Black-White Test Scores: Neighborhoods, Not Schools, Matter Most

In statistical models that include both school and neighborhood segregation, the effects of relative exposure of black and Hispanic students to their white schoolmates are uniformly small and statistically insignificant, the neighborhood composition matters more than school composition.

The large gap in student achievement, particularly between blacks and whites, has long troubled Americans. Fifty years after the Brown v. Board of Education decision, persistently large black-white differences in standardized test scores remain central to education policy.

In Racial Segregation and the Black-White Test Score Gap (NBER Working Paper No. 12078), NBER researchers David Card and Jesse Rothstein cast some fresh and perhaps surprising light on this issue. Using data from SAT records for roughly one third of test takers in the 1998-2001 high school graduation classes, they find that the black-white achievement gap is clearly linked to racial segregation.

To reach this conclusion, the authors match test-takers to information on the racial composition of their high schools and to an extensive set of family background characteristics of black and white students in their metropolitan areas. They compare the black-white achievement gap across areas with more- and less-segregated neighborhoods and schools. Within a metropolitan area, families living in integrated neighborhoods (and students attending integrated schools) may be different in a variety of unobserved ways from those in segregated neighborhoods and schools, confounding the effect of inter-racial exposure. The focus on across-area differences in segregation eliminates biases deriving from this sort of within-city sorting. Similarly, the focus on metropolitan-level black-white test score gaps removes the impact of a variety of omitted characteristics -- potentially including school quality and resource levels -- that do not vary within a city but might be correlated with inter-racial contact.

The results indicate that segregation has large, negative effects on black students' relative test scores. When a city is completely integrated, the gap in relative SAT scores between blacks and whites proves to be one quarter smaller (about 45 points) than in a city with the races fully segregated in different neighborhoods, holding family background characteristics constant.

The authors also attempt to distinguish between the effects of residential and school segregation. Considered separately, each appears to have a negative effect on the relative test scores and educational attainment of blacks students. In statistical models that include both school and neighborhood segregation, though, the effects of relative exposure of black and Hispanic students to their white schoolmates are "uniformly small and statistically insignificant." Although the authors acknowledge that the data could be consistent with equally negative effects of neighborhood and of school segregation, they write that, "Our tentative conclusion is that the neighborhood composition matters more than school composition."

These results -- both the negative effects of segregation, and the indication that neighborhood segregation matters more than does school segregation -- stand up in the face of a variety of statistical tests designed to rule out competing explanations. The segregation effects do not appear to be attributable to differential family background characteristics of black students living in more- and less-segregated cities, nor to resource differences between students' schools.

One potential explanation for the apparent lack of a school segregation effect is the prevalence of within-school segregation: if black students rarely attend class with white students even in cities with integrated schools, these cities may not post higher black test scores even though truly integrated education would have a positive effect. Indeed, the authors find a strong relationship between school integration and at least one proxy for classroom-level exposure: white students are more likely to take honors and advanced placement classes, which typically have few black students, in cities where the schools are integrated than in cities where schools are segregated. Although the authors have no way of measuring direct interactions between students of different races at school, this result suggests that school integration may not achieve high exposure rates of black to white students, potentially accounting for the lack of an integration effect on black students' test scores

-- David R. Francis