Bounds on Average and Quantile Treatment Effects of Job Corps Training on Wages

Abstract

We review and extend nonparametric partial identification results for average and quantile treatment effects in the presence of sample selection. These methods are applied to assessing the wage effects of Job Corps, United States' largest job-training program targeting disadvantaged youth. Excluding Hispanics, our estimates suggest positive program effects on wages both at the mean and throughout the wage distribution. Across the demographic groups analyzed, the statistically significant estimated average and quantile treatment effects are bounded between 4.6 and 12 percent, and 2.7 and 14 percent, respectively. We also document that the program's wage effects vary across quantiles and demographic groups.