I found an interesting article on the PBS website, that was published yesterday. It talks about the "prediction markets" for the election:
If you want a quick and meaningful rubric for understanding the likelihood of an election, one way to think about it is take every state and the District of Columbia -- the 51 Electoral College districts -- and rank them from the most likely for Romney to the most likely for Obama. Rank them all, number them 1 to 51. You have a bunch of states that are bunched up near 100 [percent likelihood] for Obama and a bunch of states bunched up near 100 [percent likelihood] for Romney. Then you have this whole middle ground from leaners [to the Democrat ticket], to tossups, to leaners [toward the Republican ticket]. And what happens is that, barring a major event, any given state is just going to start drifting, over time, towards the candidate they're most likely [to vote] for. So every day that goes by on which that state does not flip [to the other party], it's going to slowly drift in the direction of the candidate that it favors in terms of probabilities. And that's what we're seeing happen.
The article can be found here:
link