Modeling traveler choice behavior using the concepts of relative utility and relative interest
Introduction
During the last two decades, choice models have proven to be very powerful tools for policy analysis and evaluation in transportation research. The multinomial logit (MNL) model, characterized by the Independence of Irrelevant Alternatives (IIA) property, has become the most widely used choice model in transportation. Convincing examples have been put forward however to show that the IIA property is counterintuitive in many real choice situations. Thus, the development of non-IIA choice models has become a major methodological challenge in the study of individual choice behavior since the late 1970s in many disciplines. In transportation, the interest in developing the non-IIA models seems to have faded slightly as a result of the emerging field of activity-based models of travel demand, but recently a renewed interest is visible (e.g., see Wen and Koppelman, 2000).
On the other hand, individual choice behavior usually involves a complex decision-making process. Individuals utilize different heuristics that will keep the information processing demands within the bounds of their limited capacity (Payne, 1976). A wealth of literature in behavioral decision theory and psychology has shown that task complexity and choice environment affect individual choice behavior (Rushton, 1969; Swait and Adamowicz, 2001a). However, most of the existing models in transportation typically do not account for the effect of this kind of context dependence.
This paper therefore suggests an alternative random utility choice model based on the concepts of relative utility and relative interest. The concept of relative utility assumes that utility is meaningful only relative to some reference point(s). It acknowledges the fact that the individual choice behavior is context-dependent. The concept of relative interest stems from multiple-issue group decision-making theory (Coleman, 1973; Gupta, 1989), which argues that actors involved in negotiations are usually more interested in one issue than in another.
The remainder of this paper is organized as follows. First, to position the suggested choice model, Section 2 briefly reviews existing choice models with context dependence and other non-IIA choice models. Following that, Section 3 discusses the relative utility theorem, which provides the theoretic and behavioral basis for the new models. Section 4 then develops equivalents to the widely used MNL and NL models based on the concepts of relative utility and relative interest. These models were estimated in the context of the choice of destination and stop pattern. Section 5 describes the conjoint-based activity diary data for the model estimation as well as the estimation results. Finally, Section 6 concludes this study and discusses some future research.
Section snippets
Review of existing choice models
Traditional random utility maximization theory assumes that individuals choose the alternative with the highest utility, independent of context and learning etc. This is also true for most of the existing models in transportation. However, there is a wealth of literature in psychology and behavioral decision sciences, showing counter-evidence. Simonson and Tversky (1992) argued that context effects are both common and robust, representing the rule rather than the exception in choice behavior.
The concept of relative utility
The concept of relative utility has its roots in the research about income (Stadt et al., 1985). Duesenberry’s (1949) relative income hypothesis is probably the best-known example of a theory that rests on the concept of relative utility. Kapteyn (1977) developed a theory of preference formation, which assumed that utility was completely relative. Before continuing the discussion, it is necessary to define the concept of relative utility. We argue that an individual evaluates an alternative by
A new family of choice models based on the concepts of relative utility and relative interest
Although all of the three classes of non-IIA choice models, reviewed in Section 2, do not have the IIA property, each class of models has its own shortcomings and unresolved problems. Concerning the first class of non-IIA choice models, there is no way to know exactly what kind of specification will really reflect the real structure and/or distribution of error terms. One can only verify an assumed specification based on empirical analysis. Most importantly however, this class of models has no
Data
To examine the effectiveness of the r-MNL and r-NL models, a conjoint-based activity data set (Wang et al., 2000), collected in the Netherlands in 1997 for analyzing the choice of destination and stop pattern, was used. In the experiment, the respondents had to decide where, when, in what sequence and according to what types of home-based tours the activities would be conducted. The attributes assumed to influence the choices of destination and stop pattern are listed in Table 1, Table 2. Two
Conclusions and future research
The MNL and NL models have dominated in transportation for about 20 years. To avoid the IIA property of the MNL model and interdependence among alternatives in the same nest in the NL model, developing non-IIA choice models is still a major methodological challenge in the study of choice behavior. On the other hand, it has not been satisfactorily examined how to tackle the complex choice decision-making process from the perspective of context dependence.
This paper argues and illustrates that
References (43)
A heteroscedastic extreme value model of intercity travel mode choice
Transportation Research
(1995)- et al.
Joint mixed logit models of stated and revealed preferences for alternative-fuel vehicles
Transportation Research
(2000) - et al.
What is the role of consideration sets in choice modeling?
International Journal of Research in Marketing
(1995) Task complexity and contingent processing in decision making: a replication and extension
Organizational Behavior and Human Performance
(1979)- et al.
Context effects and decompositional choice modeling
Papers in Regional Science
(1991) Task complexity and contingent processing in decision making: an information search and protocol analysis
Organizational Behavior and Human Performance
(1976)Choice set generation within the generalized extreme value family of discrete choice models
Transportation Research
(2001)- et al.
Choice environment, market complexity, and consumer behavior: a theoretical and empirical approach for incorporating decision complexity into models of consumer choice
Organizational Behavior and Human Decision Processes
(2001) - et al.
Incorporating random constraints in discrete models of choice set generation
Transportation Research
(1987) - et al.
A stated choice approach to developing multi-faceted models of activity behavior
Transportation Research
(2000)
Estimating availability effects in travel choice modeling: a stated choice approach
Transportation Research Record
Discrete Choice Analysis: Theory and Application to Travel Demand
A context-sensitive model of spatial choice behavior
The Mathematics of Collective Action
Multinomial Probit: The Theory and its Applications to Demand Forecasting
Rational choice under an imperfect ability to choose
American Economic Review
Consumer preference for a no-choice option
Journal of Consumer Research
Context and task effects on choice deferral
Marketing Letters
Income, Saving and The Theory of Consumer Behavior
The dogit model
Transportation Research
Cited by (52)
Questioning the implication of the utility-maximization assumption for the estimation of deprivation cost functions after disasters
2022, International Journal of Production EconomicsCitation Excerpt :Decision-makers have a repertoire of methods or heuristics to help them identify the preferences and develop their beliefs. In the field of transportation, as well as in urban planning, there is growing interest in discrete choice models that represent different choice behaviors (Zhang et al., 2004; Arentze and Timmermans, 2007; Hess et al., 2008), and the importance of allowing for heterogeneity in the choice mechanisms used by individuals (Hensher, 2008; Zhu and Timmermans, 2010). There is a broad repertoire of choice heuristics that can be used by individuals.
Exploring the role of utility-difference threshold in choice behavior: An empirical case study of bus service choice
2020, International Journal of Transportation Science and TechnologyCitation Excerpt :This extension may make the model choice-set dependent. In this sense, the random regret model (Chorus et al., 2008) and the relative utility model (Zhang et al., 2004) may could be served as references. A challenge on this topic is that a multiple choice case makes consumers’ choice behavior much more complicated as they have to compare each pair of alternatives in the choice set.
Modeling social influence using sequential stated adaptation experiments: A study of city trip itinerary choice
2019, Transportation Research Part A: Policy and PracticeCitation Excerpt :Random regret models (e.g., Chorus et al., 2008; van Cranenburgh et al., 2015; Jang et al., 2017b) represent an attempt of capturing the effect of choice set composition by comparing the considered alternative to all foregone alternatives in a choice set. These models can be viewed as a special case of the relative utility model introduced by Zhang et al. (2004). This model states that decision makers define utility relative to a reference, which can be other choice alternatives as in regret models, but also consumption history or choices of social network members.
Residential location and commuting mode choices: Intrahousehold interaction modeling and its implications for energy policy
2019, Transport and Energy Research: A Behavioral PerspectiveSocial networks and joint/solo activity-travel behavior
2014, Transportation Research Part A: Policy and PracticeCitation Excerpt :As mentioned in the previous section, a significant part of our daily activities are performed jointly with others (Srinivasan and Bhat, 2008). While there are a number of studies on modeling household interactions and time allocation (e.g., Zhang et al., 2002, 2004), this literature review focuses on the empirical studies of joint activities, which is a major theme of the present study. Gliebe and Koppelman (2005) examined the influence of household and personal attributes on the household decisions of performing out-of-home activities independently or jointly with household members.
The role of regret minimisation in lifestyle choices affecting the risk of coronary heart disease
2013, Journal of Health Economics