Models of transportation choice with risk: an application to managed lanes

Abstract

A rapidly exploding body of recent literature recognizes the limitations of modeling travel behavior with the assumption that the traveler faces no risk. More recent models incorporate risk and can be based on expected utility theory (EUT). However, violations of key assumptions of EUT have led to the development of non-EUT models of behavior. This study focuses on managed lane (ML) choice and examines differences in the value of travel time under cumulative prospect theory (CPT) specification versus the EUT. Travelers in our sample are more risk-averse when faced with possibly being late than they are with respect to potential savings in travel time in general. Travelers exhibit optimism (underweight the probability) when the chances of facing a longer travel time are high. The CPT models performed better than traditional models and thus, may prove to be an important tool when considering the benefits of MLs.

Keywords:

• Expected utility theory
• cumulative prospect theory
• probability weighting
• risk attitude
• managed lanes

Acknowledgments

The authors would thank the Harris County Toll Road Authority, Houston Transtar, and West Houston Association for their help in spreading the word about the survey. The authors would also like to thank Richard Trey Baker for all the support he provided in hosting the survey and Prem Chand Devarasetty for his help in the survey design and coding in LimeSurvey. We are also indebted to the reviewers, who provided insightful comments; by addressing those, the paper has been significantly improved. Any errors and omissions are the responsibility of the authors.

Notes

¹ We note here that some (e.g. Rasouli and Timmermans 2014) include the RUM as an ‘expected utility’ model because expectations are taken with respect to the random error term, but we use this term to describe quite a different model, i.e. we do not characterize the conventional RUM class of models as consistent with the EUT.

² With pure uncertainty, individuals may have no guidance from science or laws of nature and therefore must form their subjective probabilities on their own.

³ More details about the estimation approach can be found in documentation for NLOGIT 5 in the NLOGIT5 Reference Guide.