Research

Work in Progress

My research focuses on the provider side of health care. Drawing on insights and approaches from labor economics and industrial organization, I study how providers decide to use new technology. I also have work investigating how the health care industry affects local health and local economies.

Medical Technologies with Comparative Advantages on Different Dimensions: Evidence from Hysterectomy

2021- 2022 Job Market Paper

Please Click Here to Download

This paper investigates why old and new medical technologies coexist. I show that the use of different surgical technologies across heterogeneous patients can be attributed to the existence of tradeoffs between multiple dimensions of health. I develop a Roy model in which physicians and their patients choose a surgical procedure, considering how the procedure affects two clinical outcomes. Laparoscopic surgery -- a newer technology in which surgeons manipulate thin instruments through small incisions -- results in shorter lengths of stay than abdominal, or open, surgery -- an older, more invasive technology -- yet not all surgical patients choose the laparoscopic procedure. The model shows that patients on the technology choice margin experience greater readmission risk under laparoscopic surgery than under abdominal surgery, although they experience shorter lengths of stay. I find evidence consistent with this prediction among Medicare-covered hysterectomy patients. I estimate the local average treatment effects and marginal treatment effects of laparoscopic surgery, relative to abdominal surgery, on patient outcomes using the difference between a patient's distance to her nearest laparoscopic surgery-performing hospital and the distance to her nearest hospital as an instrumental variable for choosing laparoscopic hysterectomy over abdominal hysterectomy. Compliers of the instrument are less likely to experience a long length of stay but more likely to have a hospital readmission under laparoscopic hysterectomy than under abdominal hysterectomy. I also use the estimated local and marginal effects to calculate the revealed preference for shorter length of stay relative to lower readmission risk, which in the absence of hospital influence over the treatment choice could be considered a marginal rate of substitution. Medical technologies do not pose tradeoffs just between cost and quality: this paper shows that they can pose tradeoffs on multiple quality dimensions.

The Effect of Subsidies for Hospital Construction on Mortality: Evidence from the Hill-Burton Program


I estimate the mortality effects of a federal program that subsidized hospital expansion. From 1948 to 1971, the Hospital Survey and Construction Act of 1946, known as the Hill-Burton program, transferred $28 billion (in 2012 dollars) from the U.S. federal government to counties in need of more hospital beds at a time when private credit markets were unprepared to do so. I use mortality data from Vital Statistics and Hill-Burton program data from the U.S. Department of Health, Education, and Welfare. Employing an event study approach, I estimate that this program's subsidies reduced overall mortality rates by 0.6 deaths per thousand residents in counties where the subsidies were awarded, which is 6% of the baseline average morality rate of 9.3 deaths per thousand.


Does Health Care Help Local Economies Weather Recessions?

with Marty Gaynor and Brian Kovak

We show that having a larger health care share of employment may have attenuated the effect of the 2006-2009 housing crisis on employment in local goods and services, or nontradable employment. We construct a model of regional economies, augmenting the Jones (1975) specific factors model, and derive that the relationship between an income shock and labor demand is attenuated by larger shares of health care employment. When health care is implicitly subsidized through a wider insurance pool such as Medicare, a larger baseline health care share of employment implies that a larger share of a region's income comes from this outside pool, causing an income shock such as the U.S. mortgage crisis to have less of an impact on the region's labor demand.

Using employment data from the U.S. Census Bureau, we find evidence consistent with this. For a county with average health care share of employment (15% of employment), the employment drop associated with a - 20 percentage point net wealth shock is 5.65 percentage points greater than the employment drop associated with a mere 1 percentage point net wealth shock. However, a county with 20% of its employment in health care (an additional standard deviation) experiences only a 4.35 percentage point greater employment drop under a large net wealth shock than under a very small net wealth shock. This means an additional standard deviation of health care's share of employment causes a 1.30 percentage point decline in the employment drop associated with the net wealth shock moving from the 10th percentile to the 90th percentile.