Are White People Unusually Self-Segregated?
Drawing on techniques from social network analysis, PRRI's 2013 American Values Survey asked respondents to identify as many as seven people with whom they had discussed important matters in the six months prior to the survey. The results reveal just how segregated white social circles are.
Overall, the social networks of whites are a remarkable [91*] percent white. ... In fact, fully three-quarters (75 percent) of whites have entirely white social networks without any minority presence. This level of social-network racial homogeneity among whites is significantly higher than among black Americans (65 percent) or Hispanic Americans (46 percent).
In an important way, as Jones writes, these numbers matter -- they show that white Americans are less likely to have interactions with members of other ethnic groups, and therefore might be less likely to understand where people of other races are coming from.
But I'm not sure that these results are surprising, or that they suggest a statistically disproportionate amount of self-segregation for whites. The U.S. population as a whole is 63 percent non-Hispanic white, 13 percent black, 17 percent Hispanic, and 5 percent Asian. This means that the baselines are different for different ethnic groups -- in an America with no sorting whatsoever, whites' social networks would still be 63 percent white, but blacks' social networks, for example, would be only 13 percent black. Similarly, just by chance, white Americans would be far more likely to fall into groups that are entirely same-race. (The odds of a white person's first friend being white as well would be 63 percent, the first two 40 percent, the first three 25 percent, and so on, multiplying by .63 each time. For blacks, the numbers would start at 13 percent and decline rapidly, to 2 percent, 0.2 percent, and so on.)
So, comparing the raw tendency to have same-race friends is highly misleading if we're trying to measure the propensity to self-segregate. And it's not clear what a better method would be. These kinds of comparisons are incredibly tricky -- especially because all the numbers are tied to each other (when one ethnic group avoids others, those other ethnic groups' social circles by definition become more homogenous too), and because the groups' sizes are so different (e.g., it's easily possible for blacks' social networks to be twice as black as the general population, but whites' networks would have to be 126 percent white for that to happen).
Robert VerBruggen is editor of RealClearPolicy. Twitter: @RAVerBruggen
* This change reflects a correction made to the Atlantic article, which originally said 93 percent.