Election 2014  
comments_image Comments

Nate Silver: The Polls Aren’t Wrong

Bias? The stats guru who nailed the 2008 election tells Salon it's very hard for this many surveys to all be off.

Continued from previous page


So if the polls right now tend to show Obama up, say by five or six, that could still be within the margin of error. Could these polls we’re looking at be wrong?

Sure. One poll that shows a five-point race doesn’t really tell you much. But since the conventions we’ve had, I think, 250 polls released — state polls, national polls — and that really reduces the margin of error to very little.

Look, you have about 10 recent national polls now, which is a fair number, certainly. But you also have 100 state polls that have been released over the last week from a more diverse set of pollsters. And again, we’re not talking about huge shifts, right? Where the state polls say Romney’s down by five instead of being down by four — it’s not that precise where that’s really all that much of a difference.

But it’s not like there is a fundamental disagreement there, exactly. Today, for example, the Gallup poll now has Obama up by six and a Bloomberg poll has him up six. So you have two more national polls to say, “All of the national polls seem to be in line now.”

When we are surrounded by predictions and drowning in polls, how should people test the validity of what they hear? What’s actual analysis, and what’s politics and spin masquerading as prediction?

On average, people should be more skeptical when they see numbers. They should be more willing to play around with the data themselves. And, yeah, that might lead to the occasional Unskewedpolls.com, but look, the application there is really wrong, but the instinct that you should be willing to take data skeptically is worthwhile.

But that’s the good thing about predictions. Someone like Dick Morris makes predictions that are specific enough that you can test them and see how they do, right? But what usually happens is people make a prediction, and then it’s right or it’s wrong — but either way it’s forgotten and you’re on to the next one.

There is no substitute for having an actual track record of having made predictions. Basically, some predictions are no better than random, right? And some are worse than random.

So why are we so fascinated with predictions? All day long, the 24-hour news networks are making projections. Sports radio and ESPN analyze the likelihood of teams making the baseball playoffs down to a tenth of a percent. We can’t get enough of trying to guess how things are going to play out.

I have two answers to this, and maybe they’re complimentary in part, contradictory in part. I think, on the one hand, you can go to a site like  FiveThirtyEight and it has that one number, right? So Obama today has a 79-point-whatever-it-is [chance of winning] — 79.6 or 79.7 — and that might save you a lot of effort; it’s like, here’s the gist, right?

I think people feel like there are all these things in our lives that we don’t really have control over. Certainly, one person’s vote doesn’t affect the political environment very much, and I think we feel like if we can predict something, we can control it. And so, getting into that data allows us to feel less alienated from the random course the world might be on, or the inability to influence it.

I also think there is a little bit more of a do-it-yourself spirit now, which helps. People don’t necessarily trust the news media to mediate information for them anymore — and maybe they shouldn’t, right? So if you show the raw numbers and are actually willing — at least in my case — to set a betting line, basically … And it’s based on a formula you’re not just making up a number everyday.

See more stories tagged with: