What Not To Notice

John Sides posted a great graph about how useless head-to-head polling is right now for predicting presidential general election results; I have a post up over at Greg's place in which I add a couple of other things not to pay attention to right now if what you want to know is whether Barack Obama will be re-elected or not. Mostly, I add that state head-to-head polls are useless now, and that they'll continue to be pretty much useless until after the party conventions next year. Why? Mostly, because state results are driven by national results. So if Obama wins by six points nationally, you can bet that he'll win, say, marginally Democratic states such as Iowa and New Mexico, regardless of what state polling indicates earlier on. Eventually, the electoral college will reflect the national results, and specific state quirks only really matter if it's a very close race. The other part of it is that while lots of organizations poll regularly at the national level, state polling tends to be a hodgepodge, and you a bad state poll is apt to be far more misleading than a bad national poll. Of course, this is the kind of stuff that Nate Silver is extremely good at, and when it starts mattering in fall 2012 I'll certainly be recommending his numbers.

As long as I'm on that subject, though, I'll add one more thing to ignore: state polling on the nomination process outside of the early states -- and even then, I'd take New Hampshire and South Carolina polling with a heavier than usual grain of salt. This is just a consequence of the stuff I was talking about earlier today...right now, rank-and-file presidential nomination preference isn't mediated by the people who are going to influence it down the line. In other words, no one is campaigning yet in California, and they won't for a long time. As 2011 goes on, Iowa and New Hampshire polling start becoming more and more meaningful as the campaign really starts happening in those places, but in states without active campaigning the results will lag behind.