Pseudo experimental study designs are often used in road safety research for the evaluation of countermeasures such as accident black spot treatments. In a traditional pseudo experimental study design, the crash history at treated sites before and after the introduction of the countermeasure being evaluated is compared with parallel changes in crash rates at suitably chosen control sites to give an estimate of the net crash effect of the countermeasure. Comparison with parallel changes in crash rates at the chosen control sites provides an estimate of the effect of all other factors, besides the countermeasure being evaluated, on crash rates at the treated sites. A method for analysing pseudo experimental study designs often used is that proposed by Tanner. A fundamental weakness in Tanner's method, however, is that it makes a number of incorrect assumptions about the distribution of the data, potentially leading to incorrect conclusions. Bruhning and Ernst (1985) have described a method similar in philosophy to that proposed by Tanner that uses the newer and more statistically rigorous loglinear regression approach, a technique that was not available at the time Tanner devised his method. This paper describes the use of the method of Bruhning and Ernst to estimate the crash effects of road safety countermeasures under pseudo experimental study designs. In addition, the use of the method to test for equivalent treatment effects across a range of treated sites is discussed. An extension of the methodology to cover analysis of cost-weighted crash data is also described. Two examples of application of the analysis method are given. The first is in the evaluation of accident black spot treatments in Victoria, illustrating analysis of both crash frequency and crash cost changes due to the program. The second example is in the evaluation of the crash effects of the rationalisation of speed zones in Victoria. (A)
Abstract