Chris Matthews and some other mainstream media voices are pushing the theory that the "Bradley effect" or the "Wilder effect" explains what happened in New Hampshire Tuesday night -- that the polls were wrong because white voters told pollsters that they'd be voting for Barack Obama even though they couldn't actually bring themselves to cast a ballot for a black man.

It's fine as theories go, but the evidence doesn't seem to support it. As the Atlantic's Marc Ambinder explains, the polls actually got Obama's level of support pretty much right. According to Pollster's "standard estimate" based on pre-primary polls -- and if you want to know what that means, there's an explanation here -- Obama was drawing support from 36.7 percent of New Hampshire Democrats. With 96 percent of the precincts reporting, Obama had 37 percent of the vote. The polls were also pretty accurate in gauging support for John Edwards (18.4 percent in the polls, 17 percent on the ballots) and Bill Richardson (5.6 percent vs. 5 percent).

So what happened?

The polls didn't overstate the support for the black man (or any man, for that matter); they understated the support for Hillary Clinton. Pollster's "standard estimate" put Clinton's support at 30.4 percent as the polls opened. When the ballots were counted, she had 39 percent.

Thus, maybe the issue this morning shouldn't be whether white New Hampshire voters hid their racist preferences from pollsters (or even the Iowa corollary -- that in public caucuses as opposed to private voting booths, Democrats wanted to make a good show of voting for the black man). Instead, perhaps it should be why the pollsters didn't have a better grasp on the support for Clinton -- or how it was that undecided and late-changing voters left the race in play long after a lot of us thought it was over.