From the issue dated August 10, 2007 

How Voters Can Protect Against Inner Biases

Pundits were shocked in 2004 when exit polls wrongly showed John F. Kerry cruising to victory over George W. Bush. In the weeks that followed the election, statisticians and political scientists scurried to explain the apparent disparity between actual vote counts and what voters told pollsters as they left the booth. That fiasco, of course, came on the heels of the major exit-poll drama four years earlier, when television networks called the race early - only to reverse themselves several hours later.

Now, with the 2008 election approaching, you can expect still more statistical craziness ahead, particularly if the contest continues to include the diverse group of candidates now jockeying for position. What's more, the racial, gender, and religious mix in the field may lead to some counterintuitive blends of politics and policy.

Research shows that respondents often underestimate their own prejudices when talking to survey takers. Many voters would probably not go so far as to lie to a pollster about their willingness to vote for, say, a black, female, or Mormon candidate. Rather, their prejudices might operate on an unconscious level. Maybe when voters finally step into the booth, a certain issue would come to mind to sway them away from the candidate of a particular race, gender, or religion. For example, a putative Hillary Clinton voter might decide, at the last minute, that security concerns had been worrying him and that he just felt more comfortable with her opponent's potential toughness on terrorism. Or perhaps the people who answered that they would have no problem voting for a white Mormon, an Italian-American mayor, or a Kenyan-American would simply not manage to find the time on Election Day to cast their ballot.

The former mayor of Los Angeles, Tom Bradley (who is African-American), learned that lesson the hard way. Polls predicted his winning the race for governor of California handily in 1982. But when the results were tallied, he lost. Ditto for Mayor David Dinkins of New York, who lost to Rudolph Giuliani in 1993 despite a pre-election lead.

Whatever the psychological dynamic, the research is clear: Voters' insistent claims that they are not prejudiced don't hold up. Their answers suffer from what sociologists call "social desirability bias" - which means that voters are not expressing their true feelings, since they know there is a socially acceptable answer (treating all candidates the same regardless of race or gender). Social-desirability bias isn't just linked to elections. For instance, when asked questions about their views of blacks, white respondents consistently appear to have grown more tolerant over the course of the last four decades.

However, recent research by Hannah Brückner and Alondra Nelson of Yale University and Ann J. Morning of New York University has shown that when they hide questions about race among other decoys, they elicit "true" responses that tend to be more racist. What they do is ask a control group how many of three statements they agree with (these are political opinions that have nothing to do with race, chosen such that most people would not agree or disagree with all of them). Then with another group, they throw in a fourth statement about race, such as, "Genetic differences contribute to income inequality between black and white people."

If the average number of statements agreed to in the first group was one and a half out of three and for the second group (with four statements presented) was two, then we can conclude that 50 percent of respondents agreed with the statement about race. Surprisingly, perhaps, the difference in results when the question is hidden - compared with asking directly about bias toward blacks - was greatest for highly educated people and for women. That is, it is these groups that were most susceptible to social-desirability bias with respect to their racial attitudes.

Likewise, the Harvard University-based Project Implicit, which uses reaction time and other subtle cues in a test of word and image pairings, has found that most people harbor some unconscious biases, despite what views they express explicitly. More than 4.5 million visitors have logged on and taken one of the project's tests so far, and results are still pouring in. One study found that about 88 percent of whites who took the test showed bias against blacks. What's even more shocking is the fact that sometimes people have hidden biases toward members of their own group. About half of blacks, for example, demonstrated antiblack bias.

There are some recent hints that polling bias could be abating. The discrepancies between pre-election numbers and actual returns for Deval L. Patrick, who was the first African-American to be elected to the Massachusetts governorship, and for the former congressman Harold Ford Jr., who failed in his bid to represent Tennessee in the U.S. Senate, were significantly smaller than in other high-profile races pitting a black candidate against a white one. But two cases in a year of Democratic landslides hardly make a solid trend, especially when we are confronted with so much experimental evidence to the contrary.

The irony is that although our unconscious biases may affect how we vote, we don't really know how the candidates we choose will perform. Race, gender, or religion often work in counterintuitive ways. We shouldn't be fooled into thinking that the color of a candidate's skin, or last name, or gender is a foolproof indicator of how he or she is going to vote on affirmative action, immigration, work-family issues, gay marriage, or any other hot-button issue of the day.

What's more, in the twisted world of American politics, it is not always straightforward who is the best person to carry the mantle of a certain cause. Wasn't it Lyndon B. Johnson - a white Southerner - who was most effective in championing civil rights? Isn't it sometimes easier for a man to argue for gender equity at work? Or a straight leader to make the case for gays in the military? Or a former prisoner of war to argue for normalization of relations with Vietnam? After all, Americans generally don't want politicians to cater to their own special interests but to speak to a broader public.

George W. Bush (or rather Karl Rove) was savvy enough to know that the left wing of American politics has painted itself into a corner with its overemphasis on identity politics. The result? He has given us a Hispanic secretary of housing and urban development, an Asian secretary of labor, a black male secretary of education (followed by a white woman in that slot), and two black secretaries of state (to name just a few) - the most diverse cabinet in American history, at least with respect to gender and race. Ideologically, of course, it's a different story. There's a lesson here for 2008 voters: Pay attention to issues, not identities.

The real question should be which candidate presents the best policies and which has the ability to carry out those plans.

So, in the media-soaked world of contemporary U.S. political discourse, how can the average voter protect him- or herself from hidden biases and the false lure of identity politics - thereby making the right choice in this important election?

Some folks remember the divergence of opinions about the first Nixon-Kennedy debate in 1960: Those who watched it on television gave JFK the nod over his less-well-groomed opponent, while those who listened to it on the radio thought Nixon had won. Such discrepancies should be all the more common in the age of YouTube - especially when we are talking about skin color or gender and not just Nixon's 5 o'clock shadow.

So here's an experiment for 2008 (yes, try this at home): If you are really concerned about which candidate will do better for blacks, Latinos, women, or any other identity group, get a friend to print out the candidates' stump speeches, interviews, and policy statements - and delete any mention of their identity with a black marker. Then read the text and decide whose policies you think would benefit disadvantaged Americans the most, who has the clearest plans, and who is most likely to be able to see those plans through. Find out whose policies you were reading and vote for that person. Or, if you don't have the energy for that, at least close your eyes when the candidates are speaking, and focus on the words. We shouldn't be fooled by either our own hidden biases or by the promise of personal identity. It's the policies (and the ability to carry them out), stupid.

Dalton Conley is chairman of the department of sociology at New York University. Among his books is Honky (University of California Press, 2000). 
Section: The Chronicle Review
Volume 53, Issue 49, Page B12