The road to happiness: Measuring Dutch car drivers’ satisfaction with travel

Transport Policy 27 (2013) 171–178

Contents lists available at SciVerse ScienceDirect

Transport Policy journal homepage: www.elsevier.com/locate/tranpol

The road to happiness: Measuring Dutch car drivers’ satisfaction with travel b,c ¨ Dick Ettema a,n, Tommy Garling , Lars E. Olsson c, Margareta Friman c, Sjef Moerdijk d a

Utrecht University, PO Box 80115, Utrecht 2508TC, The Netherlands University of Gothenburg, PO Box 500, SE-405 30 G¨ oteborg, Sweden Karlstad University, SE-651 88 Karlstad, Sweden d Centre for Transport and Navigation, PO Box 5044, 2600GA Delft, The Netherlands b c

a r t i c l e i n f o

abstract

Available online 15 March 2013

Recent research suggests that travellers’ anticipated trip utility may differ from the utility they actually experience when making the trip. This implies that it is important to investigate not only the factors underlying trip decision making, but also the actual experience of the trip. To that end, this paper presents an empirical test of the satisfaction with travel scale (STS) that was developed to measure travellers’ satisfaction with travel. STS measures travel satisfaction in terms of two affective (positive activation versus negative de-activation and positive de-activation versus negative activation) and one cognitive dimension. The STS was applied in the Netherlands in a survey of car users. The results suggest that the reliability of the measurement scales is satisfactory to good, and that they are indicative of an overarching concept of travel satisfaction. Regression analyses carried out with the three STS dimensions as dependent variables show that STS is influenced by experienced traffic safety, annoyance with other road users, the trip being tiring, being distracted by billboards, and lack of freedom to choose speed and lane. In addition, travel purpose and personal characteristics play a role. Overall, the findings provide support for the validity of the STS as a tool to measure satisfaction with travel. It is concluded that using tools such as STS may provide relevant insights into how qualitative and design-related factors influence the attractiveness of trips made by car or other travel modes. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Well-being Travel satisfaction Car Road characteristics Traffic

1. Introduction An important aim of transport and traffic policy is to influence people to travel in societally beneficial ways. While some policies aim at promoting a shift away from car use toward more sustainable travel modes (Proost and Dender, 2008), other policies, such as road pricing (Tillema et al., 2010) and reward measures (Ben-Elia and Ettema, 2011) aim at changing car drivers’ decisions about departure times or routes, leading to a reduction of congestion and local pollution. As a consequence, travel behaviour research has placed a strong emphasis on disentangling the factors that influence people’s decisions about behaviour and behavioural change. Much less attention has been given to the issue of how people experience the trips they make as a result of their decisions. It is usually assumed that the factors that influence decision making will, to the same extent, determine how the outcome of a travel choice (a trip) is experienced. In practice, this implies that the utility that can be derived from, for instance, econometric discrete choice models based on observed

n

Corresponding author. Tel.: þ31302532918; fax: þ 31302532037. E-mail addresses: [email protected] (D. Ettema), ¨ [email protected] (T. Garling), [email protected] (L.E. Olsson), [email protected] (M. Friman), [email protected] (S. Moerdijk). 0967-070X/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tranpol.2012.12.006

travel choices is assumed to be identical to the experienced utility during the trip resulting from the choice. Recently, research has been reported that questions this assumption. In general, it has been found that individuals a priori overestimate the emotions (both positive and negative) resulting from changes in their situation. This is because such positive and negative outcomes are emphasized during decision making, whereas the actual experience is affected by many other factors not considered when the decision was made. For instance, in the context of transportation, Pedersen et al. (2011) report that car drivers who voluntarily switch to using public transport evaluate their travel by public transport less negatively than they expected they would. In a similar vein, Schwartz and Xu (2011) and Xu and Schwarz (2009) report that travellers’ general perception of travel often differs from their experience of actual trips. This is because their perception is framed in an (often socially constructed) view of what travel is supposed to be like, whereas the actual trip may be affected by unforeseen circumstances and events that distract from the actual trip itself. From a theoretical point of view, Ettema et al. (2010) (see also Kahneman, 2000) argue that a distinction should be made between different forms of utility. Preceding a trip, individuals have an anticipated utility of the trip, based on previous experiences and information retrieved from others. In a more technical

172

D. Ettema et al. / Transport Policy 27 (2013) 171–178

sense, one could argue that the anticipated utilities associated with different travel options (e.g., modes) determine which choice is made. If measured by estimating formal discrete choice models, the anticipated utility is termed decision utility. The trip itself will usually consist of different stages (e.g., walking to the bus stop, waiting for the bus, sitting in the bus, etc.). Each stage (or even shorter periods within a stage) will result in momentary evaluations, termed momentary utility. After completing the trip, individuals will aggregate their experienced momentary utilities into a remembered utility. This remembered utility will influence their perception of a trip made by a specific mode as well as their anticipated utility before a next trip. To date, investigating the experience of travel has received limited attention compared to the multitude of studies of travel choices?. Yet, we feel that this area is important for the following reasons. If momentary utility influences future choices, it is important to identify factors influencing it, which may differ from the variables conventionally included in travel forecasting models. For instance, qualitative factors, such as personal safety, cleanliness and atmosphere and specific incidents occurring during the trip have been found to influence momentary utility of public transport users (Stradling et al., 2007), but are not included in travel forecasting models. In the context of driving (Novaco and Gonzalez, 2009), traffic flow conditions and road layout have been shown to influence stress levels and pleasantness of driving, whereas behavioural models of for instance route choice only consider travel times. Thus, investigating which objective or subjective factors influence travel experience brings to the fore factors that also influence travel decision making and should be subject of deliberate policy making. In addition, knowledge of the factors influencing travel experience increases our insight into how travel can be made more enjoyable in itself. Given the importance of investigating travel experience, this paper has two objectives. First, methods need to be developed to measure the momentary utility experienced during travel. Over the past years, efforts have been made to develop such approaches. Jakobsson Bergstad et al. (2011) developed a five-item scale to measure satisfaction with travel which focused mainly on cognitive evaluations of the trip. This approach to measuring travel satisfaction assumes that individuals are capable of remembering how they experienced an event (a trip in this case) earlier the same or previous day. As such it builds on the day reconstruction method (DRM; Kahneman et al., 2004) and the event reconstruction method (ERM; Schwarz et al., 2009). A close correspondence has been observed between measures obtained with reconstruction methods and momentary methods of measurement of experiences (experience sampling method; Stone et al., 1999). Jakobsson Bergstad et al. (2011) showed that satisfaction with travel correlated positively with car use and age. In a later study, Ettema et al. (2011) extended the measure, referred to the satisfaction with travel scale (STS), such that it now also includes affective dimensions. Ettema et al. (2011) tested the extended STS (see Section 2 for details) in an experimental setting. The results suggested that it has sufficient internal validity and responds to changes in activity and travel settings (time pressure, travel mode, travel time, and walk time) in the expected way. Thus, measurement tools developed to measure experienced utility appear to give satisfactory results when applied in experimental settings. However, further tests are needed to decide about the applicability of STS to measure experienced utility of travel. A first step is the application of STS to actual trips testing reliability as well as validity by investigating how STS varies across contexts. A second research challenge in the context of measuring experienced utility is to investigate the factors that influence experienced utility. As noted, this is particularly important as the factors that influence experienced utility are not necessarily the

same as included in discrete choice models assessing decision utility. There are indications that the experience of a trip is to a significant extent affected by ‘soft’ factors, such as for instance interaction with passengers, cleanliness, personal safety, use of materials, which are not easily foreseen when deciding about a trip. For instance, Jakobsson (2007) and Steg (2005) report psychological motives for car use, which refer to emotions evoked by driving a car (e.g., feelings of pleasure-to-use and freedom). Apparently, driving affects people’s mood and partly explains why the car is perceived to be attractive and satisfactory to many people (Steg, 2005). It has also been found that symbolic (selfpresentation) aspects significantly contribute to the positive utility of driving (Mokhtarian and Salomon, 2001). Studies of commute stress amongst car commuters (e.g., Novaco and Gonzalez, 2009) indicate that stress is related to affective and cognitive assessment of travel, and that higher levels of stress are associated with higher impedance, less perceived control and less predictability. Obviously, congestion levels will have a large impact on such factors, implying that policies that influence road capacity or travel demand will have an impact on stress and satisfaction with travel. In addition, factors such as road design and traffic information may add to the predictability and perceived control of car drivers. Thus, apart from the aim of the present study to test the extended STS in the context of daily travel behaviour, another aim is to test whether STS can be used to identify which policy-related factors influence car drivers’ satisfaction with travel. The rest of the paper is organized as follows. Section 2 outlines the STS that was investigated in the present study. Section 3 describes the data collection. Section 4 describes the results. Section 5 draws conclusions and discusses further research efforts.

2. Method An important aim is to test the application of STS to actual car trips and investigate the relationship of STS to external factors. The STS applied here is based on methods developed to measure subjective well-being (SWB). SWB is defined as an individual’s cognitive and emotional well-being. According to Diener et al. (1985) SWB consists of two dimensions: cognitive and affective well-being. Cognitive well-being refers to an individual’s assessment of his or her life in general, primarily based on his or her objective life circumstances. It is a judgment of one’s life in terms of how good it is, rather than directly expressing one’s emotions or mood. Still, it cannot be ruled out and it has been empirically demonstrated (Jakobsson Bergstad et al., 2012) that cognitive well-being is in part based on memory for emotional experiences. Cognitive well-being is measured using existing scales such as the satisfaction with life scale (SWLS) (Diener et al., 1985) or a single item scale (World Values Survey, 2005). Affective well-being refers to an individual’s emotional state. It may be measured by immediate self-reports of emotions or mood during execution of an activity or travel. Alternatively, affective well-being may be measured retrospectively. Schwarz et al. (2009) report that results from reconstruction methods, in which respondents recall how they felt during a specified past episode, are highly correlated with immediate reports. With respect to measurement scales, Watson et al. (1988) proposed the positive and negative affect scale (PANAS) to measure affective well-being. With this method, respondents indicate their affective experience by selfreports on a set of positive and negative adjective scales. Another method to measure affective well-being is the Swedish core affect scale (SCAS) (V¨astfj¨all et al., 2002; V¨astfj¨all and G¨arling, 2007). It is assumed in this method that emotions can be decomposed into two underlying dimensions: valence (positive versus negative) and

D. Ettema et al. / Transport Policy 27 (2013) 171–178

activation (versus de-activation). Activation refers to the extent to which in individual is stimulated by cues from his or her environment, whereas de-activation is the state related with the absence of such stimulation. Individuals’ affective states can be derived from scores on both dimensions (the ‘‘affect grid’’; see Diener et al., 1985; Russell, 1980). For instance, a high-valence, activated emotion would be ‘enthusiasm’, whereas a high-valence de-activated emotion would be ‘relaxation’. As shown in Fig. 1, this results in two dimensions obliquely to valence and activation, indicating (1) to what extent someone feels positively activated (e.g., enthusiastic) instead of negatively de-activated (e.g., bored) and (2) to what extent someone feels positively de-activated (e.g., relaxed) instead of negatively activated (e.g., stressed). It is noted that measurements of affective well-being may pertain to time horizons ranging from the immediate moment, days, weeks, or months. If affective states pertain to multi-day periods it is usually referred to as mood. The STS is designed using similar dimensions as SWB. The tenet is that satisfaction with travel can be regarded as SWB pertaining to a specific domain (travel) and should therefore be measured based on similar principles. Hence, we assume that satisfaction with travel involves both cognitive (reasoned) and affective (emotional) dimensions. Consequently, STS consists of sets of items (Table 1) enabling respondents to evaluate their trip on both cognitive and affective dimensions. The six items measuring affect (emotion) related to the trip are based on the SCAS. Each item consists of pairs of adjectives that together represent the oblique dimensions in the affect grid. We define affect in terms of combinations of valence and activation. The first three items consist of pairs of adjectives ranging from negative activation

Fig. 1. Dimensions in the Swedish core affect scale (SCAS).

Table 1 The satisfaction with travel scale: end points of scales. Negative activation—positive de-activation (items 1–3) Hurried Relaxed Worried Confident Stressed Calm Negative de-activation—positive activation (items 4–6) Tired Alert Bored Enthusiastic Fed up Engaged Cognitive evaluation (items 7–9) Travel was worst I can think of Travel was best I can think of Travel was low standard Travel was high standard Travel worked well Travel worked poor

173

(hurried, worried, stressed) to positive deactivation (relaxed, confident, calm). Together these items measure extent of positive de-activation. Items 4–6 consist of pairs of adjectives ranging from negative deactivation (tired, bored, fed up) to positive activation (alert, enthusiastic, engaged). Together, these items measure the extent of positive activation. Finally, items 7 to 9 consist of pairs of descriptions ranging from negative to positive cognitive evaluations of travel focusing on the functionality of travel. Thus, the STS has three items tapping cognitive travel satisfaction, three items tapping positive active activation (extent of engagement, enthusiasm, alertness) and three items tapping positive de-activation(extent of relaxation, confidence, calmness) (see Table 1). For all STS items presented in a counter-balanced order, respondents make ratings on 7-point scales ranging from 3 (minimum) over 0 (neutral) to 3 (maximum). Note that the STS refers to travel in general, and that it therefore may be applied directly to public transport, walking and cycling trips (see Olsson et al., 2012, for an example).

3. Data collection A survey to measure car drivers’ satisfaction with their trip was held in the autumn of 2009 on four highways in the Netherlands. The following highways were selected to vary in terms of congestion and appearance (including landscape):

1. the A12 highway near Veenendaal, which was defined as an ‘average’ highway with two lanes. 2. the A2 highway near ‘s-Hertogenbosch. This is a two-lane highway on which road construction was taking place, potentially leading to congestion and confusion. 3. The A28 highway near Harderwijk. This is a less crowded highway in a low-density area, leading to a more open landscape. 4. The A58 highway near Gilze-Rijen. This is a two-lane highway with a high density. To approach respondents, number plates of drivers on the above roads were manually recorded on Tuesdays and Thursdays between 8.00 and 11.00. Based on the number plates, names and addresses of the car owners were retrieved from the car registration authority. These car owners were contacted by mail within one week, which was deemed not too long afterwards to ask questions pertaining to the specific observed trip. In total 2015 invitations were sent out leading to 256 usable questionnaires (12.7%). The rather low response rate may partly be due to inaccuracies in recording the number plates, such that some people receiving the invitation did not actually drive on the mentioned road. This was checked in the questionnaire, where 17.4% indicating that this was indeed the case. Nevertheless, the sample obtained showed a reasonable distribution across roads, travel purposes and socio-demographics, leading to the conclusion that it serves well for the exploratory purpose of this study. The returned questionnaires were evenly distributed across the highways, 64 on the A12, 48 on the A2, 71 on the A28 end 73 on the A58. In addition to STS, the questionnaires included the questions described in the following. 3.1. Trip characteristics This part included questions about the origin and destination of the trip, specified by zipcode or street and municipality. In addition, questions were asked about the type of locations of origin and destination subsequently used to infer travel purpose. Additional questions included the frequency of the trip, company

174

D. Ettema et al. / Transport Policy 27 (2013) 171–178

during the trip, whether or not activities were undertaken during the trip (talking, telephone call, listening to music) and whether a gasoline station or rest place was visited.

of this study is exploratory and investigates the impact of sociodemographics on STS using multivariate techniques, over- or underrepresentation of certain segments is not deemed a major problem.

3.2. Road conditions Respondents were asked to indicate their subjective evaluation of the roads on specific points. In particular, respondents were asked to indicate their agreement on a 1-5 Likert scale to the following statements:

 That the traffic was crowded (CROWDED)  That their trip was unsafe (UNSAFE)  That they were limited in making decision about speed and     

lane (LIMITED) That they were annoyed by other drivers (ANNOYED) That they were insulted by other drivers (INSULT) That they had problems finding their way (WAYFINDING) That they were distracted by billboards or buildings (DISTRACT) That their trip was tiring (TIRING)

3.3. Personal characteristics To measure the variation in STS dependent on observable differences between individuals, the questionnaire included a series of questions about age, gender household type, education level, driving experience and kilometres driven per year. The answers are summarised in Table 2. The results suggest that the sample is overrepresented in the higher education categories (64% in total vs. 25% for the Dutch population), although statistics on the population level are not available for Dutch highway users. Also, the majority drives over 20,000 km per year which exceeds the national average (8900 km per year). Since the character Table 2 Sample characteristics.

4. Analyses and results 4.1. Reliability In order to test the reliability of the STS, Cronbach’s alphas were calculated for the three underlying dimensions of STS, cognitive evaluation, positive activation and positive de-activation. Cronbach’s alpha (Field, 2009) is based on the variances and covariances of the items supposed to measure each dimension and expresses the extent to which these items in fact vary in a consistent way such that it can be inferred that averaging them provides a reliable measure of a the dimension. Cronbach’s alpha ranges from 0 (totally inconsistent) to 1 (perfectly consistent), with values of 0.70 and higher considered satisfactory. Table 3 shows that reliability of the positive de-activation and cognitive dimension are satisfactory (4.70), whereas reliability of the positive activation dimension is marginally satisfactory (4.60). A confirmatory factor analysis that was carried out on another sample of data (Friman et al., 2012) indicates that all nine measurement scales load significantly on the three underlying dimensions as hypothesized. This confirms that the STS consists of three underlying dimensions and is measured adequately by the measurement scales. It also appears that the positive activation, positive de-activation and cognitive evaluation are significantly correlated (see Table 3). Hence, STS measures three correlated dimensions of satisfaction with travel capturing different aspects. The average item scores (Table 3) indicate that car drivers are positive (40) on all items. All means differ significantly from zero with 99% confidence according to a one sample t-test. Thus, overall, car drivers can on average be considered to be relaxed/ confident/calm, alert/enthusiastic/engaged and have a positive assessment of their trips. 4.2. Influencing factors

Age

%

18–25 26–35 36–45 46–55 56–65 465 Gender Male Female

4.0 12.9 24.0 25.3 25.8 8.0 65.8 34.2

Household type Single Couple without children Couple with children Single parent Living with parents Other

8.9 35.6 47.1 4.4 3.1 0.9

Education level Lower professional Middle professional MAVO/HAVO/VWO (highschool) HBO (higher professional) University

6.7 14.8 14.4 41.7 22.4

Kilometres driven per year o 5,000 5,000–10,000 10,000-20,000 420,000

4.5 13.0 26.0 56.5

In order to investigate influencing factors, regression analyses were carried out in which both overall STS (average across the three dimensions) and the three dimensions were regressed on trip characteristics, driver characteristics and road conditions (Table 4). Since all the three dimensions and the overall STS are correlated (see Table 3), when regressing an underlying dimension on the explanatory variables, one needs to control for these correlations to avoid biased results. This is done by carrying out a hierarchical regression, in which the other dimensions are entered in the first block and the influential factors in the second block to explain the residual variance. These factors include sociodemographics (see Table 2), experience of road conditions (see Section 3), travel purpose, trip frequency and driving habit (kms per year). All variables were entered in the model, but only the significant effects are presented in Table 4. The different regression analyses are discussed in the following. 4.3. Positive de-activation The positive de-activation dimension represents the degree of calm and confidence during the trip. The model suggests that 51% (R-squared¼0.51) of the variance in positive de-activation is explained by the other two dimensions (positive activation, cognitive evaluation of travel). As expected, highly significant positive correlations exist between positive de-activation and the other dimensions, suggesting that the three dimensions together

D. Ettema et al. / Transport Policy 27 (2013) 171–178

175

Table 3 Reliability of measurement scales and correlations between dimensions. Mean and standard deviation of item

STS items 1–3 (positive de-activation) Hurried relaxed Worried confident Stressed calm

1.23 (1.81) 1.00 (1.57) 1.74 (1.71)

STS items 4–6 (positive activation) Tired alert Bored enthusiastic Fed up engaged

1.38 (1.63) 0.51 (1.21) 0.82 (1.81)

STS items 7–9 (cognitive evaluation) Travel was worst/best I can think of Travel was low/high standard Travel worked poor/well

0.92 (1.49) 0.18 (2.08) 0.21 (2.23)

nn

Cronbach’s alpha

Range of inter item correlations

0.811

0.555–0.617

0.643

0.301–0.467

0.796

0.523–0.695

Correlations Positive activation

Cognitive evaluation

0.315nn

0.590nn

0.404nn

: significant at 99% confidence level.

Table 4 Results of regression analyses.

Model 1 Positive de-activation Positive activation Cognitive evaluation R2 Model 2 Positive de-activation Positive activation Cognitive evaluation Purpose¼ recreation A2 motorway Trip frequency 41/week Crowdedness Lack of autonomy Experienced safety Annoyed by road users Fatigue Difficulty in wayfinding Distracted by billboards Male R2

Positive de-activation Relaxed/confident/calm

Positive activation Alert/enthusiast/engaged

Cognitive evaluation Best/high standard/worked well

– 0.647 (0.000) 0.183 (0.001) 0.51

0.376 (0.000) – 0.355 (0.000) 0.62

0.212 (0.001) 0.706 (0.000) – 0.51

– 0.568 (0.000) 0.182 (0.012)

0.371 (0.000) – 0.386 (0.000)

0.145 (0.012) 0.471 (0.000) – 1.341 (0.044)  1.318 (0.005)

 0.689 (0.086)  0.743 (0.000)  0.595 (0.000) 0.526 (0.022)  0.386 (0.074)  0.928 (0.000)

 0.507 (0.011)  0.333 (0.047)

0.774 (0.038) 0.59

0.65

0.456 (0.013)  0.925 (0.005) 0.73

p-value in parentheses.

represent an overarching concept of satisfaction with travel. Adding additional explanatory variables increases the R-squared to 0.59, implying that they together account for another 8% of variance in positive de-activation. The variables explaining positive de-activation mostly include travellers’ evaluations of road conditions. Experienced traffic safety, ease of way finding have a positive influence. The trip being tiring has a negative effect on positive de-activation. In this respect, we assume that someone’s physical state (being tired), which may be caused by a variety of factors, including driving conditions, quality of the car and one’s physical condition, influences someone’s emotional state. On the other hand, we may not preclude the possibility that someone’s mental state (being stressed, worried or hurried) influences the feeling of fatigue. Since the regression models only allow us to prove correlations and not causalities, addressing this issue requires additional research. Of the socio-demographic variables only gender has a significant effect on positive de-activation. Men have significantly higher levels of positive de-activation than women. Marginally significant influences are found for travelling frequency (i.e., more

than once per week). In particular, frequent travellers are found to have relatively lower positive de-activation levels. 4.4. Positive activation In the analysis on positive activation, 62% (R-squared is 0.62) of the variance is explained by positive de-activation and cognitive STS, which are found to be highly significant. Adding other explanatory variables increase R-squared to 0.65. Hence, positive activation is mostly explained by the other two dimensions, and additional variables account for only 3% of the variation in positive activation. Accordingly, the majority of variables does not significantly add to the explanation of positive activation. The only factor influencing positive activation is distraction by billboards and buildings, which has a negative impact on positive activation. 4.5. Cognitive evaluation The affective dimensions (positive activation and de-activation) explain 51% (R-squared is 0.51) of the variance in cognitive

176

D. Ettema et al. / Transport Policy 27 (2013) 171–178

evaluation. Both have highly significant positive impacts. Adding additional explanatory variables increases R-squared to 0.73. The cognitive dimension is more than the other dimensions (22%) explained by road characteristics and socio-demographics. Travel purpose has a significant impact. Travellers having a recreational purpose are found to have a higher evaluation than travellers having other purposes. Previous studies (Jakobsson Bergstad et al., 2012) found that participation in recreational activities is positively related to higher SWB. Together these findings suggest that positive effects on well-being resulting from activity participation carry over to the trip made for that purpose. Evaluations of traffic conditions also have an effect on cognitive evaluation of travel. Crowdedness has a negative effect of cognitive evaluation. Thus, driving conditions are less positively evaluated in case of heavy traffic. Having limited freedom in choosing one’s own speed and lane has a negative impact on cognitive evaluation. Earlier studies have reported that sense of freedom is an important affect associated with driving a car. Therefore, it is likely that lack of freedom is evaluated as negative. Finally, distraction by billboards or buildings positively adds to cognitive evaluation. This suggests that, billboards and buildings provide a scenery that is positively evaluated. Yet, note the negative effect of buildings/billboards on positive de-activation. This suggest that buildings and billboards make drivers less relaxed, but that they still appreciate more lively surroundings. Fatigue associated with driving has negative effect on cognitive evaluation. In addition, driving on one particular highway (the A2) has a negative effect on reported cognitive well-being during driving. The A2 is a highway that was heavily reconstructed during the data collection period. Although one may argue that deteriorated driving conditions due to roadwork are reflected by other variables, such as crowdedness, lack of autonomy and fatigue, the significance of the A2 dummy suggests that the specific circumstances of reconstruction, such as temporary lanes and signs, constitute an additional factor that negatively affects the cognitive evaluation. The effect of the reconstructions (represented by the A2-dummy) was not found for the models explaining positive activation and positive de-activation. This suggests that the specific circumstances of road construction do not cause negative emotions, but are valued negatively at a cognitive level. Finally, of the socio-demographics only gender has a significant impact. Men have a relatively lower cognitive evaluation than women. Again, note that the effect of gender on positive deactivation was the reverse. This suggests that while men evaluate their trip as more relaxed, their cognitive assessment of it is lower.

5. Discussion and policy implications

support for the hypothesis that road characteristics and traffic conditions have an impact on travel satisfaction of car drivers. The results suggest that in particular the degree of positive deactivation (being relaxed or stressed) and the cognitive assessment are affected. The degree of positive activation (being bored or enthusiastic) is much less affected by road design and traffic characteristics. An important observation in the context of systems that monitor travellers’ user satisfaction is that the impact of roaddesign factors differs between the STS dimensions (positive activation, positive de-activation, cognitive evaluation). This suggests that surveys used to monitor travellers’ satisfaction in the context of changes in level-of-service should contain questions on all dimensions in order to secure a valid assessment of changes in travel satisfaction. It also suggests that certain current monitoring systems, which only use cognitive assessment scores (e.g., by asking to grade the transportation system), give an incomplete picture of potential implications of investments. 5.2. Implications for road design Our findings have implications for the design of road infrastructure. Even though some caution should be exerted since we used drivers’ perceptions of road and traffic characteristics as explanatory variables, we conclude that decisions about how roads are designed will influence travel satisfaction of drivers using these roads. The most obvious example would be increasing the capacity of congested roads. This will likely increase perceived traffic safety and freedom, and decrease perceived crowdedness, fatigue and annoyance, which all had an impact on travel satisfaction. Importantly, such positive effects of increased capacity fall outside the scope of current cost-benefit analysis methods, which treat user benefits simply in terms of travel time gains. Apart from aspects related to road capacity and crowdedness, also design aspects such as ease of way-finding and presence of buildings and billboards will influence travel satisfaction. Further research is needed to examine how roads can be designed such as to optimise way-finding and appear attractive to road users on the above dimensions. In the context of broader concerns about the environmental impacts of car traffic, it is important to realise that the potential positive impact of extended road capacity on satisfaction with travel should be balanced against the negative environmental and social effects of additional expansion of road infrastructure, such as pollution, noise and decreased traffic safety. Rather than advocating that car drivers’ STS should be maximised at any cost, we would argue that STS provides a tool to balance drivers’ benefits against potential environmental and social factors. In addition, the combination between design factors and STS may offer clues to the design of highways such that drivers’ satisfaction can be maintained under similar road capacity.

5.1. Validation of STS 5.3. Application of STS for policy support An important objective was to test the application of the satisfaction with travel scale (STS) to the evaluation of actual travel behaviour. The empirical outcomes suggest that STS measures experienced utility in a consistent and intuitively plausible way, as indicated by the reliability scores (Cronbach’s alpha) and the influence of context factors on the satisfaction with travel dimensions. In addition, as Friman et al. (2012) demonstrated by means of confirmatory factor analysis, the positive correlations found between the three STS-dimensions suggest that they together represent an overarching construct of satisfaction with travel. A second aim was to investigate to what extent STS can be used to elicit which characteristics of the trip and the infrastructure explain car drivers’ satisfaction with travel. The study provides

Having observed the potential of STS to assess the impact of road design and traffic characteristics on travel satisfaction, the question can be raised how measures of affective and cognitive satisfaction should be applied to support policy making. In this respect, we see STS results primarily as an important source of information about the actual effect of changes in (in the present study) road infrastructure on individuals’ well-being during travel. STS can be particularly useful to investigate qualitative factors influencing satisfaction with travel, which tend to go unnoticed in travel demand models. In particular, factors such as atmosphere, cleanliness and experienced traffic and personal safety may affect positive and negative affect without directly

D. Ettema et al. / Transport Policy 27 (2013) 171–178

leading to observed behavioural changes. Such knowledge is useful in designing aspects of infrastructure or transportation systems. We suggest that STS is used as a diagnostic tool (i.e., to explore salient factors) rather than as a forecasting or policy evaluation tool. Using STS to forecast changes in travel satisfaction and aggregate those over a population would be difficult given the psychological mechanisms involved in people’s response to changes in their circumstances. For instance, the treadmill effect (Frederick and Loewenstein, 1999) suggests that people become used to improvements or deterioration in their circumstances, implying that effects of improvements in travel conditions on STS are not linear and not lasting. Also, when used for policy evaluations in a similar way as economic cost-benefit measures, STS may be subject to manipulation, for instance by manipulating the order and size of changes in travel circumstances. We refer to Frey and Stutzer (2010) for a detailed discussion of this issue.

5.4. Application issues STS is developed as a generic tool for measuring satisfaction with travel. It can therefore be applied across travel modes and trip purposes using the similar nine scales as in this study (Table 1). However, given that influential factors will differ depending on the travel context (e.g., car or public transport) or the specific policy objectives, specific questions regarding travellers’ perceptions of these factors need to be included in the survey or be objectively measured. It is suggested that these factors are determined via literature search or consultation of experts. Regression models, as in the present study, in which influential factors are used to explain STS, can be used to determine the relative impact of these influential factors and determine the most import ones to address by policies. In our view, this gives a more realistic assessment of the impact of travel circumstances than asking respondents directly to report on their satisfaction with and importance of, for instance, cleanliness or safety. STS is applied to specific trips rather than to travel in general. In this study, the trip for which STS was recorded was an observed trip, identified by time and location. Measuring STS in such a specific way increases the probability that the impact of incidental or fluctuating circumstances (e.g., congestion, weather) on travel satisfaction can be assessed. Care should still be taken that STS is measured at different times in order to achieve sufficient variation in such circumstances. Likewise, STS can be used to investigate the impact of local circumstances (e.g., design aspects of highways) but this requires a careful selection of locations where data is collected. When measuring STS for a specific trip, the interval between trip making and filling out the questionnaire should be minimised, to avoid memory distortions. Comparative studies (Stone et al., 1999) suggest that reporting on affective responses at the end of the day gives result that are very similar to experience sampling methods, where affect is measured on the spot. This suggests that recall methods can be reliably used to measured individuals’ satisfaction with travel. An advantage of recall methods over experience sampling methods is that they do not interfere with the travel activity itself, which may in itself evoke affective reactions. STS can be applied to more general travel settings. For instance, in another study (Olsson et al., 2012) we asked respondents to report about STS and influential factors for a ‘typical work commute trip’ that they recalled. Also in such a context STS can be used to investigate relationships between a more general level of satisfaction with travel and general trip characteristics, such as travel mode, trip duration, average degree of crowding etc.

177

6. Summary and conclusions In this paper we have described an application of the satisfaction with travel scale (STS) to measure the satisfaction with travel by car drivers observed on four motorways in the Netherlands. The results reported suggest that the reliability of STS is satisfactory to good, and also that the underlying dimensions of STS (positive activation, positive de-activation and cognitive evaluation) are correlated, and are therefore likely to be indicative of an overarching construct of travel satisfaction. Regression analyses carried out with the three STS dimensions suggest that various variables influence STS. Positive activation during travel is positively affected by lower trip frequency, experienced traffic safety, not being annoyed by other road users, and the trip being less tiring. Also, men have higher levels of positive de-activation than women. Positive de-activation is positively affected by not being distracted by billboards. Finally, cognitive judgments of the trip is more positive if the trip purpose is recreation, not driving on the A2 motorway, freedom of choice of speed and lane is higher, fatigue is less, if one is more distracted by billboards, and if the driver is female. Overall, the findings suggest that travellers’ reported satisfaction varies with changes in travel conditions in a plausible way. This together with the results obtained by Ettema et al. (2011) provides support for the validity of the STS as a tool to measure satisfaction with travel. Various avenues for further research can be addressed. First, research is needed into the relationship between satisfaction with travel and travel choices and to longitudinal analyses of satisfaction with travel. In particular, does a change in STS lead to changes in travel choices and to what extent are changes in STS lasting over longer periods? We refer to Ettema et al. (2010) for a more extensive discussion. Regarding the relationship between STS and influencing factors (the topic of this paper), in the context of road traffic a more systematic analysis can be made with respect to design aspects of road infrastructure, including various road types and environments (urban/suburban/rural). In this respect not only design factors, but also service-related factors such as information provision by in car information systems deserve closer attention. For instance, it would be worthwhile to examine in what ways such systems influence drivers’ well-being (e.g., by increasing freedom of choice, experienced safety, etc.). In addition, segmentation of travel markets deserves more attention. For instance, satisfaction with travel may differ between regular and incidental users, requiring that interactions between influential factors and user types need to be accounted for.

References Ben-Elia, E., Ettema, D., 2011. Rewarding rush-hour avoidance: a study of commuters’ travel behaviour. Transportation Research A: Policy and Practice 45, 567–582. Diener, E., Emmons, R.A., Larsen, R.J., Griffen, S., 1985. The satisfaction with life scale. Journal of Personality Assessment 49, 71–75. ¨ Ettema, D., Garling, T., Erikson, L., Friman, M., Olsson, L.E., Fujii, S., 2011. Satisfaction with travel and subjective well-being (SWB): development and tests of a measurement tool. Transportation Research F 14, 167–175. ¨ Ettema, D., Garling, T., Olsson, L.E., Friman, M., 2010. Out-of-home activities, daily travel, and subjective well-being. Transportation Research A: Policy and Practice 44, 723–732. Field, A., 2009. Discovering Statistics using SPSS. Sage. Frederick, S., Loewenstein, G., 1999. Hedonic adaptation. In: Kahneman, D., Diener, E., Schwarz, N. (Eds.), Foundations of Hedonic Psychology: Scientific Perspectives on Enjoyment and Suffering. Russell Sage Foundation, New York, pp. 302–329. ¨ Friman, M., Fujii, S., Ettema, D., Garling, T., Olsson, L.E., 2012. Psychometric analysis of the satisfaction with travel scale. Transportation Research A: Policy and Practice. Frey, B.S., Stutzer, A., 2010. Happiness and public choice. Public Choice 144, 557–573.

178

D. Ettema et al. / Transport Policy 27 (2013) 171–178

¨ Jakobsson, C., 2007. Instrumental motives for private car use. In: Garling, T., Steg, L. (Eds.), Threats to the Quality of Urban Life from Car Traffic: Problems, Causes, and Solutions. Elsevier, Amsterdam, pp. 205–218. ¨ Jakobsson Bergstad, C., Gamble, A., Garling, T., Hagman, O., Polk, M., Ettema, D., Friman, M., Olsson, E.L., 2011. Subjective well-being related to satisfaction with daily travel. Transportation 38, 1–15. ¨ Jakobsson Bergstad, C., Gamble, A., Hagman, O., Polk, M., Garling, T., Ettema, D., Friman, M., Olsson, L.E., 2012. Influences of affect associated with routine outof-home activities on subjective well-being. Applied Research in Quality of Life 7, 49–62. Kahneman, D., 2000. Evaluation by moments: past and future. In: Kahneman, D., Tversky, A. (Eds.), Choices, Values, and Frames. Cambridge University Press, New York, pp. 693–708. Kahneman, D., Krueger, A., Schkade, D., Schwarz, N., Stone, A., 2004. A survey method for characterizing daily life experience: the day reconstruction method (DRM). Science 306, 1776–1780. Mokhtarian, P.L., Salomon, I., 2001. How derived is the demand for travel? Some conceptual and measurement considerations. Transportation Research Part A: Policy and Practice 35, 695–719. Novaco, R.W., Gonzalez, O.I., 2009. Commuting and well-being. In: AmichaiHamburger, Y. (Ed.), Technology and Well-Being. Cambridge: University Press. ¨ Olsson, L.E., Garling, T., Ettema, D., Friman, M., Fujii, S. (2012). Happiness and satisfaction with work commute. Social Indicators Research. http://dx.doi.org/ 10.1007/s11205-012-0003-2. Pedersen, T., Friman, M., Kristensson, P., 2011. Affective forecasting: predicting and experiencing satisfaction with public transport. Journal of Applied Social Psychology 41, 1926–1946. Proost, S., Dender, K.V., 2008. Optimal urban transport pricing in the presence of congestion, economies of density and costly public funds. Transportation Research A: Policy and Practice 42, 1220–1230.

Russell, J.A., 1980. A circumplex model of affect. Journal of Personality and Social Psychology 39, 1161–1178. Steg, L., 2005. Car use: lust and must. Instrumental, symbolic and affective motives for car use. Transportation Research A: Policy and Practice 39, 147–162. Stone, A.A., Shiffman, S.S., DeVries, M.W., 1999. Ecological momentary assessment. In: Kahneman, D., Diener, E., Schwarz, N. (Eds.), Well-Being: The Foundations of Hedonic Psychology. Russell-Sage, New York, pp. 61–84. Stradling, S.G., Anable, J., Carreno, M., 2007. Performance, importance and user disgruntlement: a six-step method for measuring satisfaction with travel modes. Transportation Research A: Policy and Practice 41, 98–106. Schwarz, N., Kahneman, D., Xu, J., 2009. Global and episodic reports of hedonic experience. In: Belli, R., Stafford, F., Alwin, D. (Eds.), Calendar and Diary Methods in Life Course Research. Sage, Thousand Oaks CA, pp. 157–174. Schwartz, N., Xu, J., 2011. Why don’t we learn from poor choices? The consistency of expectation, choice, and memory clouds the lessons of experience. Journal of Consumer Psychology 21, 142–145. Tillema, T., van Wee, B., Ettema, D., 2010. Road pricing and relocation decisions of Dutch households. Urban Studies 47, 3013–3033. ¨ ¨ D., Garling, ¨ Vastfj all, T., 2007. Development and aging: validation of a Swedish short self-report measure of core affect. Scandinavian Journal of Psychology 48, 233–238. ¨ ¨ D., Friman, M., Garling, ¨ Vastfj all, T., Kleiner, M., 2002. The measurement of core affect: a Swedish self-report measure derived from the affect circumplex. Scandinavian Journal of Psychology 43, 19–31. Watson, D., Clark, L.A., Tellegen, A., 1988. Development and validation of brief measures of positive and negative affect: the PANAS scales. Journal of Personality and Social Psychology 54, 1063–1070. World Values Survey. (2005). World Values Survey 2005–2006 Wave, Root version. Retrieved January 27, 2009 from /http://www.worldvaluessurvey.org/S. Xu, J., Schwarz, N., 2009. Do we really need a reason to indulge? Journal of Marketing Research 46, 25–36.