Abstract

We estimate the effect of grant aid on poor students' college graduation and earnings using student-level administrative data from four-year public colleges in Texas. To identify these effects, we exploit a discontinuity in grant generosity as a function of family income. While eligibility for additional grant aid has small contemporaneous effects on attainment, it significantly increases four-, five-, and six-year graduation rates. Corresponding to the increases in degree receipt, eligibility also generates persistent earnings gains beginning four years after entry. We project that within ten years, the additional federal income tax revenue generated from eligible students' earnings gains would be sufficient for the government to fully recoup the cost of the additional grant aid expenditures. We develop a theoretical model and a novel empirical test for treatment effects on subgroups defined by their endogenous responses to treatment that can be used in instrumental variables applications with monotonicity in the second stage. Our test rejects the traditional credit constraints model. Our theoretical framework also produces sufficient statistics for assessing the welfare implications of changes in grant generosity. While increases in grant generosity would be welfare improving in the setting we examine, our framework can be applied in other settings where welfare implications are less clear-cut.