Thankyou to Lex Pelger and his wonderful research abilities for highlighting this
Highlights
- •
We examine the relationship between marijuana laws and crime across all 50 states.
- •
We include medical (MML) and recreational (RML) laws; property and violent crimes.
- •
We use a synthetic DiD model and a dynamic model, and include covariates.
- •
Our main result is that MML reduces property crime, while RML reduces violent crime.
- •
Policymakers should be aware of this heterogeneity when designing relevant policy.
Abstract
Novel policies may generate unintended spillovers, particularly when legalizing one activity alters incentives for other forms of crime. Marijuana legalization provides a useful setting to examine such effects, given the staggered adoption of medical and recreational laws across all 50 U.S. states. Using both difference-in-differences and synthetic difference-in-differences approaches, we assess how specification choices shape estimated impacts on property crime. Initial results from both methods suggest that recreational legalization increases property crime; however, once state-specific time trends are incorporated into our preferred synthetic specification, the effect becomes negative and statistically insignificant. Overall, the findings indicate that estimated crime effects are highly sensitive to identification assumptions and do not provide robust evidence of an increase in property crime following legalization, underscoring the importance of careful empirical design in policy evaluation.
Introduction and overview
UN Reclassifies Cannabis as a Less Dangerous Drug. A United Nations commission voted on Wednesday to remove cannabis for medicinal purposes from a category of the world’s most dangerous drugs, a highly anticipated and long-delayed decision that could clear the way for an expansion of marijuana research and medical use. The New York Times, December 2, 2020.
What are the socioeconomic consequences of marijuana legalization? Some argue that legalization removes a social stigma and spurs economic
Related literature
Of course, we are not the first to study this topic, and we are interested in the effects of marijuana legalization for multiple reasons. First, we should think about the overall determinants of crime – we hope to contribute to the vast literature which examines this from many angles. Lovett et al. (2022) is a recent paper that assesses the relationship between policing and crime using data from California, finding, like several papers before it, little evidence of a relationship. From a
Data
Since we analyze state laws and crime, and need to control for confounding variables, we had to compile our data set from a variety of sources. These data were sometimes reported at different frequencies, and made available at different periods. We therefore ultimately use data for a sample period where we could collect all of the variables and controls. The main data on crime rates come from FBI Uniformed Crime Reports (UCR) arrest data from 1995 to 2019, while the data on recreational and
Empirical results
We pursue our empirical estimation with several approaches. We start with a difference in differences approach, and then estimate a dynamic model that accounts for implementation lags. Our most important results are from synthetic difference in differences estimation. We discuss each in turn.
Conclusions
We examine the relationship between marijuana legalization and crime rates, using state level panel data that we compile from a large variety of sources. Since legalization occurred in a staggered fashion across states, we can study interesting dynamics in the impact on crime. While previous research tends to focus on a few states, we consider all 50 US states. And while previous research typically examines either recreational marijuana (Dragone et al., 2019, Sabia et al., 2024, Lee, 2025) or
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Access Full Paper
