Revisiting the Impacts of Teachers

Chetty, Friedman, and Rockoff (2014a and 2014b) study value-added (VA) measures of teacher effectiveness. Exploiting teacher switches as a sort of quasi-experiment, the first paper investigates potential bias from student sorting to teachers, and concludes that any such bias is negligible. The second paper finds that VA scores are useful proxies for teachers’ effects on students’ long-run outcomes. I successfully reproduce both sets of analyses and results in data from North Carolina. But I find that teacher switching is correlated with changes in student preparedness, leading quasi-experimental estimates to understate bias in VA scores. Estimates that adjust for this indicate moderate bias. The long-run effects results are sensitive to controls and cannot support