Friday, February 25, 2011

Dogs, not quacking

It seems that polls are only talked about to the extent that they agree with the prejudices or agendas of some large group of people. (This is also true of popular psychology.) While it is inevitable and perhaps healthy that most people interested in politics in one sense should also be interested in the other, and that most "analysts" should be propagandists, one sometimes finds that genuine puzzles are widely ignored. For instance, Nate Silver has two state-by-state maps of Obama's approval rating now vs. his vote share in 2008. Here is the absolute change:

And here is the swing-relative-to-national-average:

Blue is better for Obama, red is worse. The zero has no obvious significance: approval ratings track votes but there might well be an overall offset of a couple of points. (For instance, one knows people in New England who disapprove of Obama but would be unlikely to vote for a Republican.) To the extent that this offset is different for every state one should be skeptical of treating this as an apples-to-apples comparison; but some of the effects are large and presumably real.

I find this data puzzling.

1. Re the south, one's default hypothesis is that support for Obama is inelastic -- being determined almost exclusively by demographics -- so the first graph is the relevant picture. Most states where Obama hasn't lost ground are states where he didn't have ground to lose.

1'. Silver suggests that Obama might be gaining ground among racists as they become used to having a black president. (The diminishing-salience-of-race theory.) I don't buy this theory as the pattern is evidently about deeply Republican states (NE/AK/UT), rather than southern ones in particular.

2. I don't know what's going on in upper New England. At all. Nor does anyone else seem to, at least that I've read... is there a story here that I missed? "Libertarianism" does NOT work as the Dakotas, Montana etc. do not look like NH/VT/ME.

3. Silver suggests a "swing state effect" -- states where Obama campaigned heavily in 2008 were artificially pro-Obama because he outspent McCain. I don't see it; NH looks more like VT than like Ohio or Florida.

4. If there is a pattern here -- and I'm not sure there is -- it raises exactly the opposite worry from the usual one about "working-class whites" (see Galston for another formulaic implementation of this meme, and Clive Crook for a robotic blurb). Obama has lost more ground in Colorado than in Ohio. If I had to draw a general inference from the second map it would be that (barring W. Va.) the states in red have disproportionately many cultural progressives, many of them well-off. Three obvious narratives come to mind. The first is that these people were unusually starry-eyed and are now unreasonably disillusioned. (But what about New York and Massachusetts?) The second is that, having had more to begin with, they lost more in the crisis, and are therefore unusually sour. (But Vermont and New Hampshire have low unemployment.) The third is that, having more leisure, they have been paying more attention -- this is the notion of the "high-information" voter -- and that when others really begin to tune in they will come to similar conclusions.

Though who knows what things will look like on the other side of the government shutdown.

No comments: