Causal evaluation of an urban enforcement campaign on e-bike crashes and violations: evidence from generalized linear...

The rapid expansion of electric bicycles (e-bikes) has transformed urban mobility in Chinese cities but has also introduced substantial road safety challenges. Despite widespread enforcement campaigns targeting risky riding behaviors, rigorous causal evidence on their effectiveness remains limited. This study evaluates the impact of Guangzhou's April 2023 e-bike enforcement campaign on crashes and traffic violations using a quasi-experimental time-series design. Monthly administrative data from January 2021 to December 2024 (48 observations) were analyzed, covering 27 months before and 21 months after the intervention. Two complementary causal inference approaches were applied: Generalized Linear Interrupted Time Series (GL-ITS), which estimates structural changes in outcome trajectories, and Bayesian Structural Time Series (BSTS), which constructs probabilistic counterfactual trajectories representing expected outcomes in the absence of intervention. GL‑ITS results indicate substantial reductions in violations but only minimal changes in crash outcomes. Aggregate violations declined by 89.3%, including reductions of -88.8% for risky riding, -90.8% for regulatory violations, and -86.9% for priority parking violations. In contrast, crashes exhibited an immediate level increase of 24.8%, followed by a pronounced attenuation in growth rate from roughly 3% to around 1-1.5% per month, resulting in a small cumulative change of approximately -2.2% relative to the segmented counterfactual. The BSTS analysis produced a larger positive cumulative deviation for crashes (+57.2%), reflecting differences in counterfactual assumptions between deterministic and probabilistic frameworks while still indicating that enforcement moderated, but did not reverse, the upward crash trajectory within the 21‑month evaluation period. Overall, the findings suggest that enforcement campaigns alone are insufficient for sustained safety improvements without complementary Safe System interventions.