The majority of empirical papers in the literature on school quality finds no or only small effects of class size and other school quality measures on students' outcomes. This paper analyses the effect of achievement heterogeneity and therefore the effect of the composition rather than the pure size of the class on student achievement. In this endeavor, individuallevel data from an internationally conducted standardized test, the PISA 2000 study is utilized. For the case of US schools the influence of a student's peer group is estimated in a pure endogenous effects model and a model also allowing for contextual effects. The potential endogeneity of peer group formation is addressed in an instrumental variable approach. It turns out that heterogeneous peer groups have a strong detrimental impact on individual achievement. Moreover, it becomes transparent that contextual variables are important for the extent of peer group effects and the endogeneity of peer group formation.