Understanding risky choice behaviour with travel time variability: a review of recent empirical contributions of alternative behavioural theories

ABSTRACT

Stated choice analysis has been extensively applied to understand choice behavior with travel time variability. However, under a risk-neutral and belief-neutral assumption, the conventional approach to travel time variability has a number of behavioral limitations. In this review paper, we illustrate the violation of linear utility maximization under the traditional risk-neutral assumption and the violation of linear probability weighting under the belief-neutral assumption. This paper also reviews recent stated preference studies based on behaviorally superior theories (Expected Utility Theories, Rank-Dependent Utility Theory, Cumulative Prospect Theory) that have accommodated the role of risk attitude and/or belief in understanding risky travel choice behavior, including their behavioral contributions and major advancements over previous studies. We also highlight several common errors in the majority of the existing studies including the interpretation of risk attitude, the specification of nonlinear utility and the t-test for nonlinear probability weighting, and suggest ways to avoid these problems. Several avenues for future research are set out.

KEYWORDS:

• Travel time variability
• risk attitude
• belief
• traveler behavior
• stated choice

Acknowledgments

This study was funded by the Young Talent Support Plan, Xi’an Jiaotong University. We thank the referees for comments that have been helpful in revising the paper.

Disclosure statement

On behalf of both authors, the corresponding author states that there is no conflict of interest.

Notes

1. A few studies used revealed preference (RP) to capture behavioral responses under travel time variability within alternative theoretic frameworks (see e.g. Hu, Sivakumar, and Polak 2012), which are not reviewed in this paper.

2. Risk refers to a circumstance where a decision maker has known probabilities of possible outcomes. Uncertainty refers to a situation where a decision maker is not offered such information, and has to assess the probabilities of potential outcomes with some degree of vagueness associated. Among all reviewed travel time variability studies, the probabilities of different scenarios per choice alternative are provided to the respondents. As such, what has been investigated through this type of experiment with given or designed probabilities is risky choice behavior.

3. For review studies on traditional travel time variability research, please refer to Li, Hensher, and Rose (2010), Carrion and Levinson (2012) and Taylor (2013), among others.

4. Another nonlinear utility specification is constant absolute risk aversion (CARA) which postulates an exponential specification (Blanchard and Fischer 1989).

5. If the simple power form is used for the nonlinear utility specification ($U=x^{\alpha }$), α is the corresponding risk attitude parameter.

6. There is emerging evidence on the positive utility of travel time related to Information and Communication Technology (ICT) use mainly for public transport (see Ben-Elia, Lyons, and Mokhtarian 2018). With the growing availability of smart digital technology that is already used extensively in many travel situations (including hands free when driving), there are positive sources of utility associated with traveling when the travel time is extended although on balance we would suggest that the trip travel time is a negative source of utility, albeit lower than would have been the case prior to the digital offerings.

7. If the possible outcomes are ranked from best to worst, the worst outcome would be always underweighted (i.e. optimism) or overweighted (i.e. pessimism).

8. This is further discussed in Section 5.

9. Incorrect interpretation of risk attitudes has been corrected in this study.

10. A reviewer pointed out two other important topics that relate to ways in which risk has been considered: ‘Vickrey (1969) proposed an arrival time penalty function to capture risk-aversion, which has been used in many studies since then. Second, due to the non-additivity of the standard deviation, the shortest path problem with standard deviation-based utilities is NP-hard (e.g. Khani and Boyles 2015)’.

Additional information

Funding

This work was supported by the Young Talent Support Plan, Xi’an Jiaotong University [0002019060];