Quantifying the net cost of a carbon price floor in Germany

Energy Policy 109 (2017) 685–693

Contents lists available at ScienceDirect

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

Quantifying the net cost of a carbon price floor in Germany a,b,⁎

Philipp Egli a b c

MARK

c

, Oskar Lecuyer

University of Bern, Department of Economics and Oeschger Centre for Climate Change Research, Hochschulstrasse 6, 3012 Bern, Switzerland Elektrizitätswerk der Stadt Zürich (EWZ), Tramstrasse 35, 8050 Zürich, Switzerland Agence Française de Développement (AFD), 5 Rue Roland Barthes, 75598 Paris CEDEX 12, France

A R T I C L E I N F O

A BS T RAC T

Keywords: Electricity markets Carbon price floor Germany Merit order curve EEG

The German energy and climate policy mix is failing to decarbonize electricity production until now, with only 6% overall CO2 emissions reductions since 2005. Using empirical methods and hourly market data, we estimate the aggregate supply curve of the German power market and simulate the effect of a 20€/tCO2 and 40€/tCO2 carbon price floor on the German power market and on the renewable subsidy scheme. With the 40€/tCO2 carbon price floor, median prices increase by 37€/MWh and average price peaks by 50€/MWh. At the wholesale level, the market's annual volume increases by some €18 billion to €39 billion. At the retail level, however, the net cost to consumers is moderated due to costs savings from the renewable subsidy scheme worth some €4 billion, or roughly one-fifth. The same ratio applies to a price floor at 20€/tCO2.

1. Introduction Between 2005—the start of the EU emissions trading scheme (EUETS)—and 2015, yearly CO2 emissions from the German power sector fell by 6% overall, with periods of increasing emissions (Agora, 2016). At this pace, decarbonization of the German power sector will be achieved in the middle of the 22nd century. Since the economic downturn of 2009, the EU-ETS has been structurally out of balance, with an ongoing surplus of nearly one year worth of allowances (Koch et al., 2014). More recently, the EU concluded that current reduction efforts are insufficient to reach the 2030 target of reducing emissions by 40% below 1990-levels and therefore introduced further measures known as backloading and market stability reserve (EEA, 2015; EU, 2015; Sandbag, 2015). However, Koch et al. (2016) show that prices actually fell on news from the backloading decision process, implying that announced measures are less stringent than expected. In addition to climate effectiveness, the current German energy and climate policy mix seems to under-achieve on at least two other dimensions, namely energy affordability and system adequacy. In

⁎

Germany, as in most European countries, renewable energy sources (RES) are promoted through a national feed-in-scheme, termed “EEG”, which compensates producers for the difference between market prices and their higher production costs. Although such schemes can be justified on various grounds, e.g. local or dynamic benefits (Lehmann and Gawel, 2013; Lecuyer and Quirion, 2013), the resulting policy overlap with the EU-ETS is often criticized for its inefficiency. Cludius et al. (2014) highlight the rising cost burden on households, as the EEG is financed through a levy on final electricity consumption.1 Moreover, the additional wind and solar generation fed into the system exert downward pressure on spot market prices. This lowers the economic viability of systematically important back-up generation (Würzburg et al., 2013; Traber and Kemfert, 2011; Hildmann et al., 2011), and undermines system adequacy in the long run. The experience and developments outlined above cast doubt whether the current climate policy setup can provide sufficient incentives to reduce emissions from power supply to almost zero by 2050, as set out by the EU's Roadmap (EU, 2012). It is well known that tax-like instruments are best suited to deal with the problem of climate

Corresponding author. E-mail address: [email protected] (P. Egli). E-mail address: [email protected] (O. Lecuyer). 1 At the beginning of 2016, the EEG-levy amounts to roughly three times the wholesale market price for electricity (Agora, 2016).

http://dx.doi.org/10.1016/j.enpol.2017.07.035 Received 27 June 2016; Received in revised form 5 July 2017; Accepted 16 July 2017 0301-4215/ © 2017 Elsevier Ltd. All rights reserved.

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

Germany in 2012. We use a model developed by He et al. (2013), hereafter called HHHA, to fit a deterministic supply curve to these price-load observations by using nonlinear least squares. The general idea is to estimate the aggregate heat rate curve of the entire power market from the observed price-load combinations by controlling for primary energy and CO2 prices.4 This empirically derived heat-rate curve is assumed to be static, which then allows to simulate the market outcome with other input prices, e.g. a higher carbon price. This approach has a substantial advantage over the state-of-the-art optimization models, because it does not require the mapping of individual power plants and their constraints (cost structure, ramping, etc.). Instead of reconstructing the entire market bottom-up, we derive it from historic market data. This allows us to capture real world market imperfections, such as strategic bidding, and complex technical constraints leading, among other things, to hours with prices above marginal costs or in negative territory. Thanks to its empirical foundation, the model captures and replicates such extreme situations. These market outcomes are rare, more complex to explain fundamentally and often pose a challenge for technically explicit bottom-up models. The main limitation of our approach is the assumption of a static heat-rate curve and thus a fixed supply side. By abstracting from the possibility of fuel-switching in the short term and from plant replacements in the longer term, we neglect the cost-minimizing behavior on the supply side. This means our results should be interpreted as an upper ceiling of the probable effects of a carbon price floor.

change (Newell and Pizer, 2003), and if not available, that hybrid instruments provide useful safety valves to guarantee against unbearably high costs (Roberts and Spence, 1976; Burtraw et al., 2010).2 Ironically, the lesson learned from the EU-ETS led other jurisdictions to implement price floors and ceilings from the start (World Bank, 2015; Newell et al., 2013; Goulder, 2013), but has not yet been considered at the EU-level until recently. However, the United Kingdom (UK) introduced a carbon price floor in 2013, in the form of a complimentary tax on emissions if the EU-ETS price falls below a predetermined floor (UK Parliament, 2013), and France is considering a price floor starting at 30€/tCO2. The idea of discretionary price management mechanisms recently gained traction among large utilities and some member states in the form of a reserve price in the auctions for emissions allowances in the power sector. Reserve prices are an important feature of good auction design (Ausubel and Cramton, 2004): if the market clearing price of the auction falls below the price floor, the unsold allowances automatically restrict the supply of allowances and support the carbon price.3 A carbon price floor affects marginal costs of fossil-fired plants and thereby wholesale power prices. Depending on its stringency, fuelswitching occurs in the merit order (Delarue et al., 2008), implying significant emissions reductions. Moreover, with higher wholesale prices, the costs of promoting RES decreases. We use a novel approach, proposed by He et al. (2013), to quantify the impacts of a carbon price floor on the German electricity market. We estimate the yearly supply curve of the German wholesale market using the 8760 hourly price-load observations. By controlling for fuel and carbon prices, we derive the aggregate heat rate curve of the market, a fundamental power market characteristic. This allows us to simulate the market outcome with a fabricated price vector, e.g. a higher carbon price, and to estimate the interaction with the EEG subsidy-scheme. This empirically founded fundamental approach blends accuracy (relying on actual data), with flexibility (due to the model's fundamental characteristics). Using data for the year 2012, we find that a carbon price floor of 20€/tCO2 increases median prices by 15€/MWh to 57€/MWh and peak prices, taken at the 90th percentile, by roughly 19–79€/MWh. With a price floor of 40€/tCO2, median and peak prices increase to 79€/MWh and 111€/MWh respectively. From a policy-maker's pointof-view, the carbon price floor of 40€/tCO2 increases the annual volume of the wholesale market by some €18–€39 bn. At the retail level, however, the net cost to consumers is moderated due to cost savings from the EEG-subsidy scheme worth some €4 bn, or roughly one-fifth. The same ratio applies to the lower price floor at 20€/tCO2. To our knowledge, this is the first study to assess and quantify the systematic interaction between the two main climate policy instruments. The remainder of this work is structured as follows: Section 2 describes the data, modeling and simulation. Section 3 presents the results, which are then discussed in Section 4. Section 5 concludes.

2.1. Data Daily primary energy, CO2 and hourly spot market price data are taken from European Energy Exchange (EEX) or Gaspool (GPL). Hourly load, RES-infeed,5 net exports, unavailability due to planned and unplanned outages of power plants are taken from the transparency platforms of EEX, the European Network of Transmission System Operators for Electricity (ENTSO-E) or directly from the four TSOs operating in Germany. Hourly vertical load data represents the flows from the highest level transmission grid to the lower grid levels. As virtually no RES plants are directly connected to the transmission grid (Bundesnetzagentur, 2015), vertical load represents the residual part of consumption that needs to be covered by conventional plants, after subtracting the generation of RES. By contrast, common supply represents the entire national consumption. All time series are converted to an hourly resolution, generating 8760 hourly observations per year. Table 1 reports summary statistics and sources, Fig. 1 depicts the price series. 2.2. Modeling HHHA propose several deterministic model specifications for the aggregate supply curve. We test all their specifications with our newer data. For shortness, we only present the two extreme cases, the simple exponential model and the fuel-adjusted heat rate model. In terms of fundamentals, two parameters that affect the economics of power plants and thus the slope of the merit order curve are the CO2 content of a given fuel input, ϕj, and the heat rate of a power plant type, λj, where (j = lignite , coal , gas ). The heat rate represents the amount of thermal primary energy required to generate one unit of electricity and is therefore the inverse of efficiency. The product of heat rate and CO2

2. Methods We estimate a yearly supply curve—or yearly merit order curve—on the base of the 8760 observations of hourly price-load combinations for 2 The current regime is highly asymmetric, with a control mechanism to guard against unbearably high costs (dubbed “Measures in the Event of Excessive Price Fluctuations” in the directive), but no mechanism to safeguard against too low prices. In practice, in almost all cap-and-trade programs the costs to firms have been overestimated ex ante (Lecuyer and Quirion, 2016). 3 Other propositions included the possibility to store allowances in a reserve (i.e. the MSR) and release them to the market to manage the total number of allowances in circulation Taschini (2013). As discussed by Burtraw et al. (2014), in the California trading program, allowances not sold at the price floor are withheld from the market until the price floor is exceeded for two consecutive auctions and then are incrementally added back into the market.

4 To be more precise, the model determines the market-implied merit order. Given that it is derived from observed price-load combinations, there is a layer of residual factors (e.g. mark-ups) on top of the typical stack of marginal costs. These residual factors are omitted in standard bottom-up methodology. 5 Here, RES refers only to the intermittent solar and wind energy production, excluding biomass, hydro-power or other forms of controllable renewable energy sources.

686

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

Table 1 Summary statistics of hourly time-series for Germany 2012/2013.

Table 2 Physical and technical parameters.

Time series

Source

Unit

Min.

Max.

Mean

Spot Prices S(t) Coal Prices Pcoal(t ) Gas Prices Pgas(t) CO2 Prices PCO2(t )

EEX EEX EEX, GPL EEX

€/MWhel €/MWhth €/MWhth €/tCO2

−221.99 7.90 19.71 2.72

210.00 12.41 40.27 9.31

40.19 9.59 26.17 5.92

Vertical Load LV(t) Common Supply LCS(t) Wind Infeed W(t) Solar Infeed R(t) Net Exports NX(t) Unavailabilities Unuc&lig(t )

TSO ENTSOE TSO TSO ENTSOE EEX

MW MW MW MW MW MW

5225 29201 130 0 −8188 9560

54119 75623 26041 23952 13357 23705

32713 53177 5303 3283 2768 14350

Fuel

CO2 content of fuel ϕj [tCO2/ MWhth]

Heat rate λj [MWhth/ MWhel]

Efficiency 1/ λj

CO2 intensity of power Ψj [tCO2/ MWhel]

Gas Coal Lignite

0.202a,b 0.339b 0.404b

1.98 2.54–2.73 2.86

50 37–39 35b

0.400c 0.860–0.925c 1.161b

Note: Sources hereafter, missing values based on own calculations using Ψj = ϕj *λj . a b c

From (IPCC, 2005, 194) and converted using 1 kWh = 3.6 MJ. From (UBA, 2013, 8). From (IEA, 2013, 43).

plants, L CD(t ).6 The conventional load characteristic is crucial because marginal costs of RES production are close to zero and their production is subsidized and prioritized. Thus, while spot market prices are indeed influenced by RES production, the supply curve does not reflect their marginal costs. Conventional domestic load L CD(t ) is constructed by adding vertical load LV(t), net exports NX(t) and outages of baseload plants Unuc & lig(t ):

L CD(t ) = L V (t ) + NX (t ) + Unuc & lig(t ).

(1)

The inclusion of net exports is necessary, because over time Germany oscillates between being a net importer and being a net exporter. Baseload plants are designed to run throughout the year and their outages can only be compensated by costlier mid or peak load plants.7 An extreme example is the nuclear phase out in Germany that was initiated in the spring of 2011 and led to an abrupt reduction of roughly 9GW of baseload capacity (EEX, 2014). The outage of baseload plants shifts the entire merit order curve to the left, or equivalently shifts the load curve to the right by the respective amount of outages for a given hour. Conventional domestic load is normalized by its maximum:

lCD(t ) =

L CD(t ) L CDmax(t )

(2)

In terms of the supply side, the simple exponential model, called EXPO hereafter, defines spot prices S(t) as an exponential function of load lCD(t) and an error term ε(t ):

S (t ) = e

lCD(t )− a b

+ ε(t ),

(3)

where a represents the horizontal shift and b is the scaling factor. The EXPO model captures a major characteristics of wholesale prices, namely the upward-sloping merit order, indicating that higher load levels require costlier plants to go online (Geman, 2005). However, this specification does not allow negative spot prices and ignores the possibility of a changing fuel price vector, thereby limiting its application to the very short-term. To improve on these shortcomings, the fuel-adjusted heat rate model, hereafter named FAHR consists of the product of the estimated heat rate curve fhr (lCD(t )) and a fuel-adjustment Pfuel(t):

Fig. 1. Primary energy, CO2 and baseload prices for Germany, including their yearly moving average (MA). Own elaboration based on data from (EEX, 2014).

S (t ) = fhr (lCD(t ))*Pfuel(t ) + ε(t ).

(4)

The fuel-adjustment accounts for the prices of coal Pcoal(t) and CO2 and the carbon intensity of coal ϕcoal:

Pfuel(t ) = Pcoal(t ) + PCO2(t )*ϕcoal .

content yields the overall CO2 intensity of a given electricity generation technology, Ψj. Table 2 compiles a selection of figures on typical heat rates, plant efficiency, CO2 content and intensity. Depending on the source, there are slightly differing estimations or definitions for typical characteristics. On the demand side, the raw load data requires some adjustments in order to represent the load served by conventional domestic power

(5)

Coal plants constitute the dense median part of the merit order curve, situated between lignite and gas-fired plants. Thus, coal prices are 6

Here conventional means all generation technologies except solar and wind. By contrast, peak-load plants have very limited operating hours and the outage of a particular peak-load plant is thus not strictly price increasing. 7

687

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

2.3. Simulation The FAHR model allows to simulate how the supply curve changes when introducing an arbitrary CO2 price. By doing this, we assume a fixed aggregate heat rate curve: there are no changes in the technical efficiency, the operational hours per plant type or the power plant portfolio. We also assume that the bidding behavior and the passthrough of CO2 costs stays the same.10 Building on (4) and given the CO2 content of primary energy and typical heat rates for the different generation technologies presented in Table 2, it is possible to theoretically derive the CO2 price at which a cleaner technology, for example gas, becomes cheaper than its dirtier rival technology, say coal. The implicit assumption behind the FAHR model is that the marginal cost MCj of a given generation technology j = (lignite , coal , gas ) can be approximated through the multiplication of its heat rate, and its fueladjustment. Rewriting (4) and (5) more compactly and using the technical heat rate, λj, the marginal cost of a given technology j is:

Fig. 2. Model specifications for 2012.

Table 3 In-sample performance of model specifications for 2012. 2

Specification

N

R

EXPO FAHR

8781 8781

0.579 0.682

MCj = λj *(Pj + ϕj *PCO2 ).

MAE

RMSE

AIC

BIC

6.459 6.194

12.129 10.547

4.991 4.712

4.993 4.716

Assuming λgas = 1.98, λcoal = 2.73, ϕgas = 0.202 , ϕcoal = 0.339 from Table 2 and average prices for 2012 of Pgas = 25, Pcoal = 10 , solving for PCO2 yields a switching point around 40€/tCO2. This means that with CO2 prices above 40€/tCO2, typical gas plants exhibit lower marginal costs than typical coal plants. For 2012, when CO2 prices averaged 7.4€/tCO2, (8) implies corresponding marginal costs of coal and gas-fired generation of 34€/ MWh and 52€/MWh respectively. To cross-check, using the slightly different assumptions on heat rates and fuel prices from Agora (2016), (8) yields 39€/MWh and 52€/MWh, respectively, while their own calculations yield 40€/MWh and 53€/MWh respectively. To see how a scenario with significantly higher CO2 prices might alter the entire merit order curve in Germany, we simulate the merit order curve by replacing λj in (8) with the FAHR-model's estimated heat rate curve fhr (lCD ) and consider three scenarios with hypothetical carbon prices at 7, 20 and 40€/tCO2, thus below and around the switching point computed above. Note that in reality, characteristics of power plant types are more heterogeneous and multiple switching points exist, e.g. between oldest and least efficient coal-fired and newest and most efficient gas-fired plants. This has two implications: Firstly, given the FAHR model's static supply curve, the territory beyond a switching point suffers from an upwards bias, i.e. simulated prices are too high because the model does not mimic the switching in the merit order. Secondly, since there is no all-important switching point, the model's bias is more evenly spread and not confined to scenarios with carbon prices above 40€/tCO2, although the more stringent scenarios suffer from a larger bias, undoubtedly stretching the model's application to its limits. In each scenario, we assume the carbon price is equal to the floor, while holding all other variables and parameters equal. In Scenario7 we assume a fixed carbon price at 7€/tCO2, just below the actual yearly average of 7.4€/tCO2, in order to mimic the market outcome. Scenario20 and Scenario40 represent two increasingly stringent carbon price floors at 20 and 40€/tCO2. For comparison, the EU has estimated that reducing emissions to levels consistent with reaching the 2-degree Celsius target would require a carbon price of at least €32–63 by 2030 (EU, 2012).

assumed to be the most representative single fuel and the model abstracts from gas and lignite prices.8 On the other hand, the heat rate curve is defined as a function of load lCD(t ):

fhr (lCD(t )) =

e

lCD(t )− a b

− e− 2

lCD(t )− c d

+ h,

(6)

where a and c can be interpreted as horizontal shifts, b and d as scaling factors and h as a vertical shift. Given the functional form of (4), the problem can be formulated as nonlinear least-squares in matrix notation (Hayashi, 2000, 453):

y = f (X , β ) + ε ,

(8)

(7)

where y is the vector of the 8760 hourly observations of S(t) and X is the vector of fuel prices (x1 ≡ Pcoal (t )), CO2 prices (x2 ≡ PCO2(t )) and load values (x3 ≡ lCD(t )). Minimizing the error term ε yields the coefficients or parameters β ≡ [a, b, c, d , h].9 Fig. 2 illustrates the two competing model specifications and Table 3 reports their In-sample performance in terms of fitting measures; coefficient of determination R2, Mean Absolute Error (MEA), Root Mean Squared Error (RMSE), Akaike and Bayes Information Criterion (AIC and BIC). A higher R2 indicates better model performance, the other measures are to be interpreted inversely, i.e. the lower the better. Both information criteria penalize the inclusion of parameters that fail to significantly increase the model's fit, striking a balance between better fit and model parsimony (Hayashi, 2000). While the AIC shows how close a model is from the data generating process, the BIC shows the likelihood of the model generating the observed data. Given the ratio of parameters to observations, model selection is mainly driven by the likelihood term and not the penalization term. In the annex, Table A.1 reports Out-of-sample performance and Table A.2 the estimated parameters. These findings confirm the main result of HHHA, that the FAHR model produces the best fit to the data and is robust to structural changes in the price vector, thereby enabling a simulation with a fabricated fuel price vector.

3. Results Fig. 3 shows the supply curves in Scenario20 and Scenario40 against actual market data in 2012. Horizontal lines indicate median and peak prices, the latter is assumed at the 90th percentile. Table 4 summarizes

Consistent with the findings by HHHA, we find lower model performance when using gas as the single representative fuel. 9 Given its nonlinear specification, the sum of squared residuals function is not quadratic and the solution needs to be found by numerical methods (Hansen, 2000, 188). We use Stata's “nl function evaluator program”, itself based on the Levenberg-Marquardt method to find the step increases. 8

10 Most studies estimate a cost pass-through close to 100%, such as Hintermann (2014) estimating an almost complete pass-through for the German market of 98–104%.

688

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

Fig. 3. Observed data and simulation with carbon price floor at 20 and 40€/tCO2. Based on own calculations and data for 2012. Table 5 Data and simulation results at the aggregate national level.

Table 4 Data and simulation results for median and peak power prices.

Market data Scenario7 Scenario20 Scenario40

CO2 price [€/tCO2]

Median price [€/MWh]

Peak pricea [€/MWh]

[€ bn p.a.]

Power supplya

Solarb

Windb

Biomassc

EEGSavingsd

Net costse

7.4 7 20 40

42.1 42.2 56.8 79.4

60.7 58.7 79.3 110.9

Market data Scenario7 Scenario20 Scenario40

21.2 20.9 28.1 39.2

8.6 8.6 8.2 7.5

3.0 2.9 2.3 1.5

4.8 4.8 4.3 3.6

n/a n/a 1.5 3.8

n/a n/a 5.4 14.2

Based on own calculations and data for 2012. a Total product of hourly wholesale power price and national hourly load. b Total difference between hourly market price and feed-in-tariff multiplied with hourly feed-in for onshore wind and solar generation. c Total difference between hourly market price and feed-in-tariff multiplied with average feed-in for biomass. d Total difference for all three preceding columns between the market data and the respective scenarios. e Difference between increased power supply-costs and lowered EEG-costs, with respect to current market data.

Based on own calculations and data for 2012. a Peak prices correspond to the 90th percentile.

the effects on median and peak prices. Broadly consistent with the market outcome, Scenario7 yields a median price at 42€/MWh and an average price peak at 61€/MWh. In Scenario20 , median prices increase by 15€/MWh and reach 57€/MWh, while peak prices increase by roughly 19€/MWh and reach 79€/MWh. In Scenario40 , the increase in power prices is even more pronounced, with median and peak prices roughly 38€/MWh and 52€/MWh higher than the observed market outcome. Table 5 summarizes the effects of the carbon price floor on the wholesale market's annual volume, the cost of the support scheme for renewables and the net cost effect borne by consumers.11 Without a carbon price floor, the wholesale market's volume—defined as the total product of hourly load values for common supply LCS(t) and observed hourly wholesale prices—stands at €21.2 bn for the year 2012. Data from Agora (2016) can be used as a cross-check. Using the reported average marginal costs for lignite, coal and gas and the respective yearly generation for 2012, aggregate fossil-fuel generation accounted for roughly 55% of domestic generation and valued roughly €11 bn. Assuming that wholesale prices of the remaining 45% of domestic

generation were also set by fossil fuel-fired units, the entire domestic generation volume adds up to €20 bn.12 With a carbon price floor at 20€/tCO2, the wholesale market's annual volume increases from €21.2 bn to roughly €28.1 bn, as seen in the first column in Table 5, a model-implied increase of €6.9 bn. This figure can be compared with the mechanically expected increase, when assuming the generation mix (and hence the total amount of emissions) stays constant and the pass-through of emissions costs is complete. In 2012, the German power sector emitted 325 million tCO2 (Agora, 2016). With a carbon price at 20€/tCO2 instead of 7.3€/ tCO2, the mechanically expected increase is only €4.1 bn. For both Scenario20 and Scenario40 , the computed effect with the

12 Abstracting from imprecision and rounding, the difference of roughly €1 bn may be attributed to a certain degree of market power of producers, i.e. bidding above marginal costs.

11

Retail prices typically consist of three main cost components, (i) power supply, (ii) grid infrastructure and (iii) taxes and levies.

689

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

4. Discussion

FAHR-model is some 40% higher than the mechanically expected effect. This divergence can be explained with the model's methodology, which relies on a single representative fuel price input, i.e. coal prices, and the rather uncharacteristic fuel price development in the year 2012, when gas prices increased while coal prices decreased (Fig. 1). This diverging price development means that the aggregate heat rate curve, i.e. fhr in (6), is inflated as higher spot prices cannot be attributed to coal prices. With an average of roughly 3.3 MWhth/ MWhel, our estimate for the aggregate heat rate in 2012 is relatively high compared with HHHA's figure of 2.7 MWhth/MWhel for the year 2010 and the literature in Table 2. Thus, the derivation of the aggregate heat rate curve from a single fuel price constitutes an inherent drawback of the methodology and the FAHR-model's estimate of the effect of the carbon price floor on power prices should therefore be interpreted as an upper ceiling of the price effect. The methodology is critically discussed in Section 4. Generally, as wholesale prices increase, the costs of the EEG subsidy-scheme are lowered. Fig. 4 illustrates this point and depicts the market outcome with and without the carbon price floor. Put simply, the EEG's aim is to bridge the gap between wholesale prices and production costs of the RES technology. The figures indicated as production costs are average received feed-in-tariffs of wind onshore, solar and biomass plants in 2012 (BMWi, 2015a). Thus, calculating the EEG-costs from observed hourly prices and solar and wind infeed yields €8.6 bn and €3 bn, respectively, as seen in Table 5. For biomass, a constant yearly average generation is assumed, yielding EEG-costs of €4.8 bn. Total EEG-costs for all three technologies sum to €16.4 bn. For comparison, BMWi (2015a) report so called EEG “difference-costs” of €15.5 bn for the three technologies in 2012. The second column from the right sums the EEG-cost savings for the different carbon price scenarios. The last column reports the net cost effect associated with the carbon price floor. In Scenario40, the wholesale market's annual volume increases by €18 bn (from €21.2 to €39.2 bn), but EEG-costs are lowered by €3.8 bn, resulting in a net cost effect of €14.2 bn. Assuming non-privileged consumption of 385 TW h (BMWi, 2015a) in 2012, the EEG-levy on end-users could have been lowered from 4 to 3ct/kWh.

Let us now discuss the broader implication of a carbon price floor, through the three stated objectives of the German and European climate and energy policy: reducing emissions; maintaining system adequacy; providing affordable energy. The section closes with a reference to further political considerations. 4.1. Emissions from electricity generation The main objective of the whole climate & energy policy framework is to move towards a carbon-free electricity sector. The experience of the last few years shows that the decarbonization of the German energy system is stagnating, in particular due to lignite- and coal-heavy power generation, with these two fuel types accounting for more than threequarters of the entire power sector's CO2 emissions (Agora, 2016). Emissions of the power sector are determined by the merit order and thus relative prices of fuels and CO2. Putting a price on emissions is, theoretically, a powerful lever to influence the merit order. However, with the massive surplus of emissions allowances in the EU-ETS, global fuel price variations are the predominant driver of marginal cost, while the price signal from the EU-ETS is almost negligible. The allowance surplus is testimony to the environmental effectiveness (Knopf et al., 2014): the reduction target, set out for 2020, was already reached in 2014, but the ETS fails to give adequate forward guidance for the more ambitious reductions that need to be achieved in the time after 2020. To avoid a further lock-in into a trajectory of intensive coal production, and a costly transition with massive stranded assets, a clear and reliable price signal is indispensable. Note that Germany, home of the Energiewende, has seen ten lignite and coal-fired power plants, with an aggregate capacity of 8.8 GW, being brought online between 2012 and 2015 (UBA, 2015). 4.2. System adequacy In terms of system adequacy, the simulation results need to be seen in a wider context. The historically established division of labor between base-, mid- and peak-load plants is upended by the large-scale deployment of RES. Given that marginal costs of RES are near zero and production is

Fig. 4. Actual market outcome and simulation with carbon price floor, on arbitrarily chosen time span in 2012. Based on own calculations and average production costs from (BMWi, 2015a).

690

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

outlook on futures markets, virtually all power plant types are facing years of financial distress. A carbon price floor would ensure that climateconsistency is at least factored into the process of slimming-down portfolios. On the other hand, low-emissions technologies, e.g. hydro, nuclear and to some extent also gas-fired plants, would benefit from higher wholesale prices and increased operating hours. As a concession to negatively affected groups, i.e. privileged end-users and generators with emission-intensive portfolios, additional revenues from the auctioning of CO2-permits could be partially redistributed.13

usually subsidized, wind and solar infeeds exert downward pressure on wholesale prices, often termed merit order effect. Würzburg et al. (2013, 159) conduct a meta-study and report an average merit order effect for Germany of roughly −1€/MWh for each additional GW of wind infeed. The merit order effect of solar power is particularly disruptive to system adequacy, because it systematically depresses peak prices—at midday— and thus disproportionately affects peak-load plants (Hildmann et al., 2011). Historically, peak-load plants, mostly gas-fired, used price peaks to offset their short operation spans, or in the case of pumped-storage hydro, the inherent energetic loss in the water's round trip. Thus, given the intermittent and increasing infeed from RES, the need for constantly operating baseload plants decreases, but flexible peak-load plants remain important to balance supply and demand at all times (Nicolosi, 2014). By lowering the economic viability of such plant types, wind and particularly solar are posing a challenge to system adequacy in the long term (Traber and Kemfert, 2011). This argument is at the heart of the ongoing debate about missing money and capacity remuneration mechanisms. As the simulation in Fig. 3 illustrates, a carbon price floor would induce a steeper slope of the merit order curve and thus, arguably larger intraday price spreads. As mentioned, the simulated merit order curve is imperfect because the FAHR model assumes a static aggregate heat rate in the market and does not account for the dynamics that a changed price vector would trigger. Abstracting from this, as shown in Table 4, peak prices (at P90) increase from 61€/MWh. to 116€/MWh. Such amplified price spikes and spreads would improve the economic viability of flexible peak-load plants significantly. The power sector community is divided on the question of capacity remuneration markets, either doubting their efficiency, fearing distorting incentive structures further or questioning the need altogether. Based on a series of studies, the German Ministry of Economics and Energy concludes that the existing energy-only market is preferable to a capacity remuneration mechanism, citing lower system-wide costs and a smaller risk of market design failures (BMWi, 2015b). A carbon price floor could avoid or delay the spread of capacity remuneration mechanisms and thus presents another source of significant savings, albeit one that is not quantifiable within the scope of this work.

4.4. Political considerations Theoretically, a carbon price floor could be implemented EU-wide or unilaterally, as the UK has proven. But Germany is not an island and more deeply integrated with neighboring power markets. For the moment, Germany is particularly exposed to the inefficiency of the European climate policy setup: the EEG subsidy-scheme has become a significant cost factor and an ongoing public topic, while power sector emissions are barely falling. A more ambitious policy package, coordinated with a coalition of the willing including France the UK and possibly others, could greatly increase efficiency and effectiveness compared to an unilateral measure. Meanwhile, a unilateral carbon price floor for the power sector faces a series of drawbacks, discussed in next paragraphs. The first drawback is that, given the nature of the EU-ETS, additional emissions reductions in Germany would lower the EU-ETS carbon price and thereby reduce the incentive for other countries (Sinn, 2008). In fact, as long as no final decision has been made on the allowances set aside, any additional emission reduction in a sector covered by the EU-ETS might be compensated elsewhere. However, with a more long-term view, if a German carbon price floor generates some sort of forward guidance and reduces the extent of stranded assets, the savings are not shared internationally but only accrue to Germany. The second drawback is that governance of the EU-ETS would be complicated if member states implement additional domestic climate policies, e.g. carbon price floors. But in the absence of a stringent price signal from the EU-ETS, member states have already begun to set phase-out plans for coal (Austria, UK and Netherlands), undermining the mechanism that would in time lead to higher prices again. Thus, relying only on quantitative measures but avoiding price mechanisms, e.g. a carbon price floor, it seems increasingly unrealistic that the EUETS will provide non-negligible price signals like it did at the beginning of the second trading phase in 2008 (see Fig. 1). A third drawback stems from the inefficiency of varying (carbon) prices in different sectors or countries. A carbon price signal confined to power generation hinders integration with other energy services, such as transportation or heating/cooling. In the long-term, however, ambitious emissions reductions targets can only be achieved if sectors with bigger potentials and more extensive inertia start their transition first, in order to avoid costly lock-ins (Vogt-Schilb and Hallegatte, 2014). Also, different carbon prices for ETS and non-ETS sectors have long been a reality. It might even be sensible to target the power sector unilaterally, because grid topology means generators face only limited international competition and thus regulators need not worry about carbon leakage. Finally, with an unilateral approach the price effect of a German carbon price floor would be counteracted, at least partially, by more imports from neighboring markets. From a foreign generator's perspective, more exports to Germany and marginally higher wholesale prices are welcome, from the foreign consumer's perspective the opposite is generally true. However, as consumers are set to cover the costs of (upcoming) capacity remunerations mechanisms across Europe, an unilateral carbon price floor in Germany would actually reduce the costs of a capacity remuneration mechanism in a neighboring country.

4.3. Affordability Recalling Scenario40 in Table 4, a doubling of median wholesale prices from 42€/MWh to 79€/MWh is certainly significant, however, it has to be put into context with other components of retail power prices, particularly the EEG-levy borne by end-users. Between 2009 and 2015, German retail prices increased from 228€/MWh to 291€/MWh with the EEG-levy alone responsible for a 50€/MWh increase, as it grew from 12€/MWh to 62€/MWh (Agora, 2015, 23). Given its impact on power prices, a carbon price floor would trigger some predictable objections. Currently, energy-intensive businesses may be exempt from paying the EEG-levy (Cludius et al., 2014). These so-called privileged end-users would be significantly worse-off under a regime that finances RES directly through higher wholesale prices. Such measures have been carefully avoided by policymakers on the grounds of protecting the competitiveness of energy-intensive industries and for fear of provoking carbon leakage. Note, however, that the effect of climate regulations on the competitiveness of polluting industries is still debated (Greenstone et al., 2012; Ambec et al., 2013). Another group of potential opponents are power generators with coalheavy portfolios. Since the pass-through of CO2 prices to wholesale power prices is almost complete (Hintermann, 2014), these producers would simply pass-on the increased costs to consumers. However, the idea of a stringent price floor is to provoke shifts in the merit order, i.e. fuel switching and thus emissions-intensive plants would suffer a reduction of operating hours, which may eventually trigger closures of certain plants. In 2015, an additional CO2 levy targeting only old lignite plants was rejected after public consultation. Since then, the slump in global fuel, CO2 and power prices has only become more accentuated and given the

13 According to EU regulation, “at least half of auctioning revenues [are] to be used for climate and energy related purposes” (European Commission, 2016).

691

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

The extent to which a German price floor would affect prices and trade patterns in neighboring countries is mainly a function of the interconnector capacity and the foreign market's generation fleet and size. Therefore a quantification lies beyond the scope of this work, however, an early assessment of the French proposal to implement an unilateral carbon price floor at 30€/tCO2 in France can provide a starting point. Dimantchev et al. (2016) predict slightly higher French prices, lower exports, more government revenue and only a marginal effect on prices in neighboring countries and the EU-ETS. Both countries are similarly well connected with their neighbors, with some 15–17 GW of interconnectors (ENTSO-E, 2011), but given that Germany's power mix is significantly more carbon-intensive, the price effects on its neighbors would be more pronounced than in the French example.

to be considered at the level of the electricity system, and of the climate and energy policy mix as a whole. Firstly, part of this cost increase to consumers is counteracted at the level of retail prices, because of synergies with the EEG-subsidy scheme. Secondly, it counteracts the erosion in price peaks that plagues the profitability of systematically important flexible back-up generation, thereby reducing the necessity and costs of capacity remuneration schemes. Thirdly, it generates a significant amount of public resources, that can be used to alleviate adverse effects on consumers and businesses. An important limitation of our approach is the assumption of a static heat-rate curve and thus of a fixed supply side. By abstracting from fuel-switching in the short term and from plant replacements in the longer term, our results should be interpreted as an upper ceiling of the probable effects. Despite these limitations, our approach offers a compact and comprehensive way to capture real world market imperfections and complex technical constraints. As a closing remark and topic for further investigation, we argue that the ongoing erosion of profitable operations of back-up generation could be reversed if prices in the tightest hours peak in a meaningful way. As capacity remuneration schemes are spreading all over Europe, further work should aim to quantify their system-wide costs and benefits. Such analysis would allow to better judge the potential merits of a carbon price floor.

5. Conclusion and policy implication There are two main policy levers to lower emissions in the power sector: providing additional incentives for renewable generation, e.g. the EEG-subsidy scheme in Germany, or reducing the incentive to dispatch carbon-intensive generation, e.g. the EU-ETS. As this paper argues, climate policy can be significantly more effective in reducing carbon-emissions when both levers are considered jointly. But the key contribution of this paper is to provide a first estimate of the net costs of such a policy shift to consumers. This is important, because in retail prices the EEG-subsidy scheme represents a larger cost component than the actual wholesale power supply. Using an empirical model of the aggregate hourly merit order curve, we simulate the hypothetical effects of an unilateral carbon price floor of 20€/tCO2 and 40€/tCO2 on the German power market. We find that a 40€/tCO2 carbon price floor would increase median prices by roughly 37€/MWh and peak prices (i.e. at the 90th percentile) over-proportionately by roughly 50€/MWh. This changed price structure increases the wholesale market's annual volume (i.e. total sum of hourly load multiplied with hourly prices) by some €18–€39 bn. At the retail level, however, the net cost to consumers of the carbon price floor at 40€/ tCO2 is moderated due to cost savings from the EEG-subsidy scheme worth some €4 bn, or roughly one-fifth. The same ratio applies to the lower price floor at 20€/tCO2. Such a policy shift is indeed costly to consumers, but its effects have

Acknowledgments This work has been done while affiliated to the University of Bern, and benefited from financial support from the Oeschger Center for Climate Change Research at the University of Bern. It is based on Philipp Egli's master thesis "Applied Climate Economics: How Climate Policy Reshapes the Supply Curve on Wholesale Power Markets in Germany and Italy", under supervison of Ralph Winkler and Oskar Lecuyer. Although EWZ would arguably benefit from a carbon price floor in Germany, this paper does not necessarily reflect the position of EWZ or the AFD. The views expressed here are under the sole responsibility of the authors. We thank Ralph Winkler, Philippe Quirion, Johanna Cludius, Christoph Graf, Silvia Banfi-Frost and two anonymous referees for useful comments and suggestions.

Appendix See Table A.1 and A.2.

Table A.1 Out-of-sample model performance. Specification

N

R2

MAE

RMSE

EXPO FAHR

8757 8757

0.728 0.644

7.972 7.060

10.123 10.580

Note: Own elaboration, data for 2013.

Table A.2 Estimated model parameters. Model

a

b

c

d

h

EXPO t-stat FAHR t-stat

−0.399 −34 0.781 134

0.307 103 0.080 30

0.665 146

0.068 37

3.150 66

Note: Own elaboration, data for 2012.

692

Energy Policy 109 (2017) 685–693

P. Egli, O. Lecuyer

References

the European Energy Exchange power market in Germany. IEEE Trans. Power Syst. 28 (3), 3155–3164. Hildmann, M., Ulbig, A., Andersson, G., 2011. Electricity grid in-feed from renewable sources: a risk for pumped-storage hydro plants? In: Proceedings of the 8th International Conference on the European Energy Market (EEM). IEEE, pp. 185– 190. Hintermann, B., 2014. Pass-through of CO2 Emission Costs to Hourly Electricity Prices in Germany. CESifo Working Paper 4964. CESifo Group Munich. IEA, 2013. CO2 Emissions from Fuel Combustion. Highlights. 2013 ed. Tech. Rep., International Energy Agency (IEA). Paris. IPCC, 2005. IPCC/TEAP Special report: safeguarding the ozone layer and the global climate system. Chapter 3: Methodologies. Table 3.5. Carbon Dioxide Intensities of Fuels and Electricity for Regions and Countries. Tech. Rep., Intergovernmental Panel on Climate Change IPCC. 〈https://www.ipcc.ch/pdf/special-reports/sroc/Tables/ t0305.pdf〉. Knopf, B., Koch, N., Grosjean, G., Fuss, S., Flachsland, C., Pahle, M., Jakob, M., Edenhofer, O., 2014. The European Emissions Trading System (EU ETS): Ex-post Analysis, the Market Stability Reserve and Options for a Comprehensive Reform. Koch, N., Fuss, S., Grosjean, G., Edenhofer, O., 2014. Causes of the EU ETS price drop: recession, CDM, renewable policies or a bit of everything? – New evidence. Energy Policy 73 (C), 676–685. Koch, N., Grosjean, G., Fuss, S., Edenhofer, O., 2016. Politics matters: regulatory events as catalysts for price formation under cap-and-trade. J. Environ. Econ. Manag. 78, 121–139. Lecuyer, O., Quirion, P., 2013. Can uncertainty justify overlapping policy instruments to mitigate emissions? Ecol. Econ. 93, 177-191. 〈http://www.sciencedirect.com/ science/article/pii/S092180091300178X〉. Lecuyer, O., Quirion, P., 2016. Interaction Between CO2 Emissions Trading and Renewable Energy Subsidies under Uncertainty: Feed-in Tariffs as a Safety Net Against Over-allocation. Working Paper, FAERE. Lehmann, P., Gawel, E., 2013. Why should support schemes for renewable electricity complement the EU emissions trading scheme? Energy Policy 52, 597–607, Special Section: Transition Pathways to a Low Carbon Economy. 〈http:// www.sciencedirect.com/science/article/pii/S0301421512008762〉. Newell, R.G., Pizer, W.A., 2003. Regulating stock externalities under uncertainty. J. Environ. Econ. Manag. 45 (2), 416–432 http://www.sciencedirect.com/science/ article/pii/S0095069602000165. Newell, R.G., Pizer, W.A., Raimi, D., 2013. Carbon markets 15 years after Kyoto: lessons learned, new challenges. J. Econ. Perspect. 27 (1), 123–146. Nicolosi, M., 2014. Leitstudie Strommarkt – Arbeitspaket Optimierung des Strommarktdesigns. Tech. Rep., Connect Energy Economics GmbH, Studie für Bundesministerium für Wirtschaft und Energie. Roberts, M.J., Spence, M., 1976. Effluent charges and licenses under uncertainty. J. Public Econ. 5 (3–4), 193–208 http://ideas.repec.org/a/eee/pubeco/v5y1976i34p193-208.html. Sandbag, November 27th, 2015. Europe is on Track for 30% Emissions Cuts by 2020. Tech. Rep., Sandbag. 〈https://sandbag.org.uk/site_media/pdfs/reports/EU_on_ track_for_30_cuts_by_2020_9Dec15.pdf〉. (last accessed April 10, 2016) Sinn, H.-W., 2008. Das Grüne Paradoxon. Plädoyer für eine illusionsfreie Klimapolitik. Econ-Verl. Berlin. Taschini, L., 2013. Options for Structural Measures to Improve the European Union Emissions Trading System: Response to a European Commission Consultation. Tech. Rep., Centre for Climate Change Economics and Policy and Grantham Research Institute on Climate Change and the Environment. Traber, T., Kemfert, C., 2011. Gone with the wind? – Electricity market prices and incentives to invest in thermal power plants under increasing wind energy supply. Energy Econ. 33 (2), 249–256. UBA, 2013. Entwicklung der spezifischen Kohlendioxid-Emissionen des deutschen Strommix in den Jahren 1990 bis 2012. Tech. Rep., Umweltbundesamt (UBA). UBA, 2015. Kraftwerke in Deutschland. Tech. Rep., Umweltbundesamt UBA. 〈http:// www.umweltbundesamt.de/dokument/datenbank-kraftwerke-in-deutschland〉. (last accessed April 10, 2016). UK Parliament, Nov 2013. Carbon Price Floor. Standard Note SN/SC/5927. 〈http:// www.parliament.uk/business/publications/research/briefingpapers/SN05927/ carbon-price-floor〉. (Accessed 18 April 2016). Vogt-Schilb, A., Hallegatte, S., 2014. Marginal abatement cost curves and the optimal timing of mitigation measures. Energy Policy 66 (March), 645–653. World Bank, 2015. State and Trends of the Carbon Pricing. Tech. Rep., World Bank. Würzburg, K., Labandeira, X., Linares, P., 2013. Renewable generation and electricity prices: taking stock and new evidence for Germany and Austria. Energy Econ. 40, 159–S171.

Agora, 2015. The Energiewende in the Power Sector: State of Affairs 2014. Tech. Rep., Agora Energiewende, Agora Energiewende is a joint initiative of the Mercator Foundation and the European Climate Foundation. 〈http://www.agoraenergiewende.org/topics/the-energiewende/detail-view/article/trendwende-in-derenergiewende/〉. Agora, 2016. Die Energiewende im Stromsektor: Stand der Dinge 2015. Tech. Rep., Agora Energiewende. 〈http://www.agora-energiewende.de/fileadmin/Projekte/ 2016/Jahresauswertung_2016/Agora_Jahresauswertung_2015_web.pdf〉. Ambec, S., Cohen, M.A., Elgie, S., Lanoie, P., 2013. The Porter hypothesis at 20: Can environmental regulation enhance innovation and competitiveness? Rev. Environ. Econ. Policy. http://reep.oxfordjournals.org/content/early/2013/01/04/reep. res016.abstract. Ausubel, L.M., Cramton, P., 2004. Vickrey auctions with reserve pricing. In: Assets, Beliefs, and Equilibria in Economic Dynamics. Springer, pp. 355–367. BMWi, 2015a. EEG in Zahlen: Vergütungen, Differenzkosten und EEG-Umlage 2000 bis 2016. Tech. Rep., Bundesministerium für Wirtschaft und Energie (BMWi). BMWi, 2015b. Ein Strommarkt für die Energiewende. Ergebnispapier des Bundesministeriums für Wirtschaft (Weissbuch). Tech. Rep., Bundesministerium für Wirtschaft und Energie (BMWi). 〈http://www.bmwi.de/BMWi/Redaktion/PDF/ Publikationen/weissbuch,property=pdf,bereich=bmwi2012,sprache=de,rwb=true. pdf〉. Bundesnetzagentur, 2015. EEG in Zahlen 2014. Tech. Rep. Burtraw, D., Palmer, K., Kahn, D., 2010. A symmetric safety valve. Energy Policy 38 (9), 4921–4932, Special Section on Carbon Emissions and Carbon Management in Cities with Regular Papers. 〈http://www.sciencedirect.com/science/article/pii/ S0301421510002582〉. Burtraw, D., Löfgren, Å., Zetterberg, L., 2014. A price floor solution to the allowance surplus in the EU Emissions Trading System. Issue Brief 14-02, Resources for the Future. 〈http://www.ourenergypolicy.org/wp-content/uploads/2014/01/rff-pricefloor.pdf〉. Cludius, J., Hermann, H., Matthes, F.C., Graichen, V., 2014. The merit order effect of wind and photovoltaic electricity generation in germany 2008–2016: estimation and distributional implications. Energy Econ. 44, 302-313. 〈http://www.sciencedirect. com/science/article/pii/S0140988314001042〉. Delarue, E., Voorspools, K., D'haeseleer, W., 2008. Fuel switching in the electricity sector under the EU-ETS: review and prospective. J. Energy Eng. 134 (2), 40–46. Dimantchev, E., Fjellheim, H., Qin, Y., 2016. Leading by example? Impacts of a Domestic French Carbon Price Floor. Energy Post. 〈http://energypost.eu/leading-exampleimpacts-domestic-french-carbon-price-floor/〉. EEA, 2015. Trends and Projections in the EU ETS in 2015. Tech. Rep., European Environment Agency, 〈http://www.eea.europa.eu/publications/trends-andprojections-eu-ets-2015〉. EEX, 2014. Market Data. European Energy Exchange (EEX). 〈http://www.eex.com/en/ market-data#/market-data〉; 〈http://www.transparency.eex.com/en/〉. ENTSO-E, 2011. Net Transfer Capacities in Europe. 〈https://www.entsoe.eu/fileadmin/ user_upload/_library/ntc/archive/NTC-Values-Winter-2010-2011.pdf〉. EU, 2012. Energy Roadmap 2050. Tech. Rep. COM(2011) 885 final, European Commission. Luxembourg: Publications Office of the European Union. EU, 2015. Climate Action Progress Report, Including the Report on the Functioning of the European Carbon Market and the Report on the Review of Directive 2009/31/ec on the Geological Storage of Carbon Dioxide. Report on the functioning of the European carbon market COM(2015) 576 final, European Commission. EU, December 2012. Impact Assessment and Scenario Analysis Accompanying the Energy Roadmap 2050. Commission Staff Working Document Sec(2011) 1565 final, European Commission. European Commission, 2016. Climate Action; Auctioning. At Least Half of Auctioning Revenues to Be Used for Climate and Energy Related Purposes. 〈http://ec.europa. eu/clima/policies/ets/cap/auctioning/index_en.htm〉. Geman, H., 2005. Commodities and Commodity Derivatives: Modeling and Pricing for Agriculturals, Metals and Energy. Wiley Finance, Chichester (Grande-Bretagne). Goulder, L.H., 2013. Markets for pollution allowances: what are the (new) lessons? J. Econ. Perspect. 27 (1), 87–102 http://ideas.repec.org/a/aea/jecper/ v27y2013i1p87-102.html. Greenstone, M., List, J.A., Syverson, C., September 2012. The Effects of Environmental Regulation on the Competitiveness of U.S. Manufacturing. Working Paper 18392, National Bureau of Economic Research. 〈http://www.nber.org/papers/w18392〉. Hansen, B.E., 2000. Econometrics 2013th ed.. Hayashi, Fumio, 2000. Econometrics. Princeton University, New Jersey, USA. He, Y., Hildmann, M., Herzog, F., Andersson, G., 2013. Modeling the merit order curve of

693