Driver frustration, a key negative emotion stemming from disappointment, has been linked to aggressive driving behavior, road rage, anger, and unsafe driving maneuvers. Frustration in traffic is often caused by several situational sources, such as congestion, blocked travel paths by other vehicles, frequent honking, and sporadic movement patterns of vehicles. This work aims to investigate how different driving behavior factors affect drivers’ perceived frustration levels in complex traffic scenarios while accounting for unobserved heterogeneity by conducting a questionnaire-based survey. In total, 428 valid responses were collected, considering information on driving behaviors and perceived frustration levels for four complex traffic scenarios, rated on a five-point intensity scale. Then, partial least squares structural equation modeling (PLS–SEM)-based factor analysis was employed to identify the underlying latent factors, followed by the development of a mixed-effect ordered logit model (OLM) and marginal effect analysis to fulfill the study objectives. The factor analysis identified seven driving behavior-related latent factors, which were considered as the predictors in the OLM model. The mixed-effect OLM showed better results for pseudo- R 2 and other statistical powers compared with the fixed-effect model in all scenarios. The model outcomes identified the positive and significant ( p -value < 0.1) influence of all driving behavior factors on higher frustration level perception for different scenarios. Drivers over 64 years old were the least likely to perceive a situation as highly frustrating, and female and occasional drivers reported higher frustration levels than their counterparts. These outcomes were further supported by the marginal effect analysis for each latent and sociodemographic feature. Overall, this work adopted a novel methodological approach for assessing the influence of individual perception-based driving behavior factors on their perceived level of frustration in a complex traffic scenario, incorporating heterogeneity in an ordered logit model. In addition, this work explains the complexity of calculating marginal effects in random parameter models and presents a detailed mathematical expression for each ordered level, offering valuable insights for both researchers and academicians.
Dey et al. (Tue,) studied this question.