Exploring the positive utility of travel and mode choice

Chapter 14

Exploring the positive utility of travel and mode choice: subjective well-being and travel-based multitasking during the commute Patrick A. Singleton Department of Civil and Environmental Engineering, Utah State University, Logan, UT, United States

Chapter outline 1. Introduction and background 1.1 The positive utility of travel 1.2 PUT and mode choice behavior 1.3 Research objective 2. Data and methods 2.1 Measures of travel subjective well-being

259 259 261 262 263

2.2 Measures of travel-based multitasking 2.3 Analysis methods 3. Results 4. Discussion Acknowledgments References

266 266 267 272 275 275

263

1. Introduction and background 1.1 The positive utility of travel Transportation studies frequently presume travelers seek to maximize the utilitydor, more accurately, minimize the disutility in terms of generalized cost (in minutes, dollars, or utils)dof traveling between locations. This derived-demand approach to travel behavior modeling and analysis thus assumes no benefits to traveling beyond reaching a destination and that the fundamental motivation of traveling is to perform activities in spatially-separated locations. Unfortunately, real world travel behaviors Mapping the Travel Behavior Genome. https://doi.org/10.1016/B978-0-12-817340-4.00014-0 Copyright © 2020 Elsevier Inc. All rights reserved.

259

260 PART | II New research methods and findings

rarely adhere to these theoretical axioms. People may travel out of their way to enjoy pleasant scenery or for variety (Handy et al., 2005). While some people choose to commute by bicycle to get exercise or by train to get an early start on the workday, others drive fancy or powerful cars to feel in control or to express social status (Steg, 2005). Not solely a stressful activity, commuting can be a time for reflection or preparation, improving mental health (Jain and Lyons, 2008). These instances are all examples of the Positive Utility of Travel (PUT) concept. Brought to wider attention by Mokhtarian and Salomon’s work on the “intrinsic drive for mobility” (Salomon and Mokhtarian, 1998, p. 130) and the “tripartite nature of the affinity for travel” (Mokhtarian and Salomon, 2001, p. 701), the PUT notion encompasses both intrinsic motivations for travel (and for traveling in certain ways) and non-destination-oriented benefits of the act of traveling. There are two major types of benefits: those from positive experiences while traveling, and those from using travel time for engaging in activities (Singleton, 2017). These two components are described in the following paragraphs. The travel experience itself may provide benefits that could generate travel for its own sake or motivate certain travel behaviors. Enjoyable experiences can lead to positive emotions and/or a greater sense of happiness, satisfaction, or fulfillment: for example, taking a more scenic route to enjoy the views or riding the bus as an expression of environmental values. Many of these benefits can be encapsulated by the concept of subjective well-being (SWB), a psychological construct involving elements of positive and negative emotions (affect or mood), cognitive evaluations of life and domain satisfaction, and higher-level feelings of purpose and self-actualization (De Vos et al., 2013; Ryan and Deci, 2001). Inherent modal characteristics can impact travel experiences; research consistently finds that well-being for walking and bicycling is rated as more positive than for driving an automobile, and public transit is often more negative (Singleton, 2019d). Other benefits could arise from making productive use of travel time by engaging in other activitiesdlistening to music, reading, working, and moredwhile traveling, also known as travel-based multitasking. Interest in multitasking research has increased coincident with advances in information and communications technologies (ICTs) that have facilitated activity engagement during travel, especially among public transit passengers (Kenyon and Lyons, 2007; Lyons et al., 2016). The operational requirements of certain modes naturally restricts the types of activities that can be (safely) conducted while traveling (Circella et al., 2012); indeed, research suggests that train, bus, and car passengers are most commonly reading, writing, resting, and sleeping, while car drivers are more likely to be listening to audio (Keseru and Macharis, 2018).

Exploring the positive utility of travel and mode choice Chapter | 14

261

1.2 PUT and mode choice behavior Given the modal differences discussed in the preceding paragraphs, it is reasonable to ask whether expectations of well-being or multitaskability might influence mode choice decisions (Abou-Zeid and Ben-Akiva, 2012; De Vos et al., 2016); although, positive experiences may be more likely to motivate travel than opportunities for activity participation. If so, then policies to increase the comfort and enjoyment of sustainable and active transportation modes could help increase their usage (Singleton, 2019d). Furthermore, it is important to understanding the role of multitasking on travel choices due to the potentially transformative effects of self-driving cars (Singleton, 2019a). While travel-related SWB and travel-based multitasking are important and growing areas of study in the travel behavior field (De Vos et al., 2013; Keseru and Macharis, 2018), these PUT concepts are only beginning to be considered in mode choice and other behavioral analyses (Singleton, 2017). Few studies have investigated the potential roles of expectations of positive emotions, increased SWB, or multitasking on mode choice. Studies of attitudes and noninstrumental motivations for car use suggest that positive perceptions or enjoyment of driving could make people more likely to drive (Gardner and Abraham, 2007; Zhao and Zhao, 2015). A stated preference study found that people who placed a greater importance on gaining happiness from travel were more likely to drive than to ride public transit (Duarte et al., 2010). In two stated preference studies, multitasking-related amenities of train travel over car travel (sitting down, table space, internet access, and quiet compartments) did not impact stated mode choice (van der Waerden et al., 2010), but commuters with an inclination to multitask and who listened to music had a lower sensitivity to travel time (Ettema and Verschuren, 2007). Motorized travelers in Mumbai who multitasked also had reduced values of travel time savings in a recent study (Varghese and Jana, 2018). A revealed preference study found a positive association between the perceived multitaskability of a given mode and the utility of that mode, as well as a positive relationship between the propensity to use a laptop or tablet while commuting and choosing commuter rail and carpooling (Malokin et al., 2015). As can be seen, most of this work connecting the PUT concept to mode choice involves stated preference research; even the revealed preference study (Malokin et al., 2015) used a modeled, not measured, variable (propensity to multitask). Stated preference studies can assess only a limited number of attributes that vary across alternatives (thus a narrow depiction of the PUT concept), and may they suffer from potential biases due to hypothetical choice scenarios (Hensher, 2010). One the other hand, propensity models must explain a large portion of the observed varianceda requirement unlikely to be

262 PART | II New research methods and findings

met in practicedand make the restrictive assumption that nonusers with similar characteristics would have the same multitasking behavior or emotional experiences as users of a particular mode. As a result, key questions exist regarding the validity and realism of the choice tasks and relationships being modeled. There is a critical need for additional research to fill this gap in knowledge on the effects of a PUT on observed travel behaviors, especially mode choices. This brief review highlights a key limitation: the challenge and burden of measuring travel-based multitasking and SWB for non-chosen modes. When modeling a mode choice decision for a particular trip, measures of a PUT are unique to both the decision-maker (traveler) and the choice alternative (mode); multitasking behaviors and sensations of well-being likely vary across both people and modes. Thus, as with level-of-service attributes like travel time, these PUT metrics must be gathered for all (chosen and nonchosen) modes in the choice set. From a latent variable perspective, such measures of SWB are individual-specific perceptions that also vary across alternatives, not simply attitudes related (or not) to alternatives (BahamondeBirke et al., 2017). The PUT measures required for a mode choice analysis must be trip-specificdreferring to experiences on a particular trip or a purposeemode combination (“That would have been a fun walk” or “I spent most of my bus commute reading”)dand not just mode-specific (“I like bicycling” or “It’s easy to multitask on public transit”) or about travel in general (“I like traveling” or “Making use of my travel time is very important to me”). As Bahamonde-Birke et al. (2017) note, the critical challenge is that “it is necessary to gather a new set of [perceptual] indicators for every alternative . that the individual faces . [leading] to a significant increase in the information collected .” (p. 478). The burdensome nature of such a data collection is likely the reason why so few PUT studies have analyzed mode choices and why mode choice studies almost never include measures of the PUT concept.

1.3 Research objective The purpose of this research is to examine associations with commute mode choice of measures of the PUT concept, specifically SWB and travel-based multitasking. This study overcomes the data collection challenge by utilizing a unique revealed preference dataset that measured PUT attributes not just for chosen modes but also for alternative modes considered. It uses analytical techniques that integrate discrete choice modeling and structural equation modeling in order to include multitasking measures and latent SWB variables directly into mode choice utility equations. As a result, this research adds to our knowledge of relationships between the PUT concept and travel behavior.

Exploring the positive utility of travel and mode choice Chapter | 14

263

2. Data and methods Data were obtained via a 30-min online questionnaire administered to working and commuting adults in the Portland, Oregon, region. Responses were accepted between October and December 2016, and participants were primarily recruited via email at their place of employment. Although 791 people started the survey, only 546 people completed enough questions to be used in this analysis. The sample was relatively reflective of the Portland-area working population, although it oversampled higher income workers (a result of the recruitment method) and non-auto commuters (by design). Table 14.1 presents descriptive statistics for the estimation sample. See Singleton (2017) for more information on the data collection process. An important discrete choice analysis task involves constructing realistic consideration choice sets. In this study, rather than relying upon analystspecified construction rules, choice sets of alternatives were reported by respondents, who were instructed to “select at least one other mode, but select all that you considered using.” As a result, choice sets were somewhat sparse: 70% of cases had only two modal alternatives, 27% had three alternatives, and 3% had four or five modes from which to choose. Driving an automobile was available for 76% of respondents, and public transit was an available mode for 68% of cases. About 37% of people considered bicycling and riding as a passenger in an automobile. Walking was an available mode for 14% of commuters. Around 61% of commuters chose driving when available. Bicycling, riding public transit, and walking were chosen 48%, 39%, and 31% of the time, respectively. Only 14% of travelers who had the option chose to ride as an automobile passenger. The survey asked about personal and household characteristics, typical travel patterns, and detailed information about a commuter’s most recent trip from home to work, including the chosen mode and measures of travel activities (multitasking) and travel experiences (well-being). These last two categories of PUT measuresddescribed in the following paragraphsdclassify as individual-specific perceptions that also vary across alternatives (Bahamonde-Birke et al., 2017). Notably, and in contrast to previous mode choice studies, these questions were asked for all commute modes considered, even if not chosen.

2.1 Measures of travel subjective well-being Positive and negative travel experiences related to SWB were measured by the Satisfaction with Travel Scale (STS) (Ettema et al., 2011). The STS captures two primary aspects of hedonic SWB resulting from travel: affective or emotional aspects (positive and negative feelings) and an overall cognitive evaluation. For each of nine paired statements, respondents selected a choice on a seven-point semantic differential scale that best

264 PART | II New research methods and findings

TABLE 14.1 Descriptive statistics (N ¼ 546). Categorical Variable

#

%

Mode:Walk

23

4.2

Bicycle

98

17.9

Transit

145

26.6

Auto, passenger

27

4.9

Auto, driver

253

46.3

Continuous Mean

SD

35.67

20.86

# children (age  16)

0.49

0.87

# older adults (age  65)

0.07

0.28

# cars

1.71

1.01

# bicycles

2.58

2.05

Trip characteristics

Travel time (minutes) Precipitation  0.10 in (2.5 mm)

116

21.2

Age: 18e34 years

105

19.2

35e44 years

142

26.0

45e54 years

142

26.0

55e64 years

123

22.5

65þ years

34

6.2

Gender: Female

293

53.7

Male

253

46.3

Disability

36

6.6

Education: Graduate degree

240

44.0

Traveler socio-demographics

a

Income :$0e50k

44

8.1

$50e75k

94

17.2

$75e100k

120

22.0

$100e150k

158

28.9

107

19.6

111

20.3

$150kþ b

Multifamily home

Exploring the positive utility of travel and mode choice Chapter | 14

265

TABLE 14.1 Descriptive statistics (N ¼ 546).dcont’d Categorical Variable Transit pass

#

%

234

42.9

# hours worked Flexible work schedule

351

Continuous Mean

SD

41.82

8.30

64.3

a

Income is in 2016 US Dollars. A housing unit located in a structure with other housing units, such as a duplex, apartment building, or condominium.

b

corresponded to their overall experience. These responses were then structured using confirmatory factor analysis, yielding the expected three unobserved constructsd“Positive deactivation” (PD), “Positive activation” (PA), and “Cognitive evaluation” (CE)das well as an overall STS concept “Commute satisfaction.” For more information on the STS and its construction, see Fig. 14.1 and Singleton (2019b). The survey also included unique questions probing more deeply into affective and eudaimonic dimensions of travel-related SWB (Singleton and Clifton, in progress), but those were not included in this analysis.

FIG. 14.1 Confirmatory factor analysis of the STS. From Singleton, P.A., 2019b. Validating the satisfaction with travel scale as a measure of hedonic subjective well-being for commuting in a U.S. city. Transportation Research Part F: Traffic Psychology and Behaviour. https://doi.org/10. 1016/j.trf.2018.10.029 (2019); © 2018 Elsevier Ltd.

266 PART | II New research methods and findings

2.2 Measures of travel-based multitasking Travel activity aspects were measured using questions about travel-based multitasking: Respondents were asked to first select which of 23 distinct activities they conducted while commuting, and next to provide an approximate percentage of the travel time spent doing each activity. This information, along with constructed mode-specific travel times, was used to calculate two measures of travel-based multitasking: activity participation (binary) and activity duration (minutes). Exploratory factor analysis grouped six activities into two categories: information and communication technology or “ICT”-related activities (using social websites or apps; texting, emailing, or other messaging; reading electronically), and “passive” activities (viewing scenery, watching people; thinking or daydreaming). After removing activities with low response frequencies, 14 activities or activity groups were retained. For more information on these travel activity measures of PUT, see Singleton (2018, 2019c).

2.3 Analysis methods The mode choice analysis utilized integrated choice and latent variable (ICLV) modeling, sometimes called hybrid choice modeling. ICLV modelsdwhich combine discrete choice modeling for mode choice and structural equation modeling (SEM) for psychometricsdhave become a common means of integrating latent variables like attitudes and perceptions into travel behavior analyses. The statistical methodology for specifying and estimating ICLV models was developed during the 1980s and 1990s (e.g.: McFadden, 1986; Ben-Akiva et al., 1999; Ben-Akiva et al., 2002) but did not see rapid growth until recent increases in computational power. Unlike most hybrid choice models, this study utilized individual-specific perceptions of travel experiences (the STS) that also varied across alternatives, thus requiring 20 different latent variables: four latent variables of the STS (PD, PA, CE, and overall commute satisfaction) for each of the five modes (walking, bicycling, auto driver, auto passenger, public transit). ICLV mode choice model estimation was conducted with Python Biogeme Version 2.6a (Bierlaire, 2016), using maximum simulated likelihood estimation with CFSQP nonlinear optimization and 1000 random draws according to a Modified Latin Hypercube Sampling strategy. Several types of independent variables were included in the mode choice model. Traveler socio-demographic characteristics were collected on the questionnaire and entered the model with alternative-specific coefficients (relative to the reference alternative: auto driver). Variables that were moderately-to-strongly correlated (>0.40) were removed to address multicollinearity issues; those that remained are listed in Table 14.1. Three types of alternative-varying attributes were also included in the models. Level-of-service informationdtravel time and costdfor each alternative was constructed

Exploring the positive utility of travel and mode choice Chapter | 14

267

assuming shortest-path routes and including parking costs; see Singleton (2017) for details. Measures of travel-based multitasking (activity participation and duration) had alternative-specific coefficients because inherent modal characteristics affect the range of possible in-travel activities. (Activityemode combinations had to have at least five observations to avoid multicollinearity.) The measure of travel SWB (the STS’s overall Commute satisfaction factor) had a generic coefficient because it did not vary significantly across modes.

3. Results Two different ICLV mode choice models were estimated: one using activity participation (Model A) and one using activity duration (Model B) measures of travel-based multitasking. Results for the discrete choice portions of the ICLV models are shown in Tables 14.2 and 14.3; results for the SEM portions were similar to those shown in Fig. 15.1. (Full model results can be obtained from the author.) Overall, the independent variables were relatively explanatory of mode choice behavior; the models reduced almost 70% of the null model deviance (McFadden’s pseudo R2 ¼ 0.69 and 0.68) of the discrete choice part of the ICLV model. A substantial portion of reduction this came from the PUT measures of travel activities and travel experiences: A model with just levelof-service attributes and trip/traveler characteristics had a far lower goodness-of-fit (McFadden’s pseudo R2 ¼ 0.47), and models with just SWB or just multitasking variables reduced around a quarter of the null model deviance (McFadden’s pseudo R2 ¼ 0.22e0.26). (These models are not shown but are available from the author.) Level-of-service attributes were influential in expected directions. Travel cost was negatively associated with mode choice: Every additional dollar decreased the odds of choosing a particular alternative by more than half (eB e 1 ¼ 0.59 and 0.55). The travel time coefficient was also negative and significant: A 10-min increase in travel time yielded about a 20% reduction in the odds of choosing a particular mode (e10B e 1 ¼ 0.17 and 0.23). The implied value of travel time savings (VTTS ¼ 60 ∙ BTT/BCO) was $1.22/hour in Model A and $1.95/hour in Model B. These subjective time valuations are considerably lower than is typically found in mode choice analyses; one explanation is that this study used smaller consideration choice sets. Perhaps travel time plays a bigger role in choice set construction (i.e., in deciding which modes are feasible (Singleton, 2013)) than in a mode choice decision among actually considered alternatives. In studies with less restrictive choice sets, a stronger travel time effect would likely show up as a higher value of time in mode choice. In Model B, assuming that 100% of travel time was spent doing nothing yielded a larger travel time value of $18.38/hour (VTTS ¼ 60 ∙ (BTT þ BDN)/BCO) for automobile modes.

TABLE 14.2 Results from discrete choice model portion of ICLV Model A (with activity participation). Mode Category & variable

Generic/Auto driver

Auto passenger SE

Bicycle

B

SE

p

B

p

B

SE

2.043

0.318

0.00*

2.043

ICT activities

0.061

0.971

0.95

4.517

1.186

0.00*

0.640

0.901

Passive activities

0.582

0.641

0.36

0.917

0.809

0.26

0.968

0.733

Talking with people you know

1.085

0.752

0.15

1.085

1.152

Talking with strangers

1.932

1.195

0.11

1.932

0.776

Talking on the phone

0.921

0.624

0.14

Walk P

B

SE

Transit p

B

SE

p

Travel experiences Commute satisfaction (STS)

2.043

2.043

2.043

0.48

0.640

0.537

1.083

0.62

0.19

0.968

1.692

0.754

0.02*

1.004

0.25

1.152

0.118

0.931

0.90

1.031

0.45

1.341

1.090

0.730

0.14

0.178

0.681

0.79

0.487

0.683

0.48

0.716

0.833

0.39

2.695

0.870

0.00*

2.695

0.870

0.40

0.808

0.694

0.24

2.175

1.049

0.04*

2.042

1.396

0.14

Travel activities

1.224

0.27

Reading print Listening to music, radio, audio

0.495

0.796

0.53

2.591

1.088

0.02*

0.588

1.109

0.60

1.775

0.921

0.05*

Playing game Eating; drinking

0.701

0.814

0.39

Singing; dancing

1.810

1.099

0.10w

1.644

0.687

0.02*

4.184

Exercising; physically active Planning or navigating this trip

1.644

1.119

0.00*

3.473

1.516

0.02*

3.473

2.214

0.946

0.02*

2.276

3.129

0.943

0.00*

3.129

Sleeping or snoozing

0.07w

1.132

1.218

0.35

Travel time (minutes)

0.018

0.007

0.01*

0.018

0.018

0.018

0.018

Travel cost ($)

0.901

0.164

0.00*

0.901

0.901

0.901

0.901

Doing nothing

1.132

1.260

5.995

1.471

0.00*

5.995

Level-of-service attributes

Trip and traveler characteristics Intercept

e

e

Age:18e34 years

e

4.679

35e44 years

e

e

10.450

55e64 years

e

e

e

65+ years

e

e

e

Gender: Female

3.326

1.156

0.00*

2.312

e 1.145

0.989

0.00*

0.02*

e

e

e 2.508

0.00*

e e 4.017

4.946

1.439

0.00*

e

2.253

0.06w

e

1.854

0.03*

Disability

e

2.930

1.786

0.10w

4.183

Education: Graduate degree

2.453

0.935

0.01*

1.755

0.729

0.02*

e

1.380

0.788

0.08w

# children (age  16)

2.416

0.786

0.00*

1.711

0.527

0.00*

e

1.167

0.537

0.03*

# older adults (age  65)

e

e

4.221

1.482

0.00*

Income: $0e50k

e

e

e

e

e

9.419

e

$50e75k

4.170

1.810

0.02*

2.335

0.00*

$100e150k

8.666

1.785

0.00*

e

7.850

2.533

0.00*

e

$150k+

11.355

2.035

0.00*

e

6.365

2.400

0.01*

e

e

Multifamily home

3.058

1.039

0.00*

3.320

1.207

0.01*

e

# cars

1.470

0.534

0.01*

1.516

0.526

0.00*

e

e

# bicycles

e

0.535

0.194

0.01*

e

0.482

0.214

0.02*

Transit pass

2.549

e

4.694

0.912

0.00*

# hours worked

e

Flexible work schedule

0.338

Precipitation  0.10 in (2.5 mm)

e

e

1.284

0.05*

1.397

0.840

0.10w

2.121

0.750

0.00*

5.464

1.591

0.00*

1.467

0.760

0.05*

0.068

0.00*

e

0.155

0.052

0.00*

0.120

0.044

0.01*

e

e

Coefficients (B): [blank] ¼ not included, d ¼ included but p > 0.10, B without SE/p ¼ coefficient is generic or shared with another mode. Statistical significance: * ¼ p  0.05, w ¼ p  0.10. Model statistics: N ¼ 546, null log-likelihood ¼ e449.49, model log-likelihood ¼ e137.82, McFadden R2 ¼ 0.693.

e

TABLE 14.3 Results from discrete choice model portion of ICLV Model B (with activity duration). Mode Category & variable

Generic/Auto driver

Auto passenger

B

SE

p

B

2.201

0.446

0.00*

2.201

SE

Bicycle p

B

SE

Walk P

B

SE

Transit p

B

SE

p

Travel experiences Commute satisfaction (STS)

2.201

2.201

2.201

Travel activities ICT activities

0.075

0.134

0.58

0.112

0.045

0.01*

0.003

0.080

0.97

0.003

0.042

0.019

0.03*

Passive activities

0.031

0.017

0.07w

0.105

0.037

0.00*

0.034

0.012

0.00*

0.034

0.058

0.017

0.00*

Talking with people you know

0.117

0.043

0.01*

0.117

0.057

0.030

0.06w

0.057

0.029

0.058

0.62

Talking with strangers

0.204

0.397

0.61

0.204

0.198

0.320

0.54

0.049

0.155

0.75

0.188

0.140

0.18

Talking on the phone

0.109

0.045

0.01* 0.022

0.018

0.23

0.022

0.031

0.48

0.112

0.045

0.01*

0.004

0.022

0.84

0.008

0.013

0.53

0.008

0.020

0.69

0.030

0.046

0.51

Eating; drinking

0.139

0.050

0.01*

0.087

0.065

0.18

0.157

0.063

0.01*

0.157

0.006

0.041

0.88

Singing; dancing

0.079

0.047

0.09w 0.002

0.014

0.89

0.016

0.006

0.041

0.42

0.232

0.076

0.00*

0.232

Reading print Listening to music, radio, audio Playing game

Exercising; physically active Planning or navigating this trip

0.116

0.039

0.00*

0.116

Sleeping or snoozing Doing nothing

0.218

0.059

0.00*

0.218

0.011

0.015

0.46

0.011

0.017

0.36

0.029

0.075

0.70

0.160

0.038

0.00*

0.034

0.040

0.39

Level-of-service attributes Travel time (minutes)

0.026

0.013

0.05*

0.026

0.026

0.026

0.026

Travel cost ($)

0.795

0.152

0.00*

0.795

0.795

0.795

0.795

Trip and traveler characteristics Intercept

e

e

Age:18e34 years

e

3.972

35e44 years

e

e

8.719

55e64 years

e

e

e

65+ years

e

e 1.221

0.00*

e

e

e 2.546

0.00*

e e

3.731

2.039

0.07w

5.977

1.418

0.871

0.10w

3.507

1.503

0.02*

e

e

2.814

1.455

0.05*

e

e

e

e

e

1.828

e

e

e

Gender: Female

2.074

1.226

Disability Education: Graduate degree # children (age  16)

1.605

# older adults (age  65)

e

e

Income:$0e50k

e

2.905

0.721

0.09w

0.03*

1.557

0.641

1.542

0.02*

0.06w

e

4.419

e

e

6.412

1.990

0.00*

2.233

e

3.093

1.857

0.10w

e

e

e

e

0.00*

e

e

0.505

0.01*

e

0.558

0.263

0.01*

1.719

0.00*

0.678

0.01*

1.234

0.00*

0.972

0.02*

0.323

0.08w

$50e75k

3.894

1.801

0.03*

$100e150k

6.892

1.327

0.00*

$150k+

9.611

2.099

0.00*

Multifamily home

4.347

1.234

0.00*

3.170

1.038

# cars

1.318

0.519

0.01*

1.238 0.643

e

0.377

0.215

0.08w

0.984

0.01*

e

e

3.715

0.803

0.00* 0.03*

e

# bicycles

e

Transit pass

2.399

# hours worked

e

e

4.092

1.758

0.02*

1.391

0.646

Flexible work schedule

0.313

0.075

0.00*

e

0.163

0.062

0.01*

0.134

0.052

0.01*

Precipitation  0.10 in (2.5 mm)

1.988

0.997

0.05*

1.706

1.643

0.797

0.04*

0.910

0.06w

e

Coefficients (B): [blank] ¼ not included, d ¼ included but p > 0.10, B without SE/p ¼ coefficient is generic or shared with another mode. Statistical significance: * ¼ p  0.05, w ¼ p  0.10. Model statistics: N ¼ 546, null log-likelihood ¼ e449.49, model log-likelihood ¼ e142.09, McFadden R2 ¼ 0.684.

272 PART | II New research methods and findings

The measure of SWB from the travel experiencedthe STS commute satisfaction latent variabledwas positively and significantly associated with commute mode choice. For scaling and identification purposes, the variance of this latent variable was fixed at 1.00, so coefficients can be interpreted as standardized effects. A one standard deviation increase in “Commute satisfaction” for any particular mode was associated with a greater than 650% increase (eB  1 ¼ 6.71) in the odds of using that mode in Model A; this was an 800% increase (eB  1 ¼ 8.03) in Model B. In comparison, one standard deviation was roughly the difference between the median STS scores for people who walked versus rode transit, and between those for people who bicycled versus drove (Singleton, 2019b). Surprisingly, many instances of travel-based multitasking were not or even negatively associated with mode choice. Although most of the coefficient signs were the same in both Models A and B, statistical significance tended to differ slightly. Activities with positive and significant coefficients included listening to audio for auto passengers and eating/drinking for nonmotorized commuters (in both models), as well as exercising for bicycling (Model A). Performing ICT-enabled activitiesdoften considered a productive use of time for ridersdwas actually negatively associated with mode choice for auto passengers and transit users. Other activities with a significant negative association in both models included passive activities and sleeping/snoozing for transit and planning/navigating for auto and nonmotorized travelers. Doing nothing had conflicting results: It was a positive consideration for walking/bicycling in Model A but a negative one for auto drivers/passengers in Model B. One implication of using activity durations in Model B is that values of time for travel-based multitasking can be calculated (VTTS ¼ BTA/BCO). Model estimation results imply that, on average and at the margin, commuters might be willing to pay 4e13¢/min to avoid doing passive activities, 14¢/min to avoid talking on the phone while driving, 15e29¢/min to avoid planning or navigating the trip, and 20¢/min to not sleep when on transit. Conversely, auto travelers might be willing to pay 27¢/min to avoid doing nothing. Of course, these interpretations are sensitive to small changes in the estimated parameters (particularly the cost coefficient) and should not be taken to mean that people have a literal willingness-to-pay for these things.

4. Discussion This study is among the first to demonstrate revealed preference evidence consistent with measures of the positive utility of travel (PUT) concept having a direct impact on mode choice behavior. The use of an ICLV modeling framework offers more convincing evidence than could be obtained by either a sequential estimation process or by using fewer or highly correlated explanatory PUT variables. The significant associations between both

Exploring the positive utility of travel and mode choice Chapter | 14

273

multitasking and well-being components and commute mode choice highlight the importance of these factors in travelers’ mode choice decision-making processes. Compared to a simple mode choice model with level-of-service and traveler characteristics variables, adding the PUT-related measures improved the model goodness-of-fit (McFadden’s pseudo R2) by nearly 50%. This sizable amount highlights the importance and value of accounting for these effects in a mode choice study. In particular, the strong, consistently positive, and statistically significant association between “Commute satisfaction” and mode choicedespecially via the use of a validated (and measured) multi-item metric of hedonic SWB (the STS)dis a major finding. It appears that commuters may indeed consider and make mode choice decisions based on expectations of improvements in wellbeing, mental health, and happiness as a result of the commute or at least expected differences in SWB across modes. Results for travel-based multitasking were more equivocal. The lack of significance of many traditionally productive activities (like ICT use, talking on the phone, and reading print) and the few positive but many negative associations for other activities all suggest that activity participation during travel may be more about coping with a burdensome commute than it is about making productive use of travel time. As discussed in previous work (Singleton, 2018), some activities with negative coefficientsdin this case, ICT activities, passive activities, and sleeping/snoozingdmay be more about “killing time” than making use of it. Yet, at least auto commuters preferred doing something rather than “doing nothing” (in Model B), and bicycle riders seemed to value exercise (in Model A). Overall, these results are consistent with expectations from the PUT concept that travelers may be motivated by or find benefits beyond simply reaching a destination. In fact, PUT considerations may be attenuating the willingness to pay for travel time reductions; when adding PUT measures to Model A, the implied value of travel time savings decreased slightly compared to a model (not shown) with only level-of-service attributes. The stronger association with mode choice for SWB measures than for multitasking metrics is, upon reflection, not entirely unexpected. Expectations of improved wellbeing are likely sufficient to occasionally generate undirected travel or (more frequently) affect mode choices; alternatively, the motivation for intravel activity participation is unlikely to generate any new trips. Instead, travel-based multitasking may act in other ways, such as in reducing values of time (perhaps only modestly (Singleton, 2019a)) or inhibiting motivations to reduce travel (Mokhtarian et al., 2015). This study was not without limitations; it is but an initial step toward a potentially fruitful line of research, and there are several opportunities for future work. Using more targeted questions about the quality of travel time use for various activities and reasons for activity participation during travel (Rosenfield and Zhao, 2016) could improve measures of travel-based

274 PART | II New research methods and findings

multitasking. More fundamentally, temporal issues involved in the measurement of PUT attributes for chosen versus non-chosen modes warrant more careful consideration during the survey design and data collection process. The mismatch between asking people to report what they did and felt on a recent commute versus asking them to imagine what they would have done or would have felt if they had used a different modedsimilar to discussions about prospective/expected versus experienced versus retrospective/remembered utilities (Abou-Zeid and Ben-Akiva, 2012; Singleton, 2017)dis complex and challenging to resolve. Additionally, (lack of) experience with (un)familiar modes, confirmation bias, or cognitive dissonance may also have played a role, perhaps inflating positive assessments of chosen modes. The analysis could have had more power to detect significant associations given a larger sample size, especially for walking and auto passenger modes. Panel studies of PUT factors (Rasouli and Timmermans, 2014) would also be instructive. Finally, it would be useful to replicate this study in different geographic and cultural contexts. For example, the relative lack of significant multitasking associations could result from the fact that Portland does not a have a large and longdistance rail-based transit network. External studies would help to inform our understanding of whether these relationships between the PUT concept and mode choice are more universal or depend more on local context. Overall, future research should continue investigating connections between the PUT concept and mode choice, thus adding to the sparse body of evidence. If confirmed, this study’s findings offer implications for transportation policy-making. A significant effect of travel-related SWB on mode choice would suggest that efforts to improve the traveling experience could be effective in achieving mode shifts toward more socially beneficial travel behaviors or non-auto mode split goals. For example, infrastructural and streetscape design changes for non-motorized safety or comfortdsafer street crossings, wider and more pleasant sidewalks, and protected bike lanes/ intersectionsdmight also increase enjoyment of the travel experience and encourage the use of active transportation modes. In the long run, the importance of understanding the behavioral influences of the PUT concept will only increase, as autonomous vehicles offer opportunities for automobile use to become more productive and less stressful. Yet, the lack of many positive effects of multitasking on mode choice potentially suggests that these productivity benefits could be overrated (Singleton, 2019a). The results of this research imply that reductions in stress (improvements to SWB) from not having to operate a self-driving car may outweigh the ability to engage in a greater number of travel activities (increases in multitasking); the potential size of this effect is an open question for future research. Studying the relationships between travel behaviors and dimensions of the PUT concept today can help to prepare for an uncertain future.

Exploring the positive utility of travel and mode choice Chapter | 14

275

Acknowledgments Thanks to Kelly Clifton, Jennifer Dill, Cynthia Mohr, Liming Wang, and others at Portland State University as well as several anonymous reviewers for their valuable feedback. Data collection and preliminary analysis was supported in part by a Doctoral Dissertation Fellowship from the National Institute for Transportation and Communities, a program of the Transportation Research and Education Center at Portland State University; and by a graduate fellowship from the Dwight David Eisenhower Transportation Fellowship Program, a program of the Federal Highway Administration of the US Department of Transportation. The funders had no involvement in the study design, data collection, analysis, and preparation of this manuscript. The author has no conflicts of interest to disclose. The author confirms sole responsibility for the following: study conception and design, data collection, analysis and interpretation of results, and manuscript preparation.

References Abou-Zeid, M., Ben-Akiva, M., 2012. Well-being and activity-based models. Transportation 39 (6), 1189e1207. http://doi.org/10.1007/s11116-012-9387-8. Bahamonde-Birke, F.J., Kunert, U., Link, H., de Dios Ortu´zar, J., 2017. About attitudes and perceptions: finding the proper way to consider latent variables in discrete choice models. Transportation 44 (3), 475e493. http://doi.org/10.1007/s11116-015-9663-5. Ben-Akiva, M., McFadden, D., Ga¨rling, T., Gopinath, D., Walker, J., Bolduc, D., et al., 1999. Extended framework for modeling choice behavior. Marketing Letters 10 (3), 187e203. http:// doi.org/10.1023/A:1008046730291. Ben-Akiva, M., Walker, J., Bernardino, A.T., Gopinath, D.A., Morikawa, T., Polydoropoulou, A., 2002. Integration of choice and latent variable models. In: Mahmassani, H.S. (Ed.), In Perpetual Motion: Travel Behavior Research Opportunities and Application Challenges. Pergamon, Amsterdam, NL, pp. 431e470. Bierlaire, M., 2016. PythonBiogeme: A Short Introduction (Report TRANSP-OR 160706). Ecole Polytechnique Fe´de´rale de Lausanne, Lausanne, CH. http://biogeme.epfl.ch/documents.html. Circella, G., Mokhtarian, P.L., Poff, L.K., 2012. A conceptual typology of multitasking behavior and polychronicity preferences. Electronic International Journal of Time Use Research 9 (1), 59e107. http://doi.org/10.13085/eijtur.9.1.59-107. De Vos, J., Mokhtarian, P.L., Schwanen, T., Van Acker, V., Witlox, F., 2016. Travel mode choice and travel satisfaction: bridging the gap between decision utility and experienced utility. Transportation 43 (5), 771e796. http://doi.org/10.1007/s11116-015-9619-9. De Vos, J., Schwanen, T., Van Acker, V., Witlox, F., 2013. Travel and subjective well-being: a focus on findings, methods and future research needs. Transport Reviews 33 (4), 421e442. http://doi.org/10.1080/01441647.2013.815665. Duarte, A., Garcia, C., Giannarakis, G., Lima˜o, S., Polydoropoulou, A., Litinas, N., 2010. New approaches in transportation planning: happiness and transport economics. Netnomics: Economic Research and Electronic Networking 11 (1), 5e32. http://doi.org/10.1007/s11066-0099037-2. Ettema, D., Ga¨rling, T., Eriksson, L., Friman, M., Olsson, L.E., Fujii, S., 2011. Satisfaction with travel and subjective well-being: development and test of a measurement tool. Transportation Research Part F: Traffic Psychology and Behaviour 14 (3), 167e175. http://doi.org/10.1016/j. trf.2010.11.002.

276 PART | II New research methods and findings Ettema, D., Verschuren, L., 2007. Multitasking and value of travel time savings. Transportation Research Record: Journal of the Transportation Research Board 2010, 19e25. http://doi.org/ 10.3141/2010-03. Gardner, B., Abraham, C., 2007. What drives car use? A grounded theory analysis of commuters’ reasons for driving. Transportation Research Part F: Traffic Psychology and Behaviour 10 (3), 187e200. http://doi.org/10.1016/j.trf.2006.09.004. Handy, S., Weston, L., Mokhtarian, P.L., 2005. Driving by choice or necessity? Transportation Research Part A: Policy and Practice 39 (2e3), 183e203. http://doi.org/10.1016/j.tra.2004. 09.002. Hensher, D.A., 2010. Hypothetical bias, choice experiments and willingness to pay. Transportation Research Part B: Methodological 44 (6), 735e752. https://doi.org/10.1016/j.trb.2009.12.012. Jain, J., Lyons, G., 2008. The gift of travel time. Journal of Transport Geography 16 (2), 81e89. http://doi.org/10.1016/j.jtrangeo.2007.05.001. Kenyon, S., Lyons, G., 2007. Introducing multitasking to the study of travel and ICT: examining its extent and assessing its potential importance. Transportation Research Part A: Policy and Practice 41 (2), 161e175. http://doi.org/10.1016/j.tra.2006.02.004. Keseru, I., Macharis, C., 2018. Travel-based multitasking: review of the empirical evidence. Transport Reviews 38 (2), 162e183. http://doi.org/10.1080/01441647.2017.1317048. Lyons, G., Jain, J., Weir, I., 2016. Changing timeseA decade of empirical insight into the experience of rail passengers in Great Britain. Journal of Transport Geography 57, 94e104. http:// doi.org/10.1016/j.jtrangeo.2016.10.003. Malokin, A., Circella, G., Mokhtarian, P.L., 2015. How do activities conducted while commuting influence mode choice? Testing public transportation advantage and autonomous vehicle scenarios. In: Presented at the 94th Annual Meeting of the Transportation Research Board, Washington, DC. https://trid.trb.org/view/1336974. McFadden, D., 1986. The choice theory approach to market research. Marketing Science 5 (4), 275e297. https://doi.org/10.1287/mksc.5.4.275. 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 (8), 695e719. http://doi.org/10.1016/S0965-8564(00)00013-6. Mokhtarian, P.L., Salomon, I., Singer, M.E., 2015. What moves us? An interdisciplinary exploration of reasons for traveling. Transport Reviews 35 (3), 250e274. https://doi.org/10.1080/ 01441647.2015.1013076. Rasouli, S., Timmermans, H., 2014. Judgments of travel experiences, activity envelopes, trip features and multi-tasking: a panel effects regression model specification. Transportation Research Part A: Policy and Practice 63, 67e75. https://doi.org/10.1016/j.tra.2014.02.012. Rosenfield, A., Zhao, J., 2016. Making the commute count: quality of productive travel time use. In: Presented at the 56th Annual Conference of the Association of Collegiate Schools of Planning, Portland, OR. Ryan, R.M., Deci, E.L., 2001. On happiness and human potentials: a review of research on hedonic and eudaimonic well-being. Annual Review of Psychology 52 (1), 141e166. http://doi.org/10. 1146/annurev.psych.52.1.141. Salomon, I., Mokhtarian, P.L., 1998. What happens when mobility-inclined market segments face accessibility-enhancing policies? Transportation Research Part D: Transport and Environment 3 (3), 129e140. http://doi.org/10.1016/S1361-9209(97)00038-2. Singleton, P.A., 2013. A Theory of Travel Decision-Making with Applications for Modeling Active Travel Demand (Master’s thesis). Portland State University, Portland, OR. http://doi.org/10. 15760/etd.1493.

Exploring the positive utility of travel and mode choice Chapter | 14

277

Singleton, P.A., 2017. Exploring The Positive Utility of Travel and Mode Choice (Doctoral dissertation). Portland State University, Portland, OR. http://doi.org/10.15760/etd.3447. Singleton, P.A., 2019d. Walking (and cycling) to well-being: modal and other determinants of subjective well-being during the commute. Travel Behaviour and Society 16, 249e261. https://doi.org/10.1016/j.tbs.2018.02.005. Singleton, P.A., 2018. How useful is travel-based multitasking? Evidence from commuters in Portland, Oregon. Transportation Research Record: Journal of the Transportation Research Board 2672 (50), 11e22. https://doi.org/10.1177/0361198118776151. Singleton, P.A., 2019a. Discussing the “positive utilities” of autonomous vehicles: will travellers really use their time productively? Transport Reviews 39, 50e65. https://doi.org/10.1080/ 01441647.2018.1470584. Singleton, P.A., 2019b. Validating the satisfaction with travel scale as a measure of hedonic subjective well-being for commuting in a U.S. city. Transportation Research Part F: Traffic Psychology and Behaviour 60, 399e414. https://doi.org/10.1016/j.trf.2018.10.029. Singleton, P.A., 2019c. Multimodal travel-based multitasking during the commute: who does what? International Journal of Sustainable Transportation (in-press). https://doi.org/10.1080/ 15568318.2018.1536237. Singleton, P.A., Clifton, K.J. Towards measures of affective and eudaimonic subjective well-being in the travel domain. Available from: the authors (in progress). Steg, L., 2005. Car use: lust and must. Instrumental, symbolic and affective motives for car use. Transportation Research Part A: Policy and Practice 39 (2e3), 147e162. https://doi.org/10. 1016/j.tra.2004.07.001. van der Waerden, P.J., Kemperman, A., Timmermans, H., van Hulle, R., 2010. The influence of facilities for multitasking on individual’s travel decisions in the context of work trips. In: Presented at the 89th Annual Meeting of the Transportation Research Board, Washington, DC. https://trid.trb.org/view/910535. Varghese, V., Jana, A., 2018. Impact of ICT on multitasking during travel and the value of travel time savings: empirical evidences from Mumbai, India. Travel Behaviour and Society 12, 11e22. https://doi.org/10.1016/j.tbs.2018.03.003. Zhao, Z., Zhao, J., 2015. Car pride: psychological structure and behavioral implications. In: Presented at the 94th Annual Meeting of the Transportation Research Board, Washington, DC. https://trid.trb.org/view/1336944.