Research
Doing More with Less: School Management and Education Production.
[ Abstract | Draft ]
The school superintendent is the highest-ranking executive in U.S. school districts, responsible for managing personnel decisions and overseeing regular school operations. To estimate the causal effect of superintendents on district performance, I collect data on the tenures of over 18,000 school superintendents covering over half of American public school children using a model of test score value-added that allows a superintendent’s effect to emerge over the course of their tenure within a district. Superintendent transitions between districts are leveraged to validate these estimates and to identify common practices of effective superintendents. I show that superintendents have large effects on school district performance, accounting for one- fourth of the observed differences in learning rates across districts. Top management matters most in districts where managerial flexibility is ex-ante largest: smaller districts and districts with weaker teachers unions. Effective superintendents do not change levels of district spending or staffing but instead make changes in school operations, increasing teacher turnover and reducing teacher absences. Finally, I find evidence that the link between value-added and salary for superintendents is strongest in districts with higher levels of local interdistrict competition.
3G Internet and Human Capital Development (with Ronak Jain).
[ Abstract | Draft ]
We study the impact of global expansions in mobile internet access between 2000 and 2018 on student outcomes. We link geospatial data on the rollout of 3G mobile technology with over 2 million student test scores from 82 countries. Our findings indicate that the introduction of 3G coverage leads to substantial increases in smartphone ownership and internet usage among adolescents. Moreover, changes in 3G coverage are associated with significant declines in test scores across all subjects, with magnitudes roughly equivalent to the loss of one-quarter of a year of learning. We find suggestive evidence that a reduction in feelings of belonging, ease of making friends, and self-efficacy may explain these impacts.
Trends in International Assessments and Outcomes in Adulthood.
[ Abstract | Draft ]
International assessments such as PISA and TIMSS are widely used to compare the academic proficiency of adolescents across countries and over time. Do scores on these assessments predict outcomes in adulthood? Combining data from PISA, TIMSS, PIAAC, and 18 representative global surveys, I find evidence that cohorts with higher scores on the PISA assessment---which tests the application of knowledge to "real world" scenarios---score higher on assessments of adulthood skills, obtain higher levels of education, and have higher incomes as adults. Conversely, I find little evidence for similar patterns with respect to the TIMSS assessment, which tests curricular knowledge.
School Segregation, Student Achievement, and Parental Preferences.
[ Abstract | Draft ]
For most American K-12 students, where you live determines where you go to school. At the local level, school assignment is dictated by school attendance zones: boundaries that assign households to schools. I use geospatial data on the boundaries used to assign over 10 million public school students across 1,401 districts to identify changes in school assignment and estimate how these changes affect school segregation, student achievement, and neighborhood house prices. Comparing districts that impose integrative versus segregationary rezoning schemes shows that rezoning decisions can change segregation substantially, and that reductions in segregation narrow test score gaps between White and Black students. I find evidence that house prices fall in neighborhoods that are newly-assigned to schools with higher shares of Black and Hispanic students.
Budget Hawks on the Board: School Boards, Education Finance, and Student Achievement.
[ Abstract | Draft ]
Funding for education in America is spread across multiple levels of government, but financial decision-making is handled by locally elected school boards. During elections, many candidates for board seats run on promises of reforming district finances. I identify such "budget hawks" using natural language processing methods and campaign statements from school board candidates in California. I use a regression discontinuity design to test how district outcomes evolve in the years following the narrow victory of a hawk over a non-hawk. The election of a budget hawk leads to large and prolonged cuts in district spending. Using test score data, I find suggestive evidence that students in these districts exhibit lower rates of test-based proficiency in subsequent years. Heterogeneity analyses show evidence that districts that exhibit higher reductions in spending experience larger test score declines.
Economic Shocks and Skill Acquisition: Evidence from a National Online Learning Platform at the Onset of COVID-19 (with Ina Ganguili, Jamal I. Haidar, Asim Ijaz Khwaja, and Basit Zafar). Revise and Resubmit at Labour Economics
[ Abstract | Draft | NBER WP 29921 ]
We study how large shocks impact individuals’ skilling decisions using data from the largest online learning platform in Saudi Arabia. The onset of the COVID-19 pandemic brought about a massive increase in online skilling, and demand shifted towards courses that offered skills, such as telework, likely to be immediately valuable during the pandemic. Consistent with a model where individuals trade off reskilling costs with their expectations of future labor market conditions and their duration of work, we find that shifts into telework courses were largest for older workers. In contrast, younger workers increased enrollments in courses related to new skills, such as general, occupation-specific, and computer-related skills. Using national administrative employment data, we provide suggestive evidence that these investments in skills in early 2020 helped users maintain employment over the course of the pandemic.