Both in a lengthy print piece and in a brief Corner post, I’ve made a simple point about school segregation. Here’s how I put it in the blog post, which took issue with a New York Times op-ed:
[The op-ed claims] that school segregation is on the rise. As evidence, the authors point out that black children are increasingly likely to attend schools that are either majority or overwhelmingly (more than 90 percent) nonwhite.
There’s a huge problem with such statistics [known as “exposure” or “isolation” measures], however. . . . The Hispanic and Asian populations have risen dramatically in the past several decades. Indeed, public-school students nationwide no longer have a white majority. Of course black students’ schools have a rising share of minorities in them; everyone’s schools do.
If you measure segregation in a way that accounts for this, instead of just measuring black students’ exposure to white students, you find that there’s no increase.
This has generated some pushback at places like Vox and The Atlantic. Here’s my response.
First things first: The Atlantic piece badly mischaracterizes my views. It claims that yours truly “and others say that, with slim evidence of increasing segregation, policies designed to proactively integrate schools are an obsolete form of social engineering.”
I didn’t say that at all. This assertion suggests the author, Will Stancil, didn’t actually read my print piece, and indeed didn’t even read the blog post to the end.
The longer piece does take the position that court-ordered desegregation couldn’t have lasted forever, but it explains at length why segregation is bad and the policies we could use to combat it. My suggestions skew libertarian, for sure, but they include a new approach to charter schools proposed by the liberal Century Foundation.
Now, on to the arguments that can be engaged in good faith.
The Vox piece defends the use of “exposure” or “isolation” measures despite the issue that I raised with them. I had a productive and respectful Twitter conversation with the author, Alvin Chang, and I think the simplest way to put my side of the disagreement is this: Even an extreme measure of racial isolation — such as whether schools are 90 percent minority — means a different thing over time as the minority population rises.
“Segregation” refers to the degree to which people are separated relative to a baseline where they mix freely — and if you’re measuring segregation in “percent minority” terms, that baseline moves with the minority share of the population. A 90 percent minority school is certainly segregated today, but it’s less segregated than it was decades ago. In fact, if at some point in the future the whole school-age population is 90 percent minority, a 90 percent minority school won’t be segregated at all. Thus you can’t use a fixed cutoff like 90 percent to measure trends over time.
To be sure, you can worry about 90 percent minority schools’ becoming more common, even if that’s driven by a rising minority population rather than by increased sorting among existing groups. I said as much explicitly in my print piece. But that’s not a rise in anything that deserves to be called “segregation”; it’s a nearly inevitable result of demographic change.
Stancil, meanwhile, concedes the problem with exposure measures — “when diversity increases, some measures of segregation are likely to get worse, more or less by default” — but claims that an alternative I highlighted does the same thing in reverse:
As skeptics like Ver[B]ruggen point out, some measures of segregation, especially those that focus on the prevalence of white students, tend to look worse when student diversity increases. But other measures tend to look better. For example, one statistic known as a “dissimilarity index” calculates how many people would have to swap places to achieve demographic balance. When diversity increases evenly, dissimilarity indexes will improve—because the share of minority students in the least-integrated schools will grow, making fewer swaps necessary.
The problem is, you can use a dissimilarity index to measure specific types of segregation, including black–white segregation. Adding Hispanics to schools won’t affect the degree to which blacks and whites are separated. And as you can see even in this UCLA report — which focuses on “exposure” measures in the South and argues that schools are resegregating, but provides a dissimilarity index in Table 7 — black–white segregation is holding steady, even declining slightly, when measured this way. Even in the South, which has seen the end of a lot of desegregation orders, black–white segregation ticked up in the 1990s slightly but then held steady through the next decade (echoing a different finding I noted in my print piece).
One good thing both of my critics do, to be fair, is to flesh out that you can sometimes find issues at a more granular level, such as specific cities. This is not something I ever denied, though; indeed, in the third paragraph of my print piece, I noted that the districts that lost their desegregation orders became more segregated.
My broader point was that overall segregation has not increased, and I stand behind it.