Untitled

Abstract

Does access to legal cannabis reduce alcohol involvement among fatal traffic crashes through substance substitution? I test this hypothesis using a spatial regression discontinuity design at the borders between legal and prohibition states in the western United States. Using 29,350 geocoded fatal crashes from the Fatality Analysis Reporting System (2016–2019) and 1,399 dispensary locations from OpenStreetMap, I estimate the effect of crossing from prohibition into legal territory on the share of fatal crashes involving alcohol-impaired drivers. First-stage analysis reveals that physical dispensary access does not change sharply at borders: within the RDD bandwidth, distance to the nearest dispensary is similar on both sides, likely because prohibition-state residents near borders can easily cross to purchase cannabis. The main specification yields a point estimate of 9.2 percentage points (SE = 5.9 pp), but the 95% confidence interval includes zero (p = 0.127). Border-by-border analysis and distance-to-dispensary regressions with wild cluster bootstrap also produce null results; donut RDD results are sensitive to specification, with a significant positive estimate for small donuts but null effects for larger exclusion radii. These findings suggest no robust substitution effect on alcohol involvement among fatal crashes at legal-prohibition borders during 2016–2019, though the weak first stage and specification sensitivity complicate interpretation.