摘要：This paper studies how the effectiveness of teachers varies by classroom composition, combining random assignment of teachers with rich measures of teaching practices based on a popular teacher-evaluation protocol. We find that some teaching practices are more effective in raising math achievement in classrooms with higher average prior achievement, and others are more effective in classrooms with less heterogeneity in prior achievement. We use these estimates to simulate the effects of reallocating classrooms among teachers within schools. We find substantial differences between counterfactual and actual teacher effectiveness rankings, supporting the importance of classroom composition for evaluating teachers and prescribing practice.
