Evaluating the Evidence from the Literature on the Returns to College Quality

Dan Black, Syracuse University
Jeffrey Smith, University of Maryland

This paper makes three contributions to the literature on the earnings effects of college quality. First, we present evidence on the returns to college quality for men from the NLSY. Our evidence assumes that the rich data in the NLSY suffice to control for the non-random selection of students. Second, we show that studies using only a single variable, such as mean test scores to measure quality understate its effects. Such studies ignore the fact single measures represent error-ridden proxies for the underlying quality. Third, we examine the support problem. If high quality universities have very few low quality students, then the earnings effects in studies that use linear models depend heavily on the linear functional form restriction. We find that the support problem is important but not over-whelming since there are some, but not many, low ability students at good universities and high ability students at low quality universities.