Abstract
This paper examines the effect of unreliable transit service on transit user costs with the goal of increasing the accuracy of mode choice models. The concept advanced here is to include explicitly in the formulation of mode choice models the anxiety experienced by passengers when service is unreliable because of late departure or longer-than-expected in-vehicle travel time. This anxiety is modeled as a generalized cost penalty that is added to actual in-vehicle time. The magnitude of the penalty depends on travelers' assessment of the likelihood of arriving on time at their destination. It is believed that this formulation of anxiety is behaviorally representative. To test the effects of the formulation, a simulation model is generated that quantifies the anxiety component of generalized cost for 10,000 travelers with various aversions to risk for travel between station pairs with different observed reliabilities. Results suggest that primarily for risk-averse travelers, but also for other classes, anxiety may constitute a high percentage of total generalized cost, which may explain many travelers' unwillingness to choose transit in cases in which deterministic models suggest that they will. Calibrating a model of this type presents substantial challenges. An approach is introduced that is currently being pursued to gather actual anxiety levels as a function of transit travel reliability. The paper concludes with comments on future research directions.