<em>Non-cognitive Ability, College Learning, and Student Retention</em>

Recent decades have seen a steady increase in college enrollment rates, which has not been ac- companied by a corresponding increase in graduation rates. If this discrepancy is at least partly due to insufficient "effort" exerted by students, policies that aim at rewarding effort explicitly may succeed at increasing graduation rates. This paper uses a unique and rich administrative data set to analyze the impact of the introduction of a new grading policy on performance and retention rates at Benedict College, a Historically Black College in Columbia, South Carolina. According to the new grading policy, grades for freshman and sophomore courses are determined in part by performance on tests and in part by measures of "effort" such as attendance and class participation. The policy was intended to inspire a sense of discipline in students' attitudes to- wards academic work, in the hope of improving learning and graduation rates. However, the data show that the introduction of the new grading policy was actually followed by an increase in dropout rates, disproportionately due to students with pre-college GPAs above the mean. This paper describes a simple theoretical model that illustrates how this observed change can be a result of the interaction between the increased disutility caused by the larger effort required by the new policy, and the uncertainty that relates effort to knowledge grades, an uncertainty that only disappears after the final grades are assigned. Overall, student SAT composite scores, ACT composite scores and high school GPAs are not great predictors of survival after the policy