| Summary: | This dissertation consists of essays on three inputs into the educational production function: curriculum, peers, and teachers. The chapters are linked by their focus on understanding the importance of these inputs for student achievement and by their exploitation of the exact timing of events (i.e., student mobility, receipt of special education services, teacher absences) to identify causal effects. The first chapter examines how decentralized decision-making about schools' curricula affects student achievement. Decentralized decision-making in this context involves a tradeoff, since individual schools might be better able to target the needs of their student populations, but their choices might not be aligned with the goals of the larger community. Differences in curricula may also harm the achievement of mobile students, who have to adjust to new curricula when they change schools. My analysis is based on quasi-experimental variation from New York City, which standardized math and reading curricula starting in Fall 2003. Schools received exemptions from the standardized curricula based on test score cutoffs, allowing me to use a regression discontinuity design to evaluate the causal effect of the policy change. I find that curriculum standardization has no significant effect on student achievement, either overall or for mobile students. I also assess the extent to which curriculum standardization could benefit mobile students using a novel test that exploits the timing of student mobility relative to the achievement exam: I compare the achievement losses of students who could have been affected by having to change curricula — those who changed schools before the exam — to those of students who may have similar personal circumstances but whose achievement could not have been affected by changing curricula — those who changed schools after the exam. The achievement losses of these groups of students are similar, suggesting that omitted variables associated with student mobility may explain the entirety of the mobility achievement gap and that curriculum standardization may have little scope to improve the achievement of mobile students. The second chapter examines the effects of disabled peers on student achievement and behavioral outcomes. Although disabled students might be expected to have negative effects on their peers, this expectation has not been borne out in studies that focus on the peer effects of special education students. This chapter argues that this is likely because special education ameliorates negative peer effects. To illustrate this argument, I exploit the longitudinal nature of the data to classify students into categories based on when they received special education services. I classify students who are ever observed in special education as "disabled" and those who currently receive special education services as "special education" students. Students who enter special education in the future are classified as having "undiagnosed disabilities," while those who exit special education have "declassified disabilities." I then use school-grade variation in the proportions of disabled peers to estimate their effects on non-disabled students. Consistent with special education mitigating peer effects, I find that students with undiagnosed disabilities have significant negative effects on their non-disabled peers, but students who are receiving special education services do not. Students who have been declassified also have no effects on their peers, probably because they are positively selected among disabled students. I also find that segregation is the likely mechanism by which special education mitigates negative peer effects. The third chapter, which is joint work with Jonah Rockoff and is forthcoming in the Journal of Labor Economics, examines the effect of teacher absence on student achievement. To address potential sources of bias from the endogeneity of teacher absence, we use detailed data on the timing, duration, and cause of absences. Our main specification uses within teacher variation, and in support of a causal interpretation, we show that teacher absences before the exam have a significant negative effect on achievement while absences after the exam do not. Our estimates suggest that the daily loss associated with one absence is 0.001 standard deviations in math and 0.0006 standard deviations in reading, the same as replacing an average teacher with one at the 10-20th percentile of teacher value-added. We also find evidence that the daily losses associated with an absence decline with the length of the absence spell, consistent with long-term substitutes being of higher quality or learning on the job.
|
|---|
