Ethnic Segregation and Educational Outcomes in Swedish Comprehensive Schools

We ask whether ethnic density in Swedish comprehensive schools affect teacher-assigned school grades in ninth grade (age 16). The data, based on two entire cohorts who graduated in 1998 and 1999 (188,000 pupils and 1,043 schools), link school information with Census data on social origin, and enable us to distinguish first- from second generation immigrants. Using multilevel analysis we find the proportion of first, but not the second, generation immigrant pupils in a school to depress grades in general, but particularly for (first generation) immigrant pupils. Passing a threshold of more than 40 percent immigrants reduces grades with around a fifth of a standard deviation, affecting fourteen percent of immigrant children. Our main results are robust to model specifications which address omitted variable bias both at individual- and school-level. One policy implication of our results is that desegregation policies which concentrated on the two per cent most segregated schools would probably improve school results and reduce ethnic inequality.