How robust is the evidence on the effects of college quality? Evidence from matching

We estimate the effects of college quality using propensity score matching methods and the National Longitudinal Survey of Youth 1979 cohort. Matching allows us to relax the linear functional form assumption implicit in regress ion-based estimates. We also examine the support problem by determining whether there are individuals attending low-quality colleges similar to those attending high-quality colleges, and find that the support condition holds only weakly. Thus, the linear functional form plays an important role in regression-based estimates (and matching estimates have large standard errors). Point estimates from regression and matching are similar for men but not women. (C) 2003 Elsevier B.V. All rights reserved.