The Cost of Diversifying R&D (Job Market Paper)
Presentations: American Society of Health Economics 2026, Healthcare Innovation Working Group 2026.
Discussed in forthcoming textbook by Lawton R. Burns.
Abstract: Firms in R&D-intensive industries often take multiple "shots on goal," where they invest in several product candidates to increase the likelihood that at least one reaches the market. Leveraging quasi-experimental variation and novel data on pharmaceutical R&D, I show that this diversification is costly. When firms spread their scarce resources over multiple candidates, they are less likely to successfully develop at least one product. If they do, it takes longer to reach the market. I develop a model of corporate investment in which R&D diversification arises from three mechanisms: complementarity, diminishing returns, and risk preferences. Risk aversion can explain why some firms choose diversification, even when it is productively inefficient. Structural estimation reveals that the majority of firms in my sample invest as if they are risk averse. I provide empirical evidence suggesting that such behavior arises due to principal agent frictions. My model explains observed investment patterns better than canonical IO models of innovation in which firms only maximize expected profits. The widespread existence of risk averse firms suggests that current policy incentives for innovation over-compensate firms for R&D risk.
Working Papers
Sites and Subjects: Selection in Clinical Trials
with Olivia Zhao
Presentations: American Society of Health Economics 2026
Abstract: Routine care costs of clinical trial participation are not always covered by insurance, which may shape who participates in a trial. We ask whether increased coverage expands access to medical innovation for patients and shapes which medical innovations reach the market. We assemble a 2018 to 2023 panel of U.S. oncology trial sites, linking ClinicalTrials.gov and TrialTrove records to hand-collected state coverage mandates and demographic data at the ZIP code level from ACS. A difference-in-differences design exploits the 2022 Clinical Treatment Act, which extended routine-cost coverage to Medicaid beneficiaries in states lacking prior mandates. Overall, trial and site counts did not rise, but new trial sites shifted to lower-income ZIP codes. Subsidizing participation appears to relocate research, shifting the pool of potential enrollees towards lower socioeconomic populations.
Selected Works in Progress
Make and Buy: Incomplete Markets from Own-Production
Speed vs. Power: Enrollment Selection in Clinical Trials
Presentations: American Society of Health Economics 2025
Publications
How to price the IP rights underpinning disruptive biotech innovation
with Mohan Rao
IAM Law Business Research, July 23, 2020