It is important to understand that a quality poll will normally do a very good job predicting the outcome of an election. The only thing better than one quality poll is an aggregate of multiple quality polls. That way you can effectively weed out the occasional fluke sample or mistake.
It is only because polls already do an incredibly good job of predicting the outcome of elections that Nate Silver was able to do a good job predicting election outcomes by using these polls. While it is true Silver got most of the major votes right in the 2012 election, he is by no means alone. Many other polling aggregators, like HuffPost Pollster, Realclearpolitics and PollTracker, all performed extremely well. Many of the individual polling operations, like Marist and PPP, also were spot on. Silver was successful, but in no way was his success truly unique.
Silver did create a more complicated model that has some added variables and elements, but is difficult to tell if these extras actually improved on rather basic polling averages, primarily because polling averages are very accurate. It is hard to make a noticeable improvement on something that already works very well. For example, in the UK election Silver’s more complicated model actually performed much worse than other aggregators using simpler models. The basic “secret” to Silver’s secret sauce is that polling works.
From my perspective Silver did only two things that really stood out. First, he converted a polling lead to a likelihood of winning percentage, although you can argue he compromised the integrity of his number by adding in a personal fudge factor. While a small change, it was a very smart communications move. Regular people who don’t know much about polling may not realize even a five point lead is really a big deal. Putting a percentage on it helped people understand the results.
The other thing Silver did was break into the popular narrative by being a very authoritative writer. Being a good writer is no small feat. While many people involved in politics know polling averages do a great job predicting election outcomes, Silver packages this basic knowledge in a way that spreads. Decent writing about what should be basic facts is a clear improvement over what still currently dominates much of election punditry, which is decent writing about mostly useless nonsense to make the election seem closer and more exciting.
Photo by dnkbdotcom under Creative Commons license