Andrew Sullivan and David Kuo invoke the possibility of a"Bradley" effect (or a Harvey Gantt effect, given that Gantt led in polls just before his 1990 election against Jesse Helms)--i.e., a segment of white voters lying to pollsters about their willingness to vote for a black candidate, only to vote for the white candidate in the polling booth--to explain last night's outcome.

Other factors surely contributed to Hillary Clinton's win--the effectiveness of Bill Clinton's very strong attacks on Obama, the"human" side of Hillary shown in the closing days--but it's hard to explain the stunning discrepancy in the polling (Obama's margin anywhere from eight to fifteen points below every pollster's predictions, including, it seems, the internal polls of the Clinton campaign) or the exit polling (Obama's margin running around five points below the exit poll figures) without taking into account a Bradley effect. The outcome is particularly striking in light of the last major Iowa poll (the Des Moines Register) getting the outcome correct. In caucuses, people have to identify their choice publicly.

The irony in this outcome, of course, is that Clinton has campaigned far more on the basis of identity politics than has Obama. But the New Hampshire result raises serious doubts about the validity of polls in the next major primary (South Carolina), which as of yesterday showed Obama in the lead, and running surprisingly well among white voters.