Accident prediction models for signalized intersections

Beca has developed a number of Accident Prediction Models (APM) for accidents at signalised intersections. This paper presents and discusses APMs that have been developed for signalised intersections in New Zealand. Flow-only models are now available for reported motor-vehicle, and cyclist and pedestrian versus motor-vehicle accidents. Research has now considered non-flow variables such as intersection geometry and signal phasing and has found that these can be important predictor variables. The models show that for increasing flows the accident rate per vehicle generally decreases, except for rear end motor-vehicle accidents. This effect is largest for accidents involving cyclists where there is a strong 'safety in numbers' effect. Current research into APMs for signals with high approach speeds is also outlined.