Ford School Public Events

We examine the theoretical and practical implications of ranking teachers according to a one dimensional value-added metric when teacher effectiveness is multi-dimensional. In particular, we consider the cases in which teachers teach multiple subjects or multiple student types. We outline the assumptions under which a standard value-added estimator correctly ranks teachers according to their social value. We demonstrate that these assumptions fail to hold empirically. This causes value-added based pairwise rankings of teachers to be often misleading, though the consequences of these ranking errors for students is small. We demonstrate that when teachers vary in ability across student types or subjects, student outcomes can be improved by matching teachers to students or subjects according to their comparative advantage. Our calibration suggests that these gains exceed those associated with firing the bottom 10 percent of teachers.