Confidential
Accelerating the world’s transition to sustainable transportation:
An exploratory study of what consumers want when buying an electric vehicle.
by Martin Grebhahn
January 2015
Accelerating the world’s transition to sustainable transportation:
An exploratory study of what consumers want when buying an electric vehicle.
by Martin Grebhahn
January 2015
Submitted to the:
Faculty of Economics and Business, Marketing University of Groningen Supervisor: Dr. Felix Eggers
Second Supervisor: prof. dr. J.E. Jaap Wieringa Data of the Author:
Student Number S2402181
Address: Amsteldijk 39.2, 1074 HV Amsterdam, The Netherlands Email: martin@grebhahn.de | Phone: 0031 682 008 282
“The future belongs to those who prepare for it today.”
Malcolm X
for Beate & Manfred
Executive Summary
Electric vehicles are often seen as the future of transportation by many researchers and practitioners. This research supports this notion by testing 2014, state-of-the-art electric vehicle (EV) adoption potential. Former research failed to compile statistical models that (1) can predict the adoption potential of electric vehicles in the coming years and (2) use consumer relevant attributes that manage to replicate most of the utility attached to an EV. The present paper follows and explorative approached by considering former research, latest press articles and a database of over 10,000 EV owners in order to support prior literature and to find new influencing attributes. 6 out of 7 considered attributes showed a significant impact on consumer’s choice behavior, namely charging costs, range, fast charging network density, performance, charging time and financial solution. Interestingly, charging costs have the greatest influence followed by range and fast charging network density. These findings point on the one side to car manufacturers that must further develop their cars to meet consumer preferences in terms of range and charging equipment, but on the other side they point at charging network suppliers to increase the density and speed of their charging stations.
Prior research was confirmed and updated by the conducted conjoint analysis. All attributes found in prior literature showed a similar behavior. Though, the present paper updates attribute levels to more realistic ones and finds new influencing attributes.
Further analysis showed, that the customer characteristic risk aversion significantly impacts the probability of EV adoption. Respondents that score high on risk aversion, meaning that they do not like to take risks, make their choices even more dependent on charging costs and range as compared to people that like to take risks.
The implication is that consumers must be educated about existing charging infrastructure and about latest developments in the electric car market to minimize the negative impact of risk aversion and to maximize EV adoption. This finding extends prior EV-adoption literature. The construct of risk aversion was not considered in the context of EV adoption, yet. The present paper calls for a inclusion of this construct in future EV adoption studies to strengthen results and explain differences between consumers.
The data additionally unveiled the existence of 4 customer segments. These segments place varying importance to the different attributes. E.g. one segment determines over 50% of their purchase decision on the attached costs and other segments only purchase if the fast charging network is in place. These differences call for unique strategies to target those segments as well as for focused future development plans to meet the consumers’ needs. One important conclusion is that there is only one segment that makes their decision mostly dependent on the range of the car. The other 73% of consumers have other top priorities.
This provides clarification to the often-mentioned range anxiety. It exists and the range must be in an acceptable frame, but most consumers make their decision in favor or against an EV dependent on other attributes. That is why the EV’s success is not solely dependent on the range.
Additional managerial implications form the end of the study and provide further insights. Because of the complexity of the study and the utilized statistical methods, limitations are attached. These lead to suggestions for future research.
Table of Contents
List of Tables and Graphics 7
List of Abbreviations 8
1 Introduction 1
2 Attributes Increasing Adoption Probability of Electric Vehicles 4
2.1 Method to Select Attributes 5
2.1.1 Prior Research 6
2.1.2 Exploratory – Leading Market Offering 9
2.1.3 Exploratory – Press and Article Review 10
2.1.3 Database Analysis of EV Owners 12
2.2 Conceptual Framework 13
3 Research Design and Method 14
3.1 Eliciting Consumer Preferences 14
3.2 Study Design 14
3.3 Experimental Design - Choice Design & Choice elicitation 15
3.4 The Model 16
4 Results 18
4.1 Sample Description and Procedure 18
4.2 Estimations 19
4.2.1 Aggregate-Level Analysis 19
4.2.2 Moderation Analysis 28
4.2.3 Risk Aversion Analysis 28
4.2.4 Latent Class Analysis 29
5 Discussion 35
6 Managerial Implications 37
7 Limitations 38
References 39
List of Appendices 45
Appendices 46!
List of Tables and Graphics
Graphics
Graphic 1 Conceptual Framework 13
Graphic 2 Choice-set with no-choice question 16
Graphic 3 Gender Distribution 18
Graphic 4 Age Distribution 18
Graphic 5 Attribute Level Utility Estimations 19
Graphic 6 RA interaction with Range 29
Graphic 7 RA interaction with Charging Costs 29
Graphic 8 Visualization of Information Criteria 30
Tables
Table 1 Tesla Model S competition in the US market 5
Table 2 Summary of Prior Research 6
Table 3 Car usage data 8
Table 4 Consumer characteristics 8
Table 5 Tesla Model S Value Proposition 9
Table 6 Attributes and Levels 15
Table 7 Attribute Characteristics 20
Table 8 Aggregated-Model Estimation Results 21
Table 9 Goodness of Fit Tests Results 22
Table 10 Holdout Sample Results 23
Table 11 Attribute’s Range, Importance and Rank 23
Table 12 Model S Utility 25
Table 13 Raising Charging Costs Utility Scenario 26
Table 14 Higher Charging Network Density Scenario 27
Table 15 Future EVs Utility Scenario 27
Table 16 Risk Aversion Grouped Model 28
Table 17 Segment Estimation - Information Criteria 29
Table 18 Model of 4 Segment Solution 31
Table 19 Covariates of 4 Segment Model 32
Table 20 Goodness of Fit Tests Results 32
Table 21 Segment Overview 33
List of Abbreviations
AC Alternating current
ß Beta
CBC Choice-based conjoint
DC Direct current
EV/BEV (Battery) Electric Vehicle powered solely by an onboard battery
h Hour
hp or HP Horsepower
Hybrid/HEV Hybrid Electric Vehicle - car with an electric and conventional internal combustion engine drivetrain; batteries are solely charged by regenerative braking or directly by the ICE ICE Internal Combustion Engine Vehicle running on gasoline or diesel
km Kilometer
kWh Kilowatt hour
l Liter
LL Log Likelihood
min Minutes
nPar Number of Parameters; Degrees of Freedom
PHEV Plug-in Hybrid Electric Vehicle - same as HEV with the addition that the batteries can be charged via an external power source
RA Risk Aversion
REEV Range Extended Electric Vehicle powered by an electric drive train powered by a battery with the addition of an internal combustion engine used as generator to charge the battery in case of power shortage
s Seconds
1 Introduction
Malcolm X designates the future to the individuals who prepare for it today. The automotive company Tesla Motors follows this call by preparing the future of transportation. Its mission is to “accelerating the world’s transmission to sustainable transportation” (Tesla Motors, 2014). The reason why this transmission in transportation for humanity’s future is necessary is clear: Greenhouse gas emissions, mostly CO2 (84% of greenhouse gases) are steadily rising. The average increase between 1990 and 2000 was 1% (or more) annually (EPA, 2004) and rose to 2.9% of annual increase between 2001 and 2011 (Olivier et al., 2013).
International efforts e.g. the Kyoto protocol did not achieve the necessary change in the amount of greenhouse gas emissions (EPA, 2004; RIVM, 2004). 2012 was the first year with a smaller increase of CO2 emissions since 1990 with a 1.1% increase. That could be a first effect of the increased use of sustainable energy sources, which account for a fraction of 2.4% of worldwide energy production in 2012 (Olivier et al., 2013).
90% of CO2 and therefore greenhouse gas is caused by the combustion of fossil fuels (EDGAR 4.2, JRC/PBL, 2011; EPA, 2004). An essential part of this is due to the possession and increased use of, next to others, combustion engine cars (Steg, 1999). E.g. in the US, the combustion engines in the transport sector alone account for 28% of environmentally damaging greenhouse gases, making it the second largest contributor after the conventional electricity sector (US Environmental Protection Agency, 2012). The reduction of these emissions heavily depend on the adoption of alternative powered cars, e.g. on the adoption of EVs by the consumer (Noppers et al., 2014). Hardester (2010) rightly stated that even when the EV is not emitting any exhaust fumes, the power sources used to charge the vehicle does impact the environment with its emissions. An often-used reply is, that the energy to charge EVs could be supplied by alternative sources such as wind, solar, geothermal and others (Mangram, 2012). The superchargers from Tesla Motors live up to the “sustainable pledge” by delivering green energy to the cars in over 90% of the cases (Tesla Motors, 2014). But even when consumer charge on public charging stations or at home using conventional energy obtain from gas or coal plants, EVs use the charged energy six times more efficiently than ICE and subsequently produce ten times less emissions (Eberhard & Tarpenning, 2006). Therefore, the adoption of EVs is in any case the cleaner, environmentally friendlier choice.
Researchers as well as practitioners believe this notion. For example, Eggers & Eggers (2011) consider all- electric vehicles the automobile technology of the future, which is confirmed and extended by Schuitema et al. 2013, who state that EVs have the highest potential to transform the current automotive transportation system into a sustainable one. They have even given an outlook and attached the potential of energy storing systems in a smart grid solution to current and future EVs. Despite the described great potential of the EV to be the long-term solution to decelerate greenhouse gas emissions, thus climate change, current estimations of the EV market share are at a disillusioning 10% by 2020 (McKinsey & Company, 2013). Hope that these estimations are wrong stems from practitioners like the Tesla Motors Vice President of Sales and Ownership Experience George Blankenship. He stated at the event of Motor Trend 2013 “Car of the Year Award”, which the fully electric Tesla Model S received, that the car industry is at a turning point in history towards a sustainable transportation (Motor Trend, 2013). Two years before, Google (2011) reported in its “Impact of clean energy innovation” paper that battery breakthroughs allowing for 300 km range on a single charge will be the catalyst propelling the BEV to market shares of over 30% in the total automobile industry. Google’s analysis and Motor Trend’s award underline that the mentioned turning point marks the end of ICE and the starting point for an electric future.
Tesla is not the first company who believes to mark the starting point for EVs. The Electric Car Company (ECC) believed to have reached that same point already in 1897, when starting to “mass” produce electric taxicabs in major US cities. The taxis run on rechargeable batteries (Schuitema et al. 2013) and the ECC assumed that the EV turns out to be the dominant automobile type (Rae, 1955). This assumption was based on a large set of patents on battery technology, which had the potential that the ECC could monopolize the
missed the opportunity of refining its products. Since oil was cheap and the ECC busy with buying, merging and restructuring, the gasoline powered car gained popularity and took over the EV, first in performance and then in terms of market share. Finally, the ECC went bankrupt and the EV disappeared in the drawers of the developers, concentrating on improving ICE technology (Rae, 1955).
In 2014, when Tesla Motors found itself in a comparable situation concerning the patents, it took the exact opposite direction of the ECC. In June 2014, it opened all patents to motivate competition to join the electric revolution (Vance, 2014). The patents enable companies to build competitive EVs with characteristics like long range, superior driving performance, short charging times and even conditionally grants the access to Tesla’s Supercharger Network1 to other manufacturers (Vance, 2014). That is why Tesla believes that a turning point for the whole automotive category is reached. This belief is also based on the success in the car segment, Tesla positioned its Model S. Not only in its country of origin, the United States, where it is outselling established competition (Table 1), but also in oversea markets, e.g. Norway, where the Model S is one of the top selling cars (Dagenborg, J. 2013; Tesla Motors, 2014a). All above evidences point to the fact that Tesla Motors is the first car company that managed to compile a value package that could challenge the ICE in the long run.
In contrast, current research on EV adoption mentioned above attaches mostly drawbacks and limited advantages to the possession of an EV (Mangram, 2012). Frequently, limited range, constrained driving performance, little usability, extended charging times and other shortcomings are emphasized (e.g. Schuitema et al. 2013; Noppers et al., 2014; Table 2). The present research aims at clarifying the apparent contradiction between the success of Model S and theory. Therefore, this research focuses on existing research and explores future technological enhancements in the following years. The aim is to define what value package consumers appreciate the most in order to give an indication on which attributes must be emphasized in the development and marketing of electric vehicles. This enables car manufacturers to tailor their strategy to increase the success-probability in future EV market. Additionally, it adds new attributes, levels and underlying constructs to existing literature to better predict future EV adoption.
Are competitive EVs preferred over ICE cars?
Which attributes of an EV that are found in prior research determine customer adoption?
Which attributes do consumers value that are not yet researched?
How do these attributes determine adoption?
Having a closer look, most research missed out on testing attribute and corresponding levels matching the value proposition of 2014 and future EVs (Mangram, 2012). A value proposition is a summary of how a product delivers value to its consumer and it especially includes unique points of the product stressing the difference or superiority over competitive offers (Anderson et al. 2009). The reviewed papers are not able to explain why the e.g. the EV “Tesla Model S” with its superior driving performance (e.g. faster acceleration as comparable ICE), long-range capabilities, free long-distance travel (facilitated by superchargers;
unlimited, free-of-charge, long-distance travel), price and comfort competitive to an ICE, zero emissions (one must charged from clean energy; >90% superchargers running on green energy), increased reliability (little wear of mechanical components due to 10 times fewer moving parts and simple engine construction as compared to modern ICE and hybrid - Mangram, 2012), longer guarantee as ICE competition (4 years on car and 8 years on entire drive train incl. battery) and guaranteed residual value after 36 month (circa 50% of the purchasing price) is a successful EV. For example, one major shortcoming is that only three studies were found, which included the “availability of charging”. Steg et al. (2001) noted that improvements in infrastructure might lead to decreasing resistance to car use, in our case EV use. Since the Tesla Model S comes with a free-of-charge, European-wide fast charging network, it is be interesting to see if that is a factor contributing to its success, thus if people value it. That is why this research includes both, attribute-levels of the Tesla Model S and expected attribute-levels of future EVs. This will help to refine future EV value propositions, especially it will show which attributes an EV manufacture must include in their value proposition to maximize the adoption of its market offering.
In order to answer the research question, preference measurement and in specific, a choice-based approach is utilized in order to empirically test which characteristics of EVs are valued. A vast amount of research has focused on preference measurement in the field of EVs (Table 2) or sustainable innovations. In preference measurement, each product is represented by a bundle of attribute levels. For example, a car contains a certain engine (attribute) of a certain kind (level). One consumer might prefer a diesel engine over an electric engine. When analyzing these preferences, one can assess the underlying motives of consumers’ actions and therefore predict e.g. future purchase decisions. Thus, preference measurement tries to analyze the value attached to certain attribute levels and quantify them. That is done by using an utility function that translates the characteristics of an EV into preferences. The advantage is that the function can be used to model changing market conditions (Eggers & Sattler, 2011), thus it enables to inter- and extra-polate attribute levels. Hence, future technological developments can be captured. Further, the results can be used to model any future EVs success potential.
This approach performs superior (Jun & Park, 1999) because the customer preferences are extracted out of discrete choice behavior, which represents an actual market behavior and is therefore more valid as e.g.
rankings. The realism of the results is improved by including a no-choice option to enable customers to refrain from adoption. Considering EV adoption, this option is very important as it leaves the respondent the choice to stay with ICE (Eggers & Sattler, 2011). A post-hoc latent-class analysis is used to segment the respondents based on the stated preferences enabling a grouping of customers based on the implicit differences. The received preference-levels are used to compile most preferred products for each segment.
This research updates literature on EV adoption and sets a starting point for re-defining the image of state-of- the-art EVs. In addition to the already researched attributes range, charging costs, charging network, charging time and performance, this research introduces the attributes financial solution and design in the stream of EV adoption literature. These attributes emerged when analyzing a dataset of 10,696 EV owners and conducting a press/article review (Appendix 1: Database Analysis).
The remainder of this paper is structured as follows: First, current research on EV adoption dealing with preferences of consumers is reviewed. This section identifies attribute groups, which influence consumer preferences, thus increase or decrease the EVs adoption potential. In addition, theory is used to explain why the groups have a certain effect on preferences and therefore the likelihood of adoption. Hereby, recent research and current market offerings are equally considered to develop a comprehensive overview. Each motivation ends with a hypothesis. A conceptual model summarizes the hypotheses. Second, the research method is introduced in more detail and the results of the empirical study are presented and discussed. These lead to the managerial implications that give practical advice derived from this research. The limitations and consequently future research opportunities are formulated and form the end of this paper.
2 Attributes Increasing Adoption Probability of Electric Vehicles
Adoption of EVs is defined as consumer adoption. That is a behavioral response, namely the purchase of EVs (Huijts et al., 2012). In recent times, the EV has received a growing amount of attention and the number of market introductions is increasing (Schuitema et al. 2013). There are several abbreviations for different kinds of alternative powered cars. An EV, also abbreviated as ”BEV”, is a car with an all-electric drive train powered from a battery, which is replenished from an external power supply (Schuitema et al. 2013). Other concepts comparable to hybrid electric vehicles or plug-in hybrid electric vehicles are not considered in this research because this paper supports the notion of Eggers & Eggers (2011) that all-electric vehicles are currently considered the automobile technology of the future, which excludes range-extended EVs or hybrid cars. Therefore, this paper focuses on pure EVs in comparison to its conventional ICE opponents2.
Interdisciplinary literature like marketing, transportation and sustainable consumption are analyzing factors, which could potentially influence the adoption of EVs. Noppers et al. (2014), Bergstad et al. (2011), Eggers
& Eggers (2011), Vandecasteele & Geuens (2010), Turrentine & Kurani (2007) Anable & Gatersleben (2005), Steg (2005), Steg et al. (2001), Dittmar (1992) (Table 2) found that several attribute groups have an influence on the probability of adoption of sustainable products. Sustainable products are defined as products, that are manufactured in a way that they meet quality standards and the needs of the current generation without having an impact on future generations consumption patterns, thus have long-term benefits (De Medeiros et al. 2014; World commission on Environment and Development, 1987). Research suggests that the adoption of a sustainable product like an EV is to a great extent influenced by (1) instrumental (e.g.
performance, range), (2) environmental (e.g. emissions) and (3) symbolic attributes (e.g. image) (Noppers et al., 2014, Heffner et al., 2006; Kurani et al., 2007; Skippon & Garwood, 2011; Steg, Vlek & Slotegraaf, 2001).
Instrumental attributes can be directly altered by the manufacturer and refer to the functional (positive/negative) outcome or derived utility of the possession and use of sustainable and innovative products (Dittmar, 1992, Voss et al., 2003). When customers adopt a product due to its function, in case of the EV it would be e.g. the transportation role, instrumental attributes are found to have the strongest influence on adoption because they enable the consumer to perform a certain task (e.g. Schuitema et al.
2013). Noppers et al. (2014) analysed 11 studies and concluded that most research focuses on instrumental attributes. The present research is following this motion and analysis instrumental attributes. It is found that consumers are more likely to choose an EV when they perceive more instrumental advantages in comparison to the next best alternative, e.g. the ICE (Heffner, 2007; Choo and Mokhtarian, 2002). In the specific case of EVs, Noppers et al. (2014) noted that EVs have typically less favorable instrumental attributes when comparing to ICE. These disadvantages might inhibit their adoption. Specifically, it is said that limited range (Gneezy et al., 2012; Nemry and Brons, 2010), high purchasing price (Nemry and Brons, 2010; Ashtiani et al., 2011) and concerns about the performance (Schuitema et al., 2013) are crucial barriers for the wide-scale adoption of EVs. Noppers et al. (2014) formulate it like this: “as long as electric cars have [these]
instrumental drawbacks compared to conventional, less sustainable cars, their wide-scale adoption is not likely.” In line with this, the IEA (2009) and Proost & Van Dender (2010) mentioned that the adoption of EVs is likely to be difficult as their functions differ most strongly from conventional vehicles. In other words, as long as an EV has inferior attributes levels, e.g. lower performance, shorter range, high prices, consumer derive less positive utility from the possession and use of it in comparison to an ICE. That limits its adoption potential (Noppers et al. 2014).
It is interesting to observe that recent research found that EVs have “typically less favorable instrumental attributes” while 2014 EVs, which are available on the market, winning awards for not having exact these drawbacks (Mangram, 2012). To illustrate this discrepancy, the following comparison is posed. Rogowsky (2014) from the American business magazine Forbes sees the Tesla Model S competing with BMW 7 Series, Mercedes S Class, Lexus LS, Audi A8 and Porsche Panamera. This definition of competition is supported by Valdes-Dapena (2013) from CNN and White (2013) from The Wall Street Journal (Table 1).
Table 1: Tesla Model S competition in the US market
Car Tesla
Model S
BMW 7 Series
Mercedes S Class
Lexus LS
Audi A8
Porsche Panamera
Units Sold 20131 18,000 10,932 13,303 10,727 6,300 6,421
Price (starting at)2 €75,340 €74,200 €80,920 €92,800 €76,700 €83,277
CO2 emissions | g/km2 06 225 194 249 178 196
Range | km2 502 865 987 785 1,093 952
Costs of energy for daily usage (100km)2, 3
18.1 kWh/100km
€3.865
9.25l/100km
€13.8
7.9l/100km
€11.8
10.7l/100km
€15.9
7.5l/100km
€11.2
8.4l/100km
€12.5 Costs of energy for long-
distance travel2, 3 (e.g. Munich-Berlin = circa 600km)
€0.04 €82.8 €70.6 €95.7 €67.0 €75.0
Performance (hp)2 380 320 252 387 258 310
1US figures from Forbes (Rogowsky, 2014) | 2Average figures from car manufacturer websites (Tesla Motors, 2014d, BMW, 2014; Mercedes Benz, 2014a,b; Lexus, 2014a,b; Audi, 2014a,b; Porsche, 2014) | 3Tesla Motors (2014c) - average fuel costs of €1.49 per liter/ €0.18 per kWh | 4Free supercharging, incl. in the car’s price for lifetime | 5Charging at home; to charge 18.1 kWh, 20.3kWh are needed due to losses; Tesla Motors (2014e) | 6When charging with renewable energy (over 95% of all superchargers are powered by renewable energy sources, thus CO2 free) |
Table 1 illustrates selected specifications of the competitors and the Tesla Model S. Not only is the from prior research predicted “limited adaption” not reflected in the actual sales numbers, but the specifications of the Model S outperform the competitors in most characteristics. One could argue that the range is a shortcoming because compared to the competitors it is the shortest. Opposing to that, it comes with the free- of-charge supercharger network, enabling complimentary long-distance travel with charging times of 15 to 25 minutes. In addition, it has no CO2 emissions and the price is one of the lowest. This suggests that the assumed higher price, limited range, poor performance and constrained adoption potential by former research does not hold for an EV with the characteristics similar to the Tesla Model S.
2.1 Method to Select Attributes
The EV industry is constantly changing and breakthrough innovations in e.g. battery technology is changing and will change the development of the industry (Mangram, 2012; Google, 2011; Muller, 2014). Because of that, this paper only considers research on EV adoption from the last ten years. This approach contributes to the validity of this paper by leaving aside research on immature EV technology from e.g. the nineties.
To further strengthen the validity of the selected attributes and to confirm that the selected bundle of attributes is a comprehensive set of adoption enhancing attributes of current and future EVs, this research follows a four-step approach, including an explorative and prior research section. First, prior research is reviewed to receive an overview of previously researched attributes. Second, the leading EV market offering is analysed in order receive benchmark data of a 2014 EV using experts opinion and manufacturer information (suggested by Netzer et al. 2008). Third, a press and forum review is conducted to extract most discussed features by the public (suggested by Papies et al., 2011). Fourth, the database of Tesla Model S owners provided from Tesla Motors Netherlands BV is analysed to identify what actual owners are most concerned with.
By conducting these steps, this study closes several research gaps. Until now, no paper considered a combination of prior research, current market offerings, public voice and actual EV owners’ opinion to arrive at a comprehensive set of attributes. Future attribute levels that launch in 20153 or levels that are in preparation for the next years are included. This is done to prevent this paper from being outdated shortly.
2.1.1 Prior Research
The selection of attributes includes the most frequently and most important researched attributes. It includes papers that researched electric vehicle adoption potential and used preference testing. Table 2 lists the used attributes or prior research.
Table 2: Summary of Prior Research
Article Range Fuel/charging costs Performance Charging Network Other Attributes Kim et al.
(2014) 100 km
250 km 400 km 550 km
Costs of Energy 35% more expensive 25% more expensive 15% more expensive 5% more expensive 5% less expensive 15% less expensive 25% less expensive 35% less expensive
Max Speed 80km/h 120km/h 160km/h 200km/h
Price, Share of EV in social network, reviews, recharge time, distance to charging station,
Ida &
Tanaka
(2014) 100 km
200 km 300 km 400 km
% cheaper as ICE 60%
70%
80%
90%
Price, emissions, time to find charge station
Eggers &
Eggers (2011)
150 km 250 km 350 km
Price, drivetrain technology
Hidrue et al.
(2011) 120 km
240 km 320 km 480 km
Fuel Costs
$0.50/gal
$1.00/gal
$1.50/gal
$2.00/gal
Acceleration 20% slower 5% slower 5% faster 20% faster as ICE
Price, emission reduction
Axsen et al.
(2009)
Fuel Costs
Dynamically fitted to respondent2
Horsepower 70% user HP 85% user HP 115% user HP
Price, subsidy, fuel efficiency
Ahn et al.
(2008)
Fuel Price US$0.58/L US$0.87/L US$1.16/L US$1.45/L
Fuel type, body type, maintenance cost, engine displacement, fuel efficiency
Mau et al.
(2008) Users range 120% user range 150% user range 200% user range
Fuel Costs User costs 110% user costs 125% user costs
Fuel availability all stations 1 in 5 1 in 10 1 in 20
Price, subsidy, warranty
Potoglou and Kanaroglou (2007)
Fuel Costs 20% of ICE 40% of ICE 60% of ICE 80% of ICE
Acceleration 6s to 100km/h 9s to 100km/h 12s to 100km/h 15s to 100km/h
Fuel availability
% of petrol stations 10%
25%
50%
75%
Price, maintenance cost, incentives, emission reduction
Present
Research 300 km 450 km 600 km 750 km
Costs per full Charge
€0.0
€22.5
€45.0
€67.5
HP & Top Speed 90HP - 150 km/h 180HP - 180 km/h 270HP - 210 km/h 360HP - 250 km/h
Density1
4 charging/1 petrol 2 charging/1 petrol 1 charging/2 petrol 1 charging/4 petrol
Financial solution, charging time, design
1Charging = Charging Station; Petrol = Petrol Station | 2Level shown to respondent is the product of her/his: weekly fuel costs, relative fuel efficiency of the vehicle, relative fuel price
The five studies of EV adoption include driving range of EV as important attribute. Kim et al. (2014)
km – 350 km, Hidrue et al. (2011) 120 km – 480 km and Mau et al. (2008) utilizes a user range ratio. They found that with increasing range, the derived utility in comparison to an ICE increases as well. With more utility derived, the adoption potential of EVs increases as well. Therefore, the present research includes this attribute and in addition, includes levels that become available in 2015 and in the following years. The included levels of the attribute range span from 300 km to 750 km.
H1: The higher the range, the higher the adoption potential of electric vehicles.
The literature review unveiled that charging costs and performance are also important attributes that facilitate consumer adoption of EVs because they are regularly researched and found to have a significant impact.
These findings of prior research are included further down when analyzing the leading EV market offering in 2014.
Risk Aversion Determining the Importance of Attributes
The literature review unveils that a number of attributes influence EV adoption. There are unique attributes and common attributes in each reviewed paper. For example, price and range are included in all research papers. Apparently, researchers and consumers differentiate between more- and less-important attributes.
One theory which tries to explain those differences is the “prospect theory” developed in 1976 by Kahneman
& Tversky. It is based on the assumption that consumers are prone to loss aversion and underweight opportunity costs. That means that the loss of potential gains from other (unknown, new) alternatives when one alternative is chosen are undervalued. In other words, consumers focus their attention mostly to sunk costs4 and less on future costs, meaning that they do not want to give up investments that are already made but may lead to even higher costs in the future. The extend of this effect depends on the individual customer.
Highly risk averse people do not want to take risk and undervalue possible gains from new alternatives more, as people that are not as risk averse. That is because they are uncertain about the gains (Ziamou &
Ratneshwar, 2002; Hoyer et. al, 2013). In addition, consumers’ behavior is directed by the mental coding of gains and losses, which is represented by the prospect theory value function (Thaler, 1985). This mental coding can be either explicit or implicit. One could explain this coding with “special labels” attached to gains and losses (Thaler, 1985). The aim of the theory is to describe or predict consumers’ behavior. One special label in the EV sector might be attached to range or price. Since all former research analysed the effect of range and price on EV adoption, it is concluded that a higher importance is imposed towards these attributes.
Thus range and price are attributes that have the potential to stronger influence EV adoption as other attributes.
H2: Electric vehicle adoption potential is mainly determined by attached costs and range.
H3: Highly risk-averse consumers pay more attention to costs and range than less risk averse consumer.
Covariates for Segmentation
Next to the analysis of consumers’ preferences using the car’s attributes, demographics- and car usage- data is collected. Among others, Schuitema et al. (2013), Potoglou & Kanaroglou (2007), Steg (2005) and Segal (1995) found that consumer characteristics influencing her/his adoption potential. Tables 3 and 4 summarize which characteristics were included in the present study. They are partly adapted from Schuitema et al.
(2013) and Steg (2005). When conducting the preference-based segmentation, these covariates help to characterize the resulting segments and therefore make them approachable for practitioners (Papies et al.
2010).
Table 3: Car usage data
Question Answer possibility
What is your current vehicle? Text field
For which purpose do you mainly use your car? Please select: Private/Business/ n/a How do you normally acquire your cars? Please select: Leasing/Financing/Cash How many kilometers do you drive per day? Text field
What range should a car have on a single charge/tank?
(in kilometers) Text field
How many cars do you have? Text field
When do you plan to purchase a new vehicle? not planned/ Date selection (2015-2020)
In which category do you plan to purchase your next car? City/ Compact/ Sport/ Mid-Size/ Full-Size/ Luxury/ n/a
Table 4: Consumer characteristics
Item Answer possibility
Age Drop-down selection
Gender Please select: other/female/male
Country Text field
Education Please select: Less than High School/ High school/ Undergraduate degree/ Graduate degree Income <€2000; €2000 - €5000; €5000+; prefer not to answer
Employment Please select: Full- or Part-time/ Unemployed/ Retired/ Student or Home/ other Household size Please indicate the number of persons in your household: 1; 2; 3; 4+
Please state how strongly you agree with the following statements (Risk Aversion, Bearden, 2011)
3 questions, rated on a 7-point Likert scale:
(1) I like to take chance.
(2) When it comes to taking chances, I’d rather be safe than sorry. (reverse coded) (3) I like people who are a little shocking.
The 3 items were tested with Cronbach’s alpha for reliability - Item (2) was removed; Cronbach’s alpha increased to 0.74. Scale considered as reliable (Appendix 3).
2.1.2 Exploratory – Leading Market Offering
This research defines the leading market offering based on the US market. When looking at the US market, Tesla’s Model S is the only EV that managed to outsell its ICE competition in its segment (compare with table 1), ranks best “Overall Car” by consumer reports (Plungis, J. 2014) and is the only EV with a range over 200 miles (321km) range (Ramsey, M. 2014a). On top, it achieved a 5 star rating in Euro-NCAP crash test (European crash test) and 5 stars in every NHTSA (US crash test) subcategory5 (Ramsey, M. 2014b).
Concluding from these facts and from table 1, the Tesla Model S is the leading EV market offering available in 2014. In table 5, a brief overview of its value proposition is given (based on table 1).
Table 5: Tesla Model S Value Proposition
Characteristics
● Driving performance (e.g. faster acceleration as comparable ICE)
● Low charging costs of the battery (ca. 72%6 lower costs as compared to ICE); Free long-distance travel
● Fast-charging network (called: “superchargers”; unlimited energy quantity, free-of-charge)
● Long-range, which is comparable to an ICE
● Price competitive to an ICE
● Zero emissions (one must charge from clean energy; >90% superchargers running on green energy)
Attributes, where the car stands out in comparison to an ICE are performance, costs of usage and the attached charging network.
Prior research (summary in table 2) also underlines the importance of performance and found a positive correlation between the performance and the adoption potential of an ICE. Kim et al. (2014) test performance with levels ranging from 80 km/h to 200 km/h, Hidrue et al. (2011) utilizes a comparison to the ICE and tests levels ranging from acceleration levels 20% lower to 20% higher as ICE, Axsen et al. (2009) focus on the HP of the respondent’s car and test levels from 70% to 115% of the user’s car and Potoglou and Kanaroglou (2007) use acceleration levels from 0 to 100 km/h ranging from 6 seconds up to 15 seconds. The instrumental attribute performance has a positive correlation with adoption potential because, as higher the performance of an EV, as more positive utility while using the product is derived. An higher utility for the consumer enhances the adoption potential (Dittmar, 1992, Voss et al., 2003). The present research follows the notions of Kim et al. (2014) and Axsen et al. (2009) and focuses on top-speed and HP levels. No comparison to the respondents’ car is made. That enables to include four transparent performance levels and enhances the interpretability, as results are not dependent on the respondent’s car usage.
H4: The higher the driving performance, the higher the adoption potential of electric vehicles.
The costs of usage are included in most of the reviewed research papers. The three most recent papers include the charging costs of an EV as follows: Kim et al. (2014) compares it with the respondent’s ICE car and poses ratios ranging from 35% more expensive to 35% less expensive as compared to a ICE, Ida &
Tanaka (2014) use a similar approach and use levels ranging from 60% to 90% cheaper as ICE cars and Hidrue et al. (2011) utilize the comparison to petrol prices with levels ranging from $0.5 to $2 per gallon.
Prior research found a negative correlation between the costs of usage and the adoption potential of EVs.
Research concludes that with increasing costs, the derived utility from a sustainable product decreases and therefore also the adoption potential decreases. The present research follows the approach of Hidrue et al.
(2011) and includes actual costs. In contrast, no comparison to ICE is used but total charging costs. This is done to receive a willingness-to-pay measure for charging an EV. The used levels are derived from current market offerings.
H5: The lower the charging costs are, the higher is the adoption potential of electric vehicles.
The charging network attached to the leading market offering was also included in prior research. Two studies, namely Mau et al. (2008) and Potoglou and Kanaroglou (2007), investigated the impact on adoption potential. They found a positive correlation between the availability of charging stations and adoption potential. The more charging stations are available, as higher was the derived utility from an EV. A higher utility increases the adoption probability. Both studies use a comparison to petrol stations. The present research follows this approach and includes the attribute charging network density with levels matching urban areas in 2014.
H6: An increasing charging network density increases the adoption potential of electric vehicles.
Keeping the range variable in mind, it is interesting to see market developments of charging networks. Tesla Motors delivers its Model S with an attached fast-charging service (Tesla Motors, 2014f). This service enables customers to travel long distances and quickly charge their vehicles along the route. BMW and Nissan started to follow this approach by also making first charging station available for free to their consumers (Blanco, 2014; Nissan USA, 2014). Considering the high electric range of the Model S (compare with appendix 2) in combination with the available fast-charging network, one can argue that the combination of both enhances EV’s adoption potential because it enables long distance travelling. Therefore, it can be argued that the derived utility from range and fast-charging network interact, thus the presents of both increase the adoption potential even more.
H7: High range in combination with a denser fast-charging network increases the adoption potential of electric vehicles.
Most prior research reviewed above includes the price variable. It is argued that an EV comes with a surcharge in comparison with ICE. They test how much more consumers are willing to pay for an EV. The present research is guided by reality and compares the leading EV market offering in 2014 with its competitors. Table 1 clarifies that the Tesla Model S is similar- or lower-priced than its competitors. Thus, customers that are on the scout for a new vehicle in price range of €70.000 to €100.000, are not asked to pay a surcharge to purchase an EV. That is why the present research holds the price attribute constant and therefore does not research its impact. The prospect theory used in the risk aversion section could explain the strong focus on the price variable. The theory developed by Kahneman & Tversky (1976) can be applied by assuming that former research overweight the price variable and underestimate the influence of additional attributes. That is why most prior literature includes the price variable. Since the present situation in 2014 does not show a price difference between EVs and ICEs in the analysed segment, the variable is held constant.
2.1.3 Exploratory – Press and Article Review
This section briefly lays out the results of the press/article review (Appendix 1). It aims at confirming the found attributes and finding new important attributes that have the potential to influence the utility of an EV and therefore to influence customer’s choice, thus consumer’s EV adoption. After the qualitative analysis, the final set of attributes is used to design choice-sets for the subsequent choice-based conjoint (CBC) analysis in the quantitative research section.
The review of 44 articles/forums (listed in appendix 1) between July 2014 and October 2014 showed that in addition to the attributes range and performance, which were already found in theory, the attributes design and charging time are important for consumers when considering an EV. These attributes were most frequently mentioned in the articles. The attribute design refers to the outer appearance of the car while charging time refers to the duration of time the car must be charged in order to top the battery up to a level that enables the consumer to continue his/her trip. These two attributes are therefore included in the CBC analysis.
There was no EV adoption literature found, that researched the EV design’s impact on consumers EV adoption potential using preference testing. The attribute design has a communication role and is able to
Thus, it renders or builds the outcomes of possessions/usage of sustainable innovations for one’s social status (Belk, 1988; Giddens, 1991; Sirgy, 1986) or for one’s self-identity (Griskevicius et al., 2010; Dittmar, 1992;
Roehrich, 2004). There is consensus in literature claiming that products influence how one reflects on her/himself and that consumers are interested in keeping a positive picture about themselves (Belk, 1988;
Dittmar, 1992; Giddens, 1991) and how others see them (Goffman, 1959). Attributes like design have the potential to make an impact on the purchase and use of sustainable innovations. Especially the car segment is known for high conspicuousness and symbolic connotations (Gatersleben, 2007; Heffner et al., 2007; Shove and Warde, 2002). This characteristic of the car segment can be based on several research streams, which analyse the impact since the mid-eighties: Sirgy’s (1986) self-congruity theory, Park et al.’s (1986) theory on brand concept management, Belk’s (1988) theory on the extended self, McCracken’s (1990) theory on symbolic character of consumer goods and Dittmar’s (1992) theory on the meaning of material possessions conclude that this attribute affect purchase and usage behavior of consumers. To shape a positive image of ourselves, we purchase and display products that are inline with the picture we have from ourselves or with the picture we are aiming to fulfill (Belk, 1981; Sirgy, 1985, 1986; Belk, 1988; Ericksen, 1997; Giddens, 1991; Dittmar, 1992; Fennis and Pruyn, 2007). For example, by adopting an environmentally friendly product such as an EV, one can communicate innovativeness (Brown and Venkatesh, 2005; Simonson and Nowlis, 2000; Vande Casteele and Geuens, 2010), independence (e.g. from oil products), one’s intelligence and unique character (Heffner et al., 2007). In addition, Kessmann et al. (2006) showed that the self-image congruence can explain a certain brand choice behavior. Despite that, people are not always fully aware or do not want to acknowledge that a sustainable innovation is bought and used to show-off or to feel good about themselves (Noppers et al., 2014).
A focal role in the explained construct is attached to the displaying of sustainable products. If a customer adapts a sustainable product but her/his peers do not identify it as such, it will not contribute to her/his social status. That can be one explanation why the attribute design receives a significant focus in most press articles and forum entries. It is expected that a significant different design as ICEs contributes to the adoption potential of EVs because with a different design, peers recognize the alternative product and five credits to its owner.
In order to model the attribute design, current EV market offerings are reviewed and categorized (Appendix 2). The review resulted in four attribute categories for design: (1) Same design as ICE; (2) Few design characteristics are different as compared to gasoline car; (3) Many design characteristics are different as compared to a gasoline car and (4) Significant differently designed as compared to gasoline car.
Design Expectation:
The more an EV’s design differs from an ICE’s one, the higher the EV’s adoption potential.
Kim et al. (2014) already included the attribute recharge time in their paper and emphasize that this is a crucial factor for customers when deciding for an EV. The levels tested were 5 min, 1h, 4h and 7h. They concluded that customers are not willing to spend time to charge the battery. It is found that a shorter charging time enhances adoption probability because customers derive more utility. The level 5 minutes showed positive utility while the others were estimated with a negative utility level. Looking at current market offerings, most EVs are equipped with DC fast charging technology (Appendix 2). The implication is that people are able to charge 80% of their battery in 25 to 35 minutes. The rough scale of Kim et al. (2014) does not capture available charging times and therefore fails represents current and future market offerings.
That is why the present research uses the following levels for the attribute charging time: 1-5 min; 6-10 min, 11-15 min, 16-20 min. It is expected that the shorter the charging time is, the more likely the consumers adopt EVs.
H8: The shorter the EV’s battery charging time is, the higher the consumers’ adoption potential of electric vehicles.
2.1.3 Database Analysis of EV Owners
This section briefly lays-out the results of the conducted database analysis (Appendix 1). This procedure is included to receive an up-to-date picture of how current EVs are perceived and why they are successful. The qualitative section aims at confirming the found attributes and finding new important attributes that have the potential to influence the utility of an EV and therefore to influence customer’s choice, thus consumer’s EV adoption. After the qualitative analysis, the final set of attributes is used to design choice-sets for the subsequent choice-based conjoint (CBC) analysis in the quantitative research section.
The analysed database comprises 10,696 Tesla Model S owners who bought a Tesla Model S in 2014. Most Model S owners get into contact with Tesla Motors by booking a test drive. During the booking process, customers get qualified from a inside sales advisor. This qualification results in notes about the customer and these notes entail attributes discussed in the process. After filtering out incomplete datasets, 9649 notes entailed details about discussed characteristics. Every topic discussed during the qualification was filtered out by several keywords and combined to more general attributes. All attributes and exemplary keywords can be found in appendix 1. The results of the count analysis (Appendix 1) showed that in addition to the already considered attributes from prior research (charging; range), financing and design are among the most discussed topics. A financing topic was discussed in 2054 cases and the design in 515 cases. This confirms the result of the press/forum review that design is an important factor. Financing refers to the option how a consumer can receive ownership or timed usage rights of a car. Prior literature fails to research the importance of this attribute. Since EV owners frequently discuss the financing topic when first contacting the car company, it is included in the present research to explore its effect. Analysing current EV market offerings (Appendix 2) led to the following attribute levels that represent current financial solutions: Paying cash for the car and leasing the battery; Paying cash for car and battery; Leasing car and battery; Financing car and battery. It is interesting to see that some manufacturers divide the product in car and battery and others do not. E.g. Nissan offers to lease the battery of the Nissan Leaf while e.g. Mercedes with its eSmart sells both as a package. This research aims at setting a starting point to analyze which financial solutions are beneficial for adoption of EVs. Table 6 on page 14 summarizes the final set of attributes with corresponding levels.
Further, 3 from 8 analysed papers found environmental impact as an important attribute in EV adoption. The present research analysed a database of actual EV owners and review press/forum publications. None of them unveiled the environmental impact as decisive factor. For example, from the 10,696 reviewed memos from actual EV owners, only 113 memos mention the environmental impact variable. Thus, for actual owners, the environmental impact variable is at least as uninteresting as the variables price (frequency = 134) or safety (frequency = 92). That is why these variables are fixed through all choice sets.
2.2 Conceptual Framework
The raised hypotheses in paragraph 2 except that the attributes range, charging costs, fast charging network density, charging time, financial solution, performance and design have the potential to influence the adoption potential of EVs. The conceptual framework (Graphic 1) depicts the expected effects.
Graphic 1: Conceptual Framework
