Elsevier

Journal of Transport Geography

Volume 19, Issue 6, November 2011, Pages 1155-1162
Journal of Transport Geography

Modeling experienced accessibility for utility-maximizers and regret-minimizers

https://doi.org/10.1016/j.jtrangeo.2011.02.009Get rights and content

Abstract

This paper argues that there is a discrepancy between what Logsum-measures of accessibility aim to measure (experienced-utility) and what they actually measure (decision-utility). The latter type of utility refers to the evaluation of an alternative with the aim of making a decision, while the former refers to the evaluation of a chosen alternative after the choice has been made. We argue that accessibility should preferably be conceptualized and operationalized in terms of experienced-utility, but that this type of utility is difficult to measure. Motivated by these observations we show, taking the Logsum as a starting point, how its building blocks (parameters estimated from choice patterns) can be used to construct closed-form and easy to compute accessibility-measures that provide an approximation of experienced-utility. We distinguish between decision-making based on utility-maximization and regret-minimization premises. Using a small-scale case-study building on departure time-choice data, we illustrate the working of the developed accessibility-measures and highlight how they differ from the Logsum-approach.

Highlights

► Logsum-based measures of accessibility measure decision-utility, but should ideally measure experienced utility. ► Paper presents alternatives that approximate experienced utility by allowing for preference and evaluation-rule volatility. ► Numerical analyses suggest that the different accessibility measures may imply different land use/transport strategies.

Introduction

Accessibility is “a slippery notion” (Gould, 1969). It should therefore come to no surprise that it has been measured in numerous ways (see Geurs and van Wee (2004) for a relatively recent overview). One of the more popular accessibility-measures – both in- and out-side academia – is the Logsum, which over the years has been used successfully for the appraisal of various land-use/transport policy strategies (e.g., Handy and Niemeier, 1997, Waddell et al., 2007, de Bok, 2009, Geurs et al., 2010). The Logsum has been successfully incorporated in a number of transport model systems, including TRESIS (Hensher et al., 2004) and LUSTRE (e.g., Safirova, 2007).1 It is widely acknowledged (e.g., de Jong et al., 2007, Chorus and Timmermans, 2009) that this popularity arises from the Logsum’s theoretical advantages over more ad hoc accessibility-measures. More specifically, Logsum-measures provide a closed-form expression for accessibility based on a solid foundation in discrete choice theory (Ben-Akiva and Lerman, 1985) and neo-classical consumer surplus theory (Small and Rosen, 1981, McFadden, 1981).

The Logsum2 is defined as the expected maximum utility associated with a traveler’s choice set. The expectation refers to the fact that the analyst only ‘knows’ the traveler’s utilities up to a random error. As such, he or she does not know for sure which alternative will be chosen, and what will be the exact utility associated the chosen alternative. At least implicitly, this definition suggests that the utility a traveler experiences upon executing an alternative from the choice set is measured, which would indeed constitute an intuitive measurement of accessibility. However, the Logsum-measure of accessibility is in fact not based on the utility travelers actually experience, but on the utility that has presumably driven their choice-behavior (in the form of parameters estimated from their choices). In the behavioral economics community, this latter type of utility is generally called decision-utility, while the former type of utility which is referred to as experienced-utility (Kahneman et al., 1997). The implicit assumption underlying the Logsum-notion is that decision utilities (applied by the traveler to arrive at a decision) are the same as experienced utilities (experienced by the traveler during the execution of alternatives). However, clearly these utilities refer to intrinsically different behavioral notions.

It goes without saying that accessibility-measures (or welfare measures in general) should preferably be based on experienced-utility, not decision-utility. However, it is also clear that direct measurements of experienced-utility are very hard to obtain in a sound and internally consistent way; hence most economists’ preference for working with the concept of decision-utility, as it allows them to use choices as a rigorous unit of measurement (e.g., Stigler, 1950). In sum, there appears to be a non-trivial discrepancy between what the Logsum actually measures (decision-utility), and how it is used and interpreted in the context of accessibility appraisal (as a measure of experienced-utility).

This paper aims to show how this discrepancy can be resolved to some extent: we take the Logsum as a starting point, and show how its building blocks (parameters estimated from choice patterns) can be used to construct closed-form and easy to compute accessibility-measures that provide an approximation of experienced-utility. We take a two-step approach: first, in line with a large body of literature from the field of behavioral decision theory (e.g., Payne et al., 1999, Lichtenstein and Slovic, 2006) we assume that the preferences a decision-maker uses to arrive at a decision are likely to differ to some extent from preferences used to assess the performance of a chosen alternative. Second, we allow for the situation where evaluation-rules may differ between the situation where an alternative is chosen and the situation where a chosen alternative is executed. Specifically, we propose a closed-form accessibility-measure that assumes choices may be based on a regret-minimization evaluation-rule instead of a utility-maximization evaluation-rule, while the performance of executed alternatives is evaluated based on a utility-maximization evaluation-rule.3

The choice for considering a regret-based decision-making perspective is based on two arguments: first, there is a large body of literature from various corners of the social sciences supporting the hypothesis that the minimization of anticipated regret is a very important determinant of choice-behavior (e.g., Loomes and Sugden, 1982, Simonson, 1992, Connolly and Reb, 2005). Take for example Coricelli et al. (2005) who, using neuroimaging techniques, show that the area of the human brain that is active when decision-makers experience regret after having made a (poor) choice, is also highly active split seconds before they make a choice. In their words “anticipating regret is a powerful predictor of future choices”. Second, and this is a more pragmatic reason for adopting a regret-based approach, a generic regret-based discrete choice-modeling approach for the analysis of risky as well as riskless choice4 has recently been developed and successfully applied in a variety of travel choice-contexts (Chorus, 2010). This Random Regret Minimization-approach has important formal similarities with conventional utility-based discrete choice-approaches (such as the MNL-model (McFadden, 1974)) and as such can be relatively easily combined with these conventional approaches to form integrative accessibility-measures.

The remainder of this paper is organized as follows: Section 2 discusses the Logsum-measure. Section 3 presents an accessibility-measure that is based on the notion that preferences are volatile to some extent. Section 4 presents an accessibility-measure that assumes a regret-based evaluation-rule at the level of decisions, and a utility-based evaluation-rule at the level of experiences. Section 5 illustrates the working of the developed accessibility-measures using a small-scale case-study. Section 6 presents conclusions as well as recommendations for future research.

Section snippets

The Logsum as a measure of accessibility-benefits

Assume the following choice situation: a decision-maker faces a set of J alternatives, each being described in terms of M attributes xm. Random Utility Theory (McFadden, 1974) postulates that a decision-maker chooses alternative i from the set when its random utility Ui is larger than that of all other alternatives in the set. Random utility consists of a deterministic part Vi and a random error εi, the latter representing the inability of the analyst to faultlessly assess the decision-makers’

Incorporating changes in preferences between choice and experience

A traveler’s preferences may differ between the moment of choice and the moment of execution of the chosen alternative. There may several reasons for this. First, it is widely acknowledged in the behavioral decision-making literature that people often construct their preferences when faced with a choice, rather than simply applying already existing preferences to the choice situation at hand (e.g., Payne et al., 1999, Lichtenstein and Slovic, 2006). As a result, there may be a difference

Incorporating changes in evaluation-rules between choice and experience

This section takes the difference between decision and experience one step further: in addition to the assumption that preferences are assumed to differ between the moment of decision and the moment of experience, we now also allow for the possibility that evaluations preceding decisions may be based on a different rule than evaluations during or following experiences. Although many alternative decision-rules may be applicable, we here focus on the example where the evaluation-rule used for

An illustrative case-study

This section shows how the three accessibility-measures presented in the previous sections (Eqs. (1), (2), (3) respectively) may differ from one another, in the context of estimation results based on departure time-choice data. The emphasis here is on showing that the three measures may lead to different planning decisions, although an attempt is also made to discuss in what ways and to what extent the measures differ, and what are the causes of these differences. However, since it is to be

Conclusion

This paper argues that there is a discrepancy between what Logsum-measures of accessibility aim to measure (experienced-utility) and what they actually measure (decision-utility). The former type of utility refers to the evaluation of a chosen alternative during or after execution, while the latter refers to the evaluation of an alternative during the moment a choice is made. We argue that accessibility should preferably be conceptualized and operationalized in terms of experienced-utility, but

Acknowledgements

The departure time data used for the illustrative case-study has been collected as part of a research for Rijkswaterstaat (DVS), performed by RAND Europe. We thank DVS for allowing us to re-use these data. The first author gratefully acknowledges support from the Netherlands Organization for Scientific Research (NWO), in the form of VENI-Grant 451-10-001. Comments from an anonymous referee have greatly helped improving a previous version of this paper.

References (36)

  • M.E. Ben-Akiva et al.

    The Akaike likelihood ratio index

    Transportation Science

    (1986)
  • Bierlaire, M., 2003. BIOGEME: a free package for the estimation of discrete choice models. In: Proceedings of the 3rd...
  • Bierlaire, M., 2008. An introduction to BIOGEME Version 1.7,...
  • C.G. Chorus

    A new model of random regret minimization

    European Journal of Transport and Infrastructure Research

    (2010)
  • T. Connolly et al.

    Regret in cancer-related decisions

    Health Psychology

    (2005)
  • G. Coricelli et al.

    Regret and its avoidance: a neuroimaging study of choice behaviour

    Nature Neuroscience

    (2005)
  • P. Gould

    Spatial diffusion: Commission on College Geography

    (1969)
  • S.L. Handy et al.

    Measuring accessibility: an exploration of issues and alternatives

    Environment & Planning Part A

    (1997)
  • Cited by (0)

    View full text