| Summary: | <p> Aggressive driving behavior is the cause of a large percentage of accidents and fatalities, and it is growing every year. In several cases some drivers perceive their driving as non-aggressive when in fact they drive aggressively. To investigate factors affecting perceived (self-reported) and observed (based on the data from a driving simulation experiment) aggressive driving behavior, four fixed effect bivariate ordered probit models for three categories of aggressive driving behavior (i.e., observed and perceived non-aggressive, somewhat aggressive and very aggressive driving) are estimated. The models simultaneously account for panel data effects and cross equation error correlation. To further address unobserved heterogeneity, six grouped random parameter bivariate probit models for two outcomes (observed and perceived non-aggressive and aggressive driving) are estimated. Each model type is estimated using different barriers as driving behavior separators (either physical barriers in the distribution, or basic statistical measures). The results show that different socio-demographic characteristics, driving experience and exposure, and behavioral information of the participants affect the observed and the perceived aggressive driving behavior. The proposed approach, as a whole, provides an incremental step towards better understanding the different factors that affect the observed and the perceived aggressive driving behavior.</p>
|
|---|
