The Bradley Effect

As the polls continue to show a significant lead for Barack Obama, some question whether those polls are distorted by something called the "Bradley Effect".

In 1982, the Democratic Party nominated an African American man named Tom Bradley for governor of California. Both pre-election polls, and election-day exit polls, showed Bradley ahead of his Republican opponent, George Deukmejian. But, when the vote count was completed, Deukmejian emerged as the winner.

Some speculate that a certain number of voters who did not want to vote for Bradley, because of his race, were reluctant to reveal their true intention to the pollsters. Therefore, those polls overestimated the degree of support for Bradley.

Are Obama's poll numbers similarly inflated?

It seems to me there's no way to get at that answer for sure. What are pollsters supposed to do? When someone says they're going to vote for Obama, should the follow-up question be: "did you lie?"

Here is an article by someone who says he was involved with that 1982 California campaign, arguing that the Bradley Effect does
not exist.

And here, Democratic political consultant Donna Brazile, backs up that Republican argument.

Many bloggers apply that old phrase "media hype" to this whole idea. They're referring to writings such as this blog post by an ABC-TV reporter, carrying on the speculation. Are the TV news people looking for ways to maintain interest in a race whose outcome seems more certain with every passing day? It's quite possible.

Some say the size of the distortion is around six percentage points. If I apply that rule of thumb to the analysis I published yesterday of Obama's lead in individual states, i.e., reduce Obama's lead by six points in each state, he still wins with 277 electoral votes.