[lbo-talk] race costing Obama about 6 points

Michael Pollak mpollak at panix.com
Mon Sep 22 06:07:06 PDT 2008


On Sun Sep 21, andie nachgeborenen wrote:


> Fuck. So the Bradley effect isn't going away.

I don't think this Stanford analysis supports that at all. IIUC, what this analysis says is that Obama is currently leading by 2-5%, and if he were white, he'd be leading by 7-11%. Which gives a pretty direct answer to a question everyone's been asking for 6 months: how come he's not leading by more?

But AFAICT, this has nothing to do a prediction of a Bradley effect. During the primaries as we all know the more common event was the opposite, the anti-Bradley effect, where Obama got a bigger vote than polls predicted. And for the naitonal eleciton there's no reason to expect one rather than the other. If anything, their surprising finding that a sizeable chunk of white people who harbor prejudice against blacks will end up voting for Obama that question more indeterminate than ever.

The only way this analysis would support a prediction of a Bradley effect was if they found that undecided white voters were disproportionately more prejudiced than white voters who were already decided and counted. But they didn't and they couldn't since they were using a data set that can't be aligned with current undecideds.

Michael



More information about the lbo-talk mailing list