Abstract:
We study how group composition in a training program that focuses on low-skilled adult women in Chile may impact its effectiveness. We set up an experiment within an existing training program that allows us to investigate whether separating the target population in more homogeneous groups is more efficient than running the program using the traditional allocation process of highly heterogeneous participants. We find that the program only had significant impact on self-perceived benefits and not on objective measures of labor market performance. We further find some evidence that targeting the program did not necessarily improve the benefits provided by it. We finally document significant impact of (randomly-generated) group composition, in particular in the number of classmates interested in self-employment versus formal employment.