Author: Justin A. Wong

Publisher: Stanford University

Published: 2011

Total Pages: 121

ISBN-13:

This dissertation is composed of three essays. Essay 1, "Does School Start Too Early For Student Learning?", considers the connection between school start time and student performance. Biological evidence indicates that adolescents' internal clocks are designed to make them fall asleep and wake up at later times than adults. This science has prompted widespread debate about delaying school start times in the U.S., a country which has some of the earliest start times worldwide. The debate suffers, however, from a glaring absence of evidence: the small number of prior studies has been too low powered statistically to test whether later start times improve achievement. I fill the gap by studying achievement across a large, nationally representative set of high schools that have varying start times. I identify the positive effect of later clock start times, as well as the independent effect of greater daylight at school start time. My primary empirical method is cross-sectional regression with rich controls for potentially confounding variables. The findings are confirmed by regression discontinuity analysis focused on schools close to time zone boundaries. I quantify the net gain in welfare from having an additional hour of sunlight before school starts by comparing the substantial lifetime earnings benefits for students against the likely the societal costs. Essay 2, "Student Success and Teaching Assistant Effectiveness In Large Classes", considers the impact teaching assistants (TAs) have on student performance. In universities, TAs play a crucial role by providing small group instruction in lecture courses with large enrollment. The multiplicity of TAs creates both positive opportunities and negative incentives. On the one hand, some TAs may excel at tasks--such as helping struggling students--at which other TAs fail. If so, all students may be able to learn better if they can match themselves to the TA that best suits their needs. On the other hand, the multiplicity of TAs means that students in the same class often receive instruction that varies in quality even though they are ultimately graded on the same standard. In this paper, we use data from a large lecture course in which students are conditionally randomly assigned to TAs. In addition to administrative data on scores and grades, we use survey data (which we generated) on students' initial preparation, their study habits, and their interactions with TAs. We identify the existence of variation among TAs in teaching effectiveness. We also identify how TAs vary in their effectiveness with certain subpopulations of students: the least and best prepared, students with different backgrounds, and so on. Using our parameter estimates, we simulate student achievement under scenarios such as random assignment to TAs, elimination/retraining of the least effective TAs, and matching of TAs to students based on initial information to show the potential gains in student welfare from more efficient matching. Essay 3, "A Study of Student Majors: A Historical Perspective", considers whether differing financial returns across degrees are a significant factor in a student's choice of a major. During the late 1990s, the U.S. experienced a technology boom that significantly increased the initial salary offers to engineering students, and computer science students in particular. These dramatic increases in returns provide an excellent opportunity to examine not only how students respond to salary levels, but also to salary trends. The existing literature has focused on the extent to which differing financial returns can affect a student's choice of undergraduate major. This paper extends the analysis to test if trends in salary levels also affect the share of students selecting into various majors using a comprehensive dataset of all post-secondary institutions. I find that students select into majors that offer higher salaries and have greater wage growth. Using a flexible empirical model that allows students to respond to both changes in salary levels and growth, I find that the results hold across majors and within engineering disciplines. These results help to explain why, for instance, the percentage of students choosing to major in computer science grew more rapidly than could be explained by salary level alone.