See here in the New Yorker. These models don’t tell us that Romney will lose to Obama but they certainly suggest that Romney will be the Republican nominee. Of course, you already knew that anyway, but the models help support our intuitions.
Different models will help us conclude that Romney is likely to lose in the generals if the economy continues to move in a positive direction (see here and here). If the economy tanks – given the importance of things like “rate of job growth” and “net change in the unemployment rate” – then all bets are off. See here for a discussion of some of these economic indicators in presidential election models, here for a criticism of prediction models, and here for a defense/discussion of prediction models in general. I don’t think these models tell us everything and don’t have the predictive power of models in the physical world (of course), but they do help us see a bit better in the dark.