When Barack Obama spoke with the standup comedian Marc Maron on his WTF podcast, he touched on his beliefs about progress and racism:

I always tell young people in America, “Do not say that nothing's changed when it comes to race in America unless you lived through being a black man in the 1950s or Sixties or Seventies.” It is incontrovertible that race relations have improved significantly during my lifetime and yours and that opportunities have opened up and that attitudes have changed. That is a fact. What is also true is that the legacy of slavery, Jim Crow, discrimination in almost every institution of our lives, that casts a long shadow and that's still part of our DNA that's passed on. Racism, we are not cured of. And it's not just a matter of it not being polite to say 'nigger' in public. That's not the measure of whether racism still exists or not. It's not just a matter of overt discrimination.

It's notable because the president said it, but that belief—of progress on explicit racism and explicit, de jure discrimination, with continued problems of covert and structural racism that are harder to diagnose and cure—is pretty common. And insofar as we can measure these things, it's probably true, or at least our measurements reflect that narrative.

For over 40 years, the University of Chicago's NORC has been running the General Social Survey (GSS), one of our most powerful tools for gauging what's in America's heart; among other things, it's had many questions for America about race and discrimination. And those answers have been going good: would you quit a club that refused to admit blacks? More people would. Do you support laws against interracial marriage? Fewer people would. Every measure of prejudice, at least with regard to blacks, has declined.

But, as mentioned, the president is far from alone in believing that subtler forms of racism not only exist, but are extremely powerful despite their subtlety. And there's lots of social science to back up lived experience.

These ideas cross in a fascinating study by the University of Chicago's Kerwin Charles and Northwestern's Jonathan Guryan. They took those GSS questions about prejudice, the ones showing such positive trends, and compared them to black-white wage gaps by region.

Their most immediate findings are not surprising. The most prejudiced regions (the South, basically) had the highest percentage of blacks in the workforce and the highest black-white wage gap. And overall the correlation between higher prejudice and lower black wages is strong.

But then they find something very, very interesting:

First, we show that racial wage gaps are much more closely related to the level of prejudice of the “marginal” person in the distribution than they are to average levels of prejudice. Even more striking, we show that it is only prejudice in the left tail of the prejudice distribution that seems to matter for wage gaps; wages do not vary at all with the prejudice of the most prejudiced persons in a state.

In other words, what matters to the black-white wage gap isn't how prejudiced the most prejudiced whites are, it's how prejudiced the least prejudiced whites are. Even the average prejudice doesn't matter. And it shows up in market prices: "a mechanism linking wages and prejudice, operating through the market’s ability to separate races at work, determines the racial wage gap."

That might seem intuitively obvious—if the people who are most likely to hire or promote a black employee are more prejudiced, more of those doors are closed. Meanwhile, the difference between a person who's really racist and one who's really, really racist doesn't matter, because those doors weren't going to open. (Granted, one can imagine that a place or region with a lot of very racist people might influence the least prejudiced among them into being somewhat more racist, by making lesser prejudice comparatively more acceptable.)

But consider how we often frame racism—the Confederate flag, the racial slur, representations of explicit racism, and racists. Those are important in their own way, but for this specific and very important issues of disparate wages, the findings suggest that they don't, at least not directly. It's there on the margins that the gains stand to come from.