Identifying likely voters is a problem that the polling industry is already highly engaged with, they tweak their model all the time. However, they aren’t as engaged with social desirability effects in political polls, in part b/c they’re contextual and not as consistent. That’s why people are more concerned with the second.

You have to fix both problems. But even if we didn’t see a difference in the bottom line, that doesn’t mean that there wasn’t a group of voters lying to pollsters, which is an issue for the pollsters (many phone poll workers are African American, hence the social desirability effects. However, they don’t report the data broken down by race of worker, even though we know that respondents behave differently on some key issues).

The data indicates that this effect did not exist in the dem primaries

Dem primary voters are around 20% of the party, and an unrepresentative slice at that. Generalizing from a bunch of democratic party activists to the voting population at large is a bit odd. You wouldn’t do so with the content of their opinions. I’m just saying that I’m waiting to see what happens as we study this issue further, but in my mind the jury is still out.

The whole “we will kill obsama/hunt him down” was the one moment where Obama made me cringe.

Me too. The more I think about it, the more possible similarities I see with the Johnson Presidency. Making OBL so large in his own and the public imagination, by saying this kind of thing – not to mention OBL’s own non-state actor buddies – risks creating a Che Guevara effect.

And BTW, today, 10 October, is the day Che Guevara was captured and executed in 1967 – 41 years ago, when Obama was 5+.

Data is king. The data indicates that this effect did not exist in the dem primaries, and as I pointed out in my previous comment, it is somewhat questionable whether the original Bradley loss was indeed due to a Bradley effect in the first place.

The second case is one where you have two errors that happen to be cancelling out, but you’re really doing failing to predict the behavior of two different groups wrong, and simply getting lucky about the bottom line.

I agree. But why do you want to weigh one racial effect highly (counting white voters who lie), and ignore the other (counting black voters who are not represented)?

I want to disentangle a few different issues

1. Is there a Bradley effect on the individual level, i.e. are there people who claim they will vote for Obama for social desirability reasons, but who have no intention of doing so. There are reasons to believe that this might still exist, and little reason to believe that there is an opposite effect at the individual level, i.e. people who say they are voting for McCain but really intend to vote for Obama.

2. Is there a Bradley effect at the aggregate level – i.e. will our polls overpredict the chance that a black candidate might win? Your counter-argument is that we might be underpredicting his chance of victory because we don’t know who the likely voters will be this time round, i.e. because minority turn out might be higher and b/c many new voters have been registered. I concede both of those points, but that’s a failure to predict the likely voter, something that is conceptually distinct from the Bradley effect, even if it serves as a countervailing influence in this election. It’s a bit sloppy to lump them together. Nate should know better, seeing as his father is a prof at MSU who is an expert on surveys.

The fact that our aggregate predictions have been right during the primaries may be due to multiple causes. It could be that likely primary voters are less likely to lie to pollsters for reasons of social desirability. Or it could be that they are lying, but the effect is washed out by increased turnout from minority voters. The second case is one where you have two errors that happen to be cancelling out, but you’re really doing failing to predict the behavior of two different groups wrong, and simply getting lucky about the bottom line.

He had mentioned it in the Democratic debates. I’ll need to scour through the Dem. debates transcript to find it.

Well, he couldn’t have mentioned all of it then because the news that incursions had been authorized by the President came out only 2 months ago which, I think, was after the last primary debate. But if you could find what he said, I’d be quite interested.

No, but the the airstrikes had begun to sow the seeds.

And you know, looking back at the transcripts, he did only mention it in passing. I thought he had said it much stronger than that. Maybe I was conflating a few of the other “Bush has taken my policies that I proposed even before he accepted them” speech with this issue.

But he did bring up the topic. With respect to Pakistan, I never said I would bomb Pakistan. What I said was that if we have actionable intelligence against bin Laden or other key al Qaeda officials, and we — and Pakistan is unwilling or unable to strike against them, we should. And just several days ago, in fact, this administration did exactly that and took out the third-ranking al Qaeda official.

That is the position that we should have taken in the first place. And President Musharraf is now indicating that he would generally be more cooperative in some of these efforts, we don’t know how the new legislature in Pakistan will respond, but the fact is it was the right strategy. http://www.nytimes.com/2008/02/26/us/politics/26text-debate.html?pagewanted=all

Like I said, perhaps the reason he doesn’t bring it up at all now, is because it has now become a ground assault. And it may not be the best thing to bring up. I don’t know.

But the original idea was that Obama would not adhere to Pakistani sovereignty–which still holds whether it is from the ground or from the air.

Well, he couldn’t have mentioned all of it then because the news that incursions had been authorized by the President came out only 2 months ago which, I think, was after the last primary debate. But if you could find what he said, I’d be quite interested.