RELATIONSHIP OF ACCIDENT PATTERNS TO TYPE OF INTERSECTION CONTROL

The change in accident patterns accompanying a change in intersection control was investigated. The investigation include a review of previously made studies, an analysis of before and after accident data, and a detailed statistical analysis of a large, specially assembled, nationwide accident data base. Analysis of variance and regression techniques was used to show that the relationship of accident patterns to type of control must be represented by a complex model and that a simple-signal-no-signal division cannot explain changes in accident patterns. A large number of different measures of effectiveness that describe changes in accident patterns were computed and analyzed. Hypothesis testing revealed that, although there was a definite shift in the distribution of accident types, there was no evidence that signalization, by itself, would lead to a significant decrease in net accident-related disutility, especially for traffic signals not warranted by traffic volume. No conclusive evidence was found to justify a general reduction of minimum volume requirements for rural conditions or high-accident locations.