MODELING THE INDIVIDUAL CONSIDERATION OF
TRAVEL- RELATED STRATEGY BUNDLES
UCD- ITS- RR- 04- 7
April 2004
by
Sangho Choo
Department of Civil and Environmental Engineering
and
Institute of Transportation Studies
University of California, Davis 95616, USA
Ph ( 530) 754- 7421
Fax ( 530) 752- 6572
cshchoo@ ucdavis. edu
and
Patricia L. Mokhtarian
Department of Civil and Environmental Engineering
and
Institute of Transportation Studies
University of California, Davis 95616, USA
Ph ( 530) 752- 7062
Fax ( 530) 752- 7872
plmokhtarian@ ucdavis. edu
Institute of Transportation Studies
One Shields Avenue
University of California
Davis, California 95616
Tel: 530- 752- 0247 Fax: 530- 752- 6572
http:// www. its. ucdavis. edu/
email: itspublications@ ucdavis. edu
MODELING THE INDIVIDUAL CONSIDERATION OF
TRAVEL- RELATED STRATEGY BUNDLES
Sangho Choo
Department of Civil and Environmental Engineering
One Shields Avenue
University of California, Davis
Davis, CA 95616
voice: ( 530) 754- 7421
fax: ( 530) 752- 6572
e- mail: cshchoo@ ucdavis. edu
and
Patricia L. Mokhtarian
Department of Civil and Environmental Engineering
and
Institute of Transportation Studies
One Shields Avenue
University of California, Davis
Davis, CA 95616
voice: ( 530) 752- 7062
fax: ( 530) 752- 7872
e- mail: plmokhtarian@ ucdavis. edu
April 2004
This research is funded by the University of California Transportation Center.
i
ATTITUDES TOWARD MOBILITY
Patricia L. Mokhtarian, Principal Investigator
Journal Articles Produced by this Project to Date
Cao, Xinyu and Patricia L. Mokhtarian ( 2003) How do individuals manage their personal travel?
Objective and subjective influences on the consideration of travel- related strategies. Under review
for publication.
Choo, Sangho and Patricia L. Mokhtarian ( 2004) What type of vehicle do people drive? The role
of attitude and lifestyle in influencing vehicle type choice. Transportation Research A 38( 3), 201-
222.
Choo, Sangho, Gustavo O. Collantes, and Patricia L. Mokhtarian ( forthcoming) Wanting to travel,
more or less: Exploring the determinants of a perceived deficit or surfeit of personal travel.
Transportation.
Clay, Michael J. and Patricia L. Mokhtarian ( forthcoming) Personal travel management: The
adoption and consideration of travel- related strategies. Transportation Planning and Technology.
Collantes, Gustavo O. and Patricia L. Mokhtarian ( 2002) Qualitative subjective assessments of
personal mobility: Exploring the magnifying and diminishing cognitive mechanisms involved.
Under review for publication.
Handy, Susan L., Lisa Weston, and Patricia L. Mokhtarian ( 2003) Driving by choice or necessity?
The case of the soccer mom and other stories. Paper presented at the 82nd annual meeting of the
Transportation Research Board, Washington, DC, January, draft available on conference CD- ROM.
Handy, Susan L., Lisa Weston, and Patricia L. Mokhtarian ( 2004) Driving by choice or necessity?
( later version of 2003 paper) Manuscript under review for publication.
Mokhtarian, Patricia L., Ilan Salomon, and Lothlorien S. Redmond ( 2001) Understanding the
demand for travel: It's not purely “ derived”. Innovation: The European Journal of Social Science
Research 14( 4), 355- 380.
Mokhtarian, Patricia L. and Ilan Salomon ( 2001) How derived is the demand for travel? Some
conceptual and measurement considerations. Transportation Research A 35( 8), 695- 719.
Ory, David T. and Patricia L. Mokhtarian ( 2004) When is getting there half the fun? Modeling the
liking for travel. Manuscript under review for publication.
Ory, David T., Patricia L. Mokhtarian, Ilan Salomon, Lothlorien S. Redmond, Gustavo O.
Collantes, and Sangho Choo ( 2004) When is commuting desirable to the individual? Growth and
Change 35( 3) ( Summer), special issue on Advances in Commuting Studies, Peter Nijkamp and Jan
Rouwendal, eds.
ii
Redmond, Lothlorien S. and Patricia L. Mokhtarian ( 2001) The positive utility of the commute:
Modeling ideal commute time and relative desired commute amount. Transportation 28( 2) ( May),
179- 205.
Salomon, Ilan and Patricia L. Mokhtarian ( 1999) Travel for the fun of it. Access ( a publication of
the University of California Transportation Center) 15 ( Fall), 26- 31. Available at
www. uctc. net/ access/ access15. pdf or ( without graphics) .../ access15lite. pdf.
Salomon, Ilan and Patricia L. Mokhtarian ( 1998) What happens when mobility- inclined market
segments face accessibility- enhancing policies? Transportation Research D 3( 3), 129- 140.
Salomon, Ilan and Patricia L. Mokhtarian ( 2002) Driven to travel: The identification of mobility-inclined
market segments. Chapter 22 in William R. Black and Peter Nijkamp, eds., Social Change
and Sustainable Transport. Bloomington, IN: Indiana University Press, pp. 173- 179. Included in
the Regional Futures Compendium of the Capital Region Institute ( Valley Vision), Sacramento,
California.
Schwanen, Tim and Patricia L. Mokhtarian ( forthcoming) The extent and determinants of disson-ance
between actual and preferred residential neighborhood type. Environment and Planning B.
Schwanen, Tim and Patricia L. Mokhtarian ( 2003a) Does dissonance between desired and current
neighborhood type affect individual travel behaviour? An empirical assessment from the San
Francisco Bay Area. Proceedings of the European Transport Conference ( ETC), October 8- 10,
2003, Strasbourg, France.
Schwanen, Tim and Patricia L. Mokhtarian ( 2003b) The role of attitudes toward travel and land
use in residential location behavior: Some empirical evidence from the San Francisco Bay Area.
Under review for publication.
Schwanen, Tim and Patricia L. Mokhtarian ( 2003c) What affects commute mode choice: Neigh-borhood
physical structure or preferences toward neighborhoods? Under review for publication.
Reports Produced by this Project to Date
Cao, Xinyu and Patricia L. Mokhtarian ( 2003) Modeling the Individual Consideration of
Travel- Related Strategies. Research Report UCD- ITS- RR- 03- 3, Institute of Transportation
Studies, University of California, Davis, June.
Available at www. its. ucdavis. edu/ publications/ 2003/ RR- 03- 3. pdf
Choo, Sangho and Patricia L. Mokhtarian ( 2002) The Relationship of Vehicle Type Choice to
Personality, Lifestyle, Attitudinal, and Demographic Variables. Research Report, Institute of
Transportation Studies, University of California, Davis, October.
Available at www. its. ucdavis. edu/ publications/ 2002/ RR- 02- 06. pdf.
iii
Choo, Sangho and Patricia L. Mokhtarian ( 2004) Modeling the Consideration of Travel- Related
Strategy Bundles. Research Report, Institute of Transportation Studies, University of California,
Davis, March.
Choo, Sangho, Gustavo O. Collantes, and Patricia L. Mokhtarian ( 2001) Modeling Individuals'
Relative Desired Travel Amounts. Research Report UCD- ITS- RR- 01- 13, Institute of Transpor-tation
Studies, University of California, Davis, November.
Available at www. its. ucdavis. edu/ publications/ 2001/ RR- 01- 13. pdf.
Clay, Michael J. and Patricia L. Mokhtarian ( 2002) The Adoption and Consideration of
Commute- Oriented Travel Alternatives. Research Report UCD- ITS- RR- 02- 04, Institute of
Transportation Studies, University of California, Davis, September.
Available at www. its. ucdavis. edu/ publications/ 2002/ RR- 02- 04. pdf.
Collantes, Gustavo O. and Patricia L. Mokhtarian ( 2002) Determinants of Subjective Assessments
of Personal Mobility. Research Report, Institute of Transportation Studies, University of Califor-nia,
Davis, August.
Curry, Richard W. ( 2000) Attitudes toward Travel: The Relationships among Perceived
Mobility, Travel Liking, and Relative Desired Mobility. Master’s Thesis, Department of Civil
and Environmental Engineering, University of California, Davis, June. Research Report UCD-ITS-
RR- 00- 06, Institute of Transportation Studies, University of California, Davis.
Available at www. its. ucdavis. edu/ publications/ 2000/ RR- 00- 06. pdf
Ory, David T. and Patricia L. Mokhtarian ( 2004) Who Likes Traveling? Models of the
Individual’s Affinity for Various Kinds of Travel. Research Report UCD- ITS- RR- 04- xx,
Institute of Transportation Studies, University of California, Davis, April.
Redmond, Lothlorien S. and Patricia L. Mokhtarian ( 2001) Modeling Objective Mobility: The
Impact of Travel- Related Attitudes, Personality, and Lifestyle on Distance Traveled. Research
Report UCD- ITS- RR- 01- 09, Institute of Transportation Studies, University of California, Davis,
June. Available at http:// repositories. cdlib. org/ itsdavis/ UCD- ITS- RR- 01- 09/
Redmond, Lothlorien S. ( 2000) Identifying and Analyzing Travel- related Attitudinal, Per-sonality,
and Lifestyle Clusters in the San Francisco Bay Area. Master’s Thesis, Transportation
Technology and Policy Graduate Group, Institute of Transportation Studies, University of
California, Davis, September. Research Report UCD- ITS- RR- 00- 08.
Available at www. its. ucdavis. edu/ publications/ 2000/ RR- 00- 08. pdf
iv
TABLE OF CONTENTS
LIST OF TABLES....................................................................................................................... vi
LIST OF FIGURES.................................................................................................................... vii
EXECUTIVE SUMMARY ....................................................................................................... viii
1. INTRODUCTION................................................................................................................... 1
2. DATA DESCRIPTION ........................................................................................................... 4
2.1 DATA COLLECTION ............................................................................................................... 4
2.2 TRAVEL- RELATED STRATEGIES............................................................................................. 6
2.2.1 Individual Strategies ..................................................................................................... 6
2.2.2 Strategy Bundles ........................................................................................................... 9
2.3 KEY EXPLANATORY VARIABLES ........................................................................................ 12
2.4 GENERAL HYPOTHESES....................................................................................................... 16
3. DESCRIPTIVE RELATIONSHIPS BETWEEN ADOPTION AND CONSIDERATION
............................................................................................................................... ....................... 23
3.1 DISTRIBUTION OF ADOPTION AND CONSIDERATION OF STRATEGY BUNDLES ...................... 23
3.2 DESCRIPTIVE ANALYSES OF ADOPTION AND CONSIDERATION OF STRATEGY BUNDLES ...... 28
3.2.1 Correlation Test of Adoption and Consideration ....................................................... 30
3.2.2 Clustering on Travel Attitudes.................................................................................... 32
3.2.3 Comparison of Adoption and Consideration between the Clusters............................ 33
4. MODELING THE CONSIDERATION OF STRATEGY BUNDLES............................. 37
4.1 GENERAL MODEL SPECIFICATION ISSUES ........................................................................... 37
4.2 CONCEPTUAL STRATEGY BUNDLES..................................................................................... 42
4.2.1 Travel maintaining/ increasing strategies ................................................................... 46
4.2.2 Travel reducing strategies .......................................................................................... 48
4.2.3 Major location/ lifestyle change strategies.................................................................. 52
4.3 FACTOR- BASED STRATEGY BUNDLES.................................................................................. 56
v
4.3.1 Auto improvement strategies....................................................................................... 64
4.3.2 Mobile phone .............................................................................................................. 72
4.3.3 Work- schedule change ................................................................................................ 74
4.3.4 Hire someone to do house or yard work..................................................................... 78
4.3.5 Mode Change .............................................................................................................. 82
4.3.6 Home- based work ....................................................................................................... 87
4.3.7 Residential/ employment relocation............................................................................. 92
4.3.8 Alter employment status.............................................................................................. 97
5. SUMMARY AND CONCLUSIONS.................................................................................. 102
5.1 SUMMARY ......................................................................................................................... 102
5.2 CONCLUSIONS ................................................................................................................... 107
ACKNOWLEDGEMENTS ..................................................................................................... 110
REFERENCES..................................................................................................................... .... 111
vi
LIST OF TABLES
ES- 1: Summary of Models of Consideration of Conceptual Strategy Bundles........................... xv
ES- 2: Summary of Models of Consideration of Factor- based Strategy Bundles ....................... xvi
Table 2.1: Socio- demographic Characteristics of the Sample Used in this Analysis.................... 5
Table 2.2: Distribution of Bundle Adoption and Consideration ( N = 1,283).............................. 10
Table 2.3: General Hypotheses .................................................................................................... 22
Table 3.1: Adoption and Consideration of Combinations of Conceptual Strategy Bundles
( N= 1283).......................................................................................................................... ..... 27
Table 3.2: Cross- tabulation of Adoption and Consideration Pairs for Factor- based Strategy
Bundles ( Adopters Only) ...................................................................................................... 29
Table 3.3: Correlation between Adoption and Consideration of Conceptual Strategy Bundles
( N= 1283).......................................................................................................................... .... 30
Table 3.4: Correlation between Adoption and Consideration of Factor- based Strategy Bundles
( N= 1283).......................................................................................................................... .... 32
Table 3.5: Description of Clusters for Travel Satisfaction .......................................................... 33
Table 3.6: Correlations between Adoption and Consideration of Conceptual Strategy Bundles
( Satisfied and Unsatisfied Groups) ....................................................................................... 34
Table 3.7: Correlations between Adoption and Consideration of Factor- based Strategy Bundles
( Satisfied and Unsatisfied Groups) ....................................................................................... 36
Table 4.1: Summary of Models of Consideration of Conceptual Strategy Bundles …………… 43
Table 4.2: Model of Consideration of the Travel Maintaining/ Increasing Bundle...................... 47
Table 4.3: Model of Consideration of Travel Reducing Bundle ( all respondents)...................... 49
Table 4.4: Model of Consideration of Travel Reducing Bundle ( non- adopters only)................. 52
Table 4.5: Model of Consideration of Major Location/ Lifestyle Change Bundle ( all respondents)
............................................................................................................................... ........... 53
Table 4.6: Model of Consideration of Major Location/ Lifestyle Change Bundle ( non- adopters
only) ............................................................................................................................... .. 55
Table 4.7: Summary of Models of Consideration of Factor- based Strategy Bundles ................. 62
Table 4.8: Model of Consideration of Auto Improvement ( all respondents) .............................. 66
vii
Table 4.9: Model of Consideration of Auto Improvement ( only non- adopters).......................... 69
Table 4.10: Model of Consideration of Car- replacement Bundle................................................ 71
Table 4.11: Model of Consideration of Getting a Mobile Phone ................................................ 73
Table 4.12: Model of Consideration of Work- Schedule Change ( all respondents)..................... 75
Table 4.13: Model of Consideration of Work- Schedule Change ( only non- adopters)................ 77
Table 4.14: Model of Consideration of Hiring Somebody to Do House or Yard Work ( all
respondents) ...................................................................................................................... 79
Table 4.15: Model of Consideration of “ Hire Somebody to Do House or Yard Work” ( non-adopters
only).................................................................................................................... 81
Table 4.16: Model of Consideration of Mode Change ( all respondents) .................................... 84
Table 4.17: Model of Consideration of Mode Change ( only non- adopters)................................ 86
Table 4.18: Model of Consideration of Home- based Work ( all respondents)............................. 88
Table 4.19: Model of Consideration of Home- based Work ( only non- adopters)........................ 91
Table 4.20: Model of Consideration of Residential/ Employment Relocation Bundle ( all
respondents) ...................................................................................................................... 93
Table 4.21: Model of Consideration of Residential/ employment Relocation Bundle ( only non-adopters)..................................................................................................................
......... 97
Table 4.22: Model of Consideration of Altering Employment Status Bundle ( all respondents). 99
Table 4.23: Model of Consideration of Altering Employment Status Bundle ( only non- adopters)
............................................................................................................................... ......... 100
Table 5.1: Comparison of Initial Hypotheses and Selected Results .......................................... 104
LIST OF FIGURES
Figure 2.1: Section E1 ( Adoption) from the Survey...................................................................... 7
Figure 2.2: Section E2 ( Consideration) from the Survey .............................................................. 8
Figure 2.3: Conceptual and Factor- based Bundles of the Travel- related Alternatives................ 11
Figure 3.1: Adoption and Consideration of Conceptual Strategy Bundles.................................. 24
viii
EXECUTIVE SUMMARY
For the last three decades, policy makers and transportation planners have devised a series of
policy instruments to tackle traffic congestion, starting with supply and demand controls.
Transportation Systems Management ( TSM) and Transportation Demand Management ( TDM)
programs are well- known classes of such policy strategies. Although many of these strategies
have been implemented, they have failed to reduce traffic congestion. One of the reasons for this
failure is that there is often a discrepancy, sometimes large, between the responses to congestion
that are assumed by policy makers and those that are actually adopted by individuals. This
mismatch in behavioral responses makes policies less effective, and needlessly consumes large
amounts of time and money in their trial- and- error implementation.
As one of a series of studies on individuals’ adoption and consideration of travel- related
strategies in response to congestion, this study explores the relationships between the adoption
and consideration of bundles of travel- related strategies by identifying characteristics associated
with patterns of adoption and consideration among bundles, and by developing discrete choice
( binary logit) models for individuals’ consideration of each bundle. In particular, we focus on
whether the adoption of lower- cost, short- term strategies significantly and/ or dynamically ( using
time since adoption variables) affects the consideration of higher- cost, longer- term ones. We also
investigate whether individuals with a high liking for travel, indicative of a positive utility of
travel, are resistant to higher- cost, longer- term travel- reduction strategies.
The data for this study were collected from a fourteen- page survey returned by about 1,900 adult
residents of three distinct San Francisco Bay area neighborhoods in May 1998: Concord and
Pleasant Hill represent suburban neighborhoods, and an area defined as North San Francisco
represents an urban neighborhood. The subset of 1,283 cases used in this study constitutes those
respondents identified as workers ( either part- time or full- time) who commute at least once a
month and have relatively complete responses to key questions.
From the initial study in this series, the 17 main travel- related strategies on the survey were
grouped into two sets of strategy bundles, based on conceptual and empirical similarities,
ix
respectively. The first set ( conceptual bundles) consists of three bundles that were conceptually
classified based on the generalized cost and the amount of lifestyle change for each: travel
maintaining/ increasing, travel reducing, and major location/ lifestyle change. The second set
( factor- based bundles) comprises eight bundles ( including two with only one strategy each) that
were identified by factor- analyzing the responses: auto improvement, mobile phone, work-schedule
changes, hire someone to do house or yard work, mode change, home- based work,
residential/ employment relocation, and alter employment status.
Based on these two sets of bundles, we first identified patterns of adoption and consideration
among bundles, using correlation tests. Specifically, we examined whether previous adoption is
significantly related to current consideration, and whether those relationships are different
between groups who are satisfied and unsatisfied with their current travel conditions. The highest
correlations are found in most pairs of adoption and consideration of the same bundle ( all
conceptual bundles and six of the factor- based bundles), indicating that the same or similar
strategies are likely to be considered/ adopted repeatedly throughout an individual’s life.
Additionally, the correlations of adoption and consideration have similar patterns in both
satisfied and unsatisfied groups with current travel conditions, showing that the previous
adoption is strongly associated with current consideration, more or less independently of
satisfaction with current conditions.
Furthermore, we developed discrete choice models ( binary logit models) for individuals’
consideration of each bundle in the two sets. Tables ES- 1 ( Table 4.1 in the text) and ES- 2 ( Table
4.7 in the text) summarize the significant variables in the models of conceptual and factor- based
bundles, respectively, with positive and negative signs indicating the direction of effect for each
variable. The ρ2 values of the conceptual bundle models ranged from 0.106 to 0.210, and those
of the factor- based bundle models ranged from 0.103 to 0.434. All models are significantly
better than the corresponding market share model at α << 0.001. Additionally, models of
consideration of each bundle based on non- adopters were developed for all except two bundles
( due to small sample sizes and unbalanced shares), the travel maintaining/ increasing and mobile
phone strategies. The models based on non- adopters have higher ρ2 values, ranging from 0.151
( 0.291) to 0.311 ( 0.625) for the conceptual ( factor- based) bundles. That is, the models on non-
x
adopters can explain more information in the data by eliminating the potentially heterogeneous
adopters ( for whom the previously- adopted strategy may or may not still be in force) and the
potentially opposite effects of some variables between adopters and non- adopters. As expected,
some variables in the models for non- adopters are common to the ones for the full data set, and
other variables are similar. Not surprisingly, compared to the conceptual bundle models, the
factor- based bundle models have more diverse explanatory variables and better goodness of fit
because the factor- based bundles are more finely subdivided than the conceptual ones. We
briefly summarize the key findings:
Most Objective Mobility variables are positively associated with consideration of travel- related
strategy bundles. This is consistent with our hypothesis that the higher the amount of travel the
individual does, the more likely she is to consider travel- related strategy bundles, as opposed to
doing nothing. Similar to Objective Mobility, most Subjective Mobility variables are positively
related to the consideration of the bundles. That is, the more travel the individual perceives
doing, the more likely she is to consider travel- related strategy bundles.
As hypothesized, Relative Desired Mobility variables have logically either positive or negative
effects on consideration of travel- related strategy bundles. For example, those who want to
increase commute or work travel are less likely to consider travel reducing and major
location/ lifestyle change bundles ( such as mode change and residential/ employment relocation),
whereas people with a higher desire for discretionary travel are more likely to consider them. It
is plausible that the Relative Desired Mobility variables for modes other than driving ( e. g. bus)
have negative effects on consideration of the travel maintaining/ increasing bundle.
As an indicator of a positive utility of travel, Travel Liking for long- distance personal vehicle
travel is positively related to consideration of the travel maintaining/ increasing strategy bundle,
and that for work travel is negatively associated with travel reducing and major location/ lifestyle
change bundles. These results support the idea that a positive utility of travel will motivate
people to keep or increase their current travel.
xi
Among the six Travel Attitude variables, only two are significant, collectively appearing in one
of the conceptual strategy bundle models and four of the factor- based bundle models. Logically,
pro- environmentalists are more likely to consider the travel reducing and major location/ lifestyle
change bundles ( including work- schedule change, mode change, and home- based work). On the
other hand, the individual with a higher commute benefit factor score is less likely to consider
travel reducing and major location/ lifestyle change bundles ( such as work- schedule change and
residential/ employment relocation).
Three of the four Personality factor variables are significant, collectively influencing the
consideration of one of the conceptual strategy bundles and three of the factor- based bundles.
Adventure seekers are more likely to consider commute travel reducing and major
location/ lifestyle change bundles ( such as work- schedule change and home- based work) in order
to free more time, money, and energy for adventure travel. Interestingly, loners and calm people
are less likely to consider travel reducing ( such as mode change) and major location/ lifestyle
change bundles, presumably for different but logical reasons. However, the organizer variable
did not turn out to be significant in any model.
Three of the four Lifestyle factor variables are positively associated with medium- to- high- cost
strategy bundles ( one of the conceptual strategy bundles and four of the factor- based bundles).
Frustrated people are more likely to consider the travel reducing and major location/ lifestyle
change bundles ( such as residential/ employment relocation and home- based work). Clearly,
family/ community- oriented people have a greater tendency to consider the travel reducing and
major location/ lifestyle change bundles. Similar to the organizer Personality, the workaholic
Lifestyle factor was not significant in any of the models. As expected, social status seekers are
more likely to consider the travel maintaining/ increasing bundle ( such as hiring domestic help).
As hypothesized, as a marker of preference for discretionary travel, the excess travel indicator is
positively associated with the consideration of the travel reducing and major location/ lifestyle
change bundles ( such as residential/ employment relocation and home- based work).
Mobility Constraint variables are positively associated with all three of the conceptual strategy
bundles, and five of the factor- based bundles. The individual who has limitations on driving,
xii
riding a bicycle, or vehicle availability is more likely to consider either the travel reducing and
major location/ lifestyle change bundles, or the travel maintaining one if travel is necessary.
Socio- demographic variables with respect to gender, age, household, income, and occupation are
significantly related to travel- related strategy bundles. Especially, age or number of years lived
in the U. S. ( a proxy for age) is negatively related to consideration of both the travel maintaining
and travel reducing strategies ( including two of the conceptual strategy bundles and seven of the
factor- based bundles). This suggests that younger people are more likely than older ones to
consider the lower- cost strategies against congestion, either maintaining more comfortably ( if
necessary) or reducing ( if possible) their travel. On the other hand, people in a high- income
household are more likely to consider strategies in the travel maintaining/ increasing bundle ( such
as auto improvement and hiring domestic help) but less likely to consider the travel reducing
strategy bundle. In addition, managers or administrators are positively inclined to consider the
travel maintaining/ increasing and travel reducing ( such as home- based work) bundles, while
clerical workers are more likely to consider the major location/ lifestyle change bundle ( such as
alter employment status). Interestingly, the vehicle type variable is significantly related to
consideration of the travel reducing and major location/ lifestyle change bundles. Specifically,
those who drive SUVs most often are less likely to consider the travel reducing strategy bundle
( including mode change and residential/ employment relocation), suggesting an enjoyment of
driving. Focusing on household members, people living with younger children ( under six) or
older people ( ages 65- 74) are, not surprisingly, more likely to consider the major loca-tion/
lifestyle change strategy bundle ( including alter employment status).
As hypothesized, the previous adoption of any individual strategies in a bundle positively affects
consideration of the same bundle. This indicates that the individual who previously adopted a
given strategy is more likely than others to seek either the same or another strategy in the same
bundle. Similar to the previous study, the previous adoption of lower- cost individual strategies
positively affects the consideration of the higher- cost strategy bundles, and the previous adoption
of higher- cost individual strategies positively affects consideration of lower- cost strategy
bundles.
xiii
In addition, time since adoption variables are significantly associated with consideration of
travel- related strategy bundles, with logical signs. For example, the longer ago the individual
adopted getting a better car and changing from another means to driving alone, the more likely
she is to consider the auto improvement bundle. On the other hand, the more recently the
individual adopted changing work trip departure time or hiring domestic help, the more likely
she is to consider the corresponding strategy bundles ( such as travel maintaining/ increasing
bundles), presumably to continue or resume enjoying their benefits. Interestingly, the auto
improvement bundle is more affected by the time- dependent adoption of individual strategies
than the other bundles due to the inevitable decay in the utility of a particular auto with time and
frequent use.
In modeling individuals’ consideration of travel- related strategy bundles, we found significant,
diverse variables ( such as qualitative and quantitative Mobility- related variables, Travel
Attitudes, Personality, Lifestyle, and Travel Liking), most of which have been little considered in
establishing transportation policy strategies to reduce traffic congestion. First, individuals’
subjective assessment of the amount of their travel and desire for more or less travel, play key
roles in considering which type of strategy can satisfy their travel needs. Second, Travel Liking,
representing a positive utility of travel, turns out to be resistant to strategies that could reduce
congestion. In other words, this factor can motivate individuals to maintain or increase their
current travel. Lastly, individuals’ Travel Attitudes, Personality, and Lifestyle also affect their
consideration of travel- related strategies either positively or negatively.
In addition, a couple of relationships between previous adoption and consideration of travel-related
strategy bundles can be identified in the models. The previous adoption of any individual
strategies in a bundle strongly positively affects the consideration of the same bundle, showing
an inertial or habitual response toward travel- related strategies. It suggests that a new
transportation policy at a different level may be less likely to be considered by individuals who
have never adopted it or a similar one. On the other hand, the previous adoption of any
individual strategies in a bundle can significantly increase the consideration of either lower- or
higher- cost strategy bundles, showing an unstable or cycling response toward travel- related
strategies. It is natural that individuals keep seeking a better strategy at a different time or cost
xiv
level to improve their current travel conditions, although this relationship is less often found in
our models than the former ( reconsideration of the same bundle). Further, time since adoption
variables can partially explain the dynamic nature of individuals’ responses to travel- related
strategy bundles. That is, depending on the type of travel- related strategy in a bundle, an
individual who adopted it longer ago is more ( or less) likely to consider the same bundle or
another bundle. As a general comment, it should be kept in mind that Clay and Mokhtarian
( forthcoming) found that the respondents adopted or are considering individual strategies for a
variety of reasons other than travel, although we interpreted the relationships between adoption
and consideration from the transportation point of view.
Overall, the results of this study give policy makers and planners insight into understanding the
dynamic nature of individuals’ responses to travel- related strategies as well as differences
between the responses to congestion that are assumed by policy makers and those that are
actually adopted by individuals. Our study, however, focused on individuals’ responses to the
travel- related strategy bundles ( i. e., disaggregate behaviors, not aggregate). It would be very
useful to develop aggregate approaches to explaining the Travel Attitudes, Personality, Lifestyle,
and qualitative Mobility variables that are significant in this study, to support the development
and evaluation of more effective transportation policies for reducing traffic congestion and/ or
improving mobility.
xv
ES- 1: Summary of Models of Consideration of Conceptual Strategy Bundles
Travel
maintaining/
increasing
Travel
reducing
Major
location/ life-style
change
N 1259 1220 1277
MS ρ2 0.159 0.106 0.032
ρ2
0.210 0.201 0.106
Adjusted ρ2 0.194 0.184 0.091
Variable
Objective Mobility
Frequency of commuting ( SD) +
Weekly miles to eat a meal ( SD) + +
Weekly miles by walking/ jogging/ bicycling ( SD) +
Total trips ( LD) +
Subjective Mobility
Take others where they need to go ( SD) +
Travel by personal vehicle ( SD) + +
Relative Desired Mobility
Travel by walking/ jogging/ bicycling ( SD) -
Travel by air ( LD) +
Travel Liking
Travel by personal vehicle ( LD) +
Attitudes
Pro- environmental solutions factor score +
Personality
Adventure seeker factor score +
Lifestyle
Frustrated factor score +
Family & community- oriented factor score +
Mobility Constraints
Limitations on driving during the day + +
Socio- demographics
Years lived in the U. S. - -
Manager/ administrator occupation +
Household income category -
Number of people ages under 6 in HH +
Number of people ages 65- 74 in HH +
Strategy Adoption
Buy a mobile phone -
Time since getting a fuel efficient car +
Change work trip departure time + +
Time since changing work trip departure time +
Hire somebody to do house or yard work +
Time since hiring domestic help -
Adopt compressed work week +
Change from another means to driving alone +
Buy equipment to help work from home + +
Work part- instead of full- time +
Start home- based business + +
Retire or stop working +
Major location/ lifestyle change +
Notes: SD = Short Distance, LD = Long Distance.
Shaded cells denote significant relationships between consideration of one bundle and prior adoption of strategies in
the same bundle.
xvi
ES- 2: Summary of Models of Consideration of Factor- based Strategy Bundles
Bundles
Explanatory Variables
Auto improvement
Mobile phone
Work- schedule change
Hire someone to do
house or yard work
Mode change
Home- based work
Residential/ employmen
t relocation
Alter employment
status
N 1146 1263 1204 1238 1203 1241 1222 1261
MS ρ2 0.043 0.124 0.155 0.219 0.434 0.147 0.316 0.207
ρ2
0.103 0.202 0.246 0.318 0.519 0.248 0.386 0.262
Adjusted ρ2 0.083 0.184 0.226 0.304 0.498 0.229 0.367 0.249
Objective Mobility
Frequency of commuting ( SD) +
Frequency of work/ school- related travel ( SD) +
Frequency of grocery shopping travel ( SD) +
Frequency of travel taking others where they need
to go ( SD) +
Total weekly miles ( SD) +
Weekly miles of grocery shopping travel ( SD) -
Weekly miles to eat a meal ( SD) + + +
Weekly miles of entertainment travel ( SD) +
Weekly miles of travel taking others where they
need to go ( SD) -
Weekly miles by train/ BART/ light rail ( SD) +
Weekly miles by walking/ jogging/ bicycling ( SD) -
Commute distance +
Travel miles by personal vehicle ( LD) +
Sum of log of miles for each trip by air ( LD) +
Subjective Mobility
Commute ( SD) + +
Travel for grocery shopping ( SD) + +
Travel for eating a meal ( SD) -
Travel for entertainment ( SD) +
Take others where they need to go ( SD) +
Travel by personal vehicle ( SD) + +
Travel by air ( LD) -
Relative Desired Mobility
Commute ( SD) -
Work/ school- related travel ( SD) -
Travel for grocery shopping ( SD) -
Travel for entertainment ( SD) +
Travel by personal vehicle ( SD) -
Travel by bus ( SD) -
Travel by train/ BART/ light rail ( SD) +
Travel by walking/ jogging/ bicycling ( SD) +
Travel by personal vehicle ( LD) -
Travel Liking
Work/ school- related travel ( SD) -
Travel for eating a meal ( SD) +
Travel by train/ BART/ light rail ( SD) +
Overall ( LD) +
SD = Short Distance LD = Long Distance
xvii
( ES 2 continued)
Auto improvement
Mobile phone
Work- schedule change
Hire someone to do
house or yard work
Mode change
Home- based work
Residential/ employment
relocation
Alter employment status
Attitudes
Pro- environmental solutions factor score + + +
Commute benefit factor score - -
Ideal commute time +
Personality
Adventure seeker factor score + +
Loner factor score -
Calm factor score -
Lifestyle
Frustrated factor score + +
Family & community- oriented factor score +
Status seeker factor score +
Excess Travel
Excess travel indicator + +
Mobility Constraints
Limitations on driving during the day + +
Limitations on driving on the freeway +
Limitations on riding a bicycle +
Percent of time a vehicle is available + -
Socio- demographics
Time living in the neighborhood +
Age -
Female +
Year of personal vehicle - -
Vehicle type is SUV - -
Years lived in the U. S. - + - - - +
Total workers in the household -
Full- time worker +
Manager/ administrator occupation +
Production/ construction/ craft occupation -
Clerical/ administrative support occupation +
Anyone in household needing special care + +
Personal income category + +
Number of people ages 6- 15 in HH -
Number of people ages 41- 64 in HH +
Number of people ages 65- 74 in HH +
Household with single adult -
Household with two or more adults -
SD = Short Distance LD = Long Distance
xviii
( ES 2 continued)
Auto improvement
Mobile phone
Work- schedule change
Hire someone to do
house or yard work
Mode change
Home- based work
Residential/ employment
relocation
Alter employment status
Strategy Adoption
Buy a car stereo system +
Get a better car - +
Time since getting a better car +
Buy a mobile phone -
Change work trip departure time +
Time since changing work trip departure time -
Adopt flextime +
Adopt compressed work week +
Hire somebody to do house or yard work +
Time since hiring domestic help - -
Change from driving alone to other means + +
Change from another means to driving alone +
Squared time since changing from another means
to driving alone +
Buy equipment to help work from home +
Telecommute +
Start home- based business + +
Change jobs closer to home +
Time since changing jobs closer to home -
Work part- instead of full- time +
Time since retiring or stopping working +
Work- schedule change bundle +
Alter employment status bundle +
SD = Short Distance LD = Long Distance
1
1. INTRODUCTION
Today more than two hundred million vehicles operate on highways in the U. S., and annual
vehicle miles traveled ( VMT) is more than 2.5 trillion. Traffic congestion has become a
common feature of everyday life in metropolitan areas, resulting in high social costs ( Arnott and
Small, 1994; Downs, 1992; Hanks and Lomax, 1991; The Economist, 1998). The costs of lost
time and extra fuel consumption caused by congestion were estimated to be as high as $ 78
billion in 2000, an increase of 39% over those in 1990 ( U. S. News & World Report, 2001).
For the last three decades, policy makers and transportation planners have devised a series of
policy instruments to tackle traffic congestion, starting with supply and demand controls.
Transportation Systems Management ( TSM) and Transportation Demand Management ( TDM)
programs are well- known classes of such policy strategies. A number of studies ( e. g. Downs,
1992; Giuliano and Small, 1995) have also proposed market- based pricing policies such as
congestion pricing, undergirded by the concept that users of a particular transportation facility
should pay the costs they impose on others. In addition, promoting the use of information and
communication technology ( ICT) substitutes for travel, such as telecommuting, has been
proposed as a strategy for reducing congestion ( e. g. Niles, 1994; US DOT, 1993).
Although many of these strategies have been implemented, they have failed to reduce traffic
congestion. A number of reasons have been offered for this failure. The literature on induced
demand ( e. g. Noland, 2001) argues that improved highway capacity can stimulate auto travel,
resulting in the increase of travel demand. With respect to ICT applications, substitution of
telecommunications for travel is the impact most desired from a public policy perspective, but
ICT may also have a complementary relationship to travel − generating more, on net
( Mokhtarian, 2002). These arguments suggest that there is a discrepancy, sometimes large,
between the responses to congestion that are assumed by policy makers and those that are
actually adopted by individuals. This mismatch in behavioral responses makes policies less
effective, and needlessly consumes large amounts of time and money in their trial- and- error
implementation. Giuliano ( 1992) pointed out that TDM strategies are less likely to be effective
2
without understanding individuals’ current travel behavior and preferences, from which derives
the public or political acceptability of those strategies.
Pursuant to the aim of improving our understanding of individuals’ behavior and attitudes,
Salomon and Mokhtarian ( 1997) developed a conceptual model of the behavioral response to
congestion, that incorporates the dynamics of the decision process for individuals’ choices
adjusted by costs and benefits from their previous experiences. In a subsequent empirical study,
Mokhtarian, et al. ( 1997) identified rank- based ( travel maintaining, travel reducing, and major
location/ lifestyle change) and factor- based ( auto improvement, departure time, work schedule
change, remote work, relocation, and work/ lifestyle change) tiers for a set of coping strategies
ranging from lower- cost to higher- cost, and short- term to longer- term, using rank ordering and
factor analysis, respectively. This study used data collected from 621 employees of the City of
San Diego, California in 1992. More recently, Raney, et al. ( 2000) estimated binary logit models
of the consideration of each of 15 congestion- response strategies using the same data, and found
that individuals are likely to change their responses to congestion from lower- cost, short- term
strategies to higher- cost, long- term ones when dissatisfaction remains. They also pointed out
that besides travel- related variables, various non- travel- related motivations and constraints affect
individuals’ responses.
As a sequel to the above research, a series of studies on a newer set of data explores relationships
between adoption and consideration of 17 travel- related strategies, linking them to mobility-related,
travel attitudes, personality, lifestyle, travel liking, socio- demographic, and other
variables. The first report in this series ( Clay and Mokhtarian, 2002) presented descriptive
analyses of relationships of these variables to the adoption and consideration of each individual
strategy and bundle of strategies. The second report in this series ( Cao and Mokhtarian, 2003)
developed binary logit models for the consideration of each individual strategy, taking the
adoption and time since adoption of each strategy as potential explanatory variables among
others.
Similarly, in this study, we explore the relationships between the adoption and consideration of
bundles of travel- related strategies by identifying characteristics associated with patterns of
3
adoption and consideration among bundles, and by developing discrete choice ( binary logit)
models for individuals’ consideration of each bundle. The adoption and time since adoption for
individual or bundles of strategies are included as explanatory variables in the models. In
particular, we focus on whether the adoption of lower- cost, short- term strategies significantly
and/ or dynamically ( using time since adoption variables) affects the consideration of higher- cost,
longer- term ones. We also investigate whether individuals with a high liking for travel,
indicative of a positive utility of travel, are resistant to higher- cost, longer- term travel- reduction
strategies. The data for this study were collected from a fourteen- page survey returned by about
1,900 adult residents of three distinct San Francisco Bay area neighborhoods in May 1998; the
current analysis is based on a subset of nearly 1,300 commuting workers. This study will give
policy makers and planners insight into the dynamic nature of individuals’ responses to travel-related
strategies, and help them to improve on the currently available strategies.
This report consists of five sections. The following section describes the data for this study,
explains key types of variables measured by the survey and used in this study, and suggests some
hypotheses to be tested by this study. Section 3 presents the correlations between adoption and
consideration of strategy bundles. Section 4 discusses the binary logit model results of
consideration of strategy bundles, focusing on the significant variables in the models. In the final
section, we summarize the results and suggest policy recommendations.
4
2. DATA DESCRIPTION
2.1 Data Collection
The data for this study come from a fourteen- page self- administered survey mailed in May 1998
to 8,000 randomly- selected households in three neighborhoods of the San Francisco Bay Area:
Concord and Pleasant Hill represent suburban neighborhoods, and an area defined as North San
Francisco represents an urban neighborhood. North San Francisco has more mixed land uses,
higher residential density, and a more grid- like street system compared to the suburban
examples. On the other hand, Concord has more segregated land uses and lower residential
density. Pleasant Hill was selected to represent another part of the spectrum of suburban
neighborhoods. Compared to Concord, Pleasant Hill has greater residential density, indicating
fewer single- family households.
Half of the surveys were sent to North San Francisco, and Concord and Pleasant Hill received
2,000 surveys each. Approximately 2,000 surveys were completed by a randomly- selected adult
member of the household and returned, for a 25% response rate. The subset of 1,283 cases used
in this analysis constitutes those respondents identified as workers ( either part- time or full- time)
who commute at least once a month and have relatively complete responses to key questions.
Table 2.1 presents some key socio- demographic characteristics of the study data. The sample is
relatively balanced in terms of representation by neighborhood and gender. Nearly 95% of
respondents have one or more personal vehicles in their households. Higher incomes are
overrepresented compared to Census data, as is typical for self- administered questionnaires.
The survey consists of six sections: “ Your Opinions about Travel” ( Section A), “ Your Lifestyle
as it Relates to Travel” ( B), “ The Amount You Travel” ( C), “ How You View Your Travel” ( D),
“ Your Travel- Related Choices” ( E), and “ General Information” ( F). This study mainly focuses
on Section E, which measured the adoption, time since adoption, consideration, and reasons for
adoption and consideration of various travel- related strategies. These variables are discussed in
Section 2.2. The variables from the other sections are classified into 10 categories: Objective
Mobility, Subjective Mobility, Relative Desired Mobility, Travel Liking, Attitudes, Personality,
5
Lifestyle, Mobility Constraints, Excess Travel, and Socio- demographics. These variables are
described in detail in Section 2.3. Section 2.4 presents some hypotheses to be tested by this
study.
Table 2.1: Socio- demographic Characteristics of the Sample Used in this Analysis
Category Frequency Percent
Neighborhood ( N= 1283)
Concord ( suburban) 294 22.9%
Pleasant Hill ( suburban) 346 27.0%
North San Francisco ( urban) 643 50.1%
Gender ( N= 1279)
Female 651 50.9%
Male 628 49.1%
Employment status ( N= 1283)
Full- time worker 1,080 84.2%
Part- time worker 203 15.8%
Age ( N= 1283)
18- 23 42 3.3%
24- 40 563 43.9%
41- 64 640 49.9%
> 65 38 2.9%
Personal income ( N= 1255)
< $ 15,000 91 7.3%
$ 15,000- 34,999 266 21.2%
$ 35,000- 54,999 386 30.8%
$ 55,000- 74,999 229 18.2%
$ 75,000- 94,999 126 10.0%
> $ 95,000 157 12.5%
Family status ( N= 1277)
Single 319 25.0%
2 or more adults, no children 609 47.7%
1 adult with children 28 2.2%
2 or more adults with children 321 25.1%
Number of personal vehicles in HH ( N= 1280)
0 69 5.4%
1 432 33.8%
2 505 39.5%
3 or more 274 21.3%
6
2.2 Travel- related Strategies
2.2.1 Individual Strategies
Section E of the survey comprises two pages of questions referring to travel- related alternatives
that affect the amount of individuals’ travel. Figures 1 and 2 show the original form of the
questions. The questions under E1 asked about the adoption, and E2 about the consideration, of
19 options having travel- related implications. The first column of boxes for each question was
coded as a binary variable, equal to 1 if the box was checked ( i. e. if the alternative was not
adopted), and 0 otherwise. Years since adoption was coded as whole years ( rounded to the
nearest full year, with anything less than 6 months coded as zero). Regarding the reasons for
adoption and consideration, since more than one reason could be indicated, they were coded
separately as binary variables equal to 1 if the reason was checked and 0 otherwise.
Questions “ m” and “ n” had two parts each: “ change jobs . . . closer to home” and “. . . farther
from home” ( referred to as “ m1” and “ m2,” respectively), and “ move your home . . . closer to
work” and “. . . farther from work” (“ n1” and “ n2”). The format for these two questions, shown
in Figures 2.1 and 2.2, was designed to economize on vertical space. Unfortunately, it had the
unanticipated effect of confusing many respondents ( apparently leading them to think that they
needed to respond to only one member of each pair) and resulted in a disproportionately high
number of non- responses, particularly on the second half of each question. The missing data on
the m2 and n2 alternatives for both adoption and consideration ranged from 10% to 17% of the
sample, so we did not use these alternatives to screen out cases with missing data, nor did we
attempt to fill any missing data on these variables.
In previous analyses of these data, cases with missing responses on variables of interest were
either removed or filled; this resulted in 1,904 cases containing relatively complete data for
variables other than the travel- related strategies. Since the travel- related strategies had not been
previously analyzed in depth, it was necessary to review this set of variables for missing data
before proceeding with this study.
7
Figure 2.1: Section E1 ( Adoption) from the Survey
8
Figure 2.2: Section E2 ( Consideration) from the Survey
9
For this study, any case missing more than two out of the 17 responses ( i. e. those other than m2
and n2) for either the adoption or consideration of the travel- related strategies was removed, and
stochastic data filling was used for the remaining missing responses ( see Section 3 of Clay and
Mokhtarian ( 2002) for details). In all, of the 30,328 ( 1,784 respondents × 17 alternatives) total
alternatives analyzed in the adopted section of the travel- related alternatives, responses for 277
or about 0.91% were missing and subsequently filled. For the consideration of strategies,
responses for 248 or about 0.82% were filled. Finally, consistent with the focus of previous
analyses of these data on commuting workers ( in view of the observation that they tend to have
different travel patterns and attitudes than non- commuters or non- workers) cases were removed
if the respondent did not report working part- or full- time and commuting to work at least once a
month. This reduced the final usable data set for this analysis to 1,283 cases.
2.2.2 Strategy Bundles
The initial study in this series ( Clay and Mokhtarian, 2002) grouped the 17 travel- related
strategies into two sets of strategy bundles, based on conceptual and empirical similarities,
respectively. It then related the adoption and consideration of each individual strategy and
bundle of strategies to other variables, by comparing means or frequencies between chooser and
non- chooser groups for adoption or consideration. As mentioned earlier, in this study we treat the
consideration of strategy bundles as dependent variables in discrete choice models, and the prior
adoption of strategy bundles as key explanatory variables. The bundle variables were defined as
1 if any strategy in the bundle had been adopted or considered, respectively, and 0 otherwise.
Here, we briefly summarize the two bundle identification methods ( see Section 6 of Clay and
Mokhtarian ( 2002) for a detailed discussion), with the results shown in Figure 2.3. Also, the
distributions of adoption and consideration with respect to the two sets of strategy bundles
appear in Table 2.2.
10
Table 2.2: Distribution of Bundle Adoption and Consideration ( N = 1,283)
Adoption Consideration
Bundles
Adopted Not adopted Considering Not
considering
Conceptual bundles
Travel maintaining/ increasing 1,184
( 92.3)
99
( 7.7)
926
( 72.2)
357
( 27.8)
Travel reducing 619
( 48.2)
664
( 51.8)
503
( 39.2)
780
( 60.8)
Major location/ lifestyle change 640
( 49.9)
643
( 50.1)
588
( 45.8)
695
( 54.2)
Factor- based bundles
Auto improvement 1,048
( 81.7)
235
( 18.3)
613
( 47.8)
670
( 52.2)
Mobile phone 528
( 41.2)
755
( 58.8)
380
( 29.6)
903
( 70.4)
Work- schedule change 657
( 51.2)
626
( 48.8)
369
( 28.8)
914
( 71.2)
Hire someone to do housework 392
( 30.6)
891
( 69.4)
297
( 23.1)
986
( 76.9)
Mode change 331
( 25.8)
952
( 74.2)
180
( 14.0)
1,103
( 86.0)
Home- based work 474
( 36.9)
809
( 63.1)
471
( 36.7)
812
( 63.3)
Residential/ employment relocation 448
( 34.9)
835
( 65.1)
297
( 23.1)
986
( 76.9)
Alter employment status 239
( 18.6)
1044
( 81.4)
333
( 26.0)
950
( 74.0)
Note: Number in parentheses is the percentage of 1,283.
The first method was to classify the strategies conceptually into three bundles based on the
generalized cost and the amount of lifestyle change for each. Group one includes low
( generalized) cost strategies such as getting a more comfortable car or purchasing a mobile
phone. In general, these are strategies that allow one to maintain travel more pleasantly or
productively, or may even facilitate increasing one’s travel. Group two includes more costly ( in
11
the sense of involving lifestyle changes for the individual or the household) alternatives such as
adopting a compressed workweek or telecommuting. These changes reduce one’s vehicular
travel through reducing the frequency of commuting or changing to shared- ride commute modes.
The third group consists of major location or lifestyle changes such as quitting work, working
part- time instead of full- time and moving home or work closer to the other. These strategies
reduce travel through more drastic means.
Figure 2.3: Conceptual and Factor- based Bundles of the Travel- related Alternatives
Conceptual Bundles Factor- based Bundles
Group 1:
Travel maintaining/ increasing
A. Buy a car stereo system
B. Get a mobile phone
C. Get a better car
D. Get a fuel efficient car
E. Change work trip departure time
F. Hire someone to do house or yard
work
G. Adopt flextime
J. Change from another means of
getting to work to driving alone
Group 2:
Travel reducing
H. Adopt compressed work week
I. Change from driving alone to
work to some other means
K. Buy equipment/ services to help
you work from home
L. Telecommute ( part- or full- time)
Group 3:
Major location/ lifestyle change
M. Change jobs closer to home
N. Move your home closer to work
O. Work part- time instead of full-time
P. Start home- based business or put
more effort into an existing one
Q. Retire or stop working
Group 4: Hire someone to do house or yard work ( F)
Group 2: Mobile phone ( B)
Group 1: Auto improvement ( A C, D)
Group 3: Work- schedule changes ( E, G, H)
Group 5: Mode change ( I, J)
Group 6: Home- based work ( K, L, P)
Group 7: Residential/ employment relocation ( M, N)
Group 8: Alter employment status ( O, Q)
12
The second approach to identifying bundles of strategies was to factor- analyze the responses.
This technique identifies patterns of common variation among a group of variables ( the binary
adoption and consideration variables, in this case), and as such groups the alternatives based on
the empirical similarities in responses to them. Using 36 different factor analyses ( varying the
number of factors extracted, the subsample included and whether adoption and consideration
variables were pooled or not), the strategies were classified into the eight bundles that most
commonly appeared across all the results and conceptually made the most sense. It should be
noted that bundles two and four consist of only one alternative, “ get a mobile phone” and “ hire
someone to do house or yard work”, respectively, in view of their independent factor loadings
and lack of conceptual ( or strong empirical) linkage with the other bundles.
2.3 Key Explanatory Variables
This section describes the key explanatory variables other than those based on the travel- related
strategies, by category: Objective Mobility, Subjective Mobility, Relative Desired Mobility,
Travel Liking, Attitudes, Personality, Lifestyle, Mobility Constraints, Excess Travel, and Socio-demographics.
Among them, the three mobility categories and the Travel Liking category had similar structures.
In each case, measures were obtained both overall and separately by purpose and mode, for
short- distance and long- distance travel. Consistent with the American Travel Survey, long-distance
trips were defined as those longer than 100 miles, one way. The short- distance modes
measured were: personal vehicle, bus, Bay Area Rapid Transit ( heavy rail)/ light rail/ train,
walking/ jogging/ cycling, and other. The short- distance purposes measured were: commuting to
work or school, work/ school- related, grocery shopping, eating a meal, and taking other people
where they need to go. Long- distance measures were obtained for the personal vehicle and
airplane modes, and for the work/ school- related and entertainment/ social/ recreational purposes.
Objective Mobility
These questions asked about distance and frequency of travel by mode and trip purpose, as well
as travel time for the commute trip. For short- distance trips, respondents were asked how often
they traveled for each purpose, with six categorical responses ranging from “ never” to “ 5 or
13
more times a week”. Frequency of trips by mode was not obtained ( a conscious design choice,
to reduce the burden on the respondent). Respondents were also asked to specify how many
miles they traveled each week, in total and by mode and purpose.
On one hand, reported estimations of typical travel, such as we obtained here, are not as reliable
as travel diary data. On the other hand, travel diaries can be criticized for generally
encompassing only a few days of travel and therefore potentially being unrepresentative at the
disaggregate level. Of course, these measures are respondents’ reports of the distance,
frequency, and time they are traveling, and hence are “ objective” only in the sense of referring to
those externally measurable quantities ( in contrast to the subjective measures of Subjective and
Relative Desired Mobility described below), rather than in the sense of actually being measured
through external observation.
For long- distance trips, pre- testing indicated that respondents would not be able to estimate
distances reliably. Thus, respondents were simply asked to tabulate how many trips they made
“ last year” for each mode- purpose combination ( personal vehicle/ work, personal vehicle/ enter-tainment,
etc.), to each of nine regions of the world. Those responses indicated number of trips
directly, and were also transformed to approximate measures of distance, through judgmental
average distances developed between the Bay Area and each of the nine world areas.
In addition, two transformations of the long distance objective mobility indicators are utilized in
this report: the natural log of the total miles, and the sum of the natural log of miles for each
trip1. The reason for performing a natural log transformation was to reduce the weight of long
trips, under the assumption that each additional mile traveled would have a diminishing marginal
impact ( i. e. each additional mile does not have as strong an incremental effect as the previous
mile). Also, the sum of the natural log of miles for each trip gives more weight to a larger
number of trips traveling a similar number of miles, compared to the natural log of the total
miles. For example, nine trips to Western States ( counted as 6,300 miles total) could constitute a
higher level of travel ( e. g. requiring more preparation, involving more disruption and a longer
1 Actually, ln ( miles + 1) was used to prevent combinations having zero miles from being transformed to negative
infinity ( ln [ 0]), and to return a value of 0 [= ln ( 1)] in those cases.
14
total absence) than one trip to Asia ( counted as 7,500 miles total). This higher level of travel is
captured by taking the sum of the natural log of miles for each trip: 58.96 (= 9 × ln [ 700]) for the
former case and 8.92 (= 1 × ln [ 7500]) for the latter ( Curry, 2000).
Subjective Mobility
We are interested not only in the Objective amount an individual travels, but also in how that
amount of travel is perceived. One person may consider 100 miles a week to be a lot, while
another considers it minimal. For each of the same categories as for Objective Mobility ( overall,
purpose, and mode categories for short- and long- distance), respondents were asked to rate the
amount of their travel on a five- point semantic- differential scale anchored by “ none” and “ a lot”.
Relative Desired Mobility
An individual may consider that she travels “ a lot”, but want to do even more. Thus, Relative
Desired Mobility refers to how much a person wants to travel compared to what she is doing
now. The structure of this question mirrors the structure for Subjective Mobility, with respon-dents
rating the amount of travel they want to do ( in each category) compared to the present, on a
five- point scale from “ much less” to “ much more”.
Travel Liking
Whether a respondent who already travels a lot wants to reduce it or do even more is likely to
depend on how much he enjoys traveling. To directly measure the affinity for travel, the question
was asked, “ How do you feel about traveling in each of the following categories? We are not
asking about the activity at the destination, but about the travel required to get there.” Respondents
were then asked to rate each of the same categories as Subjective Mobility on a five- point scale
from “ strongly dislike” to “ strongly like”.
Despite our attempt to alert respondents to distinguish the destination activity from the travel, it is
likely that even many of those who actually read the instructions ( and more of those who did not)
were unsuccessful at doing so. Future studies should perhaps make this distinction even more
forcefully to the respondent; interactive interviews would be one mechanism for probing answers
and helping the participant to separate these components of the utility for travel. Nevertheless, we
15
believe that the responses to this question are essentially measuring the degree of the respondent’s
affinity for travel for its own sake, even if that measurement is imperfect.
Attitudes
The survey contained 32 attitudinal statements related to travel, land use, and the environment, to
which individuals responded on the five- point Likert- type scale from “ strongly disagree” to
“ strongly agree”. Factor analysis was then used to extract the relatively uncorrelated
fundamental dimensions spanned by these 32 variables. Six underlying dimensions were
identified, using principal axis factoring with oblique rotation ( see Redmond, 2000 or
Mokhtarian, et al., 2001 for details): travel dislike, pro- environmental solutions, commute
benefit, travel freedom, travel stress, and pro- high density.
Personality
Respondents were asked to indicate how well ( on a five- point scale from “ hardly at all” to
“ almost completely”) each of 17 words and phrases described their personality. Each of these
traits was hypothesized to relate in some way to one’s orientation toward travel, or to reasons for
wanting to travel for its own sake. These 17 attributes reduced to four personality factors:
adventure- seeker, organizer, loner, and the calm personality.
Lifestyle
The survey contained 18 Likert- type scale statements relating to work, family, money, status,
and the value of time. These 18 questions comprised four lifestyle factors: status seeker,
workaholic, family/ community- oriented and a frustrated factor.
Excess Travel
Thirteen statements asked how often ( on a three- point scale: “ never/ seldom”= 0, “ sometimes”= 1,
“ often”= 2) the respondent engaged in various activities that would be considered unnecessary or
excess travel. The Excess Travel indicator is the sum of the responses to these statements,
ranging from 0 for the respondent who never/ seldom did any of them to 26 for the respondent
who often did all of them. This variable can be considered an indicator of Objective Mobility,
but also has a psychological flavor indicating an enjoyment of travel beyond the purely
16
utilitarian. The index may represent a strong desire for travel generally, or a preference for
discretionary travel which may have a negative relationship with mandatory travel for such
purposes as commuting and taking others where they need to go.
Mobility Constraints
In our study, Mobility Constraints are physical or psychological limits on travel. These
constraints may affect the amount an individual travels or her/ his enjoyment of that travel. In our
survey, these constraints are measured by questions concerning limitations on traveling by
certain modes or at certain times of day ( with ordinal response categories “ no limitation”, “ limits
how often or how long”, and “ absolutely prevents”), and the availability of an automobile when
desired.
Socio- demographics
Finally, the survey included an extensive list of Socio- demographic variables to allow for
comparison to other surveys and to Census data. These variables include neighborhood and car
type dummies, age, years in the U. S., education and employment information, and household
information such as number of people in the household, their age group, and personal and
household income.
2.4 General Hypotheses
In this section, we describe general hypotheses that represent potential relationships of the
explanatory variable categories as well as adoption variables to the consideration of strategy
bundles, particularly the conceptual strategy bundles ( because the factor- based strategy bundles
are for the most part subsets of conceptual strategy bundles). It should be emphasized that the
individual travel- related strategies, as the basis of the strategy bundles, primarily focus on
commute or work- related travel. However, discretionary travel such as recreation and
entertainment travel can directly or indirectly affect the consideration of strategy bundles. For
instance, people who desire to increase recreation travel may want to reduce their commute time,
so that they can spend more time on the desired travel. Thus, as we will see, in several cases
consideration of both travel- maintaining and the two types of travel- reducing strategies may be
positively associated with the same type of variable, for different reasons. For each category of
17
variable, the hypotheses are presented below and summarized in Table 2.2 at the end of this
section.
Objective Mobility. In general, it might seem that those who travel a lot should be more likely to
consider ways to reduce their travel. Thus, it could be expected that Objective Mobility is
positively associated with consideration of the travel reducing and major location/ lifestyle
change strategy bundles. Interestingly, the previous study ( Clay and Mokhtarian, forthcoming)
found that Objective Mobility variables are positively related to strategies in all three conceptual
bundles, based on individual t- tests. This may imply that people who have higher amounts of
travel are more likely to seek any type of travel- related strategy than to do nothing. In particular,
even travel maintaining strategies may be attractive to the heavy traveler, as a way of
ameliorating the travel that cannot be easily reduced. In view of our own expectations and these
prior findings, our hypothesis is that Objective Mobility is positively related to the consideration
of all three strategy bundles.
Subjective Mobility. Choo, et al. ( forthcoming) found that Subjective Mobility, as a
psychological assessment of the amount of travel one does, even more strongly affects
individuals’ Relative Desire to reduce their travel than Objective Mobility does. This supports
our initial hypothesis that those who perceive their travel to be a lot are more likely to consider
the travel reducing or major location/ lifestyle change strategy bundles. However, the previous
study ( Clay and Mokhtarian, forthcoming) found that Subjective Mobility variables are also
positively related to the consideration of all three bundles. Again, this implies that people with a
higher Subjective Mobility seek ways to make their travel more comfortable ( by getting a better
car) or lessen the psychological burden of travel ( by acquiring a better car stereo system or a
mobile phone), without necessarily reducing the amount of their current travel. Thus, similar to
Objective Mobility, we hypothesize that Subjective Mobility is positively related to the
consideration of all three strategy bundles.
Relative Desired Mobility. Clearly, those who generally want to increase their travel ( that is,
have a higher Relative Desired Mobility) should be more likely to consider the travel
maintaining/ increasing bundle. In contrast, people with a higher desire specifically for
18
discretionary travel may consider the travel reducing or major location/ lifestyle change strategy
bundles to reduce commute time, in order to increase the amount of time available for the desired
travel. Thus, it can be hypothesized that some Relative Desired Mobility variables are positively
related to the consideration of all three strategy bundles. Relative Desired Mobility for
commuting in particular, however, should be negatively related to the consideration of the travel
reducing and major location/ lifestyle change bundles.
Travel Liking. We first hypothesized that Travel Liking, representing a positive orientation
toward travel, would be positively associated with consideration of the travel
maintaining/ increasing bundle. That is, people who like travel are more likely to consider ways
to increase or maintain their travel. However, similar to Relative Desired Mobility ( with which it
is strongly correlated), the positive relationship of Travel Liking to the consideration of other
bundles may also be an outcome for a competitive preference for other travel than work.
Consequently, it can be hypothesized that Travel Liking is generally positively related to the
consideration of all three strategy bundles, with the same exception for commute Travel Liking
as noted for Relative Desired Mobility.
Attitudes. It is hypothesized that variables indicating a positive attitude toward ( commute) travel
( such as the commute benefit and travel freedom factor scores) are positively related to the travel
maintaining/ increasing bundle consideration, whereas variables indicating a negative attitude
toward travel ( such as the travel dislike and travel stress factor scores) are positively related to
the travel reducing or major location/ lifestyle change bundle consideration. We hypothesize that
the higher the pro- environmental solution factor score, the more likely the individual is to
consider the travel reducing or major location/ lifestyle change bundle. However, the situation
for the pro- high density attitude is more complex, with plausible hypotheses in both directions.
On the one hand, a pro- high density attitude might be a marker for not liking travel in general
( and hence wanting to live in a mixed- use neighborhood that minimizes the need to travel to
engage in desired activities). This would suggest a positive ( or, if one’s situation is already
optimized by living in a high- density area, a neutral) association with considering the travel
reducing and major location/ lifestyle change strategies. On the other hand, living in a
neighborhood where auto travel is more difficult ( congestion is higher, parking is scarce and
19
expensive) may create a sort of deprivation response, that stimulates consideration of strategies
leading to more travel ( or, stated the other way, that those with low pro- high density scores,
being travel- surfeited, are more likely to consider the travel reducing strategies). Perhaps because
of these counteracting relationships, this variable was never significant in the final models
presented here.
Personality. We hypothesize that the higher the score on the adventure seeker factor, the more
likely one is to consider the travel reducing or major location/ lifestyle change bundle. This factor
suggests a preference for entertainment travel over work, with heavily loading variables of
“ variety- seeking”, “ like being outdoors”, and “ risk- taking”. It could also be hypothesized that
those who are less calm are more bothered by congestion and hence more likely to consider
travel- related solutions, suggesting a negative relationship of the calm factor score to the
consideration of all three bundles. Hypotheses for the other two personality factor variables are
considerably weaker and more speculative. However, the variables are included in our modeling
to explore whether they significantly affect each strategy bundle.
Lifestyle. The frustrated factor represents those who are “ unsatisfied” or “ lacking control”. Thus,
people with a high score on this factor may be more likely to seek any travel- related strategy
bundles, and to change from one to another seeking more satisfaction. We expect the
family/ community oriented factor to be positively related to consideration of the travel reducing
or major location/ lifestyle change strategy bundle, permitting the individual to spend more time
with family or community by reducing commute time. Similarly, workaholics tend to want to
spend more time on work, so they may consider commuting to be wasting time that could be
better spent on work. Thus, this factor variable may positively affect the consideration of the
travel reducing strategy bundle but negatively affect the consideration of the major
location/ lifestyle change strategy bundle which includes “ retire or stop working”. On the other
hand, career- oriented professionals are often willing to accept a longer commute to a better job
( e. g., Pazy, et al., 1996), suggesting that workaholics may also be more inclined to consider
travel maintaining strategies to make more comfortable a commute that they deem necessary for
their career. We expect the status seeker factor to be positively associated with consideration of
20
the travel maintaining/ increasing bundle since people with high status seeker scores may want to
travel more to show off their cars or to buy a better car as a status symbol.
Excess Travel. As an indicator of a preference for discretionary travel, we expect Excess Travel
to be positively associated with the travel reducing or major location/ lifestyle change bundle
consideration. People with a higher Excess Travel value may have a higher Objective Mobility
and tend to want to reduce mandatory travel such as commuting.
Mobility Constraints. It can be hypothesized that Mobility Constraints are positively related to
the consideration of all strategy bundles. For example, people who have limitations on or
anxieties about driving during the day are likely to consider either travel maintaining ( changing
work trip departure time), travel reducing ( telecommuting), or major location change ( changing
jobs closer to home) strategies. That is, similar to the arguments for Objective Mobility and
Subjective Mobility, those people are more likely to seek any travel- related strategy bundles than
to do nothing to overcome their mobility constraints.
Socio- demographics. We hypothesize relationships of key socio- demographic variables to
consideration of the strategy bundles. As found in the previous related study ( Mokhtarian, et al.,
1997), we hypothesize that females are more likely to consider the more costly strategy bundles,
namely the travel reducing and major location/ lifestyle change bundles. We suggest that older
people are less likely to consider the first two travel- related strategy bundles, because they may
have been able to optimize their current circumstances or have become more accustomed to their
commute travel. On the other hand, we expect older people to be more likely to consider the
third strategy bundle, which includes changing from full- time to part- time work ( as a transition
stage to retirement) and retiring altogether. In addition, we expect that people with higher
incomes are more likely to consider all strategy bundles than to do nothing because they can
afford to buy a better car or to pay the additional costs associated with the more costly strategies.
Strategy Adoption. As suggested by Raney, et al. ( 2000), the previous adoption of a bundle or
single strategy could logically either positively or negatively affect the consideration of other
( and the same) strategies. For example, the adoption of a higher- cost strategy could reduce the
21
probability of considering a lower- cost strategy if the higher- cost strategy were effective, but it
could increase the probability of considering lower- cost strategies if the effectiveness of the
higher- cost strategy had diminished over time or was not as great as expected. In general, we
could hypothesize a progression from lower- cost to higher- cost strategies, but it is also natural to
expect some respondents to cycle within a given strategy bundle ( i. e. repeating strategies such as
getting a better car or changing work trip departure time) or to cycle back to a lower- cost strategy
after adopting a higher- cost one. Also, some strategies within a given bundle may be
complements ( so that adopting one strategy in the bundle increases the probability of considering
another one in the same bundle − e. g. buying equipment to support working from home, and
telecommuting), whereas others may be substitutes ( so that adopting one strategy in the bundle
decreases the probability of considering the same bundle − e. g. flextime and compressed work
week schedules). With respect to the time since adoption variable, we might initially expect that
people with a longer ( shorter) time since adoption of an individual strategy are more ( less) likely
to consider the corresponding bundle strategy. However, again, to the extent that strategies in a
given bundle are complements, the reverse may be true. Thus, for these variables we are in the
somewhat unaccustomed position of being able to justify virtually any relationship of prior
adoption of one strategy to the consideration of the same or a different strategy. However, it
would be of interest to identify which of the many conceptually possible relationships are
empirically dominant for this dataset. We explore this descriptively in Section 3, and analytically
through the models presented in Section 4.
22
Table 2.3: General Hypotheses
Dependent Variable ( Consideration of Strategy Bundle)
Explanatory
Variable Category Travel
maintaining/ increasing Travel reducing Major location/ lifestyle
change
Objective Mobility + + +
Subjective Mobility + + +
Relative Desired Mobility + + (- for commute) + (- for commute)
Travel Liking + + (- for commute) + (- for commute)
Attitudes
• commute benefit
• travel freedom
• travel dislike
• travel stress
• pro- environmental solutions
• pro- high density
+
+
-
+
+
+
+/-
-
+
+
+
+/-
Personality
• adventure seeker
• organized
• loner
• calm
+
undecided
undecided
-
+
undecided
undecided
-
+
undecided
undecided
-
Lifestyle
• frustrated
• family/ community oriented
• workaholic
• status seeker
+
+
+
+
+
+
+
+
-
Excess Travel + +
Mobility Constraints + + +
Socio- demographics
• female
• age
• income
-
+
+
-
+
+
+
+
Strategy Adoption
• adoption
• time since adoption
+/-
+/-
+/-
+/-
+/-
+/-
23
3. DESCRIPTIVE RELATIONSHIPS BETWEEN ADOPTION AND
CONSIDERATION
This section explores the descriptive relationships between previous adoption and current
consideration of strategy bundles, without considering the other variables. It is of interest to
explore whether the previous adoption of a strategy bundle is directly associated with the current
consideration of the corresponding or other strategy bundles. We first discuss the distribution of
previous adoption and current consideration for each set of strategy bundles, and then examine
not only their correlations but also their relationships to measures of satisfaction with current
travel conditions, using correlation tests.
3.1 Distribution of Adoption and Consideration of Strategy Bundles
As indicated in Section 2.4, Raney, et al. ( 2000) identified several possible relationships between
adoption and consideration of travel- related strategies. Given that lower- cost strategies have been
adopted, the individual is more likely to consider a higher cost strategy if she is unsatisfied with
the current strategy. On the other hand, given that lower- cost strategies have been adopted, the
individual is less likely to consider a higher cost strategy if she is satisfied with the current
strategy. In addition, it is plausible that the individual is more likely to consider the same or
another strategy in the same bundle regardless of her satisfaction. That is, if the individual has
been satisfied with the currently adopted strategy, she is more likely to keep adopting it. If not,
she may be more likely to seek another strategy in the same bundle ( particularly before
escalating to a higher- cost bundle), especially under travel time ( or cost) budget constraints.
Further, it should be emphasized that the combined adoption of more than one strategy bundle
might complicate the current consideration. If the individual is dissatisfied with the combined
adoption of strategy bundles, she may consider adding one or more strategy bundles, dropping
one or more adopted strategy bundles, or both. In fact, the Venn diagram in Figure 3.1 shows
that 68% of the sample has adopted two or more strategy bundles, and that the category for
adoption of all bundle strategies has the highest proportion. Also, it is possible that the individual
adopts more than one strategy in a given bundle. For example, 81% of the 326 respondents who
adopted only the travel maintaining/ increasing bundle have adopted more than one individual
24
strategy in that bundle. Thus, this indicates that people are likely to engage in more than one
strategy to control or reduce their work travel, with a probable synergistic effect.
Note: Numbers and percentages shown are mutually exclusive and collectively exhaustive.
Figure 3.1: Adoption and Consideration of Conceptual Strategy Bundles
For current consideration, similar to previous adoption, more than half of the respondents are
considering two or more conceptual strategy bundles. The category of the consideration of just
the travel maintaining/ increasing bundle strategy has the highest proportion of the sample ( nearly
one- fourth), and the category of the consideration of all bundles has also a high proportion ( more
than one- fifth). Interestingly, 16.3% of the sample is not considering any strategy bundles at all.
The non- consideration rate is almost five times higher than that of non- adoption. Such people
may either be so satisfied with the results of their previous adoptions that they are not motivated
Travel maintaining/
Increasing
326 ( 25.4%)
Travel reducing
12 ( 0.9%)
Major location/
lifestyle change
27 ( 2.1%)
263 ( 20.5%)
12 ( 0.9%)
338
( 26.3%)
257 ( 20.0%)
N = 1283 ( 100.0%)
Non- adoption
48 ( 3.7%)
Adoption
Travel maintaining/
Increasing
301 ( 23.5%)
Travel reducing
27 ( 2.1%)
Major location/
lifestyle change
81 ( 6.3%)
189 ( 14.7%)
40 ( 3.1%)
278
( 21.7%)
158 ( 12.3%)
N = 1283 ( 100.0%)
Non- consideration
209 ( 16.3%)
Consideration
25
to seek changes, or may be so dissatisfied with their previous adoptions that they believe nothing
they can do will improve their current travel conditions, resulting in a disinclination to pursue
new strategies.
Analyzing the survey responses ( see Figures 1 and 2 in Section 2), Clay and Mokhtarian
( forthcoming) found that the respondents adopted or are considering individual strategies for a
variety of reasons other than ( or in addition to) travel. “[ R] educing or easing travel” is the most-commonly
cited reason for only one strategy ( change from driving alone to some other means of
travel) in both adoption and consideration, and the second most- commonly cited reason for four
of the 19 strategies in adoption, and five of the 19 in consideration. However, they pointed out
( p. 15) that “ although respondents were invited to check as many reasons as applied, many
would have stopped after checking the first relevant reason. Even when they were willing to
check multiple reasons, they may not always have realized the importance of transportation to
their choices.” Keeping this in mind, we will mainly interpret the relationships between adoption
and consideration from the transportation point of view, while remembering the broader context
in which these activities take place.
Table 3.1 presents the cross- tabulation of previous adoption against current consideration of
combinations of the conceptual strategy bundles. For the 48 non- adopters ( adoption segment 1),
more than half of the respondents in this category are considering one or more strategy bundles,
especially the travel maintaining/ increasing strategy bundle. These people have likely been
mostly satisfied with their current travel conditions or are just starting to feel some
dissatisfaction, so they are more likely to consider a lower- cost strategy like those in the travel
maintaining/ increasing bundle. On the other hand, 189 ( 14.7%) respondents in the sample are not
considering any strategy bundle, despite having previously adopted one or more bundles. As
discussed before, such non- considerers might think that they have gained few ( current) benefits
from the strategy bundles they have adopted, even the higher- cost ones. Or, these people are
satisfied with their current travel conditions due to previous adoptions, so they are not motivated
to consider any strategy bundle at this time. Looking at the absolute frequencies in the final row
and column, it is reasonable that either or both of the travel reducing and major location/ lifestyle
26
change bundles are least likely to have been adopted or to be considered because of their higher
costs, compared to the other ( separate or combined) groups.
Looking at the rows of Table 3.1, for every adoption segment except segment 7 ( adoption of
Groups 2 & 3), the diagonal elements have the highest or second- highest proportion of
consideration for that category. That is, as could be expected, those who previously adopted
single or combined strategy bundles are more likely to consider the same strategy category than
to extend their consideration to other categories. For example, those who previously adopted a
single strategy in a bundle tend to consider adding another strategy in the same bundle ( or re-adopting
the same strategy), rather than changing to another bundle. Looking down the columns
and focusing on the bold numbers, Table 3.1 also shows that previous adopters of a particular
combination of bundles are generally more likely than adopters of other combinations to
consider the same combination.
Interestingly, as shown by the cross- hatched cells in Table 3.1, in contrast to the single- bundle
adopter segments 2, 3, and 4, those who adopted two strategy bundles ( segments 5, 6, and 7) tend
to consider adding another strategy bundle ( i. e. to consider all strategy bundles, as for segments 5
and 7), dropping the higher- cost one ( as for segment 6), or dropping both ( as for segment 7). It
may well be that people dissatisfied with their previously adopted strategies tend to consider
adding another strategy bundle, whereas people who are satisfied with their previously adopted
strategies tend to contemplate keeping or dropping one or more bundles.
Turning to the factor- based strategy bundles, there are a large number of combinations for all
strategy bundles ( 28= 256 possibilities), so we ( 1) consider only the respondents who have
adopted and are considering at least one strategy bundle, and ( 2) do not distinguish combinations
of bundles. That is, the adoption ( consideration) of each strategy bundle can include the adoption
( consideration) of single or multiple strategy bundles. For example, the adoption of Group 1
means the adoption of either Group 1 alone, or in combination with any other bundle( s).
27
Table 3.1: Adoption and Consideration of Combinations of Conceptual Strategy Bundles ( N= 1283)
Adoption segment Consideration
None Group
1 only
Group
2 only
Group
3 only
Groups
1 & 2
Groups
1 & 3
Groups
2 & 3
Groups
1 & 2 & 3 Total
1. Non- adoption 20
( 41.7)
12
( 25.0)
0
( 0.0)
1
( 2.1)
2
( 4.2)
7
( 14.6)
3
( 6.3)
3
( 6.3)
48
( 100.0)
2. Group 1 only:
Travel maintaining/ increasing
68
( 20.9)
107
( 32.8)
6
( 1.8)
26
( 8.0)
28
( 8.6)
57
( 17.5)
5
( 1.5)
29
( 8.9)
326
( 100.0)
3. Group 2 only: Travel reducing 2
( 16.7)
1
( 8.3)
2
( 16.7)
0
( 0.0)
1
( 8.3)
3
( 25.0)
1
( 8.3)
2
( 16.7)
12
( 100.0)
4. Group 3 only:
Major location/ lifestyle change
5
( 18.5)
3
( 11.1)
0
( 0.0)
7
( 25.9)
4
( 14.8)
5
( 18.5)
2
( 7.4)
1
( 3.7)
27
( 100.0)
5. Groups 1 & 2 37
( 14.4)
45
( 17.5)
9
( 3.5)
13
( 5.1)
48
( 18.7)
27
( 10.5)
9
( 3.5)
69
( 26.8)
257
( 100.0)
6. Groups 1 & 3 41
( 15.6)
72
( 27.4)
1
( 0.4)
21
( 8.0)
23
( 8.7)
57
( 21.7)
4
( 1.5)
44
( 16.7)
263
( 100.0)
7. Groups 2 & 3 3
( 25.0)
2
( 16.7)
1
( 8.3)
0
( 0.0)
2
( 16.7)
1
( 8.3)
0
( 0.0)
3
( 25.0)
12
( 100.0)
8. Groups 1 & 2 & 3 33
( 9.8)
59
( 17.5)
8
( 2.4)
13
( 3.8)
50
( 14.8)
32
( 9.5)
16
( 4.7)
127
( 37.6)
338
( 100.0)
Total 209
( 16.3)
301
( 23.5)
27
( 2.1)
81
( 6.3)
158
( 12.3)
189
( 14.7)
40
( 3.1)
278
( 21.7)
1283
( 100.0)
Note: The numbers in parentheses are the percents of the corresponding row category; the table focuses on the percentage of people that have
previously adopted a particular combination of bundles, who are considering each possible combination of bundles. Bold numbers indicate the
highest row percentage for that column, that is, the adoption group having proportionately the highest rate of consideration of that combination of
strategies. Cross- hatched cells indicate the highest row percentage for that row, that is, the combination of bundles most often considered by a
given adoption segment. Shaded cells simply highlight the main diagonal, i. e. the consideration of a given combination by those who have
adopted the same combination.
28
Table 3.2 shows the cross- tabulation of adoption and consideration of the factor- based strategy
bundles. Looking first at the columns, we see that, similar to the conceptual strategy bundles, in
five out of eight cases, the group most often considering a given bundle is the one who has
previously adopted it − that is, the diagonal element is the highest row percent of the column
( and is therefore bolded). To some extent, the respondents who adopted lower- cost strategy
bundles may tend to consider the next higher- cost bundle.
Turning to the rows, it is striking ( though not very surprising, in view of its low cost) that the
bundle considered most often by every adoption group except number 6 ( home- based work) is
bundle 1, auto improvement strategies. Perhaps surprisingly, the highest rate of consideration of
auto improvement comes from those who have adopted the mode change strategy bundle.
However, it is logical that those who changed from another means for commuting to driving
alone ( 82 of the 173 who adopted mode change and are considering auto improvement) are more
likely to improve their cars to make their driving commutes more comfortable. On the other hand,
those who changed from driving alone for commuting to other means ( 125 of the 173) may have
more money to invest in auto improvement strategies, because they spend less money on auto
maintenance than they would if they were commuting by driving alone.
3.2 Descriptive Analyses of Adoption and Consideration of Strategy Bundles
In this section, we conduct descriptive analyses for previous adoption and current consideration
to examine whether previous adoption is significantly related to current consideration, and
whether their relationships are significantly different between groups who are satisfied and
unsatisfied with their current travel conditions. First, a test of pairwise correlation between
adoption and consideration is carried out for each set of strategy bundles. Then, we conduct a
cluster analysis of four travel attitude factor scores − travel dislike, travel stress, commute benefit,
and travel freedom − to identify two groups, those who are unsatisfied and those who are
satisfied with their current travel conditions. Finally, we present correlation tests to explore
whether adoption and consideration are different between the two groups.
29
Table 3.2: Cross- tabulation of Adoption and Consideration Pairs for Factor- based Strategy Bundles ( Adopters Only)
Adoption ( N= Adopters) Consideration ( N= Considerers)
Group 1
( N= 613)
Group 2
( N= 380)
Group 3
( N= 369)
Group 4
( N= 297)
Group 5
( N= 180)
Group 6
( N= 471)
Group 7
( N= 297)
Group 8
( N= 333)
Group 1: Auto improvement ( N= 1048) 512
( 48.9)
317
( 30.2)
299
( 28.5)
259
( 24.7)
150
( 14.3)
385
( 36.7)
239
( 22.8)
266
( 25.4)
Group 2: Mobile phone ( N= 528) 258
( 48.9)
117
( 22.2)
154
( 29.2)
149
( 28.2)
63
( 11.9)
214
( 40.5)
123
( 23.3)
124
( 23.5)
Group 3: Work- schedule change ( N= 657) 329
( 50.1)
208
( 31.7)
254
( 38.7)
175
( 26.6)
117
( 17.8)
292
( 44.4)
194
( 29.5)
180
( 27.4)
Group 4: Hire someone to do house work
( N= 392)
189
( 48.2)
116
( 29.6)
120
( 30.6)
159
( 40.6)
58
( 14.8)
158
( 40.3)
74
( 18.9)
108
( 27.6)
Group 5: Mode change ( N= 331) 173
( 52.3)
106
( 32.0)
122
( 36.9)
75
( 22.7)
94
( 28.4)
155
( 46.8)
104
( 31.4)
78
( 23.6)
Group 6: Home- based work ( N= 474) 238
( 50.2)
151
( 31.9)
163
( 34.4)
143
( 30.2)
84
( 17.7)
299
( 63.1)
136
( 28.7)
135
( 28.5)
Group 7: Residential/ employment
relocation ( N= 448)
231
( 51.6)
149
( 33.3)
145
( 32.4)
102
( 22.8)
74
( 16.5)
191
( 42.6)
114
( 25.4)
99
( 22.1)
Group 8: Alter employment status ( N= 239) 118
( 49.4)
81
( 33.9)
71
( 29.7)
66
( 27.6)
32
( 13.4)
111
( 46.4)
56
( 23.4)
117
( 49.0)
Note: Numbers in parentheses are the percents of the adopters of the row bundle who are considering the column bundle. Respondents can
consider multiple bundles, so the sum of row percents will exceed 100. Bold numbers indicate the highest row percentage for that column. Cross-hatched
cells indicate the highest row percentage for that row. Shaded cells simply highlight the main diagonal, i. e. the consideration of a given
bundle by those who have adopted the same bundle.
30
3.2.1 Correlation Test of Adoption and Consideration
Pearson correlation tests were conducted to identify pairwise correlations between previous
adoption and current consideration. Table 3.3 presents the results of the tests for the conceptual
strategy bundles. Interestingly, except for Group 1 adoption and Group 3 consideration, previous
adoption of any bundle is significantly, positively correlated with current consideration of each
of the strategy bundles. The implication is that those who have any experience in adopting a
travel- related strategy bundle are more likely to consider another or the same bundle than are
non- adopters. The highest correlations are between adoption and consideration of the same
bundle ( the major diagonal elements), indicating that the same or similar strategies are likely to
be considered/ adopted repeatedly throughout an individual’s life. The adoption of higher- cost
strategy bundles tends to be somewhat more strongly associated with the consideration of all
three strategy bundles, compared to the adoption of the lower- cost bundles. In particular, higher-cost
bundle adopters are slightly more inclined to consider lower- cost bundles than lower- cost
bundle adopters are to consider higher- cost ones.
Table 3.3: Correlation between Adoption and Consideration of Conceptual Strategy
Bundles ( N= 1283)
Adoption Consideration
Group 1 Group 2 Group 3
Group 1: Travel maintaining/ increasing 0.127++ 0.071+, 1C, 2A
Group 2: Travel reducing 0.088++, 2C 0.336++ 0.101++, 2C
Group 3: Major location/ lifestyle change 0.080++, 2C 0.112++ 0.124++
Notes:
+: positive correlation with 0.01 < p- value ≤ 0.05, ++: positive correlation with p- value ≤ 0.01, insignificant
correlation omitted for simplicity. 1C: partial correlation becomes insignificant when Group 1 consideration is
controlled for. 2C: partial correlation becomes insignificant when Group 2 consideration is controlled for. 2A:
partial correlation becomes insignificant when Group 2 adoption is controlled for.
Additionally, we did partial correlation tests to explore whether or not a third variable
( consideration or adoption) affects a pairwise relation between adoption and consideration. That
is, if Corr( A, B) is significant ( and conceptual considerations support the causal direction A→ B
rather than B → A), but Corr( A, B | C) is not significant, it suggests that a more appropriate
model is A → C → B rather than A → B directly. The test results suggest that the impact of
31
previous adoption of bundle A on the current consideration of a lower- or higher- cost bundle B is
moderated by the simultaneous consideration of bundle A ( e. g., Group 1 adoption → Group 1
consideration → Group 2 consideration, or Group 2 adoption → Group 2 consideration → Group
1 consideration) or the simultaneous consideration of a different bundle ( Group 3 adoption →
Group 2 consideration → Group 1 consideration). In the one case in which controlling for prior
adoption significantly affected the correlation, the results are consistent with two interpretations:
Group 1 adoption → Group 2 adoption → Group 2 consideration, or Group 2 adoption affects
both Group 1 adoption and Group 2 consideration separately.
Table 3.4 shows the results of the correlation tests for the factor- based bundles. More than a third
of the pairs of adoption and consideration of strategy bundles are significantly correlated with
each other. Especially, six of the eight diagonal correlations are strongly significant. All of those
six except the mobile phone strategy are positively correlated. This is further support for the
observation that the previous adoption significantly affects the current consideration of the same
strategy bundle. As shown by the number, magnitudes and significance levels of correlations on
the upper half of the matrix compared to the lower half, the adopters of lower- cost strategy
bundles are somewhat more likely to consider higher- cost strategy bundles than the converse ( in
partial contrast to the results for the conceptual bundles). Together with the adopters of two
other strategy bundles, the adopters of the higher- cost residential/ employment relocation bundle
have the greatest number of significant correlations with consideration variables, and especially
tend to be considering the lower- cost travel maintaining bundles ( however, the correlations,
although statistically significant, are small in magnitude). It is intriguing that these adopters
consider home- based work as well, even though their travel distances for work were reduced by
moving home closer to work, or vice versa. Taken together, these results suggest that such
people may be inclined to reduce or eliminate commute trips but maintain travel for other
purposes. Adopters of medium- cost work- schedule changes and home- based work have an equal
number of significant correlations with the consideration variables, all of them representing
medium or higher- cost strategy bundles.
32
Table 3.4: Correlation between Adoption and Consideration of Factor- based Strategy
Bundles ( N= 1283)
Adoption Consideration
Group
1
Group
2
Group
3
Group
4
Group
5
Group
6
Group
7
Group
8
Group 1. Auto improvement 0.078++
Group 2. Mobile phone - 0.137-- 0.101++ 0.066+
Group 3. Work- schedule changes 0.224++ 0.085++ 0.111++ 0.164++ 0.155++
Group 4. Hire someone for
domestic help 0.274++ - 0.067-
Group 5. Mode change 0.105++ 0.244++ 0.124++ 0.116++
Group 6. Home- based work 0.095++ 0.127++ 0.081++ 0.419++ 0.101++
Group 7. Residential/ employment
relocation 0.055+ 0.058+ 0.058+ 0.090++ - 0.064-
Group 8. Alter employment status 0.097++ 0.251++
Notes:
+ (-): positive ( negative) correlation with 0.01 < p- value ≤ 0.05,
++ (--): positive ( negative) correlation with p- value ≤ 0.01,
insignificant correlations omitted for simplicity.
3.2.2 Clustering on Travel Attitudes
It is of great interest to examine whether the relationships of prior adoption to current
consideration are different between those who are satisfied with their current travel conditions,
and those who are dissatisfied. Naturally enough, we hypothesize a positive relationship to be
stronger for those who are dissatisfied, whereas those who are satisfied might have a neutral or
even negative relationship between adoption and consideration. To test these hypotheses, we
first classified the sample into two groups based on the travel attitude factor score variables ( only
one of which, commute benefit, is specific to work trips). The “ quick cluster” ( k- means)
analysis method in SPSS was employed, together with manually provided initial cluster centers.
We first tried to use all six travel attitude factor scores to classify the groups, but two variables,
pro- environmental solution and pro- high density, did not have any distinctive differences across
clusters, and in any case those two variables are less directly related to satisfaction with traveling
itself than are the remaining four. Thus, the final two clusters were created from the other four
variables: travel dislike, travel stress, commute benefit, and travel freedom. Table 3.5 presents
33
the final cluster centers with the 95% confidence interval, number of cases in each cluster, and
other work- related mobility variables. Based on the cluster centroids, people in the first group
tend to like travel, are not very stressed by it, find commuting beneficial, and feel free to go
where they want. The opposite is true for those in the second group, thus justifying the respective
“ satisfied” and “ unsatisfied” labels. With respect to the other variables, as expected, for commute
or work- related travel, on average the satisfied group has higher values for relative desired
mobility and travel liking, and lower values for subjective mobility ( specifically for commuting)
than the other group. This indicates that the satisfied group has relatively positive attitudes
toward work trips.
Table 3.5: Description of Clusters for Travel Satisfaction
Characteristics Cluster
Satisfied Unsatisfied
Number of cases 726 ( 56.6%) 557 ( 43.4%)
Final cluster centers with the 95% confidence interval
travel dislike factor - 0.510 ± 0.040 0.671 ± 0.061
travel stress factor - 0.411 ± 0.046 0.540 ± 0.060
commute benefit factor 0.414 ± 0.051 - 0.539 ± 0.063
travel freedom factor 0.194 ± 0.051 - 0.257 ± 0.060
Mean values of other variables
Subjective Mobility [ 1, 2, …, 5]
commuting to work/ school 3.46* 3.68*
work/ school- related activities 2.54 2.48
Relative Desired Mobility [ 1, 2, …, 5]
commuting to work/ school 2.48* 2.24*
work/ school- related activities 2.71* 2.59*
Travel Liking [ 1, 2, …, 5]
commuting to work/ school 2.96* 2.46*
work/ school- related activities 3.03* 2.73*
Notes:
All factor scores are standardized.
* There is a significant difference of means between the clusters at a level of α= 0.05.
3.2.3 Comparison of Adoption and Consideration between the Clusters
This section explores the statistical differences in previous adoption and current consideration for
the two clusters. We use Pearson pairwise correlation tests between adoption and consideration
for each cluster, and then compare the results between the two.
34
Table 3.6 presents the test results for the conceptual strategy bundles. Similar to the results of the
previous correlation tests, adoption and consideration are significantly, positively correlated for
most pairs of strategy bundles except one ( adoption of the travel maintaining/ increasing bundle
and consideration of the major location/ life style change bundle), in either the satisfied or the
unsatisfied group, or both. As expected, the diagonal elements have the highest correlations of
their row and column except for the major location/ life style change bundle ( especially for the
satisfied group). This suggests that one who has adopted a particular strategy is more likely to
consider either the same strategy or another strategy in the same bundle whether she is satisfied
with her current travel conditions or not.
Table 3.6: Correlations between Adoption and Consideration of Conceptual Strategy
Bundles ( Satisfied and Unsatisfied Groups)
Adoption Consideration
Group 1 Group 2 Group 3
S 0.132***
Group 1: Travel maintaining/ increasing
U 0.120** 0.107*
S 0.078* 0.346*** 0.115**
Group 2: Travel reducing
U 0.100* 0.324***
S 0.111** 0.093* 0.107**
Group 3: Major location/ lifestyle change
U 0.134** 0.149***
Notes:
* 0.01 < p- value ≤ 0.05, ** 0.001 < p- value ≤ 0.01, *** p- value ≤ 0.001 from a pairwise correlation test
statistic, insignificant correlations omitted for simplicity. S : satisfied group ( N = 726), U : unsatisfied
group ( N = 557).
Interestingly, unsatisfied people who have adopted the travel maintaining/ increasing strategy
bundle are more likely to consider the same bundle ( r = 0.120) than the higher- cost travel
reducing strategy bundle ( r = 0.107). It indicates that those people may be likely to consider
another strategy in the same bundle, without spending extra money on higher- cost strategies. It
is clear that satisfied people who have adopted this bundle tend to consider the same bundle, not
higher- cost ones. On the other hand, satisfied people who have adopted the major location/ life-style
change strategy bundle are more likely to consider the travel maintaining/ increasing
strategy bundle than unsatisfied people. That is, those adopters have already reduced their
commute distances and appear to be satisfied with the result, so they tend to try and maintain
35
their current ( reduced) work travel. Movers who remain dissatisfied, however, continue to
consider the higher- cost travel reducing and major location/ lifestyle change bundles more readily
than their satisfied counterparts.
Table 3.7 shows the test results for the factor- based strategy bundles. Similar to the results of the
previous correlation tests, nearly a third of adoption and consideration pairs are significantly
correlated in each group. As expected, diagonal correlations are strongly significant. It is also
found that the unsatisfied group has higher correlations than the satisfied one in five of six
diagonal elements. Overall, however, it does not appear that an individual’s satisfaction with her
current travel conditions plays a key role in considering a type of strategy bundle. For off-diagonal
correlations, regardless of satisfaction, lower- cost bundle adopters are more likely to
consider higher- cost bundles, and vice versa. This supports our hypotheses. Interestingly, only
the unsatisfied groups who adopted the work- schedule change and home- based work bundles
have consistently higher correlations for higher- cost strategy bundles than the corresponding
satisfied groups. Perhaps considering these strategy bundles can be more affected by individuals’
psychological assessments of their travel conditions. Similar to the conceptual bundles, those
who adopted residential/ employment relocation are more likely to consider lower- cost strategy
bundles in both the satisfied and unsatisfied groups.
Consequently, the statistical tests show that previous adoption is strongly associated with current
consideration regardless of satisfaction with current travel conditions. Actually, we do not know
whether the respondents are satisfied with their current travel conditions due to the adoption of a
certain strategy or due to other factors such as personality and socio- demographics. Thus, we do
not consider these satisfied and unsatisfied groups for the in- depth analysis of each bundle
strategy presented in Section 4.
The analysis in this section has neglected the dynamic aspect of the relationship between
adoption and consideration. For example, we would expect the impact of prior adoption on
current consideration to vary with the time since adoption, and with whether the strategy
previously adopted is still in force or has been discontinued. In the following section, we will
36
consider the time since adoption of a strategy as a key explanatory variable in modeling
consideration of a strategy bundle.
Table 3.7: Correlations between Adoption and Consideration of Factor- based Strategy
Bundles ( Satisfied and Unsatisfied Groups)
Consideration
Adoption Group
1
Group
2
Group
3
Group
4
Group
5
Group
6
Group
7
Group
8
Group 1: S
Auto improvement U 0.098*
Group 2: S - 0.132*** 0.126**
Mobile phone U - 0.141** 0.102*
Group 3: S 0.174*** 0.091* 0.102** 0.134*** 0.118**
Work- schedule changes U 0.289*** 0.124** 0.206*** 0.202***
Group 4: Hire someone for S 0.273** - 0.091*
domestic help U 0.275*** 0.095*
Group 5: S 0.112** 0.261*** 0.121** 0.143***
Mode change U 0.100* 0.227*** 0.122** 0.092*
Group 6: S 0.110** 0.388*** 0.097**
Home- based work U 0.139** 0.150*** 0.112** 0.462*** 0.108*
Group 7: Residential/ em- S 0.087* 0.100** 0.078* - 0.077*
ployment relocation U 0.115** 0.100* 0.094*
Group 8. Alter employment S 0.117** 0.240***
status U 0.266***
Notes: * 0.01 < p- value ≤ 0.05, ** 0.001 < p- value ≤ 0.01, *** p- value ≤ 0.001 from a pairwise correlation test
statistic, insignificant correlation omitted for simplicity. S : satisfied group, U : unsatisfied group.
37
4. MODELING THE CONSIDERATION OF STRATEGY BUNDLES
4.1 General Model Specification Issues
In the previous section, we discussed the descriptive relationships between previous adoption
and current consideration without involving other variables ( except for travel satisfaction, in
Section 3.2), and the results show that adoption and consideration are significantly related in
both directions, from lower- cost strategy bundles to higher- cost ones, and conversely. In this
section, we develop models for consideration of each bundle strategy, as a function not only of
adoption and time since adoption, but potentially also of the explanatory variables described in
Section 2.3. We model only consideration and not adoption, because the respondents’ adoption
takes place at various points in the past while the explanatory variables available in our cross-sectional
data set represent measures in the present. To model past adoption as a function of
present attitudes, say, would run the risk of reversing cause and effect: the present attitude is
likely to be a consequence of, rather than a cause of, the prior adoption ( Clay and Mokhtarian,
forthcoming). The dependent consideration variables are binary − 1 if the respondent seriously
considered any individual strategy in the bundle and 0 otherwise − so binary logit models were
selected for this study.
In particular, the logistic regression function of SPSS was used to estimate the models, due to its
stepwise methods of selecting significant variables for a model. For each bundle, two
specification approaches were used to obtain two ( potentially) different semi- final models. First,
based on initial specifications using various subsets of the explanatory variables, a forward
likelihood ratio method was repeatedly conducted to get a semi- final model in which all
explanatory variables were conceptually interpretable and had a significance level of 0.05 or
better. Second, based on an initial model specification containing all of the more than 200
potential variables, a semi- final model was also achieved, after manually eliminating statistically
insignificant and conceptually counter- intuitive variables step by step, allowing us to check for
any important variables that were missed through the automatic forward stepwise method due to
a marginal level of significance. After comparing semi- final models from the two methods and
testing the inclusion of variables appearing in only one of the two models into the other model,
we selected the final model. Through this procedure the final models were obtained, all of
38
whose explanatory variables were not only statistically significant, but also conceptually
interpretable.
It should be noted that a critical survey design feature affected model development for several of
the travel- related strategies. The respondents were asked in Question 1 of Part E of the survey to
indicate which alternatives they had adopted, and in Question 2 ( after stating “ even if you have
already made some of these choices, you could be thinking about making a similar change again,
or considering new options”) to indicate which they were seriously considering. This design
leaves two serious ambiguities. First, some respondents who had adopted, and were still
engaged in, a particular strategy ( such as telecommuting), may have felt uncomfortable
indicating they “ were not seriously considering” something they were in fact actually doing. On
the other hand, we have no way of ascertaining whether a strategy such as telecommuting, once
adopted, remained in place or not – it is well- established that many telecommuting engagements
are temporary ( Varma, et al., 1998).
The result of these two situations is that when someone who has previously adopted a certain
strategy indicates she is currently considering it, we do not know whether she is actually
currently doing it, or whether she has previously discontinued the strategy and is now
considering it again. Naturally these two groups of people could be quite different in terms of
explanatory variables. Similarly, “ non- considerers” who have previously adopted a strategy
comprise at least two distinct groups: those who are not considering it because their prior
adoption is still in force, and those who are not considering it because they have previously
discontinued the strategy and do not wish to re- visit it at this time. For these reasons, where
possible, we chose to estimate two models: one on the full data set, and one on non- adopters
only.
However, analyzing just the models with only non- adopters is not an ideal solution either, since
we wish to understand the behavior of adopters as well as non- adopters. Although the adopters
constitute less than half of the sample in eight of the 11 conceptual and factor- based strategy
bundles, we believe that a comparison of the full- data and non- adopter models will be fruitful,
with both the similarities and the differences between them being instructive. In such a
39
comparison, it should be kept in mind that, for the non- adopter models ( unlike the full- data
models), adoption, time since adoption of the given strategy and its quadratic term must of
necessity be excluded as potential explanatory variables. Furthermore, we have limitations on
modeling only non- adopters for a couple of strategy bundles − the travel maintaining/ in- creasing
and mobile phone strategies − due to smaller sample sizes and unbalanced shares of
consideration.
It is appropriate to comment in general on the inclusion of the adoption and time since adoption
variables in the models on the full sample. First, we used the adoption variables of either
individual or strategy bundles to exploit their potential explanatory power in the model. If we
use only the adoption of strategy bundles, we may lose significant information on the adoption of
a particular individual strategy in a given bundle. That is, due to the insignificance of the
adoption of the other individual strategies in the bundle, the adoption of the bundle strategy may
not be significant in the model, although the adoption of the particular individual strategy is
significant. In addition, it was not obvious how to define the time since adoption variables for
strategy bundles: e. g., time since adopting the most recently- chosen strategy in the bundle, time
since the most long- ago- chosen strategy, the average time since adopting a strategy in the
bundle. Thus, we used time since adoption variables only for individual strategies.
As discussed in Section 3, some evidence suggests that individuals first tend to consider or adopt
lower- impact strategies, moving to higher- impact ones if dissatisfaction still persists or returns,
and there is a weaker tendency for them to cycle back to lower- impact strategies if dissatisfaction
reoccurs after they have adopted a higher- impact one. On the other hand, if the adoption of a
strategy has met individuals’ needs, its adoption may decrease the probability of considering the
other strategies. Therefore, the former adoption of a strategy could be either positively or
negatively associated with the consideration of other strategies. By contrast, for most of the
strategies we are studying, we expect that the former adoption of a strategy positively affects the
consideration of the same strategy. Either the individual is enjoying and still wants to enjoy the
benefits from the previous adoption, or such strategies are attractive again as circumstances
change. Given that they are adopted once, it is natural to expect them to be adopted repeatedly
40
over a person’s working life. We have already seen support for this hypothesis in the pairwise
correlations analyzed in Section 3.2.
Moreover, the time since adoption of a strategy is generally expected to be positively related to
its reconsideration. That is, the longer ago an individual adopts a strategy, the more likely she is
to consider the same strategy. The time since adoption variable posed a difficulty with respect to
the treatment of non- adopters, however. Non- adopters had to be given a value for this variable
in order for them ( and this variab