One of the most interesting side stories of the US election (at least for a survey geek) is the inconsistency in the polls. In part that's due to there being so many of them and with that proliferation has come what we might call some thinning out of polling expertise. But when a venerable polling firm like Gallup raises eyebrows with some numbers that seem outside the mainstream of the other polls you really wonder what's going on. For example, before Gallup suspended their daily tracking poll on Monday because of the east coast weather disaster they had Romney leading Obama 51% to 46% among likely voters.
Enter two favs of the NewMR: predictive analytics and predictive markets.
The hands-down champ in the predictive analytics crowd is Nate Silver and his Fivethirtyeight.com blog. Nate's getting a ton of criticism right now because the right wing doesn't like what his model says is going to happen. Or maybe it's just because he works at the New York Times which makes US conservatives see red (well, blue actually). The interesting thing is that Nate's model is saying pretty much what other models are saying and what they're saying is not far off from what the predictive markets are saying. The Washington Post's Ezra Klein sums it up nicely:
As of this writing, Silver thinks Obama has a 75 percent chance of winning the election. That might seem a bit high, but note that the BetFair markets give him a 67.8 percent chance, the InTrade markets give him a 61.7 percent chance and the Iowa Electronic Markets give him a 61.8 percent chance. And we know from past research that political betting markets are biased toward believing elections are more volatile in their final weeks than they actually are. So Silver's estimate doesn't sound so off. . . Silver's model is currently estimating that Obama will win 295 electoral votes. That's eight fewer than predicted by Sam Wang's state polling meta-analysis and 37 fewer than Drew Linzer's Votamatic.
Now granted, the principal inputs to all of these approaches, including the betting being done in predictive markets, is polling data. Lots and lots of polling data with adjustments for past accuracy, known biases toward one party or the other and, in the case of the markets, individual guesses about what will happen between now and next Tuesday.
This has been a crazy election season and my nerves are pretty frayed right now, but I am cheered by a certain delicious irony. Another unnamed MR blogger who preaches the predictive analytics gospel here, there and everywhere desperately wants those good old fashioned telephone surveys to be right and for the predictive analytics to come up empty. In my case, I'm willing to take a methodological hit for the greater good of my country.