Many argue that there is no such thing as a “Bradley effect.” Others think that it has faded in recent years.
I don't believe that. My gut says there is such a thing. Not that people are disinclined to vote for blacks--people are disinclined to tell pollsters how they really intend to vote if they think their choice is less than socially acceptable. This favours blacks, women, and perceived front-runners in the polls as against the actual vote.
If there is a Bradley effect, we will not have seen it yet in this US election cycle. In the primaries, Obama was running against a woman, and a Democrat, who was expected to win. Nothing “politically incorrect” in choosing a woman Democrat. Even so, Obama most often underperformed the polls, and did best in caucus states.
But admitting one prefers a white Republican man to a black Democrat might be a different story. Even without the race factor, Republican candidates usually do better in the actual vote than in the presidential polls. And Obama is now widely expected to win. Intrade has Obama at 78.4, McCain at 22.7.
So, if there is going to be a Bradley effect, how big is it likely to be?
I looked it up. It is not that easy to calculate—it depends on which pre-election poll you assume is most reliable, and what other factors might have intervened between poll and election day. But for Bradley himself, it was “in the double digits”--the polls were more than 10% off the actual vote. It is also sometimes called the “Wilder effect”--for Douglas Wilder, it was 8.5 to 10%. For Harold Washington in Chicago, it was about 10%. For Dinkins in NYC, it was about 12%.
So—10%, on average. That means that, if it exists, Obama needs a 10% lead in the polls to win.
Currently, Real Clear Politics has him at 7.4% up.