A choice experiment on alternative fuel vehicle preferences of private car owners in the Netherlands

Transportation Research Part A 61 (2014) 199–215

Contents lists available at ScienceDirect

Transportation Research Part A journal homepage: www.elsevier.com/locate/tra

A choice experiment on alternative fuel vehicle preferences of private car owners in the Netherlands Anco Hoen a,⇑, Mark J. Koetse b a b

PBL Netherlands Environmental Assessment Agency, Oranjebuitensingel 6, 2511 VE The Hague, The Netherlands Institute for Environmental Studies (IVM), VU University Amsterdam, The Netherlands

a r t i c l e

i n f o

Article history: Received 5 October 2012 Received in revised form 20 November 2013 Accepted 31 January 2014

Keywords: Car choice Alternative fuel vehicles Electric cars Choice data Consumer preferences

a b s t r a c t This paper presents results of an online stated choice experiment on preferences of Dutch private car owners for alternative fuel vehicles (AFVs) and their characteristics. Results show that negative preferences for alternative fuel vehicles are large, especially for the electric and fuel cell car, mostly as a result of their limited driving range and considerable refueling times. Preference for AFVs increases considerably with improvements on driving range, refueling time and fuel availability. Negative AFV preferences remain, however, also with substantial improvements in AFV characteristics; the remaining willingness to accept is on average € 10,000–€ 20,000 per AFV. Results from a mixed logit model show that consumer preferences for AFVs and AFV characteristics are heterogeneous to a large extent, in particular for the electric car, additional detour time and fuel time for the electric and fuel cell car. An interaction model reveals that annual mileage is by far the most important factor that determines heterogeneity in preferences for the electric and fuel cell car. When annual mileage increases, the preference for electric and fuel cell cars decreases substantially, whilst the willingness to pay for driving range increases substantially. Other variables such as using the car for holidays abroad and the daily commute also appear to be relevant for car choice. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Concerns over climate change and reduction of greenhouse gas emissions, and dependence of economies on foreign energy sources, have become reasons for extensive research on the use of alternative fuels in transport in the last 10–15 years. Recently the European Commission announced that by 2050 a 60% cut in transport-related CO2 emissions compared to the year 2000 should be aimed for (European Commission, 2011). Alternative fuel vehicles (AFVs) are essential for reaching that goal, since passenger cars make up roughly 50% of transport-related CO2 emissions in the EU (PBL, 2009). AFVs such as electric, fuel cell, (plug-in) hybrid and flexifuel cars use non-fossil fuels and have the potential to emit only a fraction of the CO2 emissions that conventional petrol and diesel cars emit. Since AFVs are different in terms of costs and their ease of use, consumer preferences may be very different from conventional cars. This is why preferences for and market potential of AFVs have received wide attention since the mid-1970s. Since the availability of AFVs on actual markets is still very limited, stated preference research is necessary in order to obtain insight into potential barriers to AFV adoption. Since the beginning of the 1980s many choice experiment studies have ⇑ Corresponding author. Tel.: +31 (0)6 46 93 46 53. E-mail address: [email protected] (A. Hoen). http://dx.doi.org/10.1016/j.tra.2014.01.008 0965-8564/Ó 2014 Elsevier Ltd. All rights reserved.

200

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

aimed to identify relevant factors in the market penetration of alternative fuel vehicles. This study contributes to this literature by performing an online stated choice experiment among private car owners in the Netherlands. Our main goals are to obtain insight into the preferences of private car owners in the Netherlands for AFVs, to uncover the car characteristics that affect these preferences, and to find out to what extent these characteristics need to change in order to make consumers indifferent between conventional cars and AFVs. We attempt to identify the (socio-demographic) characteristics of car buyers that are currently most susceptible to buy an AFV with the aim of uncovering interesting market segments and potential early adopters. Our study has added value for several reasons. First, as we will show in the next section, most previous choice experiments on AFV preferences include a limited number of the attributes that have been shown to substantially affect AFV preferences. In particular experiments in which fuel time and fuel availability are combined are rare. Moreover, most studies compare preferences between one or two AFV types and the conventional technology. Few studies simultaneously include all AFVs that are currently considered as viable options for substantial CO2 emission reductions. In our choice experiment we both include a wide range of car attributes and most of the AFV types that are currently considered to be viable. Our study is therefore more comprehensive and allows for more accurate and reliable estimates of the impact of different (pricing) policy measures on AFV adoption and CO2 emissions. Second, most of the existing studies were carried out in the USA and Canada and only limited empirical evidence is available for Europe (Dagsvik et al., 1996; Caulfield et al., 2010; Mabit and Fosgerau, 2011; Ziegler, 2012). The results of existing studies show substantial differences in stated preferences for AFVs and AFV characteristics both across and within countries. This is an indication that stated choice results from different countries are not directly interchangeable and that country specific experiments are necessary. The Dutch case may be of particular interest since it is potentially more suitable for vehicles with limited driving range due to its particular spatial characteristics. The largest cities are relatively small (400,000–700,000 inhabitants), concentrated in a fairly small area (roughly 100 square kilometers) which also includes many medium-sized cities and towns, and connected by a relatively dense network of highways. Third, in our experiment we chose to include the current technology (petrol, diesel and LPG) in only a subset of the choice tasks. We estimate models on both the full sample and the subsample. By comparing the results for these two samples we can assess the effects of including a status quo choice option (that is, the current technology). More importantly, we are thus able to examine differences in marginal willingness to pay (WTP) and AFV market potential for the short term, with petrol, diesel and LPG cars as the current technology, and the long term, in which petrol, diesel and LPG cars may have been replaced by, for example, hybrid cars as the default technology. Fourth, we make a distinction between buyers of new and secondhand cars. Preferences of secondhand car buyers may be quite different from that of new car buyers. This is especially relevant when we consider that currently, as a result of beneficial taxation rules, most AFVs sold in the Netherlands are company cars that will enter the secondhand private market in 3–4 years. The paper is organized as follows. The next section presents the attributes and attribute levels used in our choice experiment. Section 3 explains the survey design and the data collection process. Estimation results are discussed in Sections 4 and 5, followed by a concluding section interpreting the results an indicating policy implications.

2. Choice attributes and attribute levels Findings from the existing literature on AFV preferences show that next to purchase price and operating costs, driving range (Hensher and Greene, 2001; Mau et al., 2008; Train, 2008; Beck et al., 2011; Hidrue et al., 2011; Zhang et al., 2011; Maness and Cirillo, 2012), recharge time (Hidrue et al., 2011) and fuel availability (Horne et al., 2005; Potoglou and Kanaroglou, 2007; Mau et al., 2008; Train, 2008; Ziegler, 2012) may have substantial effects on consumer preferences for AFVs. Emission reduction is also signaled as an important factor (see Ewing and Sarigöllü, 1998; Batley et al., 2004; Potoglou and Kanaroglou, 2007; Beck et al., 2011; Hidrue et al., 2011; Maness and Cirillo, 2012; Ziegler, 2012). Although the study by Beck et al. (2011) does not look at AFVs specifically, their findings show that emission charging targeted specifically at vehicle emission rates may have substantial effects on vehicle purchase decisions. Table 1 gives an overview of the various attributes used in previous choice experiments on AFV preferences. In addition to these findings from literature, consultations with policy makers from the Ministry of Energy, Agriculture and Innovation and the Ministry of Infrastructure and the Environment as well as stakeholders from the automotive sector provided information for the attribute selection process. Besides car type, which includes the conventional technology, the hybrid, plug-in hybrid, fuel cell, electric and flexifuel car, we included seven attributes in our design, that is, catalogue price, monthly costs, driving range, recharge/refueling time, additional detour time to reach a fuel or recharge station, number of available models, and policy measure. To ensure that the choice options were as close as possible to a respondents actual situation, the fuel type of the current technology and purchase price and monthly costs were made respondent specific. To this end, several questions were asked prior to the choice tasks to reveal information on the current car of respondents (that is, annual mileage, weight of the car, road tax exemption). Since characteristics of a next car may be very different from those of the current car due to job changes and changes in family or living situations, we also asked respondents to provide information on the presumed fuel type and purchase price of their next car. Below we discuss in detail the attributes and associated levels.

201

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215 Table 1 Attributes included in peer-reviewed choice experiments on consumer preferences for alternative fuel vehicles.a

Beggs et al. (1981) Calfee (1985) Bunch et al. (1993) Bunch et al. (1995) Dagsvik et al. (1996) Ewing and Sarigöllü (1998) Hensher and Greene (2001) Batley et al. (2004) Horne et al. (2005) Hess et al. (2006) Potoglou and Kanaroglou (2007) Ahn et al. (2008) Train (2008) Mau et al. (2008) Dagsvik and Liu (2009) Caulfield et al. (2010) Beck et al. (2011) Hidrue et al. (2011) Mabit and Fosgerau (2011) Zhang et al. (2011) Maness and Cirillo (2012) Ziegler (2012)

Purchase price

Fuel costb

X X X X X X X X X X X

X

X X X X X X X X X

O&M cost

X X X X X X X X

X X X X X X X X X X X X X X X X X X X X X

Range

X X X X

Fuel time

Fuel availability

Emissions

X

X X

X X

X

Incentivec

X X X

X X

X

X

X

X

X X X

X

X

X X

X X X X

X

X X

X

X X

X

a Various attributes included in the choice experiments are not included in the table, such as vehicle size, top speed, acceleration, body type, and air conditioning. b Includes variations on fuel cost, for example, fuel consumption, fuel efficiency times fuel price, etc. c This concerns government policies that try to stimulate alternative fuel technologies. Incentives used were reduced taxes, free parking, access to express lanes, and access to high occupancy vehicle lanes.

2.1. Car type The car types included were the conventional technology (petrol, diesel or LPG), hybrid, plug-in hybrid, fuel cell, electric and flexifuel car. 2.2. Purchase price Purchase price levels were made respondent specific. Prior to the choice tasks respondents were asked what the price range of their next car would presumably be, for which they could select from a drop-down menu with 17 price categories (ranging from less than € 9000 to more than € 100,000). From the price range category selected by the respondent we used the lower limit as our point of reference. To add some variation this figure was multiplied by a random number generated from a uniform distribution between 0.9 and 1.1, and rounded to the nearest hundred, resulting in the purchase price for the current technology. The purchase price of an AFV was equal to the price of the current technology plus a design-dependent mark-up, using three different mark-up levels for each AFV. The mark-up of the electric vehicle was also dependent on the vehicle driving range since higher driving range requires a larger battery pack with higher associated costs. Table 2 gives an overview of the purchase price mark-up levels for each AFV if respondents indicated their next car would be a new car. AFV mark-up levels were exactly half as high for respondents that indicated they planned to buy a secondhand car. 2.3. Monthly costs Monthly costs were comprised of three different cost elements, that is, fuel costs, maintenance costs and road taxes. Fuel costs presented in the choice tasks were respondent specific and calculated based on the vehicle weight, mileage and fuel

Table 2 Mark-up levels for alternative fuel vehicles used in the design.a

a

New cars

Level 1

Hybrid Plug-in hybrid Fuel-cell Electric Flexifuel

€ € € € €

0 0 1000 1000  (driving range/140) 500

The mark-ups for secondhand cars are exactly 50% of that of new cars.

Level 2 € € € € €

2000 2000 3000 3000  (driving range/140) 1200

Level 3 € € € € €

6000 7000 10,000 10,000  (driving range/140) 3000

202

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

type, all indicated by the respondent in questions prior to the choice tasks. The per kilometer prices for electricity, hydrogen and biofuels were varied according to the information in Table 3. Fuel prices for petrol, diesel, LPG were not varied in the design of the experiment. Accordingly, the fuel costs of hybrid cars were also not varied since hybrids can only run on conventional fuel and cannot directly tap electricity from the net. One might argue that fixing fuel prices for petrol, diesel and LPG is unrealistic since oil prices have fluctuated significantly over the past decades, a situation which is unlikely to change in the near future. We however are primarily interested in the effects of relative price differences between conventional fuels and alternative fuels. Maintenance costs were kept fixed for petrol (€ 50 a month), diesel and LPG (€ 150 a month). Three levels were adopted for electric vehicles and fuel-cell vehicles: € 20, € 30 and € 50 a month. The maintenance costs were fixed for plug-in hybrids, hybrids (both € 150 a month) and flexifuel cars (€ 100 a month). Although the information on maintenance costs of AFVs is very limited we expect the average maintenance costs for electric and fuel cell cars to be lower due to the fact that they have less moving parts that can break down. Since plug-in hybrids have both battery packs and an internal combustion engine we expect them to have higher maintenance costs. In the Netherlands, road taxes (MRB) differ for petrol, diesel and LPG vehicles and depend on the vehicle weight. In addition, some vehicles are exempt from MRB, depending on the amount of CO2 they emit per kilometer. Prior to the choice tasks respondents were asked whether they pay MRB or not. If not, then the levels for monthly costs were corrected for this to enhance the plausibility of monthly costs scenario’s for specific respondents. There were no levels adopted for MRB in the experimental design. All AFVs were exempt from MRB in the experimental design.

2.4. Driving range The driving ranges of hybrids, plug-in hybrids and flexifuel vehicles do not differ from that of conventional cars. For these four car types the driving range was kept constant at ‘same as current driving range’. The driving ranges of the electric and fuel cell car were derived from a range of studies and consultations with experts. For electric cars the current real-world driving range amounts to approximately 75 km, which we adopt as the lower bound of the level values for driving range of the electric car. Other driving ranges included were 150 km, 250 km and 350 km. For the fuel cell car driving ranges are somewhat more difficult to obtain since only prototypes exist. Based on expert consultation we included a current driving range of 250 km, and feasible medium to long-run driving ranges of 350 km, 450 km and 550 km.

2.5. Recharge/refueling time Four levels of recharge/refueling times were applied for plug-in hybrid, electric and fuel-cell vehicles. The value for the other car types was set constant at 2 min as a good proxy for the average refueling time of conventional cars. See Table 4 for a detailed overview of the car type specific recharge/refueling times. Recharge times for electric vehicles depend largely on the size of the battery and the recharging technology. For example, fast charging an electric car with a driving range of 150 km currently takes around 30 min (see for example, Grüning et al., 2011), while charging the same car using the normal electric grid takes around 8 h. We included values ranging from 30 min (fast charging) to 8 h in our experiment to allow for a wide range of possible future values and for heterogeneity in recharge technology. For the plug-in hybrid we also included a wide range but smaller values since a plug-in hybrid car has smaller batteries. The literature on refueling time for the fuel cell reports values ranging from 3 to 9 min for refueling technologies with hydrogen contained in gas or liquid form (see for example Thomas, 2009; Eberle et al., 2012). Refueling technologies with hydrogen storage contained in solid form, however, take more time, although the extent to which is uncertain (see for example Von Helmolt and Eberle, 2007). Since the refueling technology to be used in the future is uncertain we varied refueling times for the fuel cell car from 2 to 25 min. These values were based on both the existing literature and consultations with experts in the field.

Table 3 Fuel prices for the six car types (price level 2011). Fuel type

Car type

Level 1

Level 2

Level 3

Petrol Diesel LPG Petrol + electricitya Hydrogen Electricity Biofuels

Petrol, hybrid Diesel LPG Plug-in hybrid Fuel-cell Electric Flexifuel

€ 1.55/l € 1.25/l € 0.65/l 70% of petrol 65% of petrol 25% of petrol 65% of petrol

– – – 90% of petrol price 100% of petrol price 40% of petrol price 100% of petrol price

– – – 100% of petrol price 130% of petrol price 75% of petrol price 130% of petrol price

price price price price

a Plug-in hybrids drive a short distance on electricity and the remainder on petrol or diesel. The variation in the level value is based solely on assumed variation in the price of electricity.

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

203

Table 4 Recharge/refueling time for the six car types. Car type

Level 1

Level 2

Level 3

Level 4

Petrol/diesel/LPG Hybrid Plug-in hybrid Fuel-cell Electric Flexifuel

2 min 2 min 20 min 2 min 30 min 2 min

– – 35 min 10 min 1h –

– – 1h 15 min 2.5 h –

– – 3h 25 min 8h –

2.6. Additional detour time To test for differences in the availability of refueling locations we included additional detour time as an attribute. Four levels were used for fuel cell, electric and flexifuel vehicles, that is 0, 5, 15 and 30 min, and additional detour time is equal to 0 for the other car types. For electric vehicles additional detour time only appeared when recharge time was equal to 30 min (fast charging). For recharge times larger than 30 min we assumed that recharging the vehicle occurs at home. 2.7. Number of available brands/models Preferences of car buyers are substantially heterogeneous (Hoen and Geurs, 2011; Carlsson et al., 2007; Brownstone et al., 2000). If the car supply would be (much) less diversified, the chance that people would be driving the same car would become higher with higher numbers of cars sold. This might affect AFV preferences. Four attribute levels (1, 10, 50 and 200) were used for each AFV, while number of models for the current technology was always presented to the respondent as ‘‘same as the current amount’’. We argue that by including this rather general attribute we should be able to identify the existence of preferences for supply and choice. The assumption here is that respondents will realize that a higher number of models implies more choice and increases the probability of finding their preferred brand and/or vehicle type. 2.8. Policy measure Finally an attribute was added to test for respondents’ sensitivity for policy intervention. Three policy measures were included as levels for this attribute, complemented with a fourth ‘current policy’ level. The policy options were chosen in consultation with the Ministry of Energy, Agriculture and Innovation, which is responsible for the electric mobility program in the Netherlands. The chosen policy measures were (1) free parking, (2) access to bus lanes within the built up area, and (3) abolishment of the road tax exemption. Parking fees, both for residents and visitors are very common in the Netherlands and may be as high as 5 Euro per hour in densely populated (urban) areas. We explained to respondents that the ‘free parking’ measure would apply to both parking permits and parking zones throughout the country (see also Appendix A). Bus and taxi lanes are common in larger cities in the Netherlands. They allow for faster and smoother public transport through cities that are relatively difficult to access by car. Access to bus lanes for private car users could potentially decrease travel times and therefore add to AFV adoption. In 2011 energy efficient cars (depending on their CO2 emission per kilometer) were exempt from road taxes. This was a popular measure and one of the reasons many small energy efficient cars were sold in the Netherlands. Abolishment of this exemption was under discussion during the design of the experiment. 3. Survey design and data collection 3.1. Presentation, pre-tests, statistical design and pilots Information on the attributes and their levels was given to the respondent prior to the choice tasks (see Appendix A for details). Fig. 1 gives an example of a choice task. For the purpose of this paper we translated the originally Dutch wording to English. Respondents were given three options to choose from. We asked them to state their first and second most preferred choice.1 A total of eight choice tasks were given to each respondent. The order of the attributes remained the same throughout all choice tasks. Explanatory information could be accessed through ‘pop-up tooltips’ when moving the cursor over the question marks added to most attributes. We used the Sawtooth CBC software package to program and field the online questionnaire, and to generate the statistical design. The program uses a randomized design strategy and produces a design that is nearly orthogonal within respondents. The default design option in the program (called minimal overlap) ensures that attribute levels are duplicated as little as possible within choice sets. A second design option allows for slightly more overlap between attribute levels, which reduces efficiency but substantially increases efficiency when attribute interaction effects are considered (see Chrzan and Orme, 1 All results presented in this paper are based on the 1st choice data only. Although second choice data is available, in the end we felt that the second preferred choice is far more hypothetical than the first preferred choice and therefore gives far less reliable information.

204

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

Fig. 1. Choice task example (respondent values used in this example: km/year: 15,000–25,000; Tax exemption: No; Weight: 1200 kg; Next car: New; Fuel type next car: Petrol; Purchase price next car: € 21,000–€ 24,000).

2000). Since we were assured a priori of a relatively large number of respondents, efficiency on main effects was not considered to be a problem. Since interaction effects were likely to be important, we opted for the design strategy with a slight overlap between attribute levels. Using this strategy we generated an alternative-specific statistical design, consisting of 30 survey versions of 8 choice tasks each, whereby each respondent was randomly assigned to one of the 30 versions.2 In order to test the resulting design we performed simulations using the expected number of respondents.3 The D-efficiency of our design was lower than but close to the D-efficiency of the complete enumeration design. Furthermore, standard errors of the main effects that resulted from the simulations were all very acceptable and well below 0.1 and increasing the number of questionnaire versions even further did not decrease the standard errors substantially. For a comparison of different Sawtooth design strategies with other well-known design strategies we refer to Chrzan and Orme (2000). In the design around 65% of the choice tasks include the conventional technology (CT). In the remaining 35% of the choice tasks respondents had to choose between three AFVs. The reasoning was that including the CT as a choice option in every choice task could result in a large share of respondents always choosing the CT, regardless of the alternatives and their characteristics (status quo bias). In this case, potentially very little information on preferences for AFVs would result, making reliable model estimations difficult. Before fielding the questionnaire a number of consultations, tests and pilots were carried out to ensure questions were not difficult to understand and to test the level values of the attributes in order to zoom in on the most interesting parts of the utility curves. Experts and policy makers were invited to comment on the preliminary selection of attributes and attribute levels. This led to some changes in the questionnaire and design of the stated choice questions. A test version was then prepared and sent to approximately 20 experts and colleagues who commented on wording and general quality of the questionnaire. This led to additional improvements. Two consecutive pilots on small samples were fielded to finalize the testing phase; 52 respondents leading to 416 observations for pilot 1, and 51 respondents leading to 408 observations for pilot 2. The main objective of the pilots was to test the 2 3

Sawtooth does not allow for the use of priors in generating the statistical design. Simulation results for the statistical design used in this study are available upon request from the authors.

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

205

attribute level values. Several additional questions were added following the stated choice questions to determine at which level of a certain attribute respondents decided to reject a choice option. Results for pilot 1 showed expected signs on all attributes and attribute levels and were plausible in terms of magnitude. The pilots gave rise to changes on three aspects. The range levels for electric vehicles included in the first pilot were 75 km, 150 km, 250 km and 450 km. The results indicated that the difference in preference between the first three levels were minimal. We therefore decided to replace 250 km by 350 km. In a second pilot the distinction between 350 km and 450 km turned out to be minimal. In the main study we therefore included 75 km, 150 km, 250 km and 350 km, mainly because the first pilot indicated that the added value of 450 km compared to 350 km was limited. The levels for hydrogen vehicles in the pilot were 250 km, 300 km, 400 km and 600 km. Results indicated that differences in preferences for the first three levels were minimal, as were the differences in preference between 500 km and 600 km. We changed the levels to 250 km, 350 km, 450 km and 550 km in order to get a better grip on possible non-linearities. Detour times included were 2, 8 and 20 min. Results indicated that 2 min was not considered relevant by respondents and that 8 min had only limited added value. We changed detour times to 5, 15 and 30 min in order to test a wider range of detour times and to get a better grip on possible non-linearities. 3.2. Respondent panel, segmentation and selection characteristics Respondents for the choice experiment were selected from the well-established and internationally known NIPObase panel, owned by TNS-NIPO. More specifically, respondents were selected from a separate Dutch automotive panel containing more than 40,000 households with one or more cars.4 The automotive panel offered several advantages for our stated choice experiment, that is, the possibility for a priori segmentation on car type and car use, regular screenings that provide additional information on current car type and car use and familiarity of the panel members with automotive related questions. Respondents are paid for their efforts provided they fill out the entire questionnaire. We added a segmentation for owners of new and secondhand cars since their preferences for AFVs are likely to diverge due to differences in budget constraints and vehicle usage. We asked respondents whether they were planning to replace their current car with either a new or a secondhand car (to which respondents replied that they had the intention to buy another car in the next 3–4 years). Furthermore we made a segmentation on fuel type (gasoline, diesel, and LPG) as the preferences for AFVs are likely to be influenced by different tax regimes adopted for these fuel types in the Netherlands (where purchase tax, road taxes and fuel levies vary substantially between petrol, diesel and LPG cars). The reason for a priori segmentation was to obtain sufficient observations for each of the above mentioned groups so that we could test whether their preferences were indeed different. Potential bias in the results due to these segmentations will be discussed where relevant. For car owners of both new and secondhand cars we asked TNS to aim for 300 complete questionnaires for respondents with petrol, diesel and LPG cars, making a total target of 1800 completed questionnaires. Within each segment we asked TNS to aim for representative sampling on age (between 18 and 75), gender, education, and place of residence for the Dutch population. We added selection questions in the questionnaire to target the respondents who were most likely to make car choice decisions. The final version of the questionnaire was fielded in June 2011. Total response rate including only completes was 73% (approximately 11% of respondents were disqualified based on the selection questions, the remaining 16% represents non-response). After approximately 2 weeks we obtained 1903 completes, 660 for petrol, 754 for diesel and 489 for LPG, leading to a total of 15,221 observations (3 observations were missing). The share of LPG drivers is relatively low in the Netherlands (around 5%), which is why the target of 600 was not reached. Background characteristics for the 1903 respondents are presented in Table 5. There is a clear overrepresentation of male respondents in the sample with respect to the total Dutch population. The reason for this is that we added a selection question in the beginning of the questionnaire asking the respondent to indicate whether he or she was the person in the household who drove the car most frequently. If not, people were excluded from the questionnaire. In the Netherlands men drive roughly twice as many kilometers as women, and women are twice as likely to be car passengers than men, which explains the dominance of males in our sample (CBS, 2013a). The age distribution is fairly even between the age group 35–65. The age group 18–35 is somewhat underrepresented compared to the average Dutch population. The average household size (not shown) is 2.8 which is quite high compared to the national average of 2.2. Education levels are comparable to Dutch average apart from a slight overrepresentation of secondary school level 3 and bachelor graduates. 3.3. Choice frequencies Table 6 shows the car type choice frequencies for the full sample and the sample with only those choice sets that contain the conventional technology (CT sample). The conventional technology was chosen in 73% of all choices that contained the 4 Our sample does not include people that do not own a car. This experiment therefore only gives information on those people that currently have either a new or a secondhand car. The advantage is that respondents in our sample actually went through the choice process of buying a car. Apart from the practical argument that the NIPO panel does not include people who do not own a car, our choice for car owners was also based on the expectation that they have more clearly defined preferences for car and AFV attributes, can make better informed and therefore more reliable choices, and moreover are more likely to buy another car in the future than people who do not have a car.

206

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215 Table 5 Background characteristics for the 1903 respondents. Variable

Percentage share (%)

Gender Male Female

81 19

Age category 18–25 25–35 35–45 45–55 55–65 65 and older

0 9 19 24 25 23

Household size 1 Person 2 Persons 3 Persons 4 Persons or more

9 46 15 29

Highest finished education Primary school Secondary school (level 1) Secondary school (level 2) Secondary school (level 3) Secondary school (level 4) Bachelor Master/PhD Do not know/no response

2 13 10 26 10 27 11 1

Degree of urbanization Non-urbanized (less than 500 inhabitants/km2) Little urbanized (500–1000 inhabitants/km2) Moderately urbanized (1000–1500 inhabitants/km2) Urbanized (1500–2500 inhabitants/km2) Very urbanized (2500 or more inhabitants/km2)

15 21 22 27 14

conventional technology. This percentage is, by definition, lower in the full sample. What Table 6 shows us is that there is sufficient variation in car choice to reliably estimate choice models for both samples. The figures in Table 6 are not an indication of AFV preferences, because the frequency of occurrence in the choice tasks is different for each AFV. Car types that have many different levels (electric car and fuel cell car) appear more often in the choice tasks. Furthermore, the current technology was not included in all choice sets so it was not possible for a respondent to always choose the current technology. It should be noted that 924 out of 1903 respondents systematically chose the current technology whenever it was among the choice options. Since this may well reflect their true preferences, we did not exclude them from the sample. 4. AFV preferences and willingness to pay for AFV characteristics 4.1. Estimation results We estimate a mixed logit (ML) model with random parameter distributions for all attributes using NLogit 4.0.1.5 For the simulations we use a maximum of 150 iterations and 2000 Halton draws from a normal distribution to obtain sufficient reliability. Initial MNL estimations show that current fuel type only has small effects on the AFV specific constants. Therefore, not accounting for differences in preferences between petrol, diesel and LPG drivers has either only limited or no effect. Initial estimations also show that coefficients on purchase price and monthly costs are quite different for respondents with new and respondents with secondhand cars (other differences in attribute coefficients between new and secondhand car drivers were small and statistically insignificant). In the final model we therefore distinguish between new and secondhand car owners with respect to purchase price and monthly costs. Estimation results and WTP estimates for both the full and the CT sample are presented in Table 7. All coefficients have the expected sign and most are statistically significant at the usual critical significance levels. The AFV-specific constants are negative both for new and secondhand cars, and especially the willingness to accept (WTA) for the electric and fuel cell car is large. This is largely due to the limited driving ranges and long recharge/refueling times of 5 We also estimated a nested logit model, with conventional technology in a first tree and all AFVs in a second tree. Estimates and derived elasticities were very similar, both for the full sample and for the CT sample, and the two nesting coefficients were very similar and both close to one. Other nesting structures, for example with conventional technology, hybrid and plug-in hybrid in a first nest and all other AFVs in a second nest, gave comparable results. In conclusion, nested models do not appear to add much to our analyses.

207

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215 Table 6 Counts and percentages of car type choices made by respondents in the choice experiment, for the full and the CT sample. Car type

Full sample

Sample with CT in choice set

Count CT Hybrid Electric Plug-in hybrid Flexifuel Fuel cell Total

Percentage (%)

Count

Percentage (%)

7134 1009 1872 1002 1673 2531

47 7 12 7 11 17

7134 224 667 253 481 1009

73 2 7 3 5 10

15,221

100

9768

100

Table 7 Mixed logit model estimation results (monthly costs in Euro, purchase price in 1000 Euro). Full sample Beta

a

WTP new

Means of parameter distributions Hybrid Electric Plug-in hybrid Flexifuel Fuel cell Driving range electric (per km) Driving range fuel cell (per km) Recharge time electric (per minute) Recharge time plug-in hybrid (per min) Refueling time fuel cell (per min) Additional detour time (per min) Number of brands/models (per model) Free parking MRB exemption abolished Access to bus/taxi lanes

1.3432a 4.3908a 1.9846a 1.1621a 2.6648a 0.0063a 0.0024a 0.0030a 0.0048a 0.0222a 0.0286a 0.0007a 0.0461 0.2043a 0.0676

0.0781 0.1575 0.1228 0.0688 0.1788 0.0005 0.0004 0.0006 0.0017 0.0048 0.0025 0.0003 0.0566 0.0637 0.0556

€ € € € € € € € € € € € € € €

Fixed parameters Purchase price new cars Purchase price used cars Monthly costs new cars Monthly costs used cars

0.1223a 0.2289a 0.0047a 0.0077a

0.0076 0.0127 0.0004 0.0004

– – – –

Standard deviations of parameter distributions Hybrid 1.0439a Electric 2.2593a Plug-in hybrid 1.2138a Flexifuel 0.8860a Fuel cell 1.3821a Driving range electric 0.0032a Driving range fuel cell 0.0020a Recharge time electric 0.0038a Recharge time plug-in hybrid 0.0062b Refueling time fuel cell 0.0244 Additional detour time 0.0298a Number of brands/models 0.0039a Free parking 0.6444a MRB exemption abolished 0.5563a Access to bus and taxi lanes 0.4937a

0.1281 0.1024 0.1183 0.0982 0.1466 0.0009 0.0005 0.0009 0.0029 0.0132 0.0040 0.0004 0.0997 0.1333 0.1275

€ € € € € € € € € € € € € € €

NOBS Iterations completed Log-L Restricted Log-L Pseudo R2 (adjusted) b

CT sample SE

15,221 65 11,627 37,823 0.693

10,985 35,909 16,230 9504 21,793 52 19 24 39 182 234 5 377 1671 553

WTP 2nd hand € € € € € € € € € € € € € € €

5868 19,182 8670 5077 11,642 28 10 13 21 97 125 3 201 893 295

– – – – 8537 18,477 9927 7246 11,303 26 16 31 51 199 244 32 5270 4549 4038

€ € € € € € € € € € € € € € €

4560 9870 5303 3871 6038 14 9 17 27 107 130 17 2815 2430 2157

Beta

SE

WTP new

1.2201a 2.9233a 1.5592a 1.1583a 2.1634a 0.0047a 0.0024a 0.0015a 0.0029b 0.0116a 0.0137a 0.0001 0.0863 0.1495a 0.1057

0.0789 0.1427 0.1222 0.0710 0.1733 0.0006 0.0003 0.0004 0.0012 0.0042 0.0024 0.0003 0.0638 0.0740 0.0725

€ € € € € € € € € € € € € € €

0.1002a 0.1967a 0.0035a 0.0055a

0.0088 0.0144 0.0004 0.0003

– – – –

0.1407 0.5543a 0.2759 0.0149 0.4142a 0.0031a 0.0002 0.0010 0.0014 0.0037 0.0014 0.0027a 0.2102b 0.1567 0.4883a

0.1261 0.1342 0.1924 0.1014 0.0835 0.0005 0.0002 0.0006 0.0016 0.0043 0.0058 0.0005 0.0989 0.1587 0.1057

€ € € € € € € € € € € € € € €

12,180 29,182 15,565 11,563 21,597 47 24 15 29 116 136 1 862 1493 1055

WTP 2nd hand € € € € € € € € € € € € € € €

6202 14,860 7926 5888 10,997 24 12 7 15 59 69 1 439 760 537

– – – – 1405 5533 2754 148 4135 31 2 10 14 36 14 27 2099 1565 4875

€ € € € € € € € € € € € € € €

715 2818 1402 76 2106 16 1 5 7 19 7 14 1069 797 2482

9768 46 6986 24,273 0.712

Statistically significant at 1%. Statistically significant at 5%.

these car types. Focusing on new cars and the full sample, we see that for the electric car an increase in range is valued positively at 52 Euro per kilometer, implying that on average respondents are willing to pay around 4000 Euro for a doubling of the current range of electric vehicles of 75–150 km. WTP for increases in range of the fuel cell cars is lower, but this is mainly

208

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

the result of the higher starting value of the driving range level for this car type (250 km). Additional recharge time for the electric car is valued negatively at 24 Euro per minute. Interestingly, WTP for additional recharge time for the plug-in hybrid is more negative. This is counterintuitive since the plug-in car has an alternative fueling option besides electric charging. Due to this greater flexibility of the plug-in hybrid it would seem logical that the WTP for an increase in fuel time would be lower for the plug-in than for the electric car. Additional detour time due to limited recharge/refueling locations is valued negatively at 117 Euro per minute. The effects of increasing the number of available brands/models and the effects of policy incentives are relatively limited. Comparing values for new and secondhand cars shows similar patterns in preferences. The only differences in estimation results between new and secondhand cars were those on price and monthly costs. In other words, secondhand car buyers are more price sensitive than new car buyers, but preferences for other attribute levels are similar. The WTP for improvements in AFV characteristics turns out to be almost twice as low for people who indicate their next car will be a secondhand car than for people who indicate their next car will be a new car. These results are somewhat unexpected when we consider that older cars are typically driven less than new cars (even if we correct for the relatively high annual mileage of company cars) (CBS, 2013b). We would therefore expect that driving range limitations might be less of a problem for secondhand car buyers resulting ceteris paribus in higher WTP for the electric and fuel cell car. Since we find no evidence of this, the adoption of electric and fuel cell cars is likely to be similar for the new and secondhand car market. Table 7 also shows that there are interesting non-linear effects. The effects of driving range are lower for the fuel cell car than for the electric car, which suggests that the effect of increasing driving range is smaller at higher driving ranges. This is in line with results reported in the literature (see for example Dimitropoulos et al., 2011; Hidrue et al., 2011). The opposite holds true for the effects of recharge/refueling time reductions, that is, the effects are much larger for the fuel cell car than for the electric and plug-in hybrid car, while attribute levels for the fuel cell car are substantially smaller. 4.2. Effects of excluding the conventional technology: Short and long term marginal WTP estimates It is interesting that the differences in WTP estimates between the full and the CT sample are generally large, implying that choices made with respect to including or excluding a status quo alternative is non-trivial. AFV-specific constants for the full and CT sample are substantially different in most cases, but the estimates are not systematically lower or higher for either sample. The WTP for increases in driving range of the electric car is higher in the full sample, while WTP for increases in driving range of the fuel cell car is lower for the full sample. In both cases, however, the differences are small. Marginal WTP for improvements in recharge/refueling time and additional detour time are substantially higher for the full sample than for the CT sample. This holds for both the new and the secondhand market. Moreover, the differences are larger when WTP estimates are larger, that is, differences for refueling time of the fuel cell car and for additional detour time are the most substantial.6 This finding may not be very relevant in the short term when petrol and diesel engines will be the dominant technology. It may be relevant in the medium to longer term, when petrol and diesel cars would be largely replaced, arguably, by hybrid cars as the conventional technology. Modeling exercises and policies may benefit from using both sets of coefficients and WTP estimates, because they allow for short term but also for medium and long term market simulations and assessment of policy effectiveness. Differences between the two samples in terms of heterogeneity estimates are also very pronounced. For the full sample, estimated standard deviations of the random parameter distributions show that heterogeneity in preferences for most attributes is large and statistically significant at the usual critical significance levels. For the CT sample the picture changes. Judging by the WTP equivalents of the estimated standard deviations, preference heterogeneity estimates are much smaller for the CT sample than for the full sample, except for driving range of the electric car, and many are in this case statistically insignificant. Moreover, the adjusted R-squared from MNL model estimations is much higher for the CT sample (0.339) than for the full sample (0.253), while the difference is much smaller for the ML model. This suggests that accounting for heterogeneity has more added value in terms of variation explained for the full sample than for the CT sample. The underlying reasons for the differences in heterogeneity between the two samples are uncertain. It may be caused by an increase in random choices in the full sample. For example, respondents that systematically choose the conventional technology may be inclined to randomly choose an AFV when the conventional technology is not available as a choice option. It may also be caused by consumers being more doubtful about their preferences in absence of the conventional technology as a choice option, leading to larger differences in preferences over AFV attributes. In any case, the CT results show that preference heterogeneity for electric and fuel cell car constants, for driving range and recharge time of the electric car, for number of available models and for free parking and access to bus lanes, is large and statistically significant. This implies that for these characteristics potentially interesting segments and niche markets may be identified. 4.3. Effects of technological development and policy measures The coefficients in Table 7 allow us to calculate the WTP for improvements on AFV characteristics. For example, an increase in driving range of 100 km for a newly bought electric vehicle is valued at (100  € 47 =) € 4700. We use the CT sample 6

This pattern does not change when we estimate a model on a subsample of choices that do not contain the current technology.

209

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

here since, as we have argued in Section 4.2, the full sample is better suited to examine future preferences when the conventional technology has been replaced by for example hybrid cars. In this section the aim is to illustrate to what extent the gap in preferences between conventional technology and AFVs becomes smaller if technological improvements are made and what the possible effects of policy measures may be. Table 8 shows the impact on WTP for AFVs if we assume that the performance of AFVs increases substantially in terms of driving range, recharge/refueling time, additional detour time and the number of available models. If we focus on the electric car we see that an increase in range from the current 75 km to 350 km reduces WTA with almost 13,000 Euro. Reducing recharge times from 8 h to 30 min results in a further increase in WTP of 6600 Euro (assuming there is no additional detour time, that is, that sufficient recharge points are available). The impact of an increased number of models on WTP is limited on average. On the whole the negative valuation of the electric car can be reduced by more than 50%. The same holds for the fuel cell car for which driving range also has the biggest impact on total WTP. The figures in Table 8 also clearly indicate that for all AFVs the WTA can become much smaller due to improvements in performance and infrastructure, but does not disappear completely. This indicates that there are often large intrinsic negative preferences for AFVs compared to the conventional technology. This is particularly clear for the plug-in hybrid, which is surprising since it has all the benefits of the conventional car, some of benefits of the electric car (lower fuel costs) but not the disadvantage of a limited driving range. As such, policy makers should be forewarned that the plug-in hybrid might not be adopted so easily as the stepping stone to full electric vehicles as it is often assumed. Table 8 also shows the impact on WTP of abolishing the MRB (road tax) exemption that is currently in place for AFVs in the Netherlands. This policy measure decreases the WTA for all AFVs but is insufficient to overcome the intrinsic negative preferences. Coefficients for free parking and access to bus and taxi lanes are not statistically significant (see Table 7) and WTP values for these policy measures are therefore not included in Table 8. We should note that purchase price and monthly cost differences between conventional cars and AFVs have not been taken into account in Table 8. Currently, purchase costs of electric cars and particularly fuel cell cars are much higher than their current technology equivalent (measured in terms of car size and performance). On the other hand, electric and fuel cell cars are less expensive in terms of fuel, operation and maintenance costs. These purchase price and monthly cost differences would also have to be taken into account in order to assess future adoption potential of AFVs. However, future developments in purchase prices of AFVs and fuel prices for both the conventional technology and for AFVs are very uncertain. Moreover, AFV specific constants should ideally be calibrated making use of real world sales data, which are largely unavailable at this point in time. Realistic market simulations are therefore difficult to perform, and we leave this issue for future research. Nevertheless, the figures in Table 8 illustrate that monetary implications may be substantial for governments aiming to bridge the gap in preferences for conventional cars and AFVs through fiscal measures. With annual car sales of around 400,000 cars in The Netherlands, obtaining a share of 50% electric cars in new car sales would require a compensation of 3.5 billion euros per year.

Table 8 Impact of improvements on driving range, recharge/refueling time, additional detour time, number of models and abolishment of MRB exemption on AFV preferences and WTP (based on CT sample estimates).

a

Change

WTP for change

Residual WTP

Hybrid current WTP Models Policy

20–50 No MRB exemption

€ 34 € 1493

€ € €

12,157 12,123 13,616

Electric car current WTP Range Recharge timea Models Policy

75–350 km 480–30 min 10–50 No MRB exemption

€ 12,951 € 6602 € 46 € 1493

€ € € € €

36,773 23,822 17,220 17,175 18,667

Plug-in current WTP Recharge time Models Policy

180–20 min 10–50 No MRB exemption

€ 4634 € 46 € 1493

€ € € €

20,767 16,133 16,088 17,580

Flexifuel current WTP Detour time Models Policy

30–0 min 0–50 No MRB exemption

€ 4092 € 57 € 1493

€ € € €

15,655 11,563 11,506 12,999

Fuel cell current WTP Range Refueling time Detour time Models Policy

250–550 km 25–2 min 30–0 min 0–50 No MRB exemption

€ € € € €

7177 2658 4092 57 1493

€ € € € € €

22,596 15,420 12,762 8670 8614 10,106

Here we assume that fast charging an electric car in 30 min will not lead to additional detour time, that is, that sufficient infrastructure is available.

210

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

5. Market segmentations and early adopters Although mixed logit models are well suited to assess the magnitude of possible heterogeneity in consumer preferences, they do not reveal the underlying sources of heterogeneity. In this section we explore the existence of different market segments and try to identify potential early adopters. Estimating a latent class model would shed light on preferences of different latent classes. Our purpose is, however, to assess which people can be identified as potential early adopters. Inherently, latent class does not achieve this purpose. We therefore estimate a multinomial logit (MNL) model with interactions between choice experiment attributes and background and car use characteristics.7 The results for the full sample are presented in Table 9. As the results for the CT sample are very similar (unless mentioned otherwise in the text), they are not displayed separately. The main effects in Table 8 primarily serve as reference categories for the interaction effects. A first set of relevant interactions deal with the differences between respondents in annual mileage and car commuting behavior. With respect to the latter the results show that those who commute to and from work by car often (5 times or more per week) tend to have somewhat stronger negative preferences for AFVs in general. This might very well be explained by the fact that those who commute by car often, also run into the practical disadvantages of driving an AFV more often (such as finding a place to refuel and in case of electric vehicles the discomfort of having to plug in their vehicle). A factor that greatly affects preferences for electric and fuel cell cars is annual mileage. Higher annual mileage results in substantially more negative preferences for the electric car and, to a lesser extent, for the fuel cell car, implying that limited driving range is substantially less problematic for people with a relatively low annual mileage. We also see that the more people drive the larger the effect of increases in electric car driving range is, implying that preferences for the electric car for different groups of annual mileage converge when the driving range of the electric car increases. It is remarkable that we do not find this pattern for the fuel cell car. A second set of interactions show that negative preferences for the electric car are somewhat lower when the car is not used for going on holidays abroad, which is plausible since distances covered in going abroad are generally large, implying frequent recharging and waiting times for low driving ranges. Respondents that use the car for towing a caravan clearly have more negative preferences for AFVs in general than respondents who do not. This effect shows up for all AFVs and is likely the result of (some) respondents assuming that motor power of AFVs is lower than that of the conventional technology. Either that or uncertainty regarding AFV motor power is large among respondents, making them more hesitant to choose an AFV. Third, respondents who indicated that they have a permanent parking space close to their home with the possibility to charge an electric vehicle have a somewhat higher preference for the electric car. We should note that approximately 60% of respondents indicated they are able to charge close to their home. This is quite high and begs the question whether the question we presented to respondents was sufficiently clear. Free parking and access to bus lanes would be the most beneficial policies in highly urbanized areas, and although we do find such effects, they are small and statistically insignificant. In the CT sample access to bus lanes in highly urbanized areas has a more substantial positive effect on preferences. We therefore find some evidence that providing access to bus lanes may to some extent stimulate AFV adoption in highly urbanized areas. Fourth, the results show that diesel car drivers have slightly stronger negative preferences for electric cars and slightly stronger preferences for hybrid cars than petrol car drivers while LPG car drivers have stronger negative preferences than petrol car drivers for hybrid, plug-in hybrid and flexifuel cars. In all cases the differences in preferences between petrol, diesel and LPG drivers are relatively small. Finally, we tested several interactions on monthly costs and purchase price in order to assess whether some groups are more price sensitive than others. We find that the difference between those intending to buy a new car and those intending to buy a secondhand car is relevant, both in terms of monthly costs and purchase price. We therefore estimated further interactions for both new and secondhand car buyers. We find that differences between the first and second car in a household and between men and woman are relevant. Moreover, those who drive small cars and who indicate the intention to buy a relatively cheap car are more price and cost sensitive. Considering that fuel cost advantages of especially the electric car can be large when annual mileage is relatively high, people with small cars that drive a lot may benefit substantially from switching to an electric vehicle. However, people that drive a lot will also more often run into problems associated with driving range, making it difficult to assess whether this group is actually more likely to adopt electric and fuel cell cars in the future. Adding to the complexity is that there also appear to be differences in cost and price sensitivity between different annual mileage groups. Ultimately, whether people with low or high annual mileage are more likely to adopt electric and/or fuel cell cars therefore depends largely on future price, cost and driving range developments. Extensive market simulations are needed to assess the net effect of these developments.

7 For robustness purposes we also estimated a mixed logit model including heterogeneity in the means. The complexity of the mixed logit model specification is such that including all interactions from Table 6 is not possible. We therefore only estimate separate means for all AFV constants using the commuting dummy, and separate means on the electric and fuel cell car constants and driving range coefficients using the annual mileage dummies. Parameter estimates from this model show similar effects of annual mileage on electric car driving range, that is, more negative preferences with increased annual mileage and lower driving range. Higher annual mileage also results in more negative preferences for the fuel cell car. Stronger negative preferences for AFVs are found for those who often (>5 times per week) commute to and from work by car, although parameter estimates are not statistically significant at 5%. Results are available upon request of the authors.

211

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215 Table 9 MNL estimation results for a model with interaction effects (monthly costs in Euro, purchase price in 1000 Euro). Full sample Beta

SE (beta)

Main effects (reference for interaction effects) Hybrid Electric Plug-in hybrid Flexifuel Fuel cell Driving range electric Driving range fuel cell Recharge time electric Recharge time plug-in Refueling time fuel cell Additional detour time Number of brands/models Free parking MRB exemption Access to bus lanes Monthly costs new car Monthly costs used car Purchase price new car Purchase price used car

0.8587a 3.0912a 1.7162a 1.3463a 2.0283a 0.0018 0.0020a 0.0012a 0.0026a 0.0147a 0.0161a 0.0006a 0.0962b 0.0653 0.0199 0.0031a 0.0043a 0.0672a 0.1127a

0.0854 0.1852 0.1075 0.0769 0.1557 0.0009 0.0007 0.0002 0.0006 0.0030 0.0015 0.0002 0.0398 0.0434 0.0392 0.0003 0.0003 0.0065 0.0117

Interactions annual mileage and commuting Hybrid  commute P5 times per week Electric  commute P5 times per week Plug-in hybrid  commute P5 times per week Flexifuel  commute P5 times per week Fuel cell  commute P5 times per week Electric  (7500 < Yearly km < 15,000) Electric  (15,000 < Yearly km < 25,000) Electric  (25,000 < Yearly km < 35,000) Electric  (yearly km > 35,000) Fuel cell  (7500 < yearly km < 15,000) Fuel cell  (15,000 < yearly km < 25,000) Fuel cell  (25,000 < yearly km < 35,000) Fuel cell  (yearly km > 35,000) Driving range electric  (7500 < yearly km < 15,000) Driving range electric  (15,000 < yearly km < 25,000) Driving range electric  (25,000 < yearly km < 35,000) Driving range electric  (yearly km > 35,000) Driving range fuel cell  (7500 < yearly km < 15,000) Driving range fuel cell  (15,000 < yearly km < 25,000) Driving range fuel cell  (25,000 < yearly km < 35,000) Driving range fuel cell  (yearly km > 35,000)

0.1929b 0.1012 0.1958b 0.1637b 0.1748b 0.4636a 0.8562a 0.9324a 1.4541a 0.3516b 0.2085 0.6241a 0.7478a 0.0016 0.0025a 0.0023b 0.0037a 0.0006 0.0004 0.0005 0.0012

0.0968 0.0772 0.0897 0.0788 0.0703 0.1656 0.1630 0.1922 0.2228 0.1641 0.1603 0.1912 0.2161 0.0009 0.0010 0.0011 0.0012 0.0008 0.0008 0.0009 0.0010

Interactions holidays Hybrid  caravan Electric  caravan Plug-in hybrid  caravan Flexifuel  caravan Fuel cell  caravan Electric  car is not used for holidays

0.4053a 0.1886b 0.3270a 0.0615 0.2712a 0.2757a

0.1253 0.0935 0.1162 0.0985 0.0854 0.0684

0.1219 0.0769 0.2397b 0.0004

0.0833 0.0842 0.0997 0.0006

0.1910 0.2147a 0.1068 0.0769 0.1078 0.3153b 0.0115 0.3339a 0.2189b 0.0168

0.1033 0.0796 0.0962 0.0845 0.0720 0.1224 0.0882 0.1117 0.0965 0.0800

Interactions on recharging potential and policy measures Free parking  very urbanized area Access to bus lanes  very urbanized area Electric  recharging potential at home Driving range electric  recharging potential at home Interactions on current fuel type Hybrid  current fuel diesel Electric  current fuel diesel Plug-in hybrid  current fuel diesel Flexifuel  current fuel diesel Fuel cell  current fuel diesel Hybrid  current fuel LPG Electric  current fuel LPG Plug-in hybrid  current fuel LPG Flexifuel  current fuel LPG Fuel cell  current fuel LPG

(continued on next page)

212

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

Table 9 (continued) Full sample Beta Interactions on monthly costs Monthly costs new  price next car < 6000 Euro Monthly costs new  weight car < 1000 kg Monthly costs new  2nd car in household Monthly costs new  respondent is female Monthly costs new  (7500 < yearly km) Monthly costs used  price next car < 6000 Euro Monthly costs used  weight car < 1000 kg Monthly costs used  2nd car in household Monthly costs used  respondent is female Monthly costs used  (7500 < yearly km)

0.0032a 0.0038a 0.0028a 0.0018a 0.0008 0.0013a 0.0030a 0.0002 0.0015a 0.0000

0.0012 0.0008 0.0010 0.0006 0.0007 0.0004 0.0007 0.0006 0.0005 0.0005

Interactions on purchase price Purchase price new  price next car < 6000 Euro Purchase price new  weight car < 1000 kg Purchase price new  2nd car in household Purchase price new  respondent is female Purchase price new  (7500 < yearly km) Purchase price used  price next car < 6000 Euro Purchase price used  weight car < 1000 kg Purchase price used  2nd car in household Purchase price used  respondent is female Purchase price used  (7500 < yearly km)

0.0460 0.0450a 0.0310 0.0135 0.0098 0.0729a 0.0404 0.0020 0.0504b 0.0132

0.0323 0.0160 0.0226 0.0142 0.0109 0.0176 0.0240 0.0238 0.0214 0.0170

NOBS Log-L Restricted Log-L Pseudo R2 (adjusted) a b

SE (beta)

15,221 12,228 16,690 0.265

Statistically significant at 1%. Statistically significant at 5%.

6. Conclusion and discussion In this paper we have presented the results from an online stated choice experiment among Dutch private car owners. The aim has been to gain insight into the preferences of private car owners for AFVs and AFV characteristics in the Netherlands and to identify relevant market segments and potential early adopters. An important added value of this study is its comprehensiveness (most of the relevant attributes and car types are included in one experiment) and the contribution to empirical evidence for the European case, which so far has been underrepresented in existing studies. Results show that, on average and assuming current AFV characteristics, preferences for AFVs are substantially lower than those for the conventional technology. Limited driving range, long refueling times and limited availability of refueling opportunities are to a large extent responsible for this. These barriers are the most substantial for the electric car, and to a lesser extent for the fuel cell car. The average preference for AFVs increases considerably if improvements in driving range, refueling time and additional detour time are made. An increase in the number of available models from which a consumer can choose and measures such as free parking have a positive effect as well, but to a far lesser extent. The results also show that secondhand car buyers are roughly twice as price sensitive as new car buyers, while preferences for other attribute levels are very comparable for buyers of new and secondhand cars. Hence, the willingness to pay (WTP) for driving range for the electric and fuel cell car is similar for new and secondhand car buyers. This is unexpected if we take into account that older cars are driven less than new cars. We would therefore expect higher WTP of secondhand car buyers for electric and fuel cell cars. These results imply that the adoption of electric and fuel cell cars is likely to be similar for the new and secondhand car market. Insights for secondhand cars are relevant when we realize that most cars sold on the lease market are resold on the private secondhand market after 3 or 4 years, and that current tax incentives for buying AFVs are very favorable on the lease market in many European countries (ACEA, 2012). If we assume that the performance of AFVs increases substantially in terms of driving range, recharge/refueling time, additional detour time and the number of available models, negative preferences for all AFVs remain. The Dutch road tax exemption for AFVs adds to WTP but is insufficient to overcome the intrinsic negative preferences. Monetary implications may be substantial for governments that choose to bridge the gap in preferences for conventional cars and AFVs through fiscal measures. Marginal WTP for improvements in recharge/refueling time and additional detour time are substantially higher for the full sample than for a sample in which respondents could always choose at least one car with the conventional technology. This may be relevant in the medium to long term, where petrol and diesel cars could be largely replaced, arguably, by hybrid

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

213

cars as the conventional technology. Modeling exercises and policies may benefit from using both sets of coefficients and WTP estimates, because they allow for short term but also for medium and long term market simulations and assessments of policy effectiveness. With respect to market segmentation several variables, such as commuting behavior, using the car for holidays abroad, gender, weight and price of the next car, appear to be important. With respect to heterogeneity in preferences for the electric and fuel cell car, annual mileage is by far the most important factor. More specifically, preferences of people with a low annual mileage are far less negative than preferences of people with a high annual mileage. The main explanation for this pattern is that those who drive more run into problems of limited driving range sooner and more often. Because kilometers driven during a single day will often exceed the maximum electric driving range for this group, recharging would have to take place not only at night but also somewhere in between trips. Recharging potential and recharge time are limiting factors in that respect as well. Annual mileage also has a large impact on WTP for increases in driving range of the electric car, implying that preferences for the electric car of different annual mileage groups converge when electric car driving range increases. On the other hand, people who drive more also appear to be more sensitive to monthly costs, so ultimately the market potential of electric and fuel cell cars for different annual mileage groups depends largely on the net effect of future price, cost and driving range developments. The substantial preference heterogeneity and the intrinsic negative preferences for AFVs found in this study suggest that policies targeting specific groups may be more effective in speeding up AFV adoption than generic policy measures. Policy makers should be particularly aware that car users with low annual mileages are likely to be early adopters. Identifying people with a low annual mileage is possible when a road pricing system would be introduced, because such a system would require monitoring the distance driven by each car. Analogous to classic price discrimination, people with a low annual mileage could be targeted by introducing mileage-dependent AFV subsidies or tax advantages, that is, purchase price incentives that are higher for people with a low annual mileage than for people with a high annual mileage. The insight to target low mileage car users is not trivial since AFVs’ relatively high up-front investment costs and relatively low running costs would favor early adoption by those with high annual mileages (targeting this group is in fact part of public policy in the Netherlands). However, those with high annual mileages are more likely to run into driving range issues which according to our findings weighs more negatively on preferences than purchase cost differences. The purchase price differences between conventional cars and AFVs add to the substantial compensations required for AFVs, particularly for the electric and fuel cell car. If governments would aim for increased market shares of AFVs these cost differences could be partly bridged through either subsidies on AFVs or taxes on the current technology, in addition to the improvements on range, refueling/recharge times and fuel availability. We do note that fiscal incentives should be strong if the aim is to overcome the current negative preferences for AFVs. The strong negative preference for limited driving range (and to a lesser extent long recharging times) also implies that policy makers may want to focus on the development of improved battery technology and fast charging. Our results furthermore show that under current circumstances consumers view the electric car as the most unappealing choice among AFVs. In line with the often heard claim that technology neutrality is a prerequisite for effective policies (see, for example, Kemp, 1994; Bakker, 2010; Azar and Sandén, 2011), this finding suggests that current policies in, among others, the Netherlands, France and Germany, which are aimed specifically at stimulating electric car adoption, may be suboptimal and require a broader technology perspective.

Acknowledgements We gratefully acknowledge comments and suggestions by Rogier Kuin (BOVAG) on a first draft of the choice experiment set-up and questionnaire. Also comments and suggestions by Piet Rietveld (VU University), Bert van Wee (Delft University of Technology), Herman Vollebergh and Gerben Geilenkirchen (PBL Netherlands Environmental Assessment Agency) on a first draft of this paper are greatly appreciated.

Appendix A. Descriptive texts on attributes as presented to respondents Car type Full electric car: a car that is set in motion by an electric motor. Batteries provide the electric motor with energy. The car must be charged to be able to drive it and electricity from a socket is suitable. Fuel-cell car: also called hydrogen car. A car that requires to be fueled with hydrogen in order to be able to drive it. In the car the hydrogen is converted into electricity with fuel cells. An electric motor sets the car in motion. Plug-in hybrid: a car with both a petrol or diesel engine and batteries. The batteries can be charged with a plug and the car drives several tens of kilometers solely on electricity. When the batteries are empty the car will switch to using petrol/diesel. It is thus also possible to drive solely on petrol/diesel. Flexifuel car: a car that, besides petrol or diesel, can drive completely on biofuels (fuels manufactured from biological materials). It could be biodiesel, bioethanol (comparable to petrol) or biogas (comparable to natural gas).

214

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

Hybrid: a car with batteries but without a plug. The engine in the car charges the batteries during driving and braking energy is recovered as well to charge the batteries. A hybrid can drive several kilometers solely on electricity. Monthly costs A combination of fuel costs (tailored to your mileage), maintenance costs and, if applicable, road taxes. Driving range The number of kilometers you can drive at most on a full tank or fully charged batteries (in case of an electric car). Recharge/refueling time The time it takes to fully charge the car (electric or plug-in hybrid) or to fill your tank. NB. the time shown at the plug-in hybrid applies to charging time of the batteries. Additional detour time In the case that not every fueling station offers the fuel your car drives on it may be that you have to drive further to be able to refuel. As the availability of the fuel for the concerning car gets lower, the additional detour time is greater. Number of brands/models The larger the number of models, the more alternatives there are for this car type. This concerns different brands and models, and different versions regarding engine size, acceleration and size of the car. Policy measure for this car type Concerns policies with which the government aims to influence the sales of this car type. We distinguish (1) current policy, (2) free parking, which applies to both parking permits and parking zones, (3) abolishment of the road tax exemption (monthly costs are corrected for this), and (4) permission to drive on bus and taxi lanes within the built-up area. The policy only applies to the car type for which it is shown. When for example the electric car option shows ‘Free parking’, this policy measure only applies to electric cars and not to the other AFVs. When ‘Current policy’ is shown all government policies are equal to the current situation. References ACEA, 2012, ACEA Tax Guide 2012. European Automobile Manufacturers Association (ACEA), Brussels, Belgium. Ahn, J., Jeong, G., Kim, Y., 2008. A forecast of household ownership and use of alternative fuel vehicles: a multiple discrete–continuous choice approach. Energy Econ. 30, 2091–2104. Azar, C., Sandén, B.A., 2011. The elusive quest for technology-neutral policies. Environ. Innov. Soc. Transit. 1, 135–139. Bakker, S., 2010. The car industry and the blow-out of the hydrogen hype. Energy Policy 38, 6540–6544. Batley, R.P., Toner, J.P., Knight, M.J., 2004. A mixed logit model of U.K. household demand for alternative-fuel vehicles. Int. J. Transp. Econ. 31, 55–77. Beck, M.J., Rose, J.M., Hensher, D.A., 2011. Behavioural responses to vehicle emissions charging. Transportation 38, 445–463. Beggs, S., Cardell, S., Hausman, J., 1981. Assessing the potential demand for electric cars. J. Econometrics 16, 1–19. Brownstone, D., Bunch, D.S., Train, K., 2000. Joint mixed logit models of stated and revealed preferences for alternative-fuel vehicles. Transport. Res. Part A 34, 315–338. Bunch, D.S., Bradley, M., Golob, T.F., Kitamura, R., Occhiuzzo, G.P., 1993. Demand for clean-fuel vehicles in California: a discrete-choice stated preference pilot project. Transport. Res. Part A 27, 237–253. Bunch, D.S., Brownstone, D., Golob, T.F., 1995, A dynamic forecasting system for vehicle markets with clean-fuel vehicles. Working Paper UCI-ITS-WP-95-21, Institute of Transportation Studies, University of California, Irvine, USA. Calfee, J.E., 1985. Estimating the demand for electric automobiles using fully disaggregated probabilistic choice analysis. Transport. Res. Part B 19, 287–301. Carlsson, F., Johansson-Stenman, O., Martinsson, P., 2007. Do you enjoy having more than others? Survey evidence of positional goods. Economica 74, 586– 598. Caulfield, B., Farrell, S., McMahon, B., 2010. Examining individuals preferences for hybrid electric and alternatively fuelled vehicles. Transp. Policy 17, 381– 387. CBS, 2013a. Statline Web Publication. CBS Statistics Netherlands, The Hague. . CBS, 2013b, Statline Web Publication. . Chrzan, K., Orme, B., 2000. An overview and comparison of design strategies for choice-based conjoint analysis. Sawtooth Software Research Paper Series. Sequiem, USA. . Dagsvik, J.K., Liu, G., 2009. A framework for analyzing rank-ordered data with application to automobile demand. Transport. Res. Part A 43, 1–12. Dagsvik, J.K., Wetterwald, D.G., Aaberge, R., 1996. Potential Demand for Alternative Fuel Vehicles. Discussion Paper No. 165, Statistics Norway, Oslo, Norway. Dimitropoulos, A., Rietveld, P., Van Ommeren, J.N., 2011. Consumer Valuation of Driving Range: A Meta-analysis, Tinbergen Institute Discussion Paper 2011133/3. Tinbergen Institute, Amsterdam. Eberle, U., Müller, B., Von Helmolt, R., 2012. Fuel cell electric vehicles and hydrogen infrastructure: status 2012. Energy Environ. Sci. 5, 8780–8798.

A. Hoen, M.J. Koetse / Transportation Research Part A 61 (2014) 199–215

215

European Commission, 2011. Roadmap to a Single European Transport Area: Towards a Competitive and Resource Efficient Transport System (White Paper), COM (2011) 144 Final, Brussel. Ewing, G.O., Sarigöllü, E., 1998. Car fuel-type choice under travel demand management and economic incentives. Transport. Res. Part D 3, 429–444. Grüning, M., Witte, M., Boteler, B., Kantamaneni, R., Gabel, E., Bennink, D., Van Essen, H., Kampman, B., 2011. Impacts of Electric Vehicles – Deliverable 3: Assessment of the Future Electricity Sector. Publication Number 11.4058.05, CE Delft, Delft, The Netherlands. Hensher, D.A., Greene, W.H., 2001. Choosing between conventional, electric and LPG/CNG vehicles in single-vehicle households. In: The Leading Edge of Travel Behaviour Research. Pergamon Press, Oxford, pp. 725–750. Hess, S., Train, K.E., Polak, J.W., 2006. On the use of a modified Latin hypercube sampling (MLHS) method in the estimation of a mixed logit model for vehicle choice. Transport. Res. Part B 40, 147–163. Hidrue, M.K., Parsons, G.R., Kempton, W., Gardner, M.P., 2011. Willingness to pay for electric vehicles and their attributes. Resour. Energy Econ. 33, 686–705. Hoen, A., Geurs, K.T., 2011. The influence of positionality in car-purchasing behaviour on the downsizing of new cars. Transport. Res. Part D 16, 402–408. Horne, M., Jaccard, M., Tiedemann, K., 2005. Improving behavioral realism in hybrid energy-economy models using discrete choice studies of personal transportation decisions. Energy Econ. 27, 59–77. Kemp, R., 1994. Technology and the transition to environmental sustainability: the problem of technological regime shifts. Futures 26, 1023–1046. Mabit, S.L., Fosgerau, M., 2011. Demand for alternative-fuel vehicles when registration taxes are high. Transport. Res. Part D 16, 225–231. Maness, M., Cirillo, C., 2012. Measuring future vehicle preferences stated preference survey approach with dynamic attributes and multiyear time frame. Transport. Res. Rec.: J. Transport. Res. Board, No. 2285, 100–109, Transportation Research Board of the National Academies, Washington, D.C.. Mau, P., Eyzaguirre, J., Jaccard, M., Collins-Dodd, C., Tiedemann, K., 2008. The ‘Neighbor Effect’: simulating dynamics in consumer preferences for new vehicle technologies. Ecol. Econ. 68, 504–516. PBL, 2009. Getting into the Right Lane for 2050. PBL Netherlands Environmental Assessment Agency, The Hague, The Netherlands. Potoglou, D., Kanaroglou, P.S., 2007. Household demand and willingness to pay for clean vehicles. Transport. Res. Part D 12, 264–274. Thomas, C.E., 2009. Fuel cell and battery electric vehicles compared. Int. J. Hydrogen Energy 34, 6005–6020. Train, K.E., 2008. EM algorithms for nonparametric estimation of mixing distributions. J. Choice Modell. 1, 40–69. Von Helmolt, R., Eberle, U., 2007. Fuel cell vehicles: status 2007. J. Power Sources 165, 833–843. Zhang, T., Gensler, S., Garcia, R., 2011. A study of the diffusion of alternative fuel vehicles: an agent-based modeling approach. J. Prod. Innov. Manage 28, 152–168. Ziegler, A., 2012. Individual characteristics and stated preferences for alternative energy sources and propulsion technologies in vehicles: a discrete choice analysis for Germany. Transport. Res. Part A 46, 1372–1385.