The Effects of County-Level Naloxone Distribution on Drug Usage and Health Outcomes: Evidence from Indiana (Job Market Paper)
Abstract: Naloxone — a drug which can reverse an opioid overdose — has become a cornerstone of opioid harm-reduction policy, yet little is known about how institutional supply expansions affect public health, particularly in the fentanyl era. I study a statewide Indiana grant program that distributed naloxone kits to county health departments, providing the first evidence on this increasingly common distribution model. Using administrative data from 2014-2019 on naloxone kit distributions, emergency medical service (EMS) runs, health outcomes, and crime outcomes, I exploit staggered county participation in the program in a difference-in-differences framework to estimate the causal effect of the policy. I provide new first-stage evidence that program participation significantly increases naloxone administrations by 36.02%. Despite this, I find no improvements to opioid-related mortality and drug-related ED visits and hospitalizations. Using novel data on decedents' prior social services usage, I find strong evidence for a survival-reuse dynamic in which naloxone saves lives in the short run but survivors return to drug use and ultimately die: county participation in the program is associated with an 83.56% increase in the share of overdose decedents who had previously interacted with EMS. Further, opioid-related crime increases by 16.81%, consistent with increased drug-related activity. These results provide important insight into harm reduction policies during the fentanyl wave of the opioid epidemic and suggest that institutional naloxone distribution can successfully increase naloxone usage but that population-level health improvements require complementary interventions that reduce the risk of future overdose among survivors.
Difference-in-Differences with Bad Controls
Joint with: Carolina Caetano, Brantly Callaway, and Hugo Sant'Anna
Abstract: This paper considers difference-in-differences identification strategies when the parallel trends assumption holds after conditioning on covariates that may themselves be affected by the treatment (often referred to as "bad controls"). We show that common approaches such as simply dropping bad controls are often ill-advised and develop two alternative approaches that allow bad controls to function as genuine controls despite being affected by treatment. First, we derive explicit conditions that rationalize conditioning only on pre-treatment values of the bad control, leading naturally to the Callaway and Sant'Anna (2021) estimator with pre-treatment values as covariates. Second, under a covariate unconfoundedness condition, we develop imputation and double/debiased machine learning estimators that recover the average treatment effect on the treated. We extend these results to staggered treatment adoption, provide pre-tests for the identifying assumptions, and apply the methods to study the effects of job displacement on earnings.
Risky business: Pricing and participation in illicit drug markets
Joint with: Alexander Ahammer and Michael Irlacher