Pollsters Debate 'Bradley Effect'

Black candidates have increasingly won elections outside majority black districts, including the races for lieutenant governor of Colorado, a state with relatively few black people, and for attorney general of Georgia, a state with a troubled racial past.

Improved polling also may have helped produce more accurate predictions in contests such as Harold E. Ford Jr.'s losing race in 2006 for a Tennessee Senate seat and Deval L. Patrick's successful run for Massachusetts governor that year.

Dawson, however, remains skeptical about the willingness of whites to vote for a black candidate -- and the ability of polling to capture that reluctance -- in a high-profile, racially charged presidential election.

"We're talking about different levels," he said. "President is different than mayor of Chicago."

Experts agree that it is often difficult to fully tease out the extent to which race plays a factor in voting decisions. People can be reluctant to talk about their racial attitudes, and plenty of reasons -- party, age, experience, political philosophy -- can explain why voters may support or oppose a black candidate.

Still, there is little reason today, some experts contend, for people answering public opinion polls to hide their true intentions.

"For people to lie, there generally has to be a stigma attached to telling the truth," said Andrew Kohut, president of the Pew Research Center. "There is none affiliated with saying, 'I'm voting for Hillary' or 'I'm voting for McCain.' "

Kohut theorizes that polling discrepancies do not come from respondents who lie, but from people who decline to participate in polls. That is a growing problem, with studies showing that as many as half the people contacted for polls refuse to participate. Kohut recently conducted a study in which interviewers spent months repeatedly calling people back until they agreed to talk. He said that helped him see who is often missed in polling.

"Poorer, less-educated whites don't like to do these polls as much as better-educated people do," he said. "The refusals come from the same class of people who tend to be the most racially intolerant."

Anthony Greenwald, a professor of psychology at the University of Washington, said he also does not buy that people are lying to pollsters. "What I do buy," he said, "is that there were lots of undecided people who didn't have an answer before the phone rang and were generating one on the spot."

Greenwald, who has studied the primary contest between Obama and Sen. Hillary Rodham Clinton, said that when people in polls are prodded to answer a question, they know that, unlike in the voting booth, their response will have no consequences. So they may say they are supporting a candidate they have not actually decided on.

Pollsters say they build in controls to account for possible hidden racial feelings that can skew results. Kohut said he tries to elicit more-honest answers by matching the race of the interviewer and the respondent. Others try to push people to test the intensity of their backing of a particular candidate and often toss out whites who express tepid support for black candidates.

But Jon Krosnick, a professor of political science, communication and psychology at Stanford University, noted that black callers tend to get more pro-Obama answers in surveys than white callers do, no matter the race of the respondent.

"We don't have solid evidence that matching increases accuracy," said Krosnick, who does not believe the Bradley effect is real.

Harvard political scientist Daniel J. Hopkins analyzed elections involving African American candidates for governor and the Senate and found there was a Bradley effect when racially charged issues dominated the political discourse in the 1980s and early 1990s. As issues such as crime and welfare faded from the national scene in the mid-1990s, Hopkins wrote, so did the Bradley effect.

That raises the possibility that a return to racial issues could once again cause the phenomenon to reemerge, either nationally or in a key state.

"The most likely circumstance that could bring back the Bradley effect would be a racialized campaign," said Hopkins, a lecturer in Harvard's department of government. "If we spend the next month debating Jeremiah Wright or other racial issues, that would be the thing that would be on people's minds."

A spokesman for Obama said the campaign does not believe race will be much of a factor in voting. "People are more concerned about the state of the economy and our place in the world, and not so much concerned about ethnic or identity politics," said Corey Ealons.

Both those who believe the Bradley effect is a factor and those who dismiss it agree that, given the aura of history surrounding the current campaign, interest in it is high. "At literally every speech I make, I get questions on it," Newhouse said.

They also agree that this presidential election will be a highly visible test of just how real it is.

"If we don't see it now, then it's gone," Dawson said.

Polling director Jon Cohen contributed to this report.