Some people have speculated that the difference between the New Hampshire polls, which showed Barack Obama way ahead, and the actual results was due to the "Bradley effect," named after former Los Angeles Mayor Tom Bradley, an African American who was well ahead in the pre-election gubernatorial polls, but on election night in 1982 lost to Republican George Deukmejian*. I don't think so, especially if you are saying that these votes went to Hillary Clinton rather than Barack Obama. The same people who might not vote for Obama because he is black would be likely not for a woman either.
My guess is that if there was a Bradley effect at all, it showed up in people voting for John Edwards rather than Obama. If you look at Edwards' polling, he does best among voters who describe themselves as "somewhat conservative," garnering 23 percent of their vote, while getting only 15 percent among voters who describe themselves as "somewhat liberal." That 23 percent, which amounts to 23 percent of 5 percent of the Democratic electorate, or about 1.2 percent, would be the "Bradley effect." Add the five percent that Bill Richardson (who was seen as white) got among these voters and you get at best 1.5 percent. These would be voters who considered themselves Democrats and disliked George W. Bush, but were unwilling to vote for a black or a woman.
I have speculated that in a general election, Obama would have to overcome resistance from white working class voters who might otherwise vote for a Democrat, but these would not be voters who participate in the New Hampshire Democratic primary or the Iowa Caucus. They would most likely show up in some Midwestern primaries, but might not vote in primaries at all.
*Corrected at 11:24 a.m.
--John B. Judis