Youths who share a school and neighborhood often have similar academic achievement, but some studies find all or most of this apparent effect is due to sorting, not to the neighborhood itself. We present a collage of evidence from the National Educational Longitudinal Survey (NELS) indicating that a significant fraction of the apparent correlation is causal, rather than solely due to sorting. We first show that the importance of school effects is robust to very rich measures of family background. We then use the fact that the characteristics of the high school that students will attend are an additional indicator of family background. This measure can be used as an instrument to identify family background separately from neighborhood and junior high school effects. Even after this correction, the point estimate of school effects on student achievement remains large and statistically significant. Finally, we use regression and seminonparametric matching methods to show that the test scores of youths who change schools begin to converge with those of their new classmates.