International Journal of Sustainable Transportation

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/ujst20

Adoption of electric vehicles: Which factors are

really important?

Elena Higueras-Castillo , Alberto Guillén , Luis-Javier Herrera & Francisco

Liébana-Cabanillas

To cite this article: Elena Higueras-Castillo , Alberto Guillén , Luis-Javier Herrera & Francisco
Liébana-Cabanillas (2020): Adoption of electric vehicles: Which factors are really important?,
International Journal of Sustainable Transportation, DOI: 10.1080/15568318.2020.1818330

To link to this article: https://doi.org/10.1080/15568318.2020.1818330

Published online: 14 Sep 2020.

Submit your article to this journal

Article views: 199

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=ujst20
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION
https://doi.org/10.1080/15568318.2020.1818330

Adoption of electric vehicles: Which factors are really important?

Elena Higueras-Castilloa, Alberto Guill
enb, Luis-Javier Herrerab, and Francisco Li
ebana-Cabanillasa
a
Department Marketing and Market Research, Faculty of Business and Economics, University of Granada, Granada, Spain; bDepartment of
Architecture and Computer Technology, Higher Technical School of Computer and Telecommunications Engineering, University of Granada,
Granada, Spain

ABSTRACT ARTICLE HISTORY

Electric vehicles stand out for their ability to reduce pollution. However, electric vehicle sales have Received 24 November 2019
not been brisk despite their positive effects. The objective of this study is twofold: first, to deter- Revised 23 August 2020
mine the variables that predict the purchase of an electric vehicle from the implementation of an Accepted 23 August 2020
algorithm based on computational intelligence; second, to contrast these results with two panels
KEYWORDS
of experts in consumer behavior and the automobile sector. An empirical study was carried out Computational intelligence;
with 404 potential consumers in Spain with regard to their beliefs, attitudes and purchase inten- consumer behavior; electric
tion. The results show that range, incentives and reliability are the most reliable predictors of pur- vehicles adoption;
chase intention. Likewise, the experts posit that the selection of these three variables would be expert interviews
sufficient to know the purchase intention of potential buyers of electric vehicles. In addition, it
provides useful information for policy makers and private companies for decision making in the
electric vehicle marketing process.

1. Introduction noted in Figure 2, according to the Spanish Association of

Car and Lorry Manufacturers (ANFAC) EV registrations
The globally widespread use of fossil fuels causes a large
have been on the rise for the past eleven months. Among
number of environmental problems. The growing concern
the different available models, the best-selling electric
about environmental problems has led to the development of
vehicles were hybrids models with an internal combustion
environmentally friendly products. Specifically, this research
engine and an eletric motor. On the other hand, those pure
focuses on the transport sector. According to the latest Emo
electric EVs running solely on the energy provided by the
the European Environment Agency (EEA), emissions from
electric motor were among the lowest selling electric
road transport have increased by almost 30% compared to
1990. To be more specific, road transport accounts for more vehicles. This can be explained by factors such as the limited
than 70% of greenhouse gases emissions, with road vehicles driving range of EVs and the low availability of charging
being the culprit for increased pollution compared to the rest stations (Higueras-Castillo et al., 2019), turning most con-
of transportation methods (see Figure 1) (EEA, 2019). sumers to hybrid models which can rely on the internal
In this context, electric vehicles (EVs) have the ability to combustion engine to extend the driving range when bat-
reduce CO2 emissions as electricity can be produced from teries run out during a trip. In addition, it is worth noting
renewable energy sources (Mersky et al., 2016). In this sense, that the circumstances surrounding COVID-19 have also
EVs have become an alternative in the transport sector. affected the EV industry with a steep decline in the sales fig-
However, despite their positive effects, sales have been low. ures for March and April 2020.
Following She et al. (2017), the present study considers EV Wang and Liu (2015) revealed that 44% of respondents
as a vehicle which uses energy extracted from the electrical are willing to purchase an EV to replace their current
grid for part or all of its propulsion. Therefore, EV also internal combustion vehicle (ICV). However, recent studies
includes both battery electric vehicles and plug-in hybrid also indicate that less than 20% of respondents are willing to
electric vehicles. replace their ICV with an EV (Wang et al., 2018). As a
In recent years, demand for EV has grown strongly. result, new studies are needed to assess consumer behavior
However, they only represent a small part of the total of the and find new market growth formulas.
new car market in many countries. According to the
European Automobile Manufacturers Association, EV regis-
2. Aim of the research
trations increased by about 50% in 2018 over the previous
year. However, in 2018, EV represented only 2% of new Over the past few years, research has grown in the area of
vehicles registered in the European Union (ACEA, 2019). As EV adoption and purchase intention (Rezvani et al., 2015)

CONTACT Elena Higueras-Castillo ehigueras@ugr.es Department Marketing and Market Research, Faculty of Business and Economics, University of
Granada, Granada, Spain.
ß 2020 Taylor & Francis Group, LLC
2 E. HIGUERAS-CASTILLO ET AL.

Figure 1. GHG emissions derived from road transport in the context of the EU (2017). Source: EEA (2019).

Figure 2. EV registrations in Spain. Source: ANFAC (2020).

with numerous studies analyzing the main motivations and/ The objective of this work is twofold. Firstly, to determine
or barriers to purchase. This research brings together a the variables that predict the purchase of an EV from the imple-
number of significant factors from the literature review and mentation of an algorithm based on computational intelligence.
provides insight into what are the really important factors in In order to achieve it, an empirical study was conducted with
EV purchases. Therefore, governments and private compa- 404 potential consumers in Spain about their beliefs, attitudes
nies should focus on these factors first. This article contrib- and intention to purchase EV. Target group is made up of
utes to the field of research by bringing together the Spaniards over 18 years old with a driving license. Respondent
analytical experience in generating an algorithm, as well as are individuals who are in a position to purchase a vehicle and
the field work experience of the selected experts. It is worth who are aware of the existence of the EV. The data collected
noting that most studies are contextualized in other coun- were used to model consumer behavior using a model based on
tries. In this sense, Spain lacks solid empirical research on computational intelligence to determine which variables deter-
the problem under study. Therefore, the present study mine the purchase intention of an EV from the literature
presents an added value to the area of knowledge. In short, review. In order to achieve second objective. Secondly, to con-
for this research it is important to provide a framework for trast these results with two groups of experts related to the auto-
the adoption of EV in order to help companies understand mobile sector itself and also to consumer behavior and market
consumer behavior and, as a consequence, increase their research. Novelty of research lies in the methodology applied in
market share. this field of knowledge.
 INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 3

The work is structured as follows. First, a review of the

literature is conducted to identify the main variables.
Second, the research design is discussed and the data collec-
tion process is shown along with the questionnaire and the
data analysis. Third, results are assessed and the conclusions
and implications are presented. Lastly, the limitations and
avenues for future research are pointed out.

3. Literature review
The introduction of EVs in the market is an important Figure 3. Factors to be analyzed.
innovation and they are involved in the dissemination pro-
cess of innovations. This process posits that the adoption of 2. Context factors: government incentives (Higueras-Castillo
innovations tends to follow a time series model starting with et al., 2019; Wang et al., 2017), fuel price (Chu et al., 2019)
a small number of innovators and early adopters who pur- and electricity price (Dijk et al., 2013; Javid & Nejat, 2017;
chase the product relatively early, followed by most consum- Soltani-Sobh et al., 2016) and charging infrastructures
ers and ending when lagging consumers decide to adopt the (Coffman et al., 2017; Egn
er & Trosvik, 2018).
innovation (Rogers, 2010). In Spain, EVs are relatively new 3. Consumer characteristics: socio-economic (Chen et al.,
and the population still has limited knowledge about the 2020; Mukherjee & Ryan, 2020) lifestyle (Axsen et al., 2018),
product (Higueras-Castillo et al., 2019). In general, new social norms (Barth et al., 2016; Caperello et al., 2013;
technologies underpin another set of added barriers as a Jansson et al., 2017) and environmental beliefs (Krupa et al.,
result of comparing these innovations with market-dominant 2014; Schuitema et al., 2013).
designs in criteria such as price and performance (Adner,
2002). Thus, early adopters are the main target group of the Consumer characteristics are often less important than
present research study since they are usually willing to pay a vehicle price and performance attributes (Lane & Potter,
higher price or face below-average performance for the latest 2007). In addition, early adopters are more concerned about
technology (Lee et al., 2019). Widespread adoption of an technical than financial problems (Lee et al., 2019). The pre-
innovation usually requires a long period of time, even sent study focuses on technological and contextual factors.
when it has obvious advantages (Rogers, 2010). The slow Previous studies show that these variables influence the
development of EV adoption may be related to the phenom- adoption of electric vehicles, for example, She et al. (2017)
enon of slow dissemination of environmentally friendly indicate that the main barriers are financial concerns (i.e.,
innovations. In this regard, despite the significant benefits, price) and technological factors such as safety, reliability,
there are obstacles to widespread adoption. range, charging time, acceleration and the lack of charging
The purpose of this research is to analyze consumers’ infrastructure. In this research, ten factors have been ana-
intention to adopt EV. Intention is defined as an indication lyzed from those indicated in the literature. These refer to
“of how hard people are willing to try, of how much of an technological and contextual factors. In a previous analysis
effort they are planning to exert, in order to perform the it was concluded that no sociodemographic characteristics
behavior” (Ajzen, 1991). Specifically, it examines a combin- were determinant and so they were excluded from the ana-
ation of technological and contextual factors that affect con- lysis. It allowed us to focus on the rest of the factors with
sumers’ adoption intentions. In this sense, numerous greater precision. However, this is a closed set of factors and
researchers have analyzed the drivers and barriers to EV it is possible to insert additional ones. This has been consid-
adoption (e.g. Hjorthol, 2013; Rezvani et al., 2015). Most ered in the limitations of the study Specifically for this
authors focus on technological factors and consumer charac- research, the factors shown in Figure 3 have been selected.
teristics that determine the purchase of an electric vehicle
(Carley et al., 2013). Consumer perceptions of the intention
to adopt EV have also been studied (Bunce et al., 2014; She 3.1. Technological factors
et al., 2017). These factors can be classified into three broad First, consumers are sensitive to limited driving range
groups (Bjerkan et al., 2016; Sierzchula et al., 2014): (Egbue et al., 2017). There are models on the market with
different levels of range. Currently, EVs already have ranges
1. Technological factors: driving range (Haustein & of up to approximately 400 km, although it is not the norm.
Jensen, 2018; Vassileva & Campillo, 2017), charging Moreover, those models will be the most expensive ones. In
time (Abotalebi et al., 2019), noise (Rocha et al., 2016), contrast, an ICV can cover an average distance of 800 km
acceleration (Helveston et al., 2015; Skippon, 2014), with a petrol engine. Different authors have identified this
CO2 emissions (Jensen et al., 2013; Krupa et al., 2014; difference as one of the main barriers to adoption (Kumar
Li et al., 2019; Peters & D€ utschke, 2014; Zhang & Tay, & Thakur, 2020; Lim et al., 2015). Generally, improving
2017), functionality, reliability and safety (Egbue et al., range through adequate charging infrastructure leads to
2017; Schmalfuß et al., 2017) and image (Burgess greater adoption (Lim et al., 2015). On the other hand,
et al., 2013). Franke and Krems (2013) consider that range is a barrier to
4 E. HIGUERAS-CASTILLO ET AL.

adoption, however, the experience of driving EV produces et al. (2014) demonstrated in approximately 30 countries
adaptation and that would reduce the practical limitations of that the price of EVs has a negative correlation with market
low driving range. Therefore, this factor is a major concern share. Different authors point out that reducing the price
for users and it negatively influences adoption (Jensen et al., could increase the willingness to buy an EV (Junquera et al.,
2014; She et al., 2017). 2016) and its competitiveness (Feng & Figliozzi, 2013).
Second, charging time is another factor influencing EV Therefore, high cost becomes one of the main concerns of
adoption (Zhu, 2016). Most users preferred faster electric consumers (Rezvani et al., 2015). In conclusion, the high
vehicle supply equipment (Moon et al., 2018). While ICVs purchase price is a major disadvantage (Heyvaert
can be refueled in approximately four minutes, EVs require et al., 2015).
at least 30 minutes at a fast charging station and up to On the other hand, EVs have advantages in terms of
8 hours to charge at a lower power (Glerum et al., 2014; recharge price and maintenance costs, constituting one of
Kumar & Thakur, 2020). According to Sellmair and Schelo the main motivations associated with the purchase (Ozaki &
(2019), it is very important to reduce waiting times. Sevastyanova, 2011). The work of Zhang et al. (2013) identi-
However, recent studies indicate that users are willing to fies financial benefits as a driver of EV acceptance. Taking
adapt to the charging process of an EV (Schmalfuß et al., into account fuel and energy prices, the cost of charging the
2017). In conclusion, reducing charging time and increasing batteries of EVs is less than the cost of refueling ICVs
the range of EVs should significantly improve the intention (Carley et al., 2013). According to Chu et al. (2019), lower
to purchase EV (Junquera et al., 2016). recharging cost is an important motivation for buying an
With regard to the performance (range and charging EV. Thus, in comparison, the low price of electricity
time) of an EV, consumers in general are not satisfied increases the adoption of EV (Soltani-Sobh et al., 2017). It
(Chen et al., 2020; Kester et al., 2018). However, EVs have was found that there were significant savings in societal
the potential to perform better than ICVs, so this advantage costs and total cost of ownership when compared to diesel
can compensate for more unfavorable factors such as limited vehicles (Bor
en, 2019). On the other hand, electric motors
range, long charging time and high price (Skippon, 2014). It are less complex propulsion systems than ICV, which means
is recognized that performance attributes have a greater they are less costly to maintain (Taefi et al., 2016).
effect on consumer acceptance than financial or environ- Therefore, the perceived benefit has a positive effect on pur-
mental awareness factors (Zhang et al., 2013). On the other chase intention (He et al., 2018; Kim et al., 2018). However,
hand, other studies indicate that environmental benefits and despite these savings, consumers may still refuse to buy the
incentives are more relevant than performance characteris- product. It is the so-called energy-efficiency paradox or
tics (Peters & D€ utschke, 2014). energy-efficiency gap (Gillingham & Palmer, 2014).
Two attributes that differ clearly from ICVs can be high- Different studies posit that respondents do not value this
lighted: fast acceleration and low engine noise. Burgess et al. benefit very much or are not aware of this potential cost
(2013) reported that respondents were impressed by the saving (Carley et al., 2013). In this sense, consumers are
acceleration and low noise of EVs. Similarly, research by more influenced by the high purchase price and do not take
Helveston et al. (2015) respondents who were subjected to a into account the total cost of ownership of EVs (Sierzchula
choice experiment were inclined toward fast vehicle acceler- et al., 2014).
ation. In this regard, EVs produce faster acceleration at low With this in mind, in recent years, governments have
speeds and quietness compared to an ICV (Skippon, 2014). implemented different policy measures to increase interest
Also, Rocha et al. (2016) found that low noise emission is a in EV (Lieven, 2015; Sierzchula et al., 2014). Literature
key feature in achieving greater market share. shows incentives implemented in the USA (Jin et al., 2014),
On the other hand, safety and reliability are two of the Europe (Gass et al., 2014; Kley et al., 2012) and worldwide
main concerns of consumers (Thananusak et al., 2017; (Leurent & Windisch, 2011). In this regard, Sierzchula et al.
Zhang et al., 2013). In the study by She et al. (2017), safety (2014) analyzed the correlation between incentives and mar-
and reliability scored highest. In this case, it is implied that ket shares of EV in 30 countries. The results showed that
respondents do not trust EVs and are concerned about the incentives are a predictor of adoption. L
evay et al. (2017)
safety of using this technology. These results were mainly assessed the relationship between the total cost of ownership
due to the fact that the battery has caught fire in some acci- (TCO) of the EV and the impact of tax incentives on it rela-
dents. Nevertheless, reliability is highlighted as one of the tive to ICVs in eight European countries. The results show
main motivators for purchasing an EV (Higueras-Castillo that in Norway the incentives lead to a lower TCO and in
et al., 2019). In this sense, Ingeborgrud and Ryghaug (2019) the Netherlands, France and the United Kingdom the TCO
highlight that EV owners consider their vehicle a safe, silent of the EVs is close to the TCO of the ICVs. However, in the
and exciting technology. other countries, the TCO is still higher than in ICV.
Incentive measures such as reduced prices and road and
registration taxes affect the purchase and the use of EVs.
3.2. Context factors
Additional benefits include the possibility to drive on bus
The high purchase price is one of the strongest barriers to lanes and free parking spots. Figure 4 shows the different
the purchase of EV. Consumers are not willing to pay a types of incentives for EVs as well as the restrictive regula-
large premium for an EV (Larson et al., 2014). Sierzchula tions in place for traditional vehicles. Also, according to the
 INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 5

Figure 4. Incentives affecting the purchase of hybrid and pure electric vehicles. Source: adapted from Deloitte (2017).

study conducted by the European Alternative Fuels charging stations is a predictor of EV adoption (Sierzchula
Observatory (EAFO, 2018), France, Norway and Germany et al., 2014) and access to them is a key determinant of
are the European countries that stimulate the EV sector the adoption (Mersky et al., 2016). The key difference between
most, with Spain ranking as the 12th country with the electric and conventional vehicles is the charging infrastruc-
best incentives. ture, among others (Gnann et al., 2018). According to Wang
In this regard, the scientific literature highlights the et al. (2019), charger density correlates positively with a
effectiveness of each incentive according to the different tar- country’s EV market share.
get populations (Bjerkan et al., 2016; Jenn et al., 2018;
Mersky et al., 2016). However, despite the differences, the
literature shows that incentives positively influence the 4. Materials and methods
acceptance of EV (Kim et al., 2018; Langbroek et al., 2016). 4.1. Data collection
Krupa et al. (2014) suggest that increasing consumer aware-
ness of the existence of incentives could have a greater Data collection from the first sub-study was conducted
impact on perceived benefits. Also, as a consequence of low through an online survey with a structured and pre-coded
buyer motivation, Turcksin et al. (2013) confirm that questionnaire developed from April to July 2018. Data were
increasing the adoption of alternative mobility vehicles obtained through a non-probability sampling method
requires a stable and rigorous incentive policy. In this way, defined by quotas according to the structure of the popula-
incentives reduce the cost of purchasing an EV to a compar- tion. A research company specialized in sampling services
able ICV (Bjerkan et al., 2016). However, some authors was designated to select participants randomly. The sample
show in their research a negligible (Zhang et al., 2013) or consists of 404 potential consumers of EV. The sampling
weak (Harrysson et al., 2015) correlation between incentives error reached 4.874% with a 95% confidence interval. In
and consumer willingness to buy EV. According to Wang order to ensure that participants were part of the target
et al. (2019) tax incentives are no longer the cause of large population, they were asked a series of control questions
differences between countries. (see appendix B). Respondents are Spaniards over 18 years
According to Egn
er and Trosvik (2018) local incentive old with a driving license. Table 1 shows the classification
measures such as investment in infrastructure also has a sig- characteristics of respondents. 51% of respondents were
nificant impact on the adoption rate. The charging infra- male, with 60.2% of respondents below the age of 46 and
structure is essential, so its unavailability makes it an 44.1% educated to university degree level. 58.7% were in
obstacle to its adoption (Tran et al., 2012). According to She full-time employment and 56.9% had a monthly income in
et al. (2017), the lack of charging infrastructure is the great- the range of 1.100 to 2.700 euros. 76.9% of respondents had
est impediment to adoption. In their work, Jensen et al. more than 5 years of driving experience, with 61.6% driving
(2013) showed that charging stations in public places are over 12,500 km per year.
important for the purchase of EVs. While Krupa et al. In order to reduce the abandonment rate once the ques-
(2014), pointed out that having facilities at home to charge tionnaire was initiated, the following measures were imple-
the battery overnight is important for consumers, it is also mented: 1) the purpose of the research was clearly explained
important for the safety of the vehicle and the charging and declared nonprofit; 2) the researcher and the university
cable (Caperello & Kurani, 2012). Therefore, the number of were identified; 3) confidentiality and anonymity were
6 E. HIGUERAS-CASTILLO ET AL.

Table 1. Characteristics of the simple.

Gender Number Percentage Monthly income Number Percentage
Female 206 51.0 No income 37 9.2
Male 198 49.0 Less than e 1,100 73 18.1
Age bracket From e 1,100 to e 1,800 135 33.4
18–25 55 13.6 From e 1,800 to e 2,700 95 23.5
26–35 111 27.5 More than e 2,700 40 9.9
36–45 77 19.1 Don’t know/No answer 24 5.9
46–55 74 18.3 Experience as a driver
56–65 56 13.9 0–1 year 27 6.7
More than 65 31 7.7 1–3 years 34 8.4
Education 3–5 years 32 7.9
Compulsory or less 21 5.2 5–8 years 207 51.2
Professional training 114 28.2 More than 8 years 104 25.7
Bachelor’s degree 178 44.1 Annual distance driven (Km)
Postgraduate degrees 91 22.5 Up to 2,500 78 19.3
Employment status Up to 7,500 77 19.1
Unemployed 37 9.2 Up to 12,500 75 18.6
Student 43 10.6 Up to 15,000 57 14.1
Regular employed 237 58.7 Up to 20,000 64 15.8
Self-employed 39 9.7 Up to 32,500 31 7.7
Retired 48 11.9 More than 32,500 21 5.2

guaranteed; 4) it was indicated that there were no valid or amount of information that a group of variables has about
invalid answers, only the opinion of the respondents was the target variable (or variables). Its definition in the con-
interesting; 5) all questions were closed; 6) the visual design tinuous case is:
was clear and simple. Also, during the initial stage of this ð
lX:Y ðx, yÞ log ðlX:Y ðx, yÞÞ
research, the robustness of the measurement scales used was I ðX, Y Þ ¼ dxdy
lX ðxÞlY ðyÞ
tested. In addition, this study made sure that respondents
understood and approached the survey correctly. where lX:Y ðx, yÞ is the joint probability density function
(PDF) of X and Y, and lX ðxÞ is the marginal density func-
tion of the set of variables X. Its advantage in comparison
4.2. Questionnaire with other correlation criteria is that it is able to identify
This research adopted the use of a seven-point Likert scale non-linear relations among the variables involved.
(except for demographic variables). The scale ranges from 1 Although the mathematical definition is clear, computing
(strongly disagree) to 7 (strongly agree), according to the its concrete value is not an easy task and practitioners must
respondent’s level of agreement with each
ıtem. The socio- use estimations of the real MI value. There are several MI
demographic information part was completed by respond- estimators but the one presented by Kraskov (Kraskov et al.,
ents choosing the corresponding options. 2004), which is based in the k-nearest nieghbours, seems
To be more precise, the range, charging, acceleration, more robust than other alternatives based on histograms or
noise, safety and reliability scale was adapted from the work other more complex methods.
of Schmalfuß et al. (2017). Price was assessed on the basis Thanks to this criterion, it is possible to choose the
of two scales of work by He and Zhan (2018) and Petrick smallest possible set of factors that contain all the necessary
information on the target variable. In other words, the sub-
(2002). Financial benefit and incentives was adjusted from
set of variables with highest mutual information will allow
Wang, Zhao et al. (2017). Infrastructure was adapted by He
us to design a model which is able to predict the out-
and Zhan (2018) and Jansson (2011). Lastly, intention of
put accurately.
adoption was adapted from the scales of Moons and De
Once this is achieved, experts can use the variables
Pelsmacker (2012) and Barbarossa et al. (2017). Appendix A
selected for decision-making, avoiding factors that are not
shows the measurement scales used in this research.
influential in any case, or, and not of minor importance,
that are redundant with respect to others already considered.
4.3. Performing variable selection The advantages of this type of factor selection are multiple:
faster, and sometimes even more effective, subsequent com-
The selection of variables is a complex issue but one of great putational modeling, improved interpretability, easier com-
importance in an uncountable number of problems across munication between experts and cheaper and faster data
all disciplines (Guyon & Elisseeff, 2003). The basis of this gathering, among others.
type of technique is the arrangement of a measure of rele- Among the different alternatives found in the literature,
vance/correlation between variables. In the case of the pre- an algorithm adapted from Koller and Sahami (1996) was
sent study, such a measure will be mutual information, a used. This algorithm has been previously applied to diverse
non-linear correlation measure from Shannon’s Information problems where the most irrelevant and redundant factors
Theory, which was first presented in (Shannon, 1948) and of a problem are identified iteratively and in order, leaving
further discussed in the book “Elements of information the- in last place the really significant factors of the same one
ory” (Cover & Thomas, 1991). This measure estimates the (Herrera et al., 2006; Lafuente et al., 2015). The output of
 INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 7

Figure 5. Selected variables.

this algorithm is a ranking of the variables, where in first just using those may not achieve the desired accuracy to
place the most important variable is given, then the second model the output.
most important variable taking into account the first one, To make sure that the chosen variables are able to pro-
and so on. vide a proper data driven approach, a well-known modeling
This algorithm is based on the Markov blanket concept. technique has been used: Least Square Support Vector
Given a set of input variables X and an output variable Y, a Machines (LSSVM) (Suykens et al., 2002). The output vari-
set of variables Mi in X is said to be a Markov blanket for a able to be approximated was the average of the values of the
variable xi in X with respect to Y, if IðfMi [ xi g, Y ¼ respondents’ responses, being this between 1 and 7.
I ðMi , Y Þ, that is, if Mi has itself all the information that xi Before applying the method of variable selection and data
has about Y. A Markov blanket is thus, a group of variables modeling, preprocessing was performed. First, there were
subsuming the information content of a certain variable. questions whose high value implied a positive perception of
The algorithm operates in a backwards way, starting from the EV while other questions represented the opposite. In
the complete set of variables X, and iteratively discarding this sense, the results of the variables with the negative score
those which are detected to have a Markov Blanket in the were inverted so that, for each question, all the high values
remaining set of variables (named from now on XG). corresponded to a positive attitude toward the EV. In add-
The algorithm states the following steps: ition, since there were redundant questions that focused on
the same aspect, it was decided to unify by averaging the
groups of redundant questions. Thanks to this step, this
1. Calculate the MI between every pair of input variables I
study went from 42 initial variables (one per question) to
(xi, xj)
the 10 variables considered. Similarly, in order to be able to
2. Starting from the complete set of input variables XG ¼
deal with a single output, the three questions related to pur-
X, iterate:
chase intention were also grouped together, moving from
a. For each variable xi, let the candidate Markov
three possible values to one. Thus, given the data set of 404
blanket Mi be the set of p variables in XG for which
surveys, the 10 variables considered and the purchase esti-
I (xi, xj) is highest.
mate, the characteristics selection algorithm was applied,
b. Compute for each xi
obtaining the order of characteristics given in Figure 5. The
Lossi ¼ I ðfMi [ xi g, Y Þ  I ðMi , Y Þ application of different LSSVM models to the first factor,
first two factors, first three factors, etc., led to the perform-
c.Choose the xi for which Lossi is lowest and eliminate ance (expressed by the R2 metric, which indicates the pro-
xi from XG. portion of the output’s variance that can be explained by the
3. Continue with step 2 until no variables remain. chosen features) observed in Figure 6. These results were
further corroborated under different random subdivisions of
The p parameter of the algorithm (in step 2.a of the algo- training and tests over the global data set.
rithm) will take the value p ¼ 1, as recommended in previ-
ous works (Herrera et al., 2006; Koller & Sahami, 1996).
4.4. Results validation
Once the relevance ranking has been established using
the above algorithm, it is necessary to identify the number First, the individual reliability of the items is examined on
of variables required to reach the target performance. It is the basis of the simple correlations of the indicators with
possible to identify one or two very significant variables, but their respective variables. The recommended value should
8 E. HIGUERAS-CASTILLO ET AL.

Figure 6. Performance of the ranking obtained.

Table 2. Reliability of constructs.

Construct Mean SD a Construct Mean SD a
Range Charging
RAN1 4.03 1.624 0.757 CHA1 4.03 1.852 0.859
RAN2 4.30 1.722 CHA2 4.31 1.772
Perceived benefit Infrastructure
PB1 4.70 1.448 0.733 INF1 5.39 1.441 0.740
PB2 4.90 1.447 INF2 5.00 1.426
Safety Reliability
SAF1 5.38 1.422 0.821 REL1 5.36 1.304 0.921
SAF2 5.32 1.348 REL2 5.35 1.324
SAF3 5.49 1.332 REL3 5.20 1.440
Price Incentives
PRI1 5.23 1.490 0.896 INC1 5.16 1.584 0.725
PRI2 5.25 1.452 INC2 4.74 1.512
PRI3 5.47 1.443 INC3 4.92 1.675
PRI4 5.03 1.475 INC4 4.34 1.751
Acceleration Intention to adopt
ACC1 5.02 1.378 0.860 IA1 5.15 1.611 0.907
ACC2 4.86 1.301 IA2 5.07 1.618
ACC3 5.00 1.402 IA3 4.66 1.738
Low Noise
LN1 4.96 1.850 0.873
LN2 5.06 1.878
LN3 4.77 1.805
LN4 4.60 1.859
LN5 5.06 1.785

reach 0.7. This assumes that the variance shared between experts have a minimum of ten years of experience in each
the construct and its indicators is greater than the variance of their sectors.
of the error term (Barclay et al., 1995). This study uses the The validation procedure was divided into four different
Cronbach Alpha coefficient (Cronbach, 1951). Also, it shows stages across the first half of 2019 including in-depth per-
mean and standar desviation (SD) of each item, as you can sonal interviews, assessments of the different research tools
see in Table 2. and the obtained results along with consideratons of the
In order to validate the results achieved with the pro- resulting feedback. Firstly, the study conducted an in-depth
posed method, a group of experts grouped into two different interview of ten experts in the EV sector in order to detail
categories was contacted: researchers related to consumer the type of experiment and select a set of variables that they
behavior and market research with an international research considered fit to explore intention to use. The group of
curriculum and professionals from the automobile sector experts were further divided into two sub-groups: a) profes-
with experience in sales of electric vehicles. All the selected sionals from the marketing and market research sector and,
 INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 9

Table 3. Expert evaluations.

Algorithm
MARKETING AND MARKET RESEARCH OPTION 1 OPTION 2 OPTION 3 OPTION 4
SELECTED VARIABLES Range Range, incentives Range, incentives, safety Range, incentives, safety, acceleration
EXPERT 1 5 5 6 6
EXPERT 2 4 5 6 6
EXPERT 3 5 6 7 7
EXPERT 4 3 6 6 7
EXPERT 5 4 6 7 7
MEAN 4.2 5.6 6.4 6.6
PROFESSIONALS OF THE AUTOMOBILE SECTOR OPTION 1 OPTION 2 OPTION 3 OPTION 4
EXPERT 1 5 5 6 6
EXPERT 2 5 6 6 6
EXPERT 3 4 6 6 6
EXPERT 4 5 7 7 7
EXPERT 5 4 7 7 6
MEAN 4.6 6.2 6.4 6.2
TOTAL 4.4 5.9 6.4 6.4

b) directors from five different car dealers selling EVs. The best option valued by the group of researchers on
Interviews lasted 2 hours each and approached a semi-struc- consumer behavior is option 4, which includes a greater
tured questionnaire listing the purposes, methodology and number of selection variables to determine the general
the set of variables of the present research study in detail. In intention to buy EV (6.6). Although it is true that option 3
this sense, interviews helped define the final set of variables obtains a high average (6.4) by sacrificing a variable in the
that the research instruments later analyzed. choice of variables that facilitate the decision making process
Secondly, after results were obtained from the feature to buy an EV, it is also true that option 3 obtains a high
selection algorithm (see Figures 5 and 6), four options were average (6.4) by sacrificing a variable in the choice of varia-
established to be compared by the groups of experts. The bles that facilitate the decision making process to buy an
four options correspond to different restriction levels. In EV. On the other hand, in the case of professionals related
decision support systems, it is always desired a small subset to the automotive sector, they value option 3 (6.4) to a
of variables to make things easier to the human operator. greater extent instead of option 4 (6.2).
Therefore, after obtaining the ranking of the variables, 4 After a joint analysis of the results with the group of
subsets (options) were given to the experts: option 1, consid- experts in a second interview, the main reasons for the score
ering the top variable in the ranking, option 2, the two top for each of the options proposed were as follows:
ranked variables, and so on up to a maximum of 4 variables. With respect to option 1: It only includes one variable such
The reason to stop in 4 variables is mainly because the R2 as driving range (4.4) which, although relevant, is not the only
achieved did not significantly increased when adding more one that determines the purchase intention of an EV, as it
variables (even being as high as with the whole set of 10 ignores the rest of the variables. The researchers argue that
variables). Furthermore, considering more than 4 variables this option is not valid because it uses a particularly restricted
might be confusing for the human taking the decision. criterion. It is interesting to note that driving range was the
This procedure provided the groups of experts with a most relevant variable found by the feature selection algo-
predefined set of variables to be rated using a 7-point Likert rithm, so it was thus the first variable in the rest of options,
scale. Lastly, once this study categorized the obtained results following the obtained ranking. But itself did not reach
and statistical conclusions were drawn from the proposed enough performance (in terms of R2), needing more variables
methodology, the experts assessed and ranked the variables to reach the maximum R2 value (see Figures 5 and 6).
in order of importance for each approached technique. In Regarding option 2: This option includes range and incen-
addition, the level of congruence was communicated tives, and is valued more positively than the previous option
through personal interviews with the experts in order to jus- (5.9). This option is supported by both groups of experts, mak-
tify their answers and analyze the results obtained from the ing a technological element compatible with another context-
different goups (Li
ebana-Cabanillas et al., 2016). Table 3 ual element that arouses high interest among potential buyers.
details the obtained results from the evaluations conducted On this occasion, the valuation of professionals in the automo-
by the different groups of experts. bile sector surpasses the valuation of professionals in the study
In general terms, as seen in Table 2, the most valued of consumer behavior and market research (6.2 versus 5.6).
options are 3 and 4, which include the variables range, Regarding option 3: This option includes range, incen-
incentives and reliability, in the first case, and, in the tives and reliability and is assessed identically by both
second, acceleration together with the previous options groups of experts. The experts understand that this option
(average valuation ¼ 6.4). In this sense, with the same aver- achieves a balance in the valuations of the selected variables
age valuation, option 3 uses one less selection variable, so in (6.4 in both groups of experts).
terms of performance this option would be better than Finally, with respect to option 4: This last option includes
option 4. range, incentives, reliability and acceleration. In this case,
10 E. HIGUERAS-CASTILLO ET AL.

professionals studying consumer behavior and market acceleration. It is possible that experts take into account this
research value the inclusion of the last variable (acceleration) variable because it is very easy to translate into an ad, slogan
more positively than professionals in the automobile sector and advertising campaign in general. However, after conduct-
(6.6 versus 6.2). ing the objective analysis of the data plus the opinion of auto-
motive experts, it can be concluded which are the three
elements that should guide the industry in improving purchase
5. Concluding discussions
intent.
5.1. Discussion of results
For this research it is important to provide a theoretical frame 5.2. Managerial implications
of reference for the adoption of EV (battery electric vehicles
Conclusions derived from the present research study can be
and plug-in hybrid electric vehicles). The purpose of this study
extensively applied to the bussines field. In this sense, since
is twofold. Firstly, to analyze intention to adopt EV.
EV adoption is still in the early stages of adotpion with
Specifically, to determine the factors predicting purchasing
potential customers considering a significant number of var-
behavior in a reliable way. This is done by implementing a
iables while contemplating an actual purchase (Rogers,
variable selection algorithm based on computational intelli-
2010), identifying the most important factors which drive
gence. Secondly, a set of experts related to the automobile sec-
EV sales should be the main strategy for companies in this
tor and specialists in consumer behavior and market research
sector. Said factors have been already validated by both
contrast these results under their own experience and opinion.
The literature shows numerous elements that must be potential customers and proffesionals, increasing the exter-
considered to understand a consumer’s purchase intention nal validity of this apporach.
at the time of purchasing an EV. From an exhaustive review, As a result of afore mentioned implications, the following
the most representative factors were selected. They can be measures should be implemented by those agents of interest
classified into technological factors: driving range, charging willing to drive EV sales:
time, noise, acceleration, safety and reliability; and context-
ual factors: price, perceived benefit, incentives and infra-  Improve the current EV technology, especially with regard
structure availability. to the driving range which remains as a barrier and a major
Methodoloty applied allows to choose the smallest possible concern for the vast majority of customers and proffesion-
set of factors that contain all the necessary information on the als. In this sense, EV manufacturers should keep working
target variable for decision-making, avoiding factors that are on extending battery life in order to mitigate customers’
not influential in any case, or, and not of minor importance, fear of running out of power during a trip.
that are redundant with respect to others already considered.  Design a proper incentive plan. In this regard, incentives
The results of this research indicate that among all the ele- remain as a key stimulant with regard to EV sales. Since the
ments studied, three factors are the most important to con- price of EVs is higher than that of traditional internal
sider: range, incentives and reliability. This result is observed combusti
on vehicles, incentives should help motivate cus-
in Figures 5 and 6, as the R2 of the fit to predict the adoption tomers to purchase EV. Incentive plans can also be applied
of an EV does not increase when considering more than three at both company and institutional levels. Thus, private com-
variables. Thus, these three variables make it possible to model panies should implement several strategies to encourage an
all the responses given by consumers with the greatest preci- easy access to EVs based on monetary incentives and
sion. The rest of the variables do not provide significant infor- upgrade paths. On the other hand, due to the current global
mation that would allow this approximation to be improved. and European agreements (Paris Agreement), governments
In other words, the yield obtained by these factors is similar to have already different incentives in place affecting the level
the use of a greater number of factors. of adoption of EVs for each country (EAFO, 2018). In this
On the other hand, in order to validate the results sense, Spain should develop solid incentive strategies and
obtained, a series of interviews were carried out with expert plans to retain a constant, reliable source of measures avail-
researchers related to consumer behavior and market able to improve EV adoption, helping the country in the
research with an international research curriculum and pro- challenge to reach the agreements established by the Paris
fessionals from the automobile sector with experience in Agreement (European Commision, 2019).
sales of EV. From the statistical conclusions, it is observed  Reliability is also a key factor that mitigates customer’s
that a balance is reached with the inclusion of the first three fears and uncertainty, especially in the early stages of
variables, which determines that the proposed experts adoption. Since the vast majority of potential customers
understand that the selection of these three variables would are not familiar with EV technology, a signficant level of
be sufficient to know the purchase intention of potential reliability should drive sales of EVs. In this sense, trust
buyers of EVs. These experimental results corroborate the in the reliability of the new technology should improve
theoretical results of the proposed algorithm. over time as customers spread a positive word-of-motuh
It is interesting to see how the automobile industry expert (Mcknight & Chervany, 2001). Both private companies
group valued the set of variables selected by the variable selec- and public institutions can increase the added value
tion method the most. This same selection also gets a very offered by reliability by developing communication strat-
good score by marketing experts who prefer to include egies and campaigns aimed at customer needs.
 INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 11

5.3. Limitations, recommendations and avenues for Ajzen, I. (1991). The theory of planned behavior. Organizational
future research Behavior and Human Decision Processes, 50(2), 179–211. https://doi.
org/10.1016/0749-5978(91)90020-T
This research has several limitations in interpreting the ANFAC. (2020). Matriculaciones de turismos y todotorrenes el
ectricos.
results obtained which offer different lines for future Veh
ıculos alternativos. Agencia Espa~ nola de Fabricantes de
Autom
oviles y Camiones. Retrieved May 13, 2020, from https://
research. In the first place, this study examines a series of anfac.com/vehiculo-alternativo/.
factors that, according to the review of the extant scientific Axsen, J., Cairns, J., Dusyk, N., & Goldberg, S. (2018). What drives the
literature, do affect intention to adopt. However, this group pioneers? Applying lifestyle theory to early electric vehicle buyers in
of factors is limited and could have benefited from the Canada. Energy Research & Social Science, 44, 17–30. https://doi.org/
inclusion of additional factors such as customer’s emotion 10.1016/j.erss.2018.04.015
Barbarossa, C., De Pelsmacker, P., & Moons, I. (2017). Personal values,
or habits. Also, to assess the purchasing behavior with green self-identity and electric car adoption. Ecological Economics,
regard to EVs, intention to adopt is examined. In this sense, 140, 190–200. https://doi.org/10.1016/j.ecolecon.2017.05.015
intention to adopt is considered a proxy for actual behavior Barclay, D., Higgins, C., & Thompson, R. (1995). The Partial Least
and can therefore predict it (Hung et al., 2003). However, Squares (PLS) approach to causal modeling: Personal computer
this does not exempt that it may not be fully representative. adoption and use an illustration. Technology Studies, 2(2), 285–309.
Barth, M., Jugert, P., & Fritsche, I. (2016). Still underdetected–Social
At the same time, the sample is composed of potential con- norms and collective efficacy predict the acceptance of electric
sumers, most of whom have had no previous experience vehicles in Germany. Transportation Research Part F: Traffic
with EVs. Therefore, a future study should be carried out to Psychology and Behaviour, 37, 64–77. https://doi.org/10.1016/j.trf.
investigate the actual behavior of consumers, as well as to 2015.11.011
Bjerkan, K. Y., Nørbech, T. E., & Nordtømme, M. E. (2016). Incentives
analyze the beliefs and attitudes of current consumers.
for promoting battery electric vehicle (BEV) adoption in Norway.
Secondly, the sample is obtained in Spain. Considering the Transportation Research Part D: Transport and Environment, 43,
cultural differences between countries and the level of devel- 169–180. https://doi.org/10.1016/j.trd.2015.12.002
opment of the EV market, there may be problems in gener- Bor
en, S. (2019). Electric buses’ sustainability effects, noise, energy use,
alizing the results. In this sense it would be important to and costs. International Journal of Sustainable Transportation, 10,
verify the results obtained by the present study with a sam- 1–16. https://doi.org/10.1080/15568318.2019.1666324
Bunce, L., Harris, M., & Burgess, M. (2014). Charge up then charge
ple of other countries. On the other hand, data collection out? Drivers’ perceptions and experiences of electric vehicles in the
was conducted through an online survey platform. This UK. Transportation Research Part A: Policy and Practice, 59,
method may incur a sample bias as consumers who do not 278–287. https://doi.org/10.1016/j.tra.2013.12.001
use the Internet are not included in the sample. Therefore, Burgess, M., King, N., Harris, M., & Lewis, E. (2013). Electric vehicle
drivers’ reported interactions with the public: Driving stereotype
future studies should consider the inclusion of offline con-
change? Transportation Research Part F: Traffic Psychology and
sumers. Finally, this research examines all types of EV. The Behaviour, 17, 33–44. https://doi.org/10.1016/j.trf.2012.09.003
results may be different depending on the type of vehicle, so Caperello, N. D., & Kurani, K. S. (2012). Households’ stories of their
future studies may distinguish between them and then com- encounters with a plug-in hybrid electric vehicle. Environment and
pare the results. In addition, future research is invited to Behavior, 44(4), 493–508. https://doi.org/10.1177/0013916511402057
Caperello, N., Kurani, K. S., & TyreeHageman, J. (2013). Do you mind
include a number of other personal factors, including envir-
if i plug-in my car? How etiquette shapes PEV drivers’ vehicle
onmental beliefs highly valued in other works (Krupa et al., charging behavior. Transportation Research Part A: Policy and
2014) or socioeconomic beliefs (Mukherjee & Ryan, 2020) Practice, 54, 155–163. https://doi.org/10.1016/j.tra.2013.07.016
and lifestyle factors (Axsen et al., 2018). Carley, S., Krause, R. M., Lane, B. W., & Graham, J. D. (2013). Intent
to purchase a plug-in electric vehicle: A survey of early impressions
in large US cites. Transportation Research Part D: Transport and
Funding Environment, 18, 39–45. https://doi.org/10.1016/j.trd.2012.09.007
Chen, C. F., de Rubens, G. Z., Noel, L., Kester, J., & Sovacool, B. K.
This work was supported by Spanish Ministry of Economy and (2020). Assessing the socio-demographic, technical, economic and
Competitiveness (MINECO) #1 under Grant TIN2015-71873-R; behavioral factors of Nordic electric vehicle adoption and the influence
European Regional Development Fund (ERDF) #2 under Grant of vehicle-to-grid preferences. Renewable and Sustainable Energy
FPA2017-85197-P; ERDF #3 under Grant B-SEJ-209-UGR18. Reviews, 121, 109692. https://doi.org/10.1016/j.rser.2019.109692
Chu, W., Im, M., Song, M. R., & Park, J. (2019). Psychological and
behavioral factors affecting electric vehicle adoption and satisfaction:
A comparative study of early adopters in China and Korea.
References Transportation Research Part D: Transport and Environment, 76,
1–18. https://doi.org/10.1016/j.trd.2019.09.009
Abotalebi, E., Scott, D. M., & Ferguson, M. R. (2019). Why is electric Coffman, M., Bernstein, P., & Wee, S. (2017). Electric vehicles revis-
vehicle uptake low in Atlantic Canada? A comparison to leading ited: A review of factors that affect adoption. Transport Reviews,
adoption provinces. Journal of Transport Geography, 74, 289–298. 37(1), 79–93. https://doi.org/10.1080/01441647.2016.1217282
https://doi.org/10.1016/j.jtrangeo.2018.12.001 Cover, T. M., & Thomas, J. A. (1991). Elements of information theory.
ACEA. (2019). Economic and market report. EU automotive industry. In Wiley series in telecommunications. Wiley. https://doi.org/10.
Full-year 2018. European Automobile Manufacturesrs Association. 1002/0471200611
Retrieved July 2, 2019, from https://www.acea.be/uploads/statistic_ Cronbach, L. J. (1951). Coefficient alpha and the internal structure of
documents/Economic_and_Market_Report_full-year_2018.pdf. tests. Psychometrika, 16(3), 297–334. https://doi.org/10.1007/
Adner, R. (2002). When are technologies disruptive? A demand-based BF02310555
view of the emergence of competition. Strategic Management Deloitte. (2017). Un modelo de transporte descarbonizado para Espa~ na en 2050.
Journal, 23(8), 667–688. https://doi.org/10.1002/smj.246 Recomendaciones para la transici
on. Retrieved April 10, 2020, from https://
12 E. HIGUERAS-CASTILLO ET AL.

cdn2.hubspot.net/hubfs/1708142/Deloitte/Campaigns/Descarbonizacion/ consumer preferences in the US and China. Transportation Research Part

Descarbonizacion-2017/Descarbonizacion-Transporte-Monitor-Deloitte.pdf. A: Policy and Practice, 73, 96–112. https://doi.org/10.1016/j.tra.2015.01.
Dijk, M., Orsato, R. J., & Kemp, R. (2013). The emergence of an elec- 002
tric mobility trajectory. Energy Policy, 52, 135–145. https://doi.org/ Herrera, L. J., Pomares, H., Rojas, I., Verleysen, M., & Guil
en, A.
10.1016/j.enpol.2012.04.024 (2006). Effective input variable selection for function approximation.
EAFO. (2018). EV Incentives. Electric vehicle incentives across Europe. International Conference on Artificial Neural Networks (pp. 41–50).
Observatorio europeo de Combustibles Alternativos. Disponible en. Springer. https://doi.org/10.1007/11840817_5
https://www.comparethemarket.com/car-insurance/content/electric- Heyvaert, S., Coosemans, T., Mierlo, J. V., & Macharis, C. (2015).
vehicle-incentives/ Electric vehicle attitudes and purchase intention: A Flemish case
EEA. (2019). Greenhouse gas emissions from transport in Europe. study. International Journal of Electric and Hybrid Vehicles, 7(1),
European Environment Agency. Retrieved March 17, 2020, from 83–100. https://doi.org/10.1504/IJEHV.2015.068946
https://www.eea.europa.eu/data-and-maps/indicators/transport-emis- Higueras-Castillo, E., Li
ebana-Cabanillas, F. J., Mu~ noz-Leiva, F., &
sions-of-greenhouse-gases/transport-emissions-of-greenhouse-gases- Garc
ıa-Maroto, I. (2019). Evaluating consumer attitudes toward elec-
12#tab-related-briefings. tromobility and the moderating effect of perceived consumer effect-
Egbue, O., Long, S., & Samaranayake, V. A. (2017). Mass deployment iveness. Journal of Retailing and Consumer Services, 51, 387–398.
of sustainable transportation: Evaluation of factors that influence https://doi.org/10.1016/j.jretconser.2019.07.006
electric vehicle adoption. Clean Technologies and Environmental Hjorthol, R. (2013). Attitudes, ownership and use of electric vehicles–a
Policy, 19(7), 1927–1939. https://doi.org/10.1007/s10098-017-1375-4 review of literature. National Academy of Sciences.
Egn
er, F., & Trosvik, L. (2018). Electric vehicle adoption in Sweden Hung, S. Y., Ku, C. Y., & Chang, C. M. (2003). Critical factors of WAP serv-
and the impact of local policy instruments. Energy Policy, 121, ices adoption: An empirical study. Electronic Commerce Research and
584–596. https://doi.org/10.1016/j.enpol.2018.06.040 Applications, 2(1), 42–60. https://doi.org/10.1016/S1567-4223(03)00008-5
European Commision. (2019). Historial de la UE en materia de acci
on Ingeborgrud, L., & Ryghaug, M. (2019). The role of practical, cognitive
por el clima. El Pacto Verde Europeo. Retrieved April 18, 2020, from and symbolic factors in the successful implementation of battery
https://ec.europa.eu/commission/presscorner/detail/es/fs_19_6720. electric vehicles in Norway. Transportation Research Part A: Policy
Feng, W., & Figliozzi, M. (2013). An economic and technological ana- and Practice, 130, 507–516. https://doi.org/10.1016/j.tra.2019.09.045
lysis of the key factors affecting the competitiveness of electric com- Jansson, J. (2011). Consumer eco-innovation adoption: Assessing attitu-
mercial vehicles: A case study from the USA market. Transportation dinal factors and perceived product characteristics. Business Strategy
Research Part C: Emerging Technologies, 26, 135–145. https://doi.org/ and the Environment, 20(3), 192–210. https://doi.org/10.1002/bse.690
10.1016/j.trc.2012.06.007 Jansson, J., Nordlund, A., & Westin, K. (2017). Examining drivers of
Franke, T., & Krems, J. F. (2013). Interacting with limited mobility
sustainable consumption: The influence of norms and opinion leader-
resources: Psychological range levels in electric vehicle use.
ship on electric vehicle adoption in Sweden. Journal of Cleaner
Transportation Research Part A: Policy and Practice, 48, 109–122.
Production, 154, 176–187. https://doi.org/10.1016/j.jclepro.2017.03.186
https://doi.org/10.1016/j.tra.2012.10.010
Javid, R. J., & Nejat, A. (2017). A comprehensive model of regional
Gass, V., Schmidt, J., & Schmid, E. (2014). Analysis of alternative pol-
electric vehicle adoption and penetration. Transport Policy, 54,
icy instruments to promote electric vehicles in Austria. Renewable
30–42. https://doi.org/10.1016/j.tranpol.2016.11.003
Energy, 61, 96–101. https://doi.org/10.1016/j.renene.2012.08.012
Jenn, A., Springel, K., & Gopal, A. R. (2018). Effectiveness of electric
Gillingham, K., & Palmer, K. (2014). Bridging the energy efficiency
vehicle incentives in the United States. Energy policy, 119, 349–356.
gap: Policy insights from economic theory and empirical evidence.
https://doi.org/10.1016/j.enpol.2018.04.065
Review of Environmental Economics and Policy, 8(1), 18–38. https://
Jensen, A. F., Cherchi, E., & de Dios Ort
uzar, J. (2014). A long panel
doi.org/10.1093/reep/ret021
Gnann, T., Stephens, T. S., Lin, Z., Pl€ otz, P., Liu, C., & Brokate, J. survey to elicit variation in preferences and attitudes in the choice
(2018). What drives the market for plug-in electric vehicles?-A of electric vehicles. Transportation, 41(5), 973–993. https://doi.org/
review of international PEV market diffusion models. Renewable 10.1007/s11116-014-9517-6
and Sustainable Energy Reviews, 93, 158–164. https://doi.org/10. Jensen, A. F., Cherchi, E., & Mabit, S. L. (2013). On the stability of
1016/j.rser.2018.03.055 preferences and attitudes before and after experiencing an electric
Glerum, A., Stankovikj, L., Th
emans, M., & Bierlaire, M. (2014). vehicle. Transportation Research Part D: Transport and
Forecasting the demand for electric vehicles: Accounting for atti- Environment, 25, 24–32. https://doi.org/10.1016/j.trd.2013.07.006
tudes and perceptions. Transportation Science, 48(4), 483–499. Jin, L., Searle, S., & Lutsey, N. (2014). Evaluation of state-level US electric
https://doi.org/10.1287/trsc.2013.0487 vehicle incentives. The International Council on Clean Transportation.


Junquera, B., Moreno, B., & Alvarez, R. (2016). Analyzing consumer attitudes
Guyon, I., & Elisseeff, A. (2003). An introduction to variable and fea-
ture selection. Journal of Machine Learning Research, 3, 1157–1182. towards electric vehicle purchasing intentions in Spain: Technological lim-
Harrysson, S., Ulmefors, M., & Kazlova, A. (2015). Overview and ana- itations and vehicle confidence. Technological Forecasting and Social
lysis of electric vehicle incentives applied across eight selected country Change, 109, 6–14. https://doi.org/10.1016/j.techfore.2016.05.006
markets. In H. Ny (Ed.). Karlskrona. Retrieved July 3, 2019, from Kester, J., Noel, L., de Rubens, G. Z., & Sovacool, B. K. (2018). Policy
http://www.diva-portal.org/smash/get/diva2:882227/FULLTEXT01.pdf. mechanisms to accelerate electric vehicle adoption: A qualitative
Haustein, S., & Jensen, A. F. (2018). Factors of electric vehicle adop- review from the Nordic region. Renewable and Sustainable Energy
tion: A comparison of conventional and electric car users based on Reviews, 94, 719–731. https://doi.org/10.1016/j.rser.2018.05.067
an extended theory of planned behavior. International Journal of Kim, M. K., Oh, J., Park, J. H., & Joo, C. (2018). Perceived value and
Sustainable Transportation, 12(7), 484–496. https://doi.org/10.1080/ adoption intention for electric vehicles in Korea: Moderating effects
15568318.2017.1398790 of environmental traits and government supports. Energy, 159,
He, X., & Zhan, W. (2018). How to activate moral norm to adopt elec- 799–809. https://doi.org/10.1016/j.energy.2018.06.064
tric vehicles in China? An empirical study based on extended norm Kley, F., Wietschel, M., & Dallinger, D. (2012). Evaluation of European
activation theory. Journal of Cleaner Production, 172, 3546–3556. electric vehicle support schemes. Paving the road to sustainable
https://doi.org/10.1016/j.jclepro.2017.05.088 transport (pp. 75–95). Routledge.
He, X., Zhan, W., & Hu, Y. (2018). Consumer purchase intention of Koller, D., & Sahami, M. (1996, July 3–6). Toward optimal feature selection
electric vehicles in China: The roles of perception and personality. [Paper presentation]. Stanford InfoLab Proceedings of 13th International
Journal of Cleaner Production, 204, 1060–1069. https://doi.org/10. Conference on Machine Learning, Bari, Italy (pp. 284–292).
1016/j.jclepro.2018.08.260 Kraskov, A., St€ ogbauer, H., & Grassberger, P. (2004). Estimating mutual
Helveston, J. P., Liu, Y., Feit, E. M., Fuchs, E., Klampfl, E., & Michalek, J. J. information. Physical Review. E, Statistical, Nonlinear, and Soft Matter
(2015). Will subsidies drive electric vehicle adoption? Measuring Physics, 69(6 Pt 2), 66138. https://doi.org/10.1103/PhysRevE.69.066138
 INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 13

Krupa, J. S., Rizzo, D. M., Eppstein, M. J., Lanute, D. B., Gaalema, Ozaki, R., & Sevastyanova, K. (2011). Going hybrid: An analysis of
D. E., Lakkaraju, K., & Warrender, C. E. (2014). Analysis of a con- consumer purchase motivations. Energy Policy, 39(5), 2217–2227.
sumer survey on plug-in hybrid electric vehicles. Transportation https://doi.org/10.1016/j.enpol.2010.04.024
Research Part A: Policy and Practice, 64, 14–31. https://doi.org/10. Peters, A., & D€ utschke, E. (2014). How do consumers perceive electric
1016/j.tra.2014.02.019 vehicles? A comparison of German consumer groups. Journal of
Kumar, T., & Thakur, T. (2020). Electric vehicle infrastructure plan- Environmental Policy & Planning, 16(3), 359–377. https://doi.org/10.
ning: A distribution side perspective. In Novel advancements in elec- 1080/1523908X.2013.879037
trical power planning and performance (pp. 118–127). IGI Global. Petrick, J. F. (2002). Development of a multi-dimensional scale for meas-
https://doi.org/10.4018/978-1-5225-8551-0.ch005 uring the perceived value of a service. Journal of Leisure Research,
Lafuente, V., Herrera, L. J., P
erez, M. D. M., Val, J., & Negueruela, I. 34(2), 119–134. https://doi.org/10.1080/00222216.2002.11949965
(2015). Firmness prediction in Prunus persica ’Calrico’ peaches by Rezvani, Z., Jansson, J., & Bodin, J. (2015). Advances in consumer elec-
visible/short-wave near infrared spectroscopy and acoustic measure- tric vehicle adoption research: A review and research agenda.
ments using optimised linear and non-linear chemometric models. Transportation Research Part D: Transport and Environment, 34,
Journal of the Science of Food and Agriculture, 95(10), 2033–2040. 122–136. https://doi.org/10.1016/j.trd.2014.10.010
https://doi.org/10.1002/jsfa.6916 Rocha, C., Melo, S., Rolim, C., & Baptista, P. (2016). The relevance of
Lane, B., & Potter, S. (2007). The adoption of cleaner vehicles in the customer expectations and preferences: Adoption of electric vehicles
UK: Exploring the consumer attitude–action gap. Journal of Cleaner and impacts on noise, energy consumption and air pollution. A
Production, 15(11–12), 1085–1092. https://doi.org/10.1016/j.jclepro. review from Portugal. In NHV analysis techniques for design and
2006.05.026 optimization of hybrid and electric vehicles. Shaker Verlag
Langbroek, J. H., Franklin, J. P., & Susilo, Y. O. (2016). The effect of Publications. https://doi.org/10.1016/j.seta.2016.03.006
policy incentives on electric vehicle adoption. Energy Policy, 94, Rogers, E. M. (2010). Diffusion of innovations. Simon and Schuster.
94–103. https://doi.org/10.1016/j.enpol.2016.03.050 Schmalfuß, F., M€ uhl, K., & Krems, J. F. (2017). Direct experience with
Larson, P. D., Vi
afara, J., Parsons, R. V., & Elias, A. (2014). Consumer battery electric vehicles (BEVs) matters when evaluating vehicle
attitudes about electric cars: Pricing analysis and policy implications. attributes, attitude and purchase intention. Transportation Research
Transportation Research Part A: Policy and Practice, 69, 299–314. Part F: Traffic Psychology and Behaviour, 46, 47–69. https://doi.org/
https://doi.org/10.1016/j.tra.2014.09.002 10.1016/j.trf.2017.01.004
Lee, J. H., Hardman, S. J., & Tal, G. (2019). Who is buying electric Schuitema, G., Anable, J., Skippon, S., & Kinnear, N. (2013). The role
vehicles in California? Characterising early adopter heterogeneity of instrumental, hedonic and symbolic attributes in the intention to
and forecasting market diffusion. Energy Research & Social Science, adopt electric vehicles. Transportation Research Part A: Policy and
55, 218–226. https://doi.org/10.1016/j.erss.2019.05.011 Practice, 48, 39–49. https://doi.org/10.1016/j.tra.2012.10.004
Leurent, F., & Windisch, E. (2011). Triggering the development of electric Sellmair, R., & Schelo, T. (2019). Analysis of the effect of charging
mobility: A review of public policies. European Transport Research infrastructure design on electric taxi driving profiles: A case study
Review, 3(4), 221–235. https://doi.org/10.1007/s12544-011-0064-3 approach on the example of Singapore. International Journal of
L
evay, P. Z., Drossinos, Y., & Thiel, C. (2017). The effect of fiscal Sustainable Transportation, 13(7), 479–496. https://doi.org/10.1080/
incentives on market penetration of electric vehicles: A pairwise 15568318.2018.1485792
comparison of total cost of ownership. Energy Policy, 105, 524–533. Shannon, C. E. (1948). A mathematical theory of communication. Bell
https://doi.org/10.1016/j.enpol.2017.02.054 System Technical Journal, 27(3), 379–423. https://doi.org/10.1002/j.
Li, F., Ou, R., Xiao, X., Zhou, K., Xie, W., Ma, D., Liu, K., & Song, Z. 1538-7305.1948.tb01338.x
(2019). Regional comparison of electric vehicle adoption and emission She, Z. Y., Sun, Q., Ma, J. J., & Xie, B. C. (2017). What are the barriers
reduction effects in China. Resources, Conservation and Recycling, 149, to widespread adoption of battery electric vehicles? A survey of pub-
714–726. https://doi.org/10.1016/j.resconrec.2019.01.038 lic perception in Tianjin, China. Transport Policy, 56, 29–40. https://
Li
ebana-Cabanillas, F., Herrera, L. J., & Guill
en, A. (2016). Variable doi.org/10.1016/j.tranpol.2017.03.001
selection for payment in social networks: Introducing the Hy-index. Sierzchula, W., Bakker, S., Maat, K., & Van Wee, B. (2014). The influ-
Computers in Human Behavior, 56, 45–55. https://doi.org/10.1016/j. ence of financial incentives and other socio-economic factors on
chb.2015.10.022 electric vehicle adoption. Energy Policy, 68, 183–194. https://doi.org/
Lieven, T. (2015). Policy measures to promote electric mobility–A glo- 10.1016/j.enpol.2014.01.043
bal perspective. Transportation Research Part A: Policy and Practice, Skippon, S. M. (2014). How consumer drivers construe vehicle per-
82, 78–93. https://doi.org/10.1016/j.tra.2015.09.008 formance: Implications for electric vehicles. Transportation Research
Lim, M. K., Mak, H. Y., & Rong, Y. (2015). Toward mass adoption of Part F: Traffic Psychology and Behaviour, 23, 15–31. https://doi.org/
electric vehicles: Impact of the range and resale anxieties. 10.1016/j.trf.2013.12.008
Manufacturing & Service Operations Management, 17(1), 101–119. Soltani-Sobh, A., Heaslip, K., Bosworth, R., Barnes, R., & Yook, D.
https://doi.org/10.1287/msom.2014.0504 (2016). An aggregated panel data analysis to model electric vehicle
Mcknight, D. H., & Chervany, N. L. (2001). What trust means in e- adoption rates (No. 16-4010). Transportation Research Board.
commerce customer relationships: An interdisciplinary conceptual Soltani-Sobh, A., Heaslip, K., Stevanovic, A., Bosworth, R., &
typology. International Journal of Electronic Commerce, 6(2), 35–59. Radivojevic, D. (2017). Analysis of the electric vehicles adoption
https://doi.org/10.1080/10864415.2001.11044235 over the United States. Transportation Research Procedia, 22,
Mersky, A. C., Sprei, F., Samaras, C., & Qian, Z. S. (2016). 203–212. https://doi.org/10.1016/j.trpro.2017.03.027
Effectiveness of incentives on electric vehicle adoption in Norway. Suykens, J. A. K., Van Gestel, T., De Brabanter, J., De Moor, B., &
Transportation Research Part D: Transport and Environment, 46, Vandewalle, J. (2002). Least squares support vector machines. World
56–68. https://doi.org/10.1016/j.trd.2016.03.011 scientific. https://doi.org/10.1142/5089
Moon, H., Park, S. Y., Jeong, C., & Lee, J. (2018). Forecasting electricity Taefi, T. T., Kreutzfeldt, J., Held, T., Konings, R., Kotter, R., Lilley, S.,
demand of electric vehicles by analyzing consumers’ charging pat- Baster, H., Green, N., Laugesen, M. S., Jacobsson, S., & Borgqvist,
terns. Transportation Research Part D: Transport and Environment, M. (2016). Comparative analysis of European examples of freight
62, 64–79. https://doi.org/10.1016/j.trd.2018.02.009 electric vehicles schemes—a systematic case study approach with
Moons, I., & De Pelsmacker, P. (2012). Emotions as determinants of examples from Denmark, Germany, the Netherlands, Sweden and
electric car usage intention. Journal of Marketing Management, the UK. In Dynamics in logistics (pp. 495–504). Springer. https://doi.
28(3–4), 195–237. https://doi.org/10.1080/0267257X.2012.659007 org/10.1007/978-3-319-23512-7_48
Mukherjee, S. C., & Ryan, L. (2020). Factors influencing early battery Thananusak, T., Rakthin, S., Tavewatanaphan, T., & Punnakitikashem,
electric vehicle adoption in Ireland. Renewable and Sustainable Energy P. (2017). Factors affecting the intention to buy electric vehicles:
Reviews, 118, 109504. https://doi.org/10.1016/j.rser.2019.109504 Empirical evidence from Thailand. International Journal of Electric
14 E. HIGUERAS-CASTILLO ET AL.

and Hybrid Vehicles, 9 (4), 361. https://doi.org/10.1504/IJEHV.2017. Appendix A: Questionnaire

10011120
Tran, M., Banister, D., Bishop, J. D., & McCulloch, M. D. (2012).
Realizing the electric-vehicle revolution. Nature Climate Change, Adequate range and charging
2(5), 328–333. https://doi.org/10.1038/nclimate1429 The driving range of an E&HV is satisfying.
Turcksin, L., Mairesse, O., & Macharis, C. (2013). Private household The driving range of an E&HV is sufficient for my mobility needs in
demand for vehicles on alternative fuels and drive trains: A review. everyday life.
I do not mind that it takes longer to charge battery cells than to refuel.
European Transport Research Review, 5(3), 149–164. https://doi.org/
I could integrate the charging of the batteries in my everyday life without
10.1007/s12544-013-0095-z any problems.
Vassileva, I., & Campillo, J. (2017). Adoption barriers for electric
vehicles: Experiences from early adopters in Sweden. Energy, 120, Enjoyable acceleration
I would perceive the fast acceleration of EVs as pleasant.
632–664. https://doi.org/10.1016/j.energy.2016.11.119 The immediate acceleration increases the driving comfort of Evs.
Wang, N., Liu, Y. (2015). Key factors influencing consumers’ willing- I would like the racy acceleration of the EVs.
ness to purchase electric vehicles in China. The 28th International
Enjoyable low noise emission
Electric Vehicle Symposium and Exhibition.
The lack of engine noise of EVs increases the driving pleasure.
Wang, N., Tang, L., & Pan, H. (2017). Effectiveness of policy incentives I would like the low soundscape of EVs.
on electric vehicle acceptance in China: A discrete choice analysis. I would not need to change my driving style due to the noiselessness of the
Transportation Research Part A: Policy and Practice, 105, 210–218. EVs.
https://doi.org/10.1016/j.tra.2017.08.009 I believe that the lack of noise from the EV is not dangerous for road traffic.
Wang, N., Tang, L., & Pan, H. (2018). Analysis of public acceptance of The lack of engine noise would not make driving more difficult.
electric vehicles: An empirical study in Shanghai. Technological Satisfying safety and reliability
Forecasting and Social Change, 126, 284–291. https://doi.org/10. I would be as safe in an EV as in a comparable conventional compact car.
1016/j.techfore.2017.09.011 The safety in EVs is a given.
Wang, N., Tang, L., & Pan, H. (2019). A global comparison and assessment An EV will take me safely to my destination.
I rely on the new technology of EVs.
of incentive policy on electric vehicle promotion. Sustainable Cities and EVs are reliable.
Society, 44, 597–603. https://doi.org/10.1016/j.scs.2018.10.024 I can depend on an EV to reliably get me where I need to go.
Wang, Z., Zhao, C., Yin, J., & Zhang, B. (2017). Purchasing intentions of
Price
Chinese citizens on new energy vehicles: How should one respond to
E&HVs are expensive.
current preferential policy? Journal of Cleaner Production, 161, E&HVs are costly.
1000–1010. https://doi.org/10.1016/j.jclepro.2017.05.154 The price of E&HVs is higher than that of corresponding gasoline cars.
Zhang, B., & Tay, F. E. (2017). An integrated approach using data min- The price of EVs is higher than I expected.
ing and system dynamics to policy design: Effects of electric vehicle Perceived benefit
adoption on CO2 emissions in Singapore [Paper presentation]. The cost of charging the battery is very low.
Industrial Conference on Data Mining, Cham (pp. 258–268). https:// The maintenance costs are lower in EVs.
doi.org/10.1007/978-3-319-62701-4_20 Incentives
Zhang, X., Wang, K., Hao, Y., Fan, J. L., & Wei, Y. M. (2013). The It is easier to receive subsidies for the purchase of EVs than for the rest of
impact of government policy on preference for NEVs: The evidence vehicles.
from China. Energy Policy, 61, 382–393. https://doi.org/10.1016/j. The purchase of an E&HV is more cost-effective when monetary incentives are
enpol.2013.06.114 in place.
It is easier to receive subsidies for the purchase of EVs than for the rest of
Zhu, J. J. (2016). Analysis of New Zealand specific electric vehicle adop-
vehicles.
tion barriers and government policy [Thesis]. Victoria University of I am aware of the subsidies available for the purchase of EVs.
Wellington.
Infrastructure
It is hard to find a station where an EV can be charged.
It is hard to find an auto repair shop that services EVs.
Intention to adopt
Next time I buy a car, I will consider buying an E&HV.
I expect to drive an E&HV car in the near future.
I have the intention to drive an E&HV in the near future.
 INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 15

Appendix B: Control questions

Do you have your own vehicle? Yes/No

Did you know about the existence of electric and hybrid vehicles? Yes/No
Do you own an electric or hybrid vehicle? Yes/No
What’s the primary vehicle in your household? Diesel/Gasoline/Hybrid/Plug-in hybrid/Electric
Are you contemplating the purchase of an electric vehicle? Very unlikely/Not Likely/Likely/Very likely/I already own an electric vehicle