View article

This paper evaluates from traffic crashes and detailed built environment (BE) data from Porto Alegre (Brazil) the effects of BE’s contributing factors on motorcycles’ crash-injury severities. Differences among contributing factors on crash severity of traffic crashes involving motorcycles from those involving automobiles were compared. Several data sources were analyzed with GIS software tools to capture BE characteristics. Ordered and unordered discrete-choice models were estimated to analyze crash-injury severity. Two model structures were tested:(i) ordinal logit (OL) and (ii) multinomial logit (MNL). Marginal effects were computed to analyze severity level changes when a given one occurs in the contributing factors. Comparison between OL and MNL showed that standard ordered logit models overestimate the marginal effects of most variables, and forces the effect of other variables to be statistically insignificant, especially those related to the built environment, which present the highest impact. Also, MNL models present better overall fit. The results underscore the importance of population density, number of shops and services and bus-stop density on motorcycles’ crash-injury severity in both approaches. Street lighting and intersections density revealed to be key factors in motorcycles’ injury severity level in the MNL model. Results showed that BE characteristics are more related to automobiles’ crash-injury severity than to the severity of crashes involving motorcycles. The identification of these elements contributes to promoting improvements in the built environment to better address the specific safety needs for motorcycles and automobiles …