Pollster.com's Mark Blumenthal does us all an important service today by evaluating and rejecting the idea that the consistent over-estimation of Barack Obama's vote in early (i.e., unadjusted to reflect actual voting) exit polls is explained by the so-called Bradley-Wilder Effect--the reluctance of interviewees to admit they are going to vote against an African-American candidate.

As Blumental points out, the BWE, if it exists at all, is more likely to occur when voters are actually interviewed, especially "live" and in person. But exit polling is relatively private--voters are handed a questionnaire and allowed to fill it out and deposit it in a box, just like secret-ballot voting itself. So if the BWE is truly a significant factor, a pro-Obama bias would be showing up in pre-election polls as much as or more than in exit polls, and that hasn't been the case other than in isolated instances (e.g., NH).

This matters because a widespread belief in the BWE might well lead observers to discount Obama's standing in general election trial heats against John McCain.

But Blumenthal's alternative explanation of the pro-Obama bias in exit polls is interesting as well: he attributes it to the same exit polling bias in favor of younger voters that was generally accepted as the reason for the pro-Kerry inaccuracy of the initial 2004 general election exit polls. This bias was supposedly addressed and dealt with after those elections, so it's re-emergence today is certainly annoying if nothing else. Or maybe I'm just reacting as one of those Democrats who saw the early exits in 2004 and spent much of the evening calling friends and family to tell them to ignore the red sea on network election maps, because I "knew" Kerry had actually won.