Evaluating the effect of training on wages in the presence of noncompliance

In the paper, the effects of the Job Corps training program on employment and wages are evaluated, using data from a randomized study, the National Job Corps Study, and the principal stratification approach to simultaneously address the issues of noncompliance, truncation of wages by nonemployment, and missing outcomes. We conduct a likelihood-based analysis, proposing different ways of improving computational efficiency and identifiability using the theory of finite mixture models and exploiting the EM algorithm. We maintain an IV exclusion restriction assumption on the outcomes, while monotonicity of compliance holds by design. We provide estimates with and without assuming monotonicity of the truncation of wages, under the Missing at Random assumption of outcome missingness. For compliers, results show that the effect on employment is generally positive; however, there is a subgroup of participants for whom the program is detrimental on employment in the short term, although this subgroup becomes negligible in the long term. For compliers who would be employed whether trained or not, the effect of the training on wages is positive, both in the short and long term. According to our estimates, participation in the Job Corps training program is found to be effective: 208 weeks after participation, we found a causal effect on employment of about 5%, with an average effect on wages for the always employed of about 0.3 $/hour.