Students are non-randomly assigned, or sorted, into classrooms in various ways across and within schools. In this study, we use longitudinal data sets from three districts to investigate a metric for the characterization of sorting at the school level. We analyze whether non-random student assignment is associated with value-added estimates for teachers. The three longitudinal datasets come from large, urban districts but, despite this similarity, we find there is substantial variation in the degree of sorting across school districts. We see more evidence of sorting in districts with higher proportions of ELLs and students from low socioeconomic backgrounds. Despite differences in the characteristics of schools which sort, sorting as we have quantified it is only marginally related to VA estimates.