rethinkTHC Search
Menu
Study breakdown

Recreational cannabis legalization caused increased traffic fatalities, primarily after retail sales began

ObservationalModerate evidence
The takeaway

Using augmented synthetic control methods across US states, researchers found a consistent but lagged increase in traffic fatality rates after recreational cannabis legalization, primarily linked to retail availability.

Traffic safety researchers, cannabis policy designers, state legislators considering legalization

Retail availability, not legal status, drove fatality increases

What the researchers found

Using an augmented synthetic control method to generate causal inference, the study identified a consistent but lagged pattern of increased traffic fatality rates in several states following recreational marijuana legalization. The effect was primarily linked to the beginning of retail sales rather than the legalization date itself. The findings disprove conjectures dismissing the link between recreational marijuana use and fatal traffic crashes.

Why it matters

This study uses one of the most rigorous causal inference methods available for policy evaluation and finds that recreational cannabis legalization does increase traffic fatalities, but only after retail sales begin. This distinction is critical for policy design: the transition to commercial availability, not legal status alone, drives the safety impact.

The numbers in context

Multiple US states analyzed; consistent lagged increase in fatality rates post-legalization; effect primarily linked to retail availability rather than legalization date

How the study worked

Augmented synthetic control method applied to US state-level traffic fatality data. This approach creates synthetic comparison states from non-legalizing states to estimate what would have happened without legalization, adjusting for unobserved confounders. Multiple states analyzed across different legalization timelines.

What this study cannot tell us

Augmented synthetic control cannot fully account for all confounding changes that coincide with legalization. Traffic fatality data may not capture the full picture of impaired driving incidents. Cannot distinguish between cannabis-impaired driving and other behavioral changes associated with legalization.

How to read the evidence

Augmented synthetic control method provides strong causal inference for policy effects, though residual confounding and inability to measure impairment directly are limitations.

When this study was published

2025 publication

The bigger picture

The debate over whether cannabis legalization increases traffic fatalities has been contentious, with mixed findings from simpler analytical approaches. This study's use of a more rigorous causal method provides stronger evidence for the link while pinpointing retail availability as the key mechanism.

Questions still open

  • Would delaying retail sales after legalization reduce the traffic safety impact? Do states with stricter impaired driving laws for cannabis see smaller increases?

Common questions

Does legalizing cannabis increase traffic deaths?
This study found yes, but with an important nuance: the increase was lagged and primarily associated with the start of retail cannabis sales rather than the legalization date itself, suggesting that commercial availability drives the effect.
What makes this study different from previous research?
Previous studies used simpler methods like difference-in-differences that may not fully account for confounding factors. The augmented synthetic control method creates more precise comparisons by constructing synthetic control states, providing stronger causal evidence.

Read the original research

Evaluation of the causal impact of recreational marijuana legalisation on traffic safety in the US.

Accident; analysis and prevention, 220, 108106

Citation

Anupriya; McCoy, Emma; Graham, Daniel J. (2025). Evaluation of the causal impact of recreational marijuana legalisation on traffic safety in the US.. Accident; analysis and prevention, 220, 108106. https://doi.org/10.1016/j.aap.2025.108106