Over the past six years red light running caused more than 24,000 injuries in Virginia, prompting widespread interest in photored enforcement. Because previous evaluations of this technology have yielded conflicting results, the authors hypothesized that implementation may be more suitable at some locations than at others. Accordingly, this paper reports on an evaluation of photo-red's impact on crashes, with attention paid to the complex interactions of geometric and traffic characteristics at 46 sites in a jurisdiction of more than one million people. Analysis of Variance (ANOVA) and Generalized Linear Models (GLM) show that cameras significantly reduce crashes (p = 0.00) and injury crashes (p=0.05) attributable to red light running but increase rear end crashes (p = 0.00). Disconcertingly, the cameras do not reduce total injury crashes but instead show a statistically insignificant increase (p = 0.13). The paper explores two possible explanations. First, statistical results suggest that the interaction of cameras with higher speed limits and amber times substantially in excess of those recommended in national standards cause some sites to be less amenable to red light cameras. Second, certain geometric characteristics-notably the presence of frontage roads and a greater number of through lanes -can be shown to influence the propensity of red light running. The interactions of these geometric and traffic characteristics may be used to pinpoint locations where red light cameras are more likely to have a positive impact (A). For the covering abstract of the conference see E217780.
Samenvatting