Interests: Applied Microeconomics, Industrial Organization, Experiments, Machine Learning, Agentic AI and Market Design
This research explores the dynamics of Massachusetts' affordable housing market by examining the search, application, and allocation processes for rental units in new developments shaped by state policies and tax incentives targeting low- and middle-income families. By studying household behaviors within this market, the study quantifies the welfare losses resulting from costly searches, coordination failures among institutions, and time-consuming application procedures. The findings highlight existing problems in the housing allocation system and offer valuable insights for policymakers to improve the efficiency and stability of the market.
This paper introduces a synthetic survey framework for estimating housing demand by integrating census data, real estate listings, and large language models (LLMs). Using the Current Population Survey, I construct representative household personas and elicit their preferences over actual housing options through LLM-enabled surveys. These synthetic responses are used to estimate affordable housing demand in Massachusetts, demonstrating how LLMs can generate scalable and low-cost survey instruments for markets with limited preference data.
This paper develops a new approach to the housing allocation problem by linking survey-based preference elicitation with machine learning and market design. Using survey responses, I recover the preference formation process across demographics and housing attributes, and then predict how individuals would value new housing opportunities as they arrive. These predicted preferences are embedded in allocation methods, providing a framework that connects individual-level demand with the availability of affordable houses.
In this paper, I investigate the impact of holding secondary or informal jobs on the lifetime earnings of workers. By designing and administering accessible surveys to low-income households, the study aims to understand their labor market choices and the long-term consequences on their earnings potential.
This paper models the GPU allocation problem in universities across Massachusetts as a dynamic matching problem. By examining the current allocation mechanisms employed by these institutions, the study evaluates their effectiveness using welfare metrics and compares the results with empirical data.
This research analyzes the impact of a policy change in the Cambridge Housing Authority's allocation mechanism for rental public housing units. The new method offers applicants greater choice in selecting housing units when applying but introduces a costly penalty. By examining this policy shift, the study explores how increased choice coupled with penalty affects applicant behavior, allocation efficiency, and overall welfare. The findings provide insights into balancing applicant preferences with policy objectives.
The impact of the New York City Housing Authority's unique allocation mechanism for affordable housing is analyzed in this paper. Unlike traditional methods, the authority simultaneously offers applicants both previously occupied rental units and newly developed units in a unique format. By analyzing this approach, the study explores how the inclusion of used units alongside new developments affects housing allocation efficiency, applicant choices, and overall market dynamics.
In this study, we constructed and augmented a dataset of 25,000 households from the Ottoman Empire's 1845 historical archives, offering fresh insights into the empire's taxation and socio-economic dynamics. Utilizing supervised learning techniques, we uncovered evidence that the Ottoman Empire employed non-linear taxation methods and imposed additional tax mechanisms specifically targeting its non-Muslim population. Furthermore, our quantitative analysis reveals that the Muslim population had a lower average income and faced higher income inequality compared to other religious groups, highlighting significant socio-economic disparities of the era.