Sunday, November 11, 2012

Nate Silver and the Future of Polling


Nate Silver is becoming quite the celebrity. The stats wiz correctly predicted the outcome of all 50 states in the 2012 presidential election, and missed the mark on just two Senate races. In 2008, he correctly predicted the outcome of the presidential race in every state but Indiana, and correctly predicted every outcome in the Senate. The website “Is Nate Silver a Witch?” went up sometime on Wednesday this past week.  

 Silver first grew interested in politics around 2006, while following Congress’ attempt to ban online poker—he was pretty good at that too, and it was “then one of [his] main sources of income,” he explains in his recent book, The Signal and the Noise. It was around that year that he started publishing his political predictions. Silver has a clear, consistent methodology laid out on his blog, explaining the way he simulates various potential outcomes day after day with poll data and draws his predictions from the aggregate results of these simulations. While he’s received a fair amount of criticism, his accuracy, as we saw again Tuesday, is a force to be reckoned with.


 Silver has his own methods of combating different polls’ potential biases, but he’s also acknowledged problems with current polling methods in general. With more people using only cell phones and otherwise screening their calls, samples inevitably wind up skewed. In a recent interview with Reuters, Silver praised Ipsos, a partner of Reuters’ that conducted polls online this election season. It’s becoming increasingly clear to pollsters that they must find a way to reach people, and not just people with landlines who don’t use caller ID; they need a sample that can represent the electorate.


 With Twitter, Facebook, and other social media outlets, blogs like “Five Thirty Eight” spread like wildfire. Silver has proved himself and his methods at this point, so given the accessibility of the information, we should expect to see increased use of similar statistical methods in the future. It will be interesting as well to see how polling on the whole changes and adapts with our lifestyles; at least in the next few years, more online polling seems inevitable.

Rate this posting: