Research
I study industrial organization, healthcare markets, and the economics of AI using structural econometrics, causal inference, and machine learning. My work examines hospital choice, policy-driven changes in access, and how capability, price, and safety relate to AI adoption. The full record of teaching, awards, service, and conferences is in my CV.
Publications & accepted papers
Decodable but Unused: What Language Models Represent (and Ignore) When Choosing Between Two Options
Using controlled insurance and coffee-choice tasks, I study whether language models use decision-relevant information encoded in their hidden states. Probes can recover the utility-maximising option even when the tested models’ choices remain near chance, highlighting a gap between decodable information and behaviour.
Local Population Characteristics and Access Equity of 340B Contract Pharmacies
The Wizard of OZ (Opportunity Zones): Spatial Spillovers in Place-Based Programs
AI-assisted teams outperform AI-led teams but not human-only teams in assessing research reproducibility in quantitative social science
Working papers & manuscripts
Quality Considerations and Universal Healthcare
A structural model of hospital choice under limited consideration, using neural networks to measure risk-adjusted hospital quality. Counterfactuals evaluate information provision and hospital-participation incentives in a large public insurance scheme in India.
Pennies for Pads: Subsidized Menstrual Health Products and Adoption in India
Uses doubly robust difference-in-differences and linked survey, administrative, and geospatial data for nearly 500,000 women to study adoption of subsidized menstrual health products in India.
Beyond a Free Lunch: Unintended Consequences of an Out-of-School Nutrition Program on Adolescent Girls' School Enrollment and Health Outcomes
Spectrum Allocation to Public Operators: A Structural Analysis of Set-Asides versus Subsidies in Indian Telecom Auctions
Research drafts
Demand for AI Models: Capability, Price, and Safety
Studies associations between price, capability, safety, and OpenRouter usage across 164 models and 7,946 model-week observations. Part of my SPAR fellowship on the economic value of AI capability gains, mentored by Pavel Kocourek.
Research reports
Detecting Collusion Between AI Agents
The team studied behavioral and conversational evidence of collusion across 1,800 agent runs and 18,000 episodes, evaluating a rule-plus-LLM classifier against 342 annotated episodes. I contributed to experiment design, the Python labeling pipeline using Modal, and writing.
Works in progress
Safety Screens and AI Product Introduction
Ongoing research on how safety assessments interact with AI model release decisions, including timing-based and random assessment approaches.
Dynamic Competition and Frontier AI Model Release
AI Agent Swarms and Collective Risk
Who Matches with Whom: Mergers and Acquisitions in Health Systems
Allocating Bundles: An Operations Research Framework
Long Distance Relationships: Analyzing Hospital-Pharmacy Contracts
Other writing
Incentives in the Healthcare System