Using Synthetic Controls to Answer Health Policy Questions

One of the great challenges in doing empirical research is trying to determine what might have happened under different circumstances.

Take the question of whether or not legalizing recreational marijuana leads to higher crime rates. There are strong arguments suggesting a rise in marijuana use could lead to either an increase or decrease in delinquent behavior, including crime.

While examining crime rates before and after a state (or several states) legalize recreational marijuana can provide suggestive evidence, the only way to be assured of marijuana’s effect on crime is to know what would have happened to crime rates if the same state did not legalize recreational marijuana. Unfortunately, knowing this counterfactual is impossible.

However, a host of empirical techniques called quasi-experimental methods can help us isolate the contribution of single factors to changes in outcome(s). Among the most illustrative and popular of these techniques is the synthetic control method.

Developed by Abadie and Gardeazabal in 2003, the synthetic control method is a data-driven procedure used to study the effects of an intervention or policy in comparative case studies. The method creates a hypothetical counterfactual by constructing a weighted average of control units (states, countries, etc.) from a donor pool (units which were not exposed to the policy intervention), to compare a synthetic unit to the real unit. The counterfactual created by the synthetic control mimics the behavior of the unit examined before the period or intervention took place and approximates what would have happened if the treatment unit was not exposed to the policy or intervention.

The synthetic control method has many advantages. Unlike other quasi-experimental methods, the synthetic control method can account for the effects of confounders by weighting the control units to better match the treatment unit before the intervention. It combines propensity score matching and difference-in-difference methods to estimate and account for confounder effects. Synthetic controls can also compare several units in a donor pool with the treatment unit, providing a more robust comparison. Perhaps most importantly, the outcome of the study cannot be influenced by the researcher because it does not require access to postintervention outcomes.

Synthetic controls have been used to examine a variety of fascinating questions, including the effects of immigration policy on the racial composition of employment; legalized prostitution on sex crimes; right-to-carry laws on violent crime rates, and China's one child policy on fertility rates. Health economists and health policy analysts are particularly drawn to the synthetic control method because it offers a way to test for placebo effects. Recent studies have used this method to examine the impact of contraceptives on teen birth rates; medical marijuana laws on body weight; and school district nutrition policies on dietary intake and childhood obesity.

The question of how marijuana legalization affects crime was examined by Dr. Glenn Furton, a post-doctoral student at New York University. Using a synthetic control for the state of Colorado, Furton finds the legalization of recreational marijuana resulted in increased drug, property, and violent crimes.

Our current research employs a synthetic control method to examine the impact of recreational marijuana legalization on opioid abuse by contrasting the actual state of Colorado with a synthetic Colorado where recreational marijuana was never legalized. As we’ve noted in a previous post, marijuana use could be a viable option to help reduce U.S. dependency on opioids.

Our research hopes to improve our understanding of whether marijuana use can help reduce opioid abuse. Using the synthetic control method, we hope to share our findings soon. 

Meet the Authors

Elisha Kwaku Denkyirah is a graduate research assistant in the Department of Agribusiness and Applied Economics at North Dakota State University. He holds a master of philosophy degree in agribusiness from the University of Ghana.


Raymond March is a fellow at the Center for the Study of Public Choice and Private Enterprise (PCPE) and an assistant professor in the NDSU Department of Agribusiness and Applied Economics. Read his bio.

Top of page