I've always been fascinated by these predictive models, they are generally based on very complicated formulas that generally exclude everything except for economic variables. There is a good amount of dispute as to which formulas are the most accurate. I know Roy Fair has his own system and others have their own, I did see another analysis recently I think it was posted on the Atlantic Wire which also suggested a 53/47 two way split in favor of Romney. I think its based on the concept that at some point, the economic climate will create a bottom falling out effect in the incumbent share of support.
We'll see what happens. One can never tell with such models, especially when a foreign crisis, such as something happening in the ME may totally change things.