The first thought out of New Hampshire was that Obama was felled by the dreaded Bradley/Wilder effect -- the tendency of voters to overstate their support for black candidates in polls, and abandon them in the booth. But Matt Yglesias posted a series of polls showing that Obama got exactly the percentage he was predicted to receive; it was Clinton who increased her vote share among late-deciders. So maybe she just closed better. But now we find that, among late-deciders, 39 percent went for Clinton and 36 percent for Obama, so they're not the answer. According to Andrew Kohut, a pollster for Pew, the disconnect is in methodology. Clinton beat Obama among working class voters, while Obama won more affluent demographics. This distorted the polling "Poorer, less well-educated white people refuse surveys more often than affluent, better-educated whites," Kohut writes. "Polls generally adjust their samples for this tendency. But here’s the problem: these whites who do not respond to surveys tend to have more unfavorable views of blacks than respondents who do the interviews." So it's not really a mysterious effect, it's the systematic underrepresentation of a certain demographic in the polling. Or, at least, it could be. But why didn't it happen in Iowa? As the thinking goes, this effect is negated in caucuses because their public nature exerts communal pressure. But plenty of folks caucused for Clinton or Edwards without fear of public reprisal. So someone needs a plausible hypothesis of why Granite Staters are more racist than Iowans. One other possible explanation is that gender made the difference. The really salient electoral shift between Iowa and New Hampshire was that, in Iowa, there was no gender gap. Obama won women and men in exactly the same proportions. In New Hampshire, there was a significant one, and women supported Clinton in much higher numbers. This could be because John McCain, who didn't really compete in Iowa, drew Obama-supporting men from the Democratic primary and into the Republican race. It could be because the caucus effect really does exist for gender, as husbands and wives caucus together, and pressure one another to vote similarly. It could be because the campaign, in the final days, hit a lot of gendered notes, from the apparent Edwards/Obama gang-up on Hillary at the debates (interestingly, voters who said the debate was important broke for Hillary, while those who said it was unimportant went with Obama), to the media's distasteful frenzy over Clinton choking up at a New Hampshire diner. Or, the final answer could be that we just don't know, and never will. As my friend Chris Hayes put it recently, the voter's mind is the black box at the center of politics. And we've got no way to access it. Health care is so much more straightforward.
