Shankar Vedantam of National Public Radio reports on a fascinating new study by Elizabeth Ananat, Shihe Fu and Stephen L. Ross, which finds that racial wage gaps are much larger in the most populous U.S. metropolitan areas than they are in smaller cities:
Ananat explains the findings with a hypothetical example: “Say there are 1,000 black engineers in Silicon Valley, compared to 20 in Topeka, and there are 10,000 total engineers in Silicon Valley, compared to 500 in Topeka. Then blacks make up 10 percent of engineers in Silicon Valley, compared to 4 percent in Topeka.”
“A black engineer in Silicon Valley has 980 more black engineers to get spillovers from than does a black engineer in Topeka,” she writes in an email. “Meanwhile, a white engineer in Silicon Valley has 8,500 more white engineers to benefit from than a white engineer in Topeka. Thus, while both white and black engineers’ wages will be higher in Silicon Valley than in Topeka, the white engineer’s wages will increase more than the black engineer’s do — in effect, the white engineer is living in a much bigger city (of engineers) than the black engineer is, if only people within one’s own race matter for urban spillovers.”
Obviously, in the real world, social encounters are not totally segregated and other factors — including out-and-out prejudice — could play a role. But what seems to be happening, Ananat says, is that minority groups often miss out on the valuable tips and mentoring that make these ecosystems so productive and profitable. The same thing happens with other ethnic minorities, and even with whites — when they are in a minority.
Lydia DePillis, a reporter at Wonkblog, recently suggested that the findings of Ananat et al. make a compelling case for race-based college admissions. But there is a problem with this line of analysis. Consider the following from Mismatch, Richard Sander and Stuart Taylor Jr.’s study of the impact of large racial preferences and academic outcomes for underrepresented minority students:
And then there is a study by three economists that should be considered the gold standard in the field — one that had actual data on the friendship networks of tens of thousands of students attending a range of colleges from moderately selective to the most elite and that explicitly attempted to understand the role of academic indices and relative academic position on cross-racial interaction.
In “Representation versus Assimilation” economists Peter Arcidiacono, Shakeeb Khan, and Jacob Vigdor found that students are much more likely to form friendships at college with other students whose level of academic preparation is similar to their own and that this is true both for same-race friendships and cross-racial friendships. Where racial preferences are large, this directly dampens the number of cross-racial friendships. The authors point out that, given the patterns they observe and the consequences of the cascade effect, our current racial preference system sharply reduces the number of cross-racial friendships that would occur if schools used smaller preferences.
Specifically, Arcidiacono et al. find that “the interaction-maximizing degree of racial preference, while positive, is significantly weaker than that observed in practice.” That is, while DePillis might be right to suggest that racial preferences might facilitate the formation of cross-racial friendships, which can in turn contribute to cross-racial professional networks, it is crucially important that the preferences not be very large. Large preferences risk conributing to negative stereotypes about students from underrepresented backgrounds, as those who benefit from large preferences tend to have much weaker academic preparation than students who do not, and so they are more likely to struggle academically.
More broadly, I’d argue that Ananat’s findings are yet another reminder that anti-poverty policy needs to place much greater emphasis on reducing the geographical concentration of poverty, a phenomenon that tends to correspond to racial segregation. Improved crime control and higher quality schools will tend to make urban neighborhoods more attractive to the nonpoor. The relaxation of local land-use regulations, meanwhile, can see to it that “gentrification” doesn’t displace poor incumbents, but rather that it creates new job opportunities as more affluent new arrivals become a customer base for local less-skilled labor. These strategies don’t promise immediate gratification. But they could do a great deal to chip away at the barriers separating black and non-black lives over time, and they’d produce many other economic and social benefits as well.
(I should also note that the racial wage gap findings tend to reinforce the hypothesis advanced by Chenoa Flippen, namely that African Americans have been migrating from high-productivity regions to low-productivity regions in part to improve their relative status.)