Nate Silver: Stats whiz remakes political punditry
Sure, President Obama barely showed up at the first presidential debate earlier this month. But if he felt complacent, there might have been a reason: On Oct. 3, the day of that fateful debate, statistics whiz Nate Silver was predicting that Obama had an 86.1 percent chance of winning the race. Silver’s assessment helped fuel Democrats’ giddiness about Obama’s seemingly inevitable victory.
Silver, 34, is a baseball statistician who made his name in politics four years ago, when his poll-aggregation blog, FiveThirtyEight, showed that a lot of traditional political handicapping was basically worthless. He noted that, for instance, some polling firms have a better record of predicting outcomes than others, and that a presidential candidate generally doesn’t surge in one state without doing the same in others with similar demographics. He built these and other factors into his proprietary computer model, which went on to call 49 states correctly in the 2008 general election.
Since then, Silver has come to symbolize a more discriminating way of using polls — one that discourages a candidate from freaking out simply because her numbers in today’s Rasmussen poll are worse than in yesterday’s Quinnipiac. His blog now appears on The New York Times’s website. He has a new book out, “The Signal and the Noise: Why So Many Predictions Fail — but Some Don’t.” His influence may be most evident, though, in the daily habits of political junkies, who check his site obsessively for new clues about how this year’s presidential race will turn out.
Yet as Mitt Romney’s numbers steadily rose after the first debate, a gap opened up between Silver’s predictions and what some venerable polls seemed to be foretelling. After Romney cracked the 50-percent mark in the Gallup poll of likely voters last week, GOP operative Karl Rove crowed that no candidate performing at that level on that poll in mid-October has ever lost the presidential race. Yet on Thursday afternoon, when Gallup showed Romney up by 7 points, and the poll averages on other websites gave him a thin lead as well, Silver’s FiveThirtyEight still gave Obama a 65.7 percent chance of victory. (To be fair, the prediction market Intrade showed similar odds.)
Silver’s computer model will no doubt catch up if Romney’s momentum holds, but in the meantime Twitter was abuzz late last week with suggestions that maybe people should believe Gallup after all. Nate Silver can provide only so much certainty, and even an 86.1 percent probability isn’t destiny.