters and their wholesale dismissal of Mr. Silver’s analysis. They ranted that the polls were biased, assumed too great a Democratic turnout, and didn’t take enough into account the Romney momentum and that the nation was simply fed up with the last four years.
And Nate Silver agreed 100% with his detractors (actually 16%). In a brilliant column on Nov 2, Mr. Silver echoed everything his detractors were saying. If Romney wins, it will be because the polls are wrong (Well, duh!). If the polls are wrong, it will be because of bias. Maybe the sample was not representative, or maybe the pollsters mispredicted who were likely voters, or maybe the pollster’s processing of results to fit the sample to the expected universe of voters was wrong. It was interesting to read the Romney supporters ranting at Mr. Silver … when he was agreeing with them! On that day, he gave Romney a 16% chance of winning the election, based on the fact that the polls might be biased.
Some of the articles lauding Mr. Silver expressed a bit of dismay that maybe people don’t count, just the math does. But this misunderstands what Mr. Silver and his compatriots do; they predict how people are going to vote by asking them who they are going to vote for!
Mr. Silver’s results, and those of other pollsters and poll aggregators, are only as reliable as the polls themselves, and Mr. Silver’s success came from averaging polls which were themselves rather all over the map. And the reliability of polling going forward is in question. Scott Rasmussen, whose Rasmussen polling results were not too accurate, maintains that this will be the last election for phone polling, and that new techniques will need to be developed and tested. But while this may be raining on Nate’s parade at the height of his success, I’m sure he would be pleased to know that’s it’s all going to get even more interesting!