Careful! Metrics matter a great deal when estimating racial segregation in schools

The racial segregation of American schools has historically been a vivid symptom of racist attitudes, institutions, and laws. Segregation remains a highly salient issue, especially given the wide, stubborn gaps in educational outcomes by race.

The extent of segregation, its impact on achievement gaps, and the position of charter schools in particular is the subject of our new paper with Russ Whitehurst, Segregation, race, and charter schools: What do we know?

Segregation metrics produce different pictures

One of our main messages is the need to be very careful about how we measure segregation. Scholars use different indices, define their groups in different ways, and choose different geographical areas as the basis for assessing schools. One example of why this last criteria matters: in the wake of civil rights legislation, the extent of racial segregation within school districts declined, but segregation between school districts increased slightly over the same general period, partly because of “white flight” to suburban school districts.

The selection of specific measures matters a lot, too, especially given demographic changes. One commonly-used segregation metric, for example, is the proportion of black students in majority-white schools (i.e., at least 50 percent white). After rising rapidly after civil rights legislation, this “exposure” measure started falling again beginning in the late 1980s. The share of black students in majority-white schools (light blue, left axis) actually fell from 44 percent in 1988 to 23 percent in 2011:

An immediate interpretation of this chart might be: “there was some dramatic desegregation after civil rights legislation, but since 1988, the nation’s schools have started to resegregate.” Indeed, some observers have drawn this conclusion.

In some places, resegregation of black students has taken place. But a big reason for the decline is the fall in the proportion of majority-white schools (dark blue, right axis), from 81 percent in 1988 to 58 percent in 2013. This is in large part due to the influx of Hispanic students. There are, then, simply many fewer majority-white schools for black students to attend.

Sharpen the question, then select the measure

Given the diversification of U.S. schools, researchers need to be clear exactly what kind of segregation they are seeking to identify, and to then select the most appropriate measure. In order to get as complete a picture as possible, it make sense to calculate multiple metrics, as John Logan, Weiwei Zhang, and Deirdre Oakley do in a forthcoming paper:

Here, various segregation measures can be seen alongside each other. The picture that emerges is of less white isolation, less black isolation, more exposure of black and white students to Hispanic students, but no increase in the exposure of black and white students to each other. Both black and white students have then become much more likely to share classrooms with Hispanics, but not more likely to share classrooms with each other.

The separation of black and white students means that half of black students, or half of white students, in the average black student’s school district would have to move schools in order for the schools to match the demographics of the larger district.

There are many other nuances in measurement too, including the selection of geographical area, the definitions of racial categories, and the time period considered. Each has a significant impact on results. As the philosopher Bernard Williams warned us, we must always be careful not to “smuggle our answer into our question.”

Given that segregation remains, rightly, a hot topic, it is doubly important to be clear about the rationale for the approach taken to measurement. Policymakers, in particular, need to be careful in their use and interpretation of empirical measures of segregation.