This paper presents a simultaneous equations model of accident severity by level for freeway sections using five-year accident severity frequency data for 275 multilane freeway segments in the State of Washington. Accident severity is a subject of much interest in the context of freeway safety due to higher speeds of travel on freeways and the desire of transportation professionals to implement measures that could potentially reduce accident severity on such facilities. While past research has presented univariate frequency models of accident severity, this paper attempts to make a substantive contribution to the field by presenting a simultaneous equations model of accident frequency by severity level using a n-dimensional multivariate poisson regression (MVP) modeling methodology that adopts a multivariate normal heterogeneity specification to capture error correlation structures in the model system. Adopting such a modeling methodology is important as there may be unobserved factors related to driver behavior and roadway/traffic/environmental characteristics that influence accident frequencies at multiple severity levels. The paper presents the model formulation and empirical results of model estimation in the context of the Washington State dataset that suggest that error correlations are statistically significant in modeling accident frequency by severity level. These findings point to the need to adopt simultaneous equations modeling approaches such as that presented in this paper for modeling safety phenomena.
Abstract