CO2 abatement from renewables in the German electricity sector: Does a CO2 price help?

Energy Economics 40 (2013) S149–S158

Contents lists available at ScienceDirect

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

CO2 abatement from renewables in the German electricity sector: Does a CO2 price help? Hannes Weigt a,⁎, Denny Ellerman b, Erik Delarue c a b c

Forschungsstelle Nachhaltige Energie und Wasserversorgung, Wirtschaftswissenschaftliche Fakultät der Universität Basel, Peter Merian-Weg 6 Postfach, CH-4002 Basel, Switzerland Climate Policy Research Unit, Loyola de Palacio Chair, EUI, Via delle Fontanelle 10, I-50014 San Domenico di Fiesole (FI), Italy University of Leuven (KU Leuven) Energy Institute, TME Branch (Applied Mechanics and Energy Conversion), Celestijnenlaan 300A, Box 2421, 3001 Leuven, Belgium

a r t i c l e

i n f o

Available online 23 September 2013 JEL classification: L94 Q58 Keywords: Emission trading system Renewables policy Interaction Emission abatement Germany

a b s t r a c t The overlapping impact of the Emission Trading System (ETS) and renewable energy (RE) deployment targets creates a classic case of interaction effects. Whereas the price interaction is widely recognized and has been thoroughly discussed, the effect of an overlapping instrument on the abatement attributable to an instrument has gained little attention. This paper estimates the actual reduction in demand for European Union Allowances that has occurred due to RE deployment focusing on the German electricity sector, for the five years 2006 through 2010. Based on a unit commitment model we estimate that CO2 emissions from the German electricity sector are reduced by 35 to 60 Mtons, or 10% to 18% of what estimated emissions would have been without any RE policy but with the CO2 price remaining in place at the observed level. Furthermore, we find that the abatement attributable to RE injections is greater in the presence of an allowance price than otherwise. The same holds for the ETS effect in presence of RE injection. This interaction effect is consistently positive for the German electricity system, at least for the considered years, and on the order of 0.5% to 1.5% of emissions. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Through its climate and energy package, the European Union (EU) aims for a sustainable, secure and competitive energy supply. In March 2007, the EU leaders defined the so called 20-20-20 targets by 2020, i.e., a reduction of greenhouse gas emissions of 20% compared to the 1990 levels, a 20% share of renewable energy in the overall energy consumption, and a 20% increase in energy efficiency compared to a projected baseline. The climate and energy package implementing these targets was agreed by the European Parliament and Council in December 2008 and entered into law in June 2009. CO2 emissions from electricity generation and heavy industry are capped to a level of 21% below the 2005 levels in 2020 by the European Union Emissions Trading System (EU ETS). Emissions in the electricity sector are expected to be reduced more than in other sectors of the economy due to fuel switching possibilities from coal to gas and the increased displacing of conventional power plants by renewable energy sources (RES). The overlapping impact of the ETS and renewable energy (RE) deployment targets creates a classic case of interaction effects. When emissions are capped, such as in a cap-and-trade system like the EU ETS, any external factor that affects emissions changes the demand for allowances and thereby the price of allowances. When the external factor is non-policy related, such as technological change, or attributed ⁎ Corresponding author. Tel.: +41 61 267 3259; fax: +41 61 267 0496. E-mail addresses: [email protected] (H. Weigt), [email protected] (D. Ellerman), [email protected] (E. Delarue). 0140-9883/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.eneco.2013.09.013

to obtain a separate policy goal, such as phasing out nuclear, the resulting increase or decrease in the allowance price signals the greater or lesser need for abatement to achieve the unchanging cap in a least cost manner. A problem arises when the policy objectives are the same, as is arguably the case for RE incentives.1 To the extent that the purpose of these incentives is to reduce CO2 emissions as a matter of climate policy, and a cap is already in place to achieve this same objective, no additional reduction occurs. If the incentive is such as to induce deployment of RE that would not be forthcoming in response to the CO2 price, the effect is only to reduce the price of allowances and to substitute a more costly form of abatement for what would have occurred otherwise. When RE incentives are present, the reduction in the allowance price is simply signaling the reduction in demand in one part of the ETS and the consequent need for less abatement in other parts of the capped system. When instruments to promote RE deployment overlap with a capand-trade program, the potential for an interaction with the resulting allowance prices is widely recognized and has been thoroughly discussed in the literature. In Europe, the early discussion was necessarily more theoretical (Böhringer and Rosendahl, 2009; Fischer and Preonas, 2010; Hindsberger et al., 2003; Jensen and Skytte, 2003; Linares et al., 2008;

1 Besides reducing CO2 emissions, it is clear that RE targets might serve other goals as well: create a “green” industry and corresponding jobs, contribute to security of supply, provide clean electricity (free from other pollutants than CO2), overcome certain market failures, etc. For further reading on these issues, we refer to, e.g., Fischer and Newell (2008) and Kalkuhl et al. (2012).

S150

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

Meran and Wittmann, 2008; Morthorst, 2001; Skytte, 2006; Sorrel and Sijm, 2003). As the European 20-20-20 targets came to be formulated, attention turned to these specific proposals and how they would operate on existing systems. As an example, Abrell and Weigt (2008) simulated the interaction between the 20% CO2 reduction target and the 20% RE share target in a computable general equilibrium model of the German economy based on 2004 data and found that achievement of the 20% RE share target made the CO2 reduction target superfluous and thereby reduced the EUA price to zero. De Jonghe et al. (2009) widened the scope of targets to include the full range of plausible CO2 emission reduction and RE deployment targets in a simulation of the interconnected electricity systems of Germany, Benelux, and France. They similarly found that RE share targets could reduce the allowance price to zero depending on the stringency of the two targets and thereby defined a zero-price frontier. The above studies obtain emission reductions by the difference in carbon content of the fuel substitution that takes place as a result of the RE injections. Another strand of literature argues that the combination of intermittency and reduced utilization leads to decreased efficiency in the operation of fossil-generating plants and therefore to a smaller reduction of emissions than suggested by estimates that ignore these power system dynamics (Denny and O'Malley, 2006, 2007). Lang (2009) has even gone so far as to argue that the fuel inefficiencies created by intermittent injections in integrated electricity grids result in negligible if any net reductions of emissions and thus, logically, to no effect on the allowance price. Such questions about the actual reduction of emissions resulting from RE injections have led to several ex post evaluations in the US of observed emissions reductions for CO2, SO2 and NOx corresponding to actual wind injections based on hourly emission data and hourly wind data (Cullen, 2011; Kaffine et al., 2011; Novan, 2011).2 These studies provide no support to the arguments presented in Lang (2009) for they find significant emissions reductions associated with RE injections, although they do find that the reductions are less than what would be suggested by estimates based on average emission rates. In particular, they emphasize that the reductions vary considerably according to the pre-existing configuration of generation capacity and its dispatch in meeting load. For instance, Kaffine et al. (2011) find that while wind injections in the coal-dependent Upper Midwest reduce CO2 emissions by 0.92 t-CO2/MWh, the reduction drops to 0.29 t-CO2/MWh in gasdependent California and to 0.52 t-CO2/MWh in the more mixed coal and gas system in Texas. Focusing on Texas and using different years and estimation techniques, Novan (2011) and Cullen (2011) find an average CO2 reduction per MWh of electricity generated by wind of approximately 0.67 t-CO2/MWh and 0.75 t-CO2/MWh, respectively. The absence of hourly CO2 emissions data in Europe does not allow for comparable estimates that are free from counterfactual assumptions about system operations. Therefore the objective of this paper is to apply a numerical model simulation to estimate the actual reduction in demand for European Union Allowances (EUAs) that has occurred due to RE deployment in the German electricity sector for five years, 2006 through 2010. The analysis is performed focusing specifically on the impact of RE and the EUA price on electricity generation dispatch. The focus is strictly on the short-term, operational aspect of electricity generation with a fixed capital stock, that is, without regard to longer term aspects, such as the effects on investment and retirement decisions. We find that RE injections have reduced CO2 emissions from the German electricity sector by 10% to 18% in these years. In the interest of comparing the relative effectiveness of observed EUA prices and RE incentives in reducing CO2 emissions, we also estimate the CO2 emission

2

The emissions data are provided by the Continuous Emissions Monitoring System reports that all generating plants in the US are required to submit to the US EPA and which are publicly available. Wind injection data is from the electricity network operators as is the case in Europe.

reductions attributable to actual EUA prices both in the presence of RE injections and in their absence. Somewhat surprisingly, we find a second interaction effect, namely, that the abatement attributable to RE injections is greater in the presence of an allowance price than otherwise and also the emission reduction attributable to the EUA price in these years is greater with RE injections than without. Other papers have made ex post estimates of CO2 emission reductions in the electricity sector as a result of the EU ETS (Delarue et al., 2010; Ellerman and McGuinness, 2008), but these papers took RE injections as given and therefore failed to detect the interaction in abatement that we find in our simulations. While nearly all papers focusing on the interaction of RE and ETS instruments find that the use of both together leads to greater emission reductions, none have extended the analysis to find that the abatement attributable to either is greater when the other instrument is present than when it is used alone, much less to explain why the presence of the other instrument should augment the abatement effect of the instrument being considered. In our review of the literature, we have found two papers that note this effect3: Grubb et al. (2008) show that the amount of abatement associated with demand reduction (such as from conservation or RE injections) is greater with a higher carbon price because of the same dispatch effect that we identify. Capros et al. (2011) develop a three-dimensional diagram similar to what is used below and remark both that the abatement associated with a given RE value (shadow price) depends on the accompanying carbon value and that a “maximum degree of synergy” (p. 1480) exists within a certain range of these values. The present paper's contribution to the literature lies more clearly in identifying this augmenting effect and analyzing its limits, including specifically whether this augmenting effect is a general phenomenon. Furthermore, this paper is one of the few to provide quantitative estimates of the reduction in CO2 emissions in a European country resulting from RE injections based on a model calibrated on historical data and using actual load, RE injections, and fuel prices. The existence of this reinforcing effect on abatement has potentially significant policy implications concerning the one-policyone-instrument rule. Much of the paper is devoted to analyzing this interaction to determine its causes and whether it is a general case or a more special one with limited applicability and policy implication. In the remainder of the paper, Section 2 provides the numerical results obtained for the German case and the observed interaction effects. Section 3 focuses on understanding this effect with respect to abatement. Section 4 concludes. 2. Emissions and RE deployment in the German electricity market To derive insights on the interaction between emission abatement and RE deployment in the German setting a numerical model representation is designed and calculated for the observation period from 2006 and 2010. The general model setup is provided in the following section. The overall outcome in terms of generation, CO2 emissions and electricity prices is described afterwards. 2.1. Model and data description The model is formulated as a standard mixed integer linear programming problem including unit commitment restrictions and endogenous pumped storage operation. The model minimizes total generation costs including start-up costs given a fixed hourly demand

3 Linares et al. (2008) come close in noting that the effect on the marginal cost of generation will be greater when coal is on the margin than when gas is on the margin, which they show as an effect of a carbon price by differing slopes of the marginal abatement cost schedule. Still, their focus is entirely on the price effects, not on abatement although that effect could be deduced from their diagrams.

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

level, subject to the plant's capacity restrictions, start-up limits, and the storage balance.4 The underlying data set covers the power plant mix of Germany (Eurelectric, 2010; Umweltbundesamt, 2011; VGE, 2005, 2006, 2009) with installed capacities, plant type, fuel, and age. Generation costs of each plant are based on fuel and emission costs, and the plant's efficiency, which depends on plant type and age. Fuel prices for oil, gas, and coal are taken from the Federal Office of Economics and Export Control (BAFA, 2011) and vary for each month. CO2 emissions are based on the carbon content of the different fuels (IPCC, 2006) and the plant efficiency. Emission Allowance (EUA) prices are the average monthly price from the European Energy Exchange. The fixed underlying hourly demand is based on ENTSO-E (2011) and adjusted to account for import and export as well as RE injection. The model is formulated in the General Algebraic Modeling System (GAMS; Brooke et al., 2008) and solved using the CPLEX solver. All parameters are externally provided and deterministic and the model is solved with hourly time steps. The model is calibrated to historical conditions by seeking to reproduce the observed yearly generation by fuel through adjustments in the marginal generation costs of coal and gas plants and availability factors. Perfect foresight of load and RE injections is assumed in scheduling the dispatch of available plants.5 Simulations are performed on a yearly level, with time steps of 1 h for the years 2006, 2007, 2008, 2009 and 2010. As the full 8760 h of a year are not solvable within one model run, each year is divided into monthly blocks. Focus throughout these analyses is on the impact of the ETS and of RE injections on CO2 emissions. Four basic scenarios are derived: 1. OBS: This observed case represents the calibrated model approximately equivalent of the observed generation shares. The EUA prices are provided as an external cost parameter increasing the marginal costs of each power plant based on the emission intensity of the fuel and the plant efficiency. Injection from renewable sources6 is deducted from the demand as described above. 2. RES: This case reflects only the actual RE injections with EUA prices fixed at zero representing a setting without the ETS in place. 3. ETS: In this case the RE injection is fixed at zero whereas the EUA price is kept at the observed price level without adjustment for the removal of the RE injections. 4. NOPOL: This case represents the no-policy counterfactual where both the EUA price and the RE injection are fixed at zero thereby simulating conditions when neither an ETS nor an RE policy is in place. Those basic scenarios allow for an evaluation of the merit order effect (i.e., the effect on the ordering of plants in the dispatch order and therefore on marginal generating cost) of RES and ETS, the emission abatement by the ETS and by RE injection, and possible interactions of those two elements. It is important to note that the counterfactual scenarios are hypothetical. They are created by assuming that everything else is the same except for the presence of the EUA price or the RE injections, or both. For instance, if the RE injections are not present, the EUA price would in reality clearly be different from what is observed. In the absence of any empirical estimates of the effect of the price interaction, however, we have retained the observed price as if it were a tax, or fixed price. 4 The mathematical model formulation is provided in the appendix. The full model description as well as the detailed data and calibration description can be found in the working paper version (Weigt et al., 2012). 5 Thus, the effect of errors in predicting wind or PV on dispatch and emissions is not incorporated. It is, however, clear that the variation associated with this unpredictability on generation and the corresponding CO2 emissions of the conventional power system is inferior to the direct impact of these RE injection on electricity generation. As an example, Denny and O'Malley (2006) found that in Ireland the error in CO2 emissions due to deviations from expected wind forecasts was less than 1%. In the same trend, Delarue et al. (2009) indicate negligible CO2 emission differences when using different forecast accuracies for wind, in a case study for Belgium. 6 We refer to wind, biomass and solar energy as renewable energies whereas hydro generation is classified as conventional.

S151

Equivalently, the marginal abatement cost curve could be viewed as extremely elastic or flat in the vicinity of the observed price–quantity relation. This would imply that reductions of demand in one part of the ETS are compensated by increases in other parts of the ETS with little effect on the observed price of allowances. Similarly, it could be questioned whether the power plant capacity that we observe when the ETS and RE injections are present would be the same when either or both are absent. In particular, the calibrating adjustments that we make to bring modeled results in line with observed generation may not hold (as we assume in the counterfactuals) when either or both of the policy options are absent. For example, since the EUA price reduces the demand for coal generation, and the RE injections reduce the demand for coal and natural gas generation, it is reasonable to assume that more coal generation would be available when these policies are absent, either because of fewer coal plant retirements or greater availability of existing coal-fired plants. In this case, our estimates of the abatement from these policies may be understated. Overall, as in any modeling exercise, numerical results should be seen as indicative and approximate instead of being precise replications of what would occur in reality.

2.2. Basic price and quantity outcomes The general market results concerning prices, the merit order effect of RE deployment, and the fuel shift between coal and gas due to the CO2 price are in line with what would be expected.7 The emission price leads to a higher utilization of gas fired units compared to coal units since they have a lower emission factor. The RE deployment leads to a lower share of both coal and gas plants as their production is displaced by RE. Unlike the carbon price, the RE injections affect coal and natural gas generation about equally, as shown in panels a and b of Fig. 1. Depending on the year, RE injections decrease the shares of generation by coal and natural gas by 3 to 7 percentage points and 3 to 6 percentage points, respectively. The addition of a carbon price reduces the coal share of generation by another 1 to 3 percentage points (except in 2007 when the carbon price was practically zero), but raises the share of gas by an approximately equal amount. Thus plant dispatch in Germany is more affected by the RE incentives in Germany than the EU-wide carbon price. The price results are also as would be expected. The carbon price increases the cost of marginal generation, whereas the displacement of conventional generation by RE injections reduces it, as shown in panel c of Fig. 1. The EUA price acting alone would have increased prices by around 10 to 13 euros per MWh (again except in 2007), while the merit order effect of the RE injections when taken alone would lead to reductions of 12 to 20 euros per MWh. When both operate together, as in the observed case, the net effect is a decrease of 2 to 8 euros per MWh depending on the year (excluding 2007). However, as the model is calibrated with a focus on the observed generation quantities and not the market prices, the resulting simulated prices should be viewed as indicative and not necessarily what would occur in practice.8 Other model results include the resulting start-up costs and pumped storage operation. Regarding the latter, no clear relation to RES or ETS policy can be observed since its usage varies over the years without any clear trend. Start-up costs (in absolute terms) are higher in the cases without RE injection basically reflecting the need for more plants to cover demand.9 However, the share of start-up costs compared to fuel and emission costs is less than 2%. Since the model does not include 7 More detailed results with RE specific cases and estimates of resulting allowance price impacts are provided in the working paper version (Weigt et al., 2012). 8 For a review on merit-order effect estimates for Germany see Fürsch et al. (2012); in general they are estimated within a range of 5 to 10 euros per MWh for the pure RE injection effect on electricity prices. 9 The hourly structure of the model and its deterministic setting limit evaluations of the impact of RES intermittency on the results (e.g. perfect foresight has an important impact on the use of the pumped storage).

S152

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

Fig. 2. (a) Emission impact for the different scenarios and (b) impact of RES and ETS, both with and without other instrument.

Fig. 1. (a) Coal share, (b) gas share and (c) price impact for the 4 different scenarios in the 5 considered years.

an endogenous import/export calculation, changes in the European trade balances cannot be estimated.

2.3. Emission results Panel a of Fig. 2 shows the total emissions for the four scenarios over the observation period. It is evident that total emissions are significantly higher in the cases where RE injection is not in place (ETS and NOPOL). Based on those emission values it is possible to estimate the abatement attributable to each instrument when it is acting alone and in conjunction with the other instrument. This can be done by using either the OBS or the NOPOL case as starting point. For example, the impact of the EU ETS can be estimated by adding the ETS alone to the NOPOL case (comparing ETS with NOPOL), or by withdrawing the ETS from the OBS case (comparing RES with OBS). And, the same can be done

for the effect of RE injections. Panel b of Fig. 2 provides the estimates of emissions abatement for both the ETS and RES.10 Two features stand out. First, as implemented in Germany, RES policy is much more effective in reducing CO2 emissions within the German electricity sector than the EU ETS. The emission impact of the EUA price is in the range of 1 to 3% compared to an 11 to 20% impact of RE generation. Second, the abatement attributable to each instrument is greater when that instrument is deployed in conjunction with the other. Alternatively, when employed together, the abatement is greater than the sum of the individual abatement effects when each instrument is deployed individually. This reflects the reinforcing interaction effect that we have already noted and which will be discussed extensively in the next section of the paper. Table 1 provides the annual RE injections, the abatement attributable to those injections with and without the interaction effect (IA), and the corresponding average emissions intensity of the emission reduction per MWh of injected RE. CO2 abatement and the average intensity of the generation displaced by RE injections vary depending on whether the interaction term is included, i.e., whether the ETS CO2 price observed in these years is present, as also illustrated in Fig. 2.11 As can be seen, the average CO2 intensity abated varies considerably from year to year depending on the hourly interaction of load and RE injections and the ordering of dispatch as fuel prices vary (monthly in the model). When the 2007 observation for average abatement intensity is removed from the with-IA column (due to the non-existent CO2 price in that year), the annual variation is about 10% when the CO2 price is

10 Note that we refer to the total RE injection (wind, solar PV and biomass) as being completely attributed to the impact of RE policy within the analysis. This means that in the absence of an RE policy, the deployment (injection) of these RE sources would be zero. 11 The plant-specific intensity varies by fuel and plant efficiency: coal-fired power plants in the model have an average carbon intensity of 0.90 t-CO2/MWh and gas-fired plants one of 0.46 t-CO2/MWh.

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

the underlying principles. Second, the results of the German case study are used to further analyze this interaction effect.

Table 1 RE injections, CO2 abatement, and average abatement intensity. Year

2006 2007 2008 2009 2010

RE injection (TWh)

50.3 64.0 71.3 73.6 79.1

S153

CO2 abatement (Mt-CO2)

Average abatement intensity (t-CO2 per MWh RE injection)

with IA

without IA

with IA

without IA

35.3 39.3 54.0 57.5 59.8

33.3 39.2 49.0 53.7 55.6

0.70 0.61 0.75 0.78 0.76

0.66 0.61 0.69 0.73 0.70

present and about 20% when there is no carbon price (2007 is now included in the sample). These statistics are similar to those found by Novan (2011) and Kaffine et al. (2011) for CO2 abatement due to wind injections in Texas, which has a mixed configuration of coal and gas-fired power plants like that in Germany. Summing up, the results indicate that there is indeed an interacting impact of RE deployment and emission prices on the actual abatement values and not only on the overall allowance price. For Germany this interaction seems to be positive on average, meaning that the RE deployment in the years from 2006 to 2010 lead to a higher emission abatement due to the presence of the allowance prices than what would have been observed if no emission cap had been implemented. Recall that this CO2 abatement in Germany only displaces abatement elsewhere in the EU ETS because of the cap on aggregate emissions.

3. The interaction effect on abatement A salient feature of the results obtained in the above simulations is that for all years the abatement resulting from the use of both instruments together is greater than the sum of the abatement resulting from the use of either instrument alone. If generalizable, this result has significant policy implications. No longer would it be the case that overlapping instruments are either superfluous or redistributive only. There would exist an additional element, of uncertain importance, that should enter into policy deliberations. Fig. 3 depicts this interaction effect graphically. The second and third columns illustrate the abatement due to the ETS and RES when considered alone. The fourth column shows the interaction effect that occurs when both are used together. Methodologically, the same phenomenon occurs when working in reverse. If either RES or ETS is removed from the scenario in which both are present (OBS), the implied reduction of emissions due to that instrument is not that indicated by the columns showing their effect when used alone, but that impact plus the interaction effect. In the following sections we will explain why this behavior occurs in electricity systems and provide insights on the generalization of this effect. First a simplified power system is used to gain insight in

Fig. 3. Schematic representation of the observed interaction dependence.

3.1. Illustrating individual and interaction effects with the help of a methodological stacking model In this subsection, a simplified illustrative power system is used to gain more insight in the underlying principles of the observed interaction effect. This system has a generation mix that is representative for the existing capacity of the German system. It consists of one nuclear plant, one lignite plant, 5 coal fired power plants (varying in size and efficiency) and 6 gas fired power plants (also varying in size and efficiency). The two panels of Fig. 4 show a typical stacking order based on the marginal costs of these plants. The panel on the left reflects the merit order with no carbon price, while that on the right reflects the merit order with a €15/t CO2 price. On the right-hand side panel, the marginal generating cost for each segment consists of the fuel cost (bottom part, identical to the LHS panel) and the emission cost (upper part, which is higher for carbon intensive fuels like coal and lignite). The solid line on each panel reflects load in a particular hour without any RE injections, while the dashed line shows the net demand after RE injections in that hour for each of these two cases (with and without a carbon price). A carbon price changes the dispatch order moving gas-fired plants lower down in the merit order and moving coal plants up. This has an impact on the plants displaced by an RE injection. In the setting with no carbon price, mostly gas units are displaced whereas, if the carbon price is set to €15/t CO2, coal plants are moved up the dispatch order with a consequent greater probability of being displaced by the RE injection. Thus, when the carbon price re-orders dispatch in this manner, abatement from an identical RE injection is greater. Measuring abatement due to introducing a carbon price is a little more complicated since it depends on the amount of generation shifted between online and offline. If no RE injection is assumed so that the load would be that indicated by the solid line, a carbon price will both change the order in which coal and gas plants are dispatched to the left of the line and substitute some offline gas plants for some otherwise online coal plants. Only the latter substitution effect reduces emissions. If the RE injection is present, effective demand is indicated by the dotted line and there are many more gas plants offline in the absence of a carbon price (left-hand panel). Introducing a carbon price substitutes many of these offline gas plants for online coal plants. With more otherwise unutilized gas capacity moved online, the reduction of emissions from an identical carbon price is greater when the RE injection is present. Panel a of Fig. 5 shows the total emission level for the test system depending on the levels of the carbon price (0–25€/t) and RE injection (0–2500 MW). Naturally the highest emissions are observed with no CO2 price and no RE injections (a case similar to the previously considered NOPOL scenario). Increasing RE injections (moving along the x-axis) continuously reduces emissions regardless of the CO2 price, albeit at varying rates depending on the generation being displaced. Similarly, increasing the CO2 price also reduces emissions although not continuously reflecting the discrete generating units of this simple example. The other three dots indicate the points comparable with the scenario setting of the German model. The ETS and RES dots indicate the points where each instrument is used alone with the same values as in the example above. And the OBS dot represents the setting where both instruments are used simultaneously, similar to the observed case in the German model. Panel b of Fig. 5 shows a counterfactual without an interaction of ETS and RES. This surface is obtained by simply adding the individual ETS and RES effects (indicated by the upper edges in panel a, where the other instrument takes the value of zero) for the full ranges of each effect. The resulting surface, which would be obtained if there were no interaction, has a more regular shape than the preceding one.

S154

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

50

50 nuclear lignite coal natural gas demand demand after RES

marginal cost [euro/MWh]

40 35

45 40

marginal cost [euro/MWh]

45

30 25 20 15

35 30 25 20 15

10

10

5

5

0

0

2000

4000

0

6000

0

2000

power [MW]

4000

6000

power [MW]

Fig. 4. Merit order of the methodological system, 1500 MW of RE injection and CO2 price of zero (LHS figure) and 15 €/t (RHS).

The top two panels illustrate the dispatch order with and without a carbon price as in Fig. 4, only here the load is 4500 MW instead of 6000 MW. At the lower load represented by this particular hour and in the absence of a carbon price, the online plants are all coal and nearly all the offline plants are gas fired. When considered alone either the ETS or the RES cases, reduce emissions; however, when deployed together, abatement is now less than the sum of the two separate effects. For instance, and as can be readily seen in the top two panels of Fig. 6, the RE injection displaces a lot of gas when a carbon price is present instead of only coal. The resulting negative interaction effect is shown in the third

The interaction effect can now be represented by taking the differences in the values represented by the surfaces in panels a and b, as shown in panel c of Fig. 5. Again, the black dots indicate the four scenarios. The interaction effect is zero along the edges when either instrument is used alone, but the general tendency is to take on positive values as either is used in conjunction with the other. However, some areas of negative values are encountered at high levels of RE injection and low CO2 prices. Fig. 6 provides an illustration with this simple system of the conditions that would cause the abatement interaction effect to be negative.

(a) With interaction effect

(b) Without interaction effect

NOPOL

CO2 emission [ton/h]

RES

4500 4000 3500 3000 2500

OBS

2000 1500 0

0 500

1000

10 1500

2000

4500 4000 3500 3000 2500 2000 0

1500 0

500

20 2500

CO2 price [euro/ton]

RES injection [MW]

1000

10 1500

2000

RES injection [MW]

20 2500

CO2 price [euro/ton]

(c) Interaction effect CO2 emission [ton/h]

CO2 emission [ton/h]

ETS

800

OBS

600 400 200 0 −200 2000

RES 1000

RES injection [MW]

ETS

20

NOPOL

10

0

CO2 price [euro/ton]

0

Fig. 5. CO2 emission levels of the methodological system, with (a) and without (b) interaction, and interaction effect (c), all at demand level of 6200 MW.

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

(a) Merit order without CO2 price

(b) Merit order with CO2 price 50

nuclear lignite coal natural gas demand demand after RES

40

30

marginal cost [euro/MWh]

marginal cost [euro/MWh]

50

20

10

0

S155

0

1000

2000

3000

4000

5000

6000

40

30

20

10

0

7000

0

1000

2000

3000

power [MW]

4000

5000

6000

7000

power [MW]

(c)Interaction effect

NOPOL

CO2 emission [ton/h]

ETS RES

0

−500

OBS −1000 0 500 1000 1500 2000 2500

RES injection [MW]

25

20

15

10

5

0

CO2 price [euro/ton]

Fig. 6. Merit order and interaction effect, demand level 4500 MW.

panel of Fig. 6 which replicates the third panel of Fig. 5 for this lower demand case. The same negative interaction effect can be observed if RES penetration increases to a level that it is displacing gas plants that have been moved down in the dispatch order by the carbon price. While initially the interaction effect will be positive as the RE generation displaces coal plants moved up in the dispatch order, once the gas plants moved down in the dispatch order are displaced by higher levels of injection, the interaction becomes negative. This can be most easily visualized by referring to Fig. 4 and considering the effects on conventional generation and emissions as the dotted line is moved to the left. In effect, when considered continuously (instead of discretely as in these illustrations), different segments of the dispatch order will have positive, zero, or negative values depending on the effect of the carbon price on the dispatch order. In a typical mixed coal-gas system with higher gas than coal prices, the positive values for the interaction term will initially dominate, but as RE injections increase, more negative segments will be encountered and the resulting value for the interaction term will diminish and it could even turn negative. Such a case would occur if the RE injection were great enough to displace all fossil generation. Suppose that emissions are 100 and a carbon price alone reduces emissions by 20. An RE injection great enough to displace all emitting generation would reduce emissions by 100 when used alone, but by only 80 when a carbon price is present. The difference is the negative interaction effect, which in this limiting case offsets completely the abatement from the carbon price when considered alone and even though at initial levels the interaction effect was positive. In other words, the interaction effect will necessarily be the negative equivalent of the ETS impact if RES replaces all fossil generation.

More generally, the sign and value of the interaction term depend not only on the level of carbon price and RE injection, but also and as importantly on the mix of power plants, the price spread between gas and coal prices, and the demand placed on the system, all of which have a significant impact on the merit order and therefore the possibilities to obtain a positive interaction of both instruments. While no general assessment of the sign and value of the interaction effect can be made, the summed annual effect will generally be positive when the CO2 price puts coal on the margin in place of gas and the RE injection is not so large that it is displacing gas plants that have been moved down the merit order by the CO2 price and replaced coal. As we have shown, when the combination of load and RE injection is such that the interval of displaced conventional generation includes more segments in which the CO2 price has caused gas to replace coal, the positive interaction effect disappears. Eventually, as the RE injections increase in volume and displace increasing amounts of fossil generation, the interaction term will become increasingly negative. Thus, a positive interaction effect will always disappear if the RE injection is large enough. 3.2. Interaction effect in the German system We now move back from the illustrative system to our model of the German system to further explore the observed interaction effect in more detail. The simulations as presented in Section 2 feature the use of pumped storage endogenously and consider unit commitment restrictions not included in a simple stacking model. For any specific hour, electricity generation can differ between different scenarios (OBS, RES, ETS, NOPOL), due to a different use of pumped storage and

S156

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

abatement [kton/h]

(a) Merit order abatement by RES with ETS in place (2010) 30 20 10 0

0

1000

2000

3000

4000

5000

6000

7000

8000

interaction [kton/h]

abatement [kton/h]

(b) Merit order abatement by RES in absence of ETS (2010) 30 20 10 0

0

1000

2000

3000

4000

5000

6000

7000

8000

7000

8000

(c) Corresponding interaction effect (2010)

5

0

−5

0

1000

2000

3000

4000

5000

6000

hour [h] Fig. 7. Hourly abatement by RES, in case with ETS in place (a) and in absence of ETS (b), and corresponding interaction effect (c).

changes in the merit order due to unit commitment restrictions. Hence, a single interaction effect for each hour cannot be calculated from the hourly emissions of these four scenarios.12 Instead, it must be determined by setting up hourly merit orders as has been done for the illustrative system.13 The interaction effect determined this way considers every hour separately and is therefore not necessarily what would be obtained when the power system dynamics involved in operating an electricity network are taken into account. We will denote this interaction without consideration of power system dynamics in the remainder of this text as “pure” or merit order based interaction, based on the German data (demand, power plant efficiencies and fuel and CO2 prices) and simulation results (hourly electricity generation on power plant basis). In what follows, the first subsection presents and discusses the course of the pure, hourly interaction effect over the year, while the second subsection contrasts that result with the overall interaction effect (obtained from the four simulated scenarios, when power system dynamics are taken into account).

3.2.1. Variation of the “pure” interaction effect over the year Given the high variability of load, as well as RE injections, it seems plausible that the interaction effect could change sign and vary considerably over the course of any year. Fig. 7 presents the results for 2010 12 However, when aggregating on a yearly basis, the difference in pumping more or less nets out due to the similar absolute usage in all four scenarios, making the aggregated results for abatement and interaction meaningful. 13 The procedure to determine the hourly interaction effect works as follows. First, hourly merit orders are set up with the input data and the generation results from the ETS case (i.e., case with EUA price and without RE injection). The hourly abatement by RE injection is then approximated by calculating the emissions of the generation that would be displaced (according to the merit order) by the RE injection. Second, the same is done (merit order set up and abatement calculation) for the NOPOL case (i.e., no ETS price and no RE injection). The hourly interaction effect is then calculated as the difference between these two abatement levels (the abatement of RE injections when ETS is in place and the abatement of RE injections in absence of ETS prices).

when each hour is treated independently as explained above. Panel a shows the abatement due to RE injections when the ETS is present, panel b the abatement due to RE injections for the same hours when the ETS is not present, and panel c the difference between the two, i.e., the pure interaction effect. As is evident the interaction effect changes sign frequently. Two hours, one with positive interaction effect and the other with a negative effect, are selected for further analysis. These are hours 6644 (Oct. 4; 19:00–20:00) with the positive interaction and 6454 (Sept. 26; 21:00–22:00) with a negative interaction and they are indicated by the circles in panel c of Fig. 7. The hours differ in their respective peak load and RE injection. Hour 6644 has a higher residual load (60 GW vs. 53 GW) and a higher RE injection (12 GW vs. 7 GW). In the higher load case, several gas units are needed to satisfy demand regardless of the carbon price. However, with the carbon price in place, some of the more modern gas fired plants switch places with coal units and move further to the left in the merit order, thereby increasing the share of coal plants close to the margin. Those are then replaced by RE injection leading to a higher abatement than in a setting without carbon price in which mostly gas units would be at the margin. In the other selected hour (6454), the lower load level is met mainly by coal-fired units in the absence of a carbon price. When a carbon price is introduced some of the coal units are replaced by gas. Adding the RE injection in this case now replaces a mix of coal and gas units whereas it would replace mostly coal units in the case without a carbon price. Thus, when a carbon price is present, the RE injection abates less than it would have without the carbon price, thereby creating the negative interaction effect. 3.2.2. The interaction effect with system constraints and comparison for Germany 2006–2010 The discussion under Section 3.2.1 has treated each hour as if that hour was independent of dispatch in the preceding hours or expected dispatch in the succeeding hours. In any real electricity generating system with start-up restrictions, minimum down time constraints,

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

S157

Table 2 Impact of system dynamic interaction on the estimates for Germany (Mtons CO2). 2006 Abatement by RES With ETS Without ETS

With system dynamics Without system dynamics With system dynamics Without system dynamics

Interaction effect Interaction term with system dynamics Interaction term without system dynamics (= pure interaction effect) Effect of modeled system dynamics

2007

2008

2009

2010

34.06 36.05 31.73 35.13

37.31 40.97 37.14 40.79

52.64 53.97 47.60 51.56

56.04 58.06 51.83 56.17

58.30 59.52 53.96 58.33

2.33 0.92

0.17 0.19

5.04 2.42

4.21 1.88

4.34 1.19

1.42

−0.02

2.63

2.33

3.15

and pumped storage utilization, the pure interaction effect (as illustrated in the previous section) is clearly not the end of the story. For example, keeping coal units online during the night due to start-up constraints may affect the share of coal replaced by RE injection (different from what a simple one-shot merit order might suggest). The question arises then of the extent to which these system dynamics change abatement from RE injections and carbon pricing and how they affect the interaction effect that we have observed. When aggregated over all hours of the year, the pure interaction can be compared with the interaction effect obtained from the four scenario runs of the German simulation model, which include the system dynamics (see Section 2). The difference between this aggregated pure interaction and that obtained from the model runs provides an indication of the impact of system dynamics on the interaction. Table 2 sums the relevant measures for all hours for the considered years, 2006–2010, and provides results for the abatement by RE deployment with and without system dynamic constraints.14 The results for Germany indicate that the aggregate, annual interaction effect is positive for all years, although negligible in 2007 when the carbon price is low. The estimates for the share of the pure interaction effect and the influence of system dynamics vary over the years and indicate that both are significant. Again, the interaction effect is highly dependent on specific power system characteristics and the operating context (fuel prices, load, etc.); but there should be no doubt that these effects exist and that they must be taken into account in estimating the effect of RE policies and the ETS in reducing emissions. 4. Conclusion This paper started out as an exploration of the extent to which RE incentives have reduced the demand for allowances from the German electricity sector in actual practice. The answer to that question is more complicated than might be expected because abatement resulting from any given injection of RE generation depends on whether a carbon price is present. In general, a carbon price changes the merit order and thereby causes abatement from RE injections to be greater or smaller than it would otherwise be, depending on the change in merit order within the interval of generation displaced by the RE injection. We have termed this extra abatement an interaction effect, but it must be kept distinct conceptually from price interaction that has received attention in the literature. In the case of the German electricity system for the years 2006 through 2010, RE policy reduced CO2 emissions from the electricity sector between 33 and 60 Mtons, or 10% to 18% of what we estimate emissions would have been without any RE policy, depending on the

14 In the calculation of these values, combined heat and power is excluded, as the output profile of these plants is largely exogenous to the optimization. Hence, the values (with system dynamics) can differ slightly from the corresponding ones of Table 1.

year and whether the CO2 price was present or not. The German electricity sector accounts for about 15% of the emissions included within the EU ETS so that even an 18% reduction in the demand for allowances from this component translates into a system-wide reduction in demand of 2.7%. This may seem small, but when the abatement required by the cap is measured in the single digits as a percentage of what emissions would have been without a carbon price, the effect on price could be significant. For instance, and as a matter of simple geometry, if the required abatement system-wide is 5% and the marginal abatement cost curve is linear, EUA prices would be about 50% higher than what has been observed. However, such estimates are speculative absent reliable estimates of the system-wide marginal abatement cost curve, even if there are reasonable estimates of what emissions would have been in the absence of climate policy. When measured on an annual basis, the observed interaction effect is consistently positive for the German electricity system and on the order of 0.5% to 1.5% of emissions for all years, except in 2007 when the CO2 price was effectively zero. Given our estimates of No-Policy emissions, this additional effect increases the CO2 abatement due to RE incentives by 4% to 10% and that of the EUA price by about 70% to 130%, depending always on the circumstances of a particular year. However, when the year is broken down into its component hours, the interaction effect varies widely and changes sign frequently. Further analysis of the determinants of this interaction effect shows that its sign and value depend on the magnitude of the RE injection, the load in each hour and the effect of the carbon price on the merit order within the interval of generation that is displaced by the RE injection. The summed annual interaction effect turns out to be positive because the CO2 price more often replaces gas with coal than the reverse in the generation intervals that were displaced by the RE injections in these years. All research operates under limitations usually reflecting model structure and various maintained assumptions. One limitation of our research is that the CO2 price is not endogenous to the model. In effect, we treat the observed CO2 price, which reflects the demand-suppressing effect of actual ETS-wide RE injections, as a fixed price – equivalent to a tax – that does not vary with factors that would otherwise increase or lower the demand for allowances. Another limitation of our analysis concerns system interactions that are unique to electricity systems. Most of our explanation of the interaction effect is conducted on the assumption that each hour stands alone and that system constraints and characteristics do not cause the generation in any given hour to depend on generation in preceding or succeeding hours. However, the simulation model used here respects unit commitment and the effect of pumped storage in smoothing peaks and valleys. When these constraints are modeled, the aggregate annual interaction effect turns out positive. We find that in general more than half of this aggregate is due to system dynamics rather than pure merit order based effects. Finally, for all the interest that this abatement interaction effect holds, it is relatively small at least at the CO2 prices that have been observed. Our analysis suggests that it would be larger with higher

S158

H. Weigt et al. / Energy Economics 40 (2013) S149–S158

CO2 prices, but even so, the bigger factor in explaining the reduced demand for allowances is the direct reduction of CO2 emissions by RE injections. These injections are more effective in reducing CO2 emissions than the EUA price, at least at the levels observed so far in the EU ETS. Whereas a carbon price tends to substitute lower emitting gas generation for coal generation, RE injections displace whatever is on the relevant margin with a zero-CO2-emitting source. The displacement occurs without regard to the carbon content of the displaced generation, but in existing electricity systems this is nearly always CO2-emitting fossil generation. Hence, the large abatement effect and consequent reduction in the demand for allowances within the EU ETS. Acknowledgments H. Weigt partly carried out the research reported in this paper being supported as a Jean Monnet Fellow at the Florence School of Regulation (FSR) at the European University Institute. D. Ellerman acknowledges support through Service Contract 071303/2011/609668/SER/Ar from the European Commission. E. Delarue is a research fellow of the Research Foundation Flanders (FWO). The authors would further like to thank Jan Abrell, as well as the participants and discussants at the Loyola de Palacio Chair workshop “The Cost of Renewable Energy in the EU,” the EUI's Annual Climate Policy Research Conference, the 5th Atlantic Workshop on Energy and Environmental Economics (AWEEE) in A Toxa, Spain, and the inhouse seminar of the FSR for their helpful comments. We also thank 3 anonymous referees for their further remarks and suggestions. Appendix A Unit commitment model: cost ¼

min

gp;t ≥ 0

X

X

mcp;t g p;t þ

p;t

down

g p;t þ PSP t

X

max

scp g p

upp;t

objective

ð1Þ

p;t

up

¼ dt þ PSP t

∀t∈T

energy balance

ð2Þ

p

min

onp;t g p

max

≤g p;t ≤onp;t g p

upp;t ≥onp;t −onp;t−1 X

∀t ∈T; p∈ P

capacity constraint

∀t ∈ T; p ∈P ð4Þ

tþsdp

onp;t−1 −onp;t ≤1−

ð3Þ

upp;τ

∀t ∈T; p ∈P

start−up constraints

τ¼t

down

PSP tþ1 ¼ PSP t −PSP t up

PSPt ≤PSP

up

þ ηPSP PSP t

up max

∀t∈T

PSP balance

∀t ∈ T

h i dowm down max PSPt ≤ min PSP ; PSP t

ð5Þ

ð6Þ

∀t ∈ T

PSP capacity constraints

with Indexes and set

Variables

p P t T τ T

g up on PSP

Plants Time periods Time periods for start-up duration

Generation Start-up [binary variable] Online/offline [binary variable] Pumped storage generation/level

Parameters

Superscripts

η mc sc d

max up down

PSP efficiency Marginal generation costs Start-up costs Demand

Upper limit PSP in pump mode PSP in generation mode

References Abrell, J., Weigt, H., 2008. The interaction of emissions trading and renewable energy promotion. Dresden University of Technology — Chair for Energy Economics and Public Sector Management Working paper, WP-EGW-05. BAFA, 2011. Aufkommen und Export von Erdgas sowie die Entwicklung der Grenzübergangspreise, Entwicklung der Rohöleinfuhr, and Drittlandskohlepreis. Federal Office of Economics and Export Control (BAFA) (http://www.bafa.de). Böhringer, C., Rosendahl, K.E., 2009. Green serves the dirtiest: on the interaction between black and green quotas. Cesifo Working Paper No. 2837. Brooke, A., Kendrick, D., Meeraus, A., 2008. GAMS A User's Guide. Washington, GAMS Development Cooperation. Capros, P., Mantzos, L., Parousos, L., Tasios, N., Klaassen, G., Van Ierland, T., 2011. Analysis of the EU policy package on climate change and renewables. Energy Policy 39, 1476–1485. Cullen, J.A., 2011. Measuring the Environmental Benefits of Wind-Generated Electricity. Unpublished working paper dated October 2011 available at: http://www. josephcullen.com/resources/measuringwind.pdf. De Jonghe, C., Delarue, E., Belmans, R., D'haeseleer, W., 2009. Interactions between measures for the support of electricity from renewable energy sources and CO2 mitigation. Energy Policy 37, 4743–4752. Delarue, E., Luickx, P., D'haeseleer, W., 2009. The actual effect of wind power on overall electricity generation costs and CO2 emissions. Energy Convers. Manage. 50, 1450–1456. Delarue, E., Ellerman, D., D'haeseleer, W., 2010. Short-term CO2 abatement in the European power sector: 2005–2006. Clim. Change Econ. 1, 113–133. Denny, E., O'Malley, M., 2006. Wind generation, power system operation and emissions reduction. IEEE Trans. Power Syst. 21, 341–347. Denny, E., O'Malley, M., 2007. Quantifying the total net benefits of grid integrated wind. IEEE Trans. Power Syst. 22, 605–615. Ellerman, A.D., McGuinness, M., 2008. CO2 abatement in the UK power sector: evidence from the EU ETS trial period. MIT-CEEPR Working Paper 2008-010. MIT Centre for Energy and Environmental Policy Research. ENTSO-E, 2011. Hourly Load Values, European Network of Transmission System Operators for Electricity. https://www.entsoe.eu/. Eurelectric, 2010. Power Statistics 2010 Edition Full Report. Union of the Electricity Industry — Eurelectric, Brussels Belgium. Fischer, C., Newell, R.G., 2008. Environmental and technology policies for climate mitigation. J. Environ. Econ. Manag. 55, 142–162. Fischer, C., Preonas, L., 2010. Combining policies for renewable energy. Is the Whole Less than the Sum of Its Parts? Discussion Papers Dp-10-19 (Resources For the Future). Fürsch, M., Malischek, R., Lindenberger, D., 2012. Der Merit-Order-Effekt der erneuerbaren Energien - Analyse der kurzen und langen Frist. EWI Working Papers 2012-14. Energiewirtschaftliches Institut an der Universitaet zu Koeln. Grubb, M., Jamasb, T., Pollitt, M., 2008. A low-carbon electricity sector for the UK: what can be done and how much will it cost? In: Grubb, M., Jamasb, T., Pollitt, M. (Eds.), Delivering a Low-Carbon Electricity System: Technologies, Economics and Policy. Cambridge University Press, Cambridge, pp. 462–497. Hindsberger, M., Nybroe, M.H., Ravn, H.F., Schmidt, R., 2003. Co-existence of electricity, TEP and TGC markets in the Baltic Sea Region. Energy Policy 31, 85–96. IPCC, 2006. Guidelines for National Greenhouse Gas Inventories. In: Eggleston, H.S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), Institute for Global Environmental Strategies, Japan. Jensen, S.G., Skytte, K., 2003. Simultaneous attainment of energy goals by means of green certificates and emission permits. Energy Policy 31, 63–71. Kaffine, D.T., McBee, B.J., Lieskovsky, J., 2011. Emissions savings from wind power generation: Evidence from Texas, California, and the Upper Midwest. Unpublished working paper available from the authors. Kalkuhl, M., Edenhofer, O., Lessmann, K., 2012. Learning or lock-in: optimal technology policies to support mitigation. Resour. Energy Econ. 34, 1–23. Lang, P., 2009. Cost and Quantity of Greenhouse Gas Emissions Avoided by Wind Generation. Unpublished paper, dated 16 Feb 2009, available at: http://bravenewclimate. files.wordpress.com/2009/08/peter-lang-wind-power.pdf. Linares, P., Santos, F.J., Ventosa, M., 2008. Coordination of carbon reduction and renewable energy support policies. Clim. Pol. 8, 377–394. Meran, G., Wittmann, N., 2008. Green, Brown, and Now White Certificates: Are Three One Too Many? A Micromodel of Market Interaction, German Institute of Economic Research (DIW Berlin) Discussion Paper 809. Available at: http://www.diw.de/ documents/publikationen/73/diw_01.c.88343.de/dp809.pdf. Morthorst, P.E., 2001. Interactions of a tradable green certificate market with a tradable permits market. Energy Policy 29, 345–353. Novan, K., 2011. Valuing the wind: renewable energy policies and air pollution avoided. University of California Center for Energy and Environmental Economics (UCE3) Working Paper, E3 WP-027 (November 2011). Skytte, K., 2006. Interplay Between Environmental Regulation and Power Markets, EUI Working Paper RSCAS No. 2006/04 (February). Available at: http://www.eui.eu/ RSCAS/WP-Texts/06_04.pdf. Sorrel, S., Sijm, J., 2003. Carbon trading in the policy mix. Oxford Rev. Econ. Policy 19, 420–437. Umweltbundesamt, 2011. Datenbank Kraftwerke in Deutschland. Umweltbundesamt (http://www.umweltbundesamt.de). VGE, 2005. Jahrbuch der Europäischen Energie- und Rohstoffwirtschaft 2005. Verlag Glückauf GmbH, Essen. VGE, 2006. Jahrbuch der Europäischen Energie- und Rohstoffwirtschaft 2006. Verlag Glückauf GmbH, Essen. VGE, 2009. Jahrbuch der Europäischen Energie- und Rohstoffwirtschaft 2009. Verlag Glückauf GmbH, Essen. Weigt, H., Delarue, E., Ellerman, A.D., 2012. CO2 abatement from renewable energy injections in the German electricity sector: does a CO2 price help? MIT-CEEPR Working Paper 2012-003. MIT Centre for Energy and Environmental Policy Research.