Andrew Gelman, whose new book Red State, Blue State, Rich State, Poor State is a must-read for political science junkies (at least check out the working paper (pdf) that turned into the book) has a blog post up noting that historical economic trends would presage Obama victory of something in the neighborhood of six points or so, since the economy is doing badly but not, by historical standards, horrendously. This is a bit smaller than some other models, which predict an Obama win in the low double digits.
The real question here, I think, is to what degree you factor President Bush's exceptionally low approval ratings into the mix. Gelman doesn't, so his predicted margin is on the low side. Other models do. It's tricky, because while incumbent approval ratings are generally one of the main factors you'd want to incorporate into a model, never since the rise of polling have we had a president termed out of office with approval ratings this low, so in a sense we're in uncharted territory. What's more, you can make an argument that from this standpoint, having a 25 percent approval rating might not be that much different from having a 40 percent approval rating, since much of that drop-off comes among weak Republican identifiers who have become fed up with Bush, but are still going to be quite hesitant to support a Democratic presidential candidate against someone other than Bush.
The public debate (and I confess I've been guilty of this at times too) often casually assumes that discontent with Bush will ultimately have as much of an impact on the presidential race as it has on the generic party ballot. But there isn't necessarily any historical basis for believing that to be the case--the data set just isn't large enough, and the last time a president was termed out, in 2000, his approval ratings didn't translate into the level of support for his party's candidate that the models predicted. A McCain victory in November would certainly be a significant upset either way, bigger than any in recent history, but it might not be the monumentally huge upset that the current narrative suggests.
--Josh Patashnik