A Comparison of Student Growth Percentile and Value-Added Models of Teacher Performance

Abstract

School districts and state departments of education frequently must choose among a variety of methods to estimate teacher quality. This article investigates the consequences of some of these choices. We examine estimates derived from student growth percentile and commonly used value-added models. Using simulated data, we examine how well the estimators can rank teachers and avoid misclassification errors under a variety of assignment scenarios of teachers to students. We find that growth percentile measures perform worse than value-added measures that control for prior year student test scores and include teacher fixed effects when assignment of students to teachers is nonrandom. In addition, using actual data from a large diverse anonymous state, we find evidence that growth percentile measures are less correlated with value-added measures that include teacher fixed effects when there is evidence of nonrandom grouping of students in schools. This evidence suggests that the choice between estimators is most consequential under nonrandom assignment of teachers to students and that value-added measures controlling for teacher fixed effects may be better suited to estimating teacher quality in this case.