The optimal subsidy on electric vehicles in German metropolitan areas: A spatial general equilibrium analysis

The optimal subsidy on electric vehicles in German metropolitan areas: A spatial general equilibrium analysis

Energy Economics 40 (2013) 515–528 Contents lists available at ScienceDirect Energy Economics journal homepage: www.elsevier.com/locate/eneco The o...

799KB Sizes 0 Downloads 24 Views

Energy Economics 40 (2013) 515–528

Contents lists available at ScienceDirect

Energy Economics journal homepage: www.elsevier.com/locate/eneco

The optimal subsidy on electric vehicles in German metropolitan areas: A spatial general equilibrium analysis☆ Georg Hirte, Stefan Tscharaktschiew ⁎ Technische Universität Dresden, Institute of Transport & Economics, 01062 Dresden, Germany

a r t i c l e

i n f o

Article history: Received 12 December 2012 Received in revised form 30 July 2013 Accepted 3 August 2013 Available online 15 August 2013 JEL classification: H21 Q48 R13 R48 R51 Keywords: Spatial urban model Energy tax Power tax E-mobility Electric vehicles Climate change

a b s t r a c t E-mobility and diffusion of electric vehicles have become a major policy issue in many countries. For example, the German federal government pursues the strategy of achieving one million electric vehicles by 2020. In this paper we examine whether it is optimal to subsidize the use of electric vehicles by granting electric power subsidies and how large the corresponding optimal rate is. We, first, analytically derive the optimal power tax in a spatial model of a city with two zones where commuting, carbon emissions, endogenous labor supply, fuel and power taxes are considered. It is shown that in a spatial urban environment, the optimal tax rate depends in particular on transport related externalities, tax interaction effects and redistribution effects working via the urban land market. Second, we extend the model to a full spatial general equilibrium model and employ simulations to calculate sign and size of the optimal tax/subsidy rate. This model is calibrated to a typical German metropolitan area. The results show that electric vehicles should not be subsidized but taxed. The results are robust with respect to changes in the willingness to adopt electric vehicles, the costs of driving electric vehicles, and even if emissions of electric vehicles are zero. © 2013 Elsevier B.V. All rights reserved.

1. Introduction E-mobility and the impacts of the diffusion of electric vehicles (EVs) have become a topic of high interest for policymakers and scientists in many countries. Lots of governments aim at raising the share of EVs1 – e.g. either as hybrid electric vehicles, plug-in hybrid electric vehicles or full electric vehicles – in the automobile fleet to lower greenhouse gas emissions of road transport and, thus, to mitigate traffic's contribution to climate change.2 For example, Germany's federal government pursues the strategy of achieving one million electric vehicles by 2020 (see Bundesregierung, 2009). 3 However, switching to ☆ This paper was written within the framework of the project “Evaluating Measures on Climate Protection and Adaptation to Climate Change in Agglomerations (EMPACCA)” which is part of the program: “Economics of Climate Change”. Funds from the German Ministry of Education and Research (BMBF) are gratefully acknowledged. The authors also wish to thank two anonymous referees for their valuable comments. ⁎ Corresponding author. E-mail addresses: [email protected] (G. Hirte), [email protected] (S. Tscharaktschiew). 1 Smith (2010) provides an overview of the advantages and disadvantages of EVs. See also Ehsani et al. (2010). 2 According to Thiel et al. (2010), transport related greenhouse gas emissions account for more than a quarter of today's global greenhouse gas emissions where road transport is the biggest contributor to these emissions. 3 For the UK see Department for Transport (2009). 0140-9883/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.eneco.2013.08.001

EVs on account of economic incentives which lower the high costs of these cars raises questions concerning the net social benefits of these decisions and the optimal level of politically set incentives such as subsidies granted for buying or driving EVs. We explore these issues in the following by applying a spatial urban model approach not yet considered in the research on EVs. Of course, there is a large body of literature on this and other EV related issues. Generally, researchers are by far less optimistic than governments concerning the benefits or net benefits of EVs. It is even disputed whether EVs can lower CO2 emissions in passenger transport (beneficial effects of EVs are found by e.g. Karplus et al., 2010; Kazimi, 1997a, 1997b; Nanaki and Koroneos, 2013; Thiel et al., 2010; but negative effects are found by Doucette and McCulloch, 2011; Massiani and Weinmann, 2012; Öko-Institut, 2011).4 There are even some studies calculating social net benefits/costs of EVs (Baum et al., 2011; Carlsson and JohannsonStenman, 2003; Christensen et al., 2012; Funk and Rabl, 1999; Lave and 4 Also regarding environmental impacts of EVs in general, the analysis of Hawkins et al. (2013) suggests that the environmental net benefit of EVs is ambiguous, critically depending on the combination of the vehicle and electricity production impacts as well as key factors such as energy use and battery and vehicle lifetimes. For further EV related studies on emissions or environmental quality, respectively, see Hahn (1995), Lave et al. (1995), Wang (1997), King et al. (2010), Smith (2010), Brady and O'Mahony (2011), Kyle and Kim (2011), Knittel (2012), Shin et al., 2012, Bosetti and Longden (2013), Harvey (2013), and Windecker and Ruder (2013).

516

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

MacLean, 2002; Massiani and Radeke, 2013; Prud'homme and Koning, 2012). Most of them find negative social benefits of EVs. Further, demand for EVs is currently very low despite evidence in favor of a high willingness-to-pay for EVs (e.g. Axsen and Kurani, 2012; Graham-Rowe et al., 2012).5 Studies on pure private costs and benefits of EVs including life cycle cost analyses mostly find negative private net benefits which might explain low demand (e.g. Axsen and Kurani, 2009; Carley et al., 2013; Delucchi, 2005; Delucchi and Lipman, 2001; Kurani et al., 1996; Werber et al., 2009). To foster demand it might, therefore, be appealing to grant subsidies to R&D, or the purchase and use of EVs.6 However, research concerning efficient policies supporting the diffusion of EVs and the analyses of related impacts is surprisingly rare. Moreover, existing studies evaluating potential policies lack general equilibrium considerations which allow accounting for several feedback effects. This is our point of departure. We explore whether the use of EVs shall be subsidized by granting electric power subsidies and how large the corresponding subsidy rate shall be.7 In contrast to the literature, we take a more general view and consider a broad range of developments in technology, emission levels, EV prices and responsiveness of demand for EVs. We employ a fully specified spatial general equilibrium approach in a second best urban environment that allows us to consider social benefits and costs, to calculate changes in emission costs and to derive the optimal subsidy rate. Therefore, our findings are very robust with respect to many issues examined in the literature.8 The focus is on cities because we expect that the use of EVs will be particularly high in cities. They offer sufficiently short cruising ranges and enough density required for battery loading systems. However, in cities congestion is usually higher and travel related taxes/subsidies affect transport decisions. This might also influence spatial location decisions and, thus, decisions on, e.g., distances traveled.9 The general equilibrium approach is appropriate because the welfare outcome of subsidies on the use of EVs depends intuitively on a number of countervailing effects. For example, even if a higher share of EVs in the car fleet actually lowers carbon emissions there might negative side effects of this policy as well as interactions with other policy instruments. These side effects depend on the level of subsidies required to achieve a certain level of diffusion of EVs. For example, if a 5 See, further, Ewing and Sarigöllü (2000), Gardner and Abraham (2007), Lieven et al. (2011), Musti and Kockelman (2011), or Delang and Cheng (2012). 6 Actually, public funds put into related policies are often not negligible. For example, the UK grants a subsidy of up to 5000 British Pounds for buying such a car (Department for Transport, 2012). See also Peterson and Michalek (2013) for an overview on US subsidy policy. 7 There is no single and unified definition of transport subsidies across countries. According to the OECD (2005), a subsidy in general is a result of a government action that confers an advantage on consumers or producers in order to supplement their income or lower their costs. Delucchi and Murphy (2008) use the term ‘tax subsidy’ if there is a difference between actual tax payments and payments under some alternative tax baseline. Such a ‘tax subsidy’ reduces government tax revenues due to a preferential tax treatment in the form of deductions, credits, exemptions, or reduced tax rates. In our study the term subsidy refers to a tax cut that results in a power tax level below the current regular tax level. Alternative subsidization strategies can be: no purchase or value added tax on electric cars; a reduced annual tax; free or cheap use of toll roads, parking places, ferries, and bus lanes on the roads (see Klöckner et al., 2013). 8 The study most related to our paper is Carlsson and Johannson-Stenman (2003), who provide a cost–benefit analysis of policy towards electric vehicles in Sweden. Their theoretical derivation shows that net benefits are equal to net gains in external costs minus the net costs of losses in tax revenue. The latter results from substituting subsidized hybrid or electric cars for highly taxed cars. As a consequence of this negative effect on the public budget, EVs are socially not profitable. Because this effect is weaker for hybrid cars, they might be socially profitable. Similar results are found by Prud'homme and Koning (2012), though they focus on a comparison of EVs with conventional cars. Massiani and Radeke (2013) even focus on Germany by applying a simulation model designed to forecast and evaluate policies towards the diffusion of EVs in Germany. 9 Grazi and van den Bergh (2008) offer a conceptual analysis of the relationship between spatial organization, transport and greenhouse gas emissions. They emphasize the necessity of studying climate related policies (e.g. levying fees and taxes in the transport sector) from a local, spatial planning and policy perspective in order to contribute to efficient and effective mitigation of greenhouse gas emissions.

subsidy is not high enough to fully compensate for the higher vehicle costs of EVs but people switch to EVs because they have a higher willingness-to-pay for EVs, then travel costs increase. This in turn may lower congestion, labor supply and shopping activities in the city. As a consequence, emissions are reduced further but employment declines too. In contrast, if subsidies overcompensate the higher costs of EVs, traffic increases and so do emissions. In addition, financing this subsidy is likely to cause distortions. In a second-best world tax interaction effects and interaction effects among externalities matter too (see Parry and Small, 2005). There might also be spatial relocation as well as changes in the modal split. Whether this strengthens or weakens net benefits is also a priori undetermined. The overall outcome depends on the relative strength of these and other interdependent effects. As a consequence, the overall effect of subsidies to EVs can only be assessed if feedback effects working through different markets are considered. We proceed as follows: First, we analytically derive the optimal power tax in a spatial model of a city with two zones where commuting, carbon emissions, endogenous labor supply, fuel and power taxes are considered, and where we distinguish between conventional fuelpowered cars and electric vehicles. Second, we extend the model to a spatial computable general equilibrium model (CGE) in the tradition of Anas and Co-authors (see Anas and Rhee, 2006; Anas and Xu, 1999; see also Tscharaktschiew and Hirte, 2010a, 2010b, 2012) and employ simulations to calculate sign and size of the optimal subsidy or tax rate. This simulation model is calibrated to a typical German metropolitan area. The spatial CGE approach encompasses endogenous individual decisions of urban households (e.g. spatially differentiated consumption requiring shopping trips, housing, labor–leisure choice where labor supply decisions are associated with commuting trips, location decisions concerning the place of residence and employment, travel mode choice), and accounts for market distortions caused by taxes and subsidies levied by a local/federal government as well as distortions stemming from externalities caused by urban transport activities (e.g. congestion and carbon dioxide emissions). All these decisions and related effects caused by these decisions provoke feedback effects working via urban land, labor and good markets. Public policies which aim to increase the diffusion of EVs and so the share of e-mobility can then have a wide range of differentiated effects eventually affecting welfare of the economic actors. In the analytical part it is shown that, in a spatial urban environment, the optimal power tax rate depends in particular on transport related externalities, tax interaction effects and redistribution effects working via the urban land market. Because of the presence of these differentiated, occasionally countervailing, effects, the sign of the optimal tax rate is ambiguous. In the baseline simulations of the numerical part we find that the social costs of subsidizing the use of EVs exceed the social benefits, thus EVs shall not be subsidized but taxed. This refers to all tax rate levels below the current power tax rate in Germany. In the next stage, we examine whether this result also holds if assumptions are changed as much as possible in favor of EVs. We raise the willingness to adopt EVs so that a subsidy more effectively pushes diffusion of EVs implying that a smaller subsidy is sufficient for achieving the government target regarding the diffusion level of EVs. Again, the findings stay the same. Next, we assume that technological progress and scale economies reduce the average costs of EVs by thirty percent. This also does not change the findings. Eventually, we assume that neither the use of EVs nor the upstream production of power implies any carbon emissions. This might mimic a scenario where power generation exclusively comes from renewable resources. Even this does not change our findings. Hence, our analyses suggest that as long as demand for EVs only boosts if they are subsidized they are not an efficient device to achieve climate change goals as well as to improve urban welfare. 2. Optimal power tax rate in a spatial urban model In this section we analytically derive the optimal power tax in a closed city model with absentee landowners. This model is, though

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

much more simple, structurally identical to the main parts of the numerical simulation (CGE) model used in Section 3. While we calculate the optimal power tax in the simulation model by computing equivalent variations, here we use an optimal tax approach. This shall provide insights into what are the main components of an optimal power tax in a spatial urban setting. The model deviates from a standard monocentric city model because there are two discrete zones, thus, there is no continuous change in distance.10 For the sake of simplicity and to focus on the power tax, we do not consider household heterogeneity, shopping trips, some taxes (income and consumption taxes), and endogenous fuel consumption in the theoretical exercise. Looking at the optimal tax formula below, it will be straightforward how such modifications will enter this formula. 2.1. The model The city is characterized by I = 2 zones i ∈ {1,2}, one called ‘Center’ the other ‘Suburbs’. The city is populated by a given number of households N. Households differ only with respect to their idiosyncratic location preference.11 Both zones are of equal size and the homogeneous and given land area in zone i, A, is normalized so that average two-way travel distance within a specific zone is unity. Commuting is the only reason for traveling and driving by car the only travel mode available. Commuting generates two externalities: congestion and a climate change related externality on account of carbon emissions. For the time being it is assumed that all good prices are set to unity and wages are constant but differ across zones. 2.1.1. Energy choice and transport related externalities Because we do not know which kind of EV will mainly be used in the future we assume that the energy mix of vehicles can be arbitrarily chosen. We assume that a typical household drives a car with a share b of power energy12 and that changing b is free of costs. Accordingly, (1 − b) is the share the typical household drives a conventional fuelpowered car. Energy consumption differs according to the energy type used, where g units of conventional fuel are required to travel one unit of distance with an internal combustion engine vehicle13 and g e units of electric power are required to drive one unit of distance with a pure EV. We further assume that – except for the power tax rate – all travel related costs as well as the amount of energy required per vehicle distance traveled (VDT) are all exogenously fixed. The household's energy mix demand b is a function of the travel costs of an EV, pe + τ e, where pe denotes all travel costs net of the power tax, i.e. costs of owning and operating an automobile including the pure power price, and τ e is the power tax per unit of electric power levied by the government. Equivalently, p g + τ g denotes aggregate travel costs of fuel-powered (gasoline) cars where τ g is the fuel tax rate. Travel 10 Such models are common in the literature and qualitatively provide the same findings compared to models with continuous space (see the overview of Anas, 2012). 11 This might be hidden tastes for specific locations in the urban area that cannot be observed. For example, a certain household prefers more central locations for some reason while another household prefers suburban locations for other reasons which are not explicitly captured by the model's specifications. 12 Households might own different cars so that they can vary energy use over a number of trips. In fact, Klöckner et al. (2013) provide evidence for Norway that electric cars are generally bought as an additional vehicle and not as a substitute for a conventional car. Or, alternatively, one can think of urban households who initially own fuel-powered cars while car-sharing firms purchase EVs to account for the higher private willingness to adopt EVs so that the composition of their car fleet changes towards a larger share of emobility. Then households use their fuel-powered car for some kind of trips whereas an EV is used for other trips (e.g. because environmental concerns become more important regarding urban travel decisions). As a result, EV travel demand increases while traditional fuel-powered traffic decreases in the same way as the share of EVs rises in the car fleet of car-sharing suppliers. 13 Here we take fuel consumption g as given. In the simulation model, however, gasoline consumption will depend on endogenous travel speed and thus on urban traffic conditions.

517

costs pe + τe and pg + τ g are related to one unit of VDT and energy.14 Since the power tax rate is the only EV cost component varied by governmental policy, the decision on b implicitly depends solely on the tax rate of power, τ e. Thus, monetary travel costs per VDT are  e  g  e  g  c τ ¼ p þτ 1−b τ g  e e  e e ′ g g e þ p þ τ b τ g ; b b0; p ; τ ; p const:

ð1Þ

   e  e ∂c ′  e  e  e ′ e e  g g  ′ c ¼ b τ g þ b ¼ b τ g − p þ τ g − p þ τ g −b f0 : ∂b

ð2Þ We assume that b′ is negative and constant implying that a higher power tax entails a lower EV share.15 The sign of c′ is a priori ambiguous. If travel costs per VDT of EVs are higher than travel costs of fuel-powered cars, i.e. the term in square brackets on the righthand side of Eq. (2) is positive, then the relative size of [(pe + τe)g e − (pg + τg)g](−b′) to bg e determines the sign of c′. Otherwise, i.e. if (pe + τe)g e b (pg + τ g)g, it follows c′ N 0 since a higher power tax implies a switch to more expensive fuel-powered cars. Traffic per period (year) in zone i is denoted fi. Normalizing road capacity to unity, the travel time per unit of distance in zone i, ti, depends on travel speed, or inversely on traffic flow, respectively, so that t i ≡ t i ð f i Þ; t ′ i N0; t ″ i N0

∀i;

ð3Þ

where t ′ i N0 implies a congestion (time) externality. Carbon emissions per VDT are determined by accounting for energy type specific carbon intensity, i.e.  e   e   e e em τ ¼ ϕg 1−b τ g þ ϕe b τ g

ð4Þ

  ′ e ′ em ¼ ϕe g −ϕg g b f0;

ð5Þ

where ϕg and ϕe are CO2 emissions per unit of energy. Given b′ b 0, an increase in the power tax rate τ e increases emission per VDT if, and only if, ϕeg e b ϕgg, i.e. EVs emit relatively less carbon per VDT than fuel-powered cars.16 Aggregate emissions then are EM = em × f, where f is total distance traveled in the city (see below). 2.1.2. Households Households face a two-tier utility decision process. Given a specific location choice set ij, i.e. residential location i and working location j, each resident decides on local consumption, and leisure and, thus, how much labor to supply. While daily working time h is fixed, the number of workdays per year, L, and, thus, the number of (one-way) commuting trips is endogenous. Each household consumes one unit of land, thus residential lot size is normalized at unity. Household decisions are based on the random utility function   U ij ¼ u zij ; lij þ ij

∀i; j;

ð6Þ

14 Because energy required per VDT is fixed in the theoretical part, travel costs other than power or fuel can be expressed in per unit of energy. 15 Of course, demand for EVs and, thus, overall diffusion could be modeled in more detail and also depends on further factors (see Daziano and Chiew, 2012; Massiani, 2012; Massiani and Weinmann, 2012; Ozaki and Sevastyanova, 2011). Here we abstract from alternative reasons because we particularly want to focus on a specific policy variable. However, in the simulation part of this study we vary the responsiveness of demand to power subsidies to examine robustness and so we implicitly take care of variables not considered here. 16 The literature cited above suggests that this is true. However, a recent study for the German Federal Environment Agency states that this outcome depends on the kind of energy used in power generation. Only if renewable resources are used to produce a large share of the additionally required electric energy, a reduction of emissions can be expected (Öko-Institut, 2011). One may also argue that a switch to EVs does not generate (additional) carbon emissions at all because electromobility shifts energy demand from a non-capped sector to a capped sector (see Massiani and Weinmann, 2012).

518

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

where zij is consumption of the local good andlij is endogenous leisure demand of household type ij. The idiosyncratic taste constant εij represents the part of the random utility function which is independent from other utility components and variables considered. They are randomly distributed across households according to an extreme-value distribution. The monetary budget constraint of household type ij states that expenditures for consumption goods and housing equal wage income minus aggregate monetary commuting costs minus lump-sum taxes, i.e.   e  ls zij þ ri ¼ w j h−c τ δij Lij −τ ;

Because idiosyncratic tastes εij are stochastically distributed among households for each ij, the probability that a household chooses the

specific location choice set ij is Ψij ¼ Pr V ij þ εij NV e þ ε e; ∀iej≠ij , ij

Ψij ¼

ð7Þ

where ri is the endogenous land price in home zone i per residential lot, wj is the fixed wage rate in employment location j, δij denotes two-way travel distance between zone i and zone j, and τ ls is the lump-sum tax. In addition, the household is subject to a time constraint stating that total time endowment H is fully allocated to time spent on working, commuting and leisure:

ij

where   exp ΛV ij 2 X 2 X

;

ð13Þ

expðΛV ab Þ

a¼1 b¼1

is the probability that a randomly selected household most-prefers the location choice set ij.18

where tij is two-way commuting time – taken as given by urban travelers – from residential location i to employment location j defined as

2.1.3. Closing the model The model is closed by considering the remaining parts of the model, i.e. the government, firms producing local goods, an external transport sector, absentee landowners and, eventually, good, land and labor markets. The government levies a fuel, a power, and a lump-sum tax to finance fixed expenditure, S. The tax bases are total fuel consumption, G, total power consumption, Ge, and the number of households in the city, N. The government budget constraint is

    t ij f i ; f j ≡ t i ð f i Þ þ j j−ij  t j f j :

τ G þ τ G þ τ N ¼ S:

   h þ t ij f i ; f j Lij þ lij ¼ H

ð8Þ

ð9Þ

g

  V ij θij ; Y ij ¼

  max u zij ; lij

s:t: : zij þ θij lij ¼ Y ij ; zij ;lij

ð10Þ

ð11Þ

(   e c τ if i ¼ j  e ; 2c τ if i≠j

ð12Þ

It is assumed that each commuter travels the same distance within a zone i. These workers are those living and working in the same zone i, those living in zone i and leaving the zone for working in zone j, and those living in the other zone j and entering zone i for working. Two-way travel distance within a zone is normalized to unity (δii = δjj = 1), implying that two-way distance for traveling across zones is δij = δji = 2. Accordingly, a worker not traveling in both zones faces costs c for his commuting round-trip, otherwise 2c. Utility maximization (see the household's optimization program Eq. (10)) then  yields  the demand functions for consumption zij(Yij), and leisure lij θij ; Y ij , respectively. 17 Full economic income Yij is derived by solving the time constraint Eq. (8) for the number of workdays Lij and then plugging the resulting expression into the monetary budget constraint Eq. (7).

ð15Þ

e

G ¼ fbg ;

ð16Þ

depend on aggregate two-way distance traveled in the city, f = ∑i fi. Traffic flow in zone i, fi, which equals total two-way distance traveled in zone i,19 is determined by aggregating the number of (one-way) commuting trips Lij over all households traveling in zone i fi ¼ N

is the opportunity cost or shadow price of one unit of (leisure) time. It is equivalent to the average hourly net wage income. The numerator denotes daily gross wage income net of monetary commuting costs required for supplying one working day. The denominator is average daily time spent working and commuting. The two-way monetary travel costs per trip from zone i to zone j are given by  e c τ δij ¼

ð14Þ

and the power tax base e

  w j h−c τe δij   ; h þ t ij f i ; f j

ls

G ¼ f ð1−bÞg

where Yij ≡ θijH − τls − ri is full economic income, i.e. monetized time endowment net of lump-sum taxes and housing cost.17 The value of time (VOT) θij ¼

e

Both, the fuel tax base

The household's ij indirect utility function is 

e

X j

Ψij Lij þ N

X

Ψji Lji :

ð17Þ

j≠i

Transport services are supplied by external firms that use their revenue to cover costs of resources used as inputs. Further, we assume that absentee landowners own all local land. The model is, thus, fully closed by a current account that is balanced through good exports being equal to fuel and power imports plus rent payments to absentee landowners. Private consumption plus exports add up to demand (and supply) for urban commodities. Because local goods are produced with a constant-returns-to-scale production function where local labor is the only input, the good market is cleared too. While labor supply is endogenously determined, labor demands of firms are fully elastic. Hence, spatially differentiated wages are constant and labor markets are cleared too. The only markets where prices adjust to ensure market clearing are the two land markets. Land supply is fixed but location specific aggregate land demand is endogenously determined and spatially differentiated land rents adjust to clear these markets.20 The whole 18 Assuming that each εij is independent and identicallyGumbel pffiffiffi distributed (i.i.d.) with mean zero, variance σ2 and dispersion parameter Λ ¼ π= σ 6 , the choice probabilities are given by the well-known multinomial logit model. See, e.g., Train (2003) who provides a discussion on the properties of the logit probabilities. 19 This holds because intrazonal two-way travel distance (δii and δjj) is normalized to unity and, accordingly, interzonal two-way travel distance (δij and δji) is 2. Hence, each trip originating and terminating in zone i travels a total of one unit of (two-way) distance in zone i. 20 Although individual land demand is fixed at unity, aggregate land demand in a specific zone i, N∑j Ψij, depends on the endogenous spatial location decision Ψij.

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

system can thus be reduced to both land markets plus the government budget constraint. Besides choice variables of urban households, endogenous variables are land rent in the center, rent in suburbs and the tax rate used for tax revenue recycling.21 2.2. Welfare and optimal power subsidy In the following we use this model to derive the effects of a marginal increase in the power tax rate on indirect utility of a household of type ij, on urban and aggregate welfare. Eventually, we derive the optimal power tax rate to shed light on the question whether power used for driving EVs should be taxed or subsidized. In the first step, we derive the effects of an increase in the power tax rate on a household's indirect utility. Differentiating indirect utility Eq. (10) with respect to τe and applying Roy's theorem gives the marginal utility change of household type ij (in terms of income) caused by a marginal increase in the power tax rate (see Appendix A.1): i 1 dV ij dτls h e  e e g  ′ e ¼− e − bg þ τ g −τ g b f ij − λij dτ dτ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl ffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Net transfer from the goverment



dt ij θij Lij e dτffl} |fflfflfflfflffl{zfflfflfflffl

dr − ei dτ |{z}

 e e g  ′ p g −p g b f ij |fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Change in travel costs net of taxes

ð18Þ

;

Change in congestion costs Change in land costs

where λij is the marginal utility of income (MUI) of household ij and fij = δijLij is household's ij total commuting distance traveled. The increase in the power tax affects individual utility through four channels: (i) utility of a household ij increases, first, if the additional public net transfer is positive. This effect can be distinguished into the change in the lump-sum tax and the change in travel related tax payments. In fact, the change in the lump-sum tax depends on the interaction of τls with other taxes (see below). Suppose there are no interactions with other taxes, in such a case, an increase τe implies dτls/dτe b 0 since higher power tax revenue, ceteris paribus, allows the government to lower τ ls while the public budget stays balanced. As a result, welfare of household ij increases with an increase in τe. The later effect (change in travel related tax payments) reflects the higher direct power expenditure when τe raises (bg e). It is corrected by the change in tax payments induced by a smaller EV share (b′ b 0) in individual travel activities. Making the reasonable assumption that fuel tax expenditures per VDT are higher, τege b τgg, a higher power tax raises energy related tax payments due to the switch to more expensive fuel-powered cars ((τege − τ gg)b′ N 0). As a consequence, both expressions in square brackets are positive, causing a drop in individual utility. However, the net effect of the government is ambiguous because dτls/dτe b 0.22 (ii) Second, the effect of changes in commuting costs net of taxes is equivalent to the tax payment effect in (i), but the opposite way around. Provided pege N pgg, a higher power tax improves utility since b′ b 0. (iii) Third, there is the theoretically undetermined change in congestion costs, θijLijdtij/dτe. As shown in Appendix A.6, the change in commuting time with respect to the change in the power tax is composed of two terms: the response of traffic flow to changes in labor supply (thus to the change in the number of trips) and the response of traffic flow to relocation (thus to the change in travel distance). The overall effect and thus the impact on individual utility is ambiguous because neither the sign of the first nor that of the second effect is theoretically unequivocal. (iv) Eventually, there is the change in land (housing) expenditure arising from relocation decisions which affect the land market. The sign of dri/dτe is a priori unknown too.

519

These different effects imply that the direction of utility changes is theoretically ambiguous. This is also true when aggregating individual utility changes of residents to determine urban welfare. Welfare of urban households is calculated as the expected value of the maximized utilities (see e.g. Anas and Rhee, 2006; based on Small and Rosen, 1981). Under the assumption that idiosyncratic tastes εij for a specific location choice set ij are i.i.d. Gumbel distributed, welfare of a representative city household is 2 X 2 h  i 1 X   W H ¼ E maxðijÞ V ij þ εij ¼ ln exp ΛV ij : Λ i¼1 j¼1

¼

I X I X

Ψij V ij þ

i¼1 j¼1

I X I   1X Ψ −lnΨij Λ i¼1 j¼1 ij

ð19Þ

ð20Þ

Urban welfare includes all taxes, i.e. sales, income, fuel, lump-sum and power taxes, even if they are federal, because they affect individual utility and, thus, urban welfare due to their influence on household income and prices. Aggregate social welfare then additionally accounts for further effects which are linked to decisions and behavioral responses of urban households. Climate change is a global phenomenon though urban households exhaust carbon emission when traveling, thus, contributing to climate change. Therefore the city's contribution to climate change costs is not taken into account when calculating local (urban) welfare, but included in the social welfare function. Eventually, since policy induced responses of urban households might even affect land (housing) prices (e.g. via changes in spatial location decisions), social welfare has to take into account welfare changes of absentee landowners who receive land rent income from city households. Therefore, aggregate social welfare is urban welfare plus indirect utility of absentee landowners minus the climate change externality caused by carbon emissions, formally W ¼ W U þ V A −λE EM I X I I X I   X 1X Ψij V ij þ N Ψij −lnΨij þ V A −λE EM ; ¼N Λ i¼1 j¼1 i¼1 j¼1

ð21Þ

where WU = NWH, VA is indirect utility of absentee landowners, and (− λE) represents the marginal (dis-)utility of one unit of carbon dioxide emission. By applying an optimal tax approach in a spatial setting (see also Rhee (2012)) we totally differentiate each component of Eq. (21), use Eq. (18) and the government budget constraint (Eq. (14)) to get the social welfare change in terms of income with respect to a change in the power tax (see Appendix A.4 for the change in city welfare and, finally, Appendix A.5 for the total social welfare change):   31 0 2   f −b′ dW @ t em e e 5A − df þ IE MEC þ MEC −ητ g 4b þ e ¼ |{z} dτ dτ e ð−df =dτe Þ Tax interaction |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Net social marginal costs of externalities



TC b |{z}

þ RELand |fflfflffl{zfflfflffl}

Change in travel costs due to lower EV diffusion Redistribution via land market

ð22Þ

þ RETAX |fflffl{zfflffl} Redistribution via tax system

where η represents the marginal cost of public funds (MCPF) defined as (see Appendix A.2). η≡

N : N þ τ g dτdG

ð23Þ

ls

The MCPF is the welfare loss of all households per unit of tax revenue raised by a marginal change in the lump-sum tax. If NbN þ τg dτdG , i.e. η b 1, the social benefit from taxation (the denominator in Eq. (23)) exceeds the marginal private welfare loss (the numerator in Eq. (23)) caused by taxation. In contrast, if η N 1, rising public revenue ls

21

In the simulation model most of the restrictions are loosened. 22 Assuming the absence of tax interaction effects, which are, however, considered hereafter.

520

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

from the private sector causes efficiency losses due to related tax distortions.23 Eq. (22) shows that an increase in the power tax rate affects social welfare through five channels: first, the change in social net costs of externalities; second, tax interaction effects, IE; third, changes in travel costs arising from diffusion of EVs, TCb; fourth, the changes in the redistribution on the land markets, RELand; and, fifth, the redistribution via distortionary taxes, RETAX. The first channel depends, on the one hand, on the response of travel demand to the change in the power tax and, on the other hand, on the difference between social marginal externality costs (congestion MECt and carbon emission, MECem) and the social costs of the power tax. To discuss the overall contribution to the welfare change, let us look at the definitions of both marginal externalities (see Appendix A.4 and Appendix A.5) t

MEC ≡

N

XX i

j

Ψ

λij ij λ

 dt  θij Lij −dτ ij e

ð−df =dτ e Þ

  1 e ′ −b f ϕg g−ϕe g A; ≡ ω@em þ ðdf =dτe Þ

ð24Þ

MEC

ð25Þ

where ω are social costs per unit of carbon emission. The sign of the congestion externality is positive, i.e. MEC t N 0, since usually changes in household travel time move into the same direction as changes in traffic flows.24 Furthermore, because em N 0 and f(ϕgg − ϕeg e)(−b′) N 0, the sign of the marginal carbon emission costs, MECem, is strictly positive as long as higher power taxes increase total distance traveled, i.e. df/dτe N 0. Even though it is not unequivocally clear, let us assume that the sum of marginal externalities is positive and larger than the social value of the initial power tax per VDT. We consider this to be the more likely case. Then, raising the power tax (or the other way around granting a power subsidy) unequivocally lowers (increases) welfare if the tax policy implies df/dτe N 0. The second effect in Eq. (22) is the tax interaction effect (IE): IE ≡ ητ

g

 e   dG  G ∂G ′ − þ gf −b þ Θe : ls N ∂τ dτ

   g e e ′ TC b ≡ p g−p g −b f λ ;

ð26Þ

In a second-best world changes in taxation affect distortions generated by other taxes, here the fuel tax. This is true even with lump-sum tax recycling. This effect is indicated by the tax interaction effect, IE. It is composed of three terms. (i) There is an indirect effect of the power tax on fuel tax revenue via lump-sum tax recycling. Raising τe – allowing for a reduction in τls – imposes a positive income effect on the normal good leisure. As a consequence, labor supply and, associated with labor supply, the number of commuting trips, thus, fuel tax revenues decline for a given level of diffusion, i.e. (−∂ G/∂ τls) b 0. (ii) There is a direct effect of the power tax on the fuel tax base via changes in EV diffusion. Because a higher power tax implies a smaller EV diffusion, the impact on the fuel tax base is positive, i.e. ∂ G/∂ τ e = gf(−b′) N 0. (iii) Eventually, there are general equilibrium effects via relocation and changes in land rents, labor supply and distance traveled all affecting the fuel tax base (see Eq. A.12 in Appendix A.3). On account of the opposite sign of the first and second term and the undetermined sign of the third term, one cannot derive the sign of IE theoretically. Suppose, however, that the effect on the fuel tax base, ∂ G/∂ τe N 0, dominates, 23 Despite considering lump-sum taxes, η is not unity because dG/dτe contains different components one of which is ∂ G/∂ τls. 24 This is confirmed by our simulations (see the corresponding tables in Section 4). One should, however, note that in the theoretical model traffic flow and total travel distance are identical, while in the simulation model both might differ because total distance traveled could, e.g., decrease because people prefer shorter trips while they increase the number of trips. In such a case, the signs of dtij/dτe and df/dτe might differ.

ð27Þ

where f λ ¼ N∑i ∑ j Ψij λλ f ij is weighted aggregate distance traveled (total traffic flow). Since b′ b 0 and because currently pege N p gg travel costs net of taxes decline and welfare improves with an increase in the power tax. In contrast, a lower power tax rate (granting power subsidies) worsens welfare. Fourth, there are redistribution effects on the land market, RELand, resulting from differences in the value of land costs to the landowners and the city households. It is defined as ij

X dr dR RELand ≡ A i ei − λe ; dτ dτ

ð28Þ

dr where dR ¼ N∑i ∑ j Ψij λλ dτ is the weighted aggregate change in land dτ (housing) costs of city households due to an increase in the power tax. The sign of this redistribution effect is a priori ambiguous because land prices depend on relocation decisions of households that cannot be determined unequivocally without further inspection. Fifth, since welfare is converted into monetary terms by means of the average (expected) MUIs, deviations of individual MUIs from the average MUI require a correction of the welfare change measure if distortionary tax payments are not uniform across households who live and work in different locations. In that case, the tax system redistributes among city households. Theoretically, this tax redistribution effect which cannot be determined by inspection is defined as ij

λ e

0

em

then granting a power subsidy causes a negative tax interaction effect IE b 0 which lowers social welfare. Third, a change in EV diffusion affects aggregate travel costs:

i e

h   i  e e e g  ′ RETAX ≡ bg ð f − f λ Þ þ τ g −τ g −b f λ :

ð29Þ

Adding up all welfare components reveals that it is not possible to determine the overall sign of the social welfare change by theory. Therefore we apply simulations below. Nonetheless, we can derive an equation for the optimal power tax rate. Setting Eq. (22) to zero and solving for τ e yields the socially optimal power tax rate25   df  e t em − e þ Adj  IE τ ¼ Adj  MEC þ MEC − Adj  TC |fflfflfflfflffl{zfflfflfflfflffl} |fflfflfflfflfflffl{zfflfflfflfflfflffl}b dτ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Adjusted tax interaction Adjusted change in travel costs

ð30Þ

Adjusted Pigouvian tax

þ

Adj  RELand |fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl} Adjusted redistribution on the land market

þ Adj  RETAX |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} Adjusted redistribution via taxes

where the adjustment term Adj ≡

1      ηg e b −dτdf þ f −b′

ð31Þ

e

represents the inverse social value of the change in the power tax base. The socially optimal power tax rate Eq. (30) is the sum of the adjusted Pigouvian tax, the adjusted tax interaction effect, the adjusted changes in travel costs due to a change in EVs diffusion, the adjusted redistribution effect among landowners and urban households, and the adjusted redistribution via the tax system. If driving EVs is (net of power and fuel taxes) more expensive than driving fuel-powered cars, i.e. p gg b pege, it follows TCb b 0. Hence, the contribution of the pure change in travel costs to the optimal power tax level is positive since a higher power tax causes a substitution towards less expensive fuel-powered cars. Moreover, if smaller EV 25 This solution is not exactly the welfare maximum, but an approximation to it. Welfare is maximized not considering idiosyncratic utility components of the random utility approach (Anas, 2012).

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

diffusion induced by an increase in τe leads to IE N 0, the pure tax interaction effect also affects the power tax level positively since, in such a case, the positive effect of higher fuel tax revenue due to a decrease in EV diffusion is dominant in the tax interaction effect. However, the final contribution of both the travel cost effect, TCb, and the tax interaction effect, IE, to the optimal power tax level is in fact not unequivocally clear because of the adjustment term.26 Looking at the adjusted Pigouvian tax term reveals that with df/dτe N 0, leading to MECt N 0 and MECem N 0 (see Eqs. (24) and (25)), there is, not surprisingly, a downward adjustment of the optimal power tax (provided Adj N 0). As becomes clear from the optimal power tax formula Eq. (30) these and all further effects depend in particular on the sign of the undetermined response of traffic flows (= VDT in the theoretical part) to the power tax, i.e. df/dτe. As a consequence, we cannot decide only by theory whether power used for driving EVs should be taxed or subsidized. What are the consequences of these findings for our policy analysis? Suppose that the focus of policymakers regarding the promotion of EV diffusion is on emissions only. Then, a power subsidy would be adequate if the conditions described above are fulfilled, i.e. particularly df/dτe N 027 while a tax might be optimal if df/dτe b 0. However, as theory shows, other effects are also important. In the presence of congestion, the optimal power tax rate as well as welfare changes caused by power subsidy induced EV diffusion are affected by the level of congestion.28 Moreover, while redistribution effects might be small, the tax interaction effect and travel cost changes are likely to work against subsidization. In the end, however, we have to apply simulations to see the relative strength of the different effects and the overall outcome of the EV diffusion policy under consideration. 3. Simulations of the optimal power tax/subsidy 3.1. Spatial CGE model For the simulations we extend the analytical spatial model to a full spatial computable general equilibrium model29 and consider some additional aspects with respect to EVs which are important to determine sign and size of the optimal tax (subsidy) rate. In particular we add sales taxes and progressive income taxation according to the German tax tariff.30 Besides endogenous residential and employment location decision as well as labor supply (leisure demand) decisions, heterogenous households (low-skilled and high-skilled working households as well as non-working households) now additionally choose where and how much to shop (number of shopping trips),31 how much land to rent (lot size),32 and which travel mode to use. All these interdependent decisions are endogenously determined in the simulation model and implicitly determine commuting and shopping trip distances, frequencies and, along with travel speeds, 26 However, the adjustment term is supposed to be positive because even in the case of df/dτe N 0, the positive second term in square brackets of the Adj term is likely to dominate the (in such a case negative) first term. 27 MECem(−df/dτe) b 0 if df/dτe N 0 (see Eq. (30)) justifying a power subsidy (or a lower power tax, respectively). 28 Numerical simulations employed in the second part of this study support the importance of interaction and feedback effects related to congestion. For example, we find that in most of the scenarios considered subsidies to EVs serve as an internalization instrument for all three kinds of externalities associated with congestion (excess time delay, fuel consumption, CO2 emission). 29 Despite its extensions, the full spatial CGE model is structurally identical to the analytical model. 30 As long as additional taxes are fixed and not used for tax recycling they only constitute additional terms in the tax interaction effect. 31 Concerning consumption it is assumed that residents have a preference to go shopping at different locations implying that there is spatial product differentiation. Preferences for spatial consumption variety are implemented by a C.E.S. shopping subutility function. 32 Overall utility derived from spatially differentiated consumption, housing, and leisure is of the Cobb–Douglas form.

521

travel times. We further add local urban production of goods/services where labor and land are inputs into production. In the benchmark case we assume an initial e-mobility share of (roughly) zero (i.e. b = 0).33 The simulation model of the benchmark case (see Section 4) is the spatial urban general equilibrium model fully described in Tscharaktschiew and Hirte (2010a, 2012).34 E-mobility and the diffusion of EVs are then launched by assuming b N 0 (see the discussion in Section 4). The spatial CGE approach employed and the tax policy applied is a comparative static analysis. This implies that we calibrate the benchmark to some year (2009) and calculate the counterfactual equilibrium considering the policy. Most decisions are medium-term such as decisions regarding the number of labor days supplied, residential relocation and the change of the working place. On the other side very long lasting adjustments are not considered because we do not consider housing construction and land, respectively, housing supply is fixed. For these reasons, it is justified to presume that the implicit time horizon of our simulations is in the range of around ten years. This is also the perspective of the policy that aims to achieve a certain degree of diffusion of EVs until 2020. The metropolitan area under consideration now encompasses I = 9 zones (locations), where the innermost zone i = 5 is assumed to be the city center. All zones have a length of 4.5 km so that the whole urban area expands over 40.5 km. The zones 3–7 shape the city of the urban area while zones 2 and 8 (1 and 9) form the surrounding inner (outer) suburbs. In each zone i (i ∈ I) there is a given homogeneous land area available for residences, establishments (firms) and roads. Supply of land increases with distance from the city center. In the benchmark, urban residents can choose among travel modes walking, public transport, and automobile (conventional, i.e. fuelpowered car) for their commuting and shopping trips in the urban area.35 Traveling by car requires gasoline and may cause congestion (travel time delays depending on the endogenous traffic flow–road capacity ratio) and CO2 emissions. Automobile travel times, gasoline consumption and CO2 emissions are all a function of traffic congestion (and thus endogenously determined) and are specified by empirically derived functional relationships (see Appendix B.1). Hence, in contrast to the theoretical part, fuel consumption and CO2 emissions depend on travel speed and, thus, on traffic conditions. Consequently, changes in CO2 emissions can be distinguished, on the one hand, into changes in emissions stemming from a decline in travel demand (e.g. a decline in fuel-powered travel demand due to EV diffusion) and, on the other hand, changes in emissions due to improvements in fuel intensity (fuel consumption and emissions per VDT) caused by reduced congestion.36 A sufficiently large number of firms produce in each zone i zone specific commodities/services by applying a Cobb–Douglas technology that combines land and labor supplied by low-skilled and high-skilled workers. Within each zone there is competition implying price taking behavior of firms which sell their products/services at the competitive mill price (see Appendix B.2). The federal government levies progressive income taxes, sales taxes and energy (gasoline) taxes, grants transport subsidies and income transfers to non-working households and redistributes – according to fiscal interdependencies among public authorities in Germany – shares of its revenues to the local urban government. The federal tax revenues not redistributed to the urban private households and the city government are used for public consumption consisting of purchasing 33 According to Lieven et al. (2011), in 2009, the share of EVs on all new car registrations in Germany amounted to a percentage of only about 0.004 (162 EVs/3.8 million new vehicles). 34 In the following we describe the main features of the (benchmark) model (full technical details are provided in Tscharaktschiew and Hirte, 2010a, 2012). 35 Recall again that in the benchmark we assume that e-mobility is negligible, thus its share is zero. 36 In the following MECt, MECg, MECem refer to the latter effect, thus changes in externalities due to changes in congestion. The change in overall emissions is represented by ΔWem.

522

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

Table 1 Some results of the benchmark compared with empirical evidence. Average (over all locations and persons)

Urban model

Empirical evidence

Source

Gross wage [€/h] urban area Gross wage [€/h] city Average income tax rate [%] Work days [days/year] Percentage Commuters (commuting distance b10 one-way km) One-way commuting distance [km] Ratio shopping trips/commuting trips Share travel costs on disposable income VOT vs. gross wage/net wage Job–housing–balance Number of jobs in i Suburb=city Numbers of workers residing in i

19.23 19.56 20.6 219 55 12 1.29 0.09 52%/77%

20.04 (Germany 2007) 19.26 (Berlin 2007) 20.3 (2004) 215–223 (2004) 52 (2004) 12–13 1.32 (2002) 0.10 47%/78%

[1] [1] [2] [3] [4] [5] [6] [7] [8]

0.75/1.37

0.87/1.54 (Hannover) 0.79/1.33 (Hamburg) 0.86/1.39 (Munich) 0.89/1.56 (Stuttgart)

[9] [9] [9] [9]

Own-price elasticity of travel demand for private automobile with respect to gasoline price

−0.2

(−0.1)–(−0.3) −0.3 −0.2 (−0.1)–(−0.5)

[8] [10/12/13] [14/15] [11]

Own-price elasticity of travel demand for public transport with respect to transit fare

−0.7

−0.4 (on average) (−0.5)–(−0.6) (Bus) (−0.4)–(−1.0) (Metro) (−0.1)–(−1.1) (Rail) (−0.0)–(−0.8)

[8] [10] [10] [10] [11]

Cross-price elasticity of travel demand for public transport (tram) with respect to gasoline price

+0.3

+0.3 (on average) (+0.1)–(+0.8) (Range)

[10] [10]

[1] Arbeitskreis Volkswirtschaftliche Gesamtrechnungen der Länder (2009). [2] Federal Statistical Office (2008b) [3] IAB (2005) [4] Federal Statistical Office (2005). [5] Federal Ministry of Transport, Building and Urban Affairs (2009). [6] Federal Ministry of Transport, Building and Urban Affairs (2004). [7] Federal Statistical Office (2009). [8] Small and Verhoef (2007)/De Borger and Van Dender (2003) [9] Siedentop (2007). [10] Goodwin (1992) [11] Oum et al. (1992). [12] Goodwin et al. (2004) [13] Graham and Glaister (2004). [14] Hymel et al. (2010) [15] Steiner and Cludius (2010).

locally produced commodities. The city government receives its shares of federal tax revenues and levies a local lump-sum tax to finance local goods such as roads. Infrastructure costs consist of opportunity costs due to land used for infrastructure. Absentee landowners use their rent income and an external transport sector monetary travel cost revenues (except for travel related taxes) accruing from urban travel activities to purchase urban commodities. At spatial urban general equilibrium, endogenous land rents, wages and commodity prices clear the spatially differentiated markets for land, low-skilled labor, high-skilled labor and commodities. Moreover, closing the model requires that financial outflows due to e.g., tax payments and rental income of absentee landowners must be balanced by physical outflows represented by exports of urban products (see Appendix B.3). 3.2. Calibration (benchmark case) We calibrated the model to an ‘average’ German metropolitan area featuring important German data. The model calibration ensures that the benchmark case exhibits • A household composition (e.g. the relative shares of low-skilled and high-skilled workers),37

37 The total number of households in the urban area is assumed to be 1.75 million. In 2007 in Germany, the share of households with adult working age persons (the economic head of the household is at 18–65 years of age) amounted to about 70% (Federal Statistical Office, 2009). Assuming that a percentage of 5 of all households with adult working age persons is actually not working, the number of working households 1,163,750. Accordingly, the number of non-working households amounts to 586,250. The percentage of highskilled workers – reflecting an educational attainment achieved by studying at universities of applied sciences or a common university as well as advanced degrees (e.g. PhD) – is assumed to be 20% compared to a percentage of 20.5 in Munich, 17.5 in Frankfurt/Main, 20.9 in Stuttgart, or 20.3 in Dresden (Stadt Frankfurt/Main, 2009).

• Economic characteristics (e.g. rents, wages, incomes, income tax rates, federal tax revenues), • Travel characteristics (e.g. modal split, relative importance of trip purposes, travel demand elasticities, average commuting distances and average automobile travel speeds, average gasoline consumption (fuel economy) and CO2 emission per vehicle kilometer) and • Spatial patterns (e.g. residential and employment densities,38 jobhousing-balance,39 the share of urban land allocated to roads40), that are representative for a German urban area. A detailed description of the calibration including the chosen parameter values and the results of the benchmark simulation can be found in Tscharaktschiew and Hirte (2010a). Here we only repeat a comparison of some benchmark results with empirical evidence (see Table 141) showing the accuracy of the calibration procedure and suggesting that there is an appropriate fit of data.

38 In the benchmark urban area population as well as employment densities endogenously peak in the city center and decrease with distance from the center as observed in real urban areas. 39 The benchmark urban economy reflects a realistic spatial pattern regarding the job– housing-balance which is defined as the ratio of the number of jobs in a specific location to the number of employees residing in that location. According to evidence cited by Siedentop (2007) the job–housing-balance exceeds unity for central cities and falls short of unity in the suburbs which is fully in line with the spatial pattern of the benchmark city. 40 The total land area allocated to roads amounts to 15.3% in the model benchmark city (zones 3–7). For comparison, for example in the city of Berlin (Munich), the share of land area allocated to roads amounted to 15.3 % (17.2%) in 2007 (Berlin: Federal Statistical Office, 2008a; Munich: Statistical Office Munich, 2009). 41 This table is reproduced from Tscharaktschiew and Hirte (2010a).

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

523

4. Simulation results and discussion

0.3

In the simulations we calculate emissions and welfare for different power tax rates to derive the optimal subsidy or tax level. First, we present some hypotheses on the expected changes in the optimal tax (subsidy) rate in the full spatial CGE model. Thereafter we present simulations and sensitivity analyses.

0.2

4.1. Optimal power tax in the spatial CGE model Since the full CGE model extends the basic theoretical model considerably, the optimal tax formula changes and it is only possible to describe some obvious changes in comparison to the optimal tax formula Eq. (30). First, a fuel consumption externality42 enters the adjusted Pigouvian tax component. Second, sales tax and progressive income tax interactions have to be added to the tax interaction term (IE). If there would be no EV diffusion, the sign of the income tax interaction term would be the same as the sign of the fuel tax interaction term because traveling and labor supply are complements in this model. However, there is a switch to power consumption so that fuel tax interaction and income tax interaction effects might have the opposite sign. Third, redistribution effects differ because a share of landowners is living in the city (mixed landownership). This weakens the redistribution effect. Further, because we consider different household groups additional redistribution components may occur. Eventually, there are changes in traffic flows, fuel consumption, and power consumption which we cannot derive explicitly but which affect the optimal tax rate in all components in an a priori ambiguous way. 4.2. Baseline simulation We, first, present the results of the baseline simulation. In this simulation we assume that the response of the energy mix with respect to the power tax is moderate. Unfortunately there is no information on elasticities and thus on the response of travelers to the power tax because the current number of EVs is almost zero. We therefore choose the following approach: We define a function that creates the initial level of almost zero EVs at the benchmark gross power price of 0.203 €/kwh (net power price of pe = 0.182 €/kwh43 + power tax rate of τe0 = 0.021 €/kwh, see Mühlenhoff, 2011) and, then, produces an almost proportional response to a change in the power tax rate.44 e e The energy mix function is defined as a function bðτe Þ ¼ e−aðp þτ Þ − e e −aðpe0 þτe0 Þ 45 e . Since the net power price is fixed (p0 = p ) the energy mix depends only on variations in the power tax rate (starting from the current power tax rate of 0.021 €/kwh). Fig. 1 shows the plot of b(τe) for a = 1 (dashed curve), a = 0.5 (solid curve), and a = 0.1 (dotted curve). The horizontal straight line displays the (medium term) policy target in Germany, i.e. a share of about 0.025 of EVs on all cars. The elasticities of b with respect to the gross power price achieve reasonable values at the intersection between the b-function and the policy target line. For example, the response parameter a = 0.1 implies that a subsidy of 0.252 €/kwh is required to achieve this goal. In this case the gross power price drops to −0.049 €/kwh. With a higher response parameter 42 Because fuel consumption depends on congestion, an additional car driver can affect the level of fuel consumption of all other drivers on the road. 43 In the simulation part pe denotes the pure power price while travel costs other than those related to energy are modeled separately. 44 In our policy scenarios we consider a subsidy that generates a negative gross price of power and, thus, turns energy use into a source of income as not politically feasible. Hence, our maximum subsidy granted will be about 0.20 €/kwh. The energy mix function then also ensures that the EV share b is below unity at this subsidy level. 45 The energy mix function is slightly convex meaning that EV diffusion responds the stronger to a marginal increase in the subsidy the higher the subsidy and thus diffusion of EVs. This is chosen to account for e.g. positive (social) contagion, i.e. an easier switch to EVs because loading infrastructure or car technology is getting better the larger the level of diffusion or because social acceptance of EVs is getting higher the higher the diffusion of EVs.

0.1

-0.2

-0.1

0.0

0.1

0.2

Power price [€ €/kwh] a = 0.1

a = 0.5

a = 1.0

Fig. 1. Responsiveness to power subsidy.

a = 0.5 a subsidy of about 0.05 €/kwh is sufficient to achieve this policy goal. For these data the energy mix function implies that some households switch to EVs if they receive a small subsidy even if driving by EVs is more expensive. This accounts for the fact that the willingnessto-pay for EVs is high for some households and differs across households (see references in the introduction). In addition, we finally have to add central data on EVs and the use of power which become relevant for the case of b N 0. We choose the following basic figures: average costs per EV are 0.50 €/km (ADAC, 2012); traveling by EVs requires 13.15 kwh/100 km (ADAC, 2012; see also Lorf et al., 2013). Carbon emissions are set at 563 gCO2/kwh (Federal Environment Agency, 2011).46 We choose a = 0.1 as reference and calculate the baseline case with this parameter. Fig. 2 displays changes in overall welfare ΔW (the sum of the equivalent variation of urban households47 and absentee landowners plus CO2 emission cost reduction benefits). It also shows, as a separate curve, changes in CO2 emission costs48 for different levels of the tax/subsidy (horizontal axis), where positive values indicate social benefits from a policy induced reduction in CO2 emissions. Fig. 2 shows that the higher the subsidy rate on power (negative tax) the higher overall welfare costs while emission costs are reduced (increasing benefits from emission cost savings). Obviously, the other welfare components more than offset the reduction in climate emission costs. One major reason for this outcome is the strong tax interaction effect (see column IE in Table 2). Referring to our theoretical analysis, the tax interaction effect IE is found to be negative due to the relatively strong impact on the fuel tax base (see the second term in brackets of Eq. (26)).49 The subsidy implies a switch from conventional fuel-powered cars to EVs which erodes the fuel tax base and, thus, lowers fuel tax revenues. Altogether, governmental tax 46 This emission rate is based on the German energy (power) mix in 2010. Taking into consideration that in the future a larger share of power generation comes from renewable resources would result in a smaller emission rate. We therefore relax this assumption later on. 47 The equivalent variation is the amount by which the income of the urban residents must be changed in the absence of the policy to yield the same level of the expected value of the maximized utility (see Eqs. (19) and (20), respectively) as if the policy had been implemented. All changes in private decisions, public budgets and markets – such as demand quantities, labor-leisure choice, travel activities, tax levels and revenues, market prices – enter (indirect) utility via household income and prices and, thus, are reflected by the equivalent variation. 48 Assuming marginal social damage costs of 70 €/tCO2. This value is recommended by Germany's Federal Environment Agency (2007) and also used by Baum et al. (2011). It should be noted, however, that there are a lot of studies suggesting smaller values. For example, based on an evaluation of prior studies, Tol (2005) finds that when combining the studies considered the mean value of estimates is 93 $/tC which is equivalent to 25 $/tCO2 (1 tC ≈ 3.67 tCO2). He also suggests that the marginal damage costs of carbon dioxide emissions are unlikely to exceed 50 $/tC (14 $/tCO2). 49 The tax interaction effect contributes negatively to social welfare (see Eq. (22)) because we consider a tax cut.

524

-16

-14

-12

-10

-8

-6

-4

Subsidy/Tax [€-cents/kwh] Δ Wem (CO2 emission reduction benefit)

8 6 4 2 0 -2 -2 0 -4 -6 -8 -10 -12 -14 -16 -18 -20

2

million €/year

-18

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

Δ Welfare

Fig. 2. Baseline simulation: welfare effects and changes in emission costs.

revenue decreases forcing the public household to raise further taxes.50 These effects are more significant than changes in emissions. Redistribution effects between landowners and city households are small because welfare gains of the landowners are very small. Obviously, subsidizing EVs by reducing power taxes hardly affects competition on the land market, thus changes in land rents are small. In addition, the total number of trips as well as aggregate VDT51 decline which contributes to the reduction in externalities. The reason is that, even though EVs are subsidized, average travel costs per km increase because the share of EVs in the ‘average’ urban car fleet goes up.52 Although the relatively high fix costs of EVs are not fully compensated by the subsidy, there is a response towards EVs on account of the relatively high willingness to pay for EVs (see the studies cited in the introduction).53 Again, referring to our theoretical exercise, we thus find by means of simulations that df/dτe N 0, where both, change in VDT as well as traffic flow run in the same direction. In fact, because df/dτe N 0, Eq. (22) reveals that, in this case, subsidizing power use reduces externality costs which improves welfare. This in turn implies that changes in externalities constitute a countervailing force on the social welfare change compared with the negative tax interaction effect.54 To sum up, the results suggest that there is no optimal subsidy. Instead there is a corner solution at the current power tax level of 0.021 €/kwh.55 Since there is no demand for EVs beyond this level, we cannot explicitly demonstrate what happens beyond τe0. Nonetheless, taxing power is suggested to be the optimal policy.56 4.3. Willingness to adopt EVs In order to examine robustness of the results, we vary different important variables. Concerning the responsiveness of households to 50 In our approach those taxes are subsumed in the lump-sum tax. The increase in the lump-sum tax lowers disposable income which in turn implies an increase in labor supply (leisure demand ceteris paribus raises with income). 51 Recall that the total number of households amounts to 1.75 million in the benchmark urban area. Hence, a reduction in aggregate VDT by 28 million km (see the last row in Table 2) actually translates to a reduction of (only) 16 km per household and year. 52 To pick up our illustration regarding car-sharing suppliers one can imagine that carsharing firms offer their EVs at higher distance based prices in order to account for higher purchasing prices of EVs. Then, on average, the urban car fleet becomes more expensive. 53 Of course, such motives should enter the utility and, thus, constitute a positive term for welfare. However, we cannot say anything on the size of this effect and its link to utility. Instead we could assume that individuals are ordered according to their preference for EVs. A small subsidy, then, shifts the border between households using EVs and the others a little in favor of EVs. Varying the size of a varies the strength of this interaction. 54 (MECt + MECem) × (−df/dτe) b 0 whereas IE N 0 when τe is increased. 55 However, some kind of subsidy might exist even if there is a positive power tax when comparing with the optimal level of the fuel tax. 56 It should be noted, however, that all in all effects are relatively small. For example, even in the extreme case of τe = −18 €-ct./kwh, the overall welfare loss of about 12 million € is rather small when referring to the household level (≈7 € per household and year).

the subsidy we raise the parameter a – i.e. the responsiveness parameter to the adoption of EVs – to 0.5. This level implies that a subsidy of about 0.05 €/kwh generates a demand of approximately one million EVs. This response is very strong and we take it to be the upper ceiling of responsiveness.57 Fig. 3 displays the results, where the solid curves represent the case with the baseline responsiveness and the dashed curves the case with a higher responsiveness. If households are more willing to switch to EVs and if they get a subsidy the aggregate welfare loss is much higher than in the case of a lower willingness to switch. This is the case even though emission costs decrease more (there are stronger savings in social costs of carbon dioxide). This outcome is the stronger the larger the subsidy. This result is surprising. It makes clear that the social gain stemming from emission cost saving is a minor point in comparison to tax interaction effects, congestion costs, changes in travel costs and redistribution issues. Table 3 shows that the effects concerning the total number of trips, distance traveled, congestion and overall emissions is stronger if the response of households to the power tax is stronger (a = 0.5). Because households are now assumed to have a higher willingness to adopt EVs or they are more willing to accept higher average costs of operating their vehicle, respectively, on average traveling by car is getting more expensive (despite the reduction in the marginal congestion costs) because of a higher share of EVs in the urban car fleet. This reduces automobile travel demand (e.g. the length of trips) and explains the stronger reduction in car kilometers (and so the larger emission cost savings). These results suggest that policies aiming at raising the willingness to switch to EVs might help to achieve emission goals by rising the number of EVs but go along with considerable costs in the presence of EV subsidies. 4.4. Fix costs of EVs In addition to a stronger responsiveness costs could be crucial and might influence the results. Recall that our baseline assumption on average fix costs is 0.50 €/km compared with 0.30 €/km for conventional cars. However, raising the share of EV might allow to exploit economies of scale in production, provision and using EVs. Technical progress might also lower costs of producing or using EVs (see Dinger et al., 2010; Weiss et al., 2012). Therefore we assume that average costs per VDT of EVs decline to 0.35 €/km and, thus, almost achieve the level of conventional fuel-powered cars. Comparing the solid curves of the baseline case with the dashed curves in the case of lower fix costs of EVs, Fig. 4 shows that even a reduction of fix costs of EVs of around one third does hardly change the general findings. But there are changes in some important effects. Table 4 shows the results in the case of lower fix costs while Table 2 displays the results with higher fix costs (baseline). The reduction in average EV vehicle costs results in an increase in aggregate distance traveled by car. This is the outcome of relocation decisions of the households moving (on average) farther away from their working place or choosing a work location farther away from their residences.58 The increase in car kilometer driven causes congestion and – associated with congestion – the fuel consumption and emission externality to increase. This causes an increase in emissions per VDT that counteracts the reduction in emissions due to stronger diffusion of EVs (fewer fuel-powered cars). This explains why the overall benefit ΔWem stemming from the reduction in emissions is smaller than in the baseline case. Although travel costs are considerably smaller, negative welfare effects hardly 57 We conducted several additional simulations differing even more in the degree of responsiveness and, thus, the strength of EV diffusion. Generally, we find that the results found here do not qualitatively change, only the magnitude of the impacts is affected (the stronger the diffusion or the responsiveness to the policy considered, the stronger the effects). 58 Relocations affect the land market causing welfare losses of absentee landowners that almost fully neutralize benefits from the reduction of social carbon emission costs. This also shows that subsidies to EVs might have visible spatial effects on the city.

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

525

τe

ΔWem

ΔWA

IE

Δf trip

€-ct./kwh Mio. € Mio. € Mio. € Mio. 0 −2 −4 −6 −8 −10 −12 −14 −16 −18

0.3 0.5 0.8 1.0 1.3 1.5 1.8 2.0 2.3 2.5

0 −1 −1 −1 −2 −3 −4 −4 −5 −6

0.1 0.1 0.1 0.1 0.1 0.2 0.2 0.1 0.1 0.1

−0.010 −0.020 −0.029 −0.038 −0.047 −0.056 −0.065 −0.073 −0.081 −0.089

Δf δ

ΔMEC t ΔMEC g ΔMEC em

Mio. km %

%

%

−4 −7 −10 −13 −16 −18 −21 −23 −26 −28

−0.6 −1.2 −1.8 −2.4 −2.7 −3.2 −3.8 −4.1 −4.7 −5.0

−0.4 −0.6 −0.8 −1.1 −1.4 −1.7 −1.9 −2.1 −2.3 −2.5

−0.1 −0.3 −0.4 −0.5 −0.6 −0.7 −0.8 −0.9 −0.9 −1.0

-18

MEC t/MEC g/MEC em: Marginal congestion/fuel/emission externality per VDT of car driving. f trip: One-way trips [commuting + shopping] f δ: Total VDT.

change in comparison to the baseline case. Granting power subsidies now implies a stronger reduction of fuel tax revenues (compared with the baseline) because the relative cost disadvantage of EVs is now smaller, forcing households to use EVs more extensively. Consequently, taxing EVs is optimal even if producing these cars requires less resources and costs decline. Although lowering costs reduces welfare losses, granting power subsidies still causes social welfare to decrease. Note that one can hardly imagine that costs of EVs will decline more than assumed here. Considering technological progress regarding conventional fuel-powered cars too, it becomes even less likely that relative costs improve in favor of EVs in a sufficient way. Furthermore, we do not yet consider time costs or inconveniences for loading or investments required in the loading infrastructure. Hence, there is an upward bias in our calculations that actually works in favor of EVs. 4.5. Emissions of EVs Until now we have not discussed emissions and corresponding assumptions. Emissions can also be changed by technological progress or by changes in upstream emissions. Given these and other uncertainties we construct the extreme event that switching to EVs does not cause additional emission costs. Further rationales behind this assumption are that since power production is subject to emission trading in the EU increasing electric power use for passenger travel does not raise aggregate emissions if trading permits are restricted (see Massiani and Weinmann, 2012), or that power generation exclusively comes from renewable resources. Fig. 5 shows the corresponding results. The dashed curves represent the case of zero emission costs of EVs. Although there is a decline in emissions costs (benefits from emission cost reduction due to EV diffusion), additional savings are small 16

-18

-16

-14

-12

-10

-8

-6

-4

-2

0 -8 -16

0

million € /year

8

2

-24 -32 -40 -48

Subsidy/Tax [€-cents/kwh] Baseline: Higher responsiveness

ΔWem (CO2 emission reduction benefit)

-56 Δ Welfare

Fig. 3. Welfare effects and changes in emission costs (baseline and higher responsiveness).

-16

-14

-12

-10

-8

-6

-4

Subsidy/Tax [€-cents/kwh] Baseline: Lower cost

8 6 4 2 0 -2 -2 0 -4 -6 -8 -10 -12 -14 -16 -18 -20

Δ Wem (CO2 emission reduction benefit)

2

million €/year

Table 2 Results of the baseline simulation.

Δ Welfare

Fig. 4. Welfare effects and changes in emission costs (EV's average costs per km decline).

Table 3 Results of the simulation (higher willingness to adopt EVs). τe

ΔWem

IE

ΔWA

Δf trip

€-ct./kwh Mio. € Mio. € Mio. € Mio. 0 −2 −4 −6 −8 −10 −12 −14 −16 −18

1.3 2.5 3.6 4.8 6.0 7.1 8.3 9.4 10.6 11.7

−1 −3 −5 −7 −10 −13 −17 −21 −25 −30

0.2 0.4 0.5 0.6 0.7 0.7 0.7 0.7 0.6 0.5

−0.047 −0.092 −0.136 −0.179 −0.222 −0.265 −0.307 −0.348 −0.389 −0.429

Δf δ

ΔMEC t ΔMEC g ΔMEC em

Mio. km %

%

%

−16 −31 −46 −60 −74 −87 −100 −112 −124 −135

−0.1 −1.2 −1.8 −2.4 −2.7 −3.2 −3.8 −4.1 −4.7 −5.0

−0.4 −0.6 −0.8 −1.1 −1.4 −1.7 −1.9 −2.1 −2.3 −2.5

−0.1 −1.2 −1.7 −2.2 −2.7 −3.3 −3.7 −4.1 −4.6 −5.0

MEC t/MEC g/MEC em: Marginal congestion/fuel/emission externality per VDT of car driving. f trip: One-way trips [commuting + shopping] f δ: Total VDT.

compared with the baseline. Because emission costs contribute additively to welfare other decisions are not affected. Hence, adverse welfare effects of subsidies to power use do not change. This implies that even assumptions on emissions in favor of EVs do not change our general finding that granting subsidies to power use reduces aggregate social welfare.59

5. Conclusions In this paper we have analytically derived the components of the optimal power tax on the use of electric vehicles. Externality costs, tax interaction effects, changes in travel costs (or differences in travel costs between conventional fuel-powered cars and EVs) and redistribution effects determine the optimal subsidy rate. The simulations show that particularly the (negative) tax interaction effect more than offset the reduction in carbon emission costs. Even in cases where congestion costs decline the overall effect of the subsidy is negative due to the strong tax interaction effect. Accordingly, we deduce that using electric power for car traveling shall be taxed, not subsidized. Different sensitivity analyses show that our main finding is surprisingly robust with respect to strong changes in the willingness to adopt EVs, changes in costs of using EVs and even if we assume that carbon 59 Recall that, even though our assumption on marginal damage costs of CO2 (70 €/tCO2) corresponds with recommendations of the Federal Environment Agency in Germany, they are likely to be at the upper range of estimates (see Tol, 2005). In fact, the benefits from reduced emission costs when granting subsidies to EVs we found would even be smaller when assuming lower marginal damage costs of CO2.

526

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

τe

ΔWem

IE

ΔWA

Δf trip

€-ct./kwh Mio. € Mio. € Mio. € Mio. 0 −2 −4 −6 −8 −10 −12 −14 −16 −18

0.2 0.3 0.5 0.7 0.8 1.0 1.1 1.3 1.4 1.6

−1 −2 −3 −4 −5 −7 −8 −10 −11 −13

−0.1 −0.2 −0.4 −0.5 −0.6 −0.8 −0.9 −1.1 −1.3 −1.4

Δf δ

ΔMEC t ΔMEC g ΔMEC em

Mio. km %

−0.002 0 −0.004 1 −0.005 2 −0.006 3 −0.007 4 −0.008 6 −0.009 7 −0.010 9 −0.010 10 −0.010 12

0.0 0.1 0.1 0.2 0.2 0.3 0.4 0.5 0.5 0.6

%

%

0.3 0.3 0.3 0.3 0.3 0.6 0.6 0.6 0.6 0.9

0.0 0.1 0.2 0.3 0.4 0.4 0.4 0.4 0.4 0.5

MEC t/MEC g/MEC em: Marginal congestion/fuel/emission externality per VDT of car driving. f trip: One-way trips [commuting + shopping] f δ: Total VDT.

emissions of using EVs and of upstream power production are zero. The social costs of EVs exceed the benefits from the reduction in CO2 emissions in all these cases. Besides, one should additionally keep in mind that our simulations are based on two further assumptions that basically favor the use of EVs: subsidies are only financed by lump-sum taxes and additional costs of using EVs such as costs of a load infrastructure are not considered. Hence, the overall negative outcome of subsidizing EVs by granting power subsidies is likely to be even understated. In contrast, if fuel taxes (and other taxes related to fuel consumption) on combustion engine vehicles are considerably smaller our verdict might be too strong. In such a case the strong negative tax interaction effect (the main source of the welfare loss) mainly caused by lower fuel tax revenue will be much smaller.60 In such a case, power subsidies could make a contribution to an improvement in welfare. All in all, however, our finding reveals that subsidizing EVs to achieve a reduction of passenger travel related carbon emissions is inefficient and welfare diminishing. Therefore policy shall be much less optimistic concerning the use of EVs for mitigating climate change (see also Massiani and Weinmann (2012) for a corresponding discussion). There are other much more efficient instruments available: e.g. congestion tolls and emission taxes, raising energy taxes to finance public transport,61 etc. (see Tscharaktschiew and Hirte, 2010a, 2012). Concerning congestion pricing, we even looked into a combination of a Pigouvian congestion toll and a power tax subsidy to EVs just to see whether diffusion improves efficiency in the case that one of the externalities is internalized in an appropriate way. Then, subsidies to EVs should only be an instrument to internalize the emission externality. But even in that case the finding stays the same. In the presence of a Pigouvian congestion toll 60 Actually, energy (fuel) taxes are an important source of tax revenues for public authorities in Germany (energy tax revenue amounts to about 8% of public tax revenue in 2010 while income and sales tax revenues together account for about 70%). According to our results subsidizing EVs causes a strong negative tax interaction effect mainly due to a decline in energy tax revenue that is about twice as strong as the change in income plus sales tax revenue. This is supposed to apply to most European countries and some other countries, e.g. Japan, where fuel taxes are used to generate public revenue (see GIZ, 2012). Because tax revenues from income and sales taxes on general consumption run in the opposite direction compared to the decline in fuel tax revenue, the negative tax interaction effect might thus be much smaller in countries where fuel taxes including specific sales or other taxes are less important in comparison to income and consumption taxes. However, one should have in mind that the increase in labor supply results from the raise of the lumpsum tax used to finance power subsidies but also used to compensate for the loss in fuel tax revenue. Hence, with lower fuel taxes labor gains in income and sales tax revenue are much smaller too. Moreover, we do not consider other taxes, such as the property tax. This might also contribute to tax interaction effects because of land rent changes. 61 According to Storchmann (2001) higher fuel taxes might increase the deficit of public transport. Public transport ridership will gain from the price induced modal shift (reduction in distance driven by cars) mainly in the peak load traffic segment. Since this traffic is usually characterized by above-average marginal costs and below-average marginal revenues the additional peak load traffic will increase the deficit of public transport. Therefore, recycling additional fuel tax revenues (at least partly) in favor of public transport could be an opportunity to counterbalance an increasing deficit and, thus, to ensure the potential overall efficiency of such a policy.

-18

-16

-14

-12

-10

-8

-6

-4

Subsidy/Tax [€-cents/kwh] Baseline: Zero emission EVs

8 6 4 2 0 -2 -2 0 -4 -6 -8 -10 -12 -14 -16 -18 -20

Δ Wem (CO2 emission reduction benefit)

2

million €/year

Table 4 Results of the simulation (lower fix costs of EVs).

Δ Welfare

Fig. 5. Welfare effects and changes in emission costs in the case of zero emissions of EVs.

the power tax subsidy lowers aggregate welfare, too. Positive effects of EV diffusion are less extensive because congestion charges also internalize some of the externality costs associated with driving conventional fuel cars. However, we cannot exclude that other positive effects of EVs we have not considered here shift the results in favor of EVs. For example, rising oil prices which make conventional fuel more expensive could foster the demand for EVs (Diamond, 2009) by reducing the current cost disadvantage of EVs (see Dijk et al., 2013). There might be an increase in land supply due to closing of some filling stations. However, the additional available land area is likely to be rather small because only a few filling stations must be closed (the share of EVs is small even if the current policy target will be reached) and some of them might be converted to recharging stations needed to provide a sufficiently dense network for people who cannot charge at home or who are traveling longer distances. Moreover, EVs produce less noise and less local pollution. While the net effect of less noise is ambiguous because it raises safety issues, the second aspect of reduced air pollutants (see Sovacool, 2010) might be the most important benefit from e-mobility.62 Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.eneco.2013.08.001. References ADAC (Allgemeiner Deutscher Automobil-Club), 2012. Der Elektroantrieb. Anas, A., 2012. The optimal pricing, finance and supply of urban transportation in general equilibrium: a theoretical exposition. Econ. Transp. 1, 64–76. Anas, A., Rhee, H.-J., 2006. Curbing excess sprawl with congestion tolls and urban boundaries. Reg. Sci. Urban Econ. 36, 510–541. Anas, A., Xu, R., 1999. Congestion, land use, and job dispersion: a general equilibrium model. J. Urban Econ. 45, 451–473. Arbeitskreis Volkswirtschaftliche Gesamtrechnungen der Länder, 2009. Volkswirtschaftliche Gesamtrechnungen der Länder: Zusammenhänge, Bedeutung und Ergebnisse. Axsen, J., Kurani, K.S., 2009. Early US market for plug-in hybrid electric vehicles: anticipating consumer recharge potential and design priorities. Transp. Res. Rec. J. Transp. Res. Board 2139, 64–72. Axsen, J., Kurani, K.S., 2012. Who can recharge a plug-in electric vehicle at home? Transp. Res. Part D: Transp. Environ. 17, 349–353. 62 However, improvements in fuel economy and technological progress regarding conventional cars might have a slightly dampening effect on the relevance of local air pollution in the future (see e.g. Federal Environment Agency, 2012). Even in the status quo social costs of local air pollution are suggested to be in the order of climate change costs (social costs of carbon dioxide). For example, assuming CO2 emissions of 0.00017 tCO2/ km (170 gCO2/km) and marginal social costs of 70 €/tCO2 externality costs then amount to about 0.01 €/km which is almost the same the empirical literature estimates for car related local air pollution (see e.g. Baum et al., 2011; Parry and Small, 2005). As a consequence, accounting for lower overall air pollution costs due to a stronger diffusion of EVs is unlikely to qualitatively change the main finding of the present analyses.

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528 Baum, H., Dobberstein, J., Schuler, B., 2011. Kosten-Nutzen Analyse der Elektromobilität. Kölner Diskussionsbeiträge zur Verkehrswissenschaft. University of Cologne. Bosetti, V., Longden, T., 2013. Light duty vehicle transportation and global climate policy: the importance of electric drive vehicles. Energy Policy 58, 209–219. Brady, J., O'Mahony, M., 2011. Travel to work in Dublin. The potential impacts of electric vehicles on climate change and urban air quality. Transp. Res. Part D: Transp. Environ. 16, 188–193. Bundesregierung, 2009. German Federal Government's National Electromobility Development Plan. Carley, S., Krause, R.M., Lane, B.W., Graham, J.D., 2013. Intent to purchase a plug-in electric vehicle: a survey of early impressions in large US cites. Transp. Res. Part D: Transp. Environ. 18, 39–45. Carlsson, F., Johannson-Stenman, O., 2003. Costs and benefits of electric vehicles — A 2010 perspective. J. Transp. Econ. Policy 37, 1–28. Christensen, L., Kaplan, S., Jensen, T.C., Røpke, S., Olsen, A., 2012. Towards sustainable regions: the spatial distribution of electric vehicles' recharging stations from a socio-economic perspective. Paper presented at the 9th World Congress of Regional Science Association International, Timisoara, Romania. Daziano, R.A., Chiew, E., 2012. Electric vehicles rising from the dead: data needs for forecasting consumer response toward sustainable energy sources in personal transportation. Energy Policy 51, 876–894. De Borger, B., Van Dender, K., 2003. Transport tax reform, commuting, and endogenous values of time. J. Urban Econ. 53, 510–530. Delang, C.O., Cheng, W.-T., 2012. Consumers' attitudes towards electric cars: a case study of Hong Kong. Transp. Res. Part D: Transp. Environ. 17, 492–494. Delucchi, M., 2005. A Multi-Country Analysis of Lifecycle Emissions from Transportation Fuels and Motor Vehicles. Davis (WP UCD-ITS-RR-05-10). Institute of Transportation Studies, University of California. Delucchi, M., Lipman, T.E., 2001. An analysis of the retail and lifecycle of battery-powered electric vehicles. Transp. Res. Part D: Transp. Environ. 6, 371–404. Delucchi, M.A., Murphy, J.J., 2008. How large are tax subsidies to motor-vehicle users in the US? Transp. Policy 15, 196–208. Department for Transport, 2009. Low Carbon Transport: A Greener Future — A Carbon Reduction Strategy for Transport. Department for Transport (July 2009). Department for Transport, 2012. Annual Report and Accounts 2011–12. Diamond, D., 2009. The impact of government incentives for hybrid-electric vehicles: evidence from US states. Energy Policy 37, 972–983. Dijk, M., Orsato, R.J., Kemp, R., 2013. The emergence of an electric mobility trajectory. Energy Policy 52, 135–145. Dinger, A., Ripley, M., Mosquet, X., Rabl, M., Rizoulis, D., Russo, M., Sticher, G., 2010. Batteries for Electric Cars: Challenges, Opportunities, and the Outlook to 2020. Technical Report. The Boston Consulting Group. Doucette, R.T., McCulloch, M.D., 2011. Modeling the CO2 emissions from battery electric vehicles given the power generation mixes of different countries. Energy Policy 39, 803–811. Ehsani, M., Gao, Y., Gay, S.E., Emadi, A., 2010. Modern Electric, Hybrid Electric and Fuel Cell Vehicles — Fundamentals, Theory, and Design, Second edition. CRC Press. Ewing, G., Sarigöllü, E., 2000. Assessing consumer preferences for clean-fuel vehicles: a discrete choice experiment. J. Public Policy Mark. 19, 106–118. Federal Environment Agency (Umweltbundesamt), 2007. Awareness of external costs – helps to improve environmental protection: The methodological convention for estimates of environmental externalities (the energy and transport sectors as examples). Federal Environment Agency (Umweltbundesamt), 2011. FG I 2.5., Stand: Mä rz 2011. Federal Environment Agency (Umweltbundesamt), 2012. Nationale Trendtabellen f ür die deutsche Berichterstattung atmosphärischer Emissionen (Emissionsentwicklung 1990–2010, klassische Luftschadstoffe). Federal Ministry of Transport, Building and Urban Affairs, 2004. MID - Mobility in Germany 2002 — Results. (http://www.mobilitaet-in-deutschland.de.). Federal Ministry of Transport, Building and Urban Affairs, 2009. Stadtentwicklungsbericht 2008. Federal Statistical Office, 2005. Leben und Arbeiten in Deutschland: Ergebnisse des Mikrozensus 2004. Federal Statistical Office, 2008a. Statistical Yearbook 2008 for the Federal Republic of Germany. Federal Statistical Office, 2008b. Pressemitteilung Nr.305 vom 25.08.2008. Federal Statistical Office, 2009. Statistical Yearbook 2009 for the Federal Republic of Germany. Funk, K., Rabl, A., 1999. Electric versus conventional vehicles: social costs and benefits in France. Transp. Res. Part D: Transp. Environ. 4, 397–411. Gardner, B., Abraham, C., 2007. What drives car use? A grounded theory analysis of commuters' reasons for driving. Transport. Res. Part F: Traffic Psychol. Behav. 10, 187–200. GIZ (Deutsche Gesellschaft für Internationale Zusammenarbeit), 2012. International Fuel Prices 2010/2011. Goodwin, P.B., 1992. A review of new demand elasticities with special reference to short and long run effects of price changes. J. Transp. Econ. Policy 26, 155–163. Goodwin, P., Dargay, J., Hanly, M., 2004. Elasticities of road traffic and fuel consumption with respect to price and income: a review. Transp. Rev. 24, 275–292. Graham, D.J., Glaister, S., 2004. Road traffic demand elasticity estimates: a review. Transp. Rev. 24, 261–274. Graham-Rowe, E., Gardner, B., Abraham, C., Skippon, S., Dittmar, H., Hutchins, R., Stannard, J., 2012. Mainstream consumers driving plug-in battery-electric and plugin hybrid electric cars: a qualitative analysis of responses and evaluations. Transp. Res. Part A: Policy Pract. 46, 140–153.

527

Grazi, F., van den Bergh, J.C.J.M., 2008. Spatial organization, transport, and climate change: comparing instruments of spatial planning and policy. Ecol. Econ. 67, 630–639. Hahn, R.W., 1995. Choosing among fuels and technologies for cleaning up the air. J. Policy Anal. Manage. 14, 532–554. Harvey, L.D.D., 2013. Global climate-oriented transportation scenarios. Energy Policy 54, 87–103. Hawkins, T.R., Singh, B., Majeau-Bettez, G., Strømman, A.H., 2013. Comparative environmental life cycle assessment of conventional and electric vehicles. J. Ind. Ecol. 17, 53–64. Hymel, K.M., Small, K.A., Van Dender, K., 2010. Induced demand and rebound effects in road transport. Transp. Res. Part B: Methodol. 44, 1220–1241. IAB (Institute for Employment Research, Institut für Arbeitsmarkt - und Berufsforschung), 2005. Durchschnittliche Arbeitszeit und ihre Komponenten in Deutschland. Karplus, V.J., Paltsev, S., Reilly, J.M., 2010. Prospects for plug-in hybrid electric vehicles in the United States and Japan: a general equilibrium analysis. Transp. Res. Part A: Policy Pract. 44, 620–641. Kazimi, C., 1997a. Valuing alternative-fuel vehicles in southern California. Am. Econ. Rev. 87, 265–271. Kazimi, C., 1997b. Evaluating the environmental impact of alternative-fuel vehicles. J. Environ. Econ. Manag. 33, 136–185. King, D., Inderwildi, O., Carey, C., Santos, G., Yan, X., Behrendt, H., Holdway, A., Maconi, L., Owen, N., Shirvani, T., Teytelboym, A., 2010. Future of Mobility Roadmap: Ways to Reduce Emissions While Keeping Mobile. Smith School of Enterprise and the Environment, University of Oxford, England. Klöckner, C.A., Nayum, A., Mehmetoglu, M., 2013. Positive and negative spillover effects from electric car purchase to car use. Transp. Res. Part D: Transp. Environ. 21, 32–38. Knittel, C.R., 2012. Reducing petroleum consumption from transportation. J. Econ. Perspect. 26, 93–118. Kurani, K.S., Turrentine, T., Sperling, D., 1996. Testing electric vehicle demand in ‘hybrid households’ using a reflexive survey. Transp. Res. Part D: Transp. Environ. 1, 131–150. Kyle, P., Kim, S.H., 2011. Long-term implications of alternative light-duty vehicle technologies for global greenhouse gas emissions and primary energy demands. Energy Policy 39, 3012–3024. Lave, L.B., MacLean, H.L., 2002. An environmental-economic evaluation of hybrid electric vehicles: Toyota's Prius vs. its conventional internal combustion engine Corolla. Transp. Res. Part D: Transp. Environ. 7, 155–162. Lave, L.B., Hendrickson, C.T., McMichael, F.C., 1995. Environmental implications of electric cars. Science 268, 993–995. Lieven, T., Mühlmeier, S., Henkel, S., Waller, J.F., 2011. Who will buy electric cars? An empirical study in Germany. Transp. Res. Part D: Transp. Environ. 16, 236–243. Lorf, C., Martínez-Botas, R.F., Howey, D.A., Lytton, L., Cussons, B., 2013. Comparative analysis of the energy consumption and CO2 emissions of 40 electric, plug-in hybrid electric, hybrid electric and internal combustion engine vehicles. Transp. Res. Part D: Transp. Environ. 23, 12–19. Massiani, J., 2012. Using stated preferences to forecast alternative fuel vehicles market diffusion. Ital. J. Reg. Sci. Sci. Reg. 11, 93–122. Massiani, J., Radeke, J., 2013. Cost–benefit analysis of policies for the development of electric vehicles in Germany: methods and results. Working Paper No. 02/WP/20 13. Department of Economics, Ca' Foscari University of Venice. Massiani, J., Weinmann, J., 2012. Estimating electric car's emissions in Germany: an analysis through a pivotal marginal method and comparison with other methods. Econ. Policy Energy Environ. 2, 131–155. Mühlenhoff, J., 2011. Kosten und Preise für Strom. Renews Special 52/2011 (Agentur für Erneuerbare Energien). Musti, S., Kockelman, K.M., 2011. Evolution of the household vehicle fleet: anticipating fleet composition, PHEV adoption and GHG emissions in Austin, Texas. Transp. Res. Part A: Policy Pract. 45, 707–720. Nanaki, E.A., Koroneos, C.J., 2013. Comparative economic and environmental analysis of conventional, hybrid and electric vehicles — the case study of Greece. J. Clean. Prod. 53, 261–266. OECD (Organisation for Economic Co-operation and Development), 2005. Environmentally Harmful Subsidies: Challenges for Reform. Öko-Institut, 2011. OPTUM: Optimierung der Umweltentlastungspotenziale von Elektrofahrzeugen - Integrierte Betrachtung von Fahrzeugnutzung und Energiewirtschaft. Schlussbericht im Rahmen der “Förderung von Forschung und Entwicklung im Bereich der Elektromobilität” des Bundesministeriums für Umwelt, Naturschutz und Reaktorsicherheit (BMU). Oum, T.H., Waters II, W.G., Yong, J.S., 1992. Concepts of price elasticities of transport demand and recent empirical estimates: an interpretative survey. J. Transp. Econ. Policy 23, 163–187. Ozaki, R., Sevastyanova, K., 2011. Going hybrid: an analysis of consumer purchase motivations. Energy Policy 39, 2217–2227. Parry, I.W.H., Small, K.A., 2005. Does Britain or the United States have the right gasoline tax? Am. Econ. Rev. 95, 1276–1289. Peterson, S.B., Michalek, J.J., 2013. Cost-effectiveness of plug-in hybrid electric vehicle battery capacity and charging infrastructure investment for reducing US gasoline consumption. Energy Policy 52, 429–438. Prud'homme, R., Koning, M., 2012. Electric vehicles: a tentative economic and environmental evaluation. Transp. Policy 23, 60–69. Rhee, Hyok-Joo, 2012, Zoning in Cities with Traffic Congestion and Agglomeration Economies, mimeo, Seoul. Shin, J., Hong, J., Jeong, G., Lee, J., 2012. Impact of electric vehicles on existing car usage: a mixed multiple discrete-continuous extreme value model approach. Transp. Res. Part D: Transp. Environ. 17, 138–144.

528

G. Hirte, S. Tscharaktschiew / Energy Economics 40 (2013) 515–528

Siedentop, S., 2007. Auswirkungen der Beschäftigungssuburbanisierung auf den Berufsverkehr - Führt die Suburbanisierung der Arbeitsplätze zu weniger Verkehr? Informationen zur Raumentwicklung Heft 2/3 2007. The Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR). Small, K.A., Rosen, H.S., 1981. Applied welfare economics with discrete choice models. Econometrica 49, 105–130. Small, K.A., Verhoef, E.T., 2007. The Economics of Urban Transportation. Routledge. Smith, W.J., 2010. Can EV (electric vehicles) address Ireland's CO2 emissions from transport? Energy 35, 4514–4521. Sovacool, B., 2010. A transition to plug-in hybrid electric vehicles (PHEVS): why public health professionals must care. J. Epidemiol. Community Health 64, 185–187. Stadt Frankfurt/Main, 2009. Frankfurt am Main im Städtevergleich - Qualifikationsstruktur des Arbeitsangebotes 2007. Statistik Aktuell Nr. 14/2009. Statistical Office Munich, 2009. Die Gliederung des Stadtgebietes nach Bodennutzungsarten in ha. Steiner, V., Cludius, J., 2010. Ökosteuer hat zu geringerer Umweltbelastung des Verkehrs beigetragen. DIW Wochenbericht 13–14. Storchmann, K.-H., 2001. The impact of fuel taxes on public transport — an empirical assessment for Germany. Transp. Policy 8, 19–28. Thiel, C., Perujo, A., Mercier, A., 2010. Cost and CO2 aspects of future vehicle options in Europe under new energy policy scenarios. Energy Policy 38, 7142–7151.

Tol, R.S.J., 2005. The marginal damage costs of carbon dioxide emissions: an assessment of the uncertainties. Energy Policy 33, 2064–2074. Train, K.E., 2003. Discrete Choice Methods with Simulation. Cambridge University Press. Tscharaktschiew, S., Hirte, G., 2010a. The drawbacks and opportunities of carbon charges in metropolitan areas — A spatial general equilibrium approach. Ecol. Econ. 70, 339–357. Tscharaktschiew, S., Hirte, G., 2010b. How does the household structure shape the urban economy? Reg. Sci. Urban Econ. 40, 498–516. Tscharaktschiew, S., Hirte, G., 2012. Should subsidies to urban passenger transport be increased? A spatial CGE analysis for a German metropolitan area. Transp. Res. Part A: Policy Pract. 46, 285–309. Wang, M.Q., 1997. Mobile source emission control cost-effectiveness: issues, uncertainties, and results. Transp. Res. Part D: Transp. Environ. 2, 43–56. Weiss, M., Patel, M.K., Junginger, M., Perujo, A., Bonnel, P., van Grootveld, G., 2012. On the electrification of road transport — Learning rates and price forecasts for hybridelectric and battery-electric vehicles. Energy Policy 48, 374–393. Werber, M., Fischer, M., Schwartz, P.V., 2009. Batteries: lower cost than gasoline? Energy Policy 37, 2465–2468. Windecker, A., Ruder, A., 2013. Fuel economy, cost, and greenhouse gas results for alternative fuel vehicles in 2011. Transp. Res. Part D: Transp. Environ. 23, 34–40.