I had a really interesting conversation with Sam Wang last week, on his perfect performance predicting the election results. One of the things that came up was the media's inability to understand that even though the race for the presidency was close--pretty much 51-49 nationwide, and tight in the "battleground" states, it wasn't a "toss-up" or "too close to call." Sam, and 538 and other poll aggregators had really solid evidence that Obama had it won. As Sam said at one point, while there were fluctuations in Obama's lead, he was never behind. Ever.
The political experts, those guys who live in Jay Rosen's Church of the Savvy, who were all saying "too close to call" mistook the 51-49 number to mean that it was a coin toss--either candidate could win, because the margin was small. But the probability of an Obama win, as Sam was saying in the last few weeks, was no coin toss; it was in the high 90s. Sure, it was close. But more than a year of accumulated polling data made it clear that the race wasn't in doubt--that the margin was not going to change by enough to lead to an Obama loss.
What this also proved, by the way, is that being Chuck Todd--having special access to the real players who have the inside dope on the campaigns--is actually counter-productive. Saying "ZOMG who knows what will happen? Romney's key (unnamed) people say they're gonna win this. It's really tight in the battleground states. No telling what will happen." was incredibly stupid as well as completely wrong. Now I suppose it's not their fault that they know nothing about statistics and probability, but you'd think that might be a skill to seek out in choosing your analysts. Rather than who has the biggest fish in in the rolodex.
Oh, and just by the way. Sam's polling data was all publicly available. And he published all his source code. Anyone can do what he did. Even Chuck Todd.
