Cribbed this from BlueMassGroup and user CentralMassChic. I haven't read it in detail, but Alan Abramowitz of Emory University, a Poli Sci professor who claims a very good prediction record using his model, is predicting an Obama victory with 55 - 45 in the popular vote, +/- 1-2%
The paper is here for the statistically inclined:
http://www.forecastingprinciples.com...
Looks, at first glance, like a straightforward regression mode. The good professor claims that the best predictor is the popularity of the current incumbent. That'll pretty well take the lip stick off the pig. With ol' Georgie somewhere in the 20's/ 30's for approval ratings, John's done.
If this model is correct, of course. Actual mileage may vary.