Abstract
Based on the microscopic analysis of the vehicle movements before a rear end collision, the occurrence of rear end accidents is considered to have two indispensable premises in this study - one is the encountering of an obstacle vehicle, and the other is that the forthcoming vehicle driver failed to avoid the collision. Rear end accident risk is expressed as the product of the probability of the leading vehicle's deceleration and the probability of the following driver failed to respond effectively. A negative binomial regression model for estimating rear end accident risk by using accident observations is developed, and several factors were found to affect rear end accident risk significantly. (A)