Strategies for personal mobility: A study of consumer acceptance of subscription drive-it-yourself
275
1 Strategies for personal mobility: A study of consumer acceptance of subscription drive-it-yourself car services Scott Le Vine Imperial College London Department of Civil and Environmental Engineering Submitted for the Diploma of the Imperial College (DIC), PhD degree of Imperial College London
Strategies for personal mobility: A study of consumer acceptance of subscription drive-it-yourself
Appendix inc fill-ins.pdfStrategies for personal mobility: A study
of consumer acceptance of
subscription drive-it-yourself car services
Department of Civil and Environmental Engineering
Submitted for the Diploma of the Imperial College (DIC), PhD degree
of Imperial College
London
2
This study is dedicated to Natasha, and to the future
Acknowledgements
I am indebted to many for their advice, support and inspiration
that made this study
possible.
My supervisors Aruna, John, and Martin went beyond the call of duty
on many occasions,
and this research could not have proceeded if not for their
encouragement and guidance. I
would also like to thank my examiners Kay and Dan, and Alison for
placing her faith in me by
providing me the opportunity to teach whilst undertaking this
research.
This study would not have been possible if not for financial
support from the RAC
Foundation, and I would also like to thank Elizabeth for her
insights and suggestions along
the way which have led to a better project.
Warm thanks are due to Carplus, City Car Club and RAC Breakdown
Services for providing
access to customers of theirs willing to take part in the
fieldwork, and to the study
participants who generously chose to share their time. Further
gratitude is extended to
Benefit Technology Inc, DfT’s NTS team, Imperial’s HPC team,
NatCen, and SRA for their
support of this research in a variety of ways.
Discussions with a number of public sector staff, academics, and
industry partners helped to
shape the contours of this study, and I would like to thank each of
you for your time and
thought. (The usual statement applies: any remaining errors are my
responsibility.)
I could not have taken on a multi-year assignment of this nature
without the unflinching
support of my family and friends, in particular those friends who
have become like family to
me during the course of this research.
The research presented here is my own, except where the work of
others is referenced.
3
Abstract
This thesis investigates consumer acceptance of subscription
drive-it-yourself car services [SDCSs],
which are an evolution of car hire that began entering the
commercial marketplace in the mid-
1990s. The aim of this research is to develop techniques to
forecast how consumer demand for
SDCSs may develop.
On the basis of research reported in this thesis, it is argued that
a person’s [strategic] decision to
subscribe to an SDCS can be reasonably considered to have a
dependency with their expectation of
[tactically] using it to access particular out-of-home personal
activities. It is shown that people can
also be thought to view subscribing to an SDCS as part of a larger
‘portfolio’ choice of travel options.
Traditional analyses of people’s travel choices are insensitive to
these two issues.
Two datasets, one revealed-choice and the other stated-choice, were
designed in order to provide
empirical data to test the proposed ‘strategic/tactical’ and
‘portfolio’ analytical form. The revealed-
choice dataset made use of web-based data-mining techniques, whilst
the stated-choice survey is
novel in several respects to address the challenges presented by
the SDCS context.
The methodological innovations proposed in this research proved
successful in forecasting consumer
demand for SDCSs in the empirical application, and appear promising
for wider use within the
transport domain and related research fields.
4
Chapter 3: Gaming-simulation task page 48
Chapter 4: Analytical framework page 58
Chapter 5: E-NTS dataset page 75
Chapter 6: AVATAR survey page 103
Chapter 7: Independent analyses of E-NTS and AVATAR datasets page
129
Chapter 8: Substantive results page 158
Chapter 9: Summary & conclusions page 183
Appendix A: Comparison of ‘distinct’ and ‘combinatorial’ model
forms
Appendix B: Derivation of the ‘plateau effect’
Appendix C: Detailed specification and results from the analysis of
simulated data
Appendix D: Further results from joint (E-NTS/AVATAR) estimation of
parameters
Appendix E: Sample gaming-simulation survey instrument
package
Appendix F: Sample AVATAR survey instrument package
5
Figure 1.2: Workflow of this study
Figure 2.1: Schematic of the course of a person’s car club SDCS
engagement
Figure 2.2: Time trend in car club members worldwide and in the
UK
Figure 2.3: Month-on-month retention of subscribers for the
Communauto car club in 2004
Figure 2.4: Schematic of alternative trajectories for
carsharers
Figure 2.5: Overview of studies which have forecasted the potential
market penetration of SDCSs
Figure 2.6: Daily usage patterns of the Communauto car club for the
year 2004
Figure 4.1: Summary of 'Distinct' and 'Combinatorial'
structures
Figure 5.1: Map of England's Government Office Regions
Figure 5.2: Cumulative distribution plot of the number of journeys
performed by NTS respondents
that are within the E-NTS sample
Figure 5.3: Workflow of web scraping task
Figure 5.4: Screen capture of formatted HTML output from the
Journey Planner travel planning
service
Figure 5.5: Plot of observed and predicted journey speed by
time-of-day for car journeys
Figure 5.6: Plot of observed and predicted journey speed by
time-of-day for public transport
journeys
Figure 5.7: Plot of residual values from the web scraping task,
disaggregated by mode of transport
Figure 5.8: Plots of residual values from the web scraping task,
separate plots for each mode of
transport
Figure 6.1: Sample screen introducing the survey respondent to her
[his] avatar
Figure 6.2: Sample of the main survey screen (as re-designed
following field testing)
Figure 6.3: Screen capture of the request to [female] survey
respondents to advise their avatar
Figure 6.4: Sample of the main survey screen prior as pilot-tested
(prior to re-design)
Figure 6.5: Cumulative distribution of the duration of the
interviews
6
Figure 7.1: Response of likelihood function to varying the
mode-choice-level alternative-specific
constant for car passenger travel, using dataset G (Run G2)
Figure 7.2: Response of likelihood function to varying the
mode-choice-level alternative-specific
constant for taxi/minicab travel, using dataset G (Run G2)
Figure 7.3: Response of likelihood function to varying the
mode-choice-level alternative-specific
constant for driving a personal car, using dataset G (Run G2)
Figure 7.4: Response of likelihood function to varying the
portfolio-choice-level alternative-specific
constant for owning a personal car, using dataset G (Run G2)
Figure 8.1: Cumulative distribution of the change in the number of
car driving journeys per week
between the baseline scenario and Scenario #1 by the 19 people
predicted to subscribe to a car club
SDCS in scenario #1
Figure 8.2: Response of likelihood function, using the ‘portfolio’
specification and the E-NTS dataset
only, to successive increases in the ‘observed’ number of each
journeys performed by each person
Figure 8.3: Response of values of the salience (gamma) parameters
to increases in the number of
people’s journeys which are taken into account
7
Table 2.2: Examples of SDCS studies employing qualitative market
research techniques
Table 4.1: Matrix of the set of means of travel enabled by various
resource portfolios
Table 5.1: Comparison of online travel planning services
Table 5.2: Spatial match quality for journey origins and
destinations
Table 5.3: Proportion of journeys for which no itineraries for
various travel modes were reported by
the web scraping task
Table 5.4: Summary of descriptive statistics for travel time of
journey itineraries, by observed
method of travel
Table 5.5: Average residual errors (predicted minus observed)
disaggregated by source of spatial
match for journey from the web scraping task
Table 5.6: Results of diagnostic mode choice model run using the
E-NTS dataset
Table 6.1: Sample asymptotic variance/co-variance matrix,
containing dummy values
Table 6.2: Matrix of correlations from the AVATAR survey
results
Table 7.1: Comparison of data and model form characteristics for
estimating the 'mode' and
'portfolio' choice models
Table 7.2: Summary of observed portfolio choices in the E-NTS and
AVATAR datasets
Table 7.3: Summary of observed mode choices in the E-NTS and AVATAR
datasets
Table 7.4: Correlation matrix of observations (‘portfolio’ holdings
and ‘mode’ usage) from the E-NTS
dataset
Table 7.5: Correlation matrix of stated choices (‘portfolio’
holdings and ‘mode’ usage) from the
AVATAR dataset
Table 7.6: Listing of simulated datasets and characteristics
Table 7.7: Results from estimation using simulated data
Table 7.8: Comparison of target and obtained parameter values for
run F2
Table 7.9: Results from parameter estimation of ‘portfolio’ choice
using only the E-NTS dataset
Table 7.10: Results from parameter estimation of ‘portfolio’ choice
using only the AVATAR survey
dataset
8
Table 7.12: Results from parameter estimation of transport mode
choice using only the E-NTS survey
dataset
Table 7.13: Results from parameter estimation of transport mode
choice using only the AVATAR
survey dataset
Table 8.1: Results from parameter estimation of ‘portfolio’ choice
using the combined E-NTS and
AVATAR survey datasets
Table 8.2: Results from parameter estimation of mode choice using
the combined E-NTS and
AVATAR survey datasets
Table 8.3: Comparison of ‘values of time’ estimates for non-SDCS
modes of transport
Table 8.4: Comparison of ‘values of time’ estimates for SDCS modes
of transport
Table 8.5: Summary of results from baseline scenario &
Scenarios #1 through #7
Table 8.6: Correlation matrix of simulated choices (‘portfolio’
holdings and ‘mode’ usage) from the
baseline scenario (#0)
Table 8.7: Cross-tabulation of observed and predicted (baseline
scenario) ‘portfolio’ holdings
Table 8.8: Correlation matrix of simulated choices (‘portfolio’
holdings and ‘mode’ usage) from
Scenario #1
Table 8.9: Correlation matrix of simulated choices (‘portfolio’
holdings and ‘mode’ usage) from
Scenario #5
Table C.1: Target parameter values for the simulated datasets
Table C.2: Obtained parameter values for runs with simulated
dataset A
Table C.3: Obtained parameter values for runs with simulated
datasets B & C
Table C.4: Obtained parameter values for runs with simulated
datasets D & E
Table C.5: Obtained parameter values for runs with simulated
datasets F & G
Table D.1: Results from parameter estimation of ‘portfolio’ choice
using the combined E-NTS and
AVATAR survey datasets
Table D.2: Results from parameter estimation of mode choice using
the combined E-NTS and
AVATAR survey datasets
. 0 2=#
"" A( >3 # +
& 3 & BED
" &"$
# ( )#
! !
"$ !
+ ,"- + #" ,
$
# " "
. , #" ( #
( )# )
) $#*%%# +$,
#$ %%((
+ ! *+)#
("<"
? !
A .H .
?' A A A A A A A I I +1 -"-
A A
' A A A A A A A I I +<1<"(
6 A A
A A A
6' A A A A A A A I I + &"
6( A A
A A A
A' A A A A A A A I I +&'"
B A A
A A A