Transportation Research Part A 132 (2020) 214–225

Contents lists available at ScienceDirect

Transportation Research Part A

journal homepage: www.elsevier.com/locate/tra

Who will buy alternative fueled or automated vehicles: A modular,

T
behavioral modeling approach
⁎
Ioannis Tsouros , Amalia Polydoropoulou
University of the Aegean, Korai 2A, Chios 82100, Greece

A R T IC LE I N F O ABS TRA CT

Keywords: Future car purchase can determine an array of things ranging from CO2 emissions to urban life
Future car purchase quality. For this reason, models and methods predicting car purchase are valuable to policy
Alternative fuel makers. This paper examines the future car purchase choice, using modules and different levels of
Menu-based choice attributes that construct a car, and measures the effect on the purchase choice of personality
Automated vehicles
traits such as symbolic/exuberant attitudes towards vehicles. The results may enable policy
CAVs
AFVs
makers to focus on certain market segments when promoting alternative fuel and automated
vehicles. The paper proposes a hybrid choice model, with latent variables capturing the pro-
environmental, exuberant and tech-friendly attitudes of individuals. The questionnaire presented
to the respondents is in the form of a menu, from which participants may choose five different
types of vehicle characteristics (engine size, type of car, fuel type, car edition and level of au-
tomation) to construct their ideal vehicle. Results indicate a negative correlation between sym-
bolic, exuberant attitudes towards automobiles, the view of cars as symbols, and willingness to
purchase a hybrid or electric vehicle. The findings further suggest that there is a correlation
between symbol-driven exuberant attitudes and the desire to buy a larger vehicle. This paper
examines the relationship between the symbolic perception of cars and, simultaneously, a range
of characteristics, to discover which car attributes the symbolic perception affects. It also pro-
poses an integrated framework for the modeling of future car purchase, with the hypothesis that
each of the three presented latent variables can affect different modules of the individual’s ideal
car concept.

1. Introduction

Alternative fuel vehicles (AFVs) are widely regarded as more sustainable alternatives to fossil fuel vehicles and as a response to
the widespread externalities of transport, especially CO2 emissions. Additionally, the role of connected and autonomous vehicles
(CAV) in a future transport system is still unclear; a series of recent and ongoing studies (Gkartzonikas and Gkritza, 2019) explore
their potential impact on transportation, environment and society. However, the market penetration of these vehicles is not yet at a
significant level. Apart from this, AFVs have drawn controversial reviews, with public opinion often regarding them as unreliable,
unnecessary or immature. CAVs (Thomopoulos and Givoni, 2015; Zmud et al., 2019), also, have a public image of immature tech-
nology on the fringes of science fiction, and the systemic effects of such vehicles are still under considerable debate (see the work of
Bahamonde-Birke et al. (2018) for an in-detail review of the systemic effects of CAV). Thus, it is important to understand the
relationship between individuals’ perceptions of vehicles and their willingness to purchase particular cars.

⁎
Corresponding author.
E-mail addresses: jtsouros@aegean.gr (I. Tsouros), polydor@aegean.gr (A. Polydoropoulou).

https://doi.org/10.1016/j.tra.2019.11.013
Received 16 October 2018; Received in revised form 11 November 2019; Accepted 15 November 2019
Available online 25 November 2019
0965-8564/ © 2019 Elsevier Ltd. All rights reserved.
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

A number of studies, such as (Axsen and Kurani, 2013; Campbell et al., 2012; Larson et al., 2015) approach the subject of car
purchase by measuring the effect of attitudes and perceptions. Other studies investigate the relationship between the willingness to
buy an AFV and certain personality traits. Nordlund et al. (2016), in a recent article, argue that different groups of car owners
(conventional vehicles, AFVs, EVs) have different norms and sets of values, with owners of AFVs being more open to change, less
conservative and more aware of the external costs of transportation.
This paper addresses this topic by presenting an integrated modeling framework to explore future car purchase, especially that of
AFVs and CAVs, based on certain personality traits. It describes a menu-based, modular approach to car purchase, whereby the
respondent can choose different levels of various attributes in order to construct the person’s ideal vehicle. This method is distinct
from the traditional stated preference experiments, in which ready-made products appear to the respondent. In this case, the re-
spondent has the full choice set to choose from, with each attribute level having a price that sums up to the final vehicle price.
More specifically, this paper estimates a hybrid choice model to measure the effect that pro-environmental attitudes, the sym-
bolism of a car, and tech-friendly values have on individuals’ future car purchase choices. Results from the paper’s case study reveal
that using just a symbolic latent variable, which we call Exuberance, can lead to different results depending on the study area and
lifestyle, and that using two or more latent variables can enable a further and more detailed investigation of future car purchase.
The paper is composed of six sections. Section 2 presents the state of the art, focusing on relevant work, especially with respect to
AFVs. Section 3 describes the methodological framework, presenting a hybrid choice model containing the three latent variables. A
case study with its data collection methodology and descriptive statistics is presented in Section 4. Section 5 gives the model esti-
mation results. Finally, Section 6 discusses the findings and their policy implications.

2. Literature review

A series of previous studies have pointed out the importance of the symbolism-driven and instrumental perception of vehicles,
pro-environmental stance and other personality traits for car or fuel type preference. Attitudes and perceptions are regarded as
potential behavior motivators and may affect the utility of a certain alternative in given circumstances, when treated in the correct
way in the modelling process (Bahamonde-Birke et al., 2017)
The symbolic attributes of AFVs were the center of attention for a UK study (Skippon and Garwood, 2011), which used a Real-Life
experience of electric vehicles to “reduce psychological distance” on the part of the participants, who widely acknowledged the
environmental benefits of using AFVs. The study discusses a number of instrumental attributes such as willingness to pay higher
capital costs for lower running costs, performance issues, recharging time and place, and other relevant vehicle attributes. However,
it also pointed out that the symbolic attributes of the respondents that would make them ideal candidates for the purchase of an AFV
were: Openness, Conscientiousness, and Agreeableness. This pattern unfolds in many relevant studies, and it could be defined as an
alternative symbolic perception of vehicles, namely the symbolic perception of AFVs.
Instrumental perception has been discussed in many studies, focusing on the limitations of AFVs, such as higher costs and less
autonomy. The interpretation of vehicle performance is discussed by Skippon (2014), who presents an extensive literature review on
how different instrumental factors are evaluated, and groups them into independent dimensions: dynamic and cruising performance.
In dynamic performance the driver engages actively in the change of status of the vehicle, while in cruising performance the driver
maintains the condition of the vehicle. Once again, the pattern of openness to ideas, pro-environmental and pro-social values is
related to the preference for electric vehicles. EVs are perceived as offering a good cruising performance.
Pro-environmental attitudes are among the attitudes most popularly linked to the favorability of AFVs. Tsouros and
Polydoropoulou (2017) use a hybrid choice model to investigate the relationship of pro-environmental attitudes to future car pur-
chase. Results indicate that people holding pro-environmental values are more likely to buy a hybrid car independently of macro-
economic scenarios or incentives such as green taxation. However, Potoglou and Kanaroglou (2007) find that, although various
incentives such as free parking are not significant, others such as tax relief are.
Axsen et al. (2015) study preference and lifestyle heterogeneity between potential Plug-in electric vehicle (PEV) users in Canada.
Interesting results include the finding that household income does not significantly differ between the lifestyle constructed classes,
whilst a variation of interest in vehicle types, valuation of fuel savings, environmental concern and level of education were far more
determinant factors. Members of the most pro-EV class have the highest standards of technology and environmentally oriented
lifestyle. This indicates that potential PEV buyers can be hardline pro-environmentalists but also technology and gadget enthusiasts,
two social clusters that do not always overlap.
Correlation between technology enthusiasts and AFVs is discussed in other research papers as well. Noppers et al. (2015) link
earlier innovation adopters with a stronger intention to purchase an electric vehicle. Also, early adopters tend to place more value
than later adopters on the symbolic attributes of an electric car. Additionally, Petschnig et al. (2014) develop an adoption model for
AFVs in which they include a wide array of factors affecting the purchase of an AFV, from social, personal and moral obligations to
visibility of refueling stations or compatibility of the AFVs with existing driving behavior. A finding that is unexpected, according to
the authors, is the non-significant effect of trialability on attitude formation. This study points out that perceiving the car as a symbol,
projecting attitudes and emotions onto it, may result in early adoption of an AFV.
Schuitema et al. (2013), using OLS linear regression analysis, investigate the relationship of instrumental, hedonic and symbolic
attributes with the decision to purchase an electric vehicle. Study results reveal that, while previous research had concluded that
instrumental attributes of electric vehicles such as driving range were the most important factors affecting the purchase decision,
symbolic and hedonic attributes were more important. However, the perception of instrumental and symbolic attributes is vehicle
specific; for example, people in this survey tend to regard Battery Electric Vehicles as having less instrumental value (particularly

215
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

regarding their driving range) which results in less joy and consequently more negative symbolic attributes.
In their recent work, Zoepf and Keith (2016) study and model user decision making in the car-sharing context, using stated
preference experiments. Among other results, they find that car-sharing users widely prefer hybrid vehicles. More precisely, there is a
much larger exposure (400 to 1) of car sharing users to hybrid vehicle technology than of regular, household owned vehicle users.
Using scenarios and simulation to forecast future market shares for electric vehicles Jensen et al., (2017) predict a 2–6% market
share for electric vehicles in the Danish market by 2020, stating that this may be underrated based on the available data. Using a
method that combines choice models (disaggregate) with diffusion models which take into account the lifecycle of a product, they
also present a series of future scenarios, for example purchase sales of EVs by 2020 (ranging from 130 to over 5000 per month, in the
low and high diffusions scenarios). Overall this research work is important as it demonstrates the complexity of estimating future
vehicle purchase market shares based on the different and intertangled factors that affect the market shares.
The future penetration of CAVs is also a very frequent research topic, especially during recent years. As Kyriakidis et al. (2015)
reveal, using a 5000 respondent internet survey, almost 70% of respondents think that fully automated vehicles will reach 50% of the
market by 2050. To model the level of automation they use the levels designed by Wending (2014) and SAE definitions (2016). A
contribution of this study is that, unlike most of the previous studies (for example see (Howard and Dai, 2014; Payre et al., 2014)), it
presents public opinion on CAVs from a global perspective, not only that of western countries. Findings include not only the po-
pularity of manual driving, seen as the most enjoyable mode of driving by individuals, but also the fascination with automated
vehicles. Main concerns center on hacking and misuse of vehicles. Finally, people are most willing to pay for a fully automated
driving car, while the step from “partially to highly automated driving was not considered worth extra money”.
In another CAV study, Bansal et al. (2016) assess public opinion on automated vehicles and new vehicle technologies in general in
Austin, Texas. Regarding adoption timing of automated vehicles, the researchers found that older people living far from work are
more likely to adopt an AV when most of their friends do, while younger people living in urban areas are likely to adopt with less
dependence on their friends’ behavior. Also, results indicate that individuals who drive more are more apt to adopt automated
vehicles and to spend more money on a fully automated product. Finally, higher-income, tech-savvy males in urban areas are found to
be more interested in CAVs. This finding may offer a hint as to what context of the latent variable will be useful in measuring
attitudinal stances towards CAVs.
Recent years have seen an acceleration in the research exploration and the development efforts to make CAVs a reality and to
assess implications. Gkartzonikas and Gkritza (2019) provide a detailed review of survey-based studies examining potential user
preferences and behaviors regarding AVs. It is true that most studies provide a descriptive analysis of user acceptance and awareness
towards CAVs (Kyriakidis et al., 2015) or (Nordhoff et al., 2018), willingness-to-pay for CAVs (Kyriakidis et al., 2015; Talebian and
Mishra, 2018), or attitudes and perceptions towards CAVs (multiple studies, indicative examples are (Liljamo et al., 2018;
Panagiotopoulos and Dimitrakopoulos, 2018; Sanbonmatsu et al., 2018).
On the other hand, there are some studies utilizing stated-preference experiments to explore user preferences regarding CAVs,
most of which develop behavioral models to model user choices. Research ranges on the effect of attitudes and perceptions
(Haboucha et al., 2017; Krueger et al., 2016), exploration of preferences in multimodal options and last mile solutions (Yap et al.,
2016), socio-demographic differences affecting choices and user perceptions (Hohenberger et al., 2016), or other behavioral models
(Correia et al., 2019; Daziano et al., 2017).
The literature review on the future purchase choice of an AFV or a CAV revealed a broad range of information and knowledge and
also some gaps that are not yet filled due to the very recent developments in both these technologies. The relationship of attitudes and
perceptions with the effect that latent, psychological traits have on the final purchase choice of the vehicle is the core subject of many
of the studies cited in this section. Ecologically conscious individuals are more likely to be associated with the purchase of an AFV.
Also, individuals who hold a symbol-driven perception of cars are found in most studies to be prone to buying an AFV. The initial
investigation of public opinion on CAVs reveals that tech-friendly individuals are more likely to state that they would adopt an
automated vehicle. These findings from the literature led us to the decision to use three latent variables to model the intention to
purchase a vehicle: Eco-Friendliness, Exuberance, and Tech-Savviness. This paper describes a model framework for these three latent
variables and tests a part of it, using one latent variable in a case study on the Greek island of Chios.

3. Methodology

This section describes the proposed methodological framework for studying future car purchase. It initially presents the ques-
tionnaire design and implementation and then the modeling framework and related equations.

3.1. Survey design

The “Future Car Purchase” survey was developed as part of the data collection effort conducted for the “Green Transport in
Islands” study in the Greek islands of Lesvos and Chios (2013–2015). The survey was conducted in parallel with other survey
instruments which collected data regarding socio-economic characteristics, activity diaries, GPS traces, Stated Preferences experi-
ments about mode choice and tourist destination choice, spending patterns and attitudinal data.
This research uses a sample of 550 island households (roughly 2.5% of the population).
The web survey presented the respondent with a menu-based approach. A “build-your-own” section was offered to each re-
spondent in which he/she could choose between levels of different attributes such as engine size, car type, car edition and fuel type,
actively constructing his/her ideal vehicle. Then, a price for the selection would appear on the screen, and the respondent had the

216
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

Table 1
Alternative attribute variables (with variable name).
Attribute level Engine Size (CC) Car type Fuel type Car version Degree of Automation

1 1000–1199 – CC1 Small city car – CS1 Regular gas – FT1 Basic version – CE1 No Automation – AT1
2 1200–1599 – CC2 Medium city car – CS2 Diesel – FT2 Comfort version – CE2 Function-Specific Automation – AT2
3 1600–1999 – CC3 Sedan – Family car – CS3 Hybrid – FT3 “Hi-Tech” version – CE3 Combined Function Automation – AT3
4 2000 – CC4 SUV – CS4 Electric – FT4 Sports version – CE4 Limited Self-Driving Automation – AT4
5 Sports car – CS5 Natural gas – FT5 Full extra – CE5 Full Self-Driving Automation – AT5
6 Pick-up truck – CS6 Luxury – CE6

opportunity to either accept the constructed vehicle and submit his/her choice, re-construct the vehicle from the beginning, or
change the levels of some attributes. Every different attribute level had a specific price for its selection which is visible to the
respondent at all times. During the different tasks (4) the price for all levels varies. Total participants of the web survey are 600
individuals.
This paper uses a menu-based approach, which is previously used in the literature (Ben‐Akiva and Gershenfeld, 2002; Liechty
et al., 2003). The main advantage of Menu-Based (MBC) approaches, compared to CBC approaches, is that is more closely real-world
choice situations where the consumer effectively chooses between varying levels of multiple attributes in order to create their “ideal”
product, rather that choosing among a fixed number of pre-designed packages. Our opinion is that for specific choice situations such
as future car purchase or choose of a MaaS (mobility-as-a-service) plan using the Menu-Based approach can mimic the choice
situation of the user more realistically. For a detailed comparison of CBC and MBC see Table 1 in Liechty et al., (2003).
The menu-based approach can also be described as a modular approach, in which the respondent can pick different modules to
create the ideal vehicle. Table 1 presents the attributes and levels: engine and car size, fuel type and degree of automation.
It should be noted that it is possible to create a non-automated vehicle in this experiment by choosing level 1 in automation
attribute list. This possibility ensures that participants who do not want to have any level of automation on their ideal vehicle have
the chance to make such a decision. This way the experiment is “reduced” to a conventional, non-automated, alternative fuel ex-
periment. While we understand the complexity and the alienation of the CAV concept from the ordinary consumer, we believe that it
is worth trying to initially explore these preferences, especially at a time and place (insular area in Greece) where the market
penetration for such vehicles is non-existent at the time.
This survey design creates a large, exhaustive choice set, constituting a full choice set, in which every constructed alternative
varies from another by at least one attribute level. The number of alternatives in the full choice set is 3600 (4 × 6 × 5 × 6 × 5). We
coded the different alternatives starting from Alternative 1 in the following way:

Alternative1: ENGINESIZE1 (Engine Size) – CARSIZE1 (Car Type) – FUELTYPE1 (Fuel Type) – CAREDITION1 (Car Version) – AT1
(Degree of Automation)
Alternative 2: ENGINESIZE1 (Engine Size) – CARSIZE1 (Car Type) – FUELTYPE1 (Fuel Type) – CAREDITION1 (Car Version) – AT2
(Degree of Automation)
…
Alternative 3600: ENGINESIZE4 – CARSIZE6 – FUELTYPE5 – CAREDITION6 – AT5

Starting from Alternative1, AT starts from 1 and runs through 5, then CAREDITION runs from 1 to 6, then FT is set to 2, and again
CAREDITION runs from 1 to 6, until FT is set to 5. Then CARSIZE is set to 2, and the whole process is repeated until ENGINESIZE is set
to 4.
This is also the way the utilities are formed; for example:
V4 = βCC1 ∗ CC1 + βCS1 ∗ CS1 + βFT 1 ∗ FT 1 + βCE1 ∗ CE1 + βAT 4 ∗ AT 4 + βPRICE 4 ∗ PRICE 4

This methodology creates 3600 different alternatives.

3.2. Effect of latent traits

To understand the relationship of symbolic attitudes towards vehicle types we have to use a combination of latent variables.
Table 2 presents the possible user types that the two latent variables can produce. The use of one latent variable may result in very
different purchase choices. For example, someone who has exuberant attitudes towards vehicles could be a potential buyer of an AFV
if the person also has a pro-environmental stance, but could also be a regular, gasoline car buyer if he/she does not (see Table 2).
Table 3 presents the socio-demographic data of the collected sample.
This hypothesis is going to be tested using the following modeling framework. The case study tests it by creating a latent variable
called Exuberance which is similar to the definitions of symbolic view of vehicles found in the literature. The relationship of this latent
variable with different cars, based on fuel type and car size differences, will be able to verify the need for more than one latent
variable to classify potential buyers of AFVs and CAVs. It must be noted that for CAVs a third variable, measuring the “Tech-
Savviness” of the individual, may be used. Data from social media and the extensiveness of use, possession of gadgets, age, and
occupation, can be utilized as input for the structural model of the “tech-adaptation” latent variable.

217
 I. Tsouros and A. Polydoropoulou

Table 2
Latent traits and user profiles.
Tech-Saviness: Yes Tech-Saviness: No

Eco-friendliness:Yes Eco-friendliness :No Eco-friendliness:Yes Eco-friendliness :No

Exuberance: Yes Ecologically conscious, with symbolic view of Indifferent to environment, with symbolic view of Ecologically conscious, with Indifferent to environment, with symbolic view of
vehicles and tech savvy – Prone to purchase both vehicles – Prone to buy a large, gasoline fueled car or symbolic view of vehicles – Inclining vehicles – Prone to purchase a high, gasoline fueled

218
an AFV and a CAV
Exuberance: No Ecologically conscious, without symbolic view
of vehicles – Bicycle user/ Non-motorist – May
have a small chance of purchasing CAV
a CAV
Indifferent to environment, with practical view of
vehicles – Decides future vehicle purchase in a
“rational” way, taking into account more practical
attributes such as performance or price – Because of
tech savviness may be prone to buying a CAV
towards AFV
Ecologically conscious, without
symbolic view of vehicles – Bicycle
user/ Non-motorist
car but no CAV
Indifferent to environment, with practical view of
vehicles – Decides future vehicle purchase in a
“rational” way, taking into account more practical
attributes such as performance or price
Transportation Research Part A 132 (2020) 214–225
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

Table 3
Socio-demographics.
Socio-demographic characteristics n = 515

Characteristic Category Percentage

Age 18–30 42.9

31–59 40.4
60+ 16.8

Gender Female 51.3

Male 48.7

Car License Yes 79.0

No 21.0

Motorcycle License Yes 53.0

No 47.0
Occupation Employer 3.5
Freelancer 12.9
Employee 26.9
Household/Housewife 8.8
Unemployed 2.9
Pensioner 11.1
Student 20.5
Not Stated 13.5

Education No Education 3.7

Elementary 5.6
High School 13.7
Lyceum 36.6
IEK (~Community college) 9.9
AEI (University) 24.5
Master’s degree 4.3

3.3. Framework

Fig. 1 presents the overall modeling framework. The three latent variables are structured from socio-demographic data such as
gender, age, occupation, education, income, and possession of gadgets (such as a smartphone). They are then measured through
indicators. The indicators are Likert scale attitudinal questions, with I1 to I5 used to measure Exuberance while Ι6 to I10 measure Eco-
Friendliness and Ι11 to I15 measure Tech-Savviness. The various alternative attribute variables, the modules of construction of the ideal
vehicle, are inserted directly into the utility function of the alternatives, as well as the latent variables. The goal of this framework is
to measure the effect that each level of each attribute has on the car purchase choice, and also to see how the latent variables affect
the selection of specific vehicles. This framework will test our initial hypothesis that each combination of these three latent variables
will influence the future purchase choice in a different way.

3.3.1. Equations
In this section, the model equations are presented. For the case study, we use a model with three LVs: Eco-Friendliness, Exuberance
and Tech-Savviness.

3.3.2. Structural model

LVn = Zn γ + ωn ω N (0, Σω) (1)

Un = Xn β + LVn δ + εn ε N (0, Σε ) (2)

where

LVn : Latent Variable

Zn, Xn : Explanatory Variables
ωn , εn : Vectors of random disturbance terms
β , γ . δ : Unknown parameters

Un is the utility of the alternative, β is the vector of the observed variables, γ is a diagonal matrix of unknown parameters
associated with the latent variable LVn , and Un is a vector of random disturbance terms associated with the utility terms. The choice
model is assumed to be based on utility maximization:
yi = 1, if Ui = max{Ui} i = all the 3600 cars (3)

219
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

Fig. 1. Framework.

yi = 0 , otherwise
yi : the choice indicator

3.3.3. Measurement model

ILVn = α + λLVn + υn υ N (0, Συ) (4)

where ILVn is a vector of attitudes and perceptions, α is a vector of unknown parameters that indicate the association between the
responses to the scale, λ are vectors of unknown parameters that relate the random variable to the indicators, and υn is a vector of
random error terms.

3.3.4. Likelihood function

P (yn |LVn; β , δ , Σε ) f (ILV |LVn; λ, Σ υ )

f (yn , ILV |Zn, Xn; α, β , γ , δ , λ ) = ∫LVn f (LVn |Xn ; γ , Σω ) dLVn (5)

220
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

Table 4
Structural model results.
Variable name: Exuberance b t-Stat

Mean 3.83 43.61

Age (> 65) 0.038 2.86
Female −0.005 −0.23
Has Smartphone 0.02 1.85
Education: Highschool 0.099 1.98
Has motorcycle license 0.137 2.47
sigma −0.252 −2.82

Variable name: Tech-Saviness b t-Stat

Mean 7.72 18.02

Education level: University and higher 0.82 7.46
Age (as continuous variable) −0.28 −5.725
Number of cars in the household 0.40 1.552
Occupation: Private Employee 0.254 1.92
Occupation: Freelancer 0.410 2.72
sigma −0.340 −3.01

Variable name: Eco-Friednliness b t-Stat

mean 3.71 16.61

Male −0.589 −3.12
Age: Under 30 1.09 4.90
Occupation: Employee −0.952 −3.99
Education: Graduate degree 0.963 3.04
Occupation: Pensioner −0.924 −2.83
Occupation: Student 1.39 5.07
sigma 1.16 12.42

4. Descriptive statistics

4.1. Study area

The island of Chios is located in the Aegean Sea, seven miles off the coast of Asia Minor. It is the fifth-largest of the Greek islands,
with 51,320 persons living in the main city or the villages and towns of Chios. Chios has a significant amount of car ownership per
capita, and approximately 2500 new vehicles appear on the island each year. Considering the island population, this is a large
number (5% of the total population). With a population density of 427 (people/km2), Chios has 407 cars per 1000 people as of 2015,
making it the third-highest car ownership area in Greece after the metropolitan areas of Athens and Thessaloniki. It also has one of
the highest rates of motorcycle ownership in Greece.
The island of Lesvos is also located in the Aegean Sea, north of Chios. It is the third-largest Greek island with a population of
almost 86,000, one third being located in the main town of Mytilene. There were 25,632 privately owned vehicles on the island as of
2015. The effect of seasonality is high on both islands, as the population tends to double during summer months, creating a very
different transport demand situation. The presence of AFVs and CAVs is almost non-existent in both islands.

5. Model results

This section presents the results of the model and in particular, the results of the structural and measurement model of the latent
variables and the choice model.
Table 4 presents the results of the structural model presents the results of the structural model of the latent variables. As can be
viewed from these results, older, lower education, motorcycle license holders are more likely to have symbolic attitudes towards
vehicles. Also, owning a smartphone has a positive sign and gender is not statistically significant, contrary to findings in the literature
that mostly show men as having strong symbolic links to vehicles and women as tending towards more practical ones. Being a
student, younger than 30, having a graduate degree and not working as an employee are the variables that structure the Eco-
Friendliness LV. Also, having higher education, being younger and working either as an employee in the private sector or a freelancer
are the variables that form the Tech-Savviness LV.
Table 5 presents the results of the measurement model.
Table 6 presents the results of the choice model. The second level of engine size attribute (sizes from 1200 to 1599 CC) seems to be
preferable among respondents, followed by the small-sized engines, with the larger sizes coming last. The second tier of car size (a
medium sized city car) is preferred, all sizes larger than that coming second in preference and small sized city cars last. When
considering fuel type, gasoline is the preferred one, followed by hybrid, leaving diesel, electric and natural gas as third options. The
Comfort version of car edition (a level up from the basic one) is the preferred car edition, followed by all other, more expensive,
editions which are preferred to the basic edition.

221
 I. Tsouros and A. Polydoropoulou

Table 5
Measurement model results.
Eco-Friendliness Exuberance Tech-Saviness

Variable name B t-Stat Variable name B t-Stat Variable name B t-Stat

α1 3.59 0.178 a1 aa1

α2 1.98 0.214 a2 −13.4 −3.68 aa2 −22.2 −4.42
α3 1.96 0.198 a3 −12.4 −2.37 aa3 −10.9 −5.20
α4 a4 −5.17 −2.27 aa4 −9.01 −4.20
α5 2.13 0.285 a5 −13.8 −1.78 aa5 −2.82 −4.12
I believe that protecting the environment is a critical issue 0.714 12.89 I enjoy the sound of the vehicle engine a lot I rely on technology to get things done

222
I act in an environmentally friendly manner 0.783 13.23 The vehicles I own show “who I am.” 4.12 2.82 Internet is a big part of my everyday life 1.94 8.42
I recycle in a weekly basis 0.611 12.48 I may envy somebody with a nice vehicle 3.98 2.82 I own a large part of gadgets 0.72 10.92
I take part in environmental action (reforestation, cleaning of The vehicle brand is more important than the vehicle 2.08 2.75 Our civilization stops when there is no 4.90 7.40
shores, etc.) characteristics electricity
I use energy saving lightbulbs 0.874 11.12 The vehicles I own provide prestige to me 4.26 2.81 My life would be extremely hard without 3.12 2.80
technology
σ1 0.615 16.11 s1 1.94 32.03 ss1 0.33 40.02
σ2 0.526 10.54 s2 −0.916 −13.70 ss2 −0.54 −28.90
σ3 1.38 18.87 s3 −1.11 −27.15 ss3 0.74 −18.78
σ4 1.19 17.06 s4 −0.603 −31.23 ss4 0.87 −24.20
σ5 1.05 17.26 s5 −0.467 −19.38 ss5 −0.58 −12.12
Transportation Research Part A 132 (2020) 214–225
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

Table 6
Choice model results.
Variable Name B t-Stat

βCC1 5.71 2.07

βCC2 6.96 2.20
βCS2 5.58 7.20
βCSOTHER 4.12 8.81
βFTGAS 8.24 5.54
βFTHYBRID 6.77 6.44
βCE2 6.35 6.34
βCEOTHER 1.84 8.26
βprice −2.05 −9.23
γExuberance1 (specific to alternatives with hybrid fuel) −0.281 −3.68
γExuberance2 (specific to alternatives with engine size larger than 1600CC) 0.175 2.89
γTechSaviness (specific to alternatives with Automation levels 4 or 5) 4.12 8.20
γEco-Friendliness (specific to alternatives with hybrid fuel) 0.585 2.21
Rho-square-bar 0.584
Draws 10,000

The latent variable Exuberance negatively affects the choice of buying a hybrid fueled vehicle. This relationship means that
respondents holding symbolic attitudes towards vehicles do not prefer hybrid fueled cars as a purchase choice. Also, the latent
variable positively affects the choice to purchase a product of engine size above 1600 CC. The LV Tech-Saviness positively affects the
purchase of alternatives with higher levels of automation (levels 4 and 5). Finally, Eco-Friendliness LV positively affects the purchase
of hybrid vehicles (but not the purchase of electric vehicles in a statistically significant way.)

6. Findings and discussion

This paper proposed an integrated framework with which to model future car purchase of a conventional vehicle, an AFV or a
CAV, through the use of three latent variables to account for behavioral heterogeneity and latent personality traits that affect the
purchase decision. It did so by offering the respondents a modular/menu-based approach. To our knowledge, this is the first study
that attempts such an investigation using a full choice set, presenting the respondents with a menu style choice experiment whereby
they can construct their ideal vehicle based on different levels of different attributes.
The paper tested the methodology by designing and implementing a hybrid choice model in a case study of the Greek islands of
Chios and Lesvos. The results of the applied methodology include classification of these levels. Medium sized city cars, with an engine
size in the range of 1200 to 1600 CC, gas fueled and a level up from basic edition are the most popular vehicles according to the
choice model. These are the levels that achieve the higher coefficients of the model. Of course, the price of the constructed vehicle has
a negative effect on the choice. The effect of the latent variable Exuberance , which reflects symbolic attitudes towards cars, is negative
on the choice of purchasing a hybrid vehicle and positive on the choice of buying a car with a large engine.
The results for the effect of symbolic attitudes and perception on the willingness to purchase an alternative fuel vehicle are in
accordance with part of the existing literature and in disagreement with other parts of it. In a considerable number of papers, the high
symbolic view of cars (as opposed to the practical view) is associated with the willingness to purchase an alternative fueled car. In this
narrative, a person can see him/herself as having a particular identity (that of a green, environmentally friendly, tech/gadget person),
and will thus regard the purchased vehicle as something innovative, different from the standard, even anti-conformist in some cases.
However, it seems that in the case of this paper, the people who idolize cars are those who are unwilling to purchase a hybrid vehicle,
while people that perceive themselves as environmentally conscious are more likely to buy hybrid vehicles. The latter result could be
explained by creating another narrative in which people who perceive their car as a symbol (that is, as an indicator of power or
prestige) and also love non-practical aspects of the car (engine sound) are more prone to buying a “strong” gasoline car than a “soft”
alternative fueled vehicle. This confirms the hypothesis that more than one latent variable is needed to understand the different
potential buyers' profiles consistently (see proposed profiles in Table 2). Finally, technology-savvy respondents are more likely to
purchase vehicles with higher levels of automation. This is an important result as it highlights a potential market for CAVs; pin-
pointing the users that are more likely to be the first adopters of this technology.
Consequently, it could be said that assessing the effect of the symbolic on the choice may not be enough, as the symbolic as a
general concept could be interpreted in different ways by different individuals. It could even fail to indicate a consistent type of
“behavior” across countries, lifestyles, and datasets. This could be avoided by using other indicators jointly with the symbolic ones.
For example, a latent variable modeling ecological consciousness could add a more useful meaning to the analysis, in which eco-
conscious and symbolic viewers were associated more strongly with AFVs. Results are useful for both policy makers – who have a
better insight into how specific groups of people sharing common attitudinal characteristics, for example environmental conscious
individuals, may behave nowadays regarding vehicle purchase. Also, it provides an initial insight into who could be potential early
adopters of AFVs and CAVs.
A limitation of the study lies into the fact that we presented the differences between a regular and electric engine to the re-
spondents and used the CC as an approximator of engine power, instead of using the appropriate measurement which is kilowatts. A

223
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

better overall approach would be to replace engine size to horsepower, which can be consistent among the different engines. Using
the approximation, we sacrificed punctuality for simplicity, given that survey respondents are already dealing with new concepts and
mobility forms.
Opportunities for future research include a more precise investigation of family interaction and the dynamics of joint household
decisions. The family interaction could also shed more light on the effect of symbolic vs. practical attitudes and the effect they have
on the final purchase decision. A more complete and in-depth definition of the Symbolic and the Practical could disentangle the
various interpretations of different researchers.

References

Axsen, J., Bailey, J., Castro, M.A., 2015. Preference and lifestyle heterogeneity among potential plug-in electric vehicle buyers. Energy Econ. 50, 190–201. https://doi.
org/10.1016/j.eneco.2015.05.003.
Axsen, J., Kurani, K.S., 2013. Hybrid, plug-in hybrid, or electric-What do car buyers want? Energy Policy 61, 532–543. https://doi.org/10.1016/j.enpol.2013.05.122.
Bahamonde-Birke, F.J., Kickhöfer, B., Heinrichs, D., Kuhnimhof, T., 2018. A systemic view on autonomous vehicles. disP - Plan. Rev. https://doi.org/10.1080/
02513625.2018.1525197.
Bahamonde-Birke, F.J., Kunert, U., Link, H., Ortúzar, J. de D., 2017. About attitudes and perceptions: finding the proper way to consider latent variables in discrete
choice models. Transportation (Amst). https://doi.org/10.1007/s11116-015-9663-5.
Bansal, P., Kockelman, K.M., Singh, A., 2016. Assessing public opinions of and interest in new vehicle technologies: An Austin perspective. Transp. Res. Part C Emerg.
Technol. 67, 1–14. https://doi.org/10.1016/j.trc.2016.01.019.
Ben-Akiva, M., Gershenfeld, S., 2002. Multi-featured products and services: analysing pricing and bundling strategies. J. Forecast. https://doi.org/10.1002/(sici)1099-
131x(199806/07)17:3/4<175::aid-for690>3.3.co;2-e.
Campbell, A.R., Ryley, T., Thring, R., 2012. Identifying the early adopters of alternative fuel vehicles: a case study of Birmingham, United Kingdom. Transp. Res. Part A
Policy Pract. 46, 1318–1327. https://doi.org/10.1016/j.tra.2012.05.004.
Correia, G.H. de A., Looff, E., van Cranenburgh, S., Snelder, M., van Arem, B., 2019. On the impact of vehicle automation on the value of travel time while performing
work and leisure activities in a car: theoretical insights and results from a stated preference survey. Transp. Res. Part A Policy Pract. https://doi.org/10.1016/j.tra.
2018.11.016.
Daziano, R.A., Sarrias, M., Leard, B., 2017. Are consumers willing to pay to let cars drive for them? Analyzing response to autonomous vehicles. Transp. Res. Part C
Emerg. Technol. https://doi.org/10.1016/j.trc.2017.03.003.
Gkartzonikas, C., Gkritza, K., 2019. What have we learned? A review of stated preference and choice studies on autonomous vehicles. Transp. Res. Part C Emerg.
Technol. https://doi.org/10.1016/j.trc.2018.12.003.
Haboucha, C.J., Ishaq, R., Shiftan, Y., 2017. User preferences regarding autonomous vehicles. Transp. Res. Part C Emerg. Technol. https://doi.org/10.1016/j.trc.2017.
01.010.
Hohenberger, C., Spörrle, M., Welpe, I.M., 2016. How and why do men and women differ in their willingness to use automated cars? The influence of emotions across
different age groups. Transp. Res. Part A Policy Pract. https://doi.org/10.1016/j.tra.2016.09.022.
Howard, D., Dai, D., 2014. Public Perceptions of Self-driving Cars: The Case of Berkeley, California. In: 93rd Annu. Meet. Transp. Res. Board 2014 21https://doi.org/
10.1007/978-3-319-54395-6_5.
Jensen, A.F., Cherchi, E., Mabit, S.L., Ortúzar, J. de D., 2017. Predicting the potential market for electric vehicles. Transp. Sci. https://doi.org/10.1287/trsc.2015.
0659.
Krueger, R., Rashidi, T.H., Rose, J.M., 2016. Preferences for shared autonomous vehicles. Transp. Res. Part C Emerg. Technol. https://doi.org/10.1016/j.trc.2016.06.
015.
Kyriakidis, M., Happee, R., De Winter, J.C.F., 2015. Public opinion on automated driving: results of an international questionnaire among 5000 respondents. Transp.
Res. Part F Traffic Psychol. Behav. 32, 127–140. https://doi.org/10.1016/j.trf.2015.04.014.
Larson, P.D., Viáfara, J., Parsons, R.V., Elias, A., 2015. Consumer attitudes about electric cars: pricing analysis and policy implications. Transp. Res. Part A Policy Pract.
69, 299–314. https://doi.org/10.1016/j.tra.2014.09.002.
Liechty, J., Ramaswamy, V., Cohen, S.H., 2003. Choice menus for mass customization: an experimental approach for analyzing customer demand with an application
to a web-based information service. J. Mark. Res. https://doi.org/10.1509/jmkr.38.2.183.18849.
Liljamo, T., Liimatainen, H., Pöllänen, M., 2018. Attitudes and concerns on automated vehicles. Transp. Res. Part F Traff. Psychol. Behav. https://doi.org/10.1016/j.
trf.2018.08.010.
Noppers, E.H., Keizer, K., Bockarjova, M., Steg, L., 2015. The adoption of sustainable innovations: the role of instrumental, environmental, and symbolic attributes for
earlier and later adopters. J. Environ. Psychol. 44, 74–84. https://doi.org/10.1016/j.jenvp.2015.09.002.
Nordhoff, S., De Winter, J., Kyriakidis, M., Van Arem, B., Happee, R., 2018. Acceptance of driverless vehicles: results from a large cross-national questionnaire study. J.
Adv. Transp. https://doi.org/10.1155/2018/5382192.
Nordlund, A., Jansson, J., Westin, K., 2016. New transportation technology: norm activation processes and the intention to switch to an electric/hybrid vehicle,
transportation research procedia. In: Presentation at the Transport Research Arena, Warsaw, Poland, https://doi.org/10.1016/j.trpro.2016.05.334.
Panagiotopoulos, I., Dimitrakopoulos, G., 2018. An empirical investigation on consumers’ intentions towards autonomous driving. Transp. Res. Part C Emerg. Technol.
https://doi.org/10.1016/j.trc.2018.08.013.
Payre, W., Cestac, J., Delhomme, P., 2014. Intention to use a fully automated car: attitudes and a priori acceptability. Transp. Res. Part F Traff. Psychol. Behav. 27,
252–263. https://doi.org/10.1016/j.trf.2014.04.009.
Petschnig, M., Heidenreich, S., Spieth, P., 2014. Innovative alternatives take action – investigating determinants of alternative fuel vehicle adoption. Transp. Res. Part
A Policy Pract. 61, 68–83. https://doi.org/10.1016/j.tra.2014.01.001.
Potoglou, D., Kanaroglou, P.S., 2007. Household demand and willingness to pay for clean vehicles. Transp. Res. Part D Transp. Environ. https://doi.org/10.1016/j.trd.
2007.03.001.
SAE, 2016. Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles. SAE Int. J3016, 1–12. https://doi.org/10.4271/
J3016_201609.
Sanbonmatsu, D.M., Strayer, D.L., Yu, Z., Biondi, F., Cooper, J.M., 2018. Cognitive underpinnings of beliefs and confidence in beliefs about fully automated vehicles.
Transp. Res. Part F Traff. Psychol. Behav. https://doi.org/10.1016/j.trf.2018.02.029.
Schuitema, G., Anable, J., Skippon, S., Kinnear, N., 2013. The role of instrumental, hedonic and symbolic attributes in the intention to adopt electric vehicles. Transp.
Res. Part A Policy Pract. 48, 39–49. https://doi.org/10.1016/j.tra.2012.10.004.
Skippon, S., Garwood, M., 2011. Responses to battery electric vehicles: UK consumer attitudes and attributions of symbolic meaning following direct experience to
reduce psychological distance. Transp. Res. Part D Transp. Environ. 16, 525–531. https://doi.org/10.1016/j.trd.2011.05.005.
Skippon, S.M., 2014. How consumer drivers construe vehicle performance: implications for electric vehicles. Transp. Res. Part F Traff. Psychol. Behav. 23, 15–31.
https://doi.org/10.1016/j.trf.2013.12.008.
Talebian, A., Mishra, S., 2018. Predicting the adoption of connected autonomous vehicles: a new approach based on the theory of diffusion of innovations. Transp. Res.
Part C Emerg. Technol. https://doi.org/10.1016/j.trc.2018.06.005.
Thomopoulos, N., Givoni, M., 2015. The autonomous car—a blessing or a curse for the future of low carbon mobility? An exploration of likely vs. desirable outcomes.
Eur. J. Futur. Res. 3, 14. https://doi.org/10.1007/s40309-015-0071-z.

224
I. Tsouros and A. Polydoropoulou Transportation Research Part A 132 (2020) 214–225

Tsouros, I., Polydoropoulou, A., 2017. Who will buy alternative fueled or automatic vehicles: a modular, behavioral modelling approach. Transportation Research
Board 96th Annual Meeting. Issue: 17-03219.
Wending, B., 2014. Automated vehicle standards and best practices – definitions & taxonomy, 2014 automated vehicles symposium. Presentation at the Automated
Vehicles Symposium, San Francisco, CA.
Yap, M.D., Correia, G., van Arem, B., 2016. Preferences of travellers for using automated vehicles as last mile public transport of multimodal train trips. Transp. Res.
Part A Policy Pract. https://doi.org/10.1016/j.tra.2016.09.003.
Zmud, J., Diaz, F., Lavieri, P., Bhat, C., Pendyala, R., Shiftan, Y., Outwater, M., Lenz, B., 2019. Research to examine behavioral responses to automated vehicles. Road
Vehicle Autom. 5, 53–67. https://doi.org/10.1007/978-3-319-94896-6_5.
Zoepf, S.M., Keith, D.R., 2016. User decision-making and technology choices in the U.S. carsharing market. Transp. Policy 51, 150–157. https://doi.org/10.1016/j.
tranpol.2016.01.010.

225