A skewed logistic model of two-unit bicycle-vehicle hit-and-run crashes

Objective: Hit-and-run behavior in crashes is a severe offense worldwide because the identification and emergency rescue of any injured road user is delayed. A motorist’s run from the crash scene is especially serious for a cyclist who would be more prone to be physically injured in a bicycle-vehicle (BV) crash. The objective of this paper is to explore potential risk factors that contribute to the hit-and-run (HR) behavior of a driver after a two-unit BV collision. Methods: The data used in this study are extracted from traffic crash records in the city of Durham, North Carolina in 2007–2014. This study uses the skewed logistic (Scobit) model to account for the skewness of the dependent variable (i.e., HR) in the dataset. Results: The Likelihood ratio test, AIC and BIC results show that the Scobit model is preferred to the standard binary logistic model for modeling a driver’s decision to run from a two-unit BV crash scene. Estimation results indicate that, the driver’s tendency to run from a crash scene without reporting it in Durham increases if the bicyclist is a teenager or an adult, a drunk-driving or a speeding driver is involved, when the crash happens at night (19:00–6:59), on a local street, or when the automobile overtakes the bicycle. HR behavior will decrease if the cyclist is drunk, an SUV is involved, or the bicyclist fails to yield. Conclusions: The findings of this study are important and useful when developing countermeasures to prevent BV-HR crashes and to improve cycling safety.