Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

Subjective well-being (SWB) or what is also commonly labelled as happiness refers to people’s own evaluation of the quality of their life along both cognitive and affective dimensions (Andrews & Withey, 1976; Diener, Scollon, & Lucas, 2003). Research on SWB has resulted in the emergence of new fields such as hedonic psychology (Kahneman, Diener, & Schwarz, 1999), positive psychology (Seligman, 2002), and happiness economics (Bruni & Porta, 2007; Frey & Stutzer, 2002). A great number of surveys have been conducted to measure SWB, and governments in several countries (e.g. Bhutan, France, United Kingdom) have included or are including well-being as an important indicator of social progress and as a measure for guiding public policy.

Generally, satisfaction with different life domains, such as work, health, marriage, and leisure, contributes to overall SWB. On a daily level, the ability to participate in activities is crucial for SWB (Cantor & Sanderson, 1999) because it helps people satisfy their needs (Chapin, 1974; Maslow, 1970; Oishi, Diener, Lucas, & Suh, 1999). Participation in activities requires travel. The extent to which the transportation system facilitates access to activities should therefore be an important factor affecting SWB.

In recent years, the general interest in SWB research and its connection with travel, activities, and needs have motivated interest in measuring satisfaction with travel. Most of the work in this area has focused on understanding the causes and correlates of travel satisfaction or well-being (especially for commutes) and on estimating the effects on overall SWB. A smaller number of studies have modelled the relationship between well-being and travel behaviour or addressed the implications of the well-being approach for transport policy.

The purpose of this chapter is to review the application of SWB to transportation. In particular, we focus on relating SWB to travel attributes and behaviour for transportation planning and forecasting purposes. The chapter is not meant to be a comprehensive review of the SWB literature in transportation (see, for example, the reviews by Abou-Zeid, 2009; Ettema, Gärling, Olsson, & Friman, 2010). The next section covers measurement of SWB in transportation and major empirical findings relating travel to overall SWB and the dynamics of travel well-being. The following section discusses SWB in relation to travel choice models, first providing a review of studies linking SWB to activity/travel choice and attributes, and then presenting an extended random utility modelling framework that incorporates well-being measures as indicators of utility. The last section concludes.

Measurement of Subjective Well-Being Related to Travel

Empirical evidence from the measurement of SWB related to activities and travel has supported some of the findings of the general SWB research. In this section we discuss two of the main insights and the evidence supporting them. The first insight is that overall well-being or life satisfaction is influenced by satisfaction in various life domains, and travel seems to play a significant role for overall well-being. The second is that the dynamic nature of well-being is also evident in travel and has implications for the measurement of activity and travel well-being.

Transportation and Overall Subjective Well-Being

Travel affects people’s overall SWB primarily by facilitating access to activities in various life domains (work, leisure, family life) and satisfying the corresponding needs. The activities and travel are shaped by the activity and transportation systems (spatial configuration, opportunities) available to people. There may also be direct effects of travel on overall SWB if it generates psychological benefits (e.g. creating a sense of freedom in movement or providing private time) or is associated with health impacts (e.g. benefits of active travel, stress due to commuting). For further discussion of the relationship between travel and well-being, see Delbosc (2012) and Ettema et al. (2010).

SWB researchers have measured activity and travel well-being using instruments such as the Day Reconstruction method (Kahneman & Krueger, 2006; Kahneman, Krueger, Schkade, Schwarz, & Stone, 2004) and found significant variation of happiness by activity type including activities such as commuting. Furthermore, a number of studies have found evidence for the role of transportation in overall well-being. Jakobsson Bergstad et al. (2012) measured several dimensions of affect associated with activities finding a significant influence of affect experienced while performing routine out-of-home activities (which require travel) on overall weekly mood and life satisfaction, with a larger influence on mood. Similarly, using a Canadian time-use survey, Spinney, Scott, and Newbold (2009) found significant correlation between the daily exposure to different types of out-of-home activities and quality of life for elderly non-working Canadians. Duarte et al. (2010) assessed the direct influence of travel happiness on overall happiness. They found that happiness related to work trips and leisure trips positively influence overall happiness, albeit to a smaller extent than the influence of life domains such as family, social, and financial.

The main conclusion from these studies is that travel plays a vital role for overall SWB. It would be useful to gain further evidence on the separate influences of access and travel as this would have direct implications for the design of transport policies aimed at increasing overall SWB. For example, one important question is to what extent transport policies should focus on travel time savings versus improving access to closer and attractive destinations.

Dynamics of Activity/Travel Well-Being

Dolan and White (2006) argue that SWB is a temporal and iterative process involving several stages such as anticipation/planning of a behaviour, actual behaviour, the experience itself, and evaluation of the experience or behaviour, as shown schematically in Fig. 1. This process can be tapped at different points in time, resulting in different measures.

Fig. 1
figure 1

The dynamic nature of the relation of SWB to behaviour (After Dolan & White, 2006)

Full size image

Two studies lend further support to the dynamic nature of SWB measures in transportation. Abou-Zeid, Witter, Bierlaire, Kaufmann, and Ben-Akiva (2012) found that satisfaction with the commute by car of a sample of habitual car drivers was different when measured close to the moment of decision making about mode choice than when measured under routine conditions of daily commuting involving no updating of choices. This difference was attributed to stronger cognitive awareness at the moment of decision making and to possible shifts in the frame of reference when evaluating satisfaction. Pedersen, Friman, and Kristensen (2011) measured predicted, experienced, and remembered satisfaction with public transport for a sample of habitual car drivers and found that experienced satisfaction was significantly different from predicted and remembered satisfaction, similar to other results in the general SWB literature (Kahneman, Fredrickson, Schreiber, & Redelmeier, 1993; Wirtz, Kruger, Scollon, & Diener, 2003).

The main conclusion from these studies is that the time at which SWB is measured matters, and that different indicators can be used for different purposes. We will return to this point when we present a modelling framework that incorporates different types of well-being indicators.

Subjective Well-Being and Travel Choice Models

In this section, we specifically focus on transportation research that includes modelling SWB as a function of activity/travel attributes or linking it to activity/travel behaviour. The section first reviews the main findings of several studies in this area and then presents a general modelling framework for including SWB indicators in random utility models with an application to transportation.

Review of Studies Linking Subjective Well-Being to Activity/Travel Attributes and Choices

With respect to activity participation and SWB, a number of studies have found that activity happiness varies significantly by activity type and socio-economic group and is correlated with behaviour. For example, using structural equation modelling, Abou-Zeid and Ben-Akiva (2012) found significant correlations between activity happiness/travel satisfaction and the propensity to participate in activities as measured by weekly activity frequency for different types of activities: the greater the happiness derived from an activity and the satisfaction with travel to the activity, the greater the propensity to participate in the activity. Using multivariate ordinal probit models, Archer, Paleti, Konduri, Pendyala, and Bhat (2013) found that affective feelings (happiness, stress, meaningfulness, pain, tiredness, sadness) associated with different activity type-location combinations (in-home vs. out-of-home) are significantly influenced by activity attributes such as activity duration, start time, and child accompaniment.

Studies focused purely on travel well-being have also identified links between travel happiness and behaviour. For example, Duarte, Garcia, Limão, and Polydoropoulou (2009) found that experienced happiness and expected happiness (represented through cartoons depicting the travel environment) were significant attributes in models of travel mode choice. Moreover, travel well-being studies, especially those focusing on commute stress and commute satisfaction, have identified a number of modal attributes affecting these components of well-being, including travel time, cost, distance, congestion, variability/predictability of travel time, crowding, frequency of negative critical incidents, perceived control and effort, degree of arousal/boredom, symbolic and affective factors such as the perception of the car as providing independence and control, and activities conducted during travel as a coping mechanism to reduce stress (see, e.g., Koslowsky, Kluger, & Reich, 1995, for the commuting stress literature, and Abou-Zeid & Ben-Akiva, 2011; Ettema et al., 2011; Friman, Edvardsson, & Gärling, 2001; Friman & Gärling, 2001; Olsson, Gärling, Ettema, Friman, & Fujii, 2012, for commute/travel satisfaction related findings, and Ory & Mokhtarian, 2005, for a travel liking study).

Modelling Framework

Given the connection between well-being and behaviour established in both the general SWB literature and the transportation literature, our aim in this section is to provide a framework for systematically including SWB indicators in travel behaviour models and particularly as part of the widely used random utility model. The key concept is the relationship between happiness or well-being and utility. It is shown that the standard random utility modelling framework can be extended with SWB indicators in both cross-sectional and dynamic contexts. This is followed by an application of the extended framework to travel mode choice.

Happiness and Utility

McFadden (2005) summarizes the history of the study and measurement of well-being and its relationship to utility in classical and neoclassical economics and in the modern behavioural re-evaluation of the consumer. In the classical era, Bentham (1789/1948) defined utility as the experiences of pleasure and pain. Utility was related to the process, experience or sensation attached to actions rather than to their consequences. In the neoclassical era, economists viewed utility as a “black box whose inner workings were not their concern” arguing that preferences can only be inferred from choices.

In the modern behavioural re-evaluation of consumer theory, in particular Kahneman (2000) and Kahneman, Wakker, and Sarin (1997) have made significant contributions to the revival of discussions about the relationship between happiness and utility. They distinguished between experienced utility (as in Bentham’s conceptualization) and decision utility (as used by neoclassical economists). Furthermore, experienced utility can refer to remembered utility (retrospective judgement of an experience), moment utility (real-time affective experience), and predicted utility (anticipated experience). A number of studies have found that remembered utility affects decision utility in the sense that people tend to repeat experiences that are remembered more favourably.

From this conceptualization of utility and happiness, it may be concluded that happiness or SWB and utility are the same concept, but a distinction needs to be made among the different notions of utility. Consequently, one can use SWB measures as indicators of utility in random utility models, but different measures of well-being collected at different points in time may reflect different notions of utility. In the following sections, after reviewing the standard random utility model framework, we present a framework and its application that show the enrichment of random utility models with happiness data.

The Standard Random Utility Model

The discrete choice model based on random utility theory has been widely used to model travel-related decisions such as car ownership, activity participation, destination, and travel mode choice. The standard random utility modelling framework is shown in Fig. 2 (Ben-Akiva & Lerman, 1985; McFadden, 1984). In this figure and subsequent figures, solid arrows represent structural relationships while dashed arrows represent measurement relationships. Variables in rectangles are observed, while those in ellipses are latent or unobserved.

Fig. 2
figure 2

Standard discrete choice framework based on the random utility model

Full size image

The utility of every alternative is a function of measurable attributes of the alternative and random factors that are not observed. Equation (1) expresses this relationship, where \( {U_{in }} \) denotes the utility of alternative \( i \) for individual \( n \), \( {X_{in }} \) is a vector of attributes of alternative \( i \) for individual \( n \) (including interactions with characteristics of individual \( n \)), \( \beta \) is a vector of parameters, \( {\varepsilon_{in }} \) is a disturbance associated with alternative \( i \) and individual \( n \), and \( U(.) \) is a function

$$ {U_{in }}=U\left( {{X_{in }};\beta, {\varepsilon_{in }}} \right) ,\quad \forall i $$
(1)

Utility is inferred from observed choices and is used to explain these choices. That is, decision protocols based on utility maximization assume that the alternative that is chosen has the maximum utility among the alternatives in the choice set. This is reflected in Eq. (2), where \( {y_{in }} \) is a choice indicator equal to 1 if alternative \( i \) is chosen by individual \( n \) and is 0 otherwise, and \( {C_n} \) is the choice set of individual \( n \)

$$ {y_{in }}=\left\{ {\begin{array}{lll} 1 & {\mathrm{ if}\;{U_{in }}\geq {U_{jn }}} & {\forall j\in {C_n}} \\0 & {\mathrm{ otherwise}} & {} \\\end{array}} \right.,\quad \forall i $$
(2)

Choice models based on random utility theory have been criticized for their inadequate representation of the process and context of decision making (see, for example, Ben-Akiva et al., 2012). A number of developments have taken place to address these limitations, including the incorporation of attitudes and perceptions in choice models through the Hybrid Choice model (Ben-Akiva, McFadden et al., 2002a; Ben-Akiva, Walker et al., 2002b; Walker & Ben-Akiva, 2002), the use of non-expected utility theories such as prospect theory (Kahneman & Tversky, 1979), and models of social interactions in choice processes (e.g. Brock & Durlauf, 2001; de Palma, Picard, & Ziegelmeyer, 2011).

Another direction for enhancing random utility models is through the use of well-being data. In particular, the choice indicator may be an inadequate measure of the utility on its own. Well-being indicators may also capture information about the utility and can be used as additional indicators of the utility to enhance its measurement (Abou-Zeid & Ben-Akiva, 2010).

Random Utility Model with Subjective Well-Being Indicators: Cross-Sectional Framework

In a static or cross-sectional context, the framework of the standard random utility model shown in Fig. 2 can be extended as shown in Fig. 3, where both the choice and the well-being measures are indicators of the utility. Mathematically, the use of well-being measures adds equations of the following form to the standard choice model

$$ {h_{in }}=h\left( {{U_{in }};{\upsilon_{in }}} \right) ,\ \mathrm{ for}\ i\ \mathrm{ such}\ \mathrm{ that}\ {y_{in }}=1 $$
(3)

where \( {h_{in }} \) is an indicator of happiness or satisfaction with alternative \( i \) for individual \( n \), \( {\upsilon_{in }} \) is a measurement error, and \( h(.) \) is a function. Since happiness indicators would generally be collected for the chosen alternative only, the above equation applies to this alternative.

Fig. 3
figure 3

Discrete choice framework with well-being indicators in a static context (After Abou-Zeid & Ben-Akiva, 2012)

Full size image
Fig. 4
figure 4

Discrete choice framework with well-being indicators in a dynamic context (After Abou-Zeid & Ben-Akiva, 2010)

Full size image

When measuring well-being in a cross-sectional context, one issue that arises is that a happiness judgement by a respondent after the choice had been made is an indicator of remembered utility while the choice is an indicator of decision utility. The happiness measure is thus an imperfect indicator of decision utility in a cross-sectional context. In Abou-Zeid and Ben-Akiva (2012), a method is suggested for addressing this issue by collecting indicators of how different the experience is from expectations or plans and using the well-being indicators only if the experience is as expected or as planned.

Random Utility Model with Subjective Well-Being Indicators: Dynamic Framework

In a dynamic context, one can represent the different notions of utility, their interactions, and the use of well-being measures as indicators of these utilities. Such a dynamic modelling framework is shown in Fig. 4 for a given time period and is analogous to the dynamic SWB process depicted in Fig. 1. Each of the three types of utility may have its own well-being indicators which would be expressed as a function of the corresponding type of utility in a manner similar to Eq. (3). The well-being indicator of decision utility may be obtained at the time of decision making; the well-being indicator of moment utility may be obtained during the experience of the outcome of the decision; the well-being indicator of remembered utility may be obtained retrospectively. Remembered utility in one time period affects decision utility in the following time period.

Fig. 5
figure 5

Modelling framework and parameter estimates from a public transport experiment (t-statistics are shown in parentheses) (After from Abou-Zeid, 2009)

Full size image

An Application of the Extended Model

A simplified version of the dynamic framework shown in Fig. 4 was applied to data from an experiment (Abou-Zeid, 2009). A sample of habitual car drivers commuted with public transport (PT) for a few days and then had to make a choice of whether to continue using their car or to switch to PT. The following self-reported indicators were collected: pre-treatment satisfaction or happiness with the commute by car, post-treatment satisfaction with the car commute, post-treatment satisfaction with the PT commute, and the post-treatment car versus PT choice. Out of 67 participants, 20 participants cancelled their full-time parking permits post-treatment and switched to PT.

The travel mode choice model framework and estimation results are shown in Fig. 5, where curved arrows represent correlations. The framework distinguishes between decision utility (of car and PT) and remembered utility (of car) and uses different indicators of commute satisfaction to capture these notions of utility. In the pre-treatment period, the car is the habitual choice so its utility represents a remembered utility. In the post-treatment period, participants have to make a choice of whether to continue commuting by car or to switch to PT. The car and PT utilities at the moment of decision making are then decision utilities. The utilities are affected by explanatory variables including travel time and travel cost divided by income, and they are correlated with each other. With regard to the indicators, the choice of car or PT is an indicator of post-treatment or decision utility. The pre-treatment car satisfaction measure is an indicator of pre-treatment (or remembered) car utility. The post-treatment car and PT satisfaction measures are indicators of post-treatment (or decision) car and PT utilities, respectively. The detailed model formulation is available in Abou-Zeid (2009).

Among the main findings, the effect of the different utilities on the happiness indicators was positive and statistically significant, implying that the satisfaction or happiness measures are valid indicators of utility. Moreover, commuting cost was found to affect only the mode choice decision but not the satisfaction ratings. This can also be interpreted to mean that satisfaction ratings reflect the actual affective experience (time, comfort, convenience) that does not include monetary aspects.

The performance of the model with happiness indicators (extended model) was compared to that of a model of travel mode choice without happiness indicators (standard model) using several criteria. While the standard model had better goodness-of-fit, the extended model was more efficient, that is it resulted in more accurate parameter estimates. As shown in Table 1, the difference of the variance-covariance matrices of the parameter estimates (standard minus extended) is positive-definite; the variance of the systematic utility of car and PT, computed as the average variance across the sample, is substantially smaller in the extended model than in the standard model. Thus, a main benefit in using well-being measures as indicators of utility seems to be a substantial gain in efficiency.

Table 1 Efficiency of extended and standard random utility model; \( \hat{\beta} \) denotes the parameter estimates and \( \hat{V} \) denotes the fitted systematic utility component
Full size table

Conclusions

This chapter provided an overview of subjective well-being (SWB) research in the transportation field focusing on measurement and modelling efforts as they relate activity/travel well-being to travel attributes and behaviour and to SWB. It is concluded that SWB is highly relevant to activity and travel behaviour and provides a useful indicator of this behaviour. This chapter also presented the extension of random utility choice models with SWB indicators added as utility indicators with an application to travel mode choice. This extension of the standard random utility model to include SWB is part of a broader agenda for the extension of this model to better represent the process and context of decision making that would result in richer behavioural models (Ben-Akiva et al., 2012).