The question of school choice and segregation has been a common recurring theme in education policy over the past 5-10 years (most recently in response to this Associated Press story).

Critics claim that school choice, specifically charter schools and private school vouchers, exacerbate segregation by income and especially race and ethnicity, primarily because more motivated parents with greater resources will exercise choice. Choice supporters, on the other hand, counter-argue that regular public schools are highly segregated due to the lack of choice (i.e., due to school assignment based on residence), and that school choice, which severs that tie, might lead to greater integration.

Given the proliferation of charter schools (and, to a lesser extent, vouchers) over the past 20 years, this is clearly an important debate. Yet the issue of choice and segregation entails three major underlying empirical questions that are sometimes blurred. It may be useful to discuss them briefly.

First, however, it bears mentioning the fact that some form of segregation is not necessarily some unintended consequence of school choice. It is part of the deal (at least if school choice is working the way it’s supposed to). The idea of choice is that parents can identify the schools that are the best fit for their kids, thus allowing for better student/school matching. If this is, in fact, happening, then it is more than plausible that we should observe that students are sorted systematically into schools by certain traits – whether due to their exercising choice (e.g., attending a charter) or remaining in a regular public school (remember that choice-based segregation is as much about who stays as who leaves).

Consider, for example, a school that focuses on (or has a reputation for) serving special education students, or catching students up who are far behind, or serving high-performing students, or language immersion, or fostering artistic and creative skills, or any number of other measurable and hard-to-measure characteristics. Now, of course, those traits may not all be associated with the “traditional” segregation variables, most notably income and race and ethnicity (though some of them are and, as such, would affect "traditional" segregation). Still, if students with these characteristics are drawn to move to or stay in certain schools, this too is segregation, albeit a kind that is not always a "bad" thing.

To be clear (and obvious), residential segregation, particularly between districts, is the primary cause of school segregation. School choice regimes can either make it worse or better (or have no effect). It really depends on who exercises choice, where they attended school before, and where they ended up. But if you buy the idea of parents choosing certain schools to match the characteristics, needs, and preferences of themselves and their children, you have to be careful about implying that students exercising choice are randomly distributed across schools. The whole idea is that it’s not random.

That said, let’s move on to the three key empirical questions underlying this debate.

Are schools of choice (i.e., charters and vouchers) more or less segregated than regular public schools? This is the easiest of the three to address. One can compare segregation between sectors (i.e., public versus private, regular public school versus charter) in a given year, or calculate segregation between sectors (e.g., Di Carlo and Wysienska-Di Carlo 2017). And you really need to do this within a given area, such as a district. Doing it nationally or even statewide (as in the AP story linked above) is not very helpful, especially for charters, which are concentrated in cities.

(And, of course, the choice of measures must be appropriate for these comparisons – see herefor more details on this important issue.)

This can tell you, most basically, whether and how segregation is worse in one sector or the other in a given district or area at a given time. This is a descriptive exercise, but it is important and policy-relevant – if, for instance, segregation within a city’s charter sector is unusually high, policymakers might focus any desegregation efforts there.

Do school choice systems cause more or less segregation? Describing segregation and comparing it between sectors is important, but it cannot by itself support conclusions about whether choice actually causes segregation. For one thing, from a causal purist perspective, such conclusions imply the question of what the situation would be if charters and/or vouchers did not exist. If, for instance, parents in a given city or neighborhood didn’t have access to schools of choice, they might very well relocate. This issue is extremely difficult to address, and so it’s tough to make grand, universal claims about the segregation effect of choice. There are, however, decent alternatives that allow conclusions about whether existing choice systems in a given place have an effect on segregation outcomes.

Remember that the impact on segregation of a student moving from a regular public to a charter or private school depends on the characteristics of that student and her peers in her old and new schools. A white student vouchering out from a heavily minority public school to a mostly white private school exacerbates segregation, whereas an African-American student making the same move would have a desegregating effect.

So, for example, researchers with longitudinal student-level data can follow students and compare their old and new schools to calculate the “net effect” of choice-based moves on overall segregation (see, for example, Zimmer et al. [2009]). This can provide a pretty good sense of whether the flow of students between choice-defined borders is helping or hurting with segregation, at least by easily measurable characteristics such as race and ethnicity. Any claims about whether school choice “causes” segregation should rely mostly on this type of analysis, with the standard caveat that the findings may not generalize to other locations.

Why does (or doesn’t) school choice increase segregation? This brings us to our most difficult question. It is infrequently laid out explicitly, and thus may be partially responsible for some of the contentiousness and impasse in this debate.

Suppose, for example, we find that, in a given city, students are, on average, moving from regular public schools to charters or private schools (on vouchers) in a manner that increases or decreases segregation. Why is this happening? Why might it happen?

The standard procedure among too many (but not nearly all) participants in our school choice debate is to argue about the conclusions and assume the causes. For example, as mentioned above, choice opponents tend to assume that more motivated parents with greater economic and social capital are more likely to exercise choice (or are more likely to choose certain schools), thus contributing to segregation. Many advocates, in contrast, rely on the theory that decoupling residence from school assignment will lead to greater dispersion of students across schools.

Both of these assumptions are at least plausible, but difficult to prove (particularly the latter, which is more of a general causal framework than a specific mechanism). Yet they are not the only possibilities. For instance, segregation outcomes may be influenced by state-/district-level choice policy design and implementation (e.g., if navigating the choice process is particularly complex, means testing of voucher programs). School policies and practices might also have an effect (e.g., charging fees, recruitment and retention strategies). So too might concentrations of schools of choice in certain neighborhoods, or the characteristics of these schools (e.g., magnets, specialized schools, high schools), or a number of other factors.

This matters, first of all, because these factors almost certainly operate in some contexts (e.g., districts) and not others. They might also be at play concurrently, and work at cross-purposes in terms of their impact on segregation. So, in a given choice system, you might find no change in segregation outcomes simply because some factors are exacerbating segregation and others attenuating it. In this case, you might conclude that there’s no relationship between choice and segregation, when the truth is quite the opposite.

Causality also matters because different underlying causes might carry different policy implications. If, for example, choice has a segregating effect because more motivated/resourced parents are more likely to apply, then it is possible that simplifying the process would help. If it’s due, in part, to where charters are located, this might be examined and addressed via the authorization process. And so on.

(Side note: There is an argument here that private and charter and private schools, which are far less constrained by residence, are in a better position than regular public schools to address segregation directly.)

Again, it’s very difficult to isolate all these possible effects in a rigorous manner. But my purpose here is not to say “it’s complicated” and walk away. The point, rather, is that the debate about choice and segregation would benefit from more explicit and nuanced attention to why there is or might be a relationship (not to mention weighing the effects on segregation outcomes against other choice effects on student outcomes, a huge issue). When you don’t know why a problem is happening, or why you think it is happening, it is more difficult even to discuss that problem, to say nothing about addressing it.