Prospect Theoretic Contributions in Understanding Traveller Behaviour: A Review and Some Comments

Abstract

Over the last 15 years we have seen a small but growing interest in Prospect Theory (PT) as an alternative behavioural paradigm within which to represent traveller behaviour. Some elements of PT such as gains and losses have become so popular in travel choice studies that authors increasingly indicate that they are applying PT. In its strictest interpretation, PT has a number of essential elements that must be included if the link with PT can be claimed. This paper reviews recent transportation studies which report an association with PT as a way of gaining a greater appreciation of what is and what is not an application of PT. We set the scene by providing an overview of PT using studies in the fields of psychology and behavioural economics, where PT was originally established and further developed, and then identify travel behaviour studies which satisfy the PT (Original or Cumulative PT) conditions. A number of specific issues are identified to highlight the connection to PT, including empirically estimated prospect theoretic parameters and referencing. Some behavioural limitations of the reviewed transport PT studies are also presented, including the absence of willingness to pay estimates and consideration of unobserved between‐individual heterogeneity.

Acknowledgement

The very helpful comments from three referees are greatly appreciated.

Notes

1. Studies that employed a probabilistic choice rule are considered in this review paper.

2. The corresponding values estimated by Tversky and Kahneman are 0.88 for α, 0.88 for β, and 2.25 for γ.

3. Outcomes can also be ranked from best to worst (see, e.g., Diecidue and Wakker, 2001).

4. Risk averse is where a sure alternative is preferred to a risky alternative (i.e., with multiple possible outcomes) of equals expected value; risk seeking is where a risky alternative is preferred to a sure alternative of equals expected value.

5. Avineri and Prashker (2003) is one of the early studies that used logit models in a prospect‐theoretic application to travel behaviour research (also see De Blaeij and Van Vuuren, 2003).

6. As a static model, PT cannot address feedback‐based decisions with a dynamic process which involves information acquisition and learning (Barron and Erev, 2002). However, this statement is founded on the incorrect assumption that in feedback‐based decisions all people who adhere to PT’s principles frame their choice context comprehensively as probability distributions of outcomes and not myopic by adapting their decision frame between each separate choice in the succession (see van de Kaa, forthcoming for a discussion of this topic).

7. “For some parents arriving early, which we assume to constitute a gain, may imply opportunity costs, because the time that can be spent on the previous activity (presumably paid work) is restricted more severely. Unfortunately, our data did not allow us to identify for which parents and under which conditions this may be the case; this issue should be taken up in future research” (Schwanen and Ettema, 2009, pp. 521–522).

8. The suggestion of Gao et al. (2010) concerns just one example of a range of feasible reference state definitions that might be considered in that very specific travel choice context.

9. Michea and Polak (2006) estimated WTP for mean lateness (i.e., D in their equation, which is linear to utility, and hence the WTP is the ratio between the estimated parameter of mean lateness and the fare parameter). However, in their CPT model, SDE and SDL are powered by θ ₁ and θ ₂, respectively. Hence, in addition to the estimated parameters, WTP would also be influenced by the minutes of arriving earlier/later than the preferred arrival time (SDE/SDL) non‐linearly.

10. We recognize that this is not necessarily the objective of all research, where the interest in employing PT is to better understand travel behaviour dynamics.

11. There are two type of heterogeneity: between and within individuals and both can be observed and unobserved. Relative to between‐individual (or interpersonal) heterogeneity, within‐individual (or intrapersonal) heterogeneity (e.g., intra‐personal differences in behaviour between different contexts and over time) has so far been given less attention. More advanced model structures, such as MMNL, can, if carefully applied, be used to separate between‐ and within‐individual heterogeneity and to obtain a better understanding of within‐personal differences in preferences and behaviour.

12. See, e.g., Extend Prospect Theory by van de Kaa (2008) and Michea and Polak’s (2006) CPT model within an RUM framework.