The existing research and modeling applications for predicting pedestrians’ decisions to cross on red at signalized intersection most commonly model this as a function of the pedestrian’s characteristics (e.g. age, mobility restrictions), traffic conditions, and signal delay. The authors' field observations suggest that the existing literature is overlooking two key sets of variables. First of all, the physical configuration of the crosswalk should be an important predictor-pedestrians should be less likely to cross on red at major intersections. Second, the authors believe it is not only the signal delay that matters, but also the overall configuration and complexity of the signal. The authors test these hypotheses by collecting video data of pedestrian behavior at signalized intersections in Washington DC, and building a binary logit model to predict the probability of crossing on red. The authors notably find that that certain signal phases, such as protected left turns, are strongly correlated with a higher probability of crossing on red. The authors also find that the length of crosswalks is inversely correlated with the probability of crossing on red. By testing the model’s predictive accuracy under different specifications, the authors find that the design variables (e.g. geometry, types of signal phases) are key to improving the model’s predictive ability. The authors' findings also offer insights for the design of signalized intersections, and shed light on the complexity of the relationship between crossing on red and pedestrian safety, since some characteristics that are associated with a lower incidence of pedestrian crashes are also associated with more crossing on red.
↧