Teacher value-added in Chile

Abstract

This paper estimates teacher value-added measures of teacher quality in Chile. Using administrative data we link individual student test score results to teacher assignment for 6th and 8th grade students, and control for a rich vector of covariates for the value-added estimation, including previous score and tuition fees. We evaluate the degree of accuracy of our teacher value-added estimates for predicting teachers’ impacts on student achievement, by means of a teacher switching quasi-experiment devised by Chetty et al. (2014a), and find no significant bias in these estimates when controlling for our full set of variables. We evaluate next which controls are most important for the unbiasedness of our estimates, and find that previous score, and as a novelty in the literature, tuition fees, are essential for these purposes. We study the sorting on teacher value-added and report positive sorting between socioeconomic measures and teacher value-added estimates, meaning that better off students and schools get the highest performing teachers.