Electric Vehicle Range Anxiety: An Obstacle for

the Personal Transportation (R)evolution?

Dario Pevec1 , Jurica Babic1 , Arthur Carvalho2 ,
Yashar Ghiassi-Farrokhfal3 , Wolfgang Ketter3,4 , and Vedran Podobnik1
1
University of Zagreb, Faculty of Electrical Engineering and Computing, Croatia
3
Erasmus University, Rotterdam School of Management, Netherlands
2
Miami University, Farmer School of Business, USA
4
University of Cologne, Faculty of Management, Economics, and Social Sciences, Germany

Abstract—Trends in the electromobility industry, increasing increased interest in addressing the aforementioned environ-
research efforts related to alternative fueled vehicles, as well mental concerns [5] through proposing alternative means of
as growing environmental concerns are suggesting that the transportation and alternative fuel vehicles (AFV) [6], [7].
transition from the internal combustion engine technology to
electric vehicles (EV) is necessary and inevitable. To ensure and AFVs are vehicles that do not consume conventional diesel and
enable rapid market penetration of EVs, one major obstacle gas, but are rather fueled by solar power, hydrogen, biodiesel,
needs to be addressed - range anxiety, a fear of running or electricity. This paper mainly focuses on electric vehicles
out of electricity before reaching another available charging (EV) since they are perceived as clean, providing that they are
station. This research employs a survey methodology to assess sourced through renewable sources [8]. Taking into consid-
potential EV owners’ perception of range anxiety with the goal
of quantifying and explaining it through key EV parameters: eration worrying environmental challenges and technological
state of charge (i.e., a relative measure comparing the remaining advancements, together with the interest of many research
amount of energy in the EV battery with the maximum capacity) communities, the transition from internal combustion engine
and remaining range (i.e., how much distance the EV can still vehicles (ICE) to EVs is presumably the next step in the
reach without re-charging). Through the survey analysis, we evolution of private transportation.
answered two relevant research questions that fall into the range
anxiety research agenda: (i) how potential EV owners perceive The most obvious evidence of the efforts towards EV
the optimal distance between charging stations in comparison to popularization is in the corresponding market penetration. Ac-
traditional, well-developed gas station infrastructure; and (ii) how cording to the estimation by the International Energy Agency
key EV parameters influence the decision to charge as well as the [9], sales of new EVs surpassed 1 million in 2017, which
distance one is willing to travel to reach another charging station is a growth of 54% compared to 2016. It is estimated by
that may or may not be occupied. This research is beneficial
for business makers as the knowledge about range anxiety is summarizing cumulative sales from 2005 to 2017 that there
very important for making decisions about charging station are currently more than 3 million EVs on the roads around
placement, as well as for the research community since range the world. The two main reasons driving the EV sales are:
anxiety is a variable that could and should be included in various (i) battery technology advancements, which causes EVs to
research areas centered around EVs. Besides business makers and be cheaper and have larger autonomy (i.e., driving range) as
researchers, this work is beneficial to the society in general as it
may potentially have a positive impact on raising awareness about opposed to the older EVs; and (ii) government EV-related
the necessity of electrification in the transportation industry. incentives such as free charging, lower taxes, and vehicle
Index Terms—electric vehicles, green transportation, charging registration cost reductions [10].
station infrastructure, range anxiety, survey, data science However, despite the growth in the EV sales numbers,
they are still not commonly accepted among wider population
I. I NTRODUCTION of potential consumers. According to Statista Dossier [11],
most countries, besides Sweden, China, and Norway, have on
From the rising seas due to climate change [1] to climate- average less than 1% EV share. This fact shows that despite
related illnesses [2], dealing with today’s environmental con- the increase in EVs’ popularity, rising environmental concerns,
cerns is one of the most prominent challenges societies face. and technology advancement, EVs are still not present on the
One reoccurring factor in the majority of environmental issues road as they perhaps should be. Several studies aim to address
is air pollution, and the transportation sector is known to be such a specific EV adoption challenge. For example, Adnan et
one of the main contributors to greenhouse gas emissions with al. [12] discovered that the most influential factor in making
the largest greenhouse gas footprint [3]. As the fleet of private the decision of purchasing an EV is a phenomenon known as
transportation vehicles is constantly increasing [4], both the range anxiety. Range anxiety is defined as a fear of running
transportation industry and research communities are showing out of electricity before reaching an available charging station
[13]. The range anxiety challenge can be approached from
corresponding author: dario.pevec@fer.hr two different perspectives: (i) increasing the EV autonomy;
and (ii) the development of charging infrastructure. Motivated anxiety is perceived and what are the key factors that evoke
by globally underdeveloped charging station infrastructure, range anxiety. Secondly, we review research that employs a
we focused on the latter approach in a previous work [14]. similar methodology as ours (surveys) in the domain of electric
Concretely, the proposed framework combines heterogeneous vehicles.
data sources, decision maker’s input, and data science to Research by Jung and Steiner [17], as well as the research
recommend a location for placing a new charger(s). Through conducted by Franke and Krems [18], aim to explain how
sensitivity analysis, it was identified that the range anxiety the experience with EVs influences range anxiety. In their
variable could be very valuable in the process of charging research, Jung and Steiner [17] addressed how the EV’s
station deployment modelling, provided that such a variable user interface influences range anxiety. They provided 73 test
can be measured and calculated based on real-world data. participants with a 19 mile EV drive experience. The test
Considering the above-mentioned context, this work iden- scenarios were differentiated by low and high SoC at the
tifies two research questions that are focused on measuring beginning of a drive, and with low or high ambiguity of user
and analysing variables that contribute to range anxiety. The interface. The authors concluded that a low starting SoC, as
first research question is formulated as: “How do potential EV well as a high ambiguity of user interface have significant
owners perceive charging station infrastructure in comparison influence on range anxiety, trust towards the vehicle, and
to the existing gas station infrastructure?”. This research driving behavior.
question can potentially clarify how scarcity of charging Similar to the work by Jung and Steiner, Franke and Krems
stations affects range anxiety, and to what extent the scarcity [18] based their research on participants that were provided
is emphasized when compared to traditional gas stations. The with EVs for 3 months of test driving. The goal of that
second research question is formulated as: ”To what extent do research was to investigate the factors that influence range
different key EV parameters influence the range anxiety among preferences and how they change over time. Franke and Krems
non-EV owners?”. The second research question can answer [18] concluded that range preferences changed during the time
how state of charge (SoC), i.e., remaining battery capacity, of three months, which indicates that owning an EV and
influences the decision to charge the vehicle as well as whether understanding the EV basics is a highly significant factor for
range anxiety is triggered more by the SoC or by the remaining range anxiety.
range that an EV can cover. We addressed both research On the other hand, the work by Rauh et al. [16] is aiming
questions through a survey analysis, which is by the defined to describe range anxiety from the perspective of long-term
scope of this research aimed towards non-EV owners. This EV owners versus non-EV owners. The test scenario is once
design choice is due to non-EV owners having less knowledge again based on driving an EV on a predefined route with
about EVs and, therefore, they manifest a higher level of range predefined SoC. The authors concluded that the range anxiety
anxiety [15]. Besides the higher level of range anxiety, non- is significantly lower for long-term EV owners, although this
EV owners’ perception of the optimal charging infrastructure was induced by specific test parameters, which is an evidence
is important for further infrastructure development since the that range anxiety can be influenced. What was unanswered
(r)evolution of private transportation is highly dependent on was whether range anxiety can be reduced.
their transition to alternative fueled vehicles. In a follow-up Research by Bonges and Lusk [19] and by King et al.
research, we aim at collecting similar data for EV owners, [20] is trying to answer the previously-stated question by
which will enable us to generate even more relevant insights proposing theoretical frameworks. King et al. [20] proposed
about range anxiety as well as to compare preferences and access to on-demand vehicles as a service to alleviate a form
behavior of non-EV users and EV-users. of EV range anxiety. They developed a business model and
The rest of the paper is organized as follows: Section II modelled the price of the proposed business scheme, which in
positions this research against related literature. In Section III, turn can be substantially lower than the cost of accumulated
we describe our research methodology as well as the survey subsidies for EV promotion. Bonges and Lusk [19] proposed
design with emphasis on the explanation of important variables an approach for reducing range anxiety through parking layout
for answering our research questions. Section IV reports our modifications alongside policies and regulations. The main
data analysis, while Section V discusses the results as well as idea is to introduce EV-only parking spots, and that the
observed anomalies alongside explanations. Finally, Section charging station is placed to enable as many vehicles to charge
VI summarizes all the findings throughout this research and as possible, i.e., in the center of a parking lot instead of
present ideas for the future work. corners.
Unlike studies that are focused primarily on range anxiety,
II. L ITERATURE R EVIEW the work by Neuber and Wood [13] is focused on the effects
In this section, we position our research against other rel- of range anxiety and charging infrastructure on EV lifetime
evant research in this domain. Literature review is performed utility. The research was conducted employing computer sim-
from two perspectives: first, research tackling range anxiety, ulation methods using year long trip data for modelling human
i.e., the phenomenon associated with consumers’ perception driving decision behavior. The authors observed that drivers
of the available driving range [16]. This perspective is the with increased range anxiety achieved up to 8% lower utility
driver for the most relevant research in this field - how range factors. The authors also stated that an extension of the existing
public infrastructure would be beneficial for lowering range about the topic, income, or the difference in the settlement
anxiety and increasing EV lifetime utility. hierarchy based on population density. As expected, the vast
Based on the above discussion, it is evident that tackling majority of participants accessed the survey via Facebook, as
range anxiety is a very relevant challenge in the research it is arguably the most influential social network.
community. Unlike previous work in this area, our study The next phase of the methodology is data aggregation,
explores how the electric vehicle range anxiety is perceived a step in which the data collected from different sources are
in contrast to traditional ICE vehicles. More specifically, we combined into a dataset to be further analyzed. Specifically, the
explore how the charging station infrastructure is compared data aggregation step of the methodology simply instantiates
to the gas station infrastructure, as well as how key EV a new survey and stores the collected data. The phase of data
parameters influence potential consumers’ range preferences. processing precedes the final step of the flow, data analysis,
The methodology used in our research is based on data which we describe in Section IV. During the data processing
collection through a survey, which is a commonly used phase, all incomplete or inconsistent answers are removed as
methodology for data gathering in the domain of electric part of the usual data cleansing so that a high-quality dataset is
vehicles. For example, Dorcec et al. [21] and Babic et al. available for the data analysis phase. Besides removing specific
[22] used surveys to assess the willingness to pay for EV records from the dataset, this phase is used for transforming
charging service. Besides the willingness to pay assessment, variables of the dataset so that they follow an appropriate
surveys are also widely used when forecasting EV market format.
penetration (Jensen et al. [23], Lebeau et al. [24]), as well
as for understanding the factors influencing EV adoption B. Survey Design
(Higgins et al. [25], Ko and Hahn [26]). The survey comprises of three thematic sections: (i) demo-
graphic questions; (ii) a questionnaire about the satisfaction
III. R ESEARCH M ETHODOLOGY with existing gas station infrastructure and a comparison with
The methodology employed in this study is motivated by EV charging infrastructure taking into consideration settlement
relevant survey-design work [27] as well as by the well- hierarchies; and (iii) five randomly generated scenarios based
known CRISP-DM methodology, which is widely used in data on which participants should provide opinion about their
modeling and data mining [28]. We illustrate our methodology willingness to charge.
in Figure 1. The following subsections describe the end-to-end The demographic-themed section of the survey is composed
implementation of our methodology and, since we employed of a standard set of questions (e.g., gender, age, country,
a survey approach for collecting data, the structure of the income, employment, and education), and extended with a
questionnaire. couple of EV-specific questions (e.g., questions about EV
ownership, EV model, EV battery capacity, and at what state
of charge (SoC) a person on average decides to charge).
The next group of questions is designed to investigate
potential consumers’ satisfaction with the current state of gas
station infrastructures, as well as the current state of charging
station infrastructures. Throughout this set of questions, we
aim to compare two different “refueling” infrastructures from
the consumers’ comfort perspective by taking into considera-
tion settlement hierarchies.
The last set of questions is related to the range sensitivity
analysis. Through five scenarios, we ask participants two
Fig. 1: Research methodology, from data collection to data questions about their willingness to charge and about the range
analysis preferences in case they want to charge. The exact text of the
question is as follows, where the expressions inside square
brackets are replaced by numbers in real-time after being
A. Data Flow calculated based on randomly-created data for the individual
The survey was created using LimeSurvey [29] and hosted scenarios:
on our server for additional control of access. As can be When full, your EV can achieve maximal distance of
seen in Figure 1, the survey was distributed via multiple [intval((battCap ∗ 1000)/190)] km. Your current state of
communication platforms and social medias in order to maxi- charge (SoC) is [SoC%]. With that SoC you can travel the
mize reach and the diversity of the participants. In particular, maximum of [intval(((battCap ∗ 1000)/190) ∗ (SoC/100))]
we distributed the survey link using various EV-specialized km. Would you like to charge your EV in this circumstance
forums, Facebook, Reddit, and via a word-of-mouth approach. during your daily city commute?
The main reason to distribute the survey through multiple The variable battCap represents the EV battery capacity
different communication platforms is to capture more variance and has a random value between 16 and 60 kWh, i.e., the
in demographics, e.g., the age of the participants, knowledge interval that corresponds with the most popular EV battery
capacity [30]. The variable SoC represents the current EV TABLE I: Participant demographic statistics.
state of charge. The constant 190 is derived from the average Category Subcategory % of Participants [N = 134]
range of the most popular EV [30]. The main difference in Employed 67.5
Students 29.0
the five scenarios we present to survey participants is in the Working status
Retired 3.5
SoC value. In the first scenario, that value is between 5% and Unemployed 0.0
100%, i.e., it is enough to cover most of the real SoC cases. Very familiar 45.0
EV Knowledge Know something 45.0
In every following scenario, that value is lowered by 20% to Heard of 10.0
ensure the existence of a scenario that would potentially create Have 85.0
Driving licence
range anxiety. The maximal distance is calculated based on the Not have 15.0
average energy consumption per km for the most popular EVs Male 70.0
Gender
Female 30.0
today [30]. If the participant answered the question about a
hypothetical charging scenario affirmatively (i.e., (s)he wants
to charge), (s)he is then prompted with the following additional
question to express the maximal range (s)he is willing to travel The demographic characteristics of the participants are
to reach a charging station: inline with the demographics of people interested in EVs [32].
What is the acceptable additional distance (in km, 1 km In particular, most of the participants have higher education
= 0.62 miles) to travel to the charging station which may and are employed. Furthermore, EV specialized forums have
or may not be occupied, taking into consideration the time a larger percentage of males, which is also consistent with
that is needed to cover that distance? (Example: for 10 km our demographic data since the majority of the participants
in Europe, average time is 25-35 minutes.) In this situation, in our survey are male (70%). As a consequence of using the
you must carefully consider your answer since you are either Facebook social network, most (more than 80%) of our sample
unsure how far the next charging station will be or whether is between 20 and 40 years old, which corresponds with the
the next station will be unoccupied! age of the majority of Facebook users. The age and gender
The question about the range a person is willing to travel to distribution of participants can be observed in Figure 2. The
reach a charging station, which may or may not be occupied, rest of the participant demographic statistics are described in
specifically asks about how one perceives distance, which is Table I.
the reason we emphasize the words additional distance, time It is important to note that the majority of the participants
needed for greater distances, which is calculated from the (around 90%) have at least some basic knowledge about EVs,
average driving speed in populated area, and the fact that while more than 80% have a driving licence and experience
the charger may be unavailable. It is important to note that with driving a car. Driving experience is a variable of sig-
the same survey participant potentially answered this question nificant importance for this survey since the quality of our
several times, i.e., if (s)he answered affirmatively to more results is highly dependent on the correct approximation of
than one question from the set of five hypothetical charging the distance between existing gas stations and the desired
scenarios. distance between charging stations, which in turn implicitly
relies on knowledge about driving cars. The majority of the
IV. S URVEY R ESULTS participants are from Croatia (three quarters of the dataset),
The previously described survey was distributed to 170 par- where the charging infrastructure is underdeveloped and EVs
ticipants through different channels. None of the participants are not a popular option for private transportation. The last
owned EVs, but they were nonetheless interested in or have at quarter of the dataset is comprised of participants from 14
least basic knowledge about EVs. After the data pre-processing different countries, including Finland, Netherlands, Germany,
phase, 134 survey responses were considered valid, and the USA, and the United Kingdom, where EVs have significant
rest (36) were discarded due to one of the following three rea- market penetration. This distribution of countries is especially
sons: (i) incomplete answers where participants left the survey convenient for our research goals since the participants from
before finishing; (ii) inconsistent answers that were manifested the countries with underdeveloped infrastructure are arguably
through extremely different answers for very similar scenarios, more likely to be under the influence of range anxiety.
e.g., for the battery capacity that differs in less than 5 kWh and The first research question, formed as “How do poten-
SoC that results in less than 10 km difference in the remaining tial EV owners perceive charging station infrastructure in
range, some participants answered in one scenario with the comparison to existing gas station infrastructure?”, led us
lowest possible value, and with the highest possible value in to some interesting findings about the potential EV owners’
the other scenario; and (iii) trivial answers where participants preferences. As can be seen in Figure 3a, potential consumers’
always provided the same answers regardless of the differences preferences regarding gas stations are that there are currently
in scenarios. Besides automatically detected invalid answers, too many gas stations, and that the distance between them can
some of the outliers, that have significant deviation comparing be greater than it actually is. In Figure 3b, we can notice that
to the mean values, were also detected and discarded from the potential consumers are not satisfied with the current charging
final dataset, which is common procedure for noise removal station infrastructure, and they would prefer to have charging
[31]. stations deployed in a manner more similar to the existing gas
 TABLE II: Range preferences in comparison to settlement
hierarchy.
Mean Min Max Percentage
Settlement
Distance Distance Distance of
Hierarchy
[km] [km] [km] Participants
Village 8.44 2 20 6
Town 10.60 1 30 13
Large Town 8.01 1 20 7
City 6.30 1 20 12
Large City 6.00 0,5 40 52
Metropolis 4.10 1 10 10

range preferences in comparison to settlement hierarchy can

be observed in the Table II.
We attribute the previous results to the fact that refueling
infrastructure is often well-developed in bigger areas. That is,
gas stations are closer to one another in locations where there
is a higher demand for them, such as in the case of big cities.
Fig. 2: Age and gender of survey participants. Consequently, consumers are used to higher availability of
refueling stations and expect the similar service with charging
stations. A more detailed overview of the preferred distance
station infrastructure deployment, since they are accustomed to charging stations with respect to the settlement hierarchy is
to traditional gas station infrastructure planning. The resulting depicted in Figure 4.
mean distances can be observed as dashed lines in Figure 3a Based on the last part of the survey, i.e., the five different
and Figure 3b. state of charge and driving range scenarios, we can answer
Our results further show that more than 20% of the partic- our second research question: “To what extent do different
ipants prefer charging stations more closely distributed than key EV parameters influence range anxiety between non-
the gas stations are, while 50% of the participants are satisfied EV owners?”. The observed charging decision patterns (see
with the distances between existing gas station infrastructure, Figure 5) clearly indicate that the vast majority of participants
and prefer that exact distance mirrored on the charging station would like to charge when their SoC is under 30%, which
infrastructure, according to the compared answers for desired roughly corresponds to the standard low battery warning at
distances between gas stations and charging stations. On around 20% [33]. The highest concentration of scenarios
average, survey participants stated that the distance between where the participants are willing to charge is where SoC
charging stations should be 0.12 km lower than the distance is below 15% and the remaining driving range is below 50
between gas stations. It is important to mention that we did km. Figure 5 depicts that not all participants who are on
not attempt to compare existing and desired charging station less than 10% battery capacity want to charge, as well as
infrastructure since charging infrastructure might be heavily the fact that few participants who are above 90% want to
underdeveloped in some countries and, consequently, some charge. This phenomenon can be explained through different
participants might not be aware of the distances between refuelling habits, place of residence, and a poor perception
chargers. of distance expressed in kilometers. The observed behavior of
Besides the preferences about the existing and future in- outliers can also be explained through the settlement hierarchy.
frastructure, it is important to observe how the size of the Potential consumers who live in smaller areas, in houses with
living area influences the preferences about the distance be- an opportunity to install a private charger, can charge whenever
tween charging stations. Before asking the survey participants they would like to, which can result in a decision to charge
questions about preferred distances, we asked about their even on 80% of remaining SoC.
settlement hierarchy, i.e., whether they live in a village, (large) Our final research goal was to determine how SoC and
town, (large) city, or metropolis. We notice certain patterns the remaining driving range influence the preferred range
regarding the relationship between the settlement hierarchy to a charging station when the desire to charge is present.
and the desired distance to a charging station. In particular, Since we aim to observe how each participant would behave
larger settlements with high population density have a negative in different situations, each participant answered five times
impact on potential EV consumers’ desired charging station if they are willing to charge and, if they are, then how
distance. As depicted in Figure 4, in the settlement hierarchy far they would like to travel to charge. The aforementioned
with the highest population density - metropolis, potential resulted in multiple answers about range preferences from one
EV consumers tend to prefer distances of less than 5 km participant, and multiple answers from the same participant
between charging stations, with lower variance compared to should not be considered as independent as they are rather
the other levels of hierarchy settlement that have significant inter-dependent variables. Therefore, to assess the influence
deviations from the mean value. A more detailed overview of of two key variables on range preferences, we must use mixed
 (a) (b)
Fig. 3: Comparison of infrastructure preferences: (a) current and desired gas stations distances; (b) desired gas vs. charging
stations distances

Fig. 4: Consumers’ preferences regarding the distance between Fig. 5: Charging decision with different SoC and driving range.
charging stations grouped according to settlement hierarchy.

different for each participant, thus making his or her responses

effect analysis where we add a random effect to account dependent on one another, while still being independent with
for individual participant variation. We formed our regression respect to the responses of other participants. The results of
formula, using standard R notation as follows: our regression model are presented in Table III.

distance to travel ∼SoC + driving range+

(1) TABLE III: Random effect regression model: distance to the
+ (1|participant id) + 
charging station as a function of SoC and remaining range.
In Equation 1, SoC and driving range represent indepen- Random effects:
dent variables, i.e., variables with predictable influence on Groups Name Std.Dev.
a target variable, while (1|participant id) represents the participant id (Intercept) 13.27
Residual 14.73
random effect, i.e., another independent variable that captures Fixed effects:
idiosyncratic variation that is due to individual differences. (Intercept) SoC driving range
That random effect is interpreted as an intercept that is -1.84 0.08 0.22
 The results of the developed model point to the following than the corresponding mean, they nonetheless could not be
conclusions: for each unit increase in SoC, the distance partic- considered outliers. This is natural since the mean distance-
ipant is willing to travel will increase on average by 0.08km to-charger value was primarily dictated by the majority of
when everything else is held constant. Similarly, with every the participants who are from big cities and, as discussed in
unit increase of driving range, that distance will increase on Section IV, participants from smaller settlements are expected
average by 0.22 km when everything else is held constant. to drive greater distances between charging stations and,
To assess the importance of each variable, we compared our hence, affect mean distance values.
model containing both variables with a model that contains The survey was distributed strategically to attract a diverse
only one of the variables, i.e., a baseline model, using the set of participants targeting mainly non-EV owners. We have
ANOVA test. According to the tests, the SoC variable is not focused on these population because we believe that many EV
significant for the model, while the driving range is, with p- owners have a charger installed at home, which can drastically
value of 4.449e − 10. One reason for those results lies in the affect range anxiety. For example, by being able to charge
fact that those variables are partially related, i.e., lower SoC an EV at home, the scenarios where SoC is less than 20%
means lower driving range. The previously described model become unrealistic. Nonetheless, as future work, we do plan
states that if a participant owns, say, a Nissan Leaf that has to collect opinions from EV owners since some of them might
battery capacity of 30 kWh, i.e., when full this EV can achieve not have the possibility to charge at home, e.g., EV owners
a driving range of 160 km, and that battery is currently on who live in residential buildings in large cities. This would
15% SoC, i.e., 24 km remaining range, then on average a allow us to measure how the perceived range anxiety of EV-
participant will be willing to travel for additional 5.2 km to owners contrast to non-EV owners. Besides the inclusion of
reach a charging station regardless of whether it is occupied. EV owners, there is also some room for improvements in our
From the results depicted in Figure 5 and in Table III, it is experiments by removing certain survey limitations, e.g., by
obvious that potential EV owners are more influenced by the targeting more participants from a specific settlement. Results
remaining range than the SoC regarding the distance they are about the influence of the population and size of a settlement
willing to travel to reach available charging stations. This is on range anxiety would then be more precise with a balanced
the result we expected since at the same state of charge, the number of participants from each hierarchy level.
remaining driving range can differ for up to 100 km depending
on the EV’s battery capacity. For example, Mitsubishi MiEV VI. C ONCLUSION AND F UTURE W ORK
can cover 26 km on a 30% SoC, while Nissan Leaf and Tesla By providing insights into range anxiety, this research also
S 85 can cover, respectively, 48 km and more than 100 km on provides insights into the potential EV owners’ preferences
the same SoC. As for the decision to charge, both variables regarding charging station infrastructure. In particular, our
are strong indicators that a participant is most likely willing results allow one to determine how far charging stations should
to charge (90% of the cases) when the remaining SoC is less be from each other so as to reduce range anxiety. This in
than 30% or when the remaining range is less than 75 km. turn can, hopefully, lead to more potential consumers deciding
in favor of purchasing an EV, instead of a traditional ICE
V. D ISCUSSION vehicle. Taking the aforementioned into account, we argue that
According to our survey analysis, people are more influ- this research covers the range anxiety phenomenon from three
enced by the remaining driving range than the state of charge different perspectives - people, profit, and environment.
when they have to decide how far they are willing to drive Using the data collected via the survey described in Section
to reach another charging station that may or may not be III, we are able to answer both of our research questions
occupied. This is a positive result since most electric vehicles introduced in Section I. In particular, potential EV owners
currently provide an accurate prediction of the remaining driv- do not require a charging station infrastructure as dense as
ing range. However, state of charge is an important variable the current gas station infrastructure. Moreover, we are able
in the decision to charge, as our results indicate that people to infer that there currently are more gas stations than it is
are uncomfortable when their SoC is below 30%. needed. However, the survey respondents seem accustomed
One can argue that the inconsistent distance perception to the high availability of existing gas stations, and they
exhibited by the participants is one limitation with our survey seem to prefer the distance between two charging stations to
approach. Indeed, some of the participants reported as accept- be less than 5 km. As for the influence of key EV-related
able distances between chargers greater than 50 km. Some also variables, SoC and driving range, we draw the following
reported that they are willing to drive all of their remaining conclusions: (i) SoC has a more meaningful impact on making
range to reach a charging station that may not be available. the decision about whether to charge; and (ii) the driving
This observed behavior is expected since some participants range demonstrates higher influence on the decision about the
are more risk-seeking than others. To solve this issue, as distance one is willing to travel to reach a charging station.
previously mentioned, a few of the aforementioned answers For future work, we plan to expand the survey participant
were considered outliers based on their deviation from the pool with targeted audience, e.g., by including EV owners
mean value and, hence, removed from our analysis. However, in order to compare what constitutes (and what the effects
although some of the answers have significantly greater value are of) range anxiety from the perspectives of both EV
owners and those who do not (yet) own an EV since, as we [13] J. Neubauer and E. Wood, “The impact of range anxiety and home,
discussed in Section V, EV owners have, as a consequence workplace, and public charging infrastructure on simulated battery
electric vehicle lifetime utility,” Journal of power sources, vol. 257,
of experience, different understanding of key EV parameters. pp. 12–20, 2014.
We also plan on incorporating the results presented in this [14] D. Pevec, J. Babic, M. A. Kayser, A. Carvalho, Y. Ghiassi-Farrokhfal,
paper on range anxiety modelling into a broader research on and V. Podobnik, “A data-driven statistical approach for extending
electric vehicle charging infrastructure,” International journal of energy
the development of a framework for extending the existing research, vol. 42, no. 9, pp. 3102–3120, 2018.
EV charging infrastructure, partially described in the work by [15] P. Plötz, U. Schneider, J. Globisch, and E. Dütschke, “Who will buy
Pevec et al. [14]. This step is crucial to unlock smart markets electric vehicles? identifying early adopters in germany,” Transportation
Research Part A: Policy and Practice, vol. 67, pp. 96–109, 2014.
[34] in the context of the EV ecosystems. Understanding [16] N. Rauh, T. Franke, and J. F. Krems, “Understanding the impact of
and being able to quantify the range anxiety phenomenon is electric vehicle driving experience on range anxiety,” Human factors,
not only an extremely important prerequisite for realistically vol. 57, no. 1, pp. 177–187, 2015.
[17] M. F. Jung, D. Sirkin, T. M. Gür, and M. Steinert, “Displayed uncertainty
modelling charging station infrastructure planning, but also improves driving experience and behavior: The case of range anxiety in
a significant component for faster and sustainable EV accep- an electric car,” in Proceedings of the 33rd Annual ACM Conference on
tance in general. Human Factors in Computing Systems. ACM, 2015, pp. 2201–2210.
[18] T. Franke and J. F. Krems, “What drives range preferences in electric
vehicle users?” Transport Policy, vol. 30, pp. 56–62, 2013.
ACKNOWLEDGEMENTS [19] H. A. Bonges III and A. C. Lusk, “Addressing electric vehicle (ev)
sales and range anxiety through parking layout, policy and regulation,”
The authors acknowledge the support of the Croatian Sci- Transportation Research Part A: Policy and Practice, vol. 83, pp. 63–73,
ence Foundation under the grant DOK-2015-10-1777. This 2016.
[20] C. King, W. Griggs, F. Wirth, K. Quinn, and R. Shorten, “Alleviating
research has also been partly supported by the European a form of electric vehicle range anxiety through on-demand vehicle
Regional Development Fund under grants KK.01.1.1.01.0009 access,” International Journal of Control, vol. 88, no. 4, pp. 717–728,
(DATACROSS), KK.01.2.1.01.0020 (RASCO-FER-SMART- 2015.
[21] L. Dorcec, D. Pevec, H. Vdovic, J. Babic, and V. Podobnik, “How
EV) and KK.01.2.1.01.0077 (bigEVdata). do people value electric vehicle charging service? a gamified survey
approach,” Journal of Cleaner Production, vol. 210, pp. 887–897, 2019.
R EFERENCES [22] J. Babic, A. Carvalho, W. Ketter, and V. Podobnik, “Modelling electric
vehicle owners willingness to pay for a charging service,” in Proceedings
[1] R. Quinn, S. Moseman-Valtierra, K. Kroeger, R. Martin, O. Abdul-Aziz, of the Erasmus Energy Forum, 2016, pp. 1–8.
K. Ishtiaq, E. Brannon, K. Egan, and J. Tang, “The coastal squeeze: [23] A. F. Jensen, E. Cherchi, S. L. Mabit, and J. d. D. Ortúzar, “Predicting
Rising seas and upland plant invasions differentially affect vertical the potential market for electric vehicles,” Transportation Science, 2016.
exchange of greenhouse gases,” in American Geophysical Union, Ocean [24] K. Lebeau, J. Van Mierlo, P. Lebeau, O. Mairesse, and C. Macharis,
Sciences Meeting 2016, abstract# EC14B-0974, 2016. “The market potential for plug-in hybrid and battery electric vehicles
[2] J. Paavola, “Health impacts of climate change and health and social in flanders: A choice-based conjoint analysis,” Transportation Research
inequalities in the uk,” Environmental Health, vol. 16, no. 1, p. 113, Part D: Transport and Environment, vol. 17, no. 8, pp. 592–597, 2012.
2017. [25] A. Higgins, P. Paevere, J. Gardner, and G. Quezada, “Combining
[3] Y. Van Fan, S. Perry, J. J. Klemeš, and C. T. Lee, “A review on air choice modelling and multi-criteria analysis for technology diffusion: An
emissions assessment: Transportation,” Journal of cleaner production, application to the uptake of electric vehicles,” Technological Forecasting
vol. 194, pp. 673–684, 2018. and Social Change, vol. 79, no. 8, pp. 1399–1412, 2012.
[4] “Progress of eu transport sector towards its environ- [26] W. Ko and T.-K. Hahn, “Analysis of consumer preferences for electric
ment and climate objectives,” Jan 2019, available at vehicles,” IEEE Transactions on Smart Grid, vol. 4, no. 1, pp. 437–442,
https://www.eea.europa.eu/themes/transport/term/term-briefing-2018. 2013.
[5] A. Akakpo, E. A. Gyasi, B. Oduro, and S. Akpabot, “Foresight, [27] M. P. Couper, M. W. Traugott, and M. J. Lamias, “Web survey design
organization policies and management strategies in electric vehicle and administration,” Public opinion quarterly, vol. 65, no. 2, pp. 230–
technology advances at tesla,” in Futures Thinking and Organizational 253, 2001.
Policy. Springer, 2019, pp. 57–69. [28] R. Wirth and J. Hipp, “Crisp-dm: Towards a standard process model
[6] K. Valogianni, W. Ketter, J. Collins, and D. Zhdanov, “Facilitating a for data mining,” in Proceedings of the 4th international conference
sustainable electric vehicle transition through consumer utility driven on the practical applications of knowledge discovery and data mining.
pricing,” 2018. Citeseer, 2000, pp. 29–39.
[7] M. T. Kahlen, W. Ketter, and J. van Dalen, “Electric vehicle virtual [29] C. Schmitz et al., “Limesurvey: An open source survey tool,” LimeSur-
power plant dilemma: Grid balancing versus customer mobility,” Pro- vey Project Hamburg, Germany. URL http://www. limesurvey. org, 2012.
duction and Operations Management, vol. 27, no. 11, pp. 2054–2070, [30] A. Ajanovic and R. Haas, “Electric vehicles: solution or new problem?”
2018. Environment, Development and Sustainability, vol. 20, no. 1, pp. 7–22,
[8] W. Ketter, M. Peters, J. Collins, and A. Gupta, “A multiagent competitive 2018.
gaming platform to address societal challenges,” Mis Quarterly, vol. 40, [31] V. Hodge and J. Austin, “A survey of outlier detection methodologies,”
no. 2, pp. 447–460, 2016. Artificial intelligence review, vol. 22, no. 2, pp. 85–126, 2004.
[9] T. Bunsen, P. Cazzola, M. Gorner, L. Paoli, S. Scheffer, R. Schuitmaker, [32] M. K. Hidrue, G. R. Parsons, W. Kempton, and M. P. Gardner,
J. Tattini, and J. Teter, “Global ev outlook 2018: Towards cross-modal “Willingness to pay for electric vehicles and their attributes,” Resource
electrification,” International Energy Agency, 2018. and Energy Economics, vol. 33, no. 3, pp. 686–705, 2011.
[10] K. Y. Bjerkan, T. E. Nørbech, and M. E. Nordtømme, “Incentives for [33] N. Leaf, “2015 leaf owner’s manual.” [Online]. Available:
promoting battery electric vehicle (bev) adoption in norway,” Trans- https://owners.nissanusa.com/content/techpub/ManualsAndGuides/LEAF
portation Research Part D: Transport and Environment, vol. 43, pp. /2015/2015-LEAF-owner-manual.pdf
169–180, 2016. [34] M. Bichler, A. Gupta, and W. Ketter, “Research commentarydesigning
[11] Statista. (2018, may) Electric vehicles worldwide. Available at smart markets,” Information Systems Research, vol. 21, no. 4, pp. 688–
https://www.statista.com/study/11578/electric-vehicles-statista-dossier/. 699, 2010.
[12] N. Adnan, S. M. Nordin, I. Rahman, P. M. Vasant, and A. Noor, “A
comprehensive review on theoretical framework-based electric vehicle
consumer adoption research,” International Journal of Energy Research,
vol. 41, no. 3, pp. 317–335, 2017.