Borrowing Constraints and the Returns to Schooling

To a large degree, the expansion of student aid programs to potential college students over the past 25 years in the United States has been based on the presumption that borrowing constraints present an obstacle to obtaining a college education. Economists and sociologists studying schooling choices have found empirical support for college subsidies in the well-documented, large positive correlation between family income and schooling attainment. This correlation has been widely interpreted as evidence of credit constraints. Recently, however, Cameron and Heckman (1998, 2000), Keane and Wolpin (1999), and Shea (1999) have examined more closely family income's in uence on schooling and question whether borrowing constraints plays any role on college choices. Over the last 20 years, a separate literature in economics has aimed at estimating the return to schooling purged of \ability bias." Unobserved ability upwardly biases leastsquares estimates of returns to schooling. However, using instrumental variables methods to correct for the bias , researchers have employed a variety of di erent instruments and have generally found that instrumental variables produces estimated schooling returns that are larger, not smaller, than least-squares estimates. The connection between access to credit and measured returns to schooling{a connection rst noted by Becker (1972){has been explored recently by Lang (1993) and Card (1995a,2000), who argue that \discount rate bias" can explain the anomolously high instrumental variables estimates. This argument implicitely suggests that borrowing constraints are important for schooling decisions. Our paper attempts to integrate and reconcile these two literatures. Building on the seminal work of Willis and Rosen (1979), we develop a framework that allows us to study schooling determinants and returns together. Identi cation of the e ect of borrowing constraints from the fact that foregone earnings{the indirect costs of school{and the the direct costs of schooling each a ect borrowing constrained persons di erently from unconstrained individuals. We apply this idea in a number of ways using least-squares, instrumental variables regression, and a structural economic model to measure the extent of borrowing constraints on schooling choices. None of these methods produces evidence of borrowing constraints.