WORKING PAPERS

"Giving Men the Benefit of the Doubt: Credit Risk, Gendered Leniency, and Screening Incentives"  [Under Review ; Previously circulated as "Generous to Men or Harsh to Women?"]

Abstract: I study gender bias in small-business lending using a framed field experiment with 720 Egyptian loan officers evaluating past applications with randomized names. Otherwise identical applications are 6.3% more likely to be approved with male names. The gap is concentrated among profiles with repayment problems or lower credit scores, suggesting weaker responses to adverse information for men. I test interventions to reduce bias and expand women’s credit access. Increasing repayment-based incentives closes the gap through lower male approvals, without increasing female approvals. Bias feedback changes neither outcome significantly. These suggest reducing leniency toward men need not expand women’s credit access.

• Best PhD Student Paper Award (🏆Louis Pasteur Prize), Advances with Field Experiments (AFE) Conference 2025 (UChicago)

• Presentations: AEA Annual Meeting 2026 (Development Finance), University of Chicago experimental economics working group, Accelerating Growth for Women Entrepreneurs in Developing Economies conference (EBRD, We-Fi, IDB, IDB Invest, Imperial College London),  Hong Kong University of Science and Technology (Finance), KDI School of Public Policy and Management, National Taiwan University (School of Political Science and Economics, invited), Chung-Ang University (virtual), Advances with Field Experiments Conference 2025, EGSC 20th Meeting, FMA Doctoral Student Consortium 2025, Korea America Economic Association (KAEA) Job Market Conference, Korea Association of Applied Economics-DEEP (virtual), Bay Area Behavioral Economics and Experimental Workshop (USF)

"Human-in-the-Loop Bias in AI–Assisted Lending "  [Under Review]

Abstract:  The effectiveness of generative AI depends on how humans use its interactive features. I study human-in-the-loop bias in AI-assisted small business lending through a randomized experiment with 720 loan officers in Egypt. Officers evaluate past loan applications with randomized applicant names while receiving recommendations from a GPT-4o–based chatbot. While AI assistance improves overall screening accuracy, adding interactive features does not provide further gains and may even worsen outcomes by making officers more likely to override correct AI recommendations. I also document gender bias in interactions with AI: when AI incorrectly recommends approving low-quality loans, officers are more likely to question and override the recommendation for female rather than male applicants. These differential responses reallocate credit across genders, reducing approvals of low-quality loans for women while increasing them for men.

"Selection into Advice: Evidence Across and Within Individuals in Financial Decision Support?" with Robert L. Clark, Annamaria Lusardi, Olivia S. Mitchell, Hakan Ozyilmaz, and Andrea Sticha  [Under Review]

Abstract: We randomly offer 2,372 respondents opportunities to receive brief financial guidance before hypothetical financial decisions involving risk diversification, compound interest, and inflation. Among those offered advice opportunities, two-thirds opt in at least once, and randomized access to advice improves accuracy on the related financial-decision questions. We examine selection both across respondents and across decision domains within respondents. Across respondents, take-up is higher among those with greater objective financial literacy but lower among those with greater self-assessed financial knowledge. Within respondents, we find no evidence that advice seeking is associated with relative differences in knowledge across domains. Taken together, take-up is more strongly associated with differences across people than with relative knowledge gaps across domains within the same person. These patterns suggest that better targeting could involve both encouraging take-up among consumers with lower overall literacy and identifying domain-specific knowledge gaps to direct relevant assistance.

PUBLICATIONS

"From Lawn Care to Home Care: Undocumented Immigration and Aging in Place", American Journal of Health Economics, 2025, 11.3: 454–486, with Domininkas Mockus

Abstract: Elderly Americans express a strong desire to stay out of nursing homes as they age (aging in place). Undocumented immigrants play an important role in supporting the elderly who live at home. This paper estimates the effect of changes in undocumented immigration on aging in place. Identification comes from a two-stage least squares research design using the staggered rollout of the Secure Communities (SC) program between 2008 and 2013 that increased the threat of deportation for undocumented immigrants. For communities with a high level of undocumented immigrants, the change in the undocumented labor supply caused by the SC program decreases aging in place by 0.42%. As a general trend, for every 10% decrease in the undocumented labor supply, the percentage of elderly natives aging in place decreases by 0.12%.

"Does Single-Sex Schooling Improve Students' Physical Fitness?: Evidence from a Natural Experiment in South Korea", Economics Letters, 2025, 112279, with Sanghee Mun

Abstract: Leveraging a randomized natural experiment, this study examines the impact of attending single-sex middle schools on students' physical fitness, measured through standardized nationwide physical tests. In South Korea, middle school students are assigned by lottery to either single-sex or coeducational schools within their designated school zones, providing an ideal setting to evaluate the effects of single-sex schooling. Using school-level data covering all middle schools, the study finds that boys attending single-sex schools achieve significantly higher pass rates on standardized physical fitness tests, suggesting improved physical fitness compared to their peers in coeducational schools. However, no similar improvement is observed for girls attending single-sex schools. These findings suggest that single-sex schooling has differential effects by gender, highlighting the need to further research to understand the mechanisms underlying these varied outcomes.