Research

Can more precise decisions make a system less stable?

The George Washington Bridge has two decks. My grandfather thought people preferred the top one for the view, so he was adamant about taking the bottom deck, for efficiency's sake. What if everyone thought like him? The bottom deck would become the crowded one, making the top deck the better option. If everyone switched again, we would be back where we started.

I study a simple model of this problem, where people choose between two options based on how crowded they are. More precise decisions can keep the group switching back and forth rather than settling down, depending on how many people change their choices at once.

A related question arises when people have AI agents trade on their behalf: if many agents spot the same undervalued stock, their purchases can erase the opportunity. Would more capable agents make markets settle faster, or could responding together create new swings? This paper studies the simpler two-option model as a starting point.

Sharper responses, less stable adjustment: synchronization in a binary congestion game
Work in progress · Early draft (PDF)

How do you assess a borrower who has no credit history?

A person can earn a steady living and repay what they owe without having a credit file. Stemming from a hackathon project, this work explores how income records, people willing to vouch for a borrower, small-loan repayments, and saving habits might help a lender assess them. We examine what each kind of evidence can tell us, how it could be manipulated, and how to balance accuracy with fairness. The framework has not yet been tested on loan data.

The equitable financial inclusion score (EFIS): a microeconomic framework for credit assessment in underserved populations
Working paper · October 2026 · Draft (PDF)

How can we model a crowd without treating everyone as identical?

People and firms differ in how much they contribute to a market. I model participants of different sizes, each responding to the market as a whole. The question is whether their choices can settle into a balance where each is content with their own decision, given what everyone else is doing.

Granular games with $p$-liquid atomic masses
Working paper · Paper on SSRN

What role do laws play in labor automation?

Being able to automate a task is different from being allowed to do it. I am working on a measure of how exposed different occupations are to AI that also accounts for licensing, regulation, and uncertainty about future rules. The aim is to better understand the gap between what technology can do and what employers actually adopt.

Regulation-adjusted AI exposure: an occupation-level index for the United States
Work in progress

Does removing a degree requirement help people?

Removing a college-degree requirement makes more people eligible for a job. Does it change who gets hired or what they earn? I am studying these questions in state government, with particular attention to workers without a degree.

Tearing the paper ceiling for whom? The hiring and wage effects of removing degree requirements in state government
Work in progress