Tom Edsall has a nice post rounding up the electoral predictions of an array of political scientists. So far as their interpretations of the data goes, it seems pretty clear that Obama will either win or lose this election, and if he wins it, it will be either by a large or a small margin. Useful! Meanwhile, the best known forecasting model if Ray Fair's, which currently predicts a 3.8% win for Obama (slightly smaller than the difference in the polls, actually). On the other hand, Larry Bartels wrote a pretty convincing paper demolishing the statistical underpinnings of that model, which can be found here (pdf). The Bread and Peace model, which has a slightly better reputation among political scientists, (and takes into account such factors as military casualties in Iraq,) is predicting a 3.6% win for Obama. It argues, interestingly, that Iraq won't be a big deal. "Cumulative fatalities in the vicinity of 4,300 – the projected magnitude around Election Day – depress the Republican vote by approximately three-quarters of a percentage point compared to the counterfactual benchmark of zero fatalities." It's hard, of course, to know how seriously to take any of this. But it's interesting that in all the commentary about why Obama isn't ahead by more than 2%-5%, the fact that these models predict something around 3% -- and thus he's ahead by exactly what the political scientists would predict -- has barely been mentioned. For the media, of course, it makes sense: If elections were all about campaigns, Obama would be ahead by 20%. He's more exciting, more skilled, and the campaign is all about him. Political scientists err on the other side: They argue that campaigns have next to no impact at all. And right now, their argument fits the realities of the election much more closely.
