Nate Silver on political polling in the cellphone and social media age
A supporter takes a picture with his cellphone of Republican presidential candidate Mitt Romney as he prepares to speak at a rally in Zanesville, Ohio, on March 5, 2012.
Boy, could we make a lot of money if we knew now who was going win the election in two weeks. All this technology, all of this fearsome data crunching power, you'd think our predictions would be getting better. Turns out our computers too often help us make bad political predictions more quickly.
Nate Silver runs the New York Times political blog Five Thirty Eight. He's the guy who picked correctly 49 out of 50 states in the Presidential contest last time round. He's also got a new book, called "The Signal and the Noise: Why So Many Predictions Fail-But Some Don't." Part of the message in Silver's new book, which has been hailed by other number wonks as an argument for the wiz to apply his skills beyond politics, is that all the new technology we have at our fingertips doesn't guarentee better predictions.
"It has helped in a lot of fields," says Silver. "But at the same time if we don't know what to do with technology, we'll just get ourselves into more trouble. If you look, for example, at the dawn of the computer age. People were assuming computers could do all these tasks they weren't really well designed for, that we still needed human judgement and intuition to apply to instead."
The issue of assuming computers will solve all our problems is true specifically in polling as well.
"So related to technology is the other half of information technology--information," says Silver. "And now we have twenty polls that are released on a given day sometimes. But what happens is that the more polls you have the more options you have, to cherrypick the information that you look at."
You may not have to get someone's political preference over the phone any more (engineers at Facebook and Twitter's "Twindex" have endeavored to deliver useful numbers), but it still might be the most effective way. Of course, even Silver says that these days, you'll need to include cell phones, because lots of homes don't have a landline any more. But things are moving online more and more, and that includes how we take the temperature of the nation's voters.
"There's not a track record yet," says Silver. "We know what these numbers are but we don't know how to put them in any kind of context. Google has actually now begun to do surveys. It's almost the equivalent of the phone method of random digit dial; you randomly interrupt people as they're browsing the Internet and say 'you can proceed if you tell us whether you would vote for Romney or Obama. Online polls, which are becoming more common -- I'm not saying there aren't problems with them -- but they're becoming more and more integral as we live our lives more online."
We asked Silver to answer the question so many of us are asking: Who is going to win in two weeks? Well, you can always head over to his blog and check the latest prediction. But he told us last week that in most "tipping point" contests, that is to say votes in swing states like Ohio, Wisconsin, etc., his analysis predicted a win for the current president, and that means another term for Barack Obama. Of course, Silver also noted that there is still one big political event -- tonight's debate -- and that it could potentially give either candidate a last-minute advantage that would bring them a win.