The Gallup Mystery and Other Matters

 So here it is, a little after 1 on Saturday afternoon, which means that the Gallup tracking poll is out, and it again shows Romney with a five-point lead. It also shows, I see from tweeterville, that Obama's approval rating has gone from 53 percent Wednesday to 47 percent today.

Either Gallup is going to look brilliant election night or this will be a much-remembered heaping pile of egg on the organization's face. I don't claim to know at this point, although I do note, as a liberal would, that its numbers are pretty different from a lot of people's.

This approval number is fishy. Why would Obama have lost six job-approval points in three days? The only thing of any apparent moment that happened in those three days is the announcement of the GDP number, which was so-so, kind of a on-the-one-hand, on-the-other sort of tihng. So that doesn't make any actual sense.

CNN is out with a new one showing Obama up four in Ohio. We all believe the polls we want to believe. And though some of you are inevitably going to say that I would say this, you can believe me or not believe me, but I've always said: At this point, state polls are the only ones I really pay attention to. States decide it. Duh.

I also think, as I've written once, that there is reason to believe that national beauty-contest polls have a lot of noise this time around, more than usual. That's because the margins are going to be so wide in so many states. Romney is going to win in about 15 states with margins of greater than 20 points, in all likelihood. They're small states, but they add up. He's going to win some of these by 30 points. Then he's going to win two large states, Texas and Georgia, by double digits.

Similarly, Obama now leads in California, New York, and Illinois by 20 or more, sez HuffPo pollster. I haven't gone back to compare this to 2008, 2004, and 2000, but I'd bet that in general, the spreads are wider this year. In sum, there are loads of blowouts state by state, and I'm no expert but I know something about this stuff and I think it has to skew the national results.

By the same token, the "national Obama white vote" category, sometimes discussed these days, is skewed by the South. The benchmark for Obama here is 40 percent. He's below that in most polls right now. But remember, his white vote in the South is maybe 25 percent tops. So it's higher elsewhere. And it doesn't matter on Election Day how low it goes in Arkansas and Alabama and all those other places, because he's not going to win those states and he's not going to need to.