This study investigates the effects of the surrounding environment on crashes, with a focus on crash severity and at-fault drivers characterized by gender and age. Crashes where a vehicle is the guilty party are investigated. The study adopts two approaches: aggregate and disaggregate. In the aggregate approach, the numbers of crashes, classified in terms of severity (fatalities, injuries, property damages only), and gender and age of the driver (with several age groups covering the 15-100 age span), represent the variables to be investigated, and have been derived for the Central Ohio Region from the multiple files of the crash database of the Ohio Department of Public Safety, over the period 2006-2011. These data are aggregated at the level of Traffic Analysis Zones (TAZ). OLS models are first estimated, but spatial autocorrelation tests point the existence of spatial autocorrelation (SA). Spatial econometrics models are then used to eliminate the SA bias: the Spatial Autoregressive Model (SAR) and the Spatial Error Model (SEM). Subsequent analyses are conducted using the SEM estimates, as the SEM model is successful in completely eliminating spatial autocorrelation.
The aggregate approach uses a large set of explanatory variables classified into six groups: Regional and Locational, Socio-Economic, Land-Use, Public Transit and Traffic Flow, Circulation and Network, and Physical Characteristics. The results show that variables in all these groups have significant impacts on crash severity and frequencies. The disaggregate approach accounts for more variables that influence crash severity, but cannot be captured in the aggregate approach, such as weather conditions, light conditions, road conditions, type of intersection, and type of vehicle. All these variables are directly related to an individual crash. The logit model is used to explain the probability of a Bodily Injury (BI) crash at the crash scene, where the alternative is Property Damage Only (PDO) crash. Because the age of the at-fault driver is a continuous independent variable, it is possible to precisely assess the impact of age, for both male and female drivers. The results of the logit model estimation show that there is a significant relationship between the probability of a BI crash and drivers’ behavior, built environment, driving conditions, and driving situation.