The Bradley Effect has been in the news a lot recently. The “Bradley Effect” refers to fact that African American candidates in many cases have polled much better than they actually did in the final voting results. The effect is named for former Los Angeles Mayor Tom Bradley who consistently did much better in pre-election polls than he did on election day.

Many scholars attribute the “Bradley Effect” to a subtle form of racism. They believe that a percentage of white voters are unwilling to admit to pollsters that they will vote against the African American candidate for fear of being viewed as a racist. Hence, they say one thing to a pollster, but vote the other way when alone in a voting booth.

Has the Bradley Effect personally affected Barack Obama in the Democratic primaries? And is there a chance it may impact the general election?

To answer these questions, I discounted caucus states because caucuses bear no resemblence to a general vote. Obama was hugely successful in caucuses, but that shows that he is good at motivating small groups, not winning a popular vote.

The first Democractic primary was in New Hampshire. Here’s what the polls said the day before that election:

Rasmussen Obama +7
CBS Obama +7
CNN Obama +9
Zogby Obama +13

Clinton won by +3, meaning all of the above polls were way out of their margin of error. Bradley Effect? Possibly. Some attributed the late surge by Clinton to her shedding tears at a town hall meeting. Somehow, I doubt that had any impact on voters, but there is no empirical way of knowing that. All we know about New Hamsphire is that the polls were way off and everyone was surprised. Obama himself has stated in speeches that the New Hampshire primary should serve as a lesson against overconfidence.

In California’s primary on Super Tuesday, Obama lost by a landslide 10 percentage points. The results surprised me because this is what I read in the polls the day before the primary:

Rasmussen Obama +1
Zogby Obama +3
Field Clinton +2

The polls were wrong and again way out of their margin for error, and no one can really explain why. Bradley Effect? The Bradley Effect is named for a California mayor, so that seems like a possible candidate. Also, nothing happened on the eve of election day to explain a dramatic shift in voting. Again, there is no empirical way of knowing why the polls were so off.

Most of the other big state primary polls were pretty accurate. Texas, Ohio, Indiana, and Pennsylvania all finished pretty close to the margin of error of most pollsters. Also, in Virginia, Obama did 12 points better than the polls indicated. Thus, if there was any Bradley Effect at all, it occurred in two isolated instances: New Hampshire and California.

Of course, the Bradley Effect would only need to manifest itself in a handful of states to swing this election to John McCain. Just as we were all surprised in New Hamsphire and California, we might be surprised by Florida, New Mexico, Colorado, Nevada, and Pennsylvania.
Democrats should not celebrate until all the votes are counted, because the primaries showed us all that the polls just might be way off.