This report deals with two questions: (a) How to learn about the safety effect of various measures, and (b) how to go about deciding what research is worthwhile. On both questions new ground is broken. (a) To learn about the safety effect of a measure, statistical machinery have been devised which ensures that unbiased information is extracted efficiently from commonly available data (such as from uncontrolled before-and-after studies), and that the information is accumulated study-by-study. In this manner a formal "learning process" is established. The method is applied to learning about the safety effect of converting intersection control from two-way and all-way stop signs. Four recent data sets are analysed and their results joined. (b) Research on the safety effect of measures derives its value from reducing the cost which comes from implementing useless measures and the cost of not implementing useful ones. The authors show how a dollar value can be attached to a proposed research, how this can serve to guide the design of specific research, and how one can structure an entire research program. Application is to two "case studies." In the first case the authors ask how many vehicles should be in a "fleet study" designed to help Transport Canada decide on the merits of "daytime running lights." It is concluded (after several iterations) that a fleet study is not justified. The second case study is about measures aimed at reducing accidents due to vehicles stopped on freeway shoulders. (Author/publisher)
Abstract