This paper investigates the injury severity of pedestrians considering detailed road user characteristics and alternative model specification using a high-quality Danish road accident data. Such detailed and alternative modeling approach helps to assess the sensitivity of empirical inferences to the choice of these models.
The empirical analysis reveals that detailed road user characteristics such as crime history of drivers and momentary activities of road users at the time of the accident provides an interesting insight in the injury severity analysis. Likewise, the alternative analytical specification of the models reveals that some of the conventionally employed fixed parameters injury severity models could underestimate the effect of some important behavioral attributes of the accidents. For instance, the standard ordered logit model underestimated the marginal effects of some of the variables considered, and forced some important variable effects to be statistically insignificant, while they remain significant predictors in the other relatively flexible models.