A thesis submitted in partial fulfillment of the requirement for the degree of Master of Engineering, School of Civil Engineering

Traffic accidents at urban intersections result in a huge cost to society in terms of death, injury, lost productivity, and property damage. In urban areas, more than half of all accidents occurs at intersections. The objective of this study is to develop accident risk models for vehicle-to-bicycle and vehicle-to-pedestrian accidents at four-legged signalized intersections. The Poisson regression model is the natural first choice for modeling accident frequency or risk. A major limitation of this method is that the variance of the dependent variable is constrained to be equal to its mean. But practically sample accident frequencies appear to be over-dispersed and do not follow Poisson distribution. This limitation can be readily overcome by using negative binomial distribution. In this study, negative binomial distribution was used to estimate parameters of models. Each model was divided into two submodels. One is the probability of encountering an obstacle vehicle ( P0 ) model and the other is the probability of pedestrians or bicyclists fail to avoid the collision ( P1 ) model. The estimation results of this study uncovered some important interactions between geometric and traffic-related elements and accident risk. Some estimated parameters were found that tend to increase the accident risk of a leg of an intersection. Other parameters tend to decrease the risk. Parameters that tend to increase the total accident risk of a leg include: increase in average daily bicycle/pedestrian flow, increase in average daily right turning traffic volume, laceration of intersection at Central Business District (CBD), and additional right turning lane. Parameters that tend to decrease total accident risk of a leg include: four-phase traffic control pattern, median width, and prohibition of right turning movement. By altering some parameters of a particular intersection, total accident risk can be reduced significantly. For example: by changing two-phase traffic control to four-phase control, or by prohibiting right turning movement, or by combing both of these changes, vehicle-to-bicycle and vehicle-to-pedestrian accident risk of legs of a particular intersection can be abridged significantly.

Thesis (M.Eng.) - Asian Institute of Technology 2000