with A. Campos, Y. Cruz-Aguayo, C. Echeverri and N. Schady
Abstract:
We use an experiment with five consecutive rounds of random assignment and almost perfect compliance to estimate how the quality of teacher-child interactions affects child outcomes throughout elementary school. We estimate linear and CES production functions for different skills, where inputs are the quality of the teacher-child interactions in each grade. We find modest impacts of teacher behaviors on math and language achievement, and executive function, which similar across grades. These impacts fade-out over time, except for kindergarten teachers, for whom impacts increase over time. The quality of teacher-child interactions is highly substitutable across grades. Finally, we show that children do not exert more effort at school or read more at home in response to differences in teacher quality, and teachers do not exhibit different behaviors when, by chance, they are randomly assigned to children of varying characteristics.