ARTICLE IN PRESS
Energy Policy 37 (2009) 1083–1094
Contents lists available at ScienceDirect
Energy Policy
journal homepage: www.elsevier.com/locate/enpol
How much do incentives affect car purchase? Agent-based microsimulation
of consumer choice of new cars—Part II: Forecasting effects of feebates based
on energy-efficiency
Peter de Haan , Michel G. Mueller, Roland W. Scholz
ETH Zurich, Institute for Environmental Decisions, Natural and Social Science Interface, Universitaetstr. 22, CHN J73.2, 8092 Zurich, Switzerland
a r t i c l e in fo
abstract
Article history:
Received 7 August 2008
Accepted 3 November 2008
Available online 30 December 2008
In this paper, we simulate the car market in order to forecast the effects of feebate systems based on an
energy-labeling scheme using categories A to G. Very fuel-efficient (A) cars receive a cash incentive,
highly inefficient (G) cars pay additional fees. Consumers have different price elasticities and behavioral
options to react to feebates. They can switch to a smaller sized car, but as energy-efficiency varies
widely within size segments, they can also stick to the preferred size class and choose a more efficient
(smaller) engine. In addition, previously owned cars influence the next car to be chosen. We use an
agent-based microsimulation approach particularly suited to predict environmental and market effects
of feebates. Heteorogenous agents choose from a choice set drawn from a detailed fleet of new cars.
Incentives of h2000 for A-labeled cars induce an additional rated CO2 emission decrease of new car
registrations between 3.4% and 4.3%, with CO2 abatement costs between h6 and h13 per ton, and
otherwise little undesired market disturbance. The risk of rebound effects is estimated to be low. After
adopting the frequencies of consumer segments to a given country, the model presented is applicable to
all European car markets.
& 2008 Elsevier Ltd. All rights reserved.
Keywords:
Energy-efficiency policy
Consumer behavior
Policy analysis
1. Introduction
Exhaust gas after treatment and emission limits in OECD
countries have actually reduced the load of most regulated
exhaust pollutants (nitrogen oxides, hydrocarbons, etc.) or will
do so in the near future (particle emissions from diesel engines).
Still increasing however, is fossil energy demand for transportation (IEA, 2006), and resulting CO2 emissions. So after having
learnt, within the course of one generation, how to burn fuel in a
clean manner, we now face the challenge of using less of it. This
challenge is different in nature, as it will call for both, new
technical solutions, and changes in human behavior, and might
take double the time it took to control regulated pollutants.
Among all efforts to reduce our oil dependency, the reduction
of fuel use for transportation is unique in many respects. Road
transport is the second-largest sector of energy consumption in
OECD countries (IEA, 2006), and is the only sector—together with
air transport—with still growing energy demand. Energy-efficient
cars are available, but market success is often at risk. The
reduction potential is large; along with steady technological
improvements, behavioral changes (smaller engines and smaller
Corresponding author. Tel.: +41 44 632 49 78; fax: +41 44 632 10 29.
E-mail address: dehaan@env.ethz.ch (P. de Haan).
0301-4215/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enpol.2008.11.003
cars) offer large possible savings. DeCicco and Mark (1998) argue
that a large potential for improvement in energy-efficiency
persists in increasing consumer adoption of energy-efficient cars.
This delineates the mechanisms of consumer behavior in general
and of car purchase behavior in particular as decisive factors in
reducing energy consumption. However, as average car weight
and average power still increase in all car markets, it is evident
that a widespread shift of consumers towards energy-efficient
cars is still in its beginnings.
Feebate systems combine rebates awarded to products with
good environmental performance with additional fees for products which have above-average environmental impact. Rebates
might also take the form of cash incentives or tax refunds. Fees
might be either surplus taxes or a separate registration fee that is
billed separately from any other existing type of tax. Feebate
systems offer various advantages compared to measures like fuel
economy standards or fuel taxes (see Greene et al., 2005; Johnson,
2006, 2007), for example their public acceptance, the possibility
to match feebates to specific car types, and hence the potential of
feebates to target car purchase decisions. Feebate systems therefore are widely considered (Greene et al., 2005) and have become
implemented in various European countries in recent years. While
a lack of real-world experience still exists, various studies have
modeled effects of feebates (Jansen and Denis, 1999; Langer,
2005). BenDor and Ford (2006) investigate feebate systems and
ARTICLE IN PRESS
1084
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
analyze possible changes in purchasing behavior regarding
various fuel types (gasoline, alcohol, electricity and compressed
natural gas). Johnson (2007) proposed refunded tax schemes as
possible solution to the dilemma between cap-and-trade instruments on the one hand and emission taxes on the other.
In the following we argue that under feebates consumers have
different options to react in car purchase behavior, but that current
practice simulation tools do not fully capture several of these
options. Consumers may react by changing to smaller cars, by
changing to smaller (i.e., more energy-efficient) engines, by
changing fuel type, or by changing powertrain type. If given both
options, consumers are more likely to change to smaller engines
than to decrease in car size. For example, Peters et al. (2008)
report for a mail-back survey of Swiss car buyers (n ¼ 1247), on a
5-rank scale from 1 ¼ ‘‘do not agree at all’’ to 5 ¼ ‘‘do fully agree’’,
that 35% of respondents indicated they would change to a smaller
car size in exchange for a h1350 feebate (sum of ranks 4 and 5),
compared to 40.1% that would change to a smaller engine, mean
ranks being significantly different (one-sample two-tailed T test,
P ¼ 0.000). This corresponds to a two-stage decision process
when purchasing a new car. Hence in feebate simulation models it
is crucial to account for all different engine combinations within a
given car model and size class.
Most car purchase forecast models operate on a more
aggregated level; they can, following de Jong et al. (2004), be
divided into three categories: Aggregated models (total car stock
of a country depending on GDP), cohort models (same target
variable as aggregated, but single cohorts can be followed), and
disaggregated micro-economic models which operate on the
household decision level. The latter can be classified into static
(describing household car fleet at a given point of time using
discrete choice models) and dynamic approaches (vehicle transaction models, duration models, etc.).
However, at present most car purchase models give forecasts
only for very few vehicle size classes. For example, Berkovec
(1985) has fuel costs as a model parameter to predict car choice,
but uses 13 size classes, and fuel cost is assumed being constant
per size class. The year 1989 Facts model from the Netherlands
Economic Institute, as cited in de Jong et al. (2004), differentiates
between three fuel types (gasoline, diesel, LPG) and three weight
classes. Golob et al. (1997) present a car choice forecasting model
based on a stated preference survey, where fuel cost only differed
per car size class and fuel type, but different engine sizes were not
investigated. Brownstone et al. (2000) operate with 12 car body
sizes and 4 possible fuels, and equally assume that operating fuel
costs only differ between car size classes, but not within. The
model of Hayashi et al. (2001) uses a car choice model predicting
which out of 4 car size classes will be chosen. Mohammadian and
Miller (2003) have six car size classes, and treat fuel costs as
constant for each of these classes. The vehicle type choice model
of Choo and Mokhtarian (2004) distinguishes nine car size classes.
A model using a more detailed vehicle fleet is the DVTM (de
Jong, 1996), where households choose from 20 out of 1000 makemodel-age class combinations. However, this model does not
regard the different engines being available; from an energyoriented point of view, the within-class CO2 variance therefore
still remains large. Mannering et al. (2002) distinguish 175 vehicle
types, but do not account for differences in energy-efficiency for a
given vehicle type. Hocherman et al. (1983) even differentiate
between 950 vehicle types, which are characterized by make, size,
age, and fuel type, but not by engine size or powertrain.
All above mentioned models run at risk not to fully capture the
behavioral option of changing to smaller engines while keeping to
the car model. This behavioral option, however, is likely to be the
most important. In order to capture all behavioral options that car
buyers have under feebate systems, it seems beneficial to use a
decision model using a two-stage decision model, and using a
highly detailed fleet of car models.
The paper is structured as follows. In the following section we
sketch three different routes towards energy-efficient cars, the
EU’s strategy to reduce CO2 emissions from cars, and the resulting
need for fiscal policy tools to influence consumer behavior
when purchasing new cars. In Section 3 we specify the class of
policy tools under investigation, partial and full feebate systems.
Section 4 defines four feebate systems and presents results from
the microsimulation, i.e. consumer reactions, financial volume,
and environmental benefits. We discuss these findings in
Section 5. Section 6 assesses the risk for occurrence of rebound
effects, and Section 7 concludes.
2. Fiscal measures to reduce CO2 emissions from passenger cars
2.1. Three routes to energy-efficient cars
Increasingly stringent tailpipe emission standards in the US,
Europe, and Japan, and the need for higher fuel-efficiency, have
made the development of engines and powertrains for passenger
cars more costly. Accelerated by the globalization of the
automotive industry, powertrains and car platforms are therefore
often used for various makes and models on all major car markets.
In particular, the European car market, historically characterized
by strong home brands, and different legislation per country, can
today be considered as homogeneous, where all makes and
models, and almost all engine configurations, are offered on the
various national markets. On the demand side heterogeneous
preferences and constraints still lead to remarkable differences
between countries regarding the resulting fleet of new registrations. Therefore, all national car markets in Europe can be
analyzed using the same simulation model, if data on consumer
segments is available.
Over the last 30 years, the European car market has seen annual
increases in curb weight and rated power of approximately 20 kg
and 1.5 kW, respectively (see Fig. 1; note that while relative increase
since 1995 for Switzerland is lower than for EU, new car
registrations in Switzerland still show highest average weight and
rated power of all of Europe). The increase in power partly (but not
to full extent) was called for by the surplus in vehicle mass to ensure
constant driving characteristics. We therefore use relative power,
being the quotient of rated power and curb weight, as yardstick for
perceivable engine power. On average for the period 1995–2006,
relative power in Europe increased annually by 1.28%, or 0.4 kW/ton.
While in principle a strong relationship between car weight
and fuel consumption holds, today the variety in engine power
(and energy demand) is very large throughout all car size
segments. With car model we mean the traditional make-model
concept (VW Golf, Chrysler Neon, etc.), whereas a model version is
a unique technological choice with respect to body type and
powertrain (driveline, engine size, fuel type, and transmission
type). An analysis of car models with high sales numbers in
Europe (here we use the 15 best-selling passenger car models in
Switzerland, which account for over 25% of total sales in 2007),
yields an average difference of 94 g CO2/km between the most and
the least efficient version of the same passenger car model (Fig. 2).
Assuming an average total run distance of 1,60,000 km, this
amounts to 15 ton CO2. Therefore, we distinguish three routes to
increase energy-efficiency of cars:
(a) Switch to lower relative power, i.e. to less powerful engines, for
unchanged car size;
(b) Switch to smaller sized cars (with proportional decrease in
rated power so that relative power remains unchanged);
ARTICLE IN PRESS
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
rated power (ACEA)
rated power (Switzerland)
empty weight (ACEA)
empty weight (Switzerland)
relative power (ACEA)
relative power (Switzerland)
1085
relative change (1995 = 100%)
130%
120%
110%
100%
1995
1996
1997
1998
1999
2000
2001
2002
2004
2004
2005
2006
Fig. 1. Temporal evolution of rated power, curb weight, and relative power for the European car market 1995–2007 (1995 ¼ 100%). Data source: Association of European Car
Manufacturers (ACEA).
Renault Clio
Fiat Punto
Peugeot 307
Ford Focus
Audi A3
Toyota Yaris
Audi A4
Opel Zafira
VW Polo
VW Touran
Opel Astra
Skoda Octavia
VW Passat
BMW 3 Series
VW Golf
100
150
200
250
300
CO2 [g/km]
Fig. 2. Range of CO2 emissions for all different engine/gearbox combinations on the market for the 15 most-sold car models in Switzerland in 2007. Also shown are 5th,
50th, and 95th percentile of CO2 emissions (126, 187.4, and 273 g CO2/km, respectively) of new car models on sale 2007 in Switzerland.
a
route
b
c
rated power [kW]
car weight [kg]
relative power [kW/kg]
costs [ ]
Fig. 3. Main effects from three different routes to increase fuel efficiency of new
car registrations (for further explanation see text).
(c) Introduction of new technology to increase energy-efficiency,
i.e., hybrid powertrains, lightweight materials, etc.
route. Most literature focuses on the supply side (Johnson, 2007,
IEA, 2006), as already discussed in Peters et al. (2008). Route (b)
achieves its goal at the cost of decreasing car size, which is, for
itself, not a target. Route (c) uses more advanced technology to
maintain both average car size and average relative power, which
is costly. We argue that route (a) is superior in that no decrease in
car size is called for and that energy-efficient cars in fact are
cheaper. The energy reduction potential of route (a) is substantial,
as illustrated by the wide range of CO2 emissions for given car
models (Fig. 1), and given that the car market in the last 10 years
has shown an increase in average relative power (Fig. 2). We
further discuss behavioral implications of route (a) in Section 2.3.
2.2. Current EU strategy
As illustrated in Fig. 3, each route finally increases energyefficiency of new car registrations. While (a) and (b) are demandsided, behavioral routes, (c) is the supply-sided technological
The strategy of the European Union to reduce CO2 emissions
from cars (EU, 1995) consists of three pillars, the first one being
ARTICLE IN PRESS
1086
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
agreements with manufacturers to decrease average CO2 emissions of new car registrations to 140 g/km in 2008 (European car
makers) or 2009 (Korean and Japanese car makers). The strategy’s
second pillar is consumer information (including booklets
reporting fuel consumption of all car models in the market, and
compulsory information posters at the point of sale), with the
most important part being compulsory energy-labels for all new
cars being on display for sale. These labels must contain fuel
consumption and CO2 emissions; on a voluntary basis member
states may also prescribe additional information where the car in
question is rated or ranked in a certain category (see Section 3.1).
The strategy’s third pillar consists of fiscal measures that aim at
influencing the behavior of consumers when purchasing new cars.
The mid-term target of the European Commission is 120 g CO2/km.
As the observed progress slowed down from over 1% CO2
reduction annually (from 1995 to 2003) to 0.6% annually (since
2004), this target would not be reached before 2020. Therefore the
European Commission wants to introduce additional, fiscal
measures (EU, 2007). The present paper investigates one widely
debated type of measure that complies with the EU strategy’s
third pillar.
smaller cars reported higher willingness for behavioral changes.
This is investigated further in Peters et al. (2008).
2.5. Effects of incentive schemes
In this paper we distinguish and discuss the following impacts
of incentive schemes:
(a) direct change of demand due to monetary effects (price
elasticity), i.e. higher demand for those fuel-efficient cars that
are eligible for incentives; and
(b) indirect change of demand as incentive schemes also have a
normative impact, causing changes in the norms and attitudes
of consumers.
There is a variety of possible actions to reduce energy demand,
conditional to given sets of criteria. Industrialized countries may
either reduce domestic energy demand, or pay for reduction
efforts in other countries. Energy demand reductions might either
be sought within the transportation sector or in other economic
sectors. The transportation sector comprises of individual motorized mobility, commercial traffic, public transport, and transport
of goods. We defined three routes to reduce energy demand from
individual motorized mobility in Section 2.1. Switching to
alternative fuels would be another option regarding CO2 emissions, but here we focus on lowering energy demand.
We limit ourselves to revenue-neutral feebate systems to
decrease domestic energy demand from individual motorized
mobility. The income side of such incentive schemes covers the
administrative and transfer costs, and cash incentives. Other
policy tools (technical standards, higher taxes, voluntary agreements with manufacturers, information campaigns, etc.) are not
the scope of the present paper.
Feebate systems are a suitable policy tool for route (a)-type
reductions in energy demand. Buyers of cars with high energyefficiency would be eligible to a cash incentive. The efficacy of any
such feebate system depends on the elasticity, i.e. how consumers
change their purchase behavior for a given financial incentive.
In addition, long-term impacts on the supply side occur as
manufacturers adapt their research and production allocation and
ultimately bring new vehicle technology to the market. This
aspect is dealt with using macro-economic approaches and in
innovation science. Using only the economic concept of price
elasticity may be inadequate to predict consumer reactions, since
the normative character of punishments and incentives is
neglected (Langer, 2005). Our modeling approach accounts for
this in part by incorporating psychological effects which may
increase the effect of an incentive. We do not model behavioral
changes due to changes in norms or attitudes. To the authors’
knowledge there is no literature on quantitative modeling of (b),
except from Steg et al. (2001), who use a regression model
approach to assess the potential of motivational factors in the
reduction of car use.
Apart from the monetary effect, feebates based on energylabeling might take effect simply because of the information
provided. The availability of thousands of different model versions
at present might keep many consumers from informing themselves about recent advances in energy-efficiency. Feebates could
draw more attention to the underlying energy-label, and allow
consumers to choose cars that better meet their preferences. In
addition, growing awareness for energy issues might cause part of
new car buyers to change their attitudes towards more efficient
vehicles. Most visible at present is the success of hybrid cars,
which are being bought by consumers that previously owned
larger cars (de Haan et al., 2006, 2007). This is in line with the
framework from Coad et al. (submitted) that the information part
of feebates, i.e. the energy-label for cars, may be effective in
encouraging intrinsically motivated consumers to adopt green
cars, while the financial part may be more persuasive for
extrinsically motivated consumers.
2.4. Acceptance of feebate sytems
2.6. Possible benefits and drawbacks
We conducted a mail-back survey among Swiss households on
car purchase behavior, results of which are reported in Peters et al.
(2008). A sample of 3920 households was drawn from the phone
book, with a response rate of 40.3% (n ¼ 1581). Feebate systems
yielded acceptance rates equal to those for pure informational
measures (energy-labeling), being the most accepted policy
measure tested for. Acceptance on a scale from 1 (‘‘not reasonable
at all’’) to 5 (‘‘very reasonable’’) was 3.40 (SD ¼ 1.5), significantly
higher than for the purchase of CO2 emission certificates (2.46;
SD ¼ 1.4; n ¼ 1424; P ¼ 0.000 for two-sided one-sample Student’s
t-test) or higher fuel prices (2.08, SD ¼ 1.36, n ¼ 1445; P ¼ 0.000).
Willingness of respondents to change to smaller engines was
significantly higher than willingness to change to a smaller car,
which is in support for route (a). Larger households, younger
people, lower-income households and households preferring
Incentive schemes have various impacts on the new car market
and, intentionally, on the environmental load of new car
registrations. Along the advantages is the internalization of
external costs of road transport, which in principle leads to
economic gains according to the external cost concept. Other
advantages are the promotion of fuel-efficient vehicles (the direct,
intended effect) and of environment-friendly individual behavior.
The latter may also influence consumer behavior in other energyrelevant consumption field, through increased perception of
energy and climate issues (the indirect effect) in general
(Wüstenhagen et al., 2007). However, incentive schemes in
principle also have disadvantages. There is more governmental
regulation with corresponding administrative costs, and every
governmental intervention on the market economy runs at risk of
lowering the efficiency of the market segment in question.
2.3. Feebate sytems as energy policy tool
ARTICLE IN PRESS
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
Moreover, market actors experience adaptation costs, and any
regulation is exposed to the risk of not being able to adapt fast
enough to technological changes (Nilsson, 2007). Another potential disadvantage is information overload of consumers, if other
incentive schemes are active at the same time.
3. Definition of feebates systems
3.1. Energy-labeling of cars
Energy-labeling is an information tool, part of the Community
Strategy’s second pillar (Section 2.2). The label is similar in
appearance to the one for household appliances in Europe. Several
European countries use this type of label for passenger cars
(United Kingdom, Belgium, Denmark, Netherlands, France, Spain)
or consider to do so (Portugal, Germany). It shows seven arrowshaped bars labeled from A (best fuel-efficiency) to G (lowest fuelefficiency), color-coded from green (A) to red (G). Since 2003 this
label is also used in Switzerland.
Several countries now also the energy-label as basis for incentive
schemes (United Kingdom, Netherlands, Portugal) or introduced
feebates based on energy demand or CO2 emissions (Belgium,
Austria, Denmark, Sweden, Italy, France, Luxembourg, Cyprus).
The case study for which we investigate the effects of feebate
systems (Section 4) corresponds to a proposed legislation for
Switzerland. Due to the market’s homogeneity on the supply side,
most aspects of this case study will apply to all national car
markets in Europe.
3.2. Relative energy-efficiency vs. absolute energy consumption
While there is a certain degree of uniformity on the side of the
appearance of the labeling system, every country has its own basis
on which classification into categories A to G takes place. Under
the ‘‘absolute’’ notion of energy-efficiency, only the absolute level
of rated CO2 emissions of the car in question determines its
energy-efficiency label (e.g., United Kingdom, Belgium, Denmark,
France, each using different bounds between classes). Alternatively, a ‘‘relative’’ energy-efficiency may be computed using the
ratio of rated CO2 emissions (or fuel consumption in mass units, as
we do not regard alternative low-carbon fuels) to car size. Car size
may be operationalized by vehicle floorspace (e.g., Netherlands
and Spain) or by curb weight (e.g., Switzerland). This is in line
with Johnson (2007), who argues that political acceptability of
feebates might be affected by distributional effects, and suggests
distributing incentives in proportion to vehicle mass. In the case
of Switzerland energy-efficiency, ee, is defined as
ee ¼
FCm
m0 þ ma
where FCm is fuel consumption in mass units (for gasoline we
assume an average density of 745 kg m3, for diesel 829 kg m3),
m0 ¼ 600 kg is a constant in mass units, m is curb weight, and
a ¼ 0.9. The m0 constant is introduced to compensate for the fact
that small engines cannot reach the same thermodynamic
efficiency as large engines.
After choosing a policy base (absolute energy or relative
energy-efficiency), also the size of each of the categories of the
energy-label has to be defined. In Switzerland, by definition oneseventh of all car model versions being on sale is categorized as A.
In total six ee values define boundaries between categories A to G.
These boundaries are periodically adjusted in order to account for
technological progress, and to ensure that always one-seventh of
all model versions has an A label. We used the Swiss energy-label
boundaries for the period 2006–2008.
1087
3.3. Other requirements
Feebate systems can be designed in a variety of ways, general
aims being effectiveness and equity at the lowest possible
transaction and implementation costs. Careful consideration
should be given to simplicity and fairness in program design,
compatibility and coordination with other local vehicle technologyrelated and tax-related programs, the likelihood of ‘‘leakage’’ and,
most importantly, the potential for realizing significant emissions
benefits. In most cases (DeCicco et al., 1993; BenDor and Ford,
2006), feebate systems are revenue-neutral, i.e., the fees should
amount to the sum of implementation costs and total volume of
rebates.
Apart from only targeting energy(-efficiency), feebates’ policy
base can also include other greenhouse gases [GHG] expressed in
CO2 equivalents, or emissions of criteria atmospheric pollutants
(CAP) like hydrocarbons, oxides of nitrogen, and carbon monoxide
(de Haan and Keller, 2000). In the case of the latter, weights have
to be given to CAPs and GHGs (this is non-trivial, as there
are inherent trade-offs). This makes the calculation slightly
more complicated, but also rewards consumers of otherwise
‘‘clean’’ cars.
4. Microsimulation of full and partial feebate systems
4.1. Simulation model
The car market microsimulation model described by Mueller
and de Haan (2008) is employed. Following a two-stage car
purchase decision process, in a first step out of several thousand
make-model-engine combinations a choice set is selected, from
which in a second step the car to be bought is chosen using a
discrete choice model. The latter model is taken from literature
and has originally been developed for European Commission’s
Directorate General for Environment (COWI, 2002). For a detailed
description refer to Mueller and de Haan (2008).
We simulate 1 million new car sales, from which aggregated
statistics of the fleet of new passenger car registrations are
computed. In order to apply the model to a European country
other than Switzerland, only the frequencies of consumer
segments have to be adapted, followed by a model validation.
The model is static in the sense that car choice parameters,
demographic data of the population, and the car fleet being on
sale do not change. For the simulations in this paper, the
commercial consumer group from COWI (2002), that was not
regarded in Mueller and de Haan (2008), is included. Retention
rates for fuel type, car size class and gear type do not apply to
commercial agents, as we do not have market observations from
which retention rates for commercial agents could be retrieved.
Following the model validation presented in Mueller and de
Haan (2008), we calibrated the model. This was done to facilitate
communication with stakeholders as the calibrated model closely
reproduces 2005 market data that is well known to them. The
validated model shows slight underestimations of market shares
of inefficient vehicles, and overestimations for efficient vehicles,
especially for diesel cars. We interpret this as a persisting bias of
consumers against the smallest diesel engines being available. For
calibration purposes, weighting coefficients for trunk space were
modified. Trunk space as a car utility parameter has become illdefined with new body types like sport-utility vehicles, family
vans, and cross-overs. Moreover, trunk space is not recorded in the
official type approval data base and has to be parameterized,
which is imprecise. Fig. 4 depicts results for the calibrated vs. the
original simulation model. Using the statistical performance
metrics (for definition see Mueller and de Haan, 2008) normalized
ARTICLE IN PRESS
1088
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
market share
25%
0%
CO2 emission categories {<100, <120, <140, <160,
<180, <200, <220, <250, <300, ≥300 g/km }
20%
market share
0%
curb weight cat. {<1000, <1100, <1200, <1300,
<1400, <1500, <1600, <1700, <1800, ≥1800 kg }
market share
25%
0%
rated power categories {<60, <80, <100, <120,
<140, <160, <180, ≥180 kW }
amount paid out depends on the simulated total sales volume
for ‘‘A’’-labeled cars and thus has to be determined based on
iterative simulation, ensuring revenue-neutrality. A general
purchase tax already is in force, its increase does not lead to
additional implementation costs. Administrative costs comprise of transaction costs (estimated at h10 per cash incentive)
and fixed costs of 0.8 mio. h annually.
Partial feebate system with absolute policy base: same as above,
but here cars are categorized based on (absolute) energy
consumption. The change from relative to absolute policy base
means that some heavier car models leave category A and
some smaller car models with only moderate fuel-efficiency
are admitted. The changed composition of category A leads to
changes in its sales volume. By iteration, revenue-neutrality
therefore leads to a slightly different cash incentive amount.
Full feebate system with relative policy base: those 15.3% of new
registrations having lowest fuel-efficiency pay a fee of h2000
each (the amount of h2000 is fixed here; revenue-neutrality is
enforced by slight changes to the size of the A and G
categories). This allows for cash incentives to those 14.7% of
new registrations having highest fuel-efficiency, and to cover
transaction costs of h10 per incentive and h20 per fee
(accounting for debit losses).
Full feebate system with absolute policy base: same as above, but
again cars are categorized based on their (absolute) energy
consumption. Again, the changed composition of category A
(and G) makes slight modifications to A and G boundaries
necessary in order to achieve revenue-neutrality.
The model employed is steady-state, i.e. we regard the relative
difference between reference run and a policy run. Our main
target parameter is the reduction in rated CO2 emission of the
fleet of new car registrations, expressed as percentage of the rated
CO2 emission for the reference run. To quantify this relative
reduction into absolute terms for Switzerland (2005 population:
7.45 mio.), we also report total CO2 reduction, assuming that
2,60,000 new cars are sold per year and that on average each car
runs 1,60,000 km. That is, we match an incentive scheme with
implementation duration of 12 months to the total life-time effect
of the cohort of cars newly registered during those 12 months.
Fig. 4. Comparison of validated and calibrated simulation model. For further
explanations see text.
4.3. Results for partial feebate systems
mean absolute error, NMAE, mean fractional bias, MFB, and
correlation coefficient, COR, calibration improves MFB for forecasted CO2 (from 18.06% to 2.46%), curb weight (1.29% to
0.56%), and rated power (1.06% to +1.15%), whereas no
significant changes for NMAE and COR occur.
4.2. Definition of reference run and policy runs
In the following we compare the reference run to the outcomes
under four feebate systems (see Table 1 for details), all having the
same ‘‘carrot’’ component (cash incentives), but different ‘‘stick’’
components, and different policy base:
Reference run (no incentive scheme): simulation of the Swiss
car market for the year 2005.
Partial feebate system with relative policy base: purchase tax is
increased by 3% of list prices. This on average generates approx.
h770 per new registration (annual total for Switzerland being
200 mio; using a CHF/h exchange rate of 1.5). This tax revenue
(‘‘stick’’) is refunded by means of cash incentives to A-labeled
cars based on their relative energy-efficiency. The exact cash
We employ two indicators. Policy efficacy is the relative
reduction in rated CO2 emissions, efficiency is administrative costs
per abated ton CO2. In addition, market disturbance is assessed by
relating efficacy to changes in the curb weight and relative power
distributions of new registrations. Table 2 lists selected market
statistics and environmental effects. An inherent advantage to the
microsimulation approach is that any other statistic could also be
computed out of the total of simulated car purchase decisions. The
model also simulates effects on emissions of regulated pollutants
like nitrogen oxides and particulate matter (Table 2), however
their importance is in decline and will show further decline in the
future (de Haan, 2009).
Fig. 5 shows histograms of market parameters for the reference
and both partial feebates runs. Iterative simulation yielded cash
incentives of h2550 and h2670 per A-labeled car based on relative
energy-efficiency and absolute energy consumption, respectively.
The partial feebate systems lead to reduced CO2 emissions of new
registrations of 3.1–3.3%. Most notable is a shift towards car
models fitted with the smallest available diesel engine. The
market share of ‘‘A’’-labeled cars increases, as expected, compensated by only minor market share reductions for categories B to G.
The share of diesel cars increases. Due to low implementation
ARTICLE IN PRESS
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
1089
Table 1
Definition and aggregated financial figures of the simulated runs.
Reference
run
Partial feebate system
Relative base
Full feebate system
Absolute base
Relative base
Absolute base
Stick component
Purchase tax 3% for all categories
h 2000 tax surplus for ‘‘A’’ cars (15.3% of sales)
Carrot component
h2550 incentive for
‘‘A’’ cars
h2670 incentive for
‘‘A’’ cars
h 2000 incentive for ‘‘A’’ cars (14.7% of sales)
Policy base
Relative (fuel
consump./curb
weight)
Absolute (fuel
consump.)
Relative (fuel
consump./curb
weight)
Absolute (fuel
consump.)
Cash incentive volume (106h/a)
Fee volume (106h/a)
Purchase tax increase (106h/a)
Administration costs (106h/a)
0.0
0.0
0.0
0.0
195.5
0.0
196.8
1.5
190.8
0.0
192.2
1.5
78.0
82.2
0.0
4.2
79.2
83.5
0.0
4.2
Reduction CO2 emissions new reg. (%)
CO2 effect (1000 ton) per cohort (total)a
Costs per abated ton CO2(h/ton CO2)
0.0
0.0
2.8
221.9
6.9
3.3
260.7
5.9
3.9
304.1
13.7
4.3
337.5
12.5
a
Effect over vehicle life (1,60,000 km) of 2,60,000 new cars ( ¼ policy in force during 12 months).
Table 2
Environmental and market effects of the simulated runs, relative to reference run.
Market statistic
Reference run
Relative changes induced by feebate systems
Partial feebate system
Full feebate system
Relative base (%)
Absolute base (%)
Relative base (%)
Absolute base (%)
Share of diesel passenger cars (%)
Microcars market share (%)
Subcompact (%)
Compact (%)
Mid size (%)
Full size (%)
Luxury (%)
Minivan (%)
Family van (%)
4WD w/o SUV (%)
SUV (%)
Cabriolet (%)
Sports (%)
Curb weight average (kg)
Engine capacity average (ccm)
Fuel consumption average (L/100 km)
Fuel consumption (gasoline only) (L/100 km)
Fuel consumption (diesel only) (L/100 km)
CO2 emissions average (g/km)
NOx emissions average (g/km)
Particle emissions average (mg/km)
Noise level average (dB(A))
28.53
3.4
18.0
22.0
21.2
8.5
0.9
8.1
5.3
5.9
3.9
1.0
1.5
1467
1983
7.69
8.10
6.65
188.5
0.104
0.61
72.1
5.2
+1.0
+1.1
+0.6
0.6
0.2
+0.0
+0.0
0.5
0.7
0.4
0.1
0.1
0.8
1.6
3.1
1.1
4.7
2.8
+8.5
+16.5
0.0
2.0
+0.8
+1.5
+0.5
0.8
0.3
0.0
+0.1
0.4
0.7
0.4
0.1
0.1
2.1
3.2
3.3
2.6
4.1
3.3
+2.4
+5.8
0.0
4.7
3.4
18.0
22.0
21.2
8.5
0.9
8.1
5.3
5.9
3.9
1.0
1.5
1.3
3.3
4.1
3.3
3.5
3.9
+8.0
+15.3
0.0
2.0
3.4
18.0
22.0
21.2
8.5
0.9
8.1
5.3
5.9
3.9
1.0
1.5
2.7
4.6
4.2
3.8
4.6
4.3
+1.0
+6.3
0.1
Sales price
oh13,500 (%)
oh20,000 (%)
oh26,500 (%)
oh33,500 (%)
Xh33,500 (%)
13.3
27.9
24.5
13.3
21.1
+1.3
+0.6
0.3
0.5
1.1
+3.8
+1.2
2.0
1.4
1.6
+0.5
+2.8
0.1
0.8
2.4
+2.3
+3.0
0.5
1.1
3.6
costs, these incentive schemes have abatement costs of h6 per ton
CO2. Differences between relative and absolute policy bases are
present but not dominant. The absolute basis allows for slightly
higher CO2 reductions and hence offers lower CO2 abatement costs.
As can be seen from Fig. 5, changes in curb weight distribution are
more pronounced for the absolute policy base, while changes in
relative power do not show major differences between relative and
absolute policy bases. Hence the absolute feebate system reaches its
environmental effects at the cost of a higher market disturbance.
Other environmental effects (emissions of particles and nitrogen
oxides, and noise level) exhibit only very small changes, which are
fully due to the higher share of diesel cars.
ARTICLE IN PRESS
1090
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
20%
15%
10%
5%
0%
< 1000 < 1100 < 1200 < 1300 < 1400 < 1500 < 1600 < 1700 < 1800 >= 1800
curb weight classes [in kg]
20%
15%
10%
5%
0%
< 100
< 120
< 140
< 160
< 180
< 200
CO2 emission [g/km]
< 220
< 240
< 300
30%
25%
20%
15%
10%
5%
0%
<40
40-50
50-60
60-70
70-80
80-90
90-100 100-110 110-120
>120
relative power classes [kW/t]
reference run
partial feebates/rel. policy base
partial feebates/abs. policy base
Fig. 5. Distribution of new car market along curb weight classes (top panel), engine capacity classes (middle), and rated CO2 emission classes (bottom panel), for the
reference run (Swiss car market for the year 2005) and two partial feebate system runs.
4.4. Results for full feebate systems
Fig. 6 shows histograms of market parameters for the reference
run and both full feebates runs. Aggregated market parameters
are listed in Table 2 (two right-most columns). In contrast to the
partial feebate approach, the full feebate system clearly punishes
‘‘G’’-labeled ‘‘gas-guzzlers’’. This causes a clear drop in market
share for the ‘‘G’’ category. Being more selective in its steering
effect both when charging fees and when awarding rebates
allows for a higher efficacy: average CO2 emissions of new car
ARTICLE IN PRESS
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
1091
20%
15%
10%
5%
0%
< 1000 < 1100 < 1200 < 1300 < 1400 < 1500 < 1600 < 1700 < 1800 >= 1800
curb weight classes
20%
15%
10%
5%
0%
< 100
< 120
< 140
40-50
50-60
< 160
< 180
< 200
< 220
CO2emission [g per km]
< 240
< 300
30%
25%
20%
15%
10%
5%
0%
< 40
60-70
70-80
80-90
90-100 100-110 110-120
>120
relative power classes [kW/t]
reference run
full feebates/rel. policy base
full feebates/abs. policy base
Fig. 6. Same as Fig. 5, but for two full feebate system runs.
registrations drop by 3.9% and 4.3% on a relative or on a absolute
policy basis, respectively. Note that this higher efficacy is also in part
due to the psychological effects incorporated into the simulation
model (fees are perceived stronger than rebates, see Mueller and de
Haan (2008)). Due to higher implementation costs, however, the
abatement cost, at h13.6–12.3 per ton CO2, while still low, is inferior
to the one for partial feebate approaches. As in the case of partial
feebates, the absolute policy basis appears to be superior in the case
of full feebate systems as well. The resulting CO2 reduction effect
differs by 10%. Again, the absolute policy base achieves its CO2
reductions at the cost of a higher market disturbance (more
pronounced changes in curb weight distribution, see Fig. 6).
ARTICLE IN PRESS
1092
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
5. Discussion
To assess the simulated environmental and market effects of
feebates, we use policy efficacy (amount of abated CO2) and
efficiency (costs per abated ton CO2), and analyze changes in curb
weight and relative power distributions to operationalize market
disturbance. There are two main findings:
We observe high policy efficacy. Cash incentives seem to be able
to outweigh utility losses, at least for part of the consumer
groups. This is due to the fact that staying with the same car
model, while switching to a less powerful engine, causes utility
loss only for the acceleration time parameter, but utility gains
in sales price, fuel costs, plus the utility of receiving a cash
incentive.
We observe low market disturbance: In order to become eligible
for a cash incentive, consumers are not forced to switch to a
smaller car. This can been seen from Figs. 5 and 6, where the
relative power histogram reveals more policy-induced changes
compare to the curb weight histogram. Fig. 7 illustrates this in
more detail for the example of the full feebate run with relative
policy base: Gains in market share occur for market segments
with lower relative power. If within a car size class enertyefficient engines are available, total market share remains
unaffected, but within-class shifts towards lower relative
power occur. Decreases in market share do occur for those
car size classes for which no engine with A-rated energyefficiency is available (sport/luxury cars, SUV).
For the simulation run from our Swiss case study, a high policy
efficiency results. As a flat rate purchase tax already exists,
partial feebate systems have no administration costs for the
‘‘stick’’ component. A full feebate system means introducing a
new tax, which is costly. But full feebate systems have the
advantage that both the fee and the rebate have a steering
effect, where the ‘‘stick’’ component of partial feebates does
not differ between efficient and inefficient cars. Overall, partial
feebates are more efficient, but full feebate systems are more
effective.
magnitude of 20–25% (Spielmann and de Haan, 2008). As a firstorder assumption both inefficient and efficient cars have the same
amount of CO2 emissions from these pre- and post-operations life
cycle phases. In relation to the total lifecycle CO2 emission of cars,
therefore, the reduction effect will be lower than reported here. As
GHG emissions have important externalities, the overall economic
effect of incentive schemes will be positive.
As concluded by Peters et al. (2008), the comparison of
efficiency for absolute and relative policy bases calls for detailed
microsimulation of car markets, distinguishing different car buyer
groups and using a highly detailed fleet of new cars being on offer.
The simulation model employed here fulfills these requirements
and is hence capable of analyzing differences in changes of
consumer behavior under relative and absolute incentive policies.
We simulated both partial and full feebate systems on a relative as
well as on an absolute policy basis. While having same administrative costs, absolute systems have a 5–10% higher effect in CO2
reduction and hence a 5–10% better policy efficiency. While this
effect is clear within the framework of our simulation model, it
should be called in mind that acceptance and transparency of
policy measures are crucial to their success, as also discussed in
Peters et al. (2008).
With regard to reliability and validity of our modeling
approach, the validation of the model with 2005 Swiss market
data on an array of different market statistics (for details, see
Mueller and de Haan, 2008) ensures a high level of robustness of
the model results.
As with all model forecasts, we use past consumer behavior to
predict behavorial changes under a future policy instrument, and
hence assume that the norms, attitudes, preferences and decision
making strategies of new car purchases will not change. As we
only investigated incentive schemes that can be implemented
immediately, and that are already in force in several countries,
and since the financial incentives typically do not exceed 10% of
car sales prices, we believe to be well within the area of model
applicability.
6. Assessment of potential occurrence of rebound effects
The method of agent-based microsimulation that allows us to
employ a highly disaggregated fleet of over 2000 car versions on
sale is a prerequisite to account for within-model behavioral
changes. These changes are the driver of the two above findings.
Our simulation results should be regarded as minimum effect;
indirect effects as sketched in Section 3 are likely to give rise to
even higher effects. On the other hand, the manufacturing and
scrapping (Giannouli et al., 2007) of cars also causes significant
CO2 emissions within a lifecycle perspective in the order of
Car Size Class
Sports Car/Convertible
Luxury + SUV
Fullsize
Mid-Size + Van
Minivan
Compact
Subcompact
Micro
sum
sum
-0.7%
-1.8%
-0.2%
-0.6%
+0.4%
+0.7%
+1.5%
+0.6%
Rebound effects are defined as increases in demand induced by
efficiency gains (Saunders, 2000). Here we only address potential
direct rebound effects (more vehicles being purchased; increase in
average car size; more miles being driven) and neglect indirect
rebound effects (increased consumption of other goods). Hybrid
cars and other technological means to increase fuel-efficiency,
corresponding to route (c) from Section 2.1, are characterized by
higher investment costs that roughly outweigh the reduced fuel
Changes in market share of car size/relative power segments for relative full feebate run
+0.2%
+2.2%
+2.1%
<40
40-50
50-60
+0.1%
-0.7%
-0.8%
-0.8%
60-70
70-80
80-90
90-100
relative power categories [kW/t]
-1.1%
-0.7%
100-110 110-120
-0.6%
>120
Fig. 7. Changes in market share of market segments defined by car size class and relative power class. The sum row and sum column give accumulated change in market
share for the respective relative power class and car size class. Single, double and triple plus signs indicates market share changes of that specific cell of 0.1–0.2%, 0.2–0.3%,
or 40.3%, etc. Reading example: Compared to reference run, full feebate system with relative policy basis leads to 1.8% less luxury vehicles+SUV sales, and to +2.23% market
share of model versions with 40–50 kW/ton relative power. Total market share of sub-compact cars with 40–50 kW/ton relative power increased by more than 0.3%.
ARTICLE IN PRESS
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
bill (de Haan et al., 2006, 2007), so overall, rebound effects are not
likely to occur. However, feebates, that correspond to route (a),
induce a switch to more efficient, and less powerful, engines.
Consumers will thus have a reduced fuel bill, without needing to
invest more. Moreover, while on the one hand an engine with
higher energy-efficiency is more expensive to build, on the other
hand reduced rated power means lower sales prices in current
car markets. In combination, reductions in CO2 of 3.1% and
3.9% (Table 1) are associated with 2.0% and 3.0% reductions
(Table 2) in average sales price, respectively. So overall, both
purchase price and fuel bill will decrease slightly. If rebound
effects would occur, our assumption that total car sales are not
affected would not hold. We argue that revenue-neutral feebate
systems will not induce significant rebound effects for the
following reasons:
(1) Incentives of h2000 roughly correspond to the price difference
between the cheapest gasoline engine and the most energyefficient diesel engine for car models with high sales volume.
So the incentive covers additional investment costs for the
most efficient engine.
(2) Due to revenue-neutrality, average car price does not change.
Of course, as second-order effect, buyers of smaller cars have
higher price elasticity (for the 40 consumer groups from the
COWI (2002) study used in our simulation model, income
group and weighing coefficient for sales price show negative
correlation), so an asymmetric market reaction might occur
where increased sales of promoted cars are not compensated
in full by decreased sales of the cars being punished. However,
truly price-sensitive consumers can be expected to buy
second-hand cars.
(3) If (smaller, low-margin) efficient cars sell more, on the long
run car sellers will adjust their profit margins in order to
maintain profitability. This would increase prices of A-labeled
cars. From a long-term perspective it is desirable, and to be
expected, that the car market will learn to earn money with
efficient cars rather than powerful cars. Johnson (2007) also
argues that one of the main benefits of feebate systems would
be the creation of a stable environment towards cars with
higher efficiency.
(4) To large parts, the demand for car travel in industrialized
countries is not cost-constrained, but time-constrained (Schafer and Victor, 2000). Hence savings in time (e.g. through new
highways or high-speed trains) potentially have much higher
rebound potential (Spielmann et al., 2008), while changes in
car price have not. This is supported by total fuel demand in
Europe that showed almost no short-term elasticity with
regard to pronounced fuel price increases since the year 2005.
This is different in North America, where the majority of
households has more than one car and can shift to the more
efficient vehicle as short-term reaction to fuel price increases.
7. Conclusions
In order to reduce the still growing greenhouse gas emissions
from individual motorized transport, policy instruments to
influence car purchase behavior are among the most discussed.
Such policies have been implemented in several countries already.
We presented results from an agent-based microsimulation
representative for European car markets for partial and full
feebate systems. The benefit of employing this model is threefold.
First, the method of microsimulation proved to be very
successful for communication with stakeholders (Creedy, 2001).
Since we use a highly detailed fleet of make-model-engine
configurations, there is no need to use disputable car classes or
1093
‘‘average’’ vehicles. This reduces potential misunderstandings, and
has proven to be beneficial for the level of trust that the
stakeholders had towards to simulation model. Microsimulation
also enables to represent the full complexity (Creedy, 2001) of the
policy to be investigated. In stakeholder interaction, this allowed
us to implement any thinkable classification criterion (‘‘incentives
should only be paid out to diesel cars having particle filter
systems’’, ‘‘sport-utility vehicles should be banned from efficiency
category A altogether’’, etc.) and to show simulated effects almost
immediately.
Second, in real-world policy evaluation often the indicators
that will be used to compare alternative policies are not yet
known at the time of model specification. While the environmental effect in total CO2 reduction and the financial effect in
administrative costs were known, the two relative indicators for
policy efficiency (taken as costs per reduced ton of CO2) and for
market impact (taken as relative power) where put upfront only
towards the end of the research project. At any time, microsimulation enables us to compute additional indicators out of the
simulated microdata. We illustrated this strength by reporting
various indicators on market and environmental impacts.
Third, microsimulation allowed for the investigation of
detailed effects and therefore to distinguish between similar but
yet different tax base definitions. We simulated both feebate
approaches based on relative energy-efficiency and based on
(absolute) energy consumption.
Regarding the simulated results for our Swiss case study, all
feebate systems are suited to obtain substantial reductions in
energy consumption and CO2 emissions without significant
market disturbance, i.e., the statistical distribution of curb weight
(being a proxy for car size) remains merely unaffected, whereas
relative power shows more pronounced changes (Table 2). This
illustrates that car purchasers for the bigger part stay with their
preferred car size class, and change their purchase behavior by
switching to a more energy-efficient engine (Fig. 7). We expected
this based on survey results (Section 2.3). Only with a high level of
detail in car model versions, and by using different consumer
groups, it is possible to reveal such ‘‘within-car model’’ changes in
purchase behavior. A coarser vehicle fleet would have lead to an
underestimation of the overall policy effect.
On the difference between absolute and relative policy bases,
simulated performance is slightly better for feebates based on
absolute energy consumption. However, we consider the difference to relative feebates as minor in comparison to the
importance of public acceptance. Relative policy bases might
reduce distributional effects and increase political acceptability
(Johnson, 2007). We therefore conclude that as results from
microsimulation do not yield a clear picture on this question,
policy makers should be advised to adopt either a relative or a
positive policy base, depending on which will ensure higher
public acceptance.
Feebates for energy-efficient cars have a high public acceptance (Section 2.3), possible reasons being its revenue-neutrality,
low CO2 abatement costs, and the fact that CO2 reductions are
achieved without pronounced market disturbance. CO2 reductions
caused by direct monetary effects (Section 2.4) will decrease as
soon as most car purchasers have been subject to the scheme for
the first time. CO2 reductions due to indirect effects (changes in
consumers’ norms and attitudes) will then gain in importance.
Acknowledgements
We acknowledge project sponsorship by auto-schweiz (autosuisse), Erdöl-Vereinigung (Union Pétrolière) and the Swiss
Federal Office of Energy. The second author is funded by the
Swiss National Science Foundation (Grant 105212-105573).
ARTICLE IN PRESS
1094
P. de Haan et al. / Energy Policy 37 (2009) 1083–1094
References
BenDor, T., Ford, A., 2006. Simulating a combination of feebates and scrappage
incentives to reduce automobile emissions. Energy 31, 1197–1214.
Berkovec, J., 1985. Forecasting automobile demand using disaggregate choice
models. Transportation Research B 19, 315–329.
Brownstone, D., Bunch, D.S., Train, K., 2000. Joint mixed logit models of stated and
revealed preferences for alternative-fuel vehicles. Transportation Research B
34, 315–338.
Choo, S., Mokhtarian, P.L., 2004. What type of vehicle do people drive? The role of
attitude and lifestyle in influencing vehicle type choice. Transportation
Research A 38, 201–222.
Coad, A., de Haan, P., Woersdorfer, J.S., submitted. Consumer support for
environmental policies: An application to purchases of green cars. Submitted
to Ecological Economics.
COWI, 2002. Fiscal Measures to Reduce CO2 Emissions from New Passenger Cars.
Report for Directorate-General for Environment. January 2002.
Creedy, J., 2001. Tax modelling. The Economic Record 77, 189–202.
DeCicco, J., Mark, J., 1998. Meeting the energy and climate challenge for
transportation in the United States. Energy Policy 26, 395–412.
DeCicco, J.M., Geller, H.S., Morrill, J.H., 1993. Feebates for Fuel Economy: Market
Incentives for Encouraging Production and Sales of Efficient Vehicles ACEEET921. American Council for an Energy-Efficient Economy, Washington, D C.
de Haan, P., 2009. Applied comprehensive NO2 and particulate matter dispersion
modeling for Switzerland. International Journal Environment and Pollution 36
(1/2/3), 204–223.
de Haan, P., Keller, M., 2000. Emission factors for passenger cars: application of
instantaneous emission modeling. Atmospheric Environment 34, 4629–4638.
de Haan, P., Mueller, M.G., Peters, A., 2006. Does the hybrid Toyota Prius lead to
rebound effects? Analysis of previously owned cars of Swiss Prius buyers.
Ecological Economics 58, 592–605.
de Haan, P., Peters, A., Scholz, R.W., 2007. Reducing energy consumption in road
transport through hybrid vehicles: investigation of rebound effects, and
possible effects of tax rebates. Journal of Cleaner Production 15, 1076–1084.
de Jong, G., 1996. A Disaggregate model system of vehicle holding duration, type
choice and use. Transportation Research Part B 30, 263–276.
de Jong, G., Fox, J., Daly, A., Pieters, M., Smit, R., 2004. Comparison of car ownership
models. Transport Reviews 24, 379–408.
EU, 1995. Community strategy to reduce CO2 emissions from cars and to improve
fuel-efficiency. COM (95), 689.
EU, 2007. Results of the review of the Community Strategy to reduce CO2 emissions
from passenger cars and light-commercial vehicles. Brussels Feb 2007, COM
(2007) 019.
Giannouli, M., de Haan, P., Keller, M., Samaras, Z., 2007. Waste from road transport:
development of a model to predict waste from end-of-life and operation
phases of road vehicles in Europe. Journal of Cleaner Production 15, 1169–1182.
Golob, T.F., Bunch, D.S., Brownstone, D., 1997. A vehicle use forecasting model based
on revealed and stated vehicle type choice and utilization data. Journal of
Transport Economics and Policy 31, 69–92.
Greene, D.L., Patterson, P.D., Singh, M., Li, J., 2005. Feebates rebates and gas-guzzler
taxes: a study of incentives for increased fuel economy. Energy Policy 33,
757–775.
Hayashi, Y., Kato, H., Teodoro, R.V.R., 2001. A model system for the assessment of
the effects of car and fuel green taxes on CO2 emission. Transportation
Research Part D 6, 123–139.
Hocherman, I., Prashker, J.N., Ben-Akiva, M., 1983. Estimation and use of dynamic
transaction models of automobile ownership. Transportation Research Record
944, 134–141.
IEA, 2006. World Energy Outlook. IEA Publications, Paris.
Jansen, H., Denis, C., 1999. A welfare cost assessment of various policy measures to
reduce pollutant emissions from passenger road vehicles. Transportation
Research Part D 4, 379–396.
Johnson, K.C., 2006. Feebates: an effective regulatory instrument for costconstrained environmental policy. Energy Policy 34, 3965–3976.
Johnson, K.C., 2007. Refunded emission taxes: a resolution to the cap-versus-tax
dilemma for greenhouse gas regulation. Energy Policy 35, 3115–3118.
Langer, T., 2005. Vehicle efficiency incentives: an update on feebates for states.
Report Number T051. American Council for an Energy-Efficient Economy,
Washington DC.
Mannering, F., Winston, C., Starkey, W., 2002. An exploratory analysis of
automobile leasing by US households. Journal of Urban Economics 52, 154–176.
Mohammadian, A., Miller, E.J., 2003. Empirical investigation of household vehicle
type choice decisions. Transportation Research Record 1854, 99–106.
Mueller, M.G., de Haan, P., 2008. How much do incentives affect car purchase?
Agent-based microsimulation of consumer choice of new cars—Part I: Model
structure, simulation of bounded rationality, and model validation. Energy
Policy, in press, doi:10.1016/j.enpol.2008.11.002.
Nilsson, M., 2007. Red light for green paper: the EU policy on energy efficiency.
Energy Policy 35, 540–547.
Peters, A., Mueller, M.G., de Haan, P., Scholz, R.W., 2008. Feebates promoting energy
efficient cars: are counteracting effects to be expected and if so, can their
impacts be limited? Energy Policy 36, 1355–1365.
Saunders, H.D., 2000. A view from the macro side: rebound, backfire, and
Khazzoom–Brookes. Energy Policy 28, 439–449.
Schafer, A., Victor, D.G., 2000. The future mobility of the world population.
Transportation Research Part A 34, 171–205.
Spielmann M., de Haan P., 2008. Umweltindikatoren im Verkehr—Vergleich der
Verkehrsmittel anhand CO2-Emissionen, Energieaufwand und übriger Umweltauswirkungen (in German). Rüegger Verlag, Zürich, p. 76.
Spielmann, M., de Haan, P., Scholz, R.W., 2008. Environmental rebound effects of
high-speed transport technologies: a case study of climate change rebound
effects of a future underground maglev train system. Journal of Cleaner
Production 16, 1388–1398.
Steg, L., Geurs, K., Ras, M., 2001. The effects of motivational factors on car use: a
multidisciplinary modeling approach. Transportation Research A 35 (9),
789–806.
Wüstenhagen, R., Wolsink, M., Bürer, M.J., 2007. Social acceptance of renewable
energy innovation: an introduction to the concept. Energy Policy 35,
2683–2691.