When the latest Field Poll showed Proposition 8 losing by 17 points, skeptical supporters of the measure that would ban same-sex marriage in California invoked a phrase coined in the 1980s: "the Bradley effect."

Named after the late Los Angeles Mayor Tom Bradley, an African American who lost the 1982 California governor's race after polls showed him leading handily, the phrase refers to the tendency of respondents to lie to pollsters about issues of race.

The Bradley effect has been referenced in presidential polls this year as Sen. Barack Obama seeks to become the first black president. And it is also being used to describe possible disparities between polls taken on gay marriage measures and actual election results.

A new study of elections in 26 states – including California – found polls typically understate voter support for these measures.

"Because the media portrays gay marriages as being politically correct, people don't want to be seen by pollsters as being intolerant – so they hide their views," said Frank Schubert, campaign manager for the Yes on 8 campaign, which conducted the study.

As an example, Schubert cited Proposition 22, which California voters approved in 2000. The Field Poll showed the gay marriage ban – overturned in May by the state Supreme Court – was backed by 53 percent of voters right before the election. But when the votes were counted, 61 percent of voters supported the initiative.

The survey looked at measures banning same-sex marriage, dating back to the first such campaign in Hawaii in 1998. According to the study, surveys published by news media outlets before an election underestimated support for traditional marriage by an average of seven percentage points.

In only two of the 26 states did pre-election surveys accurately measure voter sentiment. Support for traditional marriage was underestimated in 23 states. In one state, Arizona, support dropped.

Schubert believes the Proposition 8 race is much closer than the Field Poll shows – a contention that's not disputed by Steve Smith, who is managing the No on 8 campaign.

Mark DiCamillo, director of the Field Poll, said they could be right because some undecided poll respondents may not have been candid.

"It's possible there could be certain segments of voters who in a similar way might be withholding their true judgment and saying, 'I'm not sure' when, in fact, they are going to be voting for the 'yes' side," DiCamillo said.

In 1982, the last Field Poll showed Bradley, a Democrat, leading by five percentage points. Yet he lost to George Deukmejian, a Republican, by more than one point.

DiCamillo said a postelection analysis conducted by his organization found "nine out of 10" undecided respondents wound up voting for Deukmejian.

The analysis also cited other factors that worked against Bradley, including an aggressive GOP vote-by-mail campaign and a gun control measure on the ballot that increased Republican turnout.

But Patrick Egan, a professor of politics at New York University, analyzed the Yes on 8 study and concluded it overstated the gap between polls and election results because it did not adequately weigh undecided voters in the polls.

Egan said that while the share of supporters for banning gay marriage increased between the polling and the balloting, so did the share of opponents. But he does not deny the existence of the Bradley effect.

"Anyone who studies survey research will tell you one of the biggest problems we encounter is this notion of social desirability bias," he said.

Poll respondents generally overreport that they vote, for example, he said. But they underreport drug use and abortions, based on vital records.

"We might expect that as it becomes less acceptable to express anti-gay attitudes, a Bradley effect of sorts would kick in," Egan said.

The central question, he said, "is to what extent are these attitudes becoming so unacceptable that people would be reluctant to state them to a researcher."

Call Aurelio Rojas, Bee Capitol Bureau, (916) 326-5545.