2015 Theses Doctoral
Three Essays on the Economics of Education
This dissertation consists of essays studying the impacts of education policies on outcomes measured at three distinct points in the high school to labor force continuum: course taking and academic performance in high school, choice of college and major, and labor market returns to completing college. The chapters are linked by their focus on understanding how these policies affect disadvantaged and under-represented populations, and by their exploitation of exogenous variation in the timing and assignment of treatments to identify causal effects.
The first chapter asks whether lack of information about ability helps explain why high-performing students from disadvantaged backgrounds tend to under-invest in their education. In the presence of uncertainty, an information shock may lead individuals to revise their beliefs and decision-making. To explore this question, I examine an individualized signal of academic aptitude known as "AP Potential'' that is provided in Preliminary SAT (PSAT) reports. The signal provides information about students' aptitude for Advanced Placement (AP), a national program that offers college-level courses and exams in high school. In the United States, participation in AP has become a key step on the path to admission into selective four-year colleges.
I begin by collecting high-frequency panel data on subjective beliefs from students in Oakland, California. Students stated their expected performance on the PSAT, beliefs about their abilities, and expectations about future academic outcomes before and after receiving their PSAT results reports. This survey data allows me to identify the information shock students experienced from the PSAT. I establish that although the PSAT is, on average, a negative information shock, the AP Potential signal itself contains valuable information: students with the same PSAT score and prior beliefs about own ability who receive the AP Potential signal experience a more positive information shock. The information shock in turn leads students to revise their beliefs about their ability, the number of AP classes they plan to take, and the likelihood that they will attend a four-year college, consistent with a Bayesian updating framework.
I focus next on estimating whether the AP Potential signal has a causal effect on the probability of participating in AP and the number of AP classes in which students actually enroll by exploiting the deterministic relationship between PSAT scores and the AP Potential signal in a Regression Discontinuity (RD) design. Both graphical and more formal non-parametric and parametric methods robustly demonstrate that surveyed students on the margin of receiving the signal enroll in approximately one more AP course their junior year, increasing the probability of participation in the AP program by at least 26 percentage points. Given the demographics and performance levels of students at the margin, this effect amounted to increasing the number of high-ability, under-represented high school students taking college-level courses in Oakland. In addition, mismatch between course enrollments and student ability decreased.
When I extend this analysis to students in other schools who did not take the survey, I find that the AP Potential signal had no effect on their course enrollment decisions. This finding is equally important, as it indicates that only students who received an explanation of their PSAT results, the AP Potential signal, and ways to use the information exhibited a behavioral response to the signal. The AP Potential message is not especially conspicuous on PSAT reports, so students who were surveyed likely received an intensified treatment. The results suggest that providing a credible, individualized signal of ability is a cost-effective means of increasing human capital investments among disadvantaged students.
The second chapter examines how men and women respond to changes in the competitiveness of university admissions. Experimental research has shown that women respond to competition differently than men, which could help explain gender gaps in math performance and selection into science, technology, engineering, and mathematics (STEM) programs. A growing body of work has found suggestive evidence that these relationships also exist in practical educational settings. However, exogenous variation in competition has been restricted to the experimental literature, leaving the differential causal effects of competitive admissions an open question.
An affirmative action policy enacted in Venezuela in 2002 provides a unique opportunity to explore how men and women's academic performance and college application decisions respond to changes in competition. The policy led to exogenous shocks in the competitiveness of the university admissions process, effectively increasing competition for socioeconomically advantaged students and decreasing competition for disadvantaged students. Students who neither belonged to the advantaged nor disadvantaged groups defined by the policy were unaffected and thus served as a control group.
I use a triple-difference approach on the universe of college applicants between 1994 and 2007 to estimate the impact of changes in the level of competition on high school GPA, math and verbal test scores, and the selectivity of college applications by gender. My results suggest that men and women respond differently to changes in the competitiveness of university admissions, consistent with experimental evidence. The results indicate that males, in both the advantaged and disadvantaged groups, did not respond to the policy change. Women, on the other hand, responded strongly, improving their performance in GPA and in the verbal test in response to both the increase in competition and the increased incentive to exert effort provided by affirmative action.
One of the main findings, however, is that increased competition led to lower math test scores for women, while reduced competition led women to increase their performance in math. A student's goal should be to maximize the academic index used for admissions, which places equal weight on verbal and math scores. Under time constraints, if women believe they are more effective at improving their verbal scores, they should allocate more time to studying for the verbal test, perhaps even allocating too much time away from studying math, which would result in lower math scores.
However, only women who experienced higher levels of competition had lower math scores. Women from the disadvantaged group who experienced less competition improved their math scores, as well as their GPA and verbal scores. The second main finding is that women in the disadvantaged group who experienced less competition applied to more selective programs. In particular, the competitiveness of their top-ranked choice, which should reflect their true preference given the assignment mechanism in place, saw the biggest increase, even net of the effect due to their improved performance. Given the persistence of a wage gender gap despite women's higher educational attainment, how competitive admissions influence sorting into specific universities and majors emerges as a key question.
The third chapter, which is a joint work with Ruth Uwaifo appearing in the Economics of Education Review, studies how expanding access to higher education affects college graduates once they reach the labor market. We focus on a major university education reform in Venezuela known as Mission Sucre, which provided free, open-access tertiary education targeted to the poor and marginalized, and its potential impact on returns to university education on non-participants. We begin by finding that returns to education decreased in Venezuela over the period Mission Sucre was introduced, despite a previous upward trend and an economic boom. Although returns to all levels of education declined during this period, the return to university education fell by over 10 percentage points more than other levels.
Motivated by these preliminary findings, we evaluate the possible role of Mission Sucre on the significant decline in returns to university education. For our main analysis, we compare the returns to university education and technical education in a difference-in-difference strategy. We focus on these particular levels of education because both are tertiary levels and are more likely to have similar general trends in returns. More importantly, Mission Sucre originally focused on only expanding university education. This allows us to classify those with university education as a treatment group and those with technical education as a potential control group. We find that Mission Sucre led to a 2.7 percentage point decrease in returns to university education of non-participants in the 23-28 age cohort between 2007 and 2008, the year the first cohort of Mission Sucre graduates entered the labor force. Further, states with higher shares of Mission Sucre students had a larger decline in the returns to university education. Specifically, a 1 percent increase in the share of Mission Sucre students led to a 0.4 percentage point decline in the returns to university education.
Although we provide ample evidence of the impact of Mission Sucre, we cannot state whether the noted effect of the program is driven solely by an excess supply of skilled labor, or a combination of the excess supply and other negative externalities of the program on nonparticipants, such as a change in the perceived overall quality of public higher education. Nevertheless, our results present a cautionary tale of the short-term effects of a rapid and large expansion in access to university education.
- Gonzalez_columbia_0054D_12671.pdf application/pdf 2.14 MB Download File
More About This Work
- Academic Units
- Thesis Advisors
- Macleod, William B.
- Ph.D., Columbia University
- Published Here
- November 5, 2017