Abstract: We propose a nonparametric approach for studying the impacts of teachers, built around the distribution of unobserved teacher value-added. Rather than assuming this distribution is normal (as standard), we show it is nonparametrically identified and can be feasibly estimated. The distribution is central to a new nonparametric estimator for individual teacher value-added that we present, and allows us to compute new metrics for assessing teacher-related policies. Simulations indicate our nonparametric approach performs very well, even in moderately-sized samples. We also show applying our approach in practice can make a significant difference to teacher-relevant policy calculations, compared with widely-used parametric estimates.
Keywords: Teacher Impacts, Teacher Value-Added, Value-Added Distribution, Nonparametric Estimation, Empirical Bayes, Education Policy, Teacher Release Policy, False Discovery Rate
JEL Classification: C11; H75; I21; J24