Answers are grounded in this paper and cite their sources.
Analyzes cyclist crash risk in intersections, suggesting implications for urban planning.
Background: Quantifying cyclist crash risk requires accounting for exposure, but data on cyclist flows are rarely available at the spatial scale of individual intersections and road segments. This paper develops a data-light approach to estimate cyclist flows from publicly available data, combines these estimates with crash data to analyse cyclist crash risk and to test whether the safety-in-numbers relationship holds at a disaggregated scale. Methods: For the ten largest Swiss agglomerations, the OpenStreetMap street network was divided into intersection and link zones. Cyclist exposure was estimated using weighted betweenness centrality, with origin and destination nodes weighted by population and employment, using six routing (resistance) schemes and several distance thresholds. The resulting centralities were scaled by the city-level cycling modal share and calibrated against cyclist counts. Bicycle crashes between 2017 and 2023 were then modelled using negative binomial regression. Estimated cyclist flows, motor vehicle volumes, infrastructure attributes (from OpenStreetMap and object detection on aerial imagery), density controls and city fixed effects were included as covariates. A two-stage residual inclusion estimator, instrumenting cycle lane provision with the expansion of district heating networks, was used to address reverse causality in infrastructure placement. Results: Cyclist flows could be reproduced from publicly available data; behaviourally calibrated route-choice weights did not outperform simpler network-based schemes. Across all specifications, the elasticity of crashes with respect to cyclist flow lay between 0.38 and 0.52---well below one---confirming a safety-in-numbers effect at the level of individual intersections and segments. Motor vehicle volume emerged as the dominant risk factor. Link segments were considerably safer than intersections, with roundabouts the most hazardous. Cycle lanes were positively associated with crashes; correcting for endogenous placement roughly halved this association, though it remained positive and significant. Conclusion: A data-light exposure model based on open network and demographic data offers a scalable method of estimating cyclist flows and analysing crash risk where GPS traces or full demand models are unavailable. The findings confirm safety-in-numbers at the micro level while underscoring motor traffic and intersection complexity as the principal determinants of cyclist crash risk.
No takes yet. Share an insight, caveat, or question.
Papini et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: