Dissertation Defense: Dario Sansone
Candidate Name: Dario Sansone
Advisor: Garance Genicot, Ph.D.
Title: Essays in Applied Microeconomics
My dissertation focuses on understanding whether and how institutions, policies and norms lead to an inefficient allocation of human capital - with specific focus on marginalized individuals - and what kind of interventions can be used to reduce such inefficiencies. In Pink Work: Same-Sex Marriage, Employment and Discrimination, I analyze how the legalization of same-sex marriage in the U.S. affected employment levels among gay and lesbian couples. Previous amendments in marriage laws, such as the introduction of unilateral divorce laws, led to substantial changes in the labor market. In addition, access to marriage led to changes in tax, health insurance and adoption laws that could have encourage some individuals to specialize in household production. Nevertheless, I compare same-sex couples living in different states over time to show increases in the individual and joint probabilities of being employed following the introduction of same-sex marriage in their state. I then provide empirical evidence suggesting that a decrease in discrimination towards sexual minorities was the driving mechanism.
In my second dissertation chapter, I combine my interest in education and gender to analyze the relationship between teacher demographic characteristics and student educational outcomes. This paper was motivated by the growing literature on the effect of teacher gender on student achievement. However, female teachers are an extremely heterogeneous group, so I wanted to probe the overwhelming positive impacts of female teachers on female students emphasized in several previous studies. In Why Does Teacher Gender Matter?, I show that the effect of high school math and science teacher gender on student interest and self-efficacy in these subjects becomes insignificant once teacher behaviors and attitudes are taken into account, thus pointing towards an omitted variable bias. Teacher beliefs about male and female ability in math and science – as well as how teachers treat boys and girls in the classroom – matter more than teacher's own gender. Scholars have been concerned that female students may perform worse than their male counterparts because of low self-confidence. Rather than hiring more female teachers, I explain how increasing the share of teachers who make their subject interesting and listen to ideas from their students could be more effective in increasing student self-efficacy and interest in science and math.
My last chapter reiterates my research philosophy of using state-of-the-art quantitative methods to analyze topics with important ramifications in the real world. Failing to graduate from high school has high individual and social costs. And yet, high schools in the U.S. tend to rely on few indicators – attendance, behavior, and past grades – in order to identify students at risk of dropping out. In Beyond Early Warning Indicators: High School Dropout and Machine Learning, I show that this parsimonious approach leads to identifying only a small subset of students who ends up dropping out. I then claim that schools can obtain more precise predictions by exploiting the available high-dimensional data jointly with machine learning techniques. I also incorporate economic theory into machine learning: the algorithms are calibrated not by selecting an ad-hoc goodness-of-fit criterion, as typically done in the literature, but by considering the goal of minimizing the expected dropout rate while respecting the school’s budget constraint. Educational organizations have access to datasets which are increasing exponentially over time: machine learning can help teachers and administrators to readily identify and target students at-risk early on, thus leading to a substantial reduction in dropout rates.
Wednesday, April 3, 2019 at 2:00pm to 4:00pm
Edward B. Bunn, S.J. Intercultural Center, 550
37th and O St., N.W., Washington