In many states, the police officer responding to a traffic accident fills out either a long form accident report if the crash involves an injury or felony, or a short form for a minor crash involving only property damage. Only long form data is maintained in statewide accident databases. This paper examines the different factors that affect crashes at signalized intersections and investigates if incomplete data can contribute to differences in the research results. Information was collected on all crashes, including minor ones, at intersections in several Florida counties. A tree regression analysis was used to relate the expected number of crashes reported on both long and short forms for each type of crash with the characteristics of the intersection. The relative importance of independent variables were determined for total crashes, angle crashes, left-turn crashes, head-on crashes, pedestrian/bicycle crashes, rear-end crashes, right-turn crashes and sideswipe crashes. Variables included factors such as number and types of lanes, speed limits, traffic volumes and presence of a median. Findings show that, when attempting to forecast the number of expected crashes, it is imperative that analyses are performed for each type of collision rather than aggregating crash types. Findings also show that, especially for angle and head-on crashes, collisions reported on the short form are important and should be documented in every crash database.
Abstract