This was the nerdiest election ever. The amount of data collected and analyzed in a public way on blogs, news sites, and television was unprecedented. And it wasn’t just the quantity of the data, but the amount of insider talk on mathematical models about how to crunch the numbers that was new this cycle.
There’s always been discussion about how to interpret polls, how to treat exit polls with caution, and so on. But this year was a whole different ballgame. Compared to 2004 when individual polls held sway (such as Zogby famously predicting, wrongly, that Kerry would win), this cycle has been dominated by meta-pollsters — analysts who aggregate together dozens of polls in various ways. Judging by the success of this method, we can expect this to be the future.
While some sites such as Pollster.com have been doing this for a while, it was newcomer Nate Smith of fivethirtyeight.com who became the poster boy and constant presence on talk shows. Smith used to work for Baseball Prospectus, a firm of math nerds that makes projections about baseball teams and players, and he famously predicted that the Tampa Bay Rays would win 90 games this season, which seemed ridiculous considering they’d lost 96 games in 2007. In fact the The Tampa Bays exceeded Silver’s projection by winning 97 games and narrowly lost the World Series.
Whereas Pollster.com tracks trend lines of polls to provided a blended assessment, fivethirtyeight.com uses a variety of methods to weight various polls, and mixes in data about demographic trends per region. Silver discusses the modeling that he does on the site (without giving away all the secrets), and the site itself is an orgy of dense charts and statistics, leavened by good writing and photography.
How did he do? Well, as of this writing three of the states are still toss-ups, but his model nailed every single other state. The chart below shows his Monday projection (including three states that would be tossups, 2 of which are still being counted, 1 day after the election), followed by a chart of the actual outcome from The New York Times. Pretty impressive.