Can Obama win if he drops below McCain in the polls?

Add to this the fact that, in primary polling there were significant percentages in some states who had no problem at all saying that race was a key factor in their vote. I think the tendency for someone who is conscious of the fact that they don’t want to vote for Obama because he’s black, wouldn’t be particularly embarassed about saying to a pollster’s face (or rather, their presence on the phone).

I think McCain still needs to watch out for third party intrusions into his base.

I pretty much agree with you, but I think you need to go back and look at what happened to Ford. IIRC, he got slightly more votes in the election than the polls were showing. He was not ahead in the polls leading up to election day.

It’s because people don’t want to seem racist, so they’ll say to the pollster they’ll vote for the black candidate, even if they won’t, because they’re afraid that if they tell the pollster they’re not voting for the black candidate, the pollster will think they’re racist.

The Bradley effect isn’t about racists. It’s about people who are concerned that they’ll be seen as racist.

I think where you get that most is that pollsters will often ask which party you are registered with. If you are registered with party A and are asked if you will be voting for that party’s candidate, and the candidate is Black, you might say you plan to vote for him even if you aren’t planning to. Whether this effect is about racists or not, I don’t know. If I had to guess, I would say it was.

I understand that it’s about people concerned they’ll be seen as racists. It just seems more likely to me that they’d say they plan to vote for the white candidate and throw in a couple lines about his policies if they feel they need justification, rather than saying they’ll vote for the black candidate.
**
John Mace’s** scenario seems a little more plausible to me. Is that how the poll questions are set up? If so, maybe that needs a little tweaking.

I was polled by Qunnipiac during the primaries, and it was 25 minutes of answering detailed questions about a specific persons record and experience. I knew exactly who I was talking about, but he was never mentioned by name. So, I’m not sure if general election pollsters are different, but I think that line of questioning was to lessen the instance of the Bradley Effect. Of course, someone could out and out lie about their positions for 25 minutes I suppose, but I doubt that enough would go through that to make a statistical difference.