In this research, we develop and apply an analytic procedure that estimates the amount of traffic congestion (vehicle hours of delay) that is caused by different types of accidents that occur on urban freeways in California. A key feature of this research is the development of a method to separate the non-recurrent delay from any recurrent delay that is present on the road at the time and place of a reported accident, in order to estimate the contribution of non-recurrent delay caused by the specific accident. Our analysis involves a case study of accidents that occurred on freeways in Orange County in 2001. The non-recurrent delay caused by the case study accidents is estimated based on inferred link speeds derived from loop data and a binary integer programming formulation to identify the temporal and spatial region affected by the accident. Computations of non-recurrent delay were successfully performed for 870 accidents that occurred on weekdays throughout the period of March through December 2001 on the six major Orange County non-toll freeways. A statistical model was estimated that describes non-recurrent delay as a function of day of week, time of day, weather, and the observable (e.g., from emergency calls and/or aerial or on-scene observation) characteristics of the accident.