Do policies that promote cross-caste contact expand economic opportunities and increase support for affirmative action? I study this question by exploiting random assignment of first-year roommates at a highly selective public university in India, using original survey data on 2,267 students across two cohorts. A theoretical framework of social integration guides my analysis: students allocate scarce social time between within-group ties, which provide belonging and social support, and cross-group ties, which can expand economic opportunities. One year after assignment, lower-caste students in mixed-caste rooms have better grades but worse mental health than their caste peers in same-caste rooms, while upper-caste students' grades decline.
Data on social interactions shed light on the mechanisms: lower-caste students gain academic help but lose social support, while upper-caste students lose same-caste academic collaborations. Consistent with these frictions, students are more likely to choose same-caste roommates when given the choice in the second year. Three years later, these costs extend to the labor market: upper-caste students receive 20% lower internship stipends, while lower-caste students are less likely to receive a summer internship offer. Cross-caste contact initially increases both groups' support for affirmative action; by year three, however, mixing reduces upper-caste support for employment reservations. Overall, cross-caste contact narrows academic gaps but entails asymmetric economic and psychological costs, shaping subsequent social choices and attitudes toward affirmative action.
This paper examines how drought affects collective action in rural India. Using the two-wave
India Human Development Survey panel (2005 and 2012) and district-level drought measures (SPEI), I
compare households in districts within the same state that experienced different numbers of dry seasons.
Wet-season (Kharif) drought is not associated with reported community cooperation on water problems,
while dry-season (Rabi) drought increases cooperation by about three percentage points, which does not
differ significantly by baseline irrigation access. Ongoing work investigates the mechanisms
Work in Progress
Family Trees of Opportunity: Vertical Peer Effects and Skill Transmission in Public Universities
with Oscar Diaz and Palaash Bhargava
How do vertical ties to older mentors and horizontal ties to peers jointly shape students’ long-term economic opportunities? We study a mentoring program at a highly selective public engineering institution in India, where incoming students are randomly assigned an older-student mentor and placed in groups of five to six co-mentees. Combining administrative records spanning 17 cohorts and 14,599 students with LinkedIn profile data, we examine how caste and specialization matches with mentors affect labor-market outcomes, and whether mentor ties complement or substitute for peer ties. The setting also allows us to map networks beyond direct ties, tracing connections through mentors’ own mentors and their other mentees to examine how indirect ties shape outcomes. The project is in the final stages of linking administrative records to LinkedIn profiles.
Enabling Access to Safe Water: Evidence from Gujarat
with Vrinda Sharma
In rural India, groundwater quality is threatened by seawater intrusion and chemical contamination, as bore wells and aquifers deepen. In partnership with an NGO in Gujarat, we are piloting interventions to improve the operation of reverse osmosis (RO) plants, which reduce salinity and remove chemical contaminants, and encourage households to use treated water. On the supply side, preliminary evidence from a pilot involving 14 plants suggests that operator training combined with financial incentives may help keep plants operational. On the demand side, we surveyed 140 households across seven randomly selected pilot villages, combining household surveys with administrative data on plant-level water consumption to measure uptake. All households in these villages received information on the health risks of unsafe water and vouchers for three months of free treated water, while a randomly selected half also received water-quality testing and individualized feedback. Preliminary comparisons suggest higher uptake in pilot than in non-pilot villages, with no detectable additional effect of testing and feedback. These early findings will inform a large randomized controlled trial on plant operations and household uptake. Because RO plants require reliable water and electricity, their locations are constrained by infrastructure availability, potentially limiting access for households in remote hamlets. We are piloting water kiosks to bring treated water closer to these communities and test whether reducing the distance households travel to collect water increases uptake.
Transfers, Income Diversification and Climate Adaptation: Long-term Evidence
with Karen Macours, Patrick Premand, and Renos Vakis
Analysis ongoing
Building Skills in Agri-Food Systems: A Meta-Analysis of Extension and Training Programs
with Karen Macours, Marcelo Gantier, and Amit Khanal
Analysis ongoing
Natural Disasters, Land Markets, and Inter-Religious Dynamics: Evidence from Bihar
with Oliver Vanden Eynde and Nitin Bharati
Analysis ongoing
Teaching
Econometrics 2A , Teaching Assistant, Analysis and Policy in Economics (APE), Paris School of Economics, 2026–27
Econometrics 2, Teaching Assistant, Master's in Economics, Université Paris 1 Panthéon-Sorbonne, 2026–27
Linear Algebra (undergraduate), Teaching Assistant, Université Paris 1 Panthéon-Sorbonne, 2026–27
Econometrics 1 , Teaching Assistant, Master 1 in Public Policy and Development (PPD), Paris School of Economics, 2022–23 and 2023–24