Essays in development economics
This dissertation contains three chapters, each investigating a question in Development Economics. The first chapter, co-authored with Pascaline Dupas and Zhongyi Tang, studies the relationship between national institutions and subnational development in Africa. Detailed data about infrastructure access in Africa is scarce, so we leverage satellite imagery and survey data to train a machine learning model that predict access to infrastructure for the whole continent with high levels of accuracy. We first use this data to implement a spatial regression discontinuity design to study how much of the heterogeneity in infrastructure access across countries comes from differences in institutional quality, finding a positive effect of a modest magnitude, reconciling previous contradicting results in this literature. We also use this data to study the role of political favoritism in explaining within country heterogeneity, finding that areas with political ties to current or former presidents have better access to infrastructure. In the second chapter, co-authored with Bridget Hoffmann and Juan Pablo Rud, we use high-frequency data on fine particulate matter air pollution (PM 2.5) to study the effects of high pollution on health outcomes in the Metropolitan Area of the Valley of Mexico. We combine hourly monitoring station data on air pollution and weather conditions with a rich dataset describing more than 10 million health episodes between 2003 and 2019, including deaths, hospitalizations, and urgent care visits. We disaggregate daily mean concentrations of PM 2.5 using the daily share of hours with PM 2.5 concentration above each WHO threshold, which allows us to uncover a positive non-linear and convex relationship between hourly air pollution concentrations and same-day respiratory health outcomes of all severities. We find that hours above IT1 have effects on respiratory health outcomes that are 20 to 30 times greater than those of hours above the Air Quality Guideline, the lowest WHO threshold. Furthermore, we find that 1 additional hour a day with PM 2.5 above IT1 has the same effects on respiratory health outcomes as increasing the daily average concentration of PM 2.5 by 41 mug/m3, which would move daily PM 2.5 concentrations of a day with no air pollution to the 95th percentile of the daily PM 2.5 distribution in the Metropolitan Area of the Valley of Mexico. In the third chapter, I study how students from an elite college program in Chile interact with each other when an affirmative action admission program is introduced. Using administrative data on undergraduate Business and Economics students at the University of Chile, I leverage their random assignment to first-semester classes to define an exogenous network of peers, to study how students interact when an affirmative action admission program is introduced. I compile data on demographic characteristics and academic outcomes for 8 cohorts of students, with around 500 students each, and use to study classroom composition, network characteristics, and heterogeneous peer effects. I find that increasing the share of special admission students in a classroom improves grades for both groups, but decreases the passing rate for regularly admitted students, while at the same time increasing passing rates for special admission students. Network characteristics play a role in shaping academic outcomes, with students benefiting from interacting with students from the other group. However, peer effects tell a different story: there are negative endogenous peer effects across admission type groups, no endogenous peer effects among affirmative action students, and positive, negative or null endogenous peer effects among regular admission students. I find limited evidence of exogenous peer effects, with peer characteristics affecting mostly the overall career GPA of students
Thesis, Dissertation, English, 2025
[Stanford University], [Stanford, California], 2025