Red light running behaviour of pedestrians at the signalized crosswalks has been one of the major causes of pedestrian-vehicle crashes. Prior studies have identified the personal, environmental and traffic factors that affect the tendency of pedestrian to violate the pedestrian signal. However, it is rare that the safety consequences of red light running behavior of pedestrians are assessed. This paper aims to estimate the risk of pedestrian-vehicle conflicts attributed to the red light running behaviour of pedestrians using a two-stage modeling framework. In the first stage, interference of the decisions between driver and pedestrian at the crosswalks is modeled as a simultaneous two-player game, with which the errors of players’ perceptions are incorporated using the quantal response equilibrium method. Then, the anticipations of pedestrian (to cross) and driver (to yield) in the game are estimated using the expected utility theory. In the second stage, risk of pedestrian-vehicle conflicts is modeled using the bivariate ordered Probit regression method, based on post encroachment time. Results indicate that both the pedestrian (i.e., gender and walking speed) and vehicle (i.e., speed, distance, and vehicle type) characteristics would affect the anticipations of driver and pedestrian, and therefore the risk of potential conflicts. For example, male and fast walking pedestrians have the higher expectations to cross. In contrast, vehicle speed increases with the anticipations of both pedestrian and driver to yield. Additionally, male have a higher risk of more severe conflicts. However, risk of more severe conflicts reduces when the walking speed and vehicle speed increase. Findings are indicative to the implementation of appropriate remedial measures including traffic management and targeted enforcement that can deter against the red light running behaviour of pedestrians. Hence, overall pedestrian safety can be improved in the long run.

行人在信号人行横道上闯红灯行为已成为行人车辆相撞的主要原因之一。先前的研究已经确定了影响行人违反行人信号的倾向的个人、环境和交通因素。然而，很少有人评估行人闯红灯行为的安全后果。本文旨在使用两阶段建模框架来估计归因于行人闯红灯行为的人车冲突风险。在第一阶段，人行横道上驾驶员和行人之间决策的干扰被建模为同时进行的两人博弈，其中使用量子响应平衡方法纳入了玩家感知的错误。然后，游戏中行人（过马路）和司机（让路）的预期是使用预期效用理论估计的。在第二阶段，基于侵占后的时间，使用二元有序 Probit 回归方法对行人车辆冲突的风险进行建模。结果表明，行人（即性别和步行速度）和车辆（即速度、距离和车辆类型）特征都会影响驾驶员和行人的预期，从而影响潜在冲突的风险。例如，男性和快速步行的行人对穿越的期望更高。相比之下，车辆速度随着行人和司机的预期让步而增加。此外，男性发生更严重冲突的风险更高。然而，当步行速度和车辆速度增加时，发生更严重冲突的风险会降低。调查结果表明会采取适当的补救措施，包括交通管理和有针对性的执法，以阻止行人闯红灯行为。因此，从长远来看，整体行人安全可以得到改善。