Abstract
A model able to predict accidents at intersections was derived from data from 110 Durban intersections. An alternative analysis showed that the use of peak hour volume data instead of 24 traffic volumes did not significantly reduce the ability of the model to predict accidents.