The Puzzling Case of School Resources and Student Achievement

Abstract

Whether additional school spending translates to overall improved student learning remains unclear. One explanation for the mixed findings in the literature is the possibility that studies confound the effects of school resources with those of unobserved variables. We show that commonly used "value-added" models are misspecified when estimated using the National Education Longitudinal Study, which raises questions about whether previous value-added studies are unbiased. We also review a more recent literature that uses instrumental variables (IV) methods to address omitted variables bias. Most IV studies suggest that additional resources typically translate to (modest) gains in test scores and that the biases associated with value-added models are large enough to be of practical importance.