Are U.S. Schools Resegregating?

Last week, the U.S. Government Accountability Office (GAO) issued a report, part of which presented an analysis of access to educational opportunities among the nation’s increasingly low income and minority public school student population. The results, most generally, suggest that the proportion of the nation's schools with high percentages of lower income (i.e., subsidized lunch eligible) and Black and Hispanic students increased between 2000 and 2013.

The GAO also reports that these schools, compared to those serving fewer lower income and minority students, tend to offer fewer math, science, and college prep courses, and also to suspend, expel, and hold back ninth graders at higher rates.

These are, of course, important and useful findings. Yet the vast majority of the news coverage of the report focused on the interpretation of these results as showing that U.S. schools are “resegregating.” That is, the news stories portrayed the finding that a larger proportion of schools serve more than 75 percent Black and Hispanic students as evidence that schools became increasingly segregated between the 2000-01 and 2013-14 school years. This is an incomplete, somewhat misleading interpretation of the GAO findings. In order to understand why, it is helpful to discuss briefly how segregation is measured.

First, however, let us be clear that segregation is about as serious an issue as there is in education today, and the question of whether or not segregation is increasing in U.S schools is quite complicated (see Reardon and Owens 2014). As mentioned below, it depends on the level one is looking at (e.g., district, metropolitan area, state, etc.), the race and ethnicity groups one examines, and the measures one uses. The point of this post is less about what’s “actually happening” than it is the interpretation of this particular report (as well as similar reports and analyses). Moreover, we should always bear in mind that, no matter how you measure it, U.S. public schools by almost any measure are heavily segregated by race and ethnicity, as well as by income, and so the discussion of whether such segregation increased during the relatively short period of 2000 to 2013 should never distract from this fundamental point.

Moving on, consider the core finding that the proportion of U.S. schools serving predominantly (i.e., more than 75 percent) Black and Hispanic students increased from 9 to 16 percent between 2000 and 2013. This obviously shows that there are many more schools serving minority student populations above this threshold, and more students attending such schools, but does it really represent “resegregation?” Not necessarily.

(Side note: Most of the discussion below also applies directly to the findings about low income students, though race- and ethnicity-based segregation bears a strong relationship with income-based segregation, and income-based school segregation is increasing, particularly between districts [see, for example, Owens at al. 2016].)

For one thing, the percentage of schools with predominantly minority student populations is a highly incomplete measure of overall segregation, particularly at the national level, even in any given year. By asserting a somewhat arbitrary cutpoint (in this case, over 75 percent), it essentially “ignores” the situation in the majority of schools - i.e., those with smaller minority populations. If, for example, these schools are diverse, U.S. public schools might actually be quite integrated, even if the nation contains a seemingly large proportion of heavily minority schools. This is generally not the case in the U.S., but it bears noting.

The biggest problem, however, is using this measure to gauge segregation trends. The increase in the proportion of heavily minority schools may be driven largely by the increase in minority representation in the student population during this time period. In other words, put simply, there may be more predominantly minority schools because there are more minority students. And this might be the case even if those “new” minority students are distributed across schools and districts in a manner reflecting a similar extent of segregation as was exhibited in the baseline year of 2000.

Again, this is not to say that changes in the student population can explain away the increase in heavily minority schools. The degree to which this is the case almost certainly varies widely by location. The big points here are: 1) such changes in the student population did occur; 2) this almost certainly played a significant role in increasing the number of heavily minority schools nationally; and, most importantly, 3) this particular measure (the percentage of heavily minority schools) cannot address the issue either way.

Underlying this whole discussion are long standing debates about how to measure segregation, both of schools and other types (e.g., residential). The most common types of measures are unevenness and isolation/exposure. The percentage of schools serving heavily minority student populations is an example of the latter, which means, as mentioned above, it is sensitive to changes in the composition of the student population. As a result, while there is no single “correct” measure, unevenness measures are arguably more appropriate when examining trends, since these measures are independent of compositional change. At the very least, any analysis of segregation trends should present both types of measures, as they often lead to different conclusions (for more discussion, see: Owens at al. 2016; Orfield et al. 2014; Massey and Denton 1988; also see our recent report on teacher segregation).

Consider also how this GAO measure provides little insight into the crucial distinction between different levels of analysis (e.g., states, metropolitan areas, districts, etc.). For instance, schools can actually become less segregated within districts even while segregation persists or even increases between districts (the converse is also theoretically true). Since the landmark Brown decision in 1954, most of the progress in racial desegregation occurred within districts, particularly during the late 1960s and early 1970s (Coleman et al. 1975), whereas segregation between districts has been far more persistent (in large part because it is tied to residential segregation, which, due in no small part to deliberate policy, has remained strong). 

As a result of these issues of choice of level and measurement, it is very difficult to characterize segregation trends with broad strokes, as results can differ depending on the level of analysis and the type of measure used (not to mention whether we're talking about segregation by race and ethnicity or by income).

So, again, the increase in the proportion of these heavily minority (and lower income) schools, which the GAO reports, does provide useful information, particularly given that these schools tend to offer fewer services (e.g., college prep courses) and suspend/expel more students than schools serving lower proportions of minority students. But it is, at best, a highly incomplete measure of segregation trends, and by itself is arguably insufficient for claims such as “resegregation.”

This may be why the GAO report itself does not really seem to portray its results on the trends in the percentage of heavily minority (or low income) as measures of segregation per se. The results are instead presented more as evidence of the impact of segregation, as these heavily poor and minority schools offer fewer services, suspend/expel more students, etc. The fact that the results were overintepreted as evidence of resegregation is certainly understandable, but it ignores the complications entailed in measuring a very complicated, important phenomenon.