Advancing the Opportunities of Underserved Students: Lessons from Child Welfare, Education, and the Labor Market

Abstract

This dissertation studies the consequences of public policy on the opportunities of underserved students. It focuses on three policy environments: child welfare, education, and the labor market. In each chapter, I use a rigorous quantitative methodology and large administrative data to understand how public institutions are---or are not---improving the wellbeing of historically disadvantaged children and youth.

The first chapter focuses on placement into foster care, which is experienced by 6% of children in the United States between birth and age eighteen. Using administrative data from Michigan, I estimate the effects of foster care on children's outcomes by exploiting the quasi-random assignment of child welfare investigators. I find that foster care reduced the likelihood of being abused or neglected in the future by 50%, increased daily school attendance by 6%, and improved math test scores by 0.34 standard deviations. Gains in safety and academics emerged after children exited the foster system when most were reunified with their birth parents, suggesting that improvements made by their birth parents was an important mechanism. Given recent federal legislation to reduce foster placements, these findings indicate that child welfare systems must invest in more effective interventions to keep vulnerable children safe and thriving in their homes.

Chapter two, with Silvia Robles and Robert W. Fairlie, examines course availability, a frequently cited yet understudied channel through which money matters for college students. Open admissions policies, binding class size constraints, and heavy reliance on state funding may make this channel especially salient at community colleges, which enroll 47% of U.S. undergraduates in public colleges and 55% of underrepresented minority students. We use administrative course registration data from a large community college in California to test this mechanism. By exploiting discontinuities in course admissions created by waitlists, we find that students stuck on a waitlist and shut out of a course section were 25% more likely to take zero courses that term relative to a baseline of 10%. Shutouts also increased transfer rates to nearby, but potentially lower quality, two-year colleges. These results document that course availability, even through a relatively small friction, can interrupt and distort community college studentsâ educational trajectories.

The third chapter, with Brian A. Jacob and Kelly Lovett, evaluates a summer youth employment program, a popular way for municipalities to provide adolescents with skills and experiences thought to improve labor market outcomes. We study the program in Detroit, Michigan using a selection on observables identification strategy. In addition to controlling for a rich set of covariates, including baseline educational measures, we match participants to their classmates of the same race and gender who applied for the program but did not participate. We find that participation is associated with a modest increase in educational attainment. Specifically, it increased the likelihood of enrolling in public school after the program by 1.5% and of graduating high school by 4%, relative to comparison means of 94.5% and 85%. Youth with the weakest academic skills benefited the most, as participation increased school enrollment by 2.2% and high school graduation by 5.5% for this group. Falsification tests of whether participation predicts pre-program characteristics, as well as robustness checks which account for omitted variable bias, as proposed in Oster (2016), suggest that our results are not driven by unobservable differences between participants and other applicants.