This paper seeks to explore possible variations in the factors that play a role in injury severity outcomes between helmet and non-helmet user motorcyclists. Both the decision of a motorcyclist to wear a helmet and the motorcyclist’s driving behavior can be largely affected by how the motorcyclist perceives risk. As such, unobserved heterogeneity is introduced in the causal mechanism of injury-severity outcomes. To account for unobserved heterogeneity and for the fact that different factors can affect the helmet and non-helmet user motorcyclist groups, mixed logit models of injury-severity likelihood are estimated. Through the use of likelihood ratio tests, the results show that separate models are warranted, as different sets of factors affect the injury-severity outcomes of helmet and non-helmet user motorcyclists.
This paper explores a possible alternative to the traditional approach of identifying safety countermeasures and forecasting high-crash locations, by studying directly the likelihood of a location having an excessive amount of crashes. To that end, a random parameters ordered probit model of highway segment crash-level likelihood is estimated. The crash level for every highway segment (i.e., low-crash, average-crash, and high-crash) is defined based on the segment’s excess-over-the-norm crash frequency and using scientific judgment and physical data barriers evident in the data distribution. Using five-year crash data from rural highways in Indiana, the results show that a number of road geometrics, pavement condition and traffic characteristics affect the highway segment crash-level likelihood-crash locations.
This paper uses driving simulation data and surveys conducted in the Spring of 2014 at Buffalo, NY, to study the factors that affect perceived (self reported, based on surveys) and actual (as measured, based on driving simulation experiments) aggressive driving behavior. To that end, a fixed effects bivariate ordered probit model is estimated. The model simultaneously accounts for panel data effects, and for cross equation error correlation. The results show that a number of sociodemographic (age, race, level of education, household income level, and growing up in a suburban or rural area), driving experience and exposure (accident history, driver experience, and willingness to drive), and behavioral and other characteristics (speeding habits, listening to music while driving, caffeine usage, and fatigue) affect how drivers perceive their driving behavior, and how they actually drive.