In 1993, the policy research firm MDRC launched a 15-year longitudinal, random-assignment evaluation of career academies , a secondary-education approach in which schools structure curricula and student opportunities around career themes. MDRC's analyses revealed important effects of the academies on labor-market outcomes eight years after high school, especially for males. A puzzle emerged, however, as they found no similar effects on academic achievement, high-school completion or college attainment, the mechanisms through which education is hypothesized to influence life outcomes. Through this research, I aim to explore the mechanisms through which the career academies impacted the subsequent lives of the adolescents involved. Chapter 1 investigates the causal impact of academy participation. MDRC's analyses focused on returns to the randomized invitation to enroll in an academy. Nevertheless, considering participation is important: nearly half of students offered the opportunity either never enrolled or participated for only a portion of high school. Utilizing a principal-stratification framework, I investigate the impacts of academy participation while maintaining the value of the experiment for drawing causal inferences. I estimate an average causal effect of treatment assignment on subsequent average monthly earnings of approximately $588 among males who would remain enrolled in an academy throughout high school, given the opportunity.

Provoked by this finding, Chapter 2 explores methodological approaches for understanding the causal impact of career-academy participation, as defined not by years of enrollment but by the self-selected take-up of salient programmatic components. Quantitative methods for modeling such mediational processes are an active area of methodological research. I discuss three approaches proposed for investigating mediational processes in an experimental context. These are: (a) multiple regression, (b) instrumental variables, and (c) principal stratification. The first two are limited by their reliance on assumptions potentially too strong to be plausible in practical settings. I discuss these limitations and present principal stratification as a framework with which researchers may overcome them. I apply each approach to the career-academies data and generate preliminary evidence to support the hypothesis that increased exposure to the world-of-work through the career-academy treatment subsequently led to improvements in average monthly earnings in the years following high school.