Highway Safety Evaluation Methods - a Comparative Analysis
T Datta, K Schattler, S Sharma, Wayne State University, US
Evaluation of highway safety improvement projects and programs is essential to identify the successes and failures of various infrastructure and traffic control related countermeasures which are often implemented to alleviate traffic crashes and injuries.
Traditionally, practitioners use a ?Before & After? (B&A) study to evaluate the effectiveness of safety improvement projects. Researchers however, tend to question the use of a B&A study methods. They argue that in many instances, high crash locations experience fluctuating crash frequencies over time. Thus, the use of one or two years of B&A data may result in a significant change in crash experience which evaluators often attribute to the countermeasures, whereas such changes may be due the ? regression to the mean? phenomenon.
During the past decade, the ?Empirical Bayes? method of evaluation of highway safety countermeasures has become quite popular in North America. This method requires the use of a large number of control sites to make the correction for ?regression-to-the-mean? effect of the study site(s)?s expected ?after? crash experience without treatment.
Some researchers use B&A with comparison groups for effectiveness evaluation, while others use a ?Comparative Parallel? study. There are issues related to all such methodologies.
A study of the effectiveness of several low cost intersection improvement projects have been completed in Michigan and their effectiveness has been tested using several of the noted methods. The results of this study indicate that there are drawbacks of using all the methods, in spite of advocates of some methods stating otherwise.
For example, the use of a simple B&A study where the ?before? crash data shows fluctuating frequencies may lead to an incorrect estimate of the ?after? crash experience without treatment. In such cases, the ?B&A with control? type of methodology will allow an evaluator to capture the true impact of the treatment. The use of a large number of control sites where all the critical independent variables of these sites are not controlled, as used by some researchers in the ?Empirical Bayes? method, may not necessarily yield realistic results. Additionally, the available human resources in state and local highway agencies in North America for performing these safety evaluation activities must be considered in selecting an appropriate evaluation methodology.
Association for European Transport