The Bradley Effect

“What’s the No. 1 reason why many people won’t vote for [Barack] Obama?” asks Dustin Glick in his comic, Dustinland. See how he narrows down the options.

Glick’s theory is that many people oppose Obama because of his race, but rationalize their position, perhaps even to themselves, by citing any of a host of other respectable reasons. But at least the fact of their opposition is clear, whatever the honesty of their stated reasons. A far bigger danger to Obama is from voters who tell pollsters that they will vote for Obama, but actually don’t. This is known as the Bradley Effect, as Andrew Hacker explains:

As I write, several polls give Obama about 47 percent and McCain about 45 percent, a decline of several points for Obama from the polls of May and July. The rest of the respondents say they are undecided. At the same time the state-by-state estimates by show Obama leading in electoral votes, with 102 such votes a “toss-up.” The numbers will probably have changed by the time you read this. Yet now and later, there’s a chance that the real percentages will be the reverse of those I’ve cited. Some people who are telling pollsters they’re for Obama could actually be lying.

Such behavior has been called the “Bradley Effect ,” after Tom Bradley, a black mayor of Los Angeles who lost his bid to be California’s governor back in 1982. While every poll showed him leading his white opponent, that isn’t how the final tally turned out. Things haven’t been far different in some other elections involving black candidates. In 1989, David Dinkins was eighteen points ahead in the polls for New York’s mayoral election, but ended up winning by only a two-point edge. The same year, Douglas Wilder was projected to win Virginia’s governorship by nine points, but squeaked in with one half of one percent of the popular vote. Nor are examples only from the past. In Michigan in 2006, the final polls forecast that the proposal to ban affirmative action would narrowly prevail by 51 percent. In fact, it handily passed with 58 percent. That’s a Bradley gap of seven points, which isn’t trivial.

Pollsters contend that respondents often change their minds at the last minute, or that conservatives are less willing to cooperate with surveys. Another twist is that more voters are mailing in absentee ballots, and it’s not clear how those early decisions are reflected in the polls. Yet the Bradley gap persists after voters have actually cast their ballots. Just out of the booth, we hear them telling white exit pollers that they supported the black candidate, whereas returns from these precincts show far fewer such votes. Thus they lie to interviewers they don’t know and will never see again. Barack Obama wants to think “white guilt [over treatment of blacks] has largely exhausted itself in America.” I’m less sure. Almost all people who reject black candidates say they have nonracial reasons for doing so. And many undoubtedly believe what they’re saying. So I’m not persuaded that the Bradley gap won’t emerge this year. The Obama campaign would do well to print signs to post prominently in all its offices: ALWAYS SUBTRACT SEVEN PERCENT!

Of course, polls can also be misleading in the other direction: for example, pollsters often ignore people with cellphones, who tend to be disproportionately young and Obama supporters. There are too many factors at play for anyone, whether a pollster or a pundit, to successfully predict what will happen in November on the basis of anything but luck.

I hope Obama wins the election, though I say that more from process of elimination than anything else: the other side makes me cringe.

(Links via email from Gautam and Sanjeev respectively.)