people might have wanted to appear to their neighbors (i.e., in an Iowa caucus) to be more open to an African-American candidate than they were in the privacy of the voting booth. A similar, though less strong effect might influence a respondent's answer to a pollster in New Hampshire: a possible reluctance to admit favoring or voting for Clinton over Obama.Hmm!.
I tend to think that the Clinton win is largely a result of people making up their mind in the last 24 hours. Pre-election polling always has a large segement of undecided voters. In New Hampshire, according to a CNN tracking poll, some 43% of the Democratic voters in New Hamsphire had not made up their minds three days before the vote. The exit poll question of when people made up their mind is somewhat instructive. Of the respondents, 38 percent of teh voters made up their mind in the last 72 hours before the election. The vote was generally split between Clinton and Obama on that segment of the vote. But Clinton had a strong lead among voters who decided before December 2007, a 17 point advantage.
So what happened with the pre-election polling. All sorts of things, errors in sampling, bad questions, untruthful responses, etc.
I still think Clinton has a long way to go (as does Obama) and some indicators are that Obama is going to be a far more formidable opponent than Clinton ever thought.