Transportation (2020) 47:1223–1250

https://doi.org/10.1007/s11116-018-9952-x

Design of a survey to assess prospects for consumer electric

mobility in Canada: a retrospective appraisal

Elnaz Abotalebi1 · Mark R. Ferguson2 · Moataz Mohamed3 · Darren M. Scott4

Published online: 20 November 2018

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract
This paper reviews the process that led to the development of a national survey instru-
ment used to gather over 20,000 observations across Canada. This survey captured aspects
of household preferences and behavioural intentions towards electric vehicles through a
choice experiment and a comprehensive suite of attitudinal questions. Background infor-
mation on demographics, residential location and context, vehicle ownership and purchase
plans, and travel patterns among other aspects were also collected. Important survey design
decisions are examined that include: the choice and implications of using a survey panel,
screening criteria for the sample, conceptualization of the observational unit for the sam-
ple, critical aspects relating to the choice experiment, and tactics employed to manage and
measure survey cognitive burden. Novel aspects associated with the survey design are dis-
cussed in this paper and these have enabled new research works on the implications of
vehicle body type on choice of powertrain and insights into spatial variation in electric
vehicle preferences. Results and insights discussed are seen as relevant for a range of sur-
vey practitioners including those with a focus on the consumer.

Keywords Attitude · Electric vehicles · Cognitive burden · Stated preferences · Survey

design

* Darren M. Scott
scottdm@mcmaster.ca
Elnaz Abotalebi
hajaboe@mcmaster.ca
Mark R. Ferguson
fergumr@mcmaster.ca
Moataz Mohamed
mmohame@mcmaster.ca
1
School of Geography and Earth Sciences, McMaster University, 1280 Main Street West, Hamilton,
ON, Canada
2
McMaster Institute for Transportation and Logistics, McMaster University, 1280 Main Street West,
Hamilton, ON, Canada
3
Department of Civil Engineering, McMaster University, 1280 Main Street West, Hamilton, ON,
Canada
4
TransLAB (Transportation Research Lab), School of Geography and Earth Sciences, McMaster
University, 1280 Main Street West, Hamilton, ON, Canada

13
Vol.:(0123456789)
1224 Transportation (2020) 47:1223–1250

Introduction

Surveys are important for research that examines human behaviour. In the field of transpor-
tation, surveys are powerful means to obtain critical information for planning and policy
making and can provide researchers and policy makers with high-quality data to evalu-
ate changes in transportation systems and regulations in response to existing transporta-
tion problems. Surveys that seek to understand consumer preferences and attitudes towards
electric and other alternative fuel vehicles have been prominent over the past 25 years, but
their use has accelerated in the past decade. Interest in this field has been driven by the
fact that electric vehicles (EV) offer a pathway to mitigate the consequences of vehicular
emissions.
With the objective to investigate the consumer electric mobility landscape in Canada, a
team of researchers and industry partners undertook an effort to develop a comprehensive
survey instrument. The acronym “SPACE” (Survey for Preferences and Attitudes of Cana-
dians towards Electric Vehicles) is used for the purposes of this paper to identify the survey
instrument. SPACE was deployed in the spring of 2015 and focused on four main power-
train types: internal combustion engines (ICE), hybrid electric vehicles along the lines of
the Toyota Prius (HEV), plug-in electric hybrid vehicles (PHEV—e.g. Chevy Volt) capa-
ble of running on gasoline or electricity and battery electric vehicles (BEV—e.g. Tesla
Model S, Nissan Leaf) powered strictly by electricity.
Canada makes for an interesting case study. It has one of the cleaner electricity genera-
tion profiles in the world, and yet has a very small market share for EVs which lags rela-
tively behind the United States (U.S.) and many European countries. While previous works
provide valuable insights about EV consumer within Canadian context (e.g. Axsen et al.
2015; Ewing and Sarigöllü 1998; Potoglou and Kanaroglou 2007b), the sample collected
with SPACE is the first of its kind in Canada with a comprehensive national scope and
with meaningful samples from all ten Canadian provinces, major metropolitan areas, and
in both official languages. A stated preference (SP) experiment is at the core of SPACE and
is supported by an array of information on demographics, residential location and context,
current vehicle ownership and vehicle purchase intentions, travel patterns and an assort-
ment of attitudinal information.
There are a number of novel aspects associated with the design of SPACE. First, the
survey is administered among a large panel of Canadians with 20,520 observations from
every corner of the country (e.g. all provinces, urban–rural settings, etc.), and hence pro-
vides more in-depth coverage of geographical variation in EV preferences than previous
studies. Second, the implementation was sensitive to respondent context (Carlsson 2010),
with SP scenarios being customized per respondent (e.g. annual mileage, replacement or
additional vehicle, and anticipated purchase price). Moreover, respondents encountered SP
scenarios associated with their preferred vehicle body type. For instance, vehicle attribute
characteristics shown to an economy buyer were different from those offered to a respond-
ent interested in a luxury sedan. Finally, the survey instrument featured a good balance of
objective socio-economics and information about perceptions, attitudes and beliefs in order
to best understand what aspects were most heavily influencing preferences.
To this point, SPACE has given rise to four studies and with others in progress.
Mohamed et al. (2016) utilized attitudinal and demographic components of this survey to
assess behavioral intention towards EVs. Higgins et al. (2017) conducted a choice mod-
eling analysis where respondents evaluated powertrain attributes that were specific to
the vehicle body types they preferred for their next vehicle, and Mohamed et al. (2018)

13
Transportation (2020) 47:1223–1250 1225

followed that up with an analysis on vehicle body type supported by structural equation
analysis. Ferguson et al. (2018) leveraged a wide range of collected variables and the full
national scope of the data to develop a latent class choice model for Canada.
Based on our literature review, Table 1 summarizes studies that have used surveys
to help understand the preferences of consumers towards different types of EVs. All the
reviewed studies are based on stated as opposed to revealed preference data and this is
likely explained by the low market penetration of electric vehicles. Within the SP experi-
ments of these studies, the number of powertrain alternatives ranges between two and
seven and several have included a conventional vehicle (gasoline or diesel) along with one
or more EV types. The number of attributes varied from five to eight, that generally include
the monetary (price, fuel cost, etc.), functional (range, acceleration, etc.), charging infra-
structure and policy-related attributes. There is also great variability in number of scenar-
ios, type of experimental design, geography and sampling frame. However, aspects of sur-
vey design are only partially discussed in previous EV studies and are mainly reported as
part of the analyses that these surveys support. As a result, the construction of the survey
instrument that underlies the analysis often remains somewhat in the background.
The present paper contributes to the literature by providing an informative overview
of the development of a comprehensive survey instrument that was applied in a Canada-
wide data collection effort to yield useful EV-oriented data and information. The overview
includes associated strategies, thought processes and significant decisions that were made
in developing and implementing the survey. The aim is for some lessons that emerge to
extend beyond those working within the EV domain. Another contribution is to address
a gap in the EV-oriented literature where many surveys have been developed to this point
around the world, but a thorough review of the thinking that went into these is not avail-
able, to the best of our knowledge.
The remainder of this paper is organized as follows. Section 2 provides an overview
of the survey instrument and describes the content of SPACE and rationale for our target
population. Having provided the necessary context, Sects. 3 and 4 move on to explore, in
some depth, the important decision-making that was linked to the survey development and
the significant and novel aspects of the main survey instrument. As will be seen, Sect. 3 is
more weighted to the important decisions and Sect. 4 is more weighted to novel aspects. A
concluding section synthesizes the results of our retrospective look at the development of
SPACE.

Overview of the survey instrument and rationale for target population

SPACE is designed to investigate factors as they relate to the choice of vehicle powertrain.
The main idea is to collect data on potential car buyers/leasers of all types, with an objec-
tive to capture a wide range of population segments and assess their interest in acquiring an
electric vehicle across these segments. The sampling approach is thus not highly targeted
compared to several past EV studies that recruited people, for example, on the basis of a
vehicle purchase within the previous or the upcoming year following survey deployment
or with a focus on new vehicle buyers (e.g. Hackbarth and Madlener 2013; Mabit and Fos-
gerau 2011; Axsen et al. 2015).
The design of SPACE is aligned with what we planned to do with the data as part
of a larger project. For example, we intended to characterize the future potential
for EVs in thousands of small geographies across Canada. The collection of 20,520

13
 Table 1  An overview of selected EV studies
1226

Study Geography No. respondents Design type Labeled/unlabeled No. attributes No. sce- No. alternatives or fuel
narios/opt-out types

13
choice?

Bunch et al. (1993) Southern California 692 Orthogonal Quasi-labeled 7 fuel type, fuel avail- 5/No 3 gasoline, alternative
ability, range, price, fuel, electric
fuel cost, pollution,
performance
Ewing and Sarigöllü Montreal, Canada 881 Suburban driver Orthogonal Labeled 8 price, maintenance 9/No 3 conventional, fuel-
(1998) commuters cost, acceleration, efficient, electric
range, refuelling rate,
emission, commuting
time, fuel and park-
ing cost
Horne et al. (2005) Canada 1150 Not Stated Labeled 6 price, fuel cost, fuel 8/Not stated 4 gasoline, natural gas,
availability, lane hybrid/electric, hydro-
access, emissions, gen fuel cell
power
Potoglou and Kan- Hamilton, Canada 482 Future car buyers Orthogonal Quasi-labeled 8 price, fuel and main- 8/No 3 gasoline, hybrid, alter-
aroglou (2007b) tenance costs, fuel native fuelled
availability, incen-
tive, acceleration,
pollution, vehicle
size, fuel type
Mabit and Fosgerau Denmark 2146 New car buyers Not stated Labeled 6 price, operation cost, 12/Not stated 2 out of 5 conventional,
(2011) range, refuelling fre- hydrogen, hybrid, bio-
quency, acceleration, diesel, electric
service dummy
Hidrue et al. (2011) US 3029 Future car buyers D-efficient Labeled 6 range, charging time, 2/Yes 2 conventional, BEV
fuel cost, pollution,
performance, price
Transportation (2020) 47:1223–1250
 Table 1  (continued)
Study Geography No. respondents Design type Labeled/unlabeled No. attributes No. sce- No. alternatives or fuel
narios/opt-out types
choice?

Achtnicht (2012) Germany 600 Future car buyers Orthogonal Quasi-labeled 6 price, fuel costs, 6/Not stated 7 gasoline, diesel,
engine power, emis- hybrid, gas, biofuel,
sions, fuel availabil- hydrogen
ity, fuel type
Ziegler (2012) Germany 598 Potential car Not Stated Quasi-labeled 5 price, power, fuel 6/Not stated 7 gasoline, diesel,
buyers costs, emissions, hybrid, gas, biofuel,
Transportation (2020) 47:1223–1250

service station avail- hydrogen, electric

ability, fuel type
Hackbarth and Madle- Germany 711 New or potential Orthogonal Not stated 8 price, fuel cost, emis- 15/Not stated 4 out of 7 gas, HEV,
ner (2013) car buyers sions, range, fuel PHEV, BEV, BV, fuel
availability, refueling cell, conventional
time, recharging
time, policy incen-
tives
Beck et al. (2013) Sydney, Australia 650 Recent car buyers D-efficient Labeled 9 price, fuel cost, 4–5/No 3 petrol, diesel, hybrid
emission charge, fuel
efficiency, engine
size, seating capacity,
manufacturer
Jensen et al. (2013) Denmark 369 Orthogonal Labeled 6 price, fuel costs, 8/Yes 2 conventional, EV
performance (top
speed), emissions,
charging possibility,
battery lifetime
1227

13
 Table 1  (continued)
1228

Study Geography No. respondents Design type Labeled/unlabeled No. attributes No. sce- No. alternatives or fuel
narios/opt-out types

13
choice?

Hoen and Koetse Netherlands 1903 Private car Orthogonal Unlabeled 8 fuel type, price, 8/No 3 out of 6 conventional,
(2014) owners monthly costs, range, hybrid, plug-in hybrid,
refueling time, addi- fuel cell, electric,
tional detour time, flex-fuel
number of brands/
models, policy
measure
Tanaka et al. (2014) Japan and US 4202 (US) Orthogonal Labeled 6 price, fuel cost, 8/Yes 3 EV, PHEV, gasoline
4000 (Japan) range, emissions, fuel
availability, home
plug-in construction
fee.
Axsen et al. (2015) Canadian provinces 1754 New car buyers Orthogonal Not Stated 5 price, fuel cost, 6/Not stated 4 conventional, HEV,
range, home recharge PHEV, BEV
access, recharge time
Helveston et al. (2015) China and the US 667 (china) Not Stated Unlabeled 6 fuel type, range, 16/No 3 out of 4 conventional,
415 (US) brand, price, fast HEV, PHEV, BEV
charging capability,
fuel cost, accelera-
tion
Present study Canada 20,520 future car D-efficient Labeled 12 price, mainte- 4/Yes 4 ICE, HEV, PHEV,
buyers nance and fuel cost, BEV
gasoline and electric
range, acceleration,
cash and non-cash
incentives, battery
warranty, emission,
charging time and
availability
Transportation (2020) 47:1223–1250
Transportation (2020) 47:1223–1250 1229

observations allows us to work with segments of the overall sample and still have a
significant and meaningful sample within each segment. It also provides a great deal
of flexibility to access the data and answer specific research questions that may come
up after the fact. Even with a large sample, it was decided not to allocate budget to
people who stated little or no likelihood to acquire a vehicle. This has implications on
the inclusion of our screening criteria as those with near zero intention of acquiring a
vehicle in the “foreseeable” future were excluded from the survey (See Sect. 3.2).
Table 2 provides an overview of the six primary survey sections and the specific
elements included in each. The survey started with an introductory script emphasizing
the aim and scope of the project as well as other general information about the survey
including sections, expected completion time, and contact information. The first sec-
tion contained the household current vehicle inventory and travel patterns. This sec-
tion gathered vehicle-specific data on make, model, year, fuel type, residential park-
ing circumstances, and number of registered vehicles for the household. A detailed
hierarchical drop-down list of all makes and models available in the Canadian new and
used vehicle markets was developed to assist participants in reporting their vehicles.
We collected this information for up to four vehicles registered by a household. This
section also collected data on the travel patterns of the household with an emphasis on
two aspects: the degree of reliance on different mobility modes, and daily/monthly trip
frequencies and driven distances for each vehicle. Note that the sample also included
households without a vehicle at the time. These households skipped this section as
vehicle inventory and travel patterns did not apply for them.
The second section covered the household purchase/lease plan including their
expected vehicle type, price, and timing of purchase/lease. This part contained a vari-
ety of supporting questions, particularly, the ones that play a more critical role in an
EV purchase/lease decision such as replacement versus additional car, and new versus
used car purchase (Massiani 2014). The time horizon for acquiring a vehicle as well
as reasons and budget for their purchase/lease were investigated here. A Likert-scale
matrix was included in this section to measure the importance of some vehicle attrib-
utes (e.g. cargo space, etc.). They were collected to obtain a separate viewpoint on
vehicle attributes from the insight that would emerge from the SP scenarios.
Attitudinal statements comprised section five of SPACE and captured the behav-
ioural intentions of participants to purchase an electric vehicle. We developed thirty
items based on an extended theory of planned behaviour that were structured around
six constructs related to EV purchase decision: environmental concern, attitude, social
norm, personal moral norm, perceived behavioural control, and behavioural intention.
Although independently the attitudinal factors provided comprehensive analyses of the
behavioural factor influencing EV adoption (Mohamed et al. 2016), it also provided a
strong ground that could be integrated in econometric analysis to explain heterogeneity
within the sample (Ferguson et al. 2018).
The last part of our survey included socioeconomic questions, which collected
potentially relevant variables that could influence the choice of vehicle powertrain,
such as place of residence and housing type (single family housing vs. apartments). We
also collected postal codes of the household residence and workplace which provided
us with valuable information for a future geodemographic analysis. Note that for the
most part, SPACE took advantage of closed-ended questions to reduce the amount of
effort needed from respondents. The question types of each section are provided in
Table 2.

13
 Table 2  SPACE sections and question types
1230

Survey sections Question types Survey elements

13
Survey start screen One-page text Introduction
Screening Single-select questions Age
If a household decision-maker or not
Likelihood to acquire a household vehicle in the future (replacement or incremental)
Section 1: Drop-down lists and Single-select questions Number of registered vehicles in household
Current vehicles and travel pattern Make/model/year of vehicles
Ownership status (e.g. owned or leased)
Fuel type
Household parking context and exposure to weather
Access to electrical outlets
Average number of trips per weekday/weekend per vehicle
Average kilometers driven per weekday/weekend per vehicle
Estimated monthly number of high mileage days
Average annual kilometers
Section 2: Single-select questions and Likert-scale matrix Number of years before vehicle replacement
Vehicle purchase/lease plan Number of years until next vehicle purchase/lease
Replacement or incremental vehicle purchase/lease
Reason for vehicle purchase/lease
Expected purchase/lease price
New or used
Importance rating of key vehicle attributes
Section 3: Three-page text and images Introduction of four powertrains and the differences
Educational part Detailed description of attributes used in the SP scenarios
Section 4: Four SP scenarios Each scenario is similar to Fig. 3
SP scenarios
Section 5: Likert-scale matrices Thirty attitudinal statements (largely linked to Theory of Planned Behaviour)
Attitudinal statements
Transportation (2020) 47:1223–1250
 Table 2  (continued)
Survey sections Question types Survey elements
Section 6: Socioeconomics Single-select questions and textboxes Province of residence
Sex of householder
Postal code (home and work)
Dwelling type
Home ownership status
Number of years living at current place of residence
Marital status of Respondent
Household size
Transportation (2020) 47:1223–1250

Number of adults (18 +) in household

Important travel modes per household member
Relation between household members
Employment status per household member
Education per household member (18 +)
Number of licensed drivers
Household income
Language spoken most often at home
1231

13
1232 Transportation (2020) 47:1223–1250

Significant survey design decisions

In the course of developing this survey, there were some decisions made that involved
choosing one direction over another. The purpose of this section is to discuss some of
the most significant decisions in terms of the thinking that went into them.

In‑house or outsourced data collection

Surveys have moved from traditional means of data collection (e.g. mail-outs) to inter-
net-based methods that are more efficient in many aspects ranging from actual collec-
tion to implementing as usable data. Associated with the maturation of internet surveys
has been the rise of the survey panel where a firm will maintain a set of thousands of
survey panelists who are willing to take internet surveys on different topics over time in
exchange for incentives. The availability of survey panels means that researchers do not
need to find ways to access actual lists of potential respondents (e.g. e-mails) which had
been historically problematic. In the absence of a survey panel, it takes longer and there
is more uncertainty involved, in researchers reaching a target number of responses and
respondents may require repeated reminders.
A survey panel can provide a geographically comprehensive and representative sam-
ple for all states/provinces/regions in a shorter period of time across a country. SPACE
has been deployed for this on a large scale across Canada. As further examples within
the EV domain, Hidrue et al. (2011) used a panel to collect a national sample in the
United States, and Hackbarth and Madlener (2013) did so for Germany. One other
important advantage is that panelists have already expressed a general willingness to
participate in surveys and are often familiar with a range of survey types including those
with choice experiments.
While the clear advantages of a survey panel are very attractive, one possible drawback
is that the cost is likely to be higher than for traditional sample recruitment. Certainly, the
use of a survey panel has to be well-budgeted ahead of time, especially if it will be a large
survey. Another limitation is that panelists who are doing many surveys may compromise
data quality by rushing through the questions to get to the end and receive the incentive.
However, panel operators adopt policies such as removing careless respondents from the
panel to increase the quality of survey responses. For instance, our panel operator dropped
those respondents from the delivered data who finished the survey ‘too quickly’, i.e. in less
than 30% of the median survey completion time. There are still issues in an online panel
regarding inattentive and fully/partially-random responses, especially within the stated
preference components of surveys (Petrik et al. 2016), and thus post-collection screening
and cleaning is important (Meade and Craig 2011).
Also, panels can be associated with ethical concerns from a university research perspec-
tive. There may be issues having to do with the level and type of incentives that are offered
to respondents, loss of incentive for an incomplete survey, or concerns with an excessive
number of questions that are mandatory as opposed to optional. Another potential draw-
back of online panels is the issue of representativeness (Szolnoki and Hoffmann 2013)
as raised in Blasius and Brandt’s (2010) study which found that their online sample was
not representative of the population on some demographic and attitudinal characteristics.
Results in Table 3 suggest that our panel operator was able to deliver a sample that was rea-
sonably representative of the population for characteristics that we could compare.

13
Transportation (2020) 47:1223–1250 1233

Table 3  Sample versus Canadian Context SPACE (%) Census (%)

households as derived from
census Age (householder)
18–24 4 3
25–34 14 14
35–44 18 17
45–54 23 20
55–64 22 20
65 + 19 22
Education (householder)
No certificate; diploma or degree 3 17
High school diploma or equivalent 18 22
College, trades certificate or diploma 34 37
University certificate, diploma or 45 23
degree at bachelor level or above
Marital status (householder)
Single 25 18
Married or common law 69 58
Other 6 23
Household income (CAD$)
Less than $25,000 5 14
$25,000–$49,999 16 21
$50,000–$99,999 37 33
$100,000 and more 29 32
Refused 14 –
Language spoken most often at home
English 77 67
French 18 21
Other 5 12
Household size
1 18 28
2 45 34
3 17 15
4 14 14
5 or more 7 8
Dwelling type
Single detached house 65 54
Townhouse or semi-detached 12 12
Apartment or condo 21 34
Other 1 1
Dwelling tenure
Owner 78 68
Renter 22 32
Province
Newfoundland and Labrador 3 1
Prince Edward Island 1 0.4
Nova Scotia 5 3
New Brunswick 4 2

13
1234 Transportation (2020) 47:1223–1250

Table 3  (continued) Context SPACE (%) Census (%)

Quebec 22 23
Ontario 29 38
Manitoba 5 4
Saskatchewan 4 3
Alberta 12 12
British Columbia 15 13

Conceptualization of respondent and implications for screening

In SPACE, we stressed “households” as opposed to individual respondents. Respond-

ents were required to represent their household and provide information on their house-
hold members. The acquisition of an expensive, relatively infrequently renewed item like
a car can often be seen as a household decision. As such, while it was individuals who
responded, those who did not consider themselves as primary household decision-makers
(effectively household heads or maintainers) were screened out. This aspect was reinforced
several times during the survey. For example, phrases such as “your next household vehi-
cle” or “members of your household” were used as the survey progressed. The primary
householder criterion aligns with the other screening criterion that excluded respondents
aged less than 18. A potential limitation to both screens is that they de-emphasize young
people who are important to the future of electric mobility.
The third criterion was the inclusion of households who had at least some intention to
purchase/lease a vehicle (new or used) in the future. In contrast to previous works, respond-
ents were not required to be recent or urgent car buyers (e.g. Achtnicht 2012; Mabit and
Fosgerau 2011). Our sample, therefore, was more general as the only group we excluded
were those who were ‘not at all likely’ to purchase a vehicle in the foreseeable future. We
did not specify an exact timeframe for that purchase since it could potentially exclude some
car buyers that consider a vehicle purchase as a long-term decision. Also, those households
currently without a vehicle were not excluded from our database, as opposed to studies
(e.g. Hoen and Koetse 2014) that included car owners only (Fig. 1).
As Fig. 2 below makes clear, a very large collection of panelists indeed (35,795) at
least considered the completion of the survey. Of all those who considered the survey, only
about 57% actually saw the survey through to completion and this translated to the 20,520
final respondents. About 8.6% evaluated the survey’s letter of information and decided not
to participate. About 17.6% were willing to participate but were screened out if they failed
to meet any one of the criteria relating to age, being a household maintainer or being inter-
ested in acquiring a vehicle in the future. Finally, 16.5% decided to withdraw from the sur-
vey at some point after beginning to answer the questions or were eliminated from the final
database by the panel operator if not passing quality control criteria. With regard to the
latter, data were not shared by the survey operator to shed light on the nature and causes of
withdrawals.
Of the 17.6% that were screened out, not being willing to acquire a vehicle in the future
(11.6%) was about twice as important as not being a household head (5.9%) in influencing
the screening outcome. The loss of potential respondents because they were aged less than
18 was essentially a non-factor in the screening (0.1%). In terms of how these screening
results were applied, age was given first priority, household head was given second and the

13
Transportation (2020) 47:1223–1250 1235

8.6% 0.1%
Declined to parcipate (3,051)
5.9%

Screened; under 18 (52)

11.6% Screened; not household head

(2,129)
Screened; not at all likely (4,140)
57.3%
Withdrew from survey (5,903)
16.5%
Final sample (20,520)

Fig. 1  Inventory of outcomes for all panelists who considered the survey, the counts of respondents for each
part are shown in parenthesis

Fig. 2  Educational materials; four powertrains and their features

desire for a future vehicle was given third priority. Furthermore, the detailed time track-
ing option enabled by our panel operator gave us the ability to search for people that were
unengaged in some sections of the survey to be eliminated from the final sample for certain
types of analysis. Overall engagement can be assessed through total time spent on the sur-
vey and tendency to “straight-line” answers to suites of attitudinal questions.
Table 3 compares our sample with the 2016 Canadian census distribution of households
(or household maintainers) across a range of variables. Education and marital status of the

13
1236 Transportation (2020) 47:1223–1250

householders are derived from census 2011. Given the focus on people with an interest to
acquire vehicle, there are “built-in” differences between the characteristics of our study
population and the entire population of Canada. There is an over-representation of the more
highly educated and two- or three-person sized households. Partly such households have a
higher on-going intention to acquire a vehicle. But the higher educated are also more likely
to participate in survey panels and sufficient recruitment of one-person household panelists
seems challenging based on these results. Similar insights are likely to apply for income,
where there is less representation from lower-income households. Under “Province”, it can
be seen that the SPACE data collection was stratified with the less-populated provinces
more heavily sampled to ensure an adequate number of respondents from all provinces.

Educating respondents on powertrains and vehicle attributes

At the time (mid-2014) when survey development was initiated, the research team was of
the opinion that most respondents would likely have at least some misconceptions about
the main vehicle types under study and many would have no knowledge about plug-in elec-
tric vehicles (Krause et al. 2013). It was clear at the time anecdotally that such misconcep-
tions applied at dealerships let alone among consumers! In the philosophical discussions
which were held about the survey, the research team came to the conclusion that this sur-
vey might well influence potential car buyers who previously had incomplete information
about them. It was decided that there was not much we could do about that but that it was
better in any case for respondents to improve their knowledge about EVs as the survey
progressed. Current knowledge about EVs was not assessed at the beginning of the survey
though some information about general environmental attitudes was collected including
the connections between vehicles and the environment.
Measures were taken to adequately educate respondents so that they could make
informed choices within the SP section of the survey grounded on correct information. The
introductory script of the survey introduced the term “electric mobility” and that it was our
research mandate to investigate the topic but definitions were not provided. In retrospect,
and as cautioned by Dillman (1978), this introductory script could have been written in a
manner that was more “powertrain neutral.” It is possible that our script attracted a dispro-
portionate share of respondents with an interest in electric vehicles. At the start of the SP
section, the survey instrument clarified the types of powertrains that respondents would
consider. Figure 2 illustrates what was briefly presented and how we covered basic knowl-
edge about how the vehicles were powered, what types of behaviours were required of the
owner, and some implications of driving these vehicles. Gasoline vehicles were explained
in a similar way. Further details about vehicles types became clear as we described the
attributes associated with vehicles (Fig. 4) that would be assessed by the respondent during
the SP scenarios. These were treated as a second layer of fundamental information about
these vehicles and covered aspects such as fuel/charging costs and battery range among
several others.

Selection of SP attributes and levels

Given the continuing small market penetration of EVs, the Stated Preference (SP)
approach has remained the most widely applied methodology and the primary source
of data for obtaining people’s preferences towards such vehicles (Abotalebi et al. 2015).
In an SP scenario, respondents choose or rank their most preferred options, with these

13
Transportation (2020) 47:1223–1250 1237

options being characterized through a series of attributes and assigned “levels” for these
attributes in accordance with an experimental design (Louviere et al. 2000). The selec-
tion and quantity of attributes for SP scenarios is an important consideration (Hensher
et al. 2005; Louviere et al. 2000). Monetary attributes (i.e. purchase price and fuel,
maintenance or operational cost) have been included in almost all reviewed studies, as
well as charging time/availability and vehicle range (See Table 1 and Liao et al. 2017).
Other attributes such as emissions and various forms of incentives have been considered
less often.
The selection of SP attributes for the present study was jointly guided by the lit-
erature (Table 1) and feedback from partners. We sought specifically to understand the
importance of technical EV attributes that include range, charging time, fuel/mainte-
nance cost savings, emission reduction, and performance; as well as the impact of pol-
icy measures including the quality of the battery warranty, government cash incentives,
and various non-cash incentives (e.g. access to high occupancy lanes, free municipal
parking and exemptions from tolls). Overall, 12 attributes were included in each SP sce-
nario covering all the items under study. On the SP screen, attributes appeared in four
main categories; cost, operational, non-cash incentives, and charging (Fig. 3). This was
preceded by another screen informing respondents about the attributes included in the
SP scenarios (Fig. 4). The information from Fig. 4 was accessible as pop-up text via
icons in Fig. 3. Hence, there was an opportunity to refresh one’s memory about the
attributes once the scenarios were under way.

Fig. 3  A sample stated preference choice scenario

13
1238 Transportation (2020) 47:1223–1250

Fig. 4  Survey overview material on selected SP vehicle attributes

A few salient points about the chosen attributes are worth noting. First, purchase price
and government cash incentives appeared as a single line item on the SP screen, as a “net”
purchase price calculation, even though they were treated as two distinct attributes in the
experimental design. The highest level for cash incentives reflected the maximum ($10 k)
offered in Canada at the time. Subsequently, even higher incentives (up to nearly $15 k)
were offered in Canada and our experimental design would have benefitted from capturing
the same.
Interestingly, range was treated in our implementation as two distinct attributes: gaso-
line and battery range. The results from a national analysis (Ferguson et al. 2018) showed
the utility of this approach. For example, willingness-to-pay analysis found that HEV-ori-
ented households placed a very high per km value on gasoline range while those inclined
towards BEVs shunned consideration of gasoline range and instead highly valued per km
increases in electric range.
Third, we distinguished between charging times for home/work and public stations, with
the latter generally being associated with faster rates of charging in the experimental design
since public charging is typically more “opportunistic” and time-constrained in nature. The
approach to separate and independently explore the primary charging contexts differed
from previous EV surveys, with the exception of Jensen et al. (2013) where home, work,
and public stations were presented as distinct attributes.

SP alternatives and their identification to respondents

As our research mandate was to focus on electric vehicles, we included EV-related tech-
nologies, hybrid (HEV), plug-in hybrid (PHEV) and battery electric vehicles (BEV), to
compete with internal combustion engines (ICE) in SP scenarios. Within the literature, a
considerable range of powertrains/fuel types are utilized (Table 1). With the exception of

13
Transportation (2020) 47:1223–1250 1239

Potoglou and Kanaroglou (2007b), ICE and BEV define two of the powertrains in all stud-
ies reviewed. Other fuel types such as natural gas, hydrogen, and biofuel that are included
in a number of studies (Achtnicht 2012; Hackbarth and Madlener 2013; Ziegler 2012)
were not included in SPACE as these were judged outside our EV-oriented mandate and
are not widely available within the Canadian vehicle market. HEV and PHEV were only
considered in a subset of past studies, however when included, they achieved considerable
predicted market shares sometimes larger than BEVs (Axsen et al. 2015; Hackbarth and
Madlener 2013; Helveston et al. 2015; Massiani 2014).
Apart from types and number of alternatives, another important decision was the pres-
entation of alternatives on the SP screen, whether in a labeled or unlabeled form. Some
previous works designed their choice sets as unlabeled experiments (Helveston et al. 2015;
Hoen and Koetse 2014; Ziegler 2012), while the choice of fuel type was nevertheless
incorporated in their modeling analyses as the dependent variable. The term quasi-labeled
has been used for such experiments (Ziegler 2012), meaning that while the names of alter-
natives convey no information to the respondents, their attributes and levels are designed in
a way that relates exactly to one powertrain or fuel type.
On a labeled choice, however, vehicle powertrains can be identified before reviewing
the attributes associated with each alternative (Hidrue et al. 2011; Jensen et al. 2013; Mabit
and Fosgerau 2011; Tanaka et al. 2014). As such, the name each alternative carries can
have a significant influence on the choices that respondents make. For instance, some peo-
ple may select a BEV simply because of their pro-environmental attitudes, without consid-
eration of attributes associated with that alternative. While the trade-offs between attrib-
utes are somewhat downgraded in this approach, labeled experiments are more in line with
choices people make in the real world where a number of branded goods or services are
considered (Hensher et al. 2005). Indeed, the label given to each alternative adds an extra
level of realism to SP surveys. Also, labeled alternatives are preferred when the focus is on
prediction and forecasting (Hensher et al. 2005).
In SPACE, the SP scenarios consisted of four fixed ‘labeled’ alternatives illustrated with
a related visual design (Fig. 3). Respondents were required to make a choice between these
four alternatives for the primary choice context, which most often was for a replacement
vehicle. A secondary choice context was also examined but this choice was for five alterna-
tives including an opt-out option. The reasoning for this had to with context sensitivity and
is thus discussed in Sect. 4.4.

Choice of design technique for sp scenarios and implications

D-efficient or Efficient experimental design was employed for generation of our SP sce-
narios (Hensher et al. 2005; Street et al. 2005). Orthogonal and D-efficient design are the
two main design types used in previous EV studies (See Table 1). Although orthogonal
has been used most often within EV research, evidence from the literature demonstrates
the outperformance of efficient design (Bliemer and Rose 2011; Carlsson and Martinsson
2003; Rose and Bliemer 2008). Bliemer and Rose (2011), for example, tested the impact of
different experimental design types (i.e. orthogonal vs. efficient) on the estimation results,
and found that efficient design lowered standard errors in the estimation of parameters. The
efficiency of experimental design can be further improved, if there is a priori information
available about coefficients of attributes (e.g., price, range, etc.), perhaps through the litera-
ture, pilot surveys, or secondary data (Bliemer and Rose 2011; Huber and Zwerina 1996;
Rose and Bliemer 2008).

13
1240 Transportation (2020) 47:1223–1250

The pre-known information about an attribute, the so-called “prior,” is an initial value
used as a coefficient when computing utilities for each alternative in a SP scenario (Carls-
son and Martinsson 2003; Huber and Zwerina 1996). In the present study, we conducted a
pilot survey using an orthogonal design (Louviere et al. 2000), based on priors being equal
to zero (we assumed no information was available). A subset of the estimated coefficients
from the pilot survey were used as priors for our final survey. We only used those coef-
ficients that were statistically significant with expected signs. For instance, the estimated
coefficient for the price attribute was a significant, negative value, and hence was used as
a prior to generate an efficient design for our final survey. We assigned no prior to those
attributes that had insignificant or counterintuitive coefficients, as it could result in a nega-
tive impact on the efficient design (Bliemer and Rose 2011).
The Ngene program was used for both pilot and final surveys to generate choice scenar-
ios for each of seven vehicle classes (See Table 4). In the experimental design, 48 scenarios
were divided into 12 blocks, and each respondent was randomly assigned to one of these
blocks, with 4 distinct scenarios. The design steps in Ngene begin by specifying the func-
tional form of the utility for each alternative and choice probabilities are calculated for each
scenario. This process is possible when priors are used, so the analyst can check for the
potential ‘dominancy’ of a certain alternative, which is considered as an issue for a more
traditional orthogonal design (Bliemer and Rose 2011). This ensures that there is a reason-
able balance between utilities of all alternatives within an SP scenario, so that respondents
actively “trade-off” when deciding on their choices (Huber and Zwerina 1996).
The efficiency of the design is assessed by the D-error which is the most widely used
measure of the goodness-of-fit for experimental design (Rose and Bliemer 2008). The
D-error does not have a unit and its magnitude depends on the units of the design attrib-
utes (Rose and Bliemer 2008). In practice, it is almost impossible to find a design with
zero D-error; therefore, researchers are satisfied if the design has a sufficiently low D-error,
depending on the units of the design attributes. This is called a D-efficient design (Huber
and Zwerina 1996; Rose and Bliemer 2008). If the error is large, the steps should be
repeated by changing the variables or priors. In SPACE, we developed the experimental
design based on attribute levels associated with each vehicle body type (See Sect. 4.1), and
the design with minimized D-error was used after several iterations, similar to the approach
adopted by Hensher et al. (2011).

Respondent cognitive burden

One challenge in the design of a survey that seeks to be comprehensive is to gather as

much high-quality information about respondents as possible while appropriately manag-
ing cognitive burden (Stopher 1998). Generally, surveys are not seen as enjoyable by many
respondents, so the degree to which a survey instrument is perceived as difficult, time con-
suming, or stressful is highly relevant. Partly the burden can be managed through economi-
cal and simple wording of questions and proper sequencing of questions.
Historically, cognitive burden has been discussed as an issue for SP scenarios (e.g.
Achtnicht 2012; Bunch et al. 1993; Potoglou and Kanaroglou 2007a), but the requirements
for other survey components are considerable and impose some burden on the respondents.
It also has to be considered that there are survey section interdependencies in burden: for
example, if a respondent is tired or frustrated at the beginning of an SP section, this can
only hurt the ultimate quality of the SP data.

13
 Table 4  Base values for SP scenarios
Transportation (2020) 47:1223–1250

Body type image Vehicle size Purchase price ($) Maintenance cost Acceleration (sec to Gasoline range CO2 tailpipe emissions Fueling/charg-
($/km) reach 100 km/h) (km) (tonnes/km) ing cost ($/km)

Economy 22,000 0.051 8.8 700 0.000143 0.07

Intermediate 25,000 0.051 9.0 700 0.000141 0.07
Full sedan 35,000 0.060 7.2 800 0.000182 0.08
Luxury sedan 59,000 0.072 6.5 700 0.000202 0.09
Minivan 33,500 0.069 5.0 700 0.000234 0.14

SUV 30,000 0.072 9.0 670 0.000234 0.09

Pickup truck 40,000 0.081 6.6 750 0.000444 0.13

1241

13
1242 Transportation (2020) 47:1223–1250

In assessing the question of survey completion times for SPACE, it became clear in
retrospect that there were issues of differential survey burden across respondents. For
example, SPACE collects several pieces of information over several questions on each
vehicle in the household. The questions gathered data such as daily trip and mileage
patterns and other aspects. Estimates were also asked regarding vehicles for which the
respondent may not have been the primary driver. A similar dynamic, though less oner-
ous, was at play in the gathering of information about household members. Respondents
were required to provide this information and some other similar inquiries (See Table 2)
for up to seven household members and as many as four vehicles registered by their
households. Hence, a respondent with many family members and several vehicles had
more to manage.
Figure 5a suggests that the questioning associated with vehicles seemed to add more
respondent burden than the questioning associated with household members and broadly
that more vehicles and more household members added to completion time. However,
another important factor in assessing the relationship is the age of the respondent as is
seen in Fig. 5b. There is evidence that regardless of the number of vehicles in the house-
hold, younger people completed the survey faster. Each age group shows that a large
number of vehicles imposes a burden but, this burden is of a large absolute magnitude
for older respondents. Older people seemed to get more “bogged down” in the questions
on vehicles and even with relatively few vehicles. Bear in mind that these results are
medians and there would be many examples of larger burden differentials on a respond-
ent-by-respondent basis.
Within the SP section, we decided on fewer scenarios (four per respondent) to further
manage cognitive burden. This also provided an opportunity to include more attributes
(12 attributes) per scenario to best capture the range of factors taken into account in an
actual vehicle purchase. We might have opted for a fewer number of attributes similar to
most previous EV studies (See Table 1), but this was not judged as in line with the real-
world complexity of a vehicle purchase decision. Appealing visual design along with
informative pop-up text for all graphic icons on the SP screen were additional elements
that sought to reduce per-scenario burden for the respondents. Also, as previously noted,
we included attributes in four main categories: cost, operational, non-cash incentives,
and charging (Fig. 3) to assist respondents in sorting through the attributes to offset the
additional burden in this respect.
Median survey completion time (min)

23 23
22 22
21 21
20 20
19 19
18 18
17 5+ HH size 17
16 4 HH size 16 3+

15 3 HH size 2 vehicles
15
14 2 HH size 1 vehicle
1 HH size
14
No vehicle
13 13
4 vehicles 3 vehicles 2 vehicles 1 vehicle No vehicle
(1,354 HH) (1,727 HH) (8,045 HH) (9,274 HH) (821 HH)
65+ 55-64 45-54 35-44 25-34 18-24

(a) (b)
Fig. 5  Median survey completion time by combination of household size, number of vehicles and age
groups

13
Transportation (2020) 47:1223–1250 1243

There were other aspects to respondent burden in the survey. Attitudinal questions (i.e.
thirty Likert scale items), were dispersed in groups through the survey to reduce monotony.
Another strategy was to leave socioeconomic inquiries to the end, as survey panelists see
these types of questions a lot and are more comfortable with them.
Overall, there is evidence from the results here that the research team, as with several
past studies, viewed the issue of survey burden too much from the viewpoint of the choice
experiment and overlooked ways that additional burden was being added for some respond-
ents in other sections of the survey. In retrospect, for example, we could have reduced the
number of questions that asked about all vehicles in the household. We could have focused
more specifically on the respondent’s vehicle and asked about other household vehicles
only for the more important points (e.g. annual km driven).

Novel aspects of the survey design

While the previous section highlighted certain design dilemmas and some challenging
decisions that were made, the current section focuses on novel features of the survey. For
these, a consensus was reached among team members that implementation would improve
the survey instrument and perhaps have our effort stand out from other work in the field.

Vehicle body type conceptualization

The preferred vehicle body type of households shaped the implementation of our SP exper-
iments. In many previous SP surveys, respondents were exposed to the attribute levels
that were not specifically designed for their vehicle body type of interest. This reduces the
ability of respondents to relate to scenarios they evaluate. In SPACE, the attribute levels
were anchored realistically on a single vehicle body type that respondents were asked to
identify prior to the SP experiment. Seven vehicle classes: economy, intermediate sedan,
full sedan, luxury sedan, minivan/crossover, sport utility vehicle (SUV), and pick-up truck
were presented to respondents. Related images and approximate prices were provided that
corresponded to a conventional gasoline version of each type (Table 4). The respondents’
selected vehicle body type applied to all scenarios and was used as the basis to derive
attribute levels per alternative (Table 5). This method added to the relevance and realism of
the attribute levels being evaluated by the respondent (Beck et al. 2013; Rose and Bliemer
2008). For instance, a household interested in an economy size vehicle would see matching
price or fuel cost levels in scenarios.
While the factor of vehicle body type has been included in a number of previous EV
studies, it was either treated as an attribute within scenarios (Potoglou and Kanaroglou
2007a) or the same attribute levels were utilized across different vehicle types. Our seg-
menting of the vehicle market into seven classes opened up the opportunity to investigate
how preferences for EVs change according to households’ preferred vehicle body types as
in the study by Higgins et al. (2017).

Emphasis on detailed geography and location

Geography played a significant role in the conceptualization of SPACE. Fundamental

in this regard, was our collection of 6-digit postal codes that could fairly precisely
locate our respondents, and identify the spatial context in which they would assess

13
 1244

13
Table 5  Relative attribute levels used in SP scenarios
Attributes Alternatives
ICE HEV PHEV BEV

Purchase price ($) − 25%, Base Base, + 50% Base, + 50%, + 100% Base, + 50%, + 100%
Cash incentive ($) 0 0 0, 5000, 10,000 0, 5000, 10,000
Maintenance cost ($/km) Base − 25%, Base − 50%, − 25%, Base − 25%, − 50%, − 75%
Acceleration (sec to reach 100 km/h) Base − 50%, Base, + 50% − 50%, Base, + 50% − 50%, Base, + 50%
Battery range (km) – – 30, 60, 90 150, 250, 350
Gasoline range (km) Base − 25%, + 25% − 25%, + 25% –
CO2 tailpipe emissions (tonnes/km) − 25%, Base − 50%, − 25% − 75%, − 50% 0
Fueling/charging cost ($/km) Base, + 20%, + 40% − 20%, Base − 60%, − 40%, − 20% − 80%, − 60%
Public charging time (h) – – 0.15, 1.5, 2.75 0.25, 3, 5.65
Home/work charging time (h) – – 2, 3, 4 3.5, 5.25, 7
Fueling/charging station availability All gas stations All gas stations 10%, Same, twice 10%, Same, twice
Battery warranty – – 3 years/58,000 km, 5 years/96,000 km, 3 years/58,000 km, 5 years/96,000 km,
8 years/160,000 km 8 years/160,000 km
Non-cash incentives – – HOV lane access, free parking, free toll HOV lane access, free parking, free toll
roads roads

Same same number of stations as number of current gas stations, Twice twice as many stations as current number of gas stations, HOV lane access high occupancy vehicle
lane access
Transportation (2020) 47:1223–1250
Transportation (2020) 47:1223–1250 1245

electric mobility. For example, whether potential EV buyers are more concentrated in
the central cities or suburban areas, also, whether or not EVs are more preferred among
single-family housings with access to a private garage as opposed to apartment occu-
pants where charging circumstances may have been less favourable. Such segmentation
analysis can be conducted at the small census area level (dissemination area) using
locational and other variables from SPACE. Such a fine level of spatial detail is not
common within the EV literature. Campbell et al. (2012) is one of the few studies that
derived a geographically-oriented segmentation system, though solely through the use
of census data; by means of variables such as income, age, and home ownership that
were considered important in characterizing the adoption of alternative fuel vehicles.
Capturing the locational context of households permits consideration of research
aspects such as: optimum locations for public charging infrastructure, identification of
places with high demand for home EV charging and resulting impact on the electric-
ity distribution network. Information on workplace location (i.e. postal code or nearest
major intersection) was also collected to derive a sense of the spatial scope of daily
activities and implications for EV range.

Emphasis on collection of attitudes

An extensive suite of Likert-based attitudinal statements was included in SPACE to

support survey analysis from multiple aspects. The focus of the statements was to
extend the Theory of Planned Behaviour (TPB), developed by Ajzen et al. (1991)
Among other studies, Egbue and Long (2012), Lane and Potter (2007), and Moons
and de Pelsmacker (2012) applied TPB in their EV analysis, though within different
research contexts since the extended TPB can be customized based on the objectives of
each study (Mohamed et al. 2016).
In SPACE, suites of attitudinal statements were developed to capture several behav-
ioural constructs and intentions as they relate to EV adoption behavior. The captured
attitudes acted as a primary outcome factor in the structural equation analysis of
Mohamed et al. (2016) that offered independent behavioural analysis of EV purchase
intentions, and were used to model choice outcomes in the SP analysis of Ferguson
et al. (2018). The inclusion of attitudinal statements combined with an emphasis to
support multiple analysis approaches is rare in the previous works. The attitudinal
component offers the possibility to be used as an explanatory role in choice model-
ling related to the SP analysis and is also useful in carrying out detailed geographical
and psychographic-based market segmentation analysis. These offer insights into how
attitudes vary over space. Altogether, the SP section and attitudinal part worked as
complementary in understanding consumer behaviours towards EVs.
A list of attitudinal statements can be found at Mohamed et al. (2016) along with
details of analysis. The developed model includes six constructs: environmental con-
cern, attitude towards adopting EVs, subjective norm, perceived behavioural control,
personal moral norm, and EV adoption intention. One possible improvement is the
inclusion of one more construct to measure the level of public awareness on EVs. This
aspect emerged from our industrial partners later on but was not emphasized during
survey development.

13
1246 Transportation (2020) 47:1223–1250

Context sensitivity

The implementation of the SP experiments in the survey instrument was rather novel in that
it was sensitive to the respondents’ specific circumstances; annual mileage, purchase plan (i.e.
incremental or replacement vehicle), preferred vehicle body type and anticipated purchase
price. In the EV literature, this is referred to as customized design, and was applied in a num-
ber of studies (e.g. Bunch et al. 1993; Hidrue et al. 2011; Potoglou and Kanaroglou 2007a). In
SPACE, the levels associated with tailpipe emissions, fueling/charging cost, and maintenance
cost were adjusted based on respondent’s annual mileage and preferred vehicle body types
reported earlier in the survey. Fueling/charging cost, for example, can vary a lot across vehicle
body types.
The choice box at the bottom of the SP screen (see Fig. 3) was also customized according
to a respondent’s previously provided information. In the example of Fig. 3, the respondent
had earlier identified that the household sought a replacement vehicle as opposed to an incre-
mental vehicle. For this reason, the first row of the choice box dealt with the powertrain that
would be selected for the replacement vehicle. As a potential source of further useful data, we
gave the respondent a second, more speculative, choice to make as to what powertrain they
would choose in the event that they did, for whatever reason, decide to add an incremental
vehicle. Note that this incremental vehicle is labelled in Fig. 3 as the “2nd household vehicle.”
This indicates that the respondent had earlier identified that the household currently operated
one single vehicle. Because this choice context was a secondary and low-probability one, the
option to choose no vehicle for that choice context was provided. For each respondent, the
presentation at the bottom of the SP screen would adjust to their household circumstances. For
the minority of households that preferred an incremental vehicle, the first choice would relate
to that circumstance and the second choice with the opt-out option would relate to a replace-
ment scenario.
To summarize, the first and primary choice context was implemented as a forced choice,
meaning that respondents had no choice for declining all four vehicles presented per scenario.
But the secondary choice context permitted an opt-out and offered five options from which
to choose. In the case of a younger household, contemplating their first vehicle purchase, the
secondary choice context would have been for a replacement vehicle. The opt-out option was
clearly needed for this non-plausible scenario. While most secondary choice contexts were at
least plausible, they were lower probability scenarios.
For the primary choice context, the one that respondents had said they were more likely to
consider, we saw the “opt out” option is in some ways being similar to a “prefer not to answer”
option and thus decided against it. The thinking was that it was better to get some insight on
preferred powertrains than none. It was important, for example, to know that a respondent
gravitated to an ICE vehicle. To the extent that even the primary choice context was not desir-
able for some respondents (e.g. maybe they tended to buy cheaper, used vehicles in real life)
there was the option to identify such respondents and remove from analyses as necessary (See
Table 2. Section 2: Vehicle purchase/lease plan). Meanwhile, we had more insight about their
vehicle preferences than we would have otherwise, had we included an opt-out in the primary
choice context.

13
Transportation (2020) 47:1223–1250 1247

Conclusions

This paper has reviewed the process that led to the development of a national survey
instrument to assess consumer prospects for electric vehicles in Canada. The survey was
deployed in the spring of 2015 to collect a consumer data set from over 20,000 households
across Canada. The benefit of hindsight has offered us an opportunity to assess the process
and what has been learned as it relates to the development of a survey instrument. Actual
data collected, including information on respondent completion durations by survey sec-
tion, has assisted in the overall assessment.
Arguably, the most important survey decision was made in 2013, long before the survey
instrument was constructed. At that time, the research team identified that a privately man-
aged consumer survey panel offered an unparalleled opportunity to explore, in consider-
able detail, Canadian preferences and attitudes toward electric vehicles. It was anticipated
that a well-constructed national sample, if large enough, would also provide meaningful
sub-samples at the level of Canadian provinces, regions, metropolitan areas and even at
the intra-urban level. Planning and budgeting well in advance were certainly important ele-
ments in the comprehensive nature of the sample that was facilitated. The decision to go
with a survey panel removed traditional researcher anxiety about issues such as response
rates and the ultimate number and nature of observations that the team would have at their
disposal. However, survey panels are not without risk. Costs are higher and there can be
issues with developing a representative sample since some types of respondents (e.g. sin-
gle-person households, the less educated, etc.) are harder to capture as survey panelists.
Some of the collected data may be suspect, and hard to diagnose as such, when survey-
hardened respondents answer inattentively.
To capture geographical locations, we collected detailed six-digit postal codes from sur-
vey respondents. As such, an important research theme to focus on detailed spatial varia-
tion in vehicle powertrain preferences was enabled and the extensive nature of the panel
itself supported this emphasis. That nature of the panel also dictated that we would deal
with individual survey respondents. We would not be able to gather data from “house-
holds” per se but we did conceptualize that respondents could be asked, subject to screen-
ing criteria, to act as representatives of the household and would provide information about
the household as a whole. This focus on household “heads” had the effect, for better or
worse, to weed out young adults still living with their parents.
Several components of the survey were novel. First, our implementation of vehicle
attributes in choice experiments was done in such a way as to be sensitive to the vehicle
body types that households preferred for their next vehicle. As such, luxury car buyers
were not seeing scenarios, for example, that might have seemed more appropriate for an
economy car buyer. Past EV literature had tended to outline more generic approaches. The
choice scenarios also benefitted from several aspects that were implemented to be context-
sensitive to the earlier answers of respondents. Another novel aspect was that the survey
simultaneously emphasized in-depth choice experiments and an extensive collection of
Likert-based attitudinal indices. Among other benefits, the two approaches offered redun-
dancy in measuring behavioural intention toward acquiring electric vehicles.
The research team paid considerable attention to cognitive burden in the development of
the survey. An important focus was not overwhelming respondents as they made their way
through the choice experiments. Ordering of survey sections was an important considera-
tion as was the core nature of the stated preference scenarios themselves. Interestingly, the
research team only really realized in retrospect that “do loops” built into the survey based

13
1248 Transportation (2020) 47:1223–1250

on the number of registered household vehicles and the number of household occupants
created some significant differential burdens in completing the survey. The survey was
much more onerous for some people depending on their circumstances. We would expect
future survey implementations by the research team to be more sensitive to this aspect.
Lessons were learned from the experience of developing SPACE. One was that the
development of a comprehensive survey instrument is quite labour intensive and numerous
iterations are required especially when partners and stakeholders play a role. The process
is not to be taken lightly. It is very important to consider carefully all the aspects that are
to be covered and even so, it is likely in retrospect that overlooked questions will be identi-
fied. While SPACE was quite comprehensive, there were omissions and probably the most
significant was in not thoroughly assessing respondent knowledge about electric vehicles.
This was an aspect that partners were requesting after the fact, since it is useful information
to benchmark over time, but the collective team including partners did not clearly identify
this at the outset. Nevertheless, some attitudinal statements included in the survey gave
some sense about aspects such as knowledge of current charging infrastructure and places
where EVs could be bought.
Another surprising oversight had to do with gathering written comments from respond-
ents. While we asked respondents to contact the research team with any comments or ques-
tions, it would have been much better to include a comment box directly within the survey.
Subsequent work with our other surveys has shown that some respondents are willing to
write significant passages about the topics examined in surveys. Not directly including this
functionality greatly reduced the amount of qualitative information that we gathered. With
respect to other survey respondents, the type that typically would be less likely to leave
detailed comments, it was learned that it is a good idea to time track respondents as this
gives a good sense of whether questions are being considered carefully. This aspect is par-
ticularly important with the rise of the survey panel where people may respond to a large
volume of surveys.

Acknowledgements We would like to thank the editor (Dr. Patricia Mokhtarian) and three anonymous
reviewers for providing insightful comments to improve our paper. This work was supported financially by
the Social Sciences and Humanities Research Council of Canada (Grant No: 886-2013-0001).

References
Abotalebi, E., Mahmoud, M., Ferguson, M.R., Kanaroglou, P.S.:Utilizing stated preference in electric vehi-
cle research; evidence from the literature. In: Proceedings of the 50th Annual Canadian Transport
Research Forum Conference, Montreal, pp. 1–19 (2015)
Achtnicht, M.: German car buyers’ willingness to pay to reduce ­CO2 emissions. Clim. Change 113, 679–697
(2012)
Ajzen, I.: The theory of planned behaviour. Org. Behav. Hum. Decis. Process. 50, 179–211 (1991)
Axsen, J., Bailey, J., Castro, M.A.: Preference and lifestyle heterogeneity among potential plug-in electric
vehicle buyers. Energy Econ. 50, 190–201 (2015)
Beck, M.J., Rose, J.M., Hensher, D.: Environmental attitudes and emissions charging: an example of policy
implications for vehicle choice. Transp. Res. Part A Policy Pract. 50, 171–182 (2013)
Blasius, J., Brandt, M.: Representativeness in online surveys through stratified samples. Bulletin de
me´thodologie sociologique 107, 5–21 (2010)
Bliemer, M.C.J., Rose, J.M.: Experimental design influences on stated choice outputs: an empirical study in
air travel choice. Transp. Res. Part A Policy Pract. 45, 63–79 (2011)
Bunch, D.S., Bradley, M., Golob, T.F., Kitamura, R., Occhiuzzo, G.P.: Demand for clean-fuel vehicles in
California: a discrete-choice stated preference pilot project. Transp. Res. Part A Policy Pract. 27, 237–
253 (1993)

13
Transportation (2020) 47:1223–1250 1249

Campbell, A.R., Ryley, T., Thring, R.: Identifying the early adopters of alternative fuel vehicles: a case
study of Birmingham, United Kingdom. Transp. Res. Part A Policy Pract. 46, 1318–1327 (2012)
Carlsson, F.: Design of stated preference surveys: is there more to learn from behavioral economics.
Environ. Resour. Econ. 46, 167–177 (2010)
Carlsson, F., Martinsson, P.: Design techniques for stated preference methods in health economics.
Health Econ. 12, 281–294 (2003)
Dillman, D.: Mail and Telephone Surveys: The Total Design Method. Wiley, New York (1978)
Egbue, O., Long, S.: Barriers to widespread adoption of electric vehicles: an analysis of consumer atti-
tudes and perceptions. Energy Policy 48, 717–729 (2012)
Ewing, G.O., Sarigöllü, E.: Car fuel-type choice under travel demand management and economic incen-
tives. Transp. Res. Part D Transp. Environ. 3, 429–444 (1998)
Ferguson, M., Mohamed, M., Higgins, C.D., Abotalebi, E., Kanaroglou, P.: How open are Canadian
households to electric vehicles? A national latent class choice analysis with willingness-to-pay and
metropolitan characterization. Transp. Res. Part D Transp. Environ. 58, 208–224 (2018)
Hackbarth, A., Madlener, R.: Consumer preferences for alternative fuel vehicles: a discrete choice analy-
sis. Transp. Res. Part D Transp. Environ. 25, 5–17 (2013)
Helveston, J.P., Liu, Y., Feit, E.M., Fuchs, E., Klampfl, E., Michalek, J.J.: Will subsidies drive elec-
tric vehicle adoption? Measuring consumer preferences in the US and China. Transp. Res. Part A
Policy Pract. 73, 96–112 (2015)
Hensher, D., Rose, J., William, W.: Applied Choice Analysis: A Primer. Cambridge University Press,
Cambridge (2005)
Hensher, D.A., Beck, M.J., Rose, J.M.: Accounting for preference and scale heterogeneity in establishing
whether it matters who is interviewed to reveal household automobile purchase preferences. Envi-
ron. Resour. Econ. 49, 1–22 (2011)
Hidrue, M.K., Parsons, G.R., Kempton, W., Gardner, M.P.: Willingness to pay for electric vehicles and
their attributes. Resour. Energy Econ. 33, 686–705 (2011)
Higgins, C.D., Mohamed, M., Ferguson, M.R.: Size matters: how vehicle body type affects consumer
preferences for electric vehicles. Transp. Res. Part A Policy Pract. 100, 182–201 (2017)
Hoen, A., Koetse, M.J.: A choice experiment on alternative fuel vehicle preferences of private car own-
ers in the Netherlands. Transp. Res. Part A Policy Pract. 61, 199–215 (2014)
Horne, M., Jaccard, M., Tiedemann, K.: Improving behavioral realism in hybrid energy-economy models
using discrete choice studies of personal transportation decisions. Energy Econ. 27, 59–77 (2005)
Huber, J., Zwerina, K.: The importance of utility balance in efficient choice designs. J. Mark. Res. XXX-
III, 307–317 (1996)
Jensen, A.F., Cherchi, E., Mabit, S.L.: On the stability of preferences and attitudes before and after expe-
riencing an electric vehicle. Transp. Res. Part D Transp. Environ. 25, 24–32 (2013)
Krause, R.M., Sanya, R.C., Lane, B.W., Graham, J.D.: Perception and reality: public knowledge of plug-
in electric vehicles in 21 U.S. cities. Energy Policy 63, 433–440 (2013)
Lane, B., Potter, S.: The adoption of cleaner vehicles in the UK: exploring the consumer attitude-action
gap. J. Clean. Prod. 15, 1085–1092 (2007)
Liao, F., Molin, E., van Wee, B.: Consumer preferences for electric vehicles: a literature review. Transp.
Rev. 37, 252–275 (2017)
Louviere, J., Hensher, D., Joffre Swait, J.: Stated Choice Methods: Analysis and Applications. Cam-
bridge University Press, Cambridge (2000)
Mabit, S.L., Fosgerau, M.: Demand for alternative-fuel vehicles when registration taxes are high. Transp.
Res. Part D Transp. Environ. 16, 225–231 (2011)
Massiani, J.: Stated preference surveys for electric and alternative fuel vehicles: are we doing the right
thing? Transp. Lett. 6, 152–160 (2014)
Meade, A.W., Craig, S.B.: Identifying careless responses in survey data. In: Proceedings of the 26th
Annual Meeting of the Society for Industrial and Organizational Psychology, Chicago (2011)
Mohamed, M., Higgins, C., Ferguson, M., Kanaroglou, P.: Identifying and characterizing potential
electric vehicle adopters in Canada: a two-stage modelling approach. Transp. Policy 52, 100–112
(2016)
Mohamed, M., Higgins, C.D., Ferguson, M., Réquia, W.J.: The influence of vehicle body type in shap-
ing behavioural intention to acquire electric vehicles: a multi-group structural equation approach.
Transp. Res. Part A 116, 54–72 (2018)
Moons, I., de Pelsmacker, P.: Emotions as determinants of electric car usage intention. J. Mark. Manag.
28, 195–237 (2012)
Petrik, O., De Abreu e Silva, J., Moura, F.: Stated preference surveys in transport demand modeling:
disengagement of respondents. Transp. Lett. Int. J. Transp. Res. 8, 13–25 (2016)

13
1250 Transportation (2020) 47:1223–1250

Potoglou, D., Kanaroglou, P.S.: An internet-based stated choices household survey for alternative fuelled
vehicles. Can. J. Transp. 1, 36–55 (2007a)
Potoglou, D., Kanaroglou, P.S.: Household demand and willingness to pay for clean vehicles. Transp. Res.
Part D Transp. Environ. 12, 264–274 (2007b)
Rose, J.M., Bliemer, M.C.J.: Stated preference experimental design strategies. In: Hensher, D.A., Button,
K.J. (eds.) Handbook of Transport Modelling, 2nd edn, pp. 151–180. Emerald Group Publishing Ltd.,
Bingley (2008)
Stopher, R.P.: A review of separate and joint strategies for the use of data on revealed and stated choices.
Transp. J. 25, 187–205 (1998)
Street, D.J., Burgess, L., Louviere, J.J.: Quick and easy choice sets: constructing optimal and nearly optimal
stated choice experiments. Int. J. Res. Mark. 22, 459–470 (2005)
Szolnoki, G., Hoffmann, D.: Online, face-to-face and telephone surveys—comparing different sampling
methods in wine consumer research. Wine Econ. Policy 2, 57–66 (2013)
Tanaka, M., Ida, T., Murakami, K., Friedman, L.: Consumers’ willingness to pay for alternative fuel vehi-
cles: a comparative discrete choice analysis between the US and Japan. Transp. Res. Part A Policy
Pract. 70, 194–209 (2014)
Ziegler, A.: Individual characteristics and stated preferences for alternative energy sources and propulsion
technologies in vehicles: a discrete choice analysis for Germany. Transp. Res. Part A Policy Pract. 46,
1372–1385 (2012)

Elnaz Abotalebi is a Ph.D. student in the School of Geography and Earth Sciences at McMaster University.
Her research focuses on consumer demand for electric vehicles (EV) in Canada using statistical modelling
to identify population segments that are most likely to consider an EV as their next vehicle purchase.

Mark R. Ferguson is Senior Research Associate at the McMaster Institute for Transportation and Logistics,
and Industry Professor in the Department of Civil Engineering at McMaster University. His research inter-
ests focus on electric mobility, the emergence and impacts of new transportation technologies and more
generally on the sustainable movement of people and goods.

Moataz Mohamed is an Assistant Professor of Smart Systems and Transportation Engineering at McMaster
University. His research focuses on the systemic impacts of disruptive technologies on mobility in general
with emphasis on electric vehicles, electric urban transit systems, ride sourcing, mobility as a service, and
infrastructure systems cascading behaviour.

Darren M. Scott is a Professor of Geography at McMaster University. As a transportation geographer, his

research interests include active transportation modes, activity-travel behavior, critical transportation infra-
structure, disruptive mobility technologies, geographic information science, sustainable transportation, and
time geography.

13