My Bias Against Measuring Bias

The latest Washington Post/ABC News poll shows Hillary Clinton solidly besting her potential Republican opponents. It even suggests by two-to-one, she benefits from her status as the potential first woman President (although far more say it makes no difference). But can we trust voters' own perceptions of their prejudices? A recent Pew Research study further highlights the difference between asking about bias, and asking about one's own bias.

Pew explores women's lack of parity in both the high levels of business and in politics. It asks respondents very directly: "Which comes closest to your own view: men generally make better (political/business) leaders, or women generally make better (political/business) leaders?" Unsurprisingly, as in the Washington Post/ABC poll, overwhelming numbers of both men and women say gender doesn't make a difference. In fact fewer than a fifth of men, and even fewer women, say men make better leaders in either industry.

These results remind me of this 2007 Washington Post/ABC News survey, which showed more would vote against a smoker than would vote against a woman, African-American, or Mormon. Results like these almost certainly reflect the socially desirability bias, rather than true bias.

Other questions from Pew can help us further detect actual bias. Far more believe women face discrimination in our society (57%). Women are particularly likely to sense the existence of gender bias than men; two-to-one among women, while men are evenly divided. Women are also far more likely than men to cite external discriminatory factors, like "being held to a higher standard," for the lack of parity in business and politics.

Gallup shows a similar pattern for race; their December survey revealed twice as many blacks as whites feel "mostly discrimination" is responsible for blacks' "inferior income, jobs, and housing situation."

One's perception of discrimination, therefore, likely depends greatly on whether you are part of the discriminated group. But even these questions are not measuring true bias, but the perception of it.

Taking all this into account means it's tricky to draw conclusions about bias from responses to explicit questions about Clinton's gender. Not everyone is equally aware of bias, and even fewer are willing to admit their own.

Further, in the current Washington Post/ABC poll, party identification -- not gender -- seems to be a bigger driver of one's view of Clinton's gender. By 13-to-one, Democrats say they are more likely to vote for her because she's a woman, while Republicans, by three-to-one, say this makes them less likely to support her. Yet both men and women say they are more likely to vote for her because of her gender. (And thanks to the kind folks at the Washington Post, we can confirm the party patterns transcend gender.) Which is at work here -- party differences in the importance of gender, or party differences in views toward a Democratic candidate?

So we'll need far more than a single question to decode the role gender has in Clinton's political fortunes. We would need to control for party, and mask that we're investigating bias. Because if voters are viewing Clinton through the lens of party, rather than gender, perhaps that may be progress after all.