Polling wizard Nate Silver has just released his first book, The Signal and the Noise: Why So Many Predictions Fail — But Some Don't, in which he examines everything from politics and weather forecasts to chess-playing computers and the stock market. We talked to Silver today about the race for president, why he doesn't play fantasy baseball anymore, and what makes Dick Morris so awful.
Has anything surprised you about presidential polling so far this year?
Well, I don't think it was necessarily expected that President Obama would get the lift he did out of his convention. Candidates typically do get some sort of convention bounce, but Obama's has persisted now for three weeks to the point that I think you have to term it a shift in the conditions and not just a temporary bounce ... We try to quantify the uncertainty, things could have gone in either direction, but if Obama does win by 5 or 6 points — and that looks possible now — that would be a little bit surprising, I think.
You write in the book about the failure to predict 9/11, and you, echoing the 9/11 Commission, write that a lack of imagination was one of the most important factors. Is there anything that you, or the political class as a whole, might be failing to imagine as you attempt to predict the outcome of the presidential race?
There's always that risk, right? There's always the risk that there are unknown unknowns ... But between the fact that there are very few undecided voters, and that [Obama] seems to lead in all or almost all of the swing states, if you had the election today, we think he'd have a 97 or 98 percent chance of winning it. There's not too much doubt. But yeah, there can absolutely be October surprises, and there have been in the past. I mean Europe, for example, is something where investors are becoming more nervous again. What happens in the Middle East, what happens in Europe — those kind of things are known unknowns, and there are also economic and foreign policy risks that can occur. But I think the question now — and I think we're close to it — is: Are we at the point where you have to have something like that occur for Romney to win?
So, are we?
I think we're pretty close to it. Most of his 15 or 20 percent chance, or whatever we're giving him ... does reflect at this point the kind of October surprise scenario.
So, basically, barring an October surprise, Romney's chances are more like nil?
Not nil, but slim. So the debates for example have proven to move the number by about 3 points in the past. So, for example, in 2000, Al Gore had about a 3-point lead against George W. Bush at this point in the election and was perceived to have lost the debates — I guess he sighed too often or something — and that was enough to reduce his lead and turn the election into what was almost literally a tie.
The problem now is that there are a lot of ways that Romney could kick a field goal, but Obama seems to be ahead by a touchdown right now. So, I guess he could kick two field goals, to stretch the metaphor out. Or maybe if Romney did pretty well in the debates, and you had a couple of bad but not extraordinarily bad jobs reports, he could put two or three things together, and then they get a strong turnout. So that's why you can't quite say that he can't come back organically, if you had a cluster of different medium-sized things happen, to get some momentum. But I'm not sure that, say, the debates alone are enough to move the numbers for him. I think he needs that and something else, or he needs some kind of acute crisis to develop.
Switching gears a little bit, I saw someone tweeted yesterday that of the 70 economists polled by Reuters, not a single one had predicted that the second-quarter GDP growth would be revised down to 1.3 percent. You devoted a whole chapter to our total haplessness at predicting the economy. Is there any hope?
People think, Oh, the GDP was revised from 1.7 to 1.3, and it's a huge error. And anyone who does that hasn't really looked at the history of these revisions, because you've had quarters in the past where you've had a 4 percent GDP quarterly number that was revised to negative growth, and also the reverse. So, having a shift of four-tenths of a point is quite normal.
And yet not a single one of those economists predicted it.
This is why when we do presidential predictions we have everything in terms of a probability. We say, Yeah, Romney could win, and we're going to give you what you think you would need to warrant making a bet on that against the spread. With economic predictions, people say, Oh, exactly 110,000 jobs will be created next month, and those forecasts are off by an average of about 70,000 jobs.
You also talk about how hard it is to beat the stock market in the book. I'm wondering if you invest in stocks at all.
I do but in a very passive way for the most part. I'm not trying to do anything too tricky ... Actually, one of the better indicators historically of how well the stock market will do is just a Gallup poll, when you ask Americans if you think it's a good time to invest in stocks, except it goes the opposite direction of what you would expect. When the markets going up, it in fact makes it more prone toward decline. So it might be kind of the thing, for example, to invest in Facebook now because the stock's gotten beaten up a lot, and the sentiment seems very negative. I talk to my friends on Wall Street and they say, "Oh, Facebook is a terrible stock." And that makes me want to buy it!
One of the things that surprised me was how much more inaccurate local weather forecasts are than, say, the Weather Channel or the National Weather Service. How do they get away with it? Shouldn’t free market competition create an incentive for accurate weather forecasts?
Well the way we perceive accuracy and what accuracy is statistically are really two different things. The risks are asymmetrical: If a weather forecaster says there's a 20 percent chance of rain and it does rain, you get really angry at them; it kind of ruins your weekend plans. The other way around is kind of a serendipitous bonus: I thought it was going to rain and it didn't. So the commercial weather forecasters deliberately bias their forecasts toward having more rain than there really is for that reason.
The Weather Channel does it just a little bit; the local forecasts aren't even close.
The Weather Channel is very careful about how they calibrate their bias. They go, We're going to do a little bit, in certain circumstances, but not too much. But in terms of the local news, you get the same bias that you get in any kind of news coverage, where every story is sensationalized. And so every hurricane is the storm of the century and everything is Snowpocalypse or derecho or whatever else. They want to create more news. And it sells more when you have a bad weather forecast than a good one. So the book really believes in free market competition, for the most part, but if you're competing for viewers who don't know any better, then it's going to facilitate a race to the bottom as far as actual accuracy goes.
At the same time, weather forecasting is one of the success stories, as you point out. It's made a lot of great progress in the last few decades. And I'm wondering what predictive field you think can be transformed by technology in the coming 50 years or so.
I've tried to specialize in fields where people aren't using metrics as much, so in baseball, kind of before the Moneyball era, or certainly in politics, where I think the average news story is not all that data driven. So I do think about thing like maybe local government, where the Bloomberg administration is getting quite data-driven, in contrast to some big cities with stodgy old bureaucratic ways of doing things. The OKCupid guys, the online dating folks, people trying to rate how we like music and food and who we want to sleep with, all that kind of stuff. Even though you're never going to have perfect answers, because it's been so subjective before, that's where you might make gains.
What do you make of Dick Morris? You mention him in your book, and he's just the absolute least accurate pundit ever, and yet he still gets a platform.
I've been thinking about starting a hedge fund, just to invest in whatever Morris says won't happen. You have to actually be quite skilled to be that bad.
It's amazing. He said the other day that Romney would win by 4 or 5 points.
I know, that would be my secret plan if I were the Romney campaign, to get Morris to make a prediction of an Obama landslide. I'd definitely lay some money on Romney at that point. I think it's just, when you make that many poor predictions — and we're not just teasing him for one, he really does have a terrible track record — are you deluding yourself? Or is it all a show to rally your base? With him I don't really know, but, I see prediction as a way to judge whether you're, frankly, in touch with reality or not. We all have our views, and we look at this information in a certain way, and you test that by means of an event that you don't know the outcome to yet and seeing if your subjective point of view matches what unfolds. So I think a slightly impolite way to put it is that at best Dick Morris is detached from reality, and at worst I guess, he knows it and is just lying to people.
You used to make some good money off of online poker, but you write in the book that at the end of 2006, you lost $75,000, "most of it in one horrible evening." What the hell happened? I'm picturing this night as an absolute disaster.
Well, I was playing games where you're betting $200 per betting round, so you start to make [laughs] some bad decisions and kind of go on tilt a little bit, and those losses add up. But poker, I played limit hold 'em, which is actually a really swingy game, you have a really small edge. So if you're off your game even a little bit — and frankly, that night, I was probably off by more than a little bit — then you're probably a favorite to lose money. And if you're just tilting then you're a favorite to lose a lot of money. But poker does teach you some life lessons. It seemed for a while when the games were easy, you're just kind of there and clicking buttons and printing money, right? But usually the free lunch scenario comes to an end, in a terribly abrupt way for me, although I was smart enough not to lose all my winnings. I kind of took the hint after I lost about a third of what I won and said, "Okay, I'm going to cash out and start writing about politics and burritos now instead."
Will your friends even play you in poker? Or are they too daunted by your predictive powers?
I think they'll play me in poker. Sometimes maybe in NCAA tournament pools and stuff, I get a little bit of pushback, right? Or fantasy football, I can intimidate people a little bit. [Laughs.]
Are you good at fantasy sports?
Uh, yeah, I don't do fantasy baseball because I find that the expectations would be — people would expect me to be perfect. And fantasy baseball is super research-intensive. But fantasy football I'm pretty good. You kind of figure out which players are overvalued or undervalued, and you take advantage of peoples' biases in that way.
I saw one review of your book, and I'm sure there are others, that compared you to Malcolm Gladwell. Do you think that's a good comparison?
I think anyone who has written a book should be flattered to be compared to Malcolm Gladwell, but I think the tone of the book is different where, this is a book that encourages people to dig into things a little bit. And it's more a book where we show the readers stuff rather than telling them "Here's the conclusion." And so it's a little bit more of a hands-on feeling than the typical Gladwell book, it doesn't have as many cute frames. It has a multiplicity of ideas, which is how I see the world. The world actually is a little bit complicated. One of the undercurrents of the book is that the devil actually is in the details, and if you think it all boils down to a couple of slogans or a couple of lessons then you probably aren't going to get that far.
Based on your personal hunch, what are the chances that Hillary Clinton runs for president in 2016?
I think the chances are pretty good, and that is, obviously, a hunch. I don't think there's any way to model that. But we know that she ran for president in 2008, so no matter what, she can't be morally opposed to the whole idea of running for president. It seems like there hasn't been that many people who we know have shown that ambition before, and I think she'd be very likely to win the Democratic nomination. So, it's like, if you run for president, you probably have a 90 percent chance of winning the Democratic nomination, and conditional upon that, let's say it's 50/50 of winning the general election. But if you have a 50 percent chance of becoming the first female president, why wouldn't you do that? She seems to be in good health, and I know it can be grueling but, I don't know, my heart says we have a decent chance of her being president someday.