Nate Silver: The Polls Aren’t Wrong
Stay up to date with the latest headlines via email.
This story was originally published at Salon.
We are drowning in polls and predictions. Whether it’s politics, sports, economics or even the weather, there’s more information and data than ever. But how much of it is white noise? How many of these predictions have rigor and mathematics behind them, and how many mask uncertainty or ideology behind the seeming exactitude of numbers?
In 2008, Nate Silver built a near-perfect model for analyzing the polls at his web site Fivethirtyeight.com. Silver called Obama over McCain in March — and ultimately nailed 49 of 50 states, got every Senate race right and predicted the popular vote within a percentage point. That’s the kind of predictive power we all dream of when we fill out NCAA tournament brackets or a lottery ticket. The New York Times added his blog to its site soon after the election. The numbers geek started a bidding war.
In his new book, “The Signal and the Noise: Why So Many Predictions Fail — But Some Don’t,” Silver tries to explain the secret to getting things right — and finds that it often turns on blind ideology and overconfidence. (He’s gotten things wrong himself; in the book, Silver admits that his much-touted baseball statistical analysis system, called PECOTA, actually fared worse than the collective wisdom of the old-school pro scouts, much-maligned in the “Moneyball” era. Hey, he still totally got Dustin Pedroia right.)
And while it’s not that important, in the end, to predict the state-by-state Electoral College winners, Silver sits down with national security and economics forecasters, as well, who have similar jobs as him in some ways — searching through the clutter and distractions to find the information that matters.
As national and state polls began swinging in President Obama’s favor this week and Romney partisans began trying to discredit the objectivity of the results, Silver met with Salon in a coffee shop downstairs from the New York Times to talk insider political baseball. (The interview has been edited and condensed.)
Fox News and Romney supporters this week have turned on the polls. As more and more of them show the president opening up a wider lead, both nationally and in many of the major swing states, they’re essentially charging that there’s liberal bias both in the way these polls are conducted and in the way they are reported. Is there any truth to that?
At that point, you literally are alleging a conspiracy, right? And you get these headlines on Fox News like: “Poll is biased toward Obama.” I think they maybe make too much of that on the conservative side.
I saw Dick Morris on the “Hannity” show last night. He wasn’t just saying Romney still has a chance; he was saying it’s a toss up, which I don’t quite believe. It’s getting a little more ridiculous the more polls that come out. But he was saying, “I think Romney will win by four points. I think he’ll win Pennsylvania and would be competitive in Michigan.” You have to be totally delusional to think that. Is he out of touch with reality? Or is he lying?
I mean, there is literally a web site right now that is devoted to the premise that the Rasmussen poll, which tends to ( favor Republicans) much more than anything else …
Right, Unskewedpolls.com …
It’s the purest example of a selective reading of the evidence that you can possibly find. Rasmussen has the race tied right now, I think. And Unskewedpolls has Romney up eight, right? When I used to play poker, if someone is that out of touch with reality, at least you could bet them money. At the same time, sure, the polls could be skewed. But they could be skewed in either direction.