Nate Silver's New YorkTimes blog, fivethirtyeight.com, is an island of rigorous thinking in an ocean of political spin. A statistician who writes with engaging clarity, Silver derives his analysis from his own mathematical models. He has just published his first book, The Signal and the Noise: Why So Many Predictions Fail — But Some Don't (Penguin Press). It's a guide to thinking about the future, a primer for the intellectually engaged. I spoke with the modestly confident and energetically curious Silver by telephone last week.
What role does intuition play in your work? What's the ratio of math to art? Art is a pretty broad term. My work is a type of science. For presidential elections it's a difficult science. You only have 16 elections since World War II. And of those elections, only the last eight or 10 have the rich, state-level poll data that fuels our model. In general, people are too often seduced by their intuition. Our first instinct tends to overrate the newest piece of information. When a new poll comes in, for example, people tend to ascribe much more meaning to it than they should.
Political professionals and journalists seem awfully damn sure of themselves these days. The larger climate of opinion, however, is uncertain, anxious, and apprehensive. One of the characteristics in the book that I associate with people who make poll predictions is that they are too certain. We think we know more than we do. And we think our perspective is the only one out there. That perspective is often wrong. My thing is to weigh in on where the bulk of the evidence points.
How do you feel when one side or the other distorts your work for immediate impact?(Laughs.) I kind of love the larger process. But let's just say sometimes I do not love individual campaigns.
Since the 9/11 terrorist attacks and then the financial meltdown, folks like you have gotten so much better at interpretation. But it seems the more we know, the less we understand. That's part of the reason I wrote this book. Information increases at an exponential rate, but over the last decade we've gotten a lot of things wrong. People have a vast menu of information, but they're not always consuming it in a way that gets them closer to the truth.
Is there anyone on the right who you find a useful corrective to what I assume is a progressive predilection on your part? I'm actually more liberal to libertarian. Sean Trende of Real Clear Politics is quite good. He's fair and thoughtful and has a lot of grounding in history. Michael Barone is very good with data and really knows every corner of the country. I'm not sure they are on the right. Center-right, maybe. These guys can separate their rooting interests from the hard data. They can step back, or step deeper.
Name some non-quantitative journalists whose work is nourishing? For "horse race" coverage: Dave Weigel at Slate and Jon Heilemann and Jonathan Chait at New York magazine. They're smart and they tend to have read the literature. What I read everyday are the political scientists who blog. The Monkey Cage is a good example. I tend to favor people versed in empirical data, not so absorbed in the Boston-New York-DC bubble. I have to say, I struggle against that myself. During the primaries, I went to a debate in New Hampshire. Every member of the campaign press corps was in a giant gymnasium, sitting next to one another, watching the debate on a screen. There's going to be group-think when a whole group thinks together.