Use of Open Data to Assess Cyclist Safety in London

Abstract

This study develops a predictive model for cycling collisions in London. Specifically, the effects of bus lanes, parking or loading facilities, and multilane roads on the risk of cycling collisions are considered. To the best of the authors’ knowledge, this is the first such predictive collision model that develops covariates to measure the characteristics of different types of road infrastructure within zones. A kernel density estimator is used to identify 90 collision hotspots. Each hotspot is populated with information regarding the highway infrastructure within it. A multiple linear regression model tests for the statistical significance of the infrastructure variables. Bus lanes, multilane roads, and 30-mph speed limits are found to affect cycle collision counts, whereas junction density has the largest impact on collision density. Speed limits of 20 mph affect collision counts to a lesser degree than 30 mph, indicating potential safety improvement from reducing speed limits. One-way roads are found to reduce the risk of collisions, along with the provision of priority junctions. This infers that other junction types, such as roundabouts and signalized junctions, present a higher collision risk. The models produce conflicting results on parking or loading provision. The models are expanded to include sociodemographic variables, such as population and employment. The combined model offers no performance improvement over the infrastructure-only model, although a potential link between public transport provision and reducing cycle collisions warrants further investigation.