Policy Issues in the Economics of Education: Lessons from Michigan

Abstract

In this dissertation, I apply administrative data from Michigan's public schools to address crucial policy questions in the economics of education. In Chapter I, I shed light on the persistent effects of attending high-quality high schools by creating a value-added model that isolates each high school's effect on students' test scores, then matching the results to students' college transcripts to determine the relationship between high-school value added and first-year college grades. I find that students who attend high schools with one standard deviation higher value added receive first-year grades about 0.09 grade points higher than their otherwise-identical counterparts. These gains in college are not driven solely by math and English, but are evenly distributed across subjects. This result is robust to adjustments for a number of potential biases that arise throughout the process, including selection into high schools and selection into college attendance. Overall, I find evidence against some of the more skeptical interpretations of test-score improvement, such as the claim that schools "teach to the test" or the concern that the content tested on standardized exams is not relevant to future learning. Human capital theory suggests that when students would graduate into a weak labor market, the opportunity cost of schooling declines, and they should instead invest in themselves and get more education. However, this assumes that they have no borrowing constraints; if students are credit-constrained and their families are hurt by the struggling labor market, then their educational options may actually diminish as they are less able to pay for college.

In Chapter II, I determine which of these effects predominates empirically using data on plant closings and mass layoffs from the Worker Adjustment and Retraining Notification ("WARN") Act, examining the impact of exposure to job losses during the senior year of high school on whether and where students attend college. A 1-standard deviation increase in per-capita job losses is associated with a small but statistically-significant 0.2-percentage point increase in the probability of attending college, driven entirely by attendance at community colleges. This result supports the argument that the opportunity cost effect dominates, as any movement out of college as a result of credit constraints and firsthand exposure to job losses is comparatively small.

Having access to an effective and experienced teacher can make a crucial difference in a student's academic achievement. In Chapter III, written with Kolby Gadd, I examine the factors that predict whether teachers will stay in their first jobs or leave for opportunities elsewhere, and then study how students perform after teachers leave, looking both at teacher turnover in general and at teacher departures for particular destinations such as new districts or the private sector. In a multinomial logit framework in which we examine each teacher's employment status in his or her fifth year, we find that the characteristics that predict departures most consistently are the fraction of Black students in the teacher's first school, the fraction of economically-disadvantaged students in the teacher's first school, the teacher's first job being in special education, and the teacher's first school being a charter. Turning to how student achievement changes in the wake of teacher turnover, we find a modest decline in test scores after teacher exits, driven by students in schools that lost teachers to other full-time teaching positions, both within and across districts.