Vous êtes sur la page 1sur 10

Energy Policy 94 (2016) 94–103

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

The effect of policy incentives on electric vehicle adoption


Joram H.M. Langbroek n, Joel P. Franklin, Yusak O. Susilo
Department of Transport Science, KTH Royal Institute of Technology, Teknikringen 10, 100 44 Stockholm, Sweden

H I G H L I G H T S

 Policy incentives have a positive influence on electric vehicle adoption.


 Being in advanced stages-of-change to EV-adoption increases likelihood to adopt EVs.
 People in more advanced stages-of-change to EV-adoption are less price-sensitive.
 People with a high self-efficacy and response efficacy are more likely to adopt EVs.

art ic l e i nf o a b s t r a c t

Article history: In order to increase the attractiveness of electric vehicles (EVs), packages of policy incentives are pro-
Received 9 December 2015 vided in many countries. However, it is still unclear how effective different policy incentives are. Also, it is
Received in revised form questionable that they have the same impact on different groups of people. In this study, based on a
29 March 2016
stated-choice experiment, the effect of several potential policy incentives on EV-adoption, as well as the
Accepted 30 March 2016
Available online 13 April 2016
influence of socio-psychological determinants are investigated, using constructs of the Transtheoretical
Model of Change (TTM) and the Protection Motivation Theory (PMT).
Keywords: The probability of stated EV-adoption increases if policy incentives are offered in the choice ex-
Electric vehicle adoption periment, which is expected because of the decrease of the generalized cost of EV-use. The high stated
Policy incentives
valuation of free parking or access to bus lanes makes those incentives an efficient alternative to ex-
stated-choice
pensive subsidies. EV-adoption probability increases for people that are further in the process of beha-
Transtheoretical Model of Change
Protection Motivation Theory vioural change. However, the responsiveness to subsidies decreases for people in more advanced stages-
of-change. People that believe EVs to be effective in decreasing the negative externalities of the current
transport system and people whose travel patterns can cope with the use of EVs also have a higher
probability to choose the EV.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction uptake of EVs is that these vehicles have a comparatively high


investment cost because of the high cost of batteries (Newbery
Electric vehicles (EVs) can decrease the dependence of the and Strbac, 2015). Therefore, in many countries in Europe policy
transport sector on fossil fuels, which has environmental ad- incentives are provided in order to stimulate the purchase of
vantages. A decrease of local exhaust emissions can increase the electric vehicles (e.g. Lieven, 2015).
air quality and decrease health problems that are related to air Several stated preference studies have been conducted in order
pollution (e.g. Vienneau et al., 2015), especially in urban areas. On to elicit people's willingness to pay for different vehicle char-
acteristics of electric vehicles (e.g. Golob et al., 1993, Hidrue et al.,
a global level, EVs may decrease the CO2-emissions related to
2011). As policy measures intervene with the generalized costs of
personal transport, especially if the electricity that is used can be
EV-use, they can be considered as attributes of EVs that can be
generated using nuclear energy or renewable energy sources such
influenced by governments. However, relatively few stated pre-
as wind, water and solar power and if many EVs charge in off-peak ference experiments have been explicitly covered packages of
electricity demand hours (Jochem et al., 2015). However, the up- policy incentives that are currently taken in order to stimulate the
take of EVs is still relatively low. One of the reasons for the slow uptake of electric vehicles (e.g. Lieven, 2015; Helveston et al., 2015;
Lin and Greene, 2012; Zhang et al., 2011). Most of these papers
n
Corresponding author.
discuss one policy measure in isolation. Helveston et al. (2015)
E-mail addresses: joram.langbroek@abe.kth.se (J.H.M. Langbroek), investigated how much subsidies would be needed to obtain a
joel.franklin@abe.kth.se (J.P. Franklin), yusak.susilo@abe.kth.se (Y.O. Susilo). specific market share of electric vehicles in the USA and in China,

http://dx.doi.org/10.1016/j.enpol.2016.03.050
0301-4215/& 2016 Elsevier Ltd. All rights reserved.
J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103 95

while Lin and Greene (2012) investigated the WTP of public 2.1. Taxonomy of policy incentives related to EV-adoption
charging infrastructure, concluding that improvements of home
charging facilities would have a higher value than improvements Policy measures that are taken to make EVs more attractive can
of public charging or charging at the workplace. A study by Lieven be of different kinds (e.g. Sierzchula et al., 2014; Lieven, 2015).
(2015) compares the effects of policy measures in twenty coun- Purchase-based and use-based incentives can be distinguished. An
tries. Moreover, cluster techniques are used in order to distinguish example of a purchase-based policy incentive is a subsidy when
different groups of travellers based on their stated preferences and buying an EV or a tax rebate when registering an EV. Examples of
demographics likely to react differently to policy incentives. use-based policy measures are providing free parking for EVs, al-
Whilst different traveller groups have different preferences, lowing EV-users to make use of bus lanes or to provide an ex-
most policy incentives to make EVs more attractive are directed to emption for congestion charging for EV users. Purchase-based
all car drivers, without any particular group to target. Moreover, incentives decrease the fixed cost of EV-use, while use-based in-
policy incentives will have implications that are likely to differ centives decrease the marginal cost of EV-use. Seen from another
between different categories of car drivers. Considering beha- perspective, policy incentives can be of a local or global kind. Local
vioural change as a process, different car drivers are in different policy incentives such as free parking in a specific city, congestion
stages-of-change towards electric vehicle use. Moreover, they have charge exemptions in a specific city or access to bus lanes are
likely to only influence people who can gain from these location-
different attitudes towards sustainable transport and electric ve-
specific incentives. On the other hand, global policy incentives
hicles in particular and different needs depending on where they
such as national tax rebates (vehicle registration tax or annual
live. All of these aspects may influence EV adoption in general, as
road taxes) or subsidies apply to everyone in a country. The ef-
well as the responsiveness to policy measures.
fectiveness and efficiency of policy incentives, as well as their
In this article, the effects of policy incentives on EV adoption
feasibility and side effects are assumed to depend on the scope
are investigated, in order to evaluate the effectiveness and effi-
and timing of incentives, as described above. In the following
ciency of different possible policy incentives. Socio-psychological
Section, each of these concepts will be shortly described.
constructs from the Transtheoretical Model of Change and the
Protection Motivation Theory, which will be described in Section
2.2. Effectiveness, efficiency, feasibility and side effects of policy
2.2, have been used. Different groups of car drivers have been
incentives
distinguished in order to investigate the probability of EV adoption
for people in different stages-of-change, people having different
It has been shown that policy incentives for EVs in general have
self-efficacy, response-efficacy and willingness to decrease their
a positive effect on EV adoption (Zhang et al., 2011). Policy in-
car use and people that live on a different distance from the city
centives can be considered to be effective if they considerably
centre of Stockholm. increase the probability of buying an electric vehicle. If policy
The hypotheses for this paper are the following: measures do not increase EV adoption, they will only imply a re-
distribution of income. The effectiveness of a policy measure can
Hypothesis 1. Policy incentives have a positive effect on EV
be measured as the difference between the number of EVs sold
adoption.
with a specific policy incentive n1 and the number of EVs sold
Hypothesis 2. People that are further in the process towards be- without that specific policy incentive n0.
havioural change according to the Transtheoretical Model of Bakker and Trip (2013) investigated the effectiveness, efficiency
Change (TTM) are more likely to adopt EVs. and feasibility of different policy measures that can be taken to
stimulate electric vehicle adoption using expert workshops. In this
Hypothesis3. The effects of policy incentives on EV adoption are study, efficiency has been defined as the amount of resources that
different for people in different stages of change. are needed to make one person adopt an EV, which can be vi-
sualized as follows:
Hypothesis 4. People that are motivated to change to EVs are
n1*costpolicymeasureXpersoldcar
more likely to adopt EVs. Efficiency =
n1 − n 0 (1)
Hypothesis 5. The effect of local policy incentives is different for where n1 ¼the number of EVs sold with policy incentive X and n0
people that live in different parts of the region of Stockholm. the number of EVs sold without policy incentive X.
The rest of this paper is structured as follows: In Section 2, the An example: suppose that a subsidy of 40,000 SEK (approxi-
theoretical background is described, followed by the Methodology mately €4300) increases the sale of EVs by 20 per cent from 1500
of this paper in Section 3. In Section 4 an overview is given of the vehicles to 1800 vehicles in a specific year.
Results, followed by the Discussion in Section 5 and Conclusions in The subsidy has to be paid for all sold EVs, and supposing the
marginal cost of public funds equals 1.3, the total cost would equal
Section 6.
93,600,000 SEK ( E€10,090,000). The effect of this incentive is an
increase in the sale of EVs by 300, which implies that the cost is
312,000 SEK ( E€33,600) for each additional vehicle, disregarding
2. Theory the benefits of this re-distribution of income for the receivers of
the subsidy.
In this section, assessment methods of different policy in- Feasibility of policy incentives consists of an evaluation of the
centives, that are used to increase the attractiveness of electric financial, social and political costs of policy measures (Bakker and
vehicles, are discussed. These concepts will be used in the Dis- Trip, 2013). For some measures, it is hard to get political and public
cussion session later on, in the light of the empirical results of this support for implementing this measure, even though they might
study. In Section 2.1, a taxonomy of policy incentives is presented. be very effective and efficient. Another issue pointed out by Bakker
In Section 2.2, different assessment methods of policy incentives and Trip was the task of governments to raise awareness. Many
are described. In Section 2.3, the connections between motivation people are currently not aware of the electric vehicle, its possibi-
and policy incentives are discussed. lities and its limitations, nor of the package of policy incentives
96 J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103

that is available, which makes them less likely to adopt an EV.


Policy incentives can have side effects: effects that are not di-
rectly related to the objectives of the policy incentives but that are
related to the kind of the policy incentives. As mentioned above,
subsidies can be considered as a re-distribution of income, which
is a side effect. Other potential side effects concern changes in
travel behaviour which may cause rebound effects. For example, in
Stockholm there is a congestion charging scheme. Before 2012,
clean vehicles such as electric vehicles or ethanol vehicles were
exempt from paying the congestion charges. Due to this exception,
many people purchased a clean vehicle and these vehicle owners
were less likely to change their travel patterns in order to decrease
congestion (Whitehead et al., 2015). When comparing fiscal policy
that is currently used in some countries to stimulate EV-use, Fig. 1. Intrinsic and extrinsic motivation.
compared to registration tax exemptions, exemptions on the an-
nual or monthly road taxes are not considered to be very effective
and efficient (Gass et al., 2014), because of the fact that these taxes vehicle. Other people might only be interested in purchasing EVs
are not up-front costs but rather returning costs that have to be in case they would have been offered huge amounts of benefits.
paid regularly, almost comparable with the costs of re-fuelling a The intrinsic motivation of someone to start using EVs depends
few times per month. As the upfront purchasing cost of EVs is on personal aspects and on attributes of the EV. Rezvani et al.
higher than for ICEVs, road tax exemptions do not target the (2015) identified technological, social and personal, contextual and
crucial disadvantage of EVs. Related to this, other incentives cur- cost factors that have an influence on electric vehicle adoption in a
rently in use in some European countries are free parking places or review of current EV adoption literature. These factors cover both
the right to make use of bus lanes, which again do not target the intrinsic and extrinsic factors that have an effect on the willingness
high upfront purchasing cost of EVs. On the other hand, these to adopt EVs. Part of the early EV adopters start using EVs because
types of policy measures might be more feasible because of the of environmental reasons (e.g. Egbue and Long, 2012; Krupa et al.,
fact that these incentives cost less and do not have to be financed 2014).
for a person at one moment but rather scattered over time. The Protection Motivation Theory (Rogers, 1975) is a theory
Green et al. (2014) postulated that instead of targeting main- that deals with motivations for behavioural change, and this the-
stream car users, policy incentives could be more efficient and ory has also been applied to EV adoption (Bockarjova and Steg,
effective if they were tailored towards specific niche markets and 2014). If the risks of the current behaviour are considered to be
early adopters. Providing incentives to certain companies or in- high and if the severity of the consequences of the current beha-
dividuals that are already more inclined to buy electric vehicles viour are considered to be high (threat appraisal), as well as if a
implies less extensive policy measures to be taken, thereby in- new behaviour is considered to decrease the threat and if someone
creasing the efficiency. In the next subsection, the relation be- has a high self-efficacy to perform the new behaviour (coping
tween motivation to adopt electric vehicles, process of behavioural appraisal), then there is a high “Protection Motivation”. Bockarjova
change towards EV-use and effectiveness and efficiency of policy and Steg (2014) found that people are more likely to adopt EVs if
incentives will be elaborated. they think that the consequences of our current travel behaviour
are severe and if they believe that EVs can decrease these
2.3. Motivation for behavioural change: intrinsic and extrinsic externalities.
Another socio-psychological model that has been developed to
The electric vehicle itself is a product that has some char- deal with behavioural changes is the Transtheoretical Model of
acteristics such as a certain design, comfort level, maximum speed, Change. This model considers behavioural change as a process
price and range. Those characteristics depend on the current state rather than an event. This process consists of different stages and
of technology and the choices that different car manufacturers assumes that people follow a certain learning curve (e.g. Pro-
have made. This core product, as well as the characteristics of the chaska, 1991). In the first stage, the Pre-contemplation stage,
potential user, provides some intrinsic motivation for somebody to people have not yet considered to change their behaviour. In the
purchase an EV. Intrinsic motivation can be defined as an actual Contemplation stage, people have started to consider changing
and internal tendency to perform an action, while extrinsic mo- behaviour, while they plan to change behaviour within a specific
tivation is about a separable consequence that is separate from the time frame in the Preparation stage. In the Action stage, the actual
actual action (Ryan and Deci, 2000). behavioural change takes place, while people in the Maintenance
Policy incentives to stimulate electric vehicle use influence the stage have had some experiences with the new behaviour.
attributes of the electric vehicle because they influence the gen- Based on a survey among active drivers in Greater Stockholm,
eralized costs of EV-use, and can therefore be seen as influencing in which socio-cognitive constructs from the Protection Motiva-
people's extrinsic motivation to switch to EV-use. Subsidies, tax tion Theory and the Transtheoretical Model of Change have been
rebates or congestion charge exemptions influence the price of the included, it has been shown that stages-of-change towards electric
electric vehicle, while allowing making use of bus lanes influences vehicle adoption are correlated with both socio-economic char-
the travel time of driving, especially in congested areas. acteristics and socio-cognitive constructs such as knowledge, at-
Fig. 1 gives an illustration of four persons, each with their own titudes and self-efficacy (Langbroek et al., 2015a). Because these
level of “intrinsic motivation”, that are offered an identical package groups are dissimilar with respect to their position in the process
of policy incentives. The amount of policy incentives that a specific of behavioural change towards electric vehicle use, we assume
person needs to be convinced to start using an electric vehicle (or that people in different groups also react differently to policy in-
in the figure, surpasses the “level of indifference”) depends on this centives offered. As people tend to be more informed about the
person's intrinsic motivation level. A person that would have characteristics of electric vehicles and tend to have an ever in-
bought an EV regardless of whether any policy incentives were creasing self-efficacy towards using EVs, some policy measures
offered has a high intrinsic motivation to purchase an electric might for example have a larger or a smaller influence on people
J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103 97

that are already considering buying an EV. People who are cur- and their stage-of-change towards electric vehicle use, going from
rently using EVs might have an even higher intrinsic motivation to Pre-contemplation (not having considered to change to EV-use) to
buy a second EV in the future, so that they depend less on policy Maintenance (having at least half a year of driving experience in
incentives. However, the question is whether they are also un- an EV).
equally responsive to policy measures that can be taken in order to When making analyses about EV adoption and the effects of
increase the attractiveness of EVs. More insight into the effects of certain potential policy incentives, the effects of people being in a
policy measures on people that are in different stages-of-change specific stage-of-change have been incorporated in two different
can contribute to more insight into EV adoption strategies. ways: either as case-specific constants or as interaction variables
between policy measures and stage-of-change.
Several Mixed Logit models will be used as an analytical tool,
3. Methodology because those models can account for panel effects due to re-
peated observations.
3.1. Survey design
3.2. Sample description
The respondents for this study have taken part in a stated
choice experiment, which was part of a larger two-stage survey. For this study, the data of a two-stage survey was used. In the
They were asked which vehicle they would like to buy in the fic- first stage, socio-cognitive questions about electric vehicles were
tive situation they were planning to buy a new personal car. One of asked. Beside this behavioural questions including a one-day travel
the vehicles was an electric vehicle and the other vehicle a con- diary and socio-economic questions were asked. In total, 294 re-
ventional vehicle. spondents completed this first part of the survey. The second stage
The attributes that were taken into consideration were the of the survey consisted of stated adaptation questions regarding
range of the car, the price of the car after subsidies, whether there EV adoption and use as well as charging behaviour. The total
was a public charging system available for users, whether EVs can number of respondents who completed both stages is 269. This
park for free or with discount and whether EV drivers are allowed means that more than 90% of the respondents participating in the
to make use of bus lanes. Those attributes were selected in order first stage also participated in the second stage. Because the same
to capture common policy incentives covering different types: respondents were used for both the first and the second wave of
global subsidies reflected in the purchasing price of EVs, local this survey, there is a possibility to link the elements of the survey
parking policy (in Stockholm), local access to bus lanes (in Stock- in different ways.
holm) and a facilitating, local network of charging infrastructure, After participation in both survey stages, the respondent got
whether or not the electricity to charge the car is free. As the range free cinema tickets or a present card and tickets for a 22 h cruise in
of EVs is a fundamental limiting factor, this attribute has been the archipelago of Stockholm.
taken into consideration as well. The respondents that do not use electric vehicles have been
The attributes of the electric vehicle (see Table 1) had attribute selected using stratified random sampling, while the respondents
levels that were composed using an orthogonal fractional factorial that use EVs have been oversampled because of their small part in
design with blocking, as described in Hensher et al. (2005). In the Stockholm population. Respondents were selected to meet a
total, there were nine blocks of three choices to make, so that 27 of set of criteria for the sample regarding the minimum number of
the in total 81 profiles in the according full factorial design were respondents per age groups, per geographic sub-area within the
selected. The conventional car always was the reference car with Greater Stockholm region, per income level and per household
the same attribute level across the study. type. In September 2014, around 300 private EV users were re-
For the analysis design, dummy coding has been used, where gistered in Stockholm's county, and all of these were asked to
attribute level 1 (see Table 1) is always used as the reference at- participate in the survey. Out of 300 EV users, 122 were willing to
tribute. In order to provide the respondents with realistic choice participate in the survey. All respondents are active drivers driving
tasks, the purchasing price of the internal combustion engine ve- a car at least once a week.
hicle (ICEV) is always lower than that of the electric vehicle, while In Table 2, a short sample description is presented. Compared
the range is always higher. Moreover, re-fuelling take much less to the population of Stockholm County, the sample consists of a
time than charging an EV, as explained in the introduction of this larger proportion of men, the sample has a relatively high average
question. The disadvantage of this approach is the fact that not all age (50 years) and a comparatively large proportion of the sample
price- and range effects can be estimated and some of the effects lives in a single family house. Also the income among the sampled
will be hidden in the alternative specific constant, as described by respondents is relatively high, with 40 per cent of the respondents
Mabit and Fosgerau (2011). having a household income over 800,000 SEK (approximately
The respondents that took part in this stated choice experiment €86,000). This is mainly due to the fact that the EV-users in the
have also answered questions related to their attitudes, knowledge sample have an on average higher income and are more likely to

Table 1
Attributes and attribute levels.

Attributes Attribute level 1 (ref.) Attribute level 2 Attribute level 3 Attribute level 4

Range 800 km 200 km 150 km 100 km


Price after subsidiesa 250,000 SEKb 300,000 SEK 350,000 SEK 400,000 SEK
Public charging No public charging available Paid public charging available Free public charging available
Parking benefits No parking benefits for EVs in the city of 50% parking discount for EVs in the city of Free parking for EVs in the city of
Stockholm Stockholm Stockholm
Use of the bus lane Not allowed Only allowed outside the inner city of Allowed
Stockholm

a
Price level autumn 2014.
b
100 SEK E €10.
98 J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103

Table 2 of MNL-models are extreme value distributed with mean zero.


Sample description. Disadvantage of the MNL-model is the fact that this model as-
sumes the error terms to be independent and identically dis-
Variable Unit Values
tributed across alternatives and decision makers. Taste variations
Gender women - number (percentage) 115 (39%) and heterogeneity of decision makers can be accounted for by
Cars in the household 1-number (percentage) 140 (52%) making use of models of the Mixed Logit family. Moreover, Mixed
2-number (percentage) 95 (35%)
Logit models can account for panel effects (Train, 2009).
3 or more- number 31 (12%)
(percentage) In Mixed Logit models, the coefficients βia, are assumed not to
Age years – average 54 years be a point estimate but follow a distribution specified by the re-
Age 30–number of respondents 22 searcher (Train, 2009). The probability for decision maker n to
65þ number of respondents 55 choose alternative i depends on the systematic part of utility of
Higher education % of respondents 64%
Living in a villa % of respondents 38%
alternative i as compared to the systematic part of utility of the
High income (above 80,000 euro) % of respondents 41% other alternatives j out of choice set J (see Eq. (3)). However, as
Pre-contemplation stage-of- number of respondents 53 some of the estimated coefficients are random coefficients, Lni is
change integrated over all possible values of βn (see Eq. (4)).
Contemplation stage-of-change number of respondents 99
Preparation stage-of-change number of respondents 8 eVni (β ′ni )
Action stage-of-change number of respondents 43 L ni (βn ) =
Maintenance stage-of-change number of respondents 66 ∑j ∈ J eVnj (β′ nj ) (3)

live in a single family house. Also the division of the respondents Pni = ∫ L ni(β n ′)
*f (β′)dβ′ (4)
over the five stages-of-change (as described in Section 2.3) is
displayed in Table 2. The number of respondents in the Prepara- In this study, the random coefficients are assumed to be nor-
tion stage is very low (8 respondents). mally distributed with mean b and covariance W, resulting in

Pni = ∫ L ni(β n ′)
*φ(β′| b′, W ′′)dβ′. (5)
3.3. Model estimation
As all respondents have made three choices, this study has
According to Lancaster (1966), the utility of a product or service repeated observations. The choices of the same individual might
does not depend on the product or service as such but rather on the be correlated (panel effects). Therefore, the probability of the se-
attributes of this product or service. When making a choice between quence of choices T (going from observation t ¼1 until observation
more different products or services, the attributes are compared and t¼3) of decision maker n are estimated as in Eq. (6):
the attribute levels of those attributes are crucial when making a T
choice between different alternatives. This concept has later been eVnit (β ′n )
L nit (βn ) = ∏
developed in the Random Utility Theory (e.g. Mc Fadden, 1974). Ac- t=1 ∑j ∈ J eVnjt (β ′n ) (6)
cording to this utility maximization framework (e.g. McFadden, 1974),
decision makers are assumed to be rational and try to maximize their By assuming the random coefficients to be normally distributed
utility when making choices. The higher utility a certain alternative is with mean b and covariance W, the probability of observing a
expected to give, the higher the chance that this alternative will be sequence of choices T for decision maker n is:
chosen by the decision maker. As the electric vehicle is an alternative T
that is often surrounded by a policy framework interfering with the Pni = ∫ ∏ L nit (β′)*φ(β′| b′, W ′′)dβ′.
generalized cost of EV-use, not only the core product-the vehicle itself- t=1 (7)
is assumed to provide utility, but supporting services and government For the models testing Hypothesis 1 until 5, different config-
incentives as well. In this study, the focus is on government incentives urations of Mixed Logit models with random coefficients and pa-
and their influence on EV-adoption. nel effects have been estimated, the effects of which are presented
Based on the utility maximization framework (e.g. Koppelman in Section 4. As a rule, only random coefficients with a statistically
and Bhat, 2006), the utility of an alternative i (Ui) consists of a significant (α ¼0.10) standard deviation have been incorporated as
systematic part Vi and an error term ε , where the systematic part random coefficients as a trade-off between information and the
consists of a vector of attributes a (e.g. price) with attribute values aim of a parsimonious model.
Xia (e.g. 400,000 SEK) for a given alternative i, and their marginal As the integral of Eq. 5 does not have a closed form, the
effect on utility βia and an Alternative Specific Constant ASC that probabilities Pni are approximated by simulation, making use of
captures systematic but non-explained variability in the data (see the distribution of the random parameters f(β). For up to 10 ran-
Eq. (2)). dom parameters, 100 scrambled Halton sequences have been
Ui = Vi + εwhereVi = β′i X ′i + ASCi (2) drawn as described in Bhat (2003).

Based on the utility maximizing framework, people tend to be


rational and maximize their utility, meaning that the chance for 4. Results
choosing option i depends on the utility of option i and on the
utility of all other alternative options in choice set J. The more In this section, five models will be discussed. These models
superior option i is compared to the other options, the higher the follow the different hypotheses that have been stated in Section 1.
probability that option i is chosen. Besides attributes of the alter- The topics of the different models is displayed in Table 3.
natives in a specific choice set, also the characteristics of decision
makers can play a role in choosing one alternative. Different 4.1. Effects of policy measures on EV adoption
functional forms or decision rules of discrete choice models have
been developed, out of which the Multinomial Logitmodel (MNL- The first model investigates whether the attributes parking
model) is a benchmark model due to its simplicity. The error terms (free/with discount), public charging (free/paid), access to bus
J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103 99

Table 3 Table 5
Description of Model 1 – Model 5. Coefficients and statistical tests for Model 2.

Model 1 Policy incentives Model 2 Estimates t-value p-value WTP (SEK)

Model 2 Policy incentives, stage-of-change and EV-adoption Alternative specific constant 11.036 3.644 0.000
Model 3 and 4 Stage specific effects of policy incentives electric vehicle
Model 5 Policy incentives, Protection Motivation and EV-adoption Contemplation 1.103 1.727 0.084
Preparation 3.903 2.568 0.010
Action 3.855 4.462 0.000
Maintenance 4.344 5.181 0.000
Table 4 Free parking 1.770 4.263 0.000 47,442
Coefficients and statistical tests for Model 1. 50% discount on parking 0.743 2.037 0.042 19,917
Access bus lanes Stockholm 1.200 2.746 0.006 32,159
Model 1 Estimates t-value p-value WTP (SEK) County
Use of bus lanes permitted out- 0.904 2.521 0.012 24,239
Alternative specific constant elec- 14.431 4.157 0.000 side the city
tric vehicle Free Charging 1.456 3.607 0.000 38,826
Free parking 1.782 3.985 0.000 52,556 Paid Charging 0.935 2.469 0.014 25,057
50% discount on parking 0.675 1.801 0.072 19,917 Range 0.018 3.976 0.000 485
Access bus lanes Stockholm 1.053 2.264 0.024 31,069 Price  0.039  6.623 0.000
County Std deviation Price 0.006 1.083 0.279
Use of bus lanes permitted out- 0.881 2.385 0.017 25,990 Std deviation Range 0.004 4.020 0.000
side the city Std deviation Free Charging 1.933 2.743 0.006
Free Charging 1.544 3.413 0.000 45,149 Std deviation Individual  1.031  2.718 0.007
Paid Charging 1.036 2.644 0.008 30,549 Values
Range 0.020 3.964 0.000 585 Log-likelihood  404.242
Price  0.039  5.641 0.000 Log-likelihood null model  559.370
Std deviation Price  0.013  2.154 0.031 Mc Fadden’s rho square 0.277
Std deviation Range  0.005  5.935 0.000 Adjusted McFadden’s rho square 0.247
Std deviation Free Charging 1.867 2.052 0.040 # iterations 248
Std deviation Individual  0.414  1.077 0.281
Values
Log-likelihood  433.66
Log-likelihood null model  559.370
Mc Fadden's rho square 0.225
Model 1 has a Mc Fadden's rho square value of 0.225 and has a
Adjusted McFadden's rho square 0.201 significantly higher explanatory power than the null model (LL
# iterations 166 Ratio test, p-value ¼0.000).

4.2. Stage-of-change and policy measures


lanes (in the whole county or outside the city), price including
subsidies and range play a role in influencing active car drivers in As the respondents in this study are in different stages in the
Stockholm to purchase an electric vehicle rather than a conven- process of behavioural change towards EV use, the second hy-
tional, internal combustion engine vehicle. For price, range and pothesis stated that decision makers that are earlier in the process
free charging policy, random parameters have been estimated, of change are less probable to adopt the electric vehicle than de-
because for these variables there was evidence of taste variability. cision makers that are further in the process of change. In Model 2,
Also a random parameter that accounts for panel effects has been this hypothesis is tested. Including stage-of-change in the condi-
estimated. The results of Model 1 are listed in Table 4. The values tional logit model increases the explanatory power of the model
in bold are statistically significant at α ¼0.10. considerably. The rho-square value increases from 0.225 to 0.277.
All included policy incentives have a statistically significant and Two clusters with more similar values can be distinguished: one
positive effect on electric vehicle adoption at a 90 per cent con- cluster with people in the Pre-contemplation and Contemplation
fidence level (see Table 3). In order to compare different policy stages and one cluster with people in the Preparation, Action and
measures and the range of the EV, the willingness-to-pay (WTP) Maintenance stages (see Table 5). The Willingness-to-Pay values
for each attribute except subsidies has been provided in Table 4. differ from the values in Model 1, but the ranking of WTP for the
Free parking in Stockholm is the non-monetary policy incentive different policy incentives remains the same.
with the highest value. The willingness to pay for free parking is People in different stages-of-change have different propensities
52,556 SEK (E€5665), which is higher than the amount of the to adopt electric vehicles. It might be that they also react differ-
current EV-subsidies that are given in Sweden (up till 40,000 SEK ently to policy measures that can be taken, in this case that a
or €4300). The second largest effect is the effect of a free charging certain policy measure has a different effect for people in different
system in Stockholm. The willingness to pay for a free charging stages-of-change. In order to investigate this phenomenon, the
system as opposed to a charging system where you have to pay is interaction effects between each stage-of-change and the stated
another 14,500 SEK (E€1560). The value of free parking according choice attribute variables have been modelled (see Table 6).
to this model is higher than the value of a range increase by 50 km. The only variable with statistically significant results for all
Range is an important variable influencing EV-adoption, but the stages-of-change is price. For the other variables, there is no such a
purely economic incentives such as free parking and free charging clear trend. Free parking has a significant result on people in the
seem to have a higher WTP. The use of bus lanes has a lower pre-contemplation, contemplation and maintenance stage. Dis-
willingness to pay than free parking or free charging. According to counts on parking only have a significant result on people in the
the estimated WTP of policy incentives, taking a large package of contemplation stage. The bus lane policy has a significant result on
policy measures that decrease the marginal cost of EV use (con- people in the pre-contemplation stage and maintenance stage, but
sisting of free parking in Stockholm, access to the bus lanes in not on people in the other stages. Free charging has a significant
Stockholm's county and an extensive and free charging system in influence on people in the contemplation stage and paid charging
Stockholm) would be similar to providing a subsidy of more than on people in the maintenance stage. Range has a significant in-
100,000 SEK (or €10,780). fluence on people in all stages but the preparation stage. Because
100 J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103

Table 6
Coefficients and statistical tests for Model 3.

Model 3 Estimate Estimate2 Estimate3 Estimate4 Estimate5


Coefficients PC C P A M

Alternative specific constant EV 7.874 7.874 7.874 7.874 7.874


Free parking 1.319 1.909 15.578 0.201 1.778
Parking with 50 per cent discount  0.425 1.576  1.075 0.379 0.476
Access bus lanes 2.569 0.741  1.275 0.511 1.355
Access bus lanes outside of Stockholm city 2.167 0.573 2.434 0.525 0.853
Free public charging 0.709 1.752 4.975 1.007 0.835
Paid public charging 0.008 1.184 4.485 0.188 1.071
Range (in km) 0.020 0.013 0.007 0.016 0.016
Price (in 1000 SEK)  0.044  0.038  0.031  0.028  0.030
Values
Log-likelihood  395.028
Log-likelihood null model  559.370
Mc Fadden’s rho square 0.294
Adjusted McFadden’s rho square 0.219
# Iterations 601

value in bold: p-value o 0.10


PC ¼ Pre-contemplation
C ¼Contemplation
P ¼ Preparation
A ¼ Action
M ¼Maintenance

of the high number of parameters, the adjusted McFadden's rho Price -sensitivity
square value of Model 3 is considerably lower than that of Model 0.000
-0.005
2.
-0.010
Even though not all stage-specific coefficients are significantly
Marginal utility

-0.015
different from zero, there are hardly any coefficients within a -0.020
specific attribute that are significantly different from each other (t- -0.025
-0.030
test for equality of parameters). This means that there is no evi-
-0.035
dence for stage-specific coefficients. An exception is price, where -0.040
many coefficients are statistically significantly different from each -0.045
-0.050
other. Therefore, a reduced model has been made with only stage- Pre -
Contemplation Preparation Action Maintenance
contemplation
specific price effects (see Table 7). Price -sensitivity -0.044 -0.040 -0.034 -0.032 -0.031
According to Model 4, the effect of price on EV-adoption di-
Fig. 2. Price-sensitivity for people in different stages of change (according to Model 4).
minishes with stage-of-change. This means that people that are in

Table 7
a more advanced stage-of-change are less price-sensitive towards
Coefficients and statistical tests for Model 4. price (see Fig. 2).
There are significant differences between six pairs of stage-
Model 4 Estimates t-value p-value specific price-coefficients (t-test for equality of parameters),
ASC 21.372 5.284 0.000
namely between the price-coefficient of people in the pre-con-
Free parking 1.663 4.347 0.000 templation stage and the coefficient of people in all four other
50% discount on parking 0.797 2.25 0.024 stages, as well as between the price-coefficient of people in the
Access bus lanes Stockholm County 1.103 2.636 0.008 contemplation stage and the price-coefficient of people in the
Use of bus lanes permitted outside the city 0.83 2.418 0.016
action and maintenance stage. In this model, the price-sensitivity
Free Charging 0.86 2.405 0.016
Paid Charging 0.016 3.752 0.000 is ever decreasing.
Range 1.325 3.319 0.001
Price Pre-contemplation  0.044  6.367 0.000 4.3. Socio-psychological determinants and EV adoption
Price Contemplation  0.04  6.64 0.000
Price Preparation  0.034  5.535 0.000
Price Action  0.032  5.745 0.000 In an earlier study, a connection was found between socio-
Price Maintenance  0.031  5.806 0.000 cognitive constructs from the Protection Motivation Theory and
Std deviation Price Pre-contemplation 0.002 1.515 0.130 being in a certain stage-of-change (Langbroek et al., 2015a). Stage-
Std deviation Price Contemplation  1.632  2.044 0.041
of-change has an influence on EV-adoption, where, controlled for
Std deviation Price Preparation 0.008 1.839 0.066
Std deviation Price Action 0.003 0.693 0.488 policy attributes in the stated choice experiment, people in more
Std deviation Price Maintenance 0.001 0.301 0.763 advanced stages-of-change are more likely to purchase an electric
Std deviation Range  0.007  2.253 0.024 vehicle rather than a conventional vehicle.
Std deviation Free Charging  0.005  1.707 0.088
For Model 5, the tested hypothesis was that constructs of the
Std deviation Individual 1.146 2.585 0.010
Values Protection Motivation Theory (threat appraisal, self-efficacy and
Log-likelihood  403.883 response efficacy) have an influence on EV-adoption. The electric
Log-likelihood null model  559.370 vehicle, able to decrease some of the environmental and health
Mc Fadden's rho square 0.278 problems that are related to the current transport system, is more
Adjusted McFadden's rho square 0.240
# Iterations 238
likely to be embraced in case people believe that the current
transport system is a danger for the environment (threat
J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103 101

Table 8 Table 9
Coefficients and statistical tests for Model 5. Comparison of nested models.

Model 5 Estimates t-value p-value Restricted Unrestricted LL-Ratio Degrees Critical Critical p-value
model model of free- value value
ASC 16.330 4.265 0.000 dom α¼ 0.05 α ¼0.01
Threat appraisal (min ¼1; max ¼7) 0.007 0.041 0.967
Self-efficacy (min ¼1; max¼ 7) 0.427 1.773 0.076 Model 1 Model 2 58.836 4 9.49 13.28 o0.01
Response efficacy (min ¼ 1; max ¼ 7) 0.388 1.884 0.060 Model 1 Model 4 59.554 8 15.51 20.09 o0.01
Free parking 1.662 4.429 0.000 Model 1 Model 5 71.278 11 19.68 24.72 o0.01
50% discount on parking 0.812 2.339 0.019 Model 4 Model 5 11.724 3 7.81 11.34 o0.01
Access bus lanes Stockholm County 0.956 2.394 0.017
Use of bus lanes permitted outside the city 0.798 2.407 0.016
Free Charging 1.288 3.487 0.000
explanatory power of the model as compared to Model 1. The
Paid Charging 0.966 2.707 0.007
Range 0.016 3.756 0.000 Likelihood Ratio values for both alternative approaches are very
Price Pre-contemplation  0.041  6.334 0.000 comparable; however including stage-specific price effects pro-
Price Contemplation  0.038  6.624 0.000 vides more information. An additional significant improvement
Price Preparation  0.034  5.532 0.000 can be made by including constructs from the Protection Motiva-
Price Action  0.032  5.901 0.000
Price Maintenance  0.031  6.058 0.000
tion Theory, that provide a significant improvement also after
Std deviation Price Pre-contemplation 0.007 2.021 0.043 stage-of-change specific price effects has been accounted for.
Std deviation Price Contemplation  0.001  0.392 0.695
Std deviation Price Preparation 0.000  0.026 0.980
Std deviation Price Action 0.006 2.205 0.027
5. Discussion
Std deviation Price Maintenance 0.002 0.614 0.539
Std deviation Range 0.004 5.790 0.000
Std deviation Free Charging 1.361 1.536 0.125 5.1. Effectiveness and efficiency
Std deviation Individual 0.027 0.045 0.964
Values Most incentives in the model significantly increase the prob-
Log-likelihood  398.021
Log-likelihood null model  559.370
ability of choosing an EV in this hypothetical setting. This was
Mc Fadden’s rho square 0.288 expected, because the generalized cost of EV-use is decreased by
Adjusted McFadden’s rho square 0.246 all of these policy incentives.
# Iterations 257 Also the hypothesis that people who are further in the process
of behavioural change (in a more advanced stage-of-change) are
more likely to choose EVs, given the provided policy package, has
appraisal), the EV has the potential to decrease the environmental been confirmed in this study. There is a non-linear positive re-
problems and health issues related to personal transport (response lationship between stage-of-change and EV adoption, controlled
efficacy) and in case people believe that they are able to satisfy for policy measures. This implies a different level of effectiveness
their travel needs with an electric vehicle rather than a conven- and efficiency of policy incentives, depending on how far people
tional vehicle (self-efficacy). When including the variables threat are in the process of behavioural change towards electric vehicle
appraisal, response efficacy and self-efficacy in the discrete choice use. The decreased price-sensitivity for people in further stages of
models that have been estimated above, there is a statistically change indicates that subsidization is less effective and therefore
significant and positive effect of response efficacy and self-efficacy less efficient for people in more advanced stages-of-change. Giving
on EV adoption (see Table 8). Threat appraisal does not have a one additional unit of subsidies to this group does not increase
significant effect on EV-adoption. their utility equally much as giving one additional unit of subsidies
to pre-contemplators. The other side of the coin is the fact that
4.4. Socio-economic characteristics and EV adoption they have a higher intrinsic motivation to purchase EVs and for
this group, the gap between the utility of conventional cars and
The explanatory power of the choice models that have been the utility of electric vehicles is smaller. The same holds for people
fitted does not increase by including socio-economic character- with a high response-efficacy towards electric vehicles and a high
istics such as gender, age, income, education level or mileage. self-efficacy.
Possibly, there is an indirect effect through the statistically sig- Ironically, if people would elicit whether they have a high
nificant attitudinal variables. Also the distance from the home Protection Motivation or whether they already consider changing
location to the city centre of Stockholm does not have a significant to EV use, governments could respond by decreasing the policy
effect on the uptake of EVs, nor on the effects of local policy incentives for this group and increase the efficiency of their EV
measures such as free parking and access to bus lanes. This has promoting policy.
been tested using a model with interaction effects between dis-
tance and certain policy incentives. The only socio-economic 5.2. Side effects
variable that has a statistically significant effect on EV adoption in
case stage-of-change has not been included deals with living in a Another argument for studying policy measures taken to in-
single-family house. crease the attractiveness of EVs is the presence of side effects that
can occur because of some policy measures taken. Side effects of
4.5. Comparison of model results policy incentives should be incorporated in an assessment of the
effectiveness of policy measures, as proposed by Givoni et al.
Model 1 is a restricted version of model 2, 4 and 5, while model (2013), which might make a case for including EV-policy in a
4 is a restricted version of model 5. Therefore, Likelihood Ratio broader “policy package” that does not only consider EV-adoption
Tests have been performed, the results of which are listed in but which takes potential rebound effects into account as well.
Table 9. Local incentives such as access to bus lanes or decreased parking
Both Model 2 and Model 4, incorporating stage-of-change ei- costs infer with the variable costs of EV-use rather than with the
ther as a dummy variable or as a specific price-effect, improve the fixed costs and might influence travel behaviour patterns. For
102 J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103

example, free parking decreases the marginal costs of EV use, bus lanes or to provide free parking. Bus lanes and parking tariffs
which are already naturally lower than the marginal costs of were originally meant to cope with excess demand on the road
conventional car use. Electricity is per unit cheaper than fossil network and on the “parking market”, thus providing road and
fuels and moreover, the engine of EVs is more efficient than the parking space for free to a part of the vehicles which is increasing
internal combustion engine, especially in urban areas where the is no sustainable option. On the other hand, if governments pro-
driving conditions imply a lot of braking and acceleration. This vide use-benefits, those policy incentives should be consistent
characteristic is advantageous for people that want to purchase over time in order to be credible and effective (White et al., 2013).
electric vehicles so they can in that way be compensated for the When making an investment, it is important for investors to
higher purchasing costs of these vehicles. However, lower mar- predict which policy measures will be taken in the future years
ginal costs of car driving might increase car driving, within the and whether current policy measures will still be in place in the
bandwidth of the available range of EVs (Langbroek et al., 2015b). future. When buying an electric vehicle, it is very important to
Bakker and Trip (2013) remarked that parking should not be free know how far into the future these use-benefits are going to stay.
because of the fact that free parking might increase the relative In this stated choice experiment, the policy measures have no end
attractiveness of the car over alternative travel modes. date, but if the government could only guarantee free parking or
Local, use-benefits that decrease the marginal cost of EV-use access to bus lanes the first year after purchase and announce that
can have a self-selection of a stimulus effect. A self-selection effect the benefit “might” stay for a longer time, the credibility and thus
would imply that people that pay a lot of parking fees are more the effectiveness of these incentives would be likely to go down.
inclined to switch to EVs if they get free parking in the city centre, Use-benefits have been found to have relatively high discount
so this policy measure targets a specific group of travellers. Tar- rates of 10–30 per cent (Greene et al., 2005).
geting at urban travellers would imply targeting in such a way that Seen from another perspective, crowding can make policy in-
the health effects are assumed to be larger. However, no effect of centives less attractive even though they are still in place. For
distance from the home location to the city centre of Stockholm on example, it has been observed that in Oslo, bus transport was af-
EV adoption has been found. A stimulus effect on the other hand fected by delays because of the high number of electric vehicles
would imply that increased EV-use is stimulated by use-benefits. making use of bus lanes (Meldalen, 2015). Crowding on bus lanes
In order to catch the highest possible amount of benefits, EV dri- decreases the time that can be saved in transport, and may
vers should drive as much as they can, park in central areas as therefore decrease the effectiveness of this policy incentive. The
much as they can and make use of bus lanes as much as they can. same counts for crowding if a limited number of free parking
All policy incentives that decrease the generalized cost of EV- places is offered.
use consist of a re-distribution of income. As the current EV-users In order to be effective, use-benefits must thus be in place
are on average wealthier than non EV-users, attention should be during a certain amount of time. During this period, the new travel
given to the equity effects of providing incentives such as subsidies behaviour that arises because of switching from an ICEV to an EV
to EV-users. becomes stable and travel habits are formed (Verplanken and
Wood, 2006). The question is how easily these changes in travel
5.3. Efficiency versus side effects habits, that are not necessarily beneficial for a sustainable trans-
port system, can be changed again when use-benefits are gradu-
Efficiency and avoiding side effects might be conflicting policy ally decreased and eventually vanished.
goals. Alternative, local policy incentives (parking/bus lane access
and charging) have a value that is higher than the current sub- 5.5. Limitations and further research
sidies in Sweden. For example, free parking has a WTP of around
50,000 SEK or €5400 and access to bus lanes 30,000 SEK (or This study is based on a sample with an over-representation of
€3200). From a cost point of view, a policy packages consisting of electric vehicle users, which are found to be more likely to adopt
local policy incentives might be tempting, because there is no high an electric vehicle in this hypothetical situation. By incorporating
upfront subsidy but rather some lost parking revenues spread out stage-of-change and socio-cognitive determinants in the model-
over some years and maybe some adaptations at signalized in- ling framework, this over-representation has been partially con-
tersections to recognize electric vehicles making use of bus lanes. trolled for. In general, stated choice experiments have the limita-
When considering the effects on EV adoption, these policy mea- tion that they reflect hypothetical decision making, which might
sures are especially effective for people in the pre-contemplation not necessarily reflect real behaviour. However, the possibilities for
and contemplation stages. However, these incentives decrease the revealed preference studies focussing on policy and EV-adoption
marginal cost of EV trips, thereby increasing the number of trips. are likely to be limited to general comparative studies across dif-
For EV users it might make sense to drive, make use of bus lanes ferent countries with different policy packages being offered.
and park for free in the city centre rather than taking public Another important issue is the fact that the respondents have
transport. been recruited within Greater Stockholm. This implies that the
As local policy incentives decrease the marginal cost of EV-use results cannot be generalized to other areas.
rather than their fixed cost, those policy measures are of little help As described above, the value of local, use-based policy in-
for the group of car buyers that can currently not afford to buy an centives depends on the travel context of the traveller. Travellers
EV. On the other hand, they stimulate EV-use by decreasing the in congested areas can gain much from being allowed to make use
absolute marginal cost as well as the marginal cost relative to of bus lanes. Travellers in areas with severe parking problems or
other transport modes. Within the range limitations of EVs, EV high parking tariffs can gain much from the allowance to park for
trips will be more attractive than other transport modes. free or to park at designated parking places. In future studies, re-
spondents from different types of areas (urban areas with different
5.4. Expected tenure of policy incentives level of congestion or less densely populated areas) should be
studied. The current hypothesis is that use-based policy incentives
Certainly, policy incentives being taken are not going to be in that are most effective are probably also having most side effects.
place forever and are meant to be in place as long as the electric The results of this study reflect the Willingness to Pay for policy
vehicle market is not able to function on its own. In case many incentives in a certain context. In this case, the context is Stock-
people switch to EVs, it is not viable to keep on giving access to holm with a relatively restrictive parking policy and a rather high
J.H.M. Langbroek et al. / Energy Policy 94 (2016) 94–103 103

level of congestion. When designing policy incentives, one should 10.1016/j.renene.2012.08.012.


be well aware of this context-dependency. Another issue is the fact Givoni, M., Macmillen, J., Banister, D., Feitelson, E., 2013. From policy measures to
policy packages. Transp. Rev. 33, 1–20. http://dx.doi.org/10.1080/
that battery electric vehicles in most cases are one of several cars 01441647.2012.744779.
in the household. This might imply complex intra-household in- Golob, T.F., Kitamura, R., Bradley, M., Bunch, D.S., 1993. Predicting the market pe-
teractions, in which the use of a certain car for a certain trip can be netration of electric and clean-fuel vehicles. Sci. Total Environ. 134, 371–381.
http://dx.doi.org/10.1016/0048–9697(93)90367-F.
influenced by use-based policy incentives. These mechanisms have Greene, D.L., Patterson, P.D., Singh, M., Li, J., 2005. Feebates, rebates and gas-guzzler
not been covered in this study and might be interesting for future taxes: a study of incentives for increased fuel economy. Energy Policy 33,
research. 757–775. http://dx.doi.org/10.1016/j.enpol.2003.10.003.
Green, E.H., Skerlos, S.J., Winebrake, J.J., 2014. Increasing electric vehicle policy ef-
ficiency and effectiveness by reducing mainstream market bias. Energy Policy
65, 562–566. http://dx.doi.org/10.1016/j.enpol.2013.10.024.
6. Conclusions and policy implications Helveston, J.P., Liu, Y., Feit, E.M., Fuchs, E., Klampfl, E., Michalek, J.J., 2015. Will
subsidies drive electric vehicle adoption? Measuring consumer preferences in
the U.S. and China. Transp. Res. Part A: Policy Pract. 73, 96–112. http://dx.doi.
Policy incentives, meant to increase the attractiveness of elec- org/10.1016/j.tra.2015.01.002.
tric vehicles, seem to be an effective way to obtain an increase in Hensher, D.A., Rose, J.M., Greene, W.H., 2005. Applied Choice Analysis: A Primer.
Cambridge University Press, Cambridge ; New York.
the sale of EVs (Hypothesis 1). People in more advanced stages-of-
Hidrue, M.K., Parsons, G.R., Kempton, W., Gardner, M.P., 2011. Willingness to pay for
change are more inclined to adopt EVs (Hypothesis 2) and are also electric vehicles and their attributes. Resour. Energy Econ. 33, 686–705. http:
less price-sensitive (Hypothesis 3). People with a higher self-effi- //dx.doi.org/10.1016/j.reseneeco.2011.02.002.
Jochem, P., Babrowski, S., Fichtner, W., 2015. Assessing CO2 emissions of electric
cacy and response efficacy are also more inclined to adopt EVs
vehicles in Germany in 2030. Transp. Res. Part A: Policy Pract. 78, 68–83. http:
(Hypothesis 4). However, no connection has been found between //dx.doi.org/10.1016/j.tra.2015.05.007.
socio-economic variables and distance between the home location Koppelman, F.S., Bhat, C., 2006. A self instructing course in mode choice modeling:
of the respondents and the city centre and EV-adoption (Hy- multinomial and nested logit models. Retrieved from https://trid.trb.org/view.
aspx?id=793000.
pothesis 5). Krupa, J.S., Rizzo, D.M., Eppstein, M.J., Brad Lanute, D., Gaalema, D.E., Lakkaraju, K.,
As use-based incentives cost less and are relatively effective Warrender, C.E., 2014. Analysis of a consumer survey on plug-in hybrid electric
compared to subsidies or registration tax rebates, these incentives vehicles. Transp. Res. Part A: Policy Pract. 64, 14–31. http://dx.doi.org/10.1016/j.
tra.2014.02.019.
might be more efficient. However, use-based incentives decrease Lancaster, K.J., 1966. A new approach to consumer theory. J. Polit. Econ., 74.
the marginal cost of driving EVs and might cause rebound effects. Langbroek, J., Franklin, J.P., Susilo, Y.O., 2015a. Electric vehicle users and their travel
At the same time, providing use-based incentives to car users gives patterns in Greater Stockholm. In: Presented at the Transportation Research
Board (TRB) 95th Annual Meeting, Washington.
a certain message favouring car use, be it electric car use, which Langbroek, J., Franklin, J.P., Susilo, Y.O., 2015b. Changing towards electric vehicle use
governments might not intend to send. in Greater Stockholm. In: Presented at the 14th International Conference on
If policy makers manage to move people towards a more ad- Travel Behaviour Research, Windsor.
Lieven, T., 2015. Policy measures to promote electric mobility – a global perspective.
vanced stage-of-change, for example by raising awareness about Transp. Res. Part A: Policy Pract. 82, 78–93. http://dx.doi.org/10.1016/j.
EVs, providing more information about the advantages and dis- tra.2015.09.008.
advantages of these vehicles and assist people to get more insight Lin, Z., Greene, D.L., 2012. Promoting the market for plug-in hybrid and battery
electric vehicles: role of recharge availability. Transp. Res. Rec., 2252.
in their own daily travel patterns, the gap between the intrinsic
Mabit, S.L., Fosgerau, M., 2011. Demand for alternative-fuel vehicles when regis-
motivation of people to purchase EVs and the cost of EVs will be tration taxes are high. Trans. Res. Part D: Transp. Environ. 16, 225–231. http:
likely to decrease. The effects are two-fold: less policy incentives //dx.doi.org/10.1016/j.trd.2010.11.001.
McFadden, D., 1974. Conditional Logit Analysis of Qualitative Choice Behavior.
are needed to convince people to start using EVs, but on the other
Academic Press, pp. 105–142.
hand, financial policy incentives are likely to become less effective Meldalen, S. G., 2015. Endrer EL-bilvedtaket: Nå Er det Bare elbiler MEd passasjerer
and efficient for people in more advanced stages-of-change. A Som får bruke kollektivfeltet på E18 i rushtida. Dagbladet.
more cost-effective strategy might be to put efforts on moving Newbery, D., Strbac, G., 2015. What is needed for battery electric vehicles to be-
come socially cost competitive? Economics of Transportation. doi:10.1016/j.
people towards a more advanced stage-of-change and focussing ecotra.2015.09.002.
on this particular group of people when designing a less extensive Prochaska, J.O., 1991. Assessing how people change. Cancer 67, 805–807.
package of policy-incentives for this specific group. Rezvani, Z., Jansson, J., Bodin, J., 2015. Advances in consumer electric vehicle
adoption research: a review and research agenda. Transp. Res. Part D: Transp.
Environ. 34, 122–136. http://dx.doi.org/10.1016/j.trd.2014.10.010.
Rogers, R.W., 1975. A Protection Motivation Theory of fear appeals and attitude
Acknowledgement change. J. Psychol. 91, 93–114. http://dx.doi.org/10.1080/
00223980.1975.9915803.
Ryan, R.M., Deci, E.L., 2000. Intrinsic and extrinsic motivations: classic definitions
This study was funded by the Swedish Energy Agency (Project and new directions. Contemp. Educ. Psychol. 25, 54–67. http://dx.doi.org/
no. 37054-1) as a part of the “Bruka Elbil” project. 10.1006/ceps.1999.1020.
Sierzchula, W., Bakker, S., Maat, K., van Wee, B., 2014. The influence of financial
incentives and other socio-economic factors on electric vehicle adoption. En-
ergy Policy 68, 183–194. http://dx.doi.org/10.1016/j.enpol.2014.01.043.
References Train, K.E., 2009. Discrete Choice Methods with Simulation. Cambridge University
Press, Cambridge, New York.
Verplanken, B., Wood, W., 2006. Interventions to break and create consumer habits.
Bakker, S., Trip, J.J., 2013. Policy options to support the adoption of electric vehicles J. Public Policy Mark. 25, 90–103. http://dx.doi.org/10.1509/jppm.25.1.90.
in the urban environment. Transp. Res. Part D: Transp. Environ. 25, 18–23. http: Vienneau, D., Perez, L., Schindler, C., Lieb, C., Sommer, H., Probst-Hensch, N., Künzli,
//dx.doi.org/10.1016/j.trd.2013.07.005. N., Röösli, M., 2015. Years of life lost and morbidity cases attributable to
Bhat, C.R., 2003. Simulation estimation of mixed discrete choice models using transportation noise and air pollution: a comparative health risk assessment for
randomized and scrambled Halton sequences. Transp. Res. Part B: Methodol. Switzerland in 2010. Int. J. Hyg. Environ. Health 218, 514–521. http://dx.doi.org/
37, 837–855. http://dx.doi.org/10.1016/S0191-2615(02)00090-5. 10.1016/j.ijheh.2015.05.003.
Bockarjova, M., Steg, L., 2014. Can protection motivation theory predict pro-en- White, W., Lunnan, A., Nybakk, E., Kulisic, B., 2013. The role of governments in
vironmental behavior? explaining the adoption of electric vehicles in the renewable energy: the importance of policy consistency. Biomass Bioenergy 57,
Netherlands. Glob. Environ. Chang. 28, 276–288. http://dx.doi.org/10.1016/j. 97–105. http://dx.doi.org/10.1016/j.biombioe.2012.12.035.
gloenvcha.2014.06.010. Whitehead, J., Franklin, J.P., Washington, S., 2015. Transitioning to energy efficient
Egbue, O., Long, S., 2012. Barriers to widespread adoption of electric vehicles: an vehicles: an analysis of the potential rebound effects and subsequent impact
analysis of consumer attitudes and perceptions. Energy Policy 48, 717–729. upon emissions. Transp. Res. Part A: Policy Pract. 74, 250–267, http://doi.org/
http://dx.doi.org/10.1016/j.enpol.2012.06.009. 10.1016/j.tra.2015.02.016.
Gass, V., Schmidt, J., Schmid, E., 2014. Analysis of alternative policy instruments to Zhang, Y., Yu, Y., Zou, B., 2011. Analyzing public awareness and acceptance of al-
promote electric vehicles in Austria. renewable energy. 61. World Renewable ternative fuel vehicles in China: the case of EV. Energy Policy 39, 7015–7024.
Energy Congress – Sweden, Linköping, Sweden, pp. 96–101. http://dx.doi.org/ http://dx.doi.org/10.1016/j.enpol.2011.07.055.

Vous aimerez peut-être aussi