- Email: [email protected]
How may incentives for electric cars affect purchase decisions?
Transport Policy 52 (2016) 113–120
Contents lists available at ScienceDirect
Transport Policy journal homepage: www.elsevier.com/locate/tranpol
How may incentives for electric cars affect purchase decisions?$ Christian Rudolph German Aerospace Center (DLR), Institute of Transport Research, Rutherfordstrasse 2, 12489 Berlin, Germany
art ic l e i nf o
a b s t r a c t
Article history: Received 3 March 2016 Received in revised form 23 June 2016 Accepted 29 July 2016
In this paper, the impact of five different incentives for buyers of zero emission vehicles (ZEV) is investigated with a stated choice experiment. The tested incentives are direct subsidies, free parking, a separate CO2 tax, an increase of fuel costs by tax elevation, and an increase of available charging infrastructure. By implementing the mobility patterns of the respondents, it was possible to simulate estimations of ecological impact and modal shift with a random utility model (mixed logit). Based on 875 complete questionnaires, the simulation results show that giving incentives to these buyers ecological rebound effects are expected: Mostly people with a low CO2-emission rate regarding their daily transportation routines (cyclists and public transport users) will exploit these incentives. They show a significantly higher likelihood of choosing alternatively propelled cars than conventional car users. Consumers that usually use a passenger car for their daily mobility routines are mostly unwilling to change to ZEV even when incentives are given. & 2016 Elsevier Ltd. All rights reserved.
Keywords: Choice modeling Alternative propulsion technology Battery electric vehicles BEV Discrete choice Stated preference Choice-based conjoint analysis
1. Introduction In the fight against global warming and forthcoming oil shortage zero emission vehicles (ZEV) such as battery electric vehicles (BEV – running only on electricity from the grid), plug-in hybrid electric vehicles (PHEV – running on electricity from the grid and on fossil fuel) or fuel cell electric vehicles (FCEV – running on hydrogen) have come into the focus of transport policies in order to decrease greenhouse gas emissions from the transportation sector. The German government for example established the German Federal Development Scheme for Electric Mobility (Federal Republic of Germany, 2009) in 2009. With this scheme the government set the ambitious target of ‘one million electric cars on German roads by 2020’. With a traditionally strong automobile industry, Germany wants to position itself as the market leader of electric cars (Federal Republic of Germany, 2009). But, even years after introduction of this scheme, purchases of ZEV show small numbers: only about 19,000 BEV and 33,000 HEV are on Germany's streets in the end of 2014 (Federal Motor Transport Authority, 2015). Due to the slow market penetration of ZEV, a debate on implementing incentive schemes to stimulate the demand for electric cars arose. Different stimulation schemes have been ☆ A discrete choice study investigating the ecological impact and modal shifts due to governmental incentives fostering the market penetration of zero emission vehicles in Germany. E-mail address: [email protected]
http://dx.doi.org/10.1016/j.tranpol.2016.07.014 0967-070X/& 2016 Elsevier Ltd. All rights reserved.
introduced in other European and non-European countries. The variety of incentives as push and pull measures is broad: tax exceptions, privileges (free parking, using bus and taxi lanes, carpooling lanes,1 etc.), direct purchase subsidy, introduction of CO2free zones are the most frequently discussed. International approaches differ strongly: While countries such as Denmark and Norway subsidize ZEV purchases and liberate from tax, countries such as Germany used to support only research and development projects. In some countries a strong increase of purchase numbers of ZEV can be observed where subsidies (buying grants and tax reductions) are given, e.g. Norway (Breivik and Volder, 2014). A recent global review about incentives for plug-in electric vehicles is given in (Zhou et al., 2015). But, purchase decisions have not been investigated in dependence of the buyer's mobility behavior. Studies so far investigate elasticies of incentives, e.g. (Jenn, 2014) or technological attributes in order to predict general sales numbers of ZEV. They do not account for the energy ratio for mobility of the buyers. This paper gives first insights about possible ecological effects of different hypothetical incentives fostering ZEV in combination with the mobility behavior of the respondents. Subsidies on the one hand may increase sales of ZEV. But on the other hand, from the ecologic perspective vehicles with a combustion engine should be exchanged by ZEV and drivers should preferably have a high 1
USA and Canada specific.
114
C. Rudolph / Transport Policy 52 (2016) 113–120
annual mileage in order to reduce the energy ratio for mobility. This is very important since electricity from the grid in Germany still is not 100% renewable. That means also ZEV emit greenhouse gases in a well-to-wheel consideration. The energy ratio would even worsen if incentives would bring cyclists and users of public transport (PT) to use ZEV since they become affordable. Additionally, these cars are considered as environmental friendly. Especially these groups showing a very low energy ratio for mobility should not be motivated to buy an individual passenger car. Sales of ZEV should not be additive but replace existing conventional vehicles. Against the background of the current discussion in Germany about the effects of incentives for ZEV and which incentives should be implemented, an online stated preference experiment which is representative for Germany was conducted.
2. Research design An online survey with 1010 respondents was conducted in the metropolitan area of Hamburg, Germany, which resulted in 875 completely finished questionnaires. Hamburg was chosen due to its spatial diversity2 (e.g. inner city and rural areas). In order to simulate hypothetical purchase decisions, a stated preference (SP) experiment using a choice-based conjoint analysis (CBCA) was conducted. The respondents were only selected by their socio economic and socio geographic attributes. The circumstance if a respondent was about purchasing a new car in the near future was not of interest, since it would bias the results against the background of the research interest investigating the effect of incentives across all citizens using different means of transport. An example of a choice set is given below in Fig. 1. The respondents' purchasing decisions are simulated with a random parameter utility model (mixed logit) to account for dependence across observations from the same respondent (Small et al., 2005). Logit models are applied to determine consumers' ‘tastes’ towards certain products, services or attributes of products (Brownstone et al., 2000; McFadden and Train, 2000; Lave and Train, 1979; Hensher et al., 2008; Train, 1998, 1986, 1980; Akiva and Lerman, 1985; Manski and Sherman, 1980). The estimation results simulated by mixed logit models should approximate reality more accurately than traditional conditional logit or multinomial logit models due to the ability to implement individual heterogeneities amongst respondents. The applied model is based on Revelt and Train (1998). To estimate the β-coefficients the mixed logit module by Hole (2007a) is applied. Hole's model uses a maximum likelihood simulation to estimate the coefficients (Hole, 2007b). For a detailed description of the methodology of the simulated maximum likelihood method please refer to (Hole, 2007a, 2007b; Train, 2003). Besides the knowledge of the preferred propulsion technology and the importance of the different incentives, it is necessary to know the mobility patterns of the respondents to draw conclusions from their decision to their future mobility. Therefore the respondents are asked during the survey about their daily mobility routines, i.e. yearly mileage (higher or lower than 15,000 km/year), possession of annual transit pass, use frequency of bicycles (daily use, sometimes, never), use frequency of PT (daily use, sometimes, never). Four groups with different mobility patterns could endogenously be clustered: car users, PT users, Multimodal users – with car affinity, Multimodal users – with PT affinity. By applying different incentives to the model, the impact of each incentive on the choice of propulsion technology could be estimated separately for each mobility group. Due to methodology 2
In further studies also spatial effects will be investigated.
the results are based on hypothetical assumptions given by the respondents, not the revealed factual behavior. In order to identify the attributes (here: purchase incentives) which should be investigated, a literature review about the state of the art survey design gave valuable hints. Green and Srinivasan (1978) state that not more than five to six different attributes should be tested within a discrete choice experiment, while Thomas (1983) report that up to 20 attributes can be applied. Ito et al. (2013) uses a CBCA to identify the willingness to pay for alternatively propelled vehicles in Japan and the resulting cost to install the charging infrastructure. He examines nine attributes: vehicle type, brand, driving range, recharging time, CO2 emission, availability of charging infrastructure, price, and operating costs. For Canada, Potoglou and Kanaroglou (2006) conducted a discrete choice study with nearly the same attributes. The model results show that direct purchasing subsidies and the emission rate of CO2 have the highest influence on the purchasing decision. Ewing and Sarigöllü (2000) conducted a CBCA (N ¼881) with commuters to examine the importance of price, costs for maintenance, acceleration, recharging time, driving range, emissions, privilege to use car-pooling lanes, fuel costs, and parking costs. They show that these measures will have only a small impact on the purchasing decision towards ZEVs and come to the conclusion to foster research and development projects instead. Axsen et al. (2009) conducted an SP analysis in California and Canada to investigate governmental incentives for HEVs, which are not in the scope of this paper. Daziano and Achtnicht (2012), Achtnicht (2009), Achtnicht et al. (2008) and Ziegler (2012) applied different models to a survey conducted in 2008 using an SP experiment with six tasks each showing seven different concepts. They investigated price, fuel costs per 100 km, engine performance, CO2 emissions, availability of charging infrastructure and type of fuel. Raich et al. (2012) finds out in a survey conducted in Austria using an SP experiment that cost-affected attributes have a much higher influence on the purchasing decision than technological attributes (such as driving range, recharging time, emission rate). Mabit and Fosgerau (2011) also conducted an SP experiment in Denmark. The experiment had 12 tasks each with five alternatives to choose from. They examined propulsion technology, fuel costs, price, driving range, number of charging operations per week, acceleration and further operating costs. The survey conducted by Dagsvik et al. (2002) with respondents from Norway comprised 15 tasks each with three vehicle concepts. They investigated price, maximum speed, driving range and energy consumption. Knockaert (2005) used in his study price, operating costs, fuel consumption, propulsion technology, driving range, emissions, volume of trunk. The literature review shows that all studies focus on investigating purchase decisions based on technological attributes of the vehicles. Purchase models are developed and the development of ZEV sales can be predicted depending of developments of these attributes. No study examines the purchase decision combined with the mobility behavior of the respondents. Therefore, rather very general statements can only be given about the ecological effect and the overall energy consumption depending from the mobility patterns of the respondents. The literature review shows also that the CBCA is the most common SP experiment design. Therefore, a CBCA is also applied for this study. Studies of the literature review show that best results are achieved with not more than three alternatives to choose from and the ‘none option’ (‘I would purchase none of the alternatives.’). Each respondent had to complete eight similar choice tasks in sequence with three alternatives and the ‘none option’, levels of attributes are changed each time. shows a sample of the display.
C. Rudolph / Transport Policy 52 (2016) 113–120
115
Fig. 1. Display of choice sets as shown in survey.
2.1. Attributes and levels used in survey Due to the high influence of brand, design, and production country on the purchasing decision, those attributes are not shown. In consequence a text-based analysis is chosen to present the choice sets. Finally a combination of the findings resulting from the literature review and the current political discussion in Germany led to the setting of the attributes and their levels. Since incentives are within the scope of this study, technological aspects were reduced to the propulsion technology. The survey investigates the purchasing decision when price, fuel costs (depending on the individual yearly mileage3), a separate CO2 tax for combustion vehicles, direct purchasing subsidies, parking fees, and the availability of charging infrastructure varies between BEV, PHEV, FCEV and ICV (internal combustion vehicles). Table 1 shows an overview of the tested attributes and the level variation. 2.2. Descriptive analysis of the respondents The data was collected by an online access panel provider. Since panel providers know comprehensive characteristics of their members, a representative study can be realised efficiently. The survey was conducted between 06/17/2013 and 06/26/2013. In order to represent the metropolitan area of Hamburg, only participants in this area were addressed. Socio-economic and sociogeographic characteristics were queried in the first part of the survey. The ratio of people living in urban and rural areas (1/2) complies with the allocation of the metropolitan area of Hamburg. In terms of sex, age, education, type of employment, net income, and population density, the sample represents the population of the metropolitan area of Hamburg very well and the population of Germany closely. An overview of the sample is given in Table 2. Emails were sent to 1010 people to participate in the survey. The survey resulted in 875 completely finished questionnaires which could be used for further analyses. The respondents were also asked if they possess a car or have access to a car, if they have a driving license, and if they have an annual transit pass for the public transport system. In terms of car access, 7.7% of the respondents do not own a car but only 3.3% of the respondents do not have any access to a car; 16.2% share a car with other household members, friends or use car sharing; and 80.5% have unlimited access to a car. A driving license was in the 3
The mileage was asked in the beginning of the survey.
Table 1 Attributes and their levels. Attribute
[No. of levels] level
Propolsion technology Price [€]
4 4
Fuel/charging costs [€/year]
3
CO2 tax [€/year] Direct subsidy [€] Parking fee [€/h] Distance to charging point [min]
4 5 4 7
BEVa, PHEVb, FCEVc, ICVd 30,000, 35,000, 40,000, 45,000 plus 20%, minus 50%, minus 65%e 0, 200, 300, 400 0, 1,000, 3,000, 5,000, 10,000 0, 3, 4, 5 0.5f, 1, 5, 10, 15, 20, 30
a
BEV¼ battery electric vehicle. PHEV ¼plug-in hybrid electric vehicle. c FCEV ¼ fuel cell electric vehicle. d ICV ¼ internal combustion engine vehicle. e Calculation according to information about the yearly mileage given by the respondent. In case no information was given, the average of 15,000 km was applied. f Walking distance to next charging infrastructure (availability of charging infrastructure). b
possession of 97.1% in total. About one fourth (25.5%) have an annual transit pass. Sex, age and the spatial allocation of the respondents comply very well with the general German statistics. The education level of the sample is generally higher than in general. Also the net household income is a bit higher than generally across Germany. This circumstance must be considered when giving recommendations. For further analysis the data will be used as is. No weights will be applied.
3. Estimation results The analysis contains the estimation of two models. Firstly, only the coefficients of the attributes are estimated (attributesonly model). With this model the effect of the different described incentives can be measured. Secondly, interaction variable that integrate daily mobility routines into the model are added. The model estimations are based on 28,000 observations from 7000 tasks (875 respondents; 8 tasks per respondent). 3.1. Attributes-only model According to Rabe-Hesketh and Skrondal (2012) and Train
116
C. Rudolph / Transport Policy 52 (2016) 113–120
Table 2 Sample description (n ¼ 875). Socio-demographic and socio-economic features of sample
Sample (n¼875) (%)
Germanya (%)
Sex Female Male
50.4 49.6
51.1 48.9
Age 18–24 25–34 35–44 45–54 55–65
12.6 19.5 23.9 25.1 18.9
13.1b 19.6b 19.8b 26.2b 21.3b
0.1 11.3 40.6
7.6c 37.0c 28.9c
48.0
25.8c 0.4c
2.3 7.2 12.0 15.3 32.6 19.4 11.2
2.0d 7.0d 18.0d 16.0d 23.0d 12.0d 1.0d
29.9 70.1
e
Education Not yet/ no school graduation Certificate of Secondary Education General Certificate of Secondary Education University-entrance diploma Not specified Net income of household [€/month] Below 500 500–1.000 (500–900)b 1000–1500 (900–1500)b 1500–2000 2000– 3000 3000–4000 more than 4000 Allocation within the metropolitan area of Hamburg City of Hamburg (urban) Metropolitan Region of Hamburg (rural)
33.0 67.0e
a The metropolitan region of Hamburg is very similar to Germany due to its size and diversity in population. b 100% equals all persons between 18 and 65. c (German Institute for International Educational Research – DIPF, 2012). d Categories according to (infas, 2010). e (Statistical Office for Hamburg and Schleswig-Holstein, 2012).
Table 3 Estimation results of attributes only model. Number of obs ¼28,000
Mixed logit model
LR chi2 (10) ¼ 4625.92 Prob4chi2 ¼0.000
Log likelihood ¼ 6914.5339 Variable
Coef.
Std. Err. z
Price BEV PHEV FCEV Fuel Subsidy Fee Tax Charge None-option Standard Deviation BEV PHEV FCEV Fuel Subsidy Fee Tax Charge None-option
1.0120 1.9665 1.6921 1.3801 0.7387 0.6687 0.1359 0.2125 0.0404 8.1162
0.0511 0.1374 0.1229 0.1589 0.0459 0.1801 0.0143 0.0185 0.0065 0.3087
2.1168 1.7028 1.5001 0.6361 3.2536 0.1443 0.1569 0.0399 3.5813
0.1251 0.1062 0.1690 0.0449 0.1927 0.0280 0.0356 0.0087 0.1857
P4|z| [95% Conf. Interval]
19.79 14.31 13.76 8.68 16.11 3.71 9.48 11.46 6.21 26.29
0.000 1.1122 0.000 2.2358 0.000 1.9330 0.000 1.6916 0.000 0.8286 0.000 0.3157 0.000 0.1640 0.000 0.2488 0.000 0.0531 0.000 8.7212
0.9118 1.6972 1.4511 1.0686 0.6489 1.0217 0.1079 0.1761 0.0276 7.5112
16.92 16.03 8.88 14.17 16.88 5.15 4.41 4.57 19.29
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
2.3620 1.9110 1.8313 0.7241 3.6313 0.0894 0.2266 0.0569 3.9453
1.8716 1.4946 1.1689 0.5481 2.8759 0.1992 0.0871 0.0228 3.2174
The sign of the estimated standard deviations is irrelevant: interpret them as being positive.
(1986) the simulation was estimated where ‘price’ is the only variable which is fixed. All other variables are random in this model. The estimation shows a value of 6914.53 for the log likelihood test with nine degrees of freedom. The likelihood ratio test for the common significance of the standard deviation is 4625.92. The values became better the more iterations were performed. The null hypothesis (all standard deviations equal 0) can be rejected. All attributes are significant on a 99% significance level (Table 3). A car with an internal combustion engine (ICV) serves as reference for the three alternative propulsion technologies BEV, PHEV and FCEV. The results show that in general ICVs are preferred against alternative propulsion technologies in the case that all other variable are controlled. As assumed, higher fuel costs, parking fees and the rise of a CO2-specific tax will have a negative effect on the purchasing decision. An increase in distance to the next charging opportunity will also decrease the probability of being chosen. An increase in direct subsidy will increase the probability if all other variables are controlled – as expected: ● ICVs are still preferred to alternatively propelled vehicles; Coefficients of BEV, PHEV and FCEV are all negative. ● An increase in costs is negatively assessed and vice versa. ● A higher availability of charging infrastructure affects the choice of electric cars positively. 3.1.1. Definition of the reference values In order to calculate the effects of each incentive on the purchasing behavior of the respondents, reference values have to be set. For each attribute a certain level serves as reference. Considering the estimated coefficients in Table 3, it is possible to determine the likelihood of choosing a ZEV according to the incentive. The references for price are €20,000 for ICV, €30,000 for BEV, €25,000 for PHEV and €45,000 for an FCEV. The references for fuel/charging costs are €1000/year for ICVs, €200/year for BEV, €600/year for PHEV and €500/year for FCEV. Parking fees are generally two euros per hour and the availability of the next charging infrastructure is within five minutes walking distance. 3.1.2. Effects of governmental incentives on the likelihood to choose a ZEV The first step is a discussion of the impact of the implementation of different incentives on the choice of car buyers related to the propulsion technology. Five different incentives to influence purchasing behavior were investigated. The likelihood for each incentive is calculated according to formula (2). The results are given in Table 4. Table 4 Likelihood of choice of each propulsion technology depending on incentive. Incentive
Likelihood [%]
Reference scenario
ICV BEV 78.7 6.7
Incentives investigated in experiment: 1) Purchasing grant for ZEV: €10,000 2) CO2 tax for ICV: €500 3) Free parking for ZEV 4) Increase of fuel costs for ICV by tax increase: 100% 5) Increase of availability of charging infrastructer for ZEV a b
PHEV FCEV none ZEVa Δb 11.7 2.3 0.6 20.7 –
9.6
0.5
37.8
þ 17.1
73.6 7.0 12.2 70.4 7.9 13.8 59.7 10.7 18.7
6.5 7.4 10.0
0.6 0.5 1.0
25.7 29.0 39.3
þ 5.0 þ 8.3 þ 18.6
73.4 7.4
5.9
0.6
26.1
þ 5.3
61.7
10.3 18.0
ZEV: Σ(BEV, PHEV, FCEV). Δ: change of likelihood due to incentive.
12.9
C. Rudolph / Transport Policy 52 (2016) 113–120
The results show that at given starting values (reference scenario) the likelihood to choose a conventional vehicle is about 79%. This value is aggregated over all respondents. The likelihood of choosing a BEV is only 7%, for a PHEV about 12%, and for an FCEV around 2.3%. The likelihood of choosing a ZEV (BEV, PHEV or FCEV) is around 21%. The likelihood of choosing none of the alternatives is only around 1%. Comparing the effect on the choice of BEV, PHEV and FCEV shows that all incentives would have the highest relative impact on FCEV but in absolute number on PHEV. In general, direct subsidies of €10,000 increases the likelihood of choosing a ZEV by 17% points due to this subsidy. CO2 tax for ICVs of about €500/year would increase the likelihood to choose a ZEV by approx. 5% points while the incentive of free parking for ZEV in inner city areas would increase the probability to choose an alternatively propelled vehicle by 8% points. The highest impact in this system shows the increase of fuel costs by 100% due to higher fuel taxation. It would increase the probability to choose a ZEV tremendously by 19% points. A rather small impact of approx. five percentage points on the choice shows an increase of available charging infrastructure for electric vehicles. In general, all tested incentives have a positive impact on the probability to choose a ZEV when introduced. The effect strongly depends on the magnitude of the incentive itself and how much money and/or time (distance to charging station) a consumer can save. The increase of fuel costs ( þ18.6% points) and a direct subsidy ( þ17.1% points) seem to have the highest influence on the choice (Table 4).
117
3.2. Integration of mobility routines As described above, the respondents' daily mobility routines are integrated into the model by introducing interaction variables. In order to avoid a dramatic increase in the number of variables, all alternatively propelled vehicle concepts are stored in only one variable (ZEV). The interaction variable is created only with this new variable. Mobility attributes are comprised by 1) vehicle miles traveled per year [km], 2) availability of an annual transit pass, 3) usage frequency of PT, and 4) usage frequency of bicycles. The simulation results in a log likelihood of about 6851. The likelihood ratio test gives 4393 with nine degrees of freedom (Table 5). By integrating the mobility features, we see significant heterogeneities between these groups in respect of choosing a ZEV as incentives are given. Respondents with a yearly mileage of up to 15,000 km might choose zero emission vehicles (ZEVs) significantly more often than respondents that drive more than 15,000 km per year. This result is unfortunate, since the highest ecological effect would be reached if people with a high yearly mileage bought ZEVs. The probability of annual transit pass holders for PT choosing a ZEV is significantly higher than of respondents without a pass. This might also be crucial, since incentives to buy a ZEV should be exploited by people using a passenger car and not PT. In respect to usage frequency of different transport means, the results show that the likelihood of choosing a ZEV by people using a bicycle increases with its usage frequency. That means the more often people use their bike the more they tend to choose a ZEV and vice versa. The risk is of making electric
Table 5 Estimation results interaction model. Number of obs ¼ 28,000
Mixed logit model
LR chi2 (9)¼ 4393.15 Prob4chi2 ¼ 0.000
Log likelihood ¼ 6850.6727 Variables Price BEV PHEV FCEV Fuel Subsidy Parking fee Emission based tax Distance to next charging point None option
Coef. 0.9930 3.4447 3.1919 2.8798 0.7200 0.7679 0.1284 0.2129 0.0397 9.1209
Std. Err. 0.0508 0.2372 0.2299 0.2539 0.0449 0.1828 0.0139 0.0187 0.4939 0.3800
z 19.55 14.52 13.88 11.34 16.05 4.20 9.23 11.41 18.47 0.46
P 4z 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.648
[95% Conf. Interval] 1.0926 3.9097 3.6426 3.3774 0.8079 0.4097 0.1557 0.2495 10.0890 0.9183
0.8934 2.9798 2.7413 2.3822 0.6320 1.1261 0.1011 0.1763 8.1529 0.5714
Interaction variables Vehicle miles traveled/year X ZEV Annual transit pass X ZEV Using bike daily X ZEV Using bike sometimes X ZEV Using PT daily X ZEV Using PT sometimes X ZEV Vehicle miles traveled/year X none option Annual transit pass X none option Using bicycle daily X none option Using bicycle never X none option Using PT daily X none option Using PT sometimes X none option
0.3654 0.6986 1.2191 0.6228 1.0584 0.6155 1.8322 0.1790 0.7584 0.2404 0.3384 0.1735
0.1568 0.2616 0.2397 0.1862 0.3426 0.1834 0.3689 0.5379 0.4311 0.3906 0.7905 0.3800
2.33 2.67 5.09 3.34 3.09 3.36 4.97 0.33 1.76 0.62 0.43 0.46
0.020 0.008 0.000 0.001 0.002 0.001 0.000 0.739 0.079 0.538 0.669 0.648
0.0581 0.1858 0.7492 0.2578 0.3869 0.2560 1.1092 1.2333 0.0865 0.5251 1.2111 0.9183
0.6728 1.2114 1.6889 0.9877 1.7300 0.9751 2.5551 0.8753 1.6033 1.0060 1.8878 0.5714
Standard deviation BEV PHEV FCEV Fuel Subsidy Parking fee Emission based tax Distance to next charging point None option
1.9817 1.7388 1.6300 0.6155 3.0777 0.1169 0.1612 0.0359 3.5287
0.1207 0.1148 0.1917 0.0459 0.1839 0.0393 0.0433 0.0112 0.2079
16.42 15.15 8.50 13.41 16.73 2.98 3.72 3.21 16.98
0.000 0.000 0.000 0.000 0.000 0.003 0.000 0.001 0.000
1.7451 1.5139 1.2543 0.5255 2.7172 0.1938 0.0763 0.0578 3.1213
2.2182 1.9638 2.0056 0.7054 3.4382 0.0400 0.2461 0.0140 3.9362
118
C. Rudolph / Transport Policy 52 (2016) 113–120
cars financially attractive to people that do not possess or use a car. Rising incentives might eventually bring more cars onto the roads. The same effect is shown observing public transport users. The likelihood of choosing a ZEV by people using PT increases significantly with the frequency of use. 3.3. Change in probability of choosing ZEVs within different mobility groups In the next step, four different behavioral homogeneous groups with distinct daily mobility routines are defined endogenously. The groups should be clearly divided by their daily mobility routines. To avoid non-realistic groups that might occur using cluster analysis methods, the groups were defined as follows: A. Everyday PT users: In this group are people who use PT usually every day. The mileage per year is below 15,000 km. They hold an annual transit pass. Their mobility routines are focused on PT. PT is the most important means of transport. B. Everyday car drivers: This group contains persons who use their car every day. Their average yearly mileage is above 15,000 km/year and they do not have an annual transit pass. They never use PT or a bicycle. The car is the most important means of transport. C. Multimodal users – with car affinity: For this group multiple means of transport are in scope. But the car is the preferred means. Like group A they have a yearly mileage of more than 15,000 km/year. Bike and Public Transport are used sometimes. People in this group do not have an annual transit pass. The car is the most important means of transport. D. Multimodal users – with PT affinity: People of this group do not drive more than 15,000 km per year. They use bicycles and cars only sometimes to come around. They hold an annual transit pass. PT is the most important means of transport. Table 6 shows the number of respondents in each mobility group and the attributes of each group. As above, the same specific incentives for each group are simulated. The change in likelihood of choosing a ZEV in respect to each incentive according to the estimated model using formula (2) is given in Table 7. The integration of interaction variables representing the respondents’ mobility behavior shows a similar result like Table 4. Incentives have different impact according to type and benefit for the consumer. The difference in between the distinct incentives may differ heavily when increasing the absolute monetary benefit for consumers. But most interestingly, the integration of these interaction variable shows, that incentive may affect different groups which are homogenic in terms of mobility behavior very distinctly. For example, direct subsidies or a sensitive increase of
Table 7 Effects of incentives on each group. Group
A
B
C
D
Reference scenario L (ZEV)a [%]
5.6
32.8
10.0
47.6
þ 5.8 þ 0.6 þ 1.5 þ 5.2
þ 18.5 þ 2.3 þ 5.9 þ 16.8
þ 9.3 þ 1.0 þ 2.6 þ 8.5
þ 18.6 þ 2.6 þ 6.4 þ 17.0
þ 0.8
þ 3.2
þ 1.4
þ 3.6
Change of L(ZEV) due to incentive [%] 1) purchasing grant for ZEV: 10,000 € 2) CO2tax for ICV: 500 € 3) free parking for ZEV 4) increase of fuel costs for ICV by tax increase: 100% 5) increase of availability of charging infrastructer for ZEV a
L(ZEV): Σ L(BEV, PHEV, FCEV); L ¼Likelihood.
fuel costs of 100% plus may increase the likelihood for group B and D by about 18.5% points while the likelihood for group C may increase by 9.3% points and for group A only by 5.8% points. That means people with a more ecological mobility behavior may be affected in a much stronger way than people using a passenger car very frequently. This phenomenon can be observed, although on different levels, with all tested incentives.
4. Implications for policy making The vehicle purchasing models which are developed in this study base on a hypothetical experiment (stated choice). In fact, stated choice experiments can never reflect real choices unlike revealed choice data reflecting choices which have been made in reality. Nevertheless, the results base on a representative number of respondents and reflect their probable choice against the background of given incentives. The main objective of this study is combining purchase decisions regarding the propulsion technology with the mobility behavior of the respondent in order to assess the ecological effect of subsidies for ZEV. The results of the model show that subsidizing ZEV may increase the likelihood to choose an alternatively propelled vehicle. All subsidies which are tested in this study (direct purchasing grant, vehicle tax reduction, emission based parking costs, fuel taxation and availability of charging infrastructure) will increase attractiveness of ZEV over conventional vehicles with combustion engines. As long as ZEV have too high market prices and the market penetration of ZEV is still marginal, subsidies can help making ZEV competitive with conventional cars. Zero emission vehicles are considered to be one solution to reduce local emissions (e.g. NOx and SOx) improving air quality in urban areas and to reduce greenhouse gas emissions (e.g. CO2). But, the results of the models show also that two rebound effects may encounter when giving subsidies for ZEV:
Table 6 Mobility groups. Attribute of mobility behavior Everyday car drivers (n¼ 35)
Everyday PT users (n¼ 21)
Multimodal users – with car affinity Multimodal users – with car affinity (n¼ 93) (n ¼36)
Yearly mileage below 15,000 km Yearly mileage above 15,000 km Annual transit pass Bicycle use daily Bicycle use sometimes Bicycle use never PT use daily PT use sometimes PT use never
0
1
0
1
1
0
1
0
0 0 0 1 0 0 1
1 0 0 1 1 0 0
0 0 1 0 0 1 0
1 0 1 0 1 0 0
C. Rudolph / Transport Policy 52 (2016) 113–120
Firstly, by increasing the attractiveness of ZEV also for people having mobility patterns with a low energy ratio (on foot, bicycle, PT) by giving subsidies may lead to higher sales amongst these mobility groups. Due to the eco label of ZEV plus economical feasible may increase the likelihood to shift to individual motorized transport (passenger car) for these people as well. The overall energy ratio for transport would then be worse for these people than before. From an ecological perspective, this result is a dilemma. The overall objective to exchange as many combustion engines as possible by ZEV may fail. Under given energy production mix in Germany CO2 emissions would also increase. Secondly, another rebound effect may encounter: urban transport policies in the last decades tried to increase the attractiveness of non-individual motorized means of transport, i.e. on foot, bicycle and PT. The models show that these mobility groups would exploit such subsidies most likely. Increasing the attractiveness of passenger cars for all mobility groups would torpedo urban transport policies aiming in decreasing individual motorized transport. It would lead to an increase of the total number of cars, the total number of vehicle miles traveled and the amount of space needed in urban areas to accommodate these vehicles – in terms of roads and parking facilities. These two effects have major implications for policy makers: In order to prevent the first rebound effect, policy makers must assure that electricity used for individual mobility comes from 100% real physical renewable sources. Using only certified green energy will improve the CO2 emission ratio within the transport sector but worsen the ratio in all other sectors as long as the energy from the grid is not 100% renewable. Only then, ZEV can also contribute to reduce greenhouse gas emissions in the transport sector. Policy makers have to ensure that electricity which is used for ZEV taken from public charging facilities as well as from private plugs have to be real green energy. In parallel, the installation of further renewable energy plants has to be successively conducted. In order to prevent the second rebound effect, policy makers should parallelly invest to increase the attractiveness for all other means of transport when fostering low emission vehicles. To increase the attractiveness making errands on foot, the local supply (e.g. small local retailers etc. to establish acity of short distances) should be fostered and the infrastructure for pedestrians should be improved. To increase the attractiveness to use the bicycle on a daily basis for commuting and making errands a safe, direct, fast, comfortable and coherent infrastructure should be provided. Finally, the public transport should be expanded steadily. Therefore, independence from individual means of transport is essential as well as vehicles which are fast, safe to use and frequently run. Recently, the German government decided to introduce a €1.2 billion stimulus program granting BEV purchases with up to €4,000 and HEV with up to €3,000 depending on the price of the vehicle (Bundesregierung). Half of the subsidy will be given by the German government; the other half will be paid by the car industry. Further 300 million Euros will be invested for the installation of 15,000 charging facilities all across Germany. Additionally, for BEV no yearly vehicle tax has to be paid for ten years after purchase. Against the background of this study, policy makers should run a research program to investigate which real mobility behavior the buyers of ZEV have had before the purchase and which ecological effects, which have been discussed above, will be induced in reality. A cap of the subsidies should be considered enabling the government to stop the program after a first pilot in case accompanying research shows that purchasers' mobility profiles show a lower energy ratio than after the ZEV purchase. In order to prevent these rebound effects counteracting decisions must be taken in advance or in parallel but not afterwards as a reaction. After all, introducing policies subsidizing expensive goods, as
119
zero emission vehicles are, means subsidizing the most affluent social class of a society. ZEV are often purchased as a second or third car. People which can’t afford a zero emission vehicle will never profit from such subsidies. Such policies are socially unjust, especially, if no money is spent to improve other means of transport. Policy maker have to consider this fact when making the tradeoff where to spend tax money on.
References Achtnicht M., Bühler G. and Hermeling C., 2008. Impact of Service Station Networks on Purchase Decisions of Alternative-fuel Vehicles, ZEW Discussion Papers No. 08-088. Achtnicht, M., 2009. German Car Buyers' Willingness to Pay to Reduce CO2 Emissions: Discussion Paper No. 09-058, ZEW Zentrum für Europäische Wirtschaftsforschung GmbH: Mannheim, p. 32. Axsen, J., Mountain, D.C., Jaccard, M., 2009. Combining stated and revealed choice research to simulate the neighbor effect: the case of hybrid-electric vehicles. Resour. Energy Econ., 221–238, Jg. 31, H. 3. Brownstone, D., Bunch, D.S., Train, K.E., 2000. Joint mixed logit models of stated and revealed preferences for alternative-fuel vehicles. Transp. Res. Part B: Methodol. 34 (5), 315–338. Dagsvik, J.K., et al., 2002. Potential demand for alternative fuel vehicles. Transp. Res. Part B: Methodol. 36 (4), 361–384. Daziano, R.A. and Achtnicht, M., 2012. Forecasting Adoption of Ultra-low-emission Vehicles Using the GHK Simulator and Bayes Estimates of a Multinomial Probit Model: Discussion Paper No. 12–017, Mannheim, p. 32. Ewing,, G.O., Sarigöllü, E., 2000. Assessing consumer preferences for clean-fuel vehicles: a discrete choice experiment. J. Public Policy Mark. 19 (1), 106–118. Federal Motor Transport Authority, 2015. Number of Passenger Vehicles in Germany 01/01/2015 According to Fuel Type (Bestand an Pkw Am 1. Januar 2015 nach ausgewählten Kraftstoffarten). Federal Republic of Germany, 2009. German Federal Government's National Electromobility Development Plan. Berlin. German Institute for International Educational Research – DIPF. Bildungsbericht 2012: Tabellenanhang. Available from: 〈http://www.bildungsbericht.de/da ten2012/tabellen_web2012.pdf〉. Green,, P.E., Srinivasan,, V., 1978. Conjoint analysis in consumer research: issues and outlook. J. Consum. Res., 103–123. Hensher,, D.A., Rose, J.M., Greene, W.H., 2008. Combining RP and SP data: biases in using the nested logit ‘trick’ – contrasts with flexible mixed logit incorporating panel and scale effects. J. Transp. Geogr. 16 (2), 126–133. Hole,, A.R., 2007a. Estimating mixed logit models using maximum simulated likelihood. Stata J. 7 (3), 388–401. Hole,, A.R., 2007b. Fitting mixed logit models by using maximum simulated likelihood. Stata J. 7 (3), 388–401, 14. Breivik, I.F. and Volder, M.O., 2014. Electric Vehicles in Norway – A Qualitative Study of the Electric vehicle Market in Norway. infas, Mobility in Germany 2008 - Report. 2010. Ito,, N., Takeuchi,, K., Managi,, S., 2013. Willingness-to-pay for infrastructure investments for alternative fuel vehicles. Transp. Res. Part D: Transp. Environ. 18, 1–8. Jenn, A.T., 2014. “Advanced and Alternative Fuel Vehicle Policies: Regulations and Incentives in the United States”, Dissertations. Paper 354. 〈http://repository. cmu.edu/dissertations/354〉. Knockaert, J., 2005. The Choice for Alternative Cars. Energy, Transport and Environment Center for Economic Studies, Leuven, Belgium. Lave, C.A., Train, K.E., 1979. A Disaggregate Model of Auto-Type Choice. Transpn Res 13A. Pergamon Press Ltd, pp. 1–9. Akiva, M.E.B. and Lerman, S.R., 1985. Discrete Choice Analysis: Theory and Application to Predict Travel Demand. Mabit,, S.L., Fosgerau,, M., 2011. Demand for alternative-fuel vehicles when registration taxes are high. Transp. Res. Part D: Transp. Environ. 16 (3), 225–231. Manski,, C.F., Sherman,, L., 1980. An empirical analysis of household choice among motor vehicles. Transp. Res. Part A: Gen. 14 (5–6), 349–366. McFadden,, D., Train,, K.E., 2000. Mixed MNL models for discrete response. J. Appl. Econ. 15 (5), 447–470. Potoglou,, D., Kanaroglou, P.S., 2006. Household Demand and Willingness to Pay for Alternative Fuelled Vehicles. Hamilton, ON; Canada, p. 31. Rabe-Hesketh,, S., Skrondal,, A., 2012. Multilevel and Longitudinal Modeling Using Stata, 3 ed., College Station; Tex: Stata Press. Revelt,, D., Train,, K.E., 1998. Mixed logit with repeated choices: households' choices of appliance efficiency level. Rev. Econ. Stat. 80 (4), 647–657. Small, K.A., Winston, C., Yan, J., 2005. Uncovering the distribution of motorists’ preferences for travel time and reliability. Econometrica 73 (4), 1367–1382. Statistical Office for Hamburg and Schleswig-Holstein., 2012. Population of Hamburg 2012 (Bevölkerung in Hamburg 2012), 10.10.2013; Available from: 〈http:// www.statistik-nord.de/uploads/tx_standocuments/SI13_117.pdf〉. Bundesregierung, Steuervorteile und Kaufprämie beschlossen (engl: tax advantages and purchase grant decided), Available from: 〈https://www.bundesregierung. de/Content/DE/Artikel/2016/05/2016–05-18-elektromobilitaet.html〉.
120
C. Rudolph / Transport Policy 52 (2016) 113–120
Thomas, L., 1983. Der Einfluß von Kindern auf Produktpräferenzen ihrer Mütter. (Schriften zum Marketing. Duncker & Humblot, Berlin, p. 6598. Train,, K.E., 1980. A structured logit model of auto ownership and mode choice. Rev. Econ. Stud. 47 (2), 357–370. Train,, K.E., 1998. Recreation demand models with taste differences over people. Land Econ., 230–239. Train,, K.E., 2003. Discrete Choice Methods with Simulation, 2nd ed. Cambridge Univ. Press, Cambridge, p. 334. Train, K.E., 1986. Qualitative Choice Analysis: Theory, Econometrics, and an
Application to Automobile Demand, Cambridge. Raich, U., Link, C., Stark, J., 2012. Möglichkeiten und Grenzen der individuellen Elektromobilität. Kauf von Elektroautos, in Technische Universität Dresden, Tagungsband Verkehrswissenschaftliche Tage. Zhou, Yan, et al., 2015. Plug-in electric vehicle market penetration and incentives: a global review. Mitig. Adapt. Strateg. Glob. Change 20 (5), 777–795. Ziegler,, A., 2012. Individual characteristics and stated preferences for alternative energy sources and propulsion technologies in vehicles: a discrete choice analysis for Germany. Transp. Res. Part A: Policy Pract. 46 (8), 1372–1385.
Contents lists available at ScienceDirect
Transport Policy journal homepage: www.elsevier.com/locate/tranpol
How may incentives for electric cars affect purchase decisions?$ Christian Rudolph German Aerospace Center (DLR), Institute of Transport Research, Rutherfordstrasse 2, 12489 Berlin, Germany
art ic l e i nf o
a b s t r a c t
Article history: Received 3 March 2016 Received in revised form 23 June 2016 Accepted 29 July 2016
In this paper, the impact of five different incentives for buyers of zero emission vehicles (ZEV) is investigated with a stated choice experiment. The tested incentives are direct subsidies, free parking, a separate CO2 tax, an increase of fuel costs by tax elevation, and an increase of available charging infrastructure. By implementing the mobility patterns of the respondents, it was possible to simulate estimations of ecological impact and modal shift with a random utility model (mixed logit). Based on 875 complete questionnaires, the simulation results show that giving incentives to these buyers ecological rebound effects are expected: Mostly people with a low CO2-emission rate regarding their daily transportation routines (cyclists and public transport users) will exploit these incentives. They show a significantly higher likelihood of choosing alternatively propelled cars than conventional car users. Consumers that usually use a passenger car for their daily mobility routines are mostly unwilling to change to ZEV even when incentives are given. & 2016 Elsevier Ltd. All rights reserved.
Keywords: Choice modeling Alternative propulsion technology Battery electric vehicles BEV Discrete choice Stated preference Choice-based conjoint analysis
1. Introduction In the fight against global warming and forthcoming oil shortage zero emission vehicles (ZEV) such as battery electric vehicles (BEV – running only on electricity from the grid), plug-in hybrid electric vehicles (PHEV – running on electricity from the grid and on fossil fuel) or fuel cell electric vehicles (FCEV – running on hydrogen) have come into the focus of transport policies in order to decrease greenhouse gas emissions from the transportation sector. The German government for example established the German Federal Development Scheme for Electric Mobility (Federal Republic of Germany, 2009) in 2009. With this scheme the government set the ambitious target of ‘one million electric cars on German roads by 2020’. With a traditionally strong automobile industry, Germany wants to position itself as the market leader of electric cars (Federal Republic of Germany, 2009). But, even years after introduction of this scheme, purchases of ZEV show small numbers: only about 19,000 BEV and 33,000 HEV are on Germany's streets in the end of 2014 (Federal Motor Transport Authority, 2015). Due to the slow market penetration of ZEV, a debate on implementing incentive schemes to stimulate the demand for electric cars arose. Different stimulation schemes have been ☆ A discrete choice study investigating the ecological impact and modal shifts due to governmental incentives fostering the market penetration of zero emission vehicles in Germany. E-mail address: [email protected]
http://dx.doi.org/10.1016/j.tranpol.2016.07.014 0967-070X/& 2016 Elsevier Ltd. All rights reserved.
introduced in other European and non-European countries. The variety of incentives as push and pull measures is broad: tax exceptions, privileges (free parking, using bus and taxi lanes, carpooling lanes,1 etc.), direct purchase subsidy, introduction of CO2free zones are the most frequently discussed. International approaches differ strongly: While countries such as Denmark and Norway subsidize ZEV purchases and liberate from tax, countries such as Germany used to support only research and development projects. In some countries a strong increase of purchase numbers of ZEV can be observed where subsidies (buying grants and tax reductions) are given, e.g. Norway (Breivik and Volder, 2014). A recent global review about incentives for plug-in electric vehicles is given in (Zhou et al., 2015). But, purchase decisions have not been investigated in dependence of the buyer's mobility behavior. Studies so far investigate elasticies of incentives, e.g. (Jenn, 2014) or technological attributes in order to predict general sales numbers of ZEV. They do not account for the energy ratio for mobility of the buyers. This paper gives first insights about possible ecological effects of different hypothetical incentives fostering ZEV in combination with the mobility behavior of the respondents. Subsidies on the one hand may increase sales of ZEV. But on the other hand, from the ecologic perspective vehicles with a combustion engine should be exchanged by ZEV and drivers should preferably have a high 1
USA and Canada specific.
114
C. Rudolph / Transport Policy 52 (2016) 113–120
annual mileage in order to reduce the energy ratio for mobility. This is very important since electricity from the grid in Germany still is not 100% renewable. That means also ZEV emit greenhouse gases in a well-to-wheel consideration. The energy ratio would even worsen if incentives would bring cyclists and users of public transport (PT) to use ZEV since they become affordable. Additionally, these cars are considered as environmental friendly. Especially these groups showing a very low energy ratio for mobility should not be motivated to buy an individual passenger car. Sales of ZEV should not be additive but replace existing conventional vehicles. Against the background of the current discussion in Germany about the effects of incentives for ZEV and which incentives should be implemented, an online stated preference experiment which is representative for Germany was conducted.
2. Research design An online survey with 1010 respondents was conducted in the metropolitan area of Hamburg, Germany, which resulted in 875 completely finished questionnaires. Hamburg was chosen due to its spatial diversity2 (e.g. inner city and rural areas). In order to simulate hypothetical purchase decisions, a stated preference (SP) experiment using a choice-based conjoint analysis (CBCA) was conducted. The respondents were only selected by their socio economic and socio geographic attributes. The circumstance if a respondent was about purchasing a new car in the near future was not of interest, since it would bias the results against the background of the research interest investigating the effect of incentives across all citizens using different means of transport. An example of a choice set is given below in Fig. 1. The respondents' purchasing decisions are simulated with a random parameter utility model (mixed logit) to account for dependence across observations from the same respondent (Small et al., 2005). Logit models are applied to determine consumers' ‘tastes’ towards certain products, services or attributes of products (Brownstone et al., 2000; McFadden and Train, 2000; Lave and Train, 1979; Hensher et al., 2008; Train, 1998, 1986, 1980; Akiva and Lerman, 1985; Manski and Sherman, 1980). The estimation results simulated by mixed logit models should approximate reality more accurately than traditional conditional logit or multinomial logit models due to the ability to implement individual heterogeneities amongst respondents. The applied model is based on Revelt and Train (1998). To estimate the β-coefficients the mixed logit module by Hole (2007a) is applied. Hole's model uses a maximum likelihood simulation to estimate the coefficients (Hole, 2007b). For a detailed description of the methodology of the simulated maximum likelihood method please refer to (Hole, 2007a, 2007b; Train, 2003). Besides the knowledge of the preferred propulsion technology and the importance of the different incentives, it is necessary to know the mobility patterns of the respondents to draw conclusions from their decision to their future mobility. Therefore the respondents are asked during the survey about their daily mobility routines, i.e. yearly mileage (higher or lower than 15,000 km/year), possession of annual transit pass, use frequency of bicycles (daily use, sometimes, never), use frequency of PT (daily use, sometimes, never). Four groups with different mobility patterns could endogenously be clustered: car users, PT users, Multimodal users – with car affinity, Multimodal users – with PT affinity. By applying different incentives to the model, the impact of each incentive on the choice of propulsion technology could be estimated separately for each mobility group. Due to methodology 2
In further studies also spatial effects will be investigated.
the results are based on hypothetical assumptions given by the respondents, not the revealed factual behavior. In order to identify the attributes (here: purchase incentives) which should be investigated, a literature review about the state of the art survey design gave valuable hints. Green and Srinivasan (1978) state that not more than five to six different attributes should be tested within a discrete choice experiment, while Thomas (1983) report that up to 20 attributes can be applied. Ito et al. (2013) uses a CBCA to identify the willingness to pay for alternatively propelled vehicles in Japan and the resulting cost to install the charging infrastructure. He examines nine attributes: vehicle type, brand, driving range, recharging time, CO2 emission, availability of charging infrastructure, price, and operating costs. For Canada, Potoglou and Kanaroglou (2006) conducted a discrete choice study with nearly the same attributes. The model results show that direct purchasing subsidies and the emission rate of CO2 have the highest influence on the purchasing decision. Ewing and Sarigöllü (2000) conducted a CBCA (N ¼881) with commuters to examine the importance of price, costs for maintenance, acceleration, recharging time, driving range, emissions, privilege to use car-pooling lanes, fuel costs, and parking costs. They show that these measures will have only a small impact on the purchasing decision towards ZEVs and come to the conclusion to foster research and development projects instead. Axsen et al. (2009) conducted an SP analysis in California and Canada to investigate governmental incentives for HEVs, which are not in the scope of this paper. Daziano and Achtnicht (2012), Achtnicht (2009), Achtnicht et al. (2008) and Ziegler (2012) applied different models to a survey conducted in 2008 using an SP experiment with six tasks each showing seven different concepts. They investigated price, fuel costs per 100 km, engine performance, CO2 emissions, availability of charging infrastructure and type of fuel. Raich et al. (2012) finds out in a survey conducted in Austria using an SP experiment that cost-affected attributes have a much higher influence on the purchasing decision than technological attributes (such as driving range, recharging time, emission rate). Mabit and Fosgerau (2011) also conducted an SP experiment in Denmark. The experiment had 12 tasks each with five alternatives to choose from. They examined propulsion technology, fuel costs, price, driving range, number of charging operations per week, acceleration and further operating costs. The survey conducted by Dagsvik et al. (2002) with respondents from Norway comprised 15 tasks each with three vehicle concepts. They investigated price, maximum speed, driving range and energy consumption. Knockaert (2005) used in his study price, operating costs, fuel consumption, propulsion technology, driving range, emissions, volume of trunk. The literature review shows that all studies focus on investigating purchase decisions based on technological attributes of the vehicles. Purchase models are developed and the development of ZEV sales can be predicted depending of developments of these attributes. No study examines the purchase decision combined with the mobility behavior of the respondents. Therefore, rather very general statements can only be given about the ecological effect and the overall energy consumption depending from the mobility patterns of the respondents. The literature review shows also that the CBCA is the most common SP experiment design. Therefore, a CBCA is also applied for this study. Studies of the literature review show that best results are achieved with not more than three alternatives to choose from and the ‘none option’ (‘I would purchase none of the alternatives.’). Each respondent had to complete eight similar choice tasks in sequence with three alternatives and the ‘none option’, levels of attributes are changed each time. shows a sample of the display.
C. Rudolph / Transport Policy 52 (2016) 113–120
115
Fig. 1. Display of choice sets as shown in survey.
2.1. Attributes and levels used in survey Due to the high influence of brand, design, and production country on the purchasing decision, those attributes are not shown. In consequence a text-based analysis is chosen to present the choice sets. Finally a combination of the findings resulting from the literature review and the current political discussion in Germany led to the setting of the attributes and their levels. Since incentives are within the scope of this study, technological aspects were reduced to the propulsion technology. The survey investigates the purchasing decision when price, fuel costs (depending on the individual yearly mileage3), a separate CO2 tax for combustion vehicles, direct purchasing subsidies, parking fees, and the availability of charging infrastructure varies between BEV, PHEV, FCEV and ICV (internal combustion vehicles). Table 1 shows an overview of the tested attributes and the level variation. 2.2. Descriptive analysis of the respondents The data was collected by an online access panel provider. Since panel providers know comprehensive characteristics of their members, a representative study can be realised efficiently. The survey was conducted between 06/17/2013 and 06/26/2013. In order to represent the metropolitan area of Hamburg, only participants in this area were addressed. Socio-economic and sociogeographic characteristics were queried in the first part of the survey. The ratio of people living in urban and rural areas (1/2) complies with the allocation of the metropolitan area of Hamburg. In terms of sex, age, education, type of employment, net income, and population density, the sample represents the population of the metropolitan area of Hamburg very well and the population of Germany closely. An overview of the sample is given in Table 2. Emails were sent to 1010 people to participate in the survey. The survey resulted in 875 completely finished questionnaires which could be used for further analyses. The respondents were also asked if they possess a car or have access to a car, if they have a driving license, and if they have an annual transit pass for the public transport system. In terms of car access, 7.7% of the respondents do not own a car but only 3.3% of the respondents do not have any access to a car; 16.2% share a car with other household members, friends or use car sharing; and 80.5% have unlimited access to a car. A driving license was in the 3
The mileage was asked in the beginning of the survey.
Table 1 Attributes and their levels. Attribute
[No. of levels] level
Propolsion technology Price [€]
4 4
Fuel/charging costs [€/year]
3
CO2 tax [€/year] Direct subsidy [€] Parking fee [€/h] Distance to charging point [min]
4 5 4 7
BEVa, PHEVb, FCEVc, ICVd 30,000, 35,000, 40,000, 45,000 plus 20%, minus 50%, minus 65%e 0, 200, 300, 400 0, 1,000, 3,000, 5,000, 10,000 0, 3, 4, 5 0.5f, 1, 5, 10, 15, 20, 30
a
BEV¼ battery electric vehicle. PHEV ¼plug-in hybrid electric vehicle. c FCEV ¼ fuel cell electric vehicle. d ICV ¼ internal combustion engine vehicle. e Calculation according to information about the yearly mileage given by the respondent. In case no information was given, the average of 15,000 km was applied. f Walking distance to next charging infrastructure (availability of charging infrastructure). b
possession of 97.1% in total. About one fourth (25.5%) have an annual transit pass. Sex, age and the spatial allocation of the respondents comply very well with the general German statistics. The education level of the sample is generally higher than in general. Also the net household income is a bit higher than generally across Germany. This circumstance must be considered when giving recommendations. For further analysis the data will be used as is. No weights will be applied.
3. Estimation results The analysis contains the estimation of two models. Firstly, only the coefficients of the attributes are estimated (attributesonly model). With this model the effect of the different described incentives can be measured. Secondly, interaction variable that integrate daily mobility routines into the model are added. The model estimations are based on 28,000 observations from 7000 tasks (875 respondents; 8 tasks per respondent). 3.1. Attributes-only model According to Rabe-Hesketh and Skrondal (2012) and Train
116
C. Rudolph / Transport Policy 52 (2016) 113–120
Table 2 Sample description (n ¼ 875). Socio-demographic and socio-economic features of sample
Sample (n¼875) (%)
Germanya (%)
Sex Female Male
50.4 49.6
51.1 48.9
Age 18–24 25–34 35–44 45–54 55–65
12.6 19.5 23.9 25.1 18.9
13.1b 19.6b 19.8b 26.2b 21.3b
0.1 11.3 40.6
7.6c 37.0c 28.9c
48.0
25.8c 0.4c
2.3 7.2 12.0 15.3 32.6 19.4 11.2
2.0d 7.0d 18.0d 16.0d 23.0d 12.0d 1.0d
29.9 70.1
e
Education Not yet/ no school graduation Certificate of Secondary Education General Certificate of Secondary Education University-entrance diploma Not specified Net income of household [€/month] Below 500 500–1.000 (500–900)b 1000–1500 (900–1500)b 1500–2000 2000– 3000 3000–4000 more than 4000 Allocation within the metropolitan area of Hamburg City of Hamburg (urban) Metropolitan Region of Hamburg (rural)
33.0 67.0e
a The metropolitan region of Hamburg is very similar to Germany due to its size and diversity in population. b 100% equals all persons between 18 and 65. c (German Institute for International Educational Research – DIPF, 2012). d Categories according to (infas, 2010). e (Statistical Office for Hamburg and Schleswig-Holstein, 2012).
Table 3 Estimation results of attributes only model. Number of obs ¼28,000
Mixed logit model
LR chi2 (10) ¼ 4625.92 Prob4chi2 ¼0.000
Log likelihood ¼ 6914.5339 Variable
Coef.
Std. Err. z
Price BEV PHEV FCEV Fuel Subsidy Fee Tax Charge None-option Standard Deviation BEV PHEV FCEV Fuel Subsidy Fee Tax Charge None-option
1.0120 1.9665 1.6921 1.3801 0.7387 0.6687 0.1359 0.2125 0.0404 8.1162
0.0511 0.1374 0.1229 0.1589 0.0459 0.1801 0.0143 0.0185 0.0065 0.3087
2.1168 1.7028 1.5001 0.6361 3.2536 0.1443 0.1569 0.0399 3.5813
0.1251 0.1062 0.1690 0.0449 0.1927 0.0280 0.0356 0.0087 0.1857
P4|z| [95% Conf. Interval]
19.79 14.31 13.76 8.68 16.11 3.71 9.48 11.46 6.21 26.29
0.000 1.1122 0.000 2.2358 0.000 1.9330 0.000 1.6916 0.000 0.8286 0.000 0.3157 0.000 0.1640 0.000 0.2488 0.000 0.0531 0.000 8.7212
0.9118 1.6972 1.4511 1.0686 0.6489 1.0217 0.1079 0.1761 0.0276 7.5112
16.92 16.03 8.88 14.17 16.88 5.15 4.41 4.57 19.29
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
2.3620 1.9110 1.8313 0.7241 3.6313 0.0894 0.2266 0.0569 3.9453
1.8716 1.4946 1.1689 0.5481 2.8759 0.1992 0.0871 0.0228 3.2174
The sign of the estimated standard deviations is irrelevant: interpret them as being positive.
(1986) the simulation was estimated where ‘price’ is the only variable which is fixed. All other variables are random in this model. The estimation shows a value of 6914.53 for the log likelihood test with nine degrees of freedom. The likelihood ratio test for the common significance of the standard deviation is 4625.92. The values became better the more iterations were performed. The null hypothesis (all standard deviations equal 0) can be rejected. All attributes are significant on a 99% significance level (Table 3). A car with an internal combustion engine (ICV) serves as reference for the three alternative propulsion technologies BEV, PHEV and FCEV. The results show that in general ICVs are preferred against alternative propulsion technologies in the case that all other variable are controlled. As assumed, higher fuel costs, parking fees and the rise of a CO2-specific tax will have a negative effect on the purchasing decision. An increase in distance to the next charging opportunity will also decrease the probability of being chosen. An increase in direct subsidy will increase the probability if all other variables are controlled – as expected: ● ICVs are still preferred to alternatively propelled vehicles; Coefficients of BEV, PHEV and FCEV are all negative. ● An increase in costs is negatively assessed and vice versa. ● A higher availability of charging infrastructure affects the choice of electric cars positively. 3.1.1. Definition of the reference values In order to calculate the effects of each incentive on the purchasing behavior of the respondents, reference values have to be set. For each attribute a certain level serves as reference. Considering the estimated coefficients in Table 3, it is possible to determine the likelihood of choosing a ZEV according to the incentive. The references for price are €20,000 for ICV, €30,000 for BEV, €25,000 for PHEV and €45,000 for an FCEV. The references for fuel/charging costs are €1000/year for ICVs, €200/year for BEV, €600/year for PHEV and €500/year for FCEV. Parking fees are generally two euros per hour and the availability of the next charging infrastructure is within five minutes walking distance. 3.1.2. Effects of governmental incentives on the likelihood to choose a ZEV The first step is a discussion of the impact of the implementation of different incentives on the choice of car buyers related to the propulsion technology. Five different incentives to influence purchasing behavior were investigated. The likelihood for each incentive is calculated according to formula (2). The results are given in Table 4. Table 4 Likelihood of choice of each propulsion technology depending on incentive. Incentive
Likelihood [%]
Reference scenario
ICV BEV 78.7 6.7
Incentives investigated in experiment: 1) Purchasing grant for ZEV: €10,000 2) CO2 tax for ICV: €500 3) Free parking for ZEV 4) Increase of fuel costs for ICV by tax increase: 100% 5) Increase of availability of charging infrastructer for ZEV a b
PHEV FCEV none ZEVa Δb 11.7 2.3 0.6 20.7 –
9.6
0.5
37.8
þ 17.1
73.6 7.0 12.2 70.4 7.9 13.8 59.7 10.7 18.7
6.5 7.4 10.0
0.6 0.5 1.0
25.7 29.0 39.3
þ 5.0 þ 8.3 þ 18.6
73.4 7.4
5.9
0.6
26.1
þ 5.3
61.7
10.3 18.0
ZEV: Σ(BEV, PHEV, FCEV). Δ: change of likelihood due to incentive.
12.9
C. Rudolph / Transport Policy 52 (2016) 113–120
The results show that at given starting values (reference scenario) the likelihood to choose a conventional vehicle is about 79%. This value is aggregated over all respondents. The likelihood of choosing a BEV is only 7%, for a PHEV about 12%, and for an FCEV around 2.3%. The likelihood of choosing a ZEV (BEV, PHEV or FCEV) is around 21%. The likelihood of choosing none of the alternatives is only around 1%. Comparing the effect on the choice of BEV, PHEV and FCEV shows that all incentives would have the highest relative impact on FCEV but in absolute number on PHEV. In general, direct subsidies of €10,000 increases the likelihood of choosing a ZEV by 17% points due to this subsidy. CO2 tax for ICVs of about €500/year would increase the likelihood to choose a ZEV by approx. 5% points while the incentive of free parking for ZEV in inner city areas would increase the probability to choose an alternatively propelled vehicle by 8% points. The highest impact in this system shows the increase of fuel costs by 100% due to higher fuel taxation. It would increase the probability to choose a ZEV tremendously by 19% points. A rather small impact of approx. five percentage points on the choice shows an increase of available charging infrastructure for electric vehicles. In general, all tested incentives have a positive impact on the probability to choose a ZEV when introduced. The effect strongly depends on the magnitude of the incentive itself and how much money and/or time (distance to charging station) a consumer can save. The increase of fuel costs ( þ18.6% points) and a direct subsidy ( þ17.1% points) seem to have the highest influence on the choice (Table 4).
117
3.2. Integration of mobility routines As described above, the respondents' daily mobility routines are integrated into the model by introducing interaction variables. In order to avoid a dramatic increase in the number of variables, all alternatively propelled vehicle concepts are stored in only one variable (ZEV). The interaction variable is created only with this new variable. Mobility attributes are comprised by 1) vehicle miles traveled per year [km], 2) availability of an annual transit pass, 3) usage frequency of PT, and 4) usage frequency of bicycles. The simulation results in a log likelihood of about 6851. The likelihood ratio test gives 4393 with nine degrees of freedom (Table 5). By integrating the mobility features, we see significant heterogeneities between these groups in respect of choosing a ZEV as incentives are given. Respondents with a yearly mileage of up to 15,000 km might choose zero emission vehicles (ZEVs) significantly more often than respondents that drive more than 15,000 km per year. This result is unfortunate, since the highest ecological effect would be reached if people with a high yearly mileage bought ZEVs. The probability of annual transit pass holders for PT choosing a ZEV is significantly higher than of respondents without a pass. This might also be crucial, since incentives to buy a ZEV should be exploited by people using a passenger car and not PT. In respect to usage frequency of different transport means, the results show that the likelihood of choosing a ZEV by people using a bicycle increases with its usage frequency. That means the more often people use their bike the more they tend to choose a ZEV and vice versa. The risk is of making electric
Table 5 Estimation results interaction model. Number of obs ¼ 28,000
Mixed logit model
LR chi2 (9)¼ 4393.15 Prob4chi2 ¼ 0.000
Log likelihood ¼ 6850.6727 Variables Price BEV PHEV FCEV Fuel Subsidy Parking fee Emission based tax Distance to next charging point None option
Coef. 0.9930 3.4447 3.1919 2.8798 0.7200 0.7679 0.1284 0.2129 0.0397 9.1209
Std. Err. 0.0508 0.2372 0.2299 0.2539 0.0449 0.1828 0.0139 0.0187 0.4939 0.3800
z 19.55 14.52 13.88 11.34 16.05 4.20 9.23 11.41 18.47 0.46
P 4z 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.648
[95% Conf. Interval] 1.0926 3.9097 3.6426 3.3774 0.8079 0.4097 0.1557 0.2495 10.0890 0.9183
0.8934 2.9798 2.7413 2.3822 0.6320 1.1261 0.1011 0.1763 8.1529 0.5714
Interaction variables Vehicle miles traveled/year X ZEV Annual transit pass X ZEV Using bike daily X ZEV Using bike sometimes X ZEV Using PT daily X ZEV Using PT sometimes X ZEV Vehicle miles traveled/year X none option Annual transit pass X none option Using bicycle daily X none option Using bicycle never X none option Using PT daily X none option Using PT sometimes X none option
0.3654 0.6986 1.2191 0.6228 1.0584 0.6155 1.8322 0.1790 0.7584 0.2404 0.3384 0.1735
0.1568 0.2616 0.2397 0.1862 0.3426 0.1834 0.3689 0.5379 0.4311 0.3906 0.7905 0.3800
2.33 2.67 5.09 3.34 3.09 3.36 4.97 0.33 1.76 0.62 0.43 0.46
0.020 0.008 0.000 0.001 0.002 0.001 0.000 0.739 0.079 0.538 0.669 0.648
0.0581 0.1858 0.7492 0.2578 0.3869 0.2560 1.1092 1.2333 0.0865 0.5251 1.2111 0.9183
0.6728 1.2114 1.6889 0.9877 1.7300 0.9751 2.5551 0.8753 1.6033 1.0060 1.8878 0.5714
Standard deviation BEV PHEV FCEV Fuel Subsidy Parking fee Emission based tax Distance to next charging point None option
1.9817 1.7388 1.6300 0.6155 3.0777 0.1169 0.1612 0.0359 3.5287
0.1207 0.1148 0.1917 0.0459 0.1839 0.0393 0.0433 0.0112 0.2079
16.42 15.15 8.50 13.41 16.73 2.98 3.72 3.21 16.98
0.000 0.000 0.000 0.000 0.000 0.003 0.000 0.001 0.000
1.7451 1.5139 1.2543 0.5255 2.7172 0.1938 0.0763 0.0578 3.1213
2.2182 1.9638 2.0056 0.7054 3.4382 0.0400 0.2461 0.0140 3.9362
118
C. Rudolph / Transport Policy 52 (2016) 113–120
cars financially attractive to people that do not possess or use a car. Rising incentives might eventually bring more cars onto the roads. The same effect is shown observing public transport users. The likelihood of choosing a ZEV by people using PT increases significantly with the frequency of use. 3.3. Change in probability of choosing ZEVs within different mobility groups In the next step, four different behavioral homogeneous groups with distinct daily mobility routines are defined endogenously. The groups should be clearly divided by their daily mobility routines. To avoid non-realistic groups that might occur using cluster analysis methods, the groups were defined as follows: A. Everyday PT users: In this group are people who use PT usually every day. The mileage per year is below 15,000 km. They hold an annual transit pass. Their mobility routines are focused on PT. PT is the most important means of transport. B. Everyday car drivers: This group contains persons who use their car every day. Their average yearly mileage is above 15,000 km/year and they do not have an annual transit pass. They never use PT or a bicycle. The car is the most important means of transport. C. Multimodal users – with car affinity: For this group multiple means of transport are in scope. But the car is the preferred means. Like group A they have a yearly mileage of more than 15,000 km/year. Bike and Public Transport are used sometimes. People in this group do not have an annual transit pass. The car is the most important means of transport. D. Multimodal users – with PT affinity: People of this group do not drive more than 15,000 km per year. They use bicycles and cars only sometimes to come around. They hold an annual transit pass. PT is the most important means of transport. Table 6 shows the number of respondents in each mobility group and the attributes of each group. As above, the same specific incentives for each group are simulated. The change in likelihood of choosing a ZEV in respect to each incentive according to the estimated model using formula (2) is given in Table 7. The integration of interaction variables representing the respondents’ mobility behavior shows a similar result like Table 4. Incentives have different impact according to type and benefit for the consumer. The difference in between the distinct incentives may differ heavily when increasing the absolute monetary benefit for consumers. But most interestingly, the integration of these interaction variable shows, that incentive may affect different groups which are homogenic in terms of mobility behavior very distinctly. For example, direct subsidies or a sensitive increase of
Table 7 Effects of incentives on each group. Group
A
B
C
D
Reference scenario L (ZEV)a [%]
5.6
32.8
10.0
47.6
þ 5.8 þ 0.6 þ 1.5 þ 5.2
þ 18.5 þ 2.3 þ 5.9 þ 16.8
þ 9.3 þ 1.0 þ 2.6 þ 8.5
þ 18.6 þ 2.6 þ 6.4 þ 17.0
þ 0.8
þ 3.2
þ 1.4
þ 3.6
Change of L(ZEV) due to incentive [%] 1) purchasing grant for ZEV: 10,000 € 2) CO2tax for ICV: 500 € 3) free parking for ZEV 4) increase of fuel costs for ICV by tax increase: 100% 5) increase of availability of charging infrastructer for ZEV a
L(ZEV): Σ L(BEV, PHEV, FCEV); L ¼Likelihood.
fuel costs of 100% plus may increase the likelihood for group B and D by about 18.5% points while the likelihood for group C may increase by 9.3% points and for group A only by 5.8% points. That means people with a more ecological mobility behavior may be affected in a much stronger way than people using a passenger car very frequently. This phenomenon can be observed, although on different levels, with all tested incentives.
4. Implications for policy making The vehicle purchasing models which are developed in this study base on a hypothetical experiment (stated choice). In fact, stated choice experiments can never reflect real choices unlike revealed choice data reflecting choices which have been made in reality. Nevertheless, the results base on a representative number of respondents and reflect their probable choice against the background of given incentives. The main objective of this study is combining purchase decisions regarding the propulsion technology with the mobility behavior of the respondent in order to assess the ecological effect of subsidies for ZEV. The results of the model show that subsidizing ZEV may increase the likelihood to choose an alternatively propelled vehicle. All subsidies which are tested in this study (direct purchasing grant, vehicle tax reduction, emission based parking costs, fuel taxation and availability of charging infrastructure) will increase attractiveness of ZEV over conventional vehicles with combustion engines. As long as ZEV have too high market prices and the market penetration of ZEV is still marginal, subsidies can help making ZEV competitive with conventional cars. Zero emission vehicles are considered to be one solution to reduce local emissions (e.g. NOx and SOx) improving air quality in urban areas and to reduce greenhouse gas emissions (e.g. CO2). But, the results of the models show also that two rebound effects may encounter when giving subsidies for ZEV:
Table 6 Mobility groups. Attribute of mobility behavior Everyday car drivers (n¼ 35)
Everyday PT users (n¼ 21)
Multimodal users – with car affinity Multimodal users – with car affinity (n¼ 93) (n ¼36)
Yearly mileage below 15,000 km Yearly mileage above 15,000 km Annual transit pass Bicycle use daily Bicycle use sometimes Bicycle use never PT use daily PT use sometimes PT use never
0
1
0
1
1
0
1
0
0 0 0 1 0 0 1
1 0 0 1 1 0 0
0 0 1 0 0 1 0
1 0 1 0 1 0 0
C. Rudolph / Transport Policy 52 (2016) 113–120
Firstly, by increasing the attractiveness of ZEV also for people having mobility patterns with a low energy ratio (on foot, bicycle, PT) by giving subsidies may lead to higher sales amongst these mobility groups. Due to the eco label of ZEV plus economical feasible may increase the likelihood to shift to individual motorized transport (passenger car) for these people as well. The overall energy ratio for transport would then be worse for these people than before. From an ecological perspective, this result is a dilemma. The overall objective to exchange as many combustion engines as possible by ZEV may fail. Under given energy production mix in Germany CO2 emissions would also increase. Secondly, another rebound effect may encounter: urban transport policies in the last decades tried to increase the attractiveness of non-individual motorized means of transport, i.e. on foot, bicycle and PT. The models show that these mobility groups would exploit such subsidies most likely. Increasing the attractiveness of passenger cars for all mobility groups would torpedo urban transport policies aiming in decreasing individual motorized transport. It would lead to an increase of the total number of cars, the total number of vehicle miles traveled and the amount of space needed in urban areas to accommodate these vehicles – in terms of roads and parking facilities. These two effects have major implications for policy makers: In order to prevent the first rebound effect, policy makers must assure that electricity used for individual mobility comes from 100% real physical renewable sources. Using only certified green energy will improve the CO2 emission ratio within the transport sector but worsen the ratio in all other sectors as long as the energy from the grid is not 100% renewable. Only then, ZEV can also contribute to reduce greenhouse gas emissions in the transport sector. Policy makers have to ensure that electricity which is used for ZEV taken from public charging facilities as well as from private plugs have to be real green energy. In parallel, the installation of further renewable energy plants has to be successively conducted. In order to prevent the second rebound effect, policy makers should parallelly invest to increase the attractiveness for all other means of transport when fostering low emission vehicles. To increase the attractiveness making errands on foot, the local supply (e.g. small local retailers etc. to establish acity of short distances) should be fostered and the infrastructure for pedestrians should be improved. To increase the attractiveness to use the bicycle on a daily basis for commuting and making errands a safe, direct, fast, comfortable and coherent infrastructure should be provided. Finally, the public transport should be expanded steadily. Therefore, independence from individual means of transport is essential as well as vehicles which are fast, safe to use and frequently run. Recently, the German government decided to introduce a €1.2 billion stimulus program granting BEV purchases with up to €4,000 and HEV with up to €3,000 depending on the price of the vehicle (Bundesregierung). Half of the subsidy will be given by the German government; the other half will be paid by the car industry. Further 300 million Euros will be invested for the installation of 15,000 charging facilities all across Germany. Additionally, for BEV no yearly vehicle tax has to be paid for ten years after purchase. Against the background of this study, policy makers should run a research program to investigate which real mobility behavior the buyers of ZEV have had before the purchase and which ecological effects, which have been discussed above, will be induced in reality. A cap of the subsidies should be considered enabling the government to stop the program after a first pilot in case accompanying research shows that purchasers' mobility profiles show a lower energy ratio than after the ZEV purchase. In order to prevent these rebound effects counteracting decisions must be taken in advance or in parallel but not afterwards as a reaction. After all, introducing policies subsidizing expensive goods, as
119
zero emission vehicles are, means subsidizing the most affluent social class of a society. ZEV are often purchased as a second or third car. People which can’t afford a zero emission vehicle will never profit from such subsidies. Such policies are socially unjust, especially, if no money is spent to improve other means of transport. Policy maker have to consider this fact when making the tradeoff where to spend tax money on.
References Achtnicht M., Bühler G. and Hermeling C., 2008. Impact of Service Station Networks on Purchase Decisions of Alternative-fuel Vehicles, ZEW Discussion Papers No. 08-088. Achtnicht, M., 2009. German Car Buyers' Willingness to Pay to Reduce CO2 Emissions: Discussion Paper No. 09-058, ZEW Zentrum für Europäische Wirtschaftsforschung GmbH: Mannheim, p. 32. Axsen, J., Mountain, D.C., Jaccard, M., 2009. Combining stated and revealed choice research to simulate the neighbor effect: the case of hybrid-electric vehicles. Resour. Energy Econ., 221–238, Jg. 31, H. 3. Brownstone, D., Bunch, D.S., Train, K.E., 2000. Joint mixed logit models of stated and revealed preferences for alternative-fuel vehicles. Transp. Res. Part B: Methodol. 34 (5), 315–338. Dagsvik, J.K., et al., 2002. Potential demand for alternative fuel vehicles. Transp. Res. Part B: Methodol. 36 (4), 361–384. Daziano, R.A. and Achtnicht, M., 2012. Forecasting Adoption of Ultra-low-emission Vehicles Using the GHK Simulator and Bayes Estimates of a Multinomial Probit Model: Discussion Paper No. 12–017, Mannheim, p. 32. Ewing,, G.O., Sarigöllü, E., 2000. Assessing consumer preferences for clean-fuel vehicles: a discrete choice experiment. J. Public Policy Mark. 19 (1), 106–118. Federal Motor Transport Authority, 2015. Number of Passenger Vehicles in Germany 01/01/2015 According to Fuel Type (Bestand an Pkw Am 1. Januar 2015 nach ausgewählten Kraftstoffarten). Federal Republic of Germany, 2009. German Federal Government's National Electromobility Development Plan. Berlin. German Institute for International Educational Research – DIPF. Bildungsbericht 2012: Tabellenanhang. Available from: 〈http://www.bildungsbericht.de/da ten2012/tabellen_web2012.pdf〉. Green,, P.E., Srinivasan,, V., 1978. Conjoint analysis in consumer research: issues and outlook. J. Consum. Res., 103–123. Hensher,, D.A., Rose, J.M., Greene, W.H., 2008. Combining RP and SP data: biases in using the nested logit ‘trick’ – contrasts with flexible mixed logit incorporating panel and scale effects. J. Transp. Geogr. 16 (2), 126–133. Hole,, A.R., 2007a. Estimating mixed logit models using maximum simulated likelihood. Stata J. 7 (3), 388–401. Hole,, A.R., 2007b. Fitting mixed logit models by using maximum simulated likelihood. Stata J. 7 (3), 388–401, 14. Breivik, I.F. and Volder, M.O., 2014. Electric Vehicles in Norway – A Qualitative Study of the Electric vehicle Market in Norway. infas, Mobility in Germany 2008 - Report. 2010. Ito,, N., Takeuchi,, K., Managi,, S., 2013. Willingness-to-pay for infrastructure investments for alternative fuel vehicles. Transp. Res. Part D: Transp. Environ. 18, 1–8. Jenn, A.T., 2014. “Advanced and Alternative Fuel Vehicle Policies: Regulations and Incentives in the United States”, Dissertations. Paper 354. 〈http://repository. cmu.edu/dissertations/354〉. Knockaert, J., 2005. The Choice for Alternative Cars. Energy, Transport and Environment Center for Economic Studies, Leuven, Belgium. Lave, C.A., Train, K.E., 1979. A Disaggregate Model of Auto-Type Choice. Transpn Res 13A. Pergamon Press Ltd, pp. 1–9. Akiva, M.E.B. and Lerman, S.R., 1985. Discrete Choice Analysis: Theory and Application to Predict Travel Demand. Mabit,, S.L., Fosgerau,, M., 2011. Demand for alternative-fuel vehicles when registration taxes are high. Transp. Res. Part D: Transp. Environ. 16 (3), 225–231. Manski,, C.F., Sherman,, L., 1980. An empirical analysis of household choice among motor vehicles. Transp. Res. Part A: Gen. 14 (5–6), 349–366. McFadden,, D., Train,, K.E., 2000. Mixed MNL models for discrete response. J. Appl. Econ. 15 (5), 447–470. Potoglou,, D., Kanaroglou, P.S., 2006. Household Demand and Willingness to Pay for Alternative Fuelled Vehicles. Hamilton, ON; Canada, p. 31. Rabe-Hesketh,, S., Skrondal,, A., 2012. Multilevel and Longitudinal Modeling Using Stata, 3 ed., College Station; Tex: Stata Press. Revelt,, D., Train,, K.E., 1998. Mixed logit with repeated choices: households' choices of appliance efficiency level. Rev. Econ. Stat. 80 (4), 647–657. Small, K.A., Winston, C., Yan, J., 2005. Uncovering the distribution of motorists’ preferences for travel time and reliability. Econometrica 73 (4), 1367–1382. Statistical Office for Hamburg and Schleswig-Holstein., 2012. Population of Hamburg 2012 (Bevölkerung in Hamburg 2012), 10.10.2013; Available from: 〈http:// www.statistik-nord.de/uploads/tx_standocuments/SI13_117.pdf〉. Bundesregierung, Steuervorteile und Kaufprämie beschlossen (engl: tax advantages and purchase grant decided), Available from: 〈https://www.bundesregierung. de/Content/DE/Artikel/2016/05/2016–05-18-elektromobilitaet.html〉.
120
C. Rudolph / Transport Policy 52 (2016) 113–120
Thomas, L., 1983. Der Einfluß von Kindern auf Produktpräferenzen ihrer Mütter. (Schriften zum Marketing. Duncker & Humblot, Berlin, p. 6598. Train,, K.E., 1980. A structured logit model of auto ownership and mode choice. Rev. Econ. Stud. 47 (2), 357–370. Train,, K.E., 1998. Recreation demand models with taste differences over people. Land Econ., 230–239. Train,, K.E., 2003. Discrete Choice Methods with Simulation, 2nd ed. Cambridge Univ. Press, Cambridge, p. 334. Train, K.E., 1986. Qualitative Choice Analysis: Theory, Econometrics, and an
Application to Automobile Demand, Cambridge. Raich, U., Link, C., Stark, J., 2012. Möglichkeiten und Grenzen der individuellen Elektromobilität. Kauf von Elektroautos, in Technische Universität Dresden, Tagungsband Verkehrswissenschaftliche Tage. Zhou, Yan, et al., 2015. Plug-in electric vehicle market penetration and incentives: a global review. Mitig. Adapt. Strateg. Glob. Change 20 (5), 777–795. Ziegler,, A., 2012. Individual characteristics and stated preferences for alternative energy sources and propulsion technologies in vehicles: a discrete choice analysis for Germany. Transp. Res. Part A: Policy Pract. 46 (8), 1372–1385.