Research

Work in Progress

Market Size and Trade in Medical Services (with Jonathan Dingel, Josh Gottlieb, and Pauline Mourot)

We document substantial interregional trade in medical services and investigate whether regional economies of scale explain it. In Medicare data, one-fifth of production involves a doctor treating a patient from another region. Larger regions produce greater quantity, quality, and variety of medical services, which they “export” to patients from smaller regions. These patterns reflect scale economies: greater demand enables larger regions to improve quality, so they attract patients from elsewhere. Contrary to concerns that production is too concentrated, we estimate that larger regions have higher marginal returns. We study counterfactual policies that would lower travel costs rather than relocating production.

Resources: BFI Summary, BibTex, Public Data
Media: Washington Post, The Center SquareChicago Booth ReviewFor All

Knowledge Growth and Organization of Expert Work: Evidence from Oncologists

Many fields, such as computer science, molecular biology, and medicine, have a rapidly growing knowledge base. Do expert workers respond to growth in knowledge by becoming more specialized? We study this empirically in the context of oncology, where we document explosive growth in knowledge. Using Medicare claims data and novel panel of historical cancer treatment guidelines, we test if oncologists exposed to greater knowledge growth become more specialized. We proxy knowledge growth in each cancer subfield with the increase in the length of clinical guidelines. Exposure to knowledge growth causes ex-ante specialized oncologists to become even more specialized but does not affect general oncologists. Growth in specialization over time occurs entirely among specialists in large markets. This pattern leads to growing geographic inequality in specialization and suggests that knowledge growth increases economies of scale in expert work.
Accuracy and Interpretability in Government Payment Algorithms
I empirically investigate the trade-off between accuracy and interpretability in Medicare Advantage risk adjustment models. I introduce a formal metric for model complexity in payment policy, which equates complexity to the number of coefficients in a model, a factor central to stakeholder interpretation of payment rates. Machine learning models significantly improve prediction accuracy and robustness to upcoding but also dramatically increase complexity. An analysis of policymakers’ preferences reveals that these models likely do not justify their additional complexity. Future research should explore aligning machine learning advances with payment policy constraints.
The Art of Medicine as Economics: Modeling Clinical Tradeoffs in the Production of Health
Standard economic models assume that monetary costs are the primary constraint to the use of medical care. However, clinical medicine abounds with situations where the use of effective medical treatments is not constrained by monetary costs but by the negative side effects of treatments, which I refer to as “health costs”. Navigating these trade-offs is referred to clinically as “the art of medicine.” I model these trade-offs by making a simple but substantive change to standard economic models of healthcare consumption. Medical treatments have a health cost. Treatment improves health in one dimension but harms it in another, constraining healthcare use. Patients whose medical care is constrained by health costs are unresponsive to monetary price changes. The model provides further predictions about situations in which price will be an effective policy lever and in which moral hazard will be limited. It also provides economists with a framework for thinking about clinical tradeoffs, a central driver of patterns of healthcare consumption.