If the race is close on Election Day — less than a point — there really could be a national vote/electoral vote split. Since the overall tendency is for the Electoral College and popular vote to match up, you wouldn’t give more than 50-50 odds of it occurring. I’d call it about 1 in 3 chance, which just happens to be the number of times we ended up with a split decision in the three races decided by less than a point (although there is an argument that Kennedy lost the popular vote — an argument that does not include conspiracy theories about Chicago vote counting –, but that is a story for another day).
What if the popular vote margin is larger than a point? I think the odds are slim that there would be a disconnect. This year doesn’t really seem akin to other years where we’ve seen large mismatches between the popular vote and where the electoral vote number 270 is located. It’s not a blowout like 1964 or 1972, nor is there a substantial third party presence, as there was in 1968 and 1980.
Still, given the result in 2008, I don’t think that we can completely write off the possibility that Romney could lose the Electoral College even if he were to win by more than a point. Now, 2008 was a borderline-landslide year; this year will almost certainly be decided by a closer margin than that election. It’s also difficult to say whether the discrepancy we noted in 2008 was a result of noise in the data or whether the Obama campaign’s efforts actively created such a split.
The tendency over the course of this cycle has been for the popular vote in the states to trend toward the national vote. Given this tendency, and the overall history of the Electoral College, the smart money suggests that these state polls will revert to a mean somewhere around the popular vote in relatively short order. If there aren’t any more large shifts in the next two weeks and we see the same split persist, we will have to revisit this. But for now, I wouldn’t give an electoral vote/popular vote split of more than a point any better than a 1 percent chance of occurring.