Essays in the Economics of Education
Name
AguilarLlanes-saloagui-phd-econ-2026-thesis.pdf
Size
3.33 MB
Format
Adobe PDF
Checksum (MD5)
325a184e9b0b7a1c44628e69834d5c38
Author(s)
Aguilar Llanes, Salome
Advisor(s)
Schilbach, Frank
Duflo, Esther
Mullainathan, Sendhil
Date Issued
May 2026
Publisher
Massachusetts Institute of Technology
Abstract
Expanding access to high-quality education is critical for reducing educational disparities. Online learning has emerged as a powerful tool to scale interventions that develop both academic and social skills, and represents a growing trend worldwide, with tutoring among its major components. Rapid advances in AI and natural language processing have further opened new windows for research into educational interactions and classroom dynamics. This thesis consists of three chapters studying different aspects of an online tutoring program in Mexico, developed and implemented by the researchers who coauthored these papers during our time as PhD students at MIT. Each chapter applies AI-based methods to study class dynamics alongside traditional experimental designs. The first chapter describes the tutoring program and estimates the results of a randomized controlled trial (RCT) measuring the impact on math learning. The program pairs university student volunteers with children for free online math tutoring. We find that students assigned to tutoring improved their math scores by 0.14 standard deviations. Beyond program evaluation, we showcase how the platform serves as a research tool: by randomly assigning tutors to students, it enables studies of tutor-student matching and teacher value-added. The second chapter focuses on social-emotional learning (SEL). Social skills predict earnings, teamwork, and productivity, and are in increasing demand in the labor market. Using the same tutoring platform, we implement an RCT randomizing delivery of a structured SEL curriculum of ten one-hour modules against a control group receiving ten one-hour history modules. The SEL group reported improved tutor-student rapport and increased math learning by 0.085 standard deviations. Applying a novel AI-based hypothesis generation method to session transcripts, we find that SEL-treated classes featured a slower pace, less technical language, less frustration, more personal topics, and a less formal overall tone in the opening minutes of class. The third chapter studies the impact of tutors’ traditional gender beliefs (TGB) on girls’ math learning. We conduct a randomized controlled trial in Mexico in which students are cross-randomized into single-gender or mixed-gender tutoring groups and assigned to tutors whose gender beliefs are measured using the survey instrument of alan2018. We find that girls in mixed-gender groups learn significantly less when assigned a TGB tutor, by 0.10 to 0.14 standard deviations, while girls in single-gender groups are substantially less affected, suggesting that the negative effect of TGB tutor assignment operates through differential attention toward boys in mixed-gender settings rather than through uniform reductions in teaching effort. We rule out belief transmission as the operative channel: post-intervention gender beliefs do not differ across treatment arms. Conversely, when tutors do not hold TGB, girls in single-gender groups underperform relative to those in mixed-gender groups by 0.08 to 0.13 standard deviations, indicating that the effectiveness of single-gender grouping as a policy intervention depends critically on the gender beliefs of instructors. Using an AI-based hypothesis generation approach applied to session transcripts aguilar2025recording, we find that TGB tutors deliver less group-level instruction, pose fewer questions to the class as a whole, and are more likely to address students individually and by name, a pattern consistent with differential allocation of attention across genders in mixed-gender settings, and one that attenuates when groups are single-gender. JEL Classification: I21, I24, C93, O15 Keywords: online tutoring, randomized controlled trial, math learning, Mexico
MIT Department
Massachusetts Institute of Technology. Department of Economics
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
