2008/1. Evaluating the impact of average cost based contracts on the industrial sector in the European emission trading scheme

Transcription

1 2008/1 Evauating the impact of average cost based contracts on the industria sector in the European emission trading scheme Giorgia Oggioni and Yves Smeers

2 CORE DISCUSSION PAPER 2008/1 Evauating the impact of average cost based contracts on the industria sector in the European emission trading scheme Giorgia OGGIONI 1 and Yves SMEERS 2 January 2008 Abstract The inception of the Emission Trading System in Europe (EU-ETS) has made power price more expensive. This affects the competitiveness of eectricity intensive industria consumers and may force them to eave Europe. Taking up of a proposa of the industria sector, we expore the possibe appication of specia contracts, based on the average cost pricing system, which woud mitigate the impact of CO 2 cost on their eectricity price. The mode supposes fixed generation capacities. A companion paper treats the case with capacity expansion. We first consider a reference mode representing a perfecty competitive market where a consumers (househods and industries) are price-takers and buy eectricity at the short-run margina cost. We then change the market design assuming that arge industria consumers pay power either at a singe or at a noda average cost price. The anaysis of these probems is conducted with simuation modes appied to the Northwestern European market. The equiibrium modes deveoped are impemented in the GAMS environment. Keywords: average cost pricing, compementarity conditions, EU-ETS, Northwestern Europe market. AMS Cassification: 46N10, 90C33, 91B24, 91B42 1 Universita degi Studi di Bergamo, Itay and CORE, Université cathoique de Louvain, Begium. E- mai: giorgia.oggioni@unibg.it 2 Tractebe, Department of Mathematica Engineering and CORE, Université cathoique de Louvain, Begium. E-mai: yves.smeers@ucouvain.be. The author is aso member of ECORE, the newy created association between CORE and ECARES. This work has been supported by PRIN 2006, Generaized monotonicity: modes and appications, Itay. Nationa Responsibe: Prof. E. Aevi. We wish to thank Thierry Bréchet and a the attendances of the Environmenta Workshops for their comments and advices. This paper presents research resuts of the Begian Program on Interuniversity Poes of Attraction initiated by the Begian State, Prime Minister's Office, Science Poicy Programming. The scientific responsibiity is assumed by the authors.

3 1 Introduction Starting from January 2005, Member States of the European Union have introduced the European Emission Trading Scheme (EU-ETS hereafter) in order to curb CO 2 emissions and tacke cimate change. It is widey admitted today that the functioning of the EU-ETS shoud be improved. The degradation of the competitiveness of part of the European industry is progressivey emerging as one of the possibe consequences of the combination of the EU-ETS and the restructuring of the eectricity sector. Eectricity intensive industria users strongy encourage the review of the ETS to mitigate its negative effects on their cost baance. This negative impact is twofod: industries need not ony to abate emissions; they have aso to pay a higher eectricity price. This second effect resuts from the practice of the power sector to charge the CO 2 costs into eectricity prices. Industria sectors can adapt to these cost increases by accepting a reduction of their profits or by increasing their product prices. They expain that this endangers their competitiveness on internationa markets. This wi eventuay impy a oss of saes and market share, especiay for companies extremey exposed to foreign competition. Some studies (see [2, 9, 11, 13]) show that the industria sectors exposure to the EU-ETS depends 1) on the industry s abiity to pass the extra carbon cost to consumers, 2) on the internationa trade openness, 3) on the energy intensity and the possibiity to abate carbon and, ast, 4) on the aowance aocation method 1. These factors vary with the industria sector considered. Large industria consumers argue that this probem may force some of them to move part of their production activities in extra-community countries where emission poicies are ess restrictive. This entais a serious oss of wefare for the European countries with an additiona environmenta damage due to more enient norms in non-european countries (carbon eakage 2 ). Large industria consumers further caim that power producers exercise market power in order to maintain high eectricity prices, a phenomenon that exacerbates the threat on competitiveness. A recent move from giant stee company Arceor Mitta iustrates this CO 2 probem. The company is refusing to re-open its bast furnace in Liège (Begium) if it does not receive the necessary permits free. Lakshmi Mitta, the CEO of the company, asks Begian pubic authorities to provide the carbon permits needed, otherwise he wi move company s stee production outside Europe. In an interview given to the Begian newspaperl Echo, the CEO argues we must address the probem gobay and not penaize the sector in Europe. The risk is a reocation of stee production to areas without CO 2 constraints. He simpy requires good conditions before investing in Europe 3. Under these conditions, industries expicity ask for either specia contracts whereby they can procure eectricity at the average cost or specia reguation that woud prevent generators to embed the emission aowance price in the margina cost of eectricity. 1 The aowance aocation method and the amount of aowances distributed affect operationa costs. For instance, grandfathering essens the cost imposed by the ETS system. 2 Carbon eakage strongy depends on pants re-ocation. It measures the compensation of an industry s greenhouse gas reduction by an increase in the same industry s emissions in regions without carbon constraint. 3 Sources: L Echo, 5/12/2007. Avaiabe at sans quotas CO2 je ne reance pas Liege

4 In this paper, we expore the first of these proposas, i.e. the possibe introduction of specia contracts based on the average cost pricing system, and we test to which extend they woud hea some of the difficuties of the industry. We iustrate this issue by the means of sma market simuations appied to a simpified eectricity network of the Northwestern Europe (France, Germany, Begium and The Netherands). With the aim of investigating the impications of the carbon market on eectricity prices and demand, we first simuate a perfecty competitive market where a consumers (househods and industries) are price-takers and are assumed to pay an identica eectricity price based on margina costs. This represents the reference case. We then mode two innovative pricing mechanisms. Assuming that industria consumers can harness the financia means to acquire and operate part of base oad power pants, we consider an aternative contractua organization whereby they pay the fu cost of those instaations and, hence, face average cost based prices. This new pricing poicy is tested in two forms: there is a singe average cost price in the first case; noda average cost prices are introduced in the second one. Both scenarios impy segmenting the market and sharing the existing generation capacity between the two consumer groups in the sense that part of the generation capacity is fuy dedicated to the arge industria consumers. We find that the impact of average cost based prices on industria consumers differs depending on their ocation and the price arrangement (singe or noda) appied. Moreover, the technoogy mix use to produce eectricity pays a decisive roe in industria power prices. 2 Input Data and Assumptions We conduct the anaysis on a styized representation of the Northwestern Europe of eectricity market, where eectricity is provided by eight generators 4 and a fringe, which assembes the remaining sma generators. Generating companies are supposed to produce eectricity by operating eight different technoogies 5. These are characterized by the existing capacities, the efficiencies and fue costs (see Tabes 14 and 15 in Appendix B). We use staircase margina cost curves to represent the operation of generators. The margina cost curves describe the (endogenousy determined) merit order, which account emission factors and fue costs of each pant. Moreover, we assume some renewabe poicy whereby TSO smooths out the variation of wind power and transforms it into a base oad band. This poicy has obviousy a cost that we assume spread through a feed in tariff that a consumers pay. This type of poicy is in pace in some Member States such as Germany. It is introduced here as a proxy for poicies favoring renewabes and shoud obviousy be modeed in more detai in a true case study. The power system covers 15 nodes ocated in four countries (see Figure 1). They are connected by 28 arcs with imited capacity. The grid is modeed by DC oad fow approximation and represented using a Power Transfer Distribution Factor (P T DF ) matrix provided by ECN (see [4] and Figure 1 for the vaues). 4 E.ON Energie AG, Eectrabe SA, Eectricité de France, ENBW Energieversorgung Baden- Württemberg, Essent Energie Productie BV, Nuon, RWE Energie AG, Vattenfa Europe AG. 5 Hydro (running-of-river pants), wind, nucear, ignite, coa, CCGT, other-gas and oibased pants. 3

5 Suppy and demand are ocated at seven nodes: two in Begium (Merchtem and Gramme), three in the Netherands (Krimpen, Maastricht and Zwoe), one in Germany ( D ) and, finay, one in France ( F ). The remaining German and French nodes are passive and are ony used to transfer eectricity. We consider two periods: winter peak and summer off-peak, measured in hours per year, with different durations. They are required to be identica in a countries incuded in the network. We state that the off-peak asts seven months (5,136 h) and the peak the remaining five months (3,624 h). We fuy acknowedge that this is not sufficient to get a good representation of the system. But it is sufficient to iustrate the phenomenon at work whie keeping the mode simpe. Figure 1: Northwestern European market and Network ine capacities We distinguish two consumer groups: househods (representing sma consumers of eectricity and tertiary) and arge industries with an intensive use of eectricity. Their demand functions are assumed to be inear and differ over nodes. Moreover, it seems reaistic to suppose a constant industria consumption eve over the year whie sma consumers demand more eectricity in winter than in summer (see Tabe 13 in Appendix B). As aready indicated, we imit ourseves to seasons assumed to respectivey represent the peak and the off-peak periods. Demand curves are caibrated through a reference point and an easticity at that point. A whoesae reference price of 40 e/mwh is appied to sma and industria consumers demand functions in both periods. Since our priority consists in anaysing consumers reactions to the introduction of the ETS, we mode ong-run demand function. Sma consumers are expected to behave ess fexibe and then we assume that their demand easticity is -0.1 in the reference point. Contrariy, we set industria consumers demand easticity at -1 in order to account for their abiity to eave Europe in case of too high eectricity prices 6. We adopt an emission cap E of about 397 Mio ton p.a. We assume that the permit market is restricted to the power sector and we do not mode permit trade among industries. The justification of this restriction is practica: there is no information avaiabe on the demand for aowance by arge industria consumers. We sha offer a forma extension of the current mode that embeds the whoe set of sectors covered by the EU-ETS. Given this restriction, the emission cap E corresponds exacty to the sum of the NAPs of the eectricity generating 6 Since there is amost compete ack of information of demand response of industria consumers, we took this -1 vaue from a side of Newbery. 4

6 companies incuded in the simuation tests. It defines the amount of emissions aowed in the power market. A detaied ist of the aowance aocations per country and generating company is reported in Tabe 16 in Appendix B. Finay, market simuations are caibrated with data updated to 2005 (see Appendix B). 3 Statement of the Reference Mode The basic modes simuate a perfecty competitive market with three main agents: power companies, consumers (househods and industries) and the Transmission System Operator (TSO hereafter). A agents are supposed to be pricetakers and maximize their surpuses. Power companies profit from seing eectricity to both consumer groups. There is no price discrimination in perfecty competitive markets. Eectricity price are thus identica in both market segments. Their generation activity is subject to a production baance and a technoogica constraint. The first ensures that the quantity of eectricity produced is at east equa or greater than the amount sod. The second accounts for the imited production capacity of power pants. Consumers maximize their surpus supposing that the amount of eectricity demand is positive. Reca that, for the sake of reaism, we assume that industria consumption is constant over the year. To this aim, we add a constraint to the industria optimization probem, imposing the equaity of the houry consumption of eectricity during summer and winter. The TSO maximizes the merchandising surpus accruing from transmission operations. As expained above, we incude network constraints in P T DF form. The transmission market is thus a noda system operating under perfect competition. Last, we represent the aowance market through an emission baance equation and introduce a market cearing price for that market. The market cearing price on the aowance market is thus an output of the mode. The perfecty competitive scenario incuding both the emission and the transmission constraints is our reference mode. In Appendix A.1 we describe the mathematica formuation of this reference mode. It is expressed in compementarity condition form. 4 Resuts of the Reference Mode In order to assess the ong-run impacts of the EU-ETS on industries eectricity prices and demand, we compare the resuts of the reference mode with and without emission constraint. Our resuts are certainy in ine with industries caim: arge industria consumers decrease their eectricity consumption by -11% after the inception of the EU-ETS. This hods in both periods modeed 7. This is accompanied by a cut of amost -23% in their annua surpus. Tabe 1 shows industria eectricity houry demand before and after the introduction of the emission constraint. 7 This fa in their eectricity consumption is significant, but it is reay driven by our assumptions. The reader shoud indeed keep in mind that we mode industries ong run behavior and hence assume a demand easticity of -1. This mode assumption is seected to fit the threat, exposed by arge consumers, of reocating their production capacities. 5

7 MW Without ETS With ETS Reative Changes Germany 32,214 25,095-22% France 25,015 24, % Merchtem 3,573 3,538-1% Gramme 2,029 1,963-3% Krimpen 2,722 2,603-4% Maastricht % Zwoe 1,800 1,615-10% Tota 68,294 60,613-11% Tabe 1: Industry s houry demand without and with EU-ETS The inception of CO 2 restrictive poicies aso introduces some contrasts in the inter-tempora eectricity prices. They are respectivey reported in Tabes 2 and 3 and are identica for industries and househods. Transmission costs infuence off-peak power prices. In fact, some of the grid connections are congested. In particuar, ines are satured between France and Germany and between France and Begium. The interconnection capacity between the Begian node Merchtem and the Dutch node Krimpen is congested and the same happens between Gramme and Maastricht. The direction of the fows reveas that France exports both to Germany and Begium, which, in turn, suppies the Netherands. Since a great part of the nucear eectricity generated in France is exported, the congestion of the ines, combined with the margina cost pricing impicit in this mode, reduces the French power price to a eve cose to its margina operating cost (5.07 e/mwh). This happens even though the price of the neighboring countries is higher. The congestion cost makes the difference. In peak, instead, a consumers pay the price set SUMMER e/mwh Without ETS With ETS Reative Changes Germany % France % Merchtem % Gramme % Krimpen % Maastricht % Zwoe % Tabe 2: Off-peak eectricity prices of industry and househods in e/mwh without and with ETS at the hub node because the transmission grid is no onger congested. This is ony true in the carbon unconstrained version of the mode. In fact, without carbon restriction, a ine connecting France with Begium is congested in peak period and, then, the standard economy of noda transmission system makes eectricity prices different in a nodes. Under the EU-ETS, consumers gobay reduce their eectricity demand and the amount of power exported in winter decreases, avoiding the congestion of the grid. It is then important to recognize the rea nature of the phenomenon. The EU-ETS decreases congestion because it reduces industria demand. 6

8 The comparison reported in Tabe 2 shows that, with the EU-ETS, eectricity prices increase in off-peak. The merit order of the pants can expain these seemingy strange resuts. We reca that, in perfect competition and when there are no transmission constraints, the most expensive pant defines eectricity price at the hub node. With the impementation of the EU-ETS, generators are encouraged to expoit ow emitting technoogies, ike CCGT, and reduce the utiization of more pouting ones (notaby coa and ignite). In particuar, hydro, wind and nucear capacities are saturated in both periods; whie CCGT pants are fuy run in winter. In this case, CCGT technoogies, together with the pass through of the aowance price, set the summer eectricity price at a higher eve than coa based pants woud have done without the EU- ETS. In peak, we observe a reverse behavior: under the emission trading poicy, energy prices become ower. The combination of the change of pant merit order and the eectricity demand cuts expains this outcome. The EU-ETS obiges generators to modify their fue mix and to switch ower emitting pants in order to achieve their emission target. For a these reasons, CCGT power units determine the winter eectricity price (47.36 e/mwh), repacing the natura gas and oi based instaations used in the case with no environmenta reguation. This makes power cheaper than before. The requirement that industria demand remains constant throughout expains how demand and price can simutaneousy decrease. This is confirmed by the behavior of sma consumers in the peak period: in presence of emission imitations, they increase by 1% their eectricity consumptions, since prices are a itte bit ower in each node. In off-peak, instead, they essen their energy utiization (amost -3%) as a consequence of the raised power prices. Tabe 3 reports the associated price reative changes. WINTER e/mwh Without ETS With ETS Reative Changes Germany % France % Merchtem % Gramme % Krimpen % Maastricht % Zwoe % Tabe 3: Peak eectricity prices of industry and househods in e/mwh without and with ETS The EU-ETS has thus two effects: first, it adds a carbon component to summer prices; second, it removes expensive and inefficient units in winter. These two effects interact with the constant (endogenous) demand eve in the industria sectors. Under the EU-ETS restrictions, the aowance price found is e/ton. This positive vaue signas a tight emission cap. This contradicts the ow aowance price that prevaied in the second part of the first compiance period after goba NAPs were commony judged excessive. Our figures can easiy be confirmed through the common wisdom about the tota NAPs in the first compiance period. Reinaud (see [12]) and Smeers (see [18]) expain that CCGT 7

9 and coa pants become competitive when CO 2 price reaches a certain tapping point. In particuar, Smeers in [18] argues that this shift between CCGT and coa happens for CO 2 price of around e/ton. Reinaud in [12] indicates a CO 2 cost of 19 e/ton. Our resuts are aigned with these studies: ony with a high permit price (ike e/ton), one forces eectricity generating companies to repace coa technoogy with ess emitting CCGT stations. This is exacty what we have observed 8. With the appication of the environmenta poicy, carbon emissions gobay reduce by -14% from a eve of about 464 Mio ton p.a. to 397 annua Mio ton. This is exacty the CO 2 emission ceiing imposed on the power sector in the mode. Parae to the goba reduction in eectricity consumption, one observes a decreasing emission eve in amost a the nodes of the mode. There are two exceptions through: they occur in the Dutch ocations Maastricht and Zwoe. In the first of these two nodes, the goba poution eve increases with respect to the unconstrained carbon case. Power producers, in fact, raise the operation of their CCGT pants in order to reduce their eectricity imports, as a consequence of the congestion of the ine between Maastricth and the Begian node Gramme. In Zwoe, instead, the emissions remain constant, since both the capacity and the fue mix used to produce energy do not change with respect to the unconstrained carbon case. Finay, under these conditions, eectricity generating companies profit gobay increases by 16% with respect to the unconstrained carbon case. A these compy with industria position: there might be a probem of competitiveness and demand destruction. This justifies to expore their soution proposa. 5 Average Cost Pricing Mechanism Our resuts compy with the thesis that the impementation of the EU-ETS increases eectricity prices and induce a corresponding reduction of consumers demand, at east in the ong run. This effect is not reay surprising and resuts from basic economic phenomenons. Carbon eakage indeed impies both increase of emissions and oss of economic activity. The phenomenon serves neither the EU nor the environment. In order to expore a possibe aternative to that outcome, we modify the basic mode and assume that industria consumers either pay fu average generation cost or directy contro part of the generating capacity instaed in the network. This approach has two impications: the market is segmented into two sub markets respectivey representing the eectricity intensive industries and househods. As a direct consequence, the generation system is aso spit between these two market segments. This subdivision is endogenousy determined since 8 Smeers and Reinaud indicate different CO 2 switching price. This because their studies are based on different input data. Reinaud assumes that coa and gas fue prices are respectivey 1.5 e/gj and 3.5 e/gj; whie those of Smeers study are higher: 2.3 e/gj (coa) and 3.5 e/gj (gas). Efficiency rates are identica (0.37 (coa) and 0.49 (gas)) and aso gas emission factor (0.412 ton/mwh). Smeers supposes that coa emits more than Reinaud does (0.99 ton/mwh vs ton/mwh). In order to check the consistency of these two studies, we compute the CO 2 switching price using Smeers formua and Reinaud s input data. The resut obtained is e/ton that is aigned with Renaiud s vaue. Our input data are more simiar to those adopted by Smeers. In particuar, we assume that coa costs 2.22 e/gj with an emission rate of and efficiency rate of 0.37; whie gas price is 4.15 e/gj with an emission factor of and efficiency rate of For more detais, see Tabe 15 and [1]. 8

10 the fina demand of the arge industria consumers is aso endogenousy determined. We sha see that the principe underying this aocation is to equaize the margina vaue of the capacities aocated to the two segments. This aso impicity amounts to maximizing the tota capacity vaue. Within this market segmentation, we assume that sma consumers are sti priced at the short-run margina costs. In contrast, eectricity intensive industria consumers pay eectricity at the average costs corresponding to the fu cost of the power pants reserved for them. We consider two particuar appications of this view. We first represent a case where industria consumers can purchase eectricity at the same average cost price in any node by the means of a power purchase consortium. The assumption corresponds to a request of the arge consumers to achieve a singe price on the continenta copper pate through extended countertrading. Large industria consumers (and aso seemingy the European Consumers) indeed see it desirabe, bearing sufficient investment in transmission, to deveop countertrading and to a singe eectricity price in regiona market. In this scenario, the fina eectricity price they faced by arge industria consumers woud incude both the average production and the average transmission costs, which consumers generay pay to the TSO. We aso mode a second case, where we suppose that industria consumers buy ony the eectricity produced in the node where they are ocated. This eads to a noda average cost based price system. In this way, industria consumers are reieved of paying transmission costs, but they are subject to the various constraints that affect generation at that node. In other words, they are not free to choose the technoogy used to suppy them. This makes eectricity prices stricty depending on the fue mix at the node. Subsection 5.1 describes the resuts of the average cost modes. The average cost price formuation makes the mathematica probem more compex because averaging processes destroy the monotonicity properties of the mode. Average cost pricing may aso make the probem infeasibe. This wi be in particuar the case if the fixed charge aocated to the price increases this atter too much. In order to attempt to mitigate numerica difficuties, we sove the average cost based modes as a sequence of two different sub-probems. For each average cost price modes, we first simuate a market with capacity spitting and demand segmentation where both consumer segments are priced at the same short-run margina costs (the so-caed preiminary probem ). This aways has a soution. This preiminary step amounts to simuating a standard competitive mode. Its soution is empoyed as starting point for soving the probems with average cost pricing system. We then run the average cost pricing modes. These probems may not have a soution 9. The surpus maximization probem representing the sma consumers is identica in both the preiminary and the average cost probems; changes concern ony industria consumers equations. For the sake of simpicity, we do not report the formuation of the preiminary modes. The reader shoud simpy note that these preiminary modes are quite simiar to the average cost probems presented in the foowing sections: one just repaces the average cost with a margina cost based price. 9 In our specific cases, we can find soution to the singe and the noda average cost pricing scenarios both with or without the preiminary mode. Nevertheess, these two approaches give different resuts. As indicated, this is perfecty possibe in theory because of non monotonicity; it aso occurs in practice. 9

11 Both average cost pricing mechanisms adopt the same representation of perfecty competitive transmission and emission markets as the reference mode. Moreover, the main structure and the constraints of the average cost based probems are quite simiar to the those of the reference case. Power producers desire to maximize their annua profits under the standard production and capacity constraints. On the other side, consumers maximize their goba surpuses. The compementarity conditions of the singe and the noda average cost pricing systems are presented respectivey in Appendices A.2 and A Resuts of the Average Cost Pricing Modes Our anaysis shows that singe and noda average cost based contracts have different impacts on industries. These depend on their ocation and the technoogica mix used to produce eectricity in each node. Tabes 4 and 6 report their vaues. The singe average cost price amounts to e/mwh. Fixed costs contribute for the argest part, foowed by the fue and the emission charges. Industries tend to congest the network and for this reason they pay transmission costs to the TSO. The appication of the singe average cost pricing system resuts in a goba increase of +5% of the industria eectricity consumption with respect to the reference case. This is the desired objective of the poicy. However, not a industria consumers benefit from the appication of this innovative pricing scheme. In Germany, in the Netherands and in the Begian ocation Merchtem industries pay eectricity at ower prices. Reative changes are between -19% (in Merchtem, Krimpen and Maastricht) and -17% (in Germany). Instead, in France and in the Begian node Gramme the comparison with the margina cost case shows that a singe average cost price eads to an increase of +69% and +6% respectivey. The direct consequence is that industria consumers require ess eectricity. The decreases of -22% (in France) and -1% (in Gramme) are compensated by the increase of industria eectricity demand in the other nodes. Tabe 5 reports in absoute vaues industries houry eectricity demand under different pricing scenarios. Cost Components e/mwh Fue Transmission 2.74 Emission 7.32 Capacity Average cost price Tabe 4: Singe average cost price The endogenous aocation of the instaed capacity may offer a pausibe expanation to these resuts. A great part of the tota base-oad technoogies (hydro (60%), nucear (52%) and ignite (62%)) existing in the network is now reserved for industries. Aso 71% of the avaiabe wind capacity is dedicated to industria consumers. Both in France and in Gramme, eectricity generating companies reserve ony cean technoogies (namey hydro, wind and nucear) for industries. As expected, France pays an important roe in this market segment, since it covers amost the entire industria eectricity demand that the 10

12 bordering countries are not abe to satisfy. In fact, the nucear capacity that French generators reserved for the industry is arger than the amount they reay need for their own (industria) consumers and then anyone can have access to the exceeding part. This aows meeting the industria market energy baance. This has an obvious side effect; this affects the French industries competitive positions, which now have to buy eectricity at a higher price. Moreover, because of the capacity spit, French sma consumers have a imited access to nucear power. This has aso a further negative consequence on their eectricity prices, especiay in winter, when their energy consumption is higher. In fact, during the peak period the capacity reserved for French sma consumers is not sufficient to meet their entire demand and therefore they have to import more expensive eectricity. One encounters a simiar situation in Gramme, where oca industries have to share their hydro, wind and nucear pants with foreign arge consumers. With respect to the reference case, househods face higher eectricity prices both in peak and off-peak periods. This is partiay due to the endogenous spit in capacity that reserves cheap and base-oad technoogies to industria consumers. In off-peak, CCGT pants set eectricity prices after accounting for the corresponding carbon cost. In peak, instead, natura gas and oi based instaations become again active and determine the prices. Moreover, in this case, permits are more expensive (+17%) than in the reference scenario. In fact, power companies have to produce more eectricity to fuy cover industria requests. Athough sma consumers face eectricity price increases, their goba demand does not vary too much with respect to the reference case 10. The industria demand effect prevais over househod consumption. This resuts in a growth of the amount of eectricity produced and, consequenty, of the demand for aowances. Since carbon emissions are capped, the aowance price increases 11. MW Reference case Singe Average case Noda Average case Germany 25,095 31,065 26,913 France 24,910 19,408 29,002 Merchtem 3,538 4,511 2,176 Gramme 1,963 1,939 2,601 Krimpen 2,603 3,319 2,119 Maastricht 889 1, Zwoe 1,615 2,033 1,113 Tota 60,613 63,408 64,543 Tabe 5: Comparison of the industria consumers houry demand under different pricing scenarios The appication of the noda average cost based poicy has a goba positive effect on industries: they increase their eectricity consumption with respect to both the reference (+6%) and the singe average cost price (2%). Looking at the houry demand vaues reported in Tabe 5, one can easiy notice that industria consumers assume different positions in reation to the node where they are 10 Totay sma consumers reduce their demand by -0.4% in off-peak and by -2.8% in peak. We reca that we assume that they have a sma price easticity. 11 With respect to the reference case, aowance price augments by +17%. In fact, it reaches a cost of e/ton starting from a vaue of e/ton. 11

13 ocated. In France and in the Begian node Gramme, the noda average cost pricing system represents the best poicy to hea industria difficuties. This is mainy a resut of the technoogica structure in these ocations. In a country, ike France, where nucear is the main power source, noda average cost pricing contracts perfecty suit industria consumers needs, since they have wide access to this cheap and cean technoogy without sharing it with foreign consumers 12. On the other side, the situation of industries paced in nodes where eectricity is mosty produced by CCGT or coa based technoogies is more critica. This is what happens in the Netherands and in the Begian node Merchtem, where industries are reay damaged by this new contractua poicy 13. In Germany, instead, the industria eectricity consumption decreases by -13% with respect to the singe average cost case, but it is higher in comparison to the reference eve (+7%). e/mwh Fue Emission Capacity Average cost price Germany France Merchtem Gramme Krimpen Maastricht Zwoe Tabe 6: Noda average cost price The contribution of the three components of the noda average prices (fue, emission and fixed costs) shown in Tabe 6 may expain these resuts. They depend on the noda technoogy mix. For instance, in France, the industria eectricity price is e/mwh, of which 4.50 e/mwh are the average fue costs and e/mwh are the fixed charges. Eectricity producers expoit ony nucear pants 14 to cover French industria demand. Nucear is an environmentafriendy technoogy and then French average cost based price does not incude emission burdens. Fue and capacity charges correspond exacty to the costs of a nucear pants, as indicated in Tabes 15 and 17 in Appendix B. In Gramme, industries are mainy suppied by hydro, wind and nucear. Moreover, 31% and 100% of the avaiabe CCGT and natura gas pants are empoyed, but these proportions correspond to a few MW of capacity (respectivey 373 MW and 170 MW). Hydro and wind reduce the fue average cost; whie emissions caused by CCGT and natura gas stations add to this cost. Contrariy, industria consumers in Merchtem, the other Begian node, face the highest noda average cost price of the market. Eectricity generating companies, in fact, use the entire amount of coa capacity (1,564 MW) and 24% (612 MW) of the CCGT pants instaed in the node to cover their eectricity demand. A the cean power stations (namey hydro, wind and nucear) are dedicated to househods. As indicated in Tabe 6, this impies high emission costs. Dutch industries in Maastricht and in Zwoe are ony suppied by CCGT stations; in Krimpen, instead, 12 It is exacty the opposite of what happens in the singe average cost pricing scenario. 13 Their eectricity cuts are as foows: -39% (reference case) and -52% (singe average) in Merchtem, -19% (reference case) and -36% (singe average) in Krimpen, -30% (reference case) and -45% in Maastricht and, finay, - 31% (reference case) and -45% (singe average) in Zwoe. 14 Precisey 64% of the nucear capacity instaed corresponding to 29,002 MW. 12

14 the set of the technoogies given to industries is composed of wind, nucear, coa and CCGT. Finay, German industries are mosty suppied by ignite technoogies, accompanied by nucear, ignite, hydro and wind : this is the reason why emission charges have the major weight in their prices. GWh Househods Industry Tota Summer Winter Tota Reference case ,176 Singe Average case ,189 Noda Average case ,188 Tabe 7: Consumers annua demand under different pricing scenarios The comparison of the noda average cost case with the reference mode shows that the appication of this average cost pricing mechanism has a goba negative infuence on househods. We reca that, in a transmission constraint free system, sma consumers prices are determined by the ast power station used to produce eectricity augmented by the associated carbon cost. Since, both in the reference and in the noda average cost modes, CCGT pants are the margin; differences of eectricity prices depend ony on carbon cost. In fact, in the reference case, aowances are cheaper than in the noda average cost case (24.44 e/ton vs e/ton). On the other side, the anaysis of the two average cases highights that sma consumers demand assumes different trends in reation to the node and the period considered: apart in Merchtem, they gobay reduce their eectricity consumption in off-peak (-4.76%); whie in peak ony French househods essen their eectricity consumption (-3.3%). These resuts are sti affected both by the permit price (that in this case is e/ton) and the fue mix. Biion e Ref. NO EU-ETS Ref. EU-ETS Singe A. Noda A. Benefit Industry Househods Consumers Generators (29.246) a (32.943) a (36.785) a TSO Wefare Tabe 8: Wefare anaysis under different pricing scenarios a Generators profits accounting for the vaue of the carbon permits received for free. Tabe 7 compares consumers annua demand under the pricing scenario considered. The resuts therein reported confirm the tendency described. The appication of the average cost pricing systems induces a wefare redistribution among market payers. This is shown in Tabe 8. Sma consumers 13

15 maximize their surpus under a perfecty competitive market. On the other side, industries benefit a ot from buying eectricity at the noda average cost price. Their surpus increases by +20% and +16% compared to the reference and the singe average cost case respectivey. With the inception of the EU-ETS and without free aowances, that is if aowances are auctioned, generators profits become ower 15. This is due the fact that, power companies have now to account for the additiona burdens caused by emissions. This hods in a cases except for the noda average cost scenario where power producers increase their profit by +1% with respect to the mode without EU-ETS reguation. This extra gains come from seing eectricity to househods. Reca that this wefare anaysis does not incude the vaue of the CO 2 permits that eectricity producers woud sti receive for free. However, in Tabe 8, we report aso the case when free aowances are incuded in the computation of generators profits. Note that, under this assumption, generators benefits are much higher. These resuts support the theory of the so caed windfa profits (see [14]). Finay, the TSO s merchandising surpus depends on network utiization. For instance, in the reference case with carbon restriction, the TSO ony makes profit in summer, when the network is congested. In contrast, the construction of the noda average cost case impies that the TSO ony receives payments from househods. Moreover, these payments are very ow with respect to the other cases because ony one ine, connecting Germany to France, is congested and this occurs ony in winter. One can note that one fas back on the standard resut of economic theory that perfect competition maximizes goba wefare. 6 Sensitivity and Robustness Anaysis In this Section, we conduct a sensitivity anaysis in order to check how industria consumers demand and emissions vary when one modifies the assumptions reguating the permit market. In our modes, we endogenousy determine carbon price and expore its modification under different eectricity pricing mechanisms. We now simpify our modes by directy introducing an exogenous permit price. We test two cases when: 1. Permit price is set at 10 e/ton (hereafter EU-ETS PP 10 case) 2. Permit price is set at 20 e/ton (hereafter EU-ETS PP 20 case) and we compare the resuts with those of the origina version of our modes. This test requires a sma modification of the mode. An exogenous aowance price dispenses with the need to expicity mode the market of aowances. This impies that there is no need to retain the emission constraint in the mode. Power companies buy and se permits on the market at exogenousy given prices. In contrast with we have done before, we first define a permit price and then we compute the amount of emissions generated. Note that the endogenous permit prices found in the previous sections are higher than those tested here. Obviousy, the ower carbon charges are and the higher CO 2 emitted wi be. 15 The comparison among the perfect competition mode without carbon constraint and the other scenarios resuts in the foowing profit cuts: -23% (reference case with EU-ETS) and -14% (singe average cost case). 14

16 Reca that the reference mode without emission restrictions, industries gobay consume 68,294 MW of eectricity per hour. This amount decreases and fas to 60,613 MW when we set a cap of 397 ton p.a. The endogenous vaue of the associated permit price is e/ton. Instead, when we fix a price of 10 e/ton, industria consumption increases by +4% with respect to the reference case with EU-ETS reguation, because eectricity becomes cheaper 16. Aso sma consumers demand raises. As indicated in Tabe 9, annua emissions are about 419 Mio ton (+6% with respect to the reference case). This is due to the fact that eectricity generating companies expoit more ignite/coa and CCGT pants in order to cover the increased whoesae demand. With such a ow permit price, ignite/coa stations remain more competitive than CCGT stations. A simiar reasoning hods when we appy a permit price of 20 e/ton. Industries demand 61,946 MW of eectricity each hour. In this case, the increase with respect to reference endogenous case is 3%, ower than in the EU-ETS PP 10 scenario. Imposing a permit price of 20 e/ton makes eectricity more expensive. Under this assumptions, the goba amount of CO 2 emitted is 407 Mio ton (see Tabe 9). A simiar reasoning can be appied to the sensitivity anaysis conducted on the singe average cost pricing modes. Fixing the permit price at 10 e/ton entais a huge increase (+10%) of the industria consumers eectricity consumption. In fact, starting from a houry consumption of 63,408 MW, when the carbon price is endogenous, industries now demand 69,723 MW. This depends on the fact that the singe average cost price is set at e/mwh, that is, 11% ess than the singe average cost price of the origina mode (38.10 e/mwh). Imposing a permit price of 20 e/ton has sti a positive impact on industries, even if the raise in their goba eectricity consumption is ony +3%. We reca that in the origina version of this mode, CO 2 permits cost e/ton. The more enient emission prices in combination with increased eectricity demand expain why emission eves now reach the high eve of 464 Mio ton and 441 Mio ton respectivey in the EU-ETS PP 10 and EU-ETS PP 20 scenarios (see Tabe 9). Note that 464 Mio ton corresponds exacty to the poution eve without any environmenta reguation. The appication of a mid environmenta poicy, ike EU-ETS PP 10 aso induces industries to consume a ot of power in the noda average cost mode. The consumption increase with respect to the origina version of the mode is about 11% (71,840 MW). In the EU-ETS PP 20, the goba eectricity consumption ony increases by +3% (66,619 MW). Again, this impies higher CO 2 emissions as indicated in Tabe 9. In particuar, when the permit price is fixed at 10 e/ton, goba emissions amount to 471 Mio ton, that is even higher than the tota carbon emitted in absence of an environmenta reguation. This suggests that the Reguator shoud impose a restrictive cap to emissions in order to achieve the EU-ETS goa and make cean technoogies more competitive. We further check the robustness of our resuts by anaysing the case when industria consumers demand is ess eastic. Starting from the same reference demand point (see Section 2), we set the easticity at that point at -0.8 instead of -1. This obviousy modifies the resuts (for a agents incuded in the modes), but keeps consumption and price trends amost identica. Tabes 10 reports annua consumers demand. As in Tabe 7, industries increase their eectricity 16 This is due the fact that the CO 2 contribution to the power price is reduced. 15

17 Mio ton PP endogenous PP 10 PP 20 Reference case before EU-ETS 464 Reference case after EU-ETS Singe Average case Noda Average case Tabe 9: Emission eve under different pricing scenarios consumption in the average cost pricing system. Again, the increment is higher with noda average cost based contracts. Sma consumers, instead, suffer from pricing discrimination and reduce their power demand. Nevertheess, this new industria easticity assumption modifies some resuts of the noda average cost mode. GWh Househods Industry Tota Summer Winter Tota Reference case ,169 Singe Average case ,188 Noda Average case ,192 Tabe 10: Consumers annua demand under different pricing scenarios. Industria easticity 0.8. Biion e Ref. NO EU-ETS Ref. EU-ETS Singe A. Noda A. Benefit Industry Househods Consumers Generators (29.791) a (32.770) a (32.463) a TSO Wefare Tabe 11: Wefare anaysis under different pricing scenarios. Industria easticity a Generators profits accounting for the vaue of the carbon permits received for free. In particuar, househods eectricity demand is higher than in the singe average cost pricing case. In the origina version of the mode (industria easticity set at -1), one faces the reverse situation. The comparison among the resuts reported in Tabes 10 and 7 shows that industria eectricity demand is reduced. Consequenty, they need ess capacity that now becomes avaiabe for sma consumers. Househods therefore increase their eectricity consumption in both periods. This raises network congestion and aows TSO to increase its merchandising surpus as indicated in Tabe 11, where we report the resuts 16

18 of the wefare anaysis conducted under the industria easticity assumption of Concusion The specia contracts tested in this paper represent one response proposa of European industria consumers to the new environment created by the EU-ETS. It impies changing the pricing system to mitigate the increase of eectricity prices caused by the EU-ETS. We test two different average cost pricing poicies (singe and noda ones) that have different effects on industries. A common point, discussed in Section 5, is that average cost based pricing encourages industries to maintain their activities (here represented by consumption of eectricity) with respect to the reference eve at east under the condition retained in this mode (exogenous capacities and efficient transmission market). However, neither the singe nor the noda average cost pricing mechanisms competey mitigate the burdens imposed by the EU-ETS on the industria sector. The first poicy negativey affects French and part of the Begian eectricity intensive users, who, instead, profit of the second average strategy. In Germany, in Merchtem and in a Dutch nodes, industries face the opposite situation. The concusion is that the impact of these specia contracts on eectricity intensive consumers depends on the particuar pricing scheme impemented (singe or noda) and on the fue mix adopted to produce eectricity in the countries where industries operate. This is obviousy the key factor, which defines power average cost based prices. Finay, the high emission aowance price reveas the stress that the generation system is currenty subject to. This suggests that investments in renewabe power technoogies, the improvement of the efficiency of the existing eectricity units and the repacing of od power units are very needed. But ooking at this probem requires capacity expansion mode. References [1] K. Davis and URS Corporation, Greenhouse Gas Emission Factor Review, Edison Mission Energy, [2] D. Demaiy and P. Quirion, CO 2 abatement, competitiveness and eakage in the European cement industry under the EU-ETS: grandfathering versus output-based aocation, Cimate Poicy, 6(2006), pp [3] S.P. Dirkse and M.C. Ferris, The PATH sover: a nonmomotone stabiization scheme for mixed compementarity probems, Optm. Methods Software, 5(1995), pp [4] ECN, Evauation of modes for market power in eectricity networks Test data, Retrieved December Avaiabe at http : // services/modes and instruments/competes/mode evauation/. 17

19 [5] European Commission, Nationa Reports on Verified Emission and Surrendered Aowances, Retrieved Apri Avaiabe at http : //ec.europa.eu/environment/cimat/emission/review e n.htm. [6] Eurostat, Consumption of eectricity by industry, transport activities and househods/services, Retrieved June Avaiabe at http : //epp.eurostat.ec.europa.eu. [7] F. Facchinei and J. S. Pang, Finite-Dimensiona Variationa Inequaities and Compementarity Probems, Springer, Voume I, [8] IEA, Projected Costs of Generating Eectricity. Internationa Energy Agency, [9] McKinsey and Ecofys, EU ETS Review, Report on Internationa Competitiveness, European Commission, DG Environment, [10] A. Nagurney, Network Economics, A Variationa Inequaities Approach, Kumer Academic Pubishers, [11] U. Oberndorfer and K. Reggings, The Impacts of the European Union Emissions Trading Scheme on the Competitiveness in Europe, Discussion Paper No , ZEW Centre for European Economic Research, [12] J. Reinaud, Emission Trading and Its Possibe Impacts on Investment Decisions in the Power Sector, IEA Information Paper, [13], Industria Competitiveness under the European Union Emission Trading Scheme, IEA Information Paper, [14] J.P.M. Sijm, K. Neuhoff and Y. Chen, Cost Pass Thought and Windfa Profits in the Power Sector, Cimate Poicy 5 (2006), pp [18] Y. Smeers, Assessment of EU CO 2 Reguation, Ifri Energy Roundtabe, January 30, [15] UCTE, Nationa eectricity consumption 2005 and peak oad, Retrieved May Avaiabe at http : // [16] UCTE, Monthy Consumption of a specific country for a specific range of time, Retrieved May Avaiabe at http : // Appendix A. Compementarity Probem The modes used in this study are a formuated as compementarity probems. First, we briefy present the mathematica formuation of the compementarity probems (CPs) and, then, we describe the modes impemented (see [7, 10] for a presentation of compementarity probems). From a mathematica point of view, CPs are defined as foows: 18