The ‘Bradley effect’ myth

Much has been said and written in recent weeks about the so-called Bradley effect, the supposed phenomenon of white voters who secretly harbor racist animosity toward black candidates but who are too embarrassed to admit that to pollsters and thus lie about whom they intend to support. The effect is said to have lulled then-Los Angeles Mayor Tom Bradley into complacency in his 1982 run for governor of California, and to have produced misleading poll results that suggested Bradley would win. Instead, he lost by 1 percentage point, about 100,000 votes.

There is no doubt that racism lingers in American politics, but much more has been made of the Bradley effect than is justified -- take it from us, we were here the night it was born. In fact, the lessons of the Bradley campaign against George Deukmejian have far less to do with the viability of black candidates than they do with the limits of polling.

In that campaign, Bradley supporters begged gun-control advocates not to place a restrictive gun measure on the same ballot as the governor’s race, but their pleadings were ignored. The gun measure drew large numbers of first-time voters, many from the interior of the state, who went to the polls specifically to protect their guns and, while there, rejected Bradley as well. Because so many were novice voters, they were underrepresented in preelection polls that highlighted “likely voters,” a generally sound approach but one that missed the gun surge in 1982. They also were not picked up by exit polls because many distrusted pollsters and refused to answer their questions.

That election also was one of the first in California to feature a large absentee vote, commonplace today but relatively novel then. With so many ballots mailed in, particularly from conservative areas, the exit polls were sidestepped and the results skewed.

Regarding today’s election, there are two more reasons to doubt whether the Bradley effect, if it exists, will play much of a role. First, those who oppose Barack Obama because of his race can certainly find other ways to dress that up for a pollster. There are, after all, ample reasons to support John McCain, whether because of his experience -- or Obama’s lack thereof -- or his positions on issues from Iraq to taxes. Voters’ inclinations thus should be captured accurately, even if their motivations might not be. Second, this election may feature a jump in the number of younger voters who cast ballots. They appear to tilt heavily toward Obama and are more likely to rely on cellphones, which pollsters have yet to figure out how to contact. Thus any racists who conceal themselves from pollsters may be counterbalanced by voters who are simply unavailable to them.

Polls can be wrong. Those that show Obama ahead as election day opens could be as misleading as the 2004 predictions that John Kerry was the likely winner. If McCain wins, it won’t likely be proof of the Bradley effect, which demonstrated not so much that respondents lie to pollsters as that pollsters can’t know what they don’t know to look for.