Recent evidence has established that non-cognitive skills (e.g., persistence and selfcontrol) are valuable in the labor market and are malleable throughout adolescence. Some recent high school interventions have been developed to foster these skills, but there is little evidence on whether they are effective. Using administrative data, we apply two methods to evaluate an intervention called OneGoal, which attempts to help disadvantaged students attend and complete college in part by teaching non-cognitive skills. First, we compare the outcomes of participants and non-participants with similar pre-program cognitive and non-cognitive skills. In doing so, we develop and validate a measure of non-cognitive skill that is based on readily available data and rivals standard measures of cognitive skill in predicting educational attainment. Second, we use an instrumental variable difference-in-difference approach that exploits the fact that OneGoal was introduced into different schools at different times. We estimate that OneGoal improves academic indicators, increases college enrollment by 10–20 percentage points, and reduces arrest rates by 5 percentage points for males. We demonstrate that improvements in non-cognitive skill account for 15–30 percent of the treatment effects.