Opioids for the masses: Welfare tradeoffs in the regulation of narcotic pain medications (Job Market Paper)
Abstract: Use of prescription opioid pain relievers to manage pain has increased fourfold since 1999, as medical guidelines have increasingly emphasized that appropriate pain management is required for an acceptable standard of care. However, a concomitant rapid rise in opioid abuse, addiction, overdose, and death has led to recent efforts to crack down on opioid prescribing. This paper sheds light on the tradeoffs of public policies that reduce the supply of medical opioids by investigating their health, labor, and welfare ramifications. I exploit state-level variation in the introduction of Prescription Monitoring Program (PMP) laws, and make use of several rich data sources, documenting that PMPs reduce the distribution of opioids, and achieve a key policy goal by reducing opioid overdose deaths by about 12%. I also find substantial costs resulting from these policies, including increased pain in the hospital setting, more missed days for injured and disabled individuals, and substitution towards more expensive medical care. A rough back-of-the-envelope welfare calculation suggests the welfare losses and gains from regulation are on the same order of magnitude - approximately $12.1 billion per year in increased costs from inpatient and outpatient medical spending plus lost wages, compared to $7.3 billion per year in benefits from lives saved from opioid and heroin overdose.
The Boston Globe, "With opioids, can you fight addiction without causing pain?"
The Chicago Tribune, "Column: Fewer opiates mean more suffering"
Marginal Revolution, "Opioids for the masses?"
Research in Progress
Understanding recent expansion in the demand for heroin
Abstract: Heroin overdose deaths have increased by 40% yearly since 2010, increasing from 3,048 in 2010 to 8,291 in 2013, the most recent year data are available. I investigate the determinants of this marked expansion in heroin overdose death, focusing on two potential explanations. First, I examine substitution by users between prescription opioid pain relievers and heroin. I study the effect of county-level illicit opioid availability, exploiting variation from state-level prescription crackdowns, and using a county border strategy that tests for cross-state spillovers from crackdowns in neighboring states. Second, I examine the role of lower heroin prices in heroin demand. I exploit variation in drug trafficking networks from south of the border through the United States, where east coast white powder is usually supplied by Colombia, and west coast black tar is typically supplied by Mexico, and utilize intertemporal differences in interdiction efforts in Colombia and Mexico. I find evidence that heroin is a short-run substitute and long-run complement for prescription opioids, and that opioids prescribed through the medical care system have therefore played an important role in driving the current expansion in heroin use. Research on the role of lower heroin prices in increased demand is ongoing.
Heterogeneous causal effects in the presence of many observables: An application to the treatment of pain
Abstract: Growing interest in “precision medicine” necessitates a better understanding of heterogeneous responses to medical treatment. While traditional econometrics techniques can be used to estimate average treatment effects conditional on observable characteristics, medical settings present implementation challenges due to the large number of observed covariates that typically characterize an individual’s disease state. In this paper I apply new methods pioneered by Athey and Imbens (2015), who import techniques from machine learning into contexts with quasi-experimental variation, to intelligibly characterize treatment effect heterogeneity in the use of opioids for the treatment of pain. Opioids are a natural setting for exploring this question: patient heterogeneity in response to opioid pain relievers has been identified as a key policy issue in the medical literature, and addressing this question in the context of opioids is particularly challenging given their prevalence across a wide variety of conditions and diseases. The approach used in this study will be generalizable to many other medical settings with similarly rich data and complex disease states.