Motorcycle safety outcomes on light-segregated corridors: An observational assessment of São Paulo's Blue Lane

Light-segregated motorcycle facilities are expanding in motorcycle-intensive cities, yet their safety effects remain uncertain when corridors are selected because of prior traffic volume and conflict. This study evaluates São Paulo's Blue Lane using a balanced monthly panel of 12,820 normalized named-road identifiers from January 2019 to October 2025, totaling 1,051,240 road-months and 37 treated identifiers. The analysis prioritizes within-treated-road estimates with small-cluster corrections and complements them with citywide two-way fixed-effects benchmarks, event-time and pseudo‑lead diagnostics, Callaway–Sant'Anna and Sun–Abraham estimators, count models, sensitivity analyses, and a System Dynamics model. Within treated roads, the CR2 estimate was imprecise (β = 0.595; SE = 0.494; 95% CI = −0.450 to 1.640; p = 0.246), with an 80% minimum detectable effect of 1.384 motorcycles per road-month, approximately 33% of the treated-road pre-treatment mean; CR3 inference was similarly inconclusive (p = 0.265). Although the citywide TWFE estimate was positive (β = 0.902; p = 0.008), a three-month pseudo‑lead reproduced nearly the entire association (β = 0.889; p = 0.002), a twelve-month pseudo-treatment was also positive (β = 0.513; p = 0.003), and pre-treatment trends were nonparallel (p < 0.001). Heterogeneity-robust estimates were inconclusive, while severity results were specification-sensitive and underpowered. The System Dynamics model shows that lower crash risk per unit of exposure could coexist with unchanged or higher aggregate crashes if motorcycle exposure increased, but it does not establish that this occurred. Overall, the evidence does not identify a post-implementation change attributable to the Blue Lane or demonstrate reduced motorcycle crash outcomes.