You’ll hear over and over again in the coming weeks that the national vote doesn’t matter in presidential elections, that all that really matters is what’s happening in the states, especially the swing states. Technically, that’s true.
But in practice, the only people who need to pay attention to state polls, even at this late date, would be those who make decisions about where to deploy campaign resources and those who plan to enter a pool that requires predictions of individual states. If you want to know who will win the election, however, you’re still better off focusing on a good average of national polls.
It’s pretty simple. If either candidate wins the national vote by three percentage points or more, it’s virtually certain he will also win the electoral college.
If either candidate wins the national vote by as much as a full percentage point, it’s highly likely — probably better than an 80 percent chance — that the vote-winner will also win the electoral college.
And if the national vote is very close, then it’s certainly possible that the electoral college will differ . . . but the direction in which it will differ is close to impossible to tell because the polls just aren’t definitive.
This all has to do with the degree to which individual states “swing” with the national polls. If we truly had perfect “uniform swing” — if Romney gaining four points in national polls was reflected by a gain of four points for Romney in Democratic Rhode Island, Republican Utah, and swing-state Ohio — then, we would know exactly what would happen to the electoral college for any particular national lead. If we were nowhere close to uniform swing – if all the states moved independent of the national polls – then the national polls would in fact be useless and we would just want to look at state polls. What we have, instead, is mostly uniform swing. But that’s almost impossible to measure; polls are pretty good at predicting final results, but they’re not quite good enough to discern the fine differences in swings at the state level. We typically see small discrepancies: One state may move 5% farther towards Romney than the national average, while another may move 5% less than the national average, just enough to make it not clear from election to election which swing state will give a candidate the electoral college win.(I think a split is more likely than Sean Trende does, but he has an excellent discussion of why the state polls can be difficult to interpret).
So while all those people telling you that it’s only the states that count aren’t exactly wrong, the truth is that you’re going to have a better idea of who will win from just picking two or three of the poll-averaging sites out there and watching their national poll-of-polls or forecasts. Those are going to be your best bets.