Are young novice driver injury-severity determinants transferable across contexts? Evidence from three international cities

Source abstract

Young novice drivers remain overrepresented in serious-injury crashes worldwide, yet most evidence guiding safety policy is derived from single jurisdictions, limiting understanding of whether injury-severity determinants and related policy insights are transferable across different regulatory and operating contexts. This study examines the contextual stability of young novice driver injury-severity relationships by comparing crashes involving drivers aged 18-25 across three cities with contrasting licensing systems, enforcement regimes and speed environments: London (UK), Philadelphia (USA), and Melbourne (Australia). Using police-reported crash data from 2022-2023, city-specific random parameters logit models with heterogeneity in means are estimated to capture unobserved variation in injury-severity outcomes. Pairwise likelihood-ratio tests are then used to assess the cross-jurisdiction stability of the city-specific model specifications. The tests reject model-level transferability across all city pairs, indicating substantial cross-context instability in the estimated injury-severity relationships. At the variable level, only a small number of determinants show full empirical consistency across all three cities. Female drivers are consistently associated with a lower probability of severe injury, whereas weekend crashes are consistently associated with a higher probability. By contrast, adverse weather and first-point-of-impact variables are significant across jurisdictions but exhibit differences in effect direction, while lighting conditions, speed environments, vehicle age, and traffic-control characteristics show additional city-specific variation. These findings suggest that some injury-severity relationships are more consistent across contexts, whereas others vary substantially across cities and therefore warrant local validation. Overall, the findings caution against assuming that injury-severity evidence and the policy responses informed by that evidence can be transferred directly across jurisdictions without local validation.