Election Predictions

I wouldn’t take this prediction to the bank if I were a betting man. But, like co-blogger David Bernstein, I give Obama a slight edge, perhaps a 60-65 percent chance of victory. In the contest between national polls favoring Romney and battleground state polls favoring Obama, I give slightly greater credence to the latter. My main reason for doing so is that their results have been more consistently favorable to Obama than the national polls have been for Romney. In addition, there is a nontrivial chance that Obama could win the electoral college while narrowly losing the popular vote. I also give some weight to the majority view among mainstream pollsters, which seems to be that Obama is more likely to win than not. On technical questions like this, I try to give some deference to expert opinion, unless there is strong evidence of bias or ulterior motives. And I am skeptical of claims by some conservatives that the professional pollsters are in the tank for Obama.

On the other hand, it’s certainly possible that the pollsters’ likely voter models are just slightly skewed in Obama’s favor. In a very close election like this one, even a 1-2 point skew could lead to an incorrect prediction as to the outcome. Dan McLaughlin of Red State makes an interesting case for that view in a series of posts (see here and here). Notice that McLaughlin is not claiming that Nate Silver and other analysts who predict an Obama victory are a bunch of idiots whose models are radically deficient, or a bunch of shills for the Democrats. Rather, he seems to be saying that Silver has a pretty good model that is slightly off – enough to make a wrong prediction in a close election. Silver perhaps has a similar take on McLaughlin’s prediction.

Turning to the other races, I predict that the Republicans will retain control of the House with a majority fairly close to the one they have now. My guess is that they will probably fall just short of taking control of the Senate, with perhaps a total of 48-49 seats. But there are enough close races that I could easily be wrong on this one. The more you believe that state polls are understating Republican turnout, the more you should expect that the GOP might take more close Senate races than most pollsters expect.

I also expect that Virginia Question 1 will pass by a comfortable margin, as has every other post-Kelo property rights reform referendum, except for a few that were tied to unpopular measures such as abolishing rent control.

With the exception of the predictions on the House and Question 1, I am far from confident about any of these calls. The election is so close that both the popular vote margin and that in key battleground states is within the margin of error of most polls. The same goes for a number of the close Senate races. That magnifies the potential impact of even small methodological errors by pollsters.

UPDATE: Some commenters correctly point out that the RCP average of national polls has Obama and Romney effectively tied (an 0.1 point differential). That strengthens the case for Obama. The most recent two polls (one by Rassmussen and one by the Washington Post and ABC) include one that is tied (Rassmussen) and one that gives Romney a 1 point lead (WP-ABC). But for reasons Ted Frank outlines, a tie does give an implicit edge to a challenger because undecideds are more likely to oppose the incumbent, despite recent efforts to overturn that conventional wisdom. Given the most recent national polls, it may be that 60-65 for Obama should really be a little higher, such as 65-70. But I still see the national polls as very slightly favoring Romney, though certainly less so than a week or two ago.