Among voters lacking strong party preferences, Obama faces 20 percent handicap due to race bias

"Although they may not determine the election outcome, race biases are having a strong anti-Obama effect among the least politically partisan voters," said Anthony Greenwald, a University of Washington psychology professor who conducted the survey. "If present pre-election polling is accurate, the effect of racial attitudes will have their effect on Barack Obama's winning margin but not on the election outcome."

Most recent polls show Obama in the lead. This suggests that although race biases are a "hill he has to climb, the polls indicate that he's actually climbing it," Greenwald said. "People who have race biases against Obama may still believe he's preferable to Mitt Romney for other reasons, and so race attitudes do not appear to be potent enough to overcome the other sources of favorability for Obama."

About 8,600 eligible voters participated in Greenwald's latest online survey, collected from July through September. Eighty percent of the respondents were white, approximating the electorate. However, because the sample had a substantial majority of liberals it should not be considered representative of the American electorate.

The study included the Implicit Association Test to measure implicit racial attitudes, which are preferences that people may not realize they have. In the past 10 years, various adaptations of the tool have been used to reveal unconscious attitudes about race, gender, sexuality, ethnicities and other topics.

In addition to implicit and self-reported measures on race attitudes and demographic characteristics, Greenwald included various measures of political beliefs, including stances on policy issues such as taxes on the wealthy, immigration and health care.

In his sample of voters, Greenwald found that 25 percent of the most ideologically polarized voters appeared to be already settled on their preferred candidate. In this group, race attitudes appeared to influence votes for only about 2.4 percent of participants.

"These people strongly favor one candidate over another, and race attitudes have only a relatively small effect on their vote," Greenwald said.

But then he took a closer look at the race attitudes for the 25 percent who were least polarized -- having no strong affiliations with either political party. The research showed that race attitudes were influencing the choice of candidate for nearly 10 percent of these voters. If 10 percent of voters switched their vote for or against a candidate, that would cause an increase of 20 percent in the gap between candidates.

"Our method was like putting a magnifying glass on the distribution of voters, allowing us a close-up look at various segments," Greenwald said. Those at the far ends strongly favoring Obama or Romney were relatively little affected by racial attitudes. But those closer to the middle of the distribution showed a considerably greater influence of racial biases.

Greenwald points out that any Democratic presidential candidate, including white ones, faces a race attitude handicap. This is because voters with white racial preferences are typically opposed to social and economic policies favored by national Democratic nominees.

Greenwald's findings are consistent with data he released in May, showing that voters' race attitudes leading up to the Republican primary already had a more pronounced role in predicting their vote than in the 2008 presidential contest.

He will continue to collect survey data as part of the year-long Decision 2012 IAT project. Data collected through October will be used to determine whether race attitudes still have as large of an effect on undecided voters.

Other collaborators on the project are Brian Nosek and Sriram Natarajan, at the University of Virginia, and Teri Kirby, Kaiyuan Xu and Sianna Ziegler, all at UW.

Anyone can participate in the online study, which will be available until a few days beyond Election Day, Nov. 6: https://implicit.harvard.edu/implicit/demo/featuredtask.html.