How did they get it so wrong?
Katherine Q. Seelye at the Times politics blog has this simple but depressing explanation. "[The Iowa result] may lead to speculation that in a secret ballot, white voters will not vote for an African American. In Iowa, which Mr. Obama won, the caucuses are public." In other words, white voters may tell a pollster that they're for the African-American candidate in order to appear enlightened, but when alone in a ballot booth their baser instincts take over.
Ms. Seelye's colleague Janet Elder had previously published a longer piece on this topic, the first paragraph of which foresaw what happened last night in Iowa:
In high-profile contests where one of the major party candidates is black, pre-election telephone polls have often been wrong, overstating the strength of the black candidate. In polling circles this is known as the “Bradley effect” or the “Wilder effect” or the “Dinkins effect.” Will it also be known as the Obama effect?I don't purport to know much about polling, but this argument and last night's result is going to affect (can I say color?) the way I look at prediction and results from now through November.
*UPDATE* I see this morning that Andrew Sullivan blogged essentially the same theory last night. Complete with Wikipedia link no less.