1 Estimating Treatment Effects with Multiple Proxies of Academic Aptitude

Existing studies of college student learning gains due to pedagogical practices typically rely on a wide variety of proxy variables for academic aptitude. This study questions the wisdom of this approach because each proxy, or particular combinations of proxies, measures latent ability with distinctly different substantial errors. We consider four econometric approaches to identify the best proxies for latent academic aptitude: principle components, factor analysis, post hoc estimator, and instrumental variables. Our estimates suggest that collegiate GPA best controls for the students ’ ability to learn economics and the existing literature overestimates the treatment effect of faculty pedagogical practices