We use a randomized experiment to evaluate a large-scale active labor market policy: Turkey's vocational training programs for the unemployed. A detailed follow-up survey of a large sample with low attrition enables precise estimation of treatment impacts and their heterogeneity. The average...