Recreational Cannabis Legalization and Traffic Fatality Rates in the United States, 2005-2024: An Imputation-Based...

Background Recreational cannabis legalization has expanded rapidly in the United States; however, population-level analyses examining its association with traffic fatalities yield mixed results. These heterogeneous findings arise from methodological limitations, including short follow-up windows, a focus on early-adopting states, and methods that do not adequately account for staggered policy adoption. Methods We analyzed a 50-state panel (2005-2024) using an imputation-based difference-in-differences estimator designed for staggered policy adoption. The outcome was the log traffic fatality rate per one billion vehicle miles traveled. We estimated effects from the year of legalization through year four, with four pre-treatment years to assess parallel trends. The primary model included state- and year-fixed effects and controlled for demographic composition, weather (precipitation and annual average temperature), unemployment, traffic safety laws, and COVID-19 mortality. Sensitivity analyses estimated robustness by varying the sample composition (excluding early adopters, pandemic years, and Florida), modifying the treatment definition (legalization effective date vs. retail sales start), and including Poisson count-model replication. Results The panel included 1,000 state-year observations; the primary difference-in-differences analysis used 951 observations meeting all inclusion criteria. Post-legalization, traffic fatality rates increased by approximately 4.2% in the legalization year (τ=0: 0.041; 95% confidence interval (CI): 0.006, 0.077; p=0.022) and 6.1% by year four (τ=4: 0.060; 95% CI: 0.007, 0.112; p=0.027). Pre-treatment trends were parallel between treatment and control groups (joint test: χ²(4)=4.93; p=0.295), and the post-legalization increase was statistically significant overall (joint test: χ²(5)=11.56; p=0.041). Findings remained consistent when excluding early-adopting states, pandemic years, or Florida, and a Poisson count-model replication produced the same direction and magnitude. When treatment was instead indexed to the start of retail sales, no increase was observed in the first year (τ = 0: 0.011; 95% CI: -0.027, 0.050), and the overall post-retail joint test did not reach conventional significance (p = 0.067). Conclusion Legalizing recreational cannabis was followed by a rise in traffic deaths during the first four years after the law took effect. The increase began at legalization rather than at the start of retail sales. This suggests that the change in law and shifting public attitudes had a greater impact on traffic safety than the availability of cannabis in stores. States considering legalization should prepare traffic safety measures in advance, while those that have already legalized should reinforce existing efforts to address ongoing risks.