The dynamic impact of carbon reduction and renewable support policies on the electricity sector

Utilities Policy 28 (2014) 28e41

Contents lists available at ScienceDirect

Utilities Policy journal homepage: www.elsevier.com/locate/jup

The dynamic impact of carbon reduction and renewable support policies on the electricity sector Riccardo Fagiani*, Jörn C. Richstein, Rudi Hakvoort, Laurens De Vries Faculty of Technology, Policy and Management, Delft University of Technology, Jaffalaan 5, 2600 GA Delft, The Netherlands

a r t i c l e i n f o

a b s t r a c t

Article history: Received 1 August 2013 Received in revised form 19 November 2013 Accepted 29 November 2013

Carbon reduction and renewable energy policies are implemented in Europe to improve the sustainability of the electricity sector while achieving security of supply. We investigate the interactions between these policies using a dynamic investment model. Our analysis indicates that both policies are necessary to achieve a sustainable power sector. However, renewable energy generation significantly affects carbon markets and could lead to very low prices. These would attract investments in carbon intensive technologies, locking the sector into future higher emissions. To contrast this effect, policy makers may introduce a floor price in the carbon market or adjust the emissions quota periodically. Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: Electricity sector EU ETS Renewable energy policy Dynamic policy interaction

1. Introduction The European Union set an ambitious target to lower the emissions of greenhouse gases (GHG) and established challenging goals for the production of energy from renewable energy sources (RES). An Emissions Trading Scheme (ETS) was established at the European level while mechanisms supporting investments in RES are implemented at a national level. A significant sector affected by these policies is the power industry, since it is one of the primary sectors emitting GHG and many RES technologies are electricity generators. In this paper we investigate the dynamic interactions between carbon reduction and renewable energy policies. Both policies affect operational and investment decisions concerning renewable and conventional generation in the electricity market. Carbon policy adds to the variable cost of conventional generators. This affects their revenue streams and may change the merit order. Renewable energy producers also affect generation dispatch and thus the (future) revenues of the other generators in the market. While interactions have been studied for equilibrium conditions, the dynamic feedback loops between the two policies over time have been less investigated. Jensen and Skytte (2003) discuss the impact of the correlation between the consumer price and the

* Corresponding author. Tel.: þ31 655795492. E-mail addresses: [email protected], (R. Fagiani).

[email protected]

0957-1787/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jup.2013.11.004

renewable energy quota on the interactions between carbon reduction and green certificate markets. Linares et al. (2008) present an oligopolistic partial-equilibrium model simulating the Spanish electricity sector under different energy policy scenarios. Amundsen and Nese (2009) investigate the interaction between carbon and renewable energy policies in the Scandinavian region implementing an analytical equilibrium model. De Jonghe et al. (2009) use a welfare maximization simulation in order to find equilibrium states for combinations of renewable and carbon policy in a three zone system. This paper presents a bottom-up investment model which simulates the evolution of a hypothetical electricity sector, with characteristics close to the Spanish system, under different policy scenarios. This work addresses the following research question: In which manner do CO2 reduction policies and renewable energy support mechanisms dynamically affect each other? This research adds insights to the analysis of the interactions existing between carbon reduction and renewable energy policy. We apply a simulation approach that combines elements of agent-based and system-dynamic modeling. The purpose of our model is to analyze how energy policy instruments affect the investment decisions of generating companies by changing the profit and risk profiles of investment projects (Gross et al., 2010). We apply the notion of bounded rationality (Simon, 1957), recognizing that investors are not fully rational when making decisions and do not necessarily optimize but rather satisfice. This means that investors’ decisions may not be optimal, but adequate to comply with their expectations. This reflects the fact that investors have

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

informational, intellectual, and computational limitations. Hence, in our model the agents base their investment decisions on available information and on expectations, trying to maximize the trade-off between risks and profits. Agent behavior is also limited by their past investment choices, which affect their current generation portfolios, balance sheets and cash positions, reflecting path dependency. By simulating the impact of carbon reduction and renewable energy policies on investors’ choices, we model how energy policy shapes the evolution of the electricity sector (Chappin, 2011). Our results indicate that while both carbon reduction and renewable support policies are necessary for improving the sustainability of the electricity sector, an aggressive renewable energy policy may reduce the effectiveness of a carbon market in attracting investments in carbon-intensive technologies. Renewable electricity generation reduces carbon emissions and may therefore lead to lower carbon prices; this is part of the reason by the EU ETS is currently experiencing a period of low prices (in addition to a decline in energy demand and industrial activity) (Rathmann, 2007; Lecuyer and Quirion, 2013). If this causes generation companies to invest in coal-fired generators, this may lock the system into higher emissions in the future. In the opposite direction, the EU ETS does not impose negative side-effects on the national renewable support mechanisms; rather, a higher CO2 price reduces the need for RES subsidies. This paper is organized as follows. Section 2 provides an overview of policy instruments for supporting renewable energy and reducing carbon emissions, and describes how these are implemented in Europe. Section 3 presents the details of the model. Section 4 describes and discuss the results of the simulations. Finally, Section 5 concludes with policy recommendations. 2. Renewable and carbon policy in the European power sector Carbon reduction and renewable energy policy mechanisms can be categorized into price-based and quantity-based policies, a distinction made by Weitzman (1974) in a seminal paper on the generic case of regulating a particular economic variable. In quantity-based instruments, the desired level of outcome is set and an artificial market is created in which participants trade certificates to fulfill the policy target. This yields a price for the regulated variable. Examples are the emission trading schemes (ETS) for GHG, in which emitters buy emission allowances (and therefore need to pay for emissions) and tradable green certificates (TGC) for the promotion of RES, which are sold by power producers. In pricebased policies, on the other hand, the regulator sets a price for a specific variable, thus, levying a tax on or paying a subsidy to a producer. Ideally, this is a Pigovian tax, which means it is equal to the externality cost of the variable. Examples are technologyspecific feed-in tariffs for RES technologies and a carbon tax on GHG emissions. Combinations of these policy instruments are possible as well; however, they represent an increased level of complexity (Hepburn, 2006). While many publications compare TGC markets with feed-in tariffs (FITs), the question of which policy leads to preferable results for society is still debated. Feed-in tariffs have proven to be effective in reaching policy targets but they suffer from poorer costeffectiveness (Menanteau et al., 2003). They can be used to stimulate technological change, as they can be designed to have a technological specific component. This can be used to promote technologies in their early stages of development, which may possibly lead to higher dynamic efficiency by inducing technological learning (Del Rio, 2012). Feed-in tariffs may also limit the windfall profits that cheaper RES generators experience in a TGC scheme (Haas et al., 2011) (since cheap RES technologies receive the

29

same remuneration as the marginal technology), and do not cause generators to charge a risk-premium as they do in a TGC market due to the volatility of the green certificate price. This was investigated in a precursor to this study which incorporated generator’s risk attitudes and that found that the efficiency of TGC schemes’ depended on the risk attitudes towards RES technologies (Fagiani et al., 2013). The general arguments made by Weitzman (1974) were applied to carbon policy by Grubb and Newberry (2007). They concluded that while a well-set, slowly rising carbon tax would probably be more efficient because it provided more investment certainty (temporal price stability), only a CO2 market in the form of the European ETS was politically viable and credible for such a large area, due to easier negotiations on quantities than prices and because it offered a better promise for global integration. Chappin et al. (2010) presented an agent-based simulation in which they found that a carbon tax was more efficient in reducing emissions at similar costs to an ETS. Both studies also conclude that setting an appropriate tax is very difficult, due to lack of information. Nonetheless, Chappin et al. (2010) suggest a relatively low starting tax, with a commitment that it will only be adjusted upwards periodically, to provide policy flexibility while limiting investor uncertainty. In the tradition of Tinbergen (1952), the European Union implemented different policy instruments for the different policy goals. The establishment of the EU ETS, which was established by Directive 2003/87/EC, has involved three trading phases, from 2005 to 2007 (Phase I), 2008e2012 (Phase II), and Phase III since in 2013. The main difference between the phases was in the increasing number of sectors that were covered and the allocation of allowances, freely or in an auction. While the banking of allowances was not allowed between Phase I and Phase II, from Phase II they may be taken over to Phase III. Renewable energy policy, on the other hand, developed at the initiative of member states. Directive 2009/28/EC imposed legally binding national renewable targets for 2020, which differ between the states and were implemented via National Renewable Action Plans (NREAP) (European Commission, 2009; European Commission, 2010). However, no instruction was provided as to the choice of policy instrument for reaching these targets (Haas et al., 2011). In addition, targets for both carbon reduction and renewable energy were established for 2050 (European Commission, 2011). 3. Model description Our model simulates the evolution of a hypothetical power sector with characteristics similar to the Spanish system from 2012 to 2050. The model is written and run in Matlab R2011a and comprehends elements of agent-based and system-dynamics methodologies. A previous version of the model was used to evaluate renewable energy support mechanisms, confronting pricebased and quantity-based mechanisms under different riskaversion investors’ behavior (Fagiani et al., 2013). That analysis did not consider the interaction between carbon and renewable policies; instead, an increasing pigovian tax on carbon emission was defined as an exogenous variable. Also, the green certificate market was modeled assuming a steady state equilibrium with its price reflecting the difference between the average total generation cost of the marginal renewable plant in the market and the average electricity price. For the purpose of this analysis, we added to the model a carbon market which covers the power sector exclusively, introducing the carbon price as an endogenous variable. The green certificate and carbon prices are modeled to reflect both short-term and long-term expectations of the generation companies in order to better reflect

30

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

price volatility, and includes the possibility of banking excess green certificates to comply with future quota obligations. Similarly to the previous model, the simulation flow consists of three main blocks which are repeated every simulated year as presented in Fig. 1:  A market block in which the model clears the electricity, the carbon, and the green certificate markets;  A forecasting block in which the model centrally estimates future prices for each of these markets;  An investment block in which the different generating companies make decisions about investments and dismantling existing power plants. Three renewable energy and three carbon policy options are implemented, resulting in a total of nine model runs for each simulated scenario. The simulation includes these renewable support mechanisms:  No RES policy;  A technology specific feed-in tariff system (price-based);  A technology neutral green certificate market (quantity-based). And the following carbon policy options:  No CO2 policy;  An increasing tax imposed on CO2 emissions (price-based);  An emissions allowances scheme limiting the discharge of CO2 (quantity-based). We continue by describing in Sections 3.1, 3.2, and 3.3 respectively how the model clears the electricity, the green certificate, and carbon markets.

3.1. The electricity market Some measures needed to be taken reduce the computational complexity of the model and the time required to run simulations. The annual load duration curve is approximated with 365 steps, each one including 24 h with similar electricity demand. Electricity demand is assumed to grow at a constant 1.5% annual rate until 2020 and increases to 1.9% afterward. The installed generation capacity at the beginning of the simulation period corresponds to the generation mix of Spain in 2012 as presented in Table 1. We assume that the electricity companies have no market power, thus generators’ bids reflect their marginal costs (including their cost of carbon). For each section of the load-duration curve, the market is cleared by intersecting the supply curve with demand, which is assumed to be inelastic. For the sake of simplicity, the model disregards ramping constraints of thermal generators and the possible congestion of transmission lines. Nonetheless, to account for spike prices during peak load hours, we introduced a constraint which prevents electricity companies to cover the two demand sections with the highest loads by starting up power plants which have long start-up times. As a consequence, only gas turbines and generators that are already dispatched are allowed to bid in the two peak load periods. Every year, the model updates fuel prices. To reflect the performance decrease of old power plants, the model increases the fixed O& cost of generators by 1% every year after the end of their expected service life. Technological development is simulated by updating the characteristics of available technologies annually according to an exogenous learning curve, which we assume to be independent from the simulation.

Fig. 1. Flow scheme of the simulation.

3.2. Renewable energy support mechanisms In this section we explain how the model simulates feed-in tariffs and tradeable green certificates. In both cases the model ends the renewable energy support mechanism in 2050 or if a longterm capacity target is reached. This long-term renewable objective is an input variable of the simulation; it corresponds to the final quota of the green certificate market. For the base scenario a value of 65% is used. This constraint can be interpreted as a technical limit to the integration of intermittent generators or as a limit imposed by policy makers to limit the cost of subsidizing RES. This is especially necessary in case of a feed-in mechanism without a quantity limit. Under a feed-in tariff system, generators are guaranteed a fixed electricity price during their expected life time, bidding at a null price in the electricity market (this distinction is important for RES technologies with marginal costs greater than zero, i.e., biomass power plants). After reaching the end of their expected service lives or if no support is given to renewable energy, RES generators behave like conventional generators bidding at marginal cost and receiving the electricity price until they are dismantled. We assume the feed-in tariff mechanism to be an open-budget scheme, so there is no limit for new generators to apply for subsidy until the long-term renewable target is reached. This reflects the current policy implementation in several European countries. Tariffs are technology-specific and reflect the regulator’s estimate of the average generating cost of each technology. The regulator has a biased knowledge of generation costs. Tariffs for new technologies are calculated by multiplying the exact generation cost with a biasing factor as indicated in Equation (1), where FDTjt indicates the j feed in tariff level for technology j at year t,Costt the average generating cost of a plant built in year t with technology j, and BF the biasing factor which is an input parameter of the model and remains constant during the simulated period.

j

j

FDTt ¼ Costt  BF

(1)

The tariff level has a strong impact on the effectiveness of the mechanism. If tariffs are set too low, investors are unlikely to adopt RES technologies and the mechanism is ineffective. High tariffs are effective but may result in unneccessarily high subsidy costs to society. Before running the simulation, we performed a sensitivity analysis to find a moderate biasing factor which would stimulate investments in RES without leading to an excessive subsidy cost. Fig. 2 indicates that there is a notable shift in renewable energy production between a biasing factor of 7.5% and 10%, as many renewable technologies become competitive with conventional generators in that range. A further increase from 10% to 30% only attracts slightly more investment. Therefore, we choose a biasing factor of 15% for our base case scenario, as it represents a good compromise between policy effectiveness and cost-efficiency. Under a green certificate policy, renewable energy generators receive a green certificate for each MWh of electricity produced

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

31

Table 1 Installed capacity in Spain in year 2010 (Red Electrica de Espana, 2012).1

GW %

Hydro

Nuclear

Coal

Fuel/Gas

Combined cycle

Wind on-shore

Small hydro

Biomass

PV

Other

17.564 16.6%

7.777 7.3%

12.210 11.5%

4.376 4.1%

27.123 25.6%

21.239 20.1%

2.041 1.9%

0.859 0.8%

4.249 4.0%

8.450 8.0%

during the expected service period, independently of the technology used. The corresponding electricity that is generated is offered in the market at a zero price, so the volume of green certificate always reflects actual renewable electricity production. Existing hydro generators are treated the same way. In order to be comparable with the feed-in tariff mechanism in which only new generators are subsidized for a period equivalent to their expected life time, renewable energy plants that have reached their expected life times do not receive green certificates and are treated like conventional plants.1 The green certificate and the electricity markets are simulated together. In order to obtain an approximation of the electricity price, the electricity market is cleared first, assuming a null green certificate price. The renewable energy generators use this electricity price to compute their bids for the green certificate market (as explained further below). After the green certificate market is cleared, renewable energy generators bid their marginal generation cost minus the green certificate price and the electricity market is cleared again. This process is not iterated again since the impact of the certificate price on the electricity price is negligible. Regarding the bidding behavior in the green certificate market, if certificates were not bankable, renewable energy generators would bid the difference between the marginal generation cost (CostM t ) and the electricity price (Et), as indicated in (2).

Ct ¼ CostM t  Et

(2)

This would result in many generators offering green certificates at zero price in the market due to zero marginal generation cost of wind and PV generators. As a consequence, the certificate price would fluctuate between zero and the marginal cost of biomassfueled generators, depending on the supply and demand of green certificates. Instead, we assume the validity of green certificates to be unlimited. Market participants are allowed to bank their certificates and use them to comply with future quota obligations. In this case, producers are not forced to sell their certificates in the market during periods of over-supply when the price would be close to zero.2 Certificate banking thus guarantees a certain bargaining power to producers, who could try to offer green certificates at a price higher than their marginal cost in the market in an attempt to raise green certificate prices. In the long run, however, in the absence of barriers to entry, the threat of new entrants would cause the green certificate price to converge towards the difference between the average costs of renewable generators and the electricity price, as indicated in Equation (3).

CLT ¼ CostM LT  ELT

(3) CostM LT

Here CLT indicates the long-term certificate price, the long-term average generation cost of the marginal technology and ELT the long-term average electricity price.

To reflect the ability of renewable energy producers to arbitrage between present and future price expectations, the model calculates the fundamental value of a green certificate by considering shortterm and long-term equilibrium price expectations. The procedure used to calculate the fundamental value of a green certificate is explained in the section on forecasting. Generators bid the higher of the fundamental green certificate price valuation and the difference between the marginal generation cost and the electricity price calculated as in (2). The latter is important for biomass-fired generators who are willing to run only if the sum of electricity and green certificate prices is equal to or higher than their marginal cost, and may require a price above the fundamental value to run. This introduces a certain elasticity to the short-term green certificate supply. The green certificate price is obtained by intersecting supply and demand. The demand for green certificates is set as a percentage of the electricity demand, corresponding in 2012 to a quota of 30% (including hydro power) which constantly increases untill 2050. The final quota level corresponds to the long-term renewable energy objective, which is a scenario variable of the simulation (see Table 3). In case the supply of green certificates is below the quota obligation, the price is capped at 100 V/MWh. If supply exceeds demand, the model dispatches the RES generators whose offers are accepted in the green certificate market and those with null marginal cost, such as wind and PV generators. All unsold green certificates are stored in producers’ accounts and offered in successive years at the fundamental value. 3.3. Carbon reduction policies We will now explain how the carbon tax and the ETS mechanisms are implemented in the model. Under both policy mechanisms, generators internalize the cost of CO2 emissions in their cost functions when bidding into the electricity market. For the carbon market, we assume that the regulator constantly decreases the cap, down to a value of 20% of the 1990 emissions level in 2050, as indicated by the European Commission. In order to compare the two policies, we calculated a tax that achieves a similar reduction of emissions. The tax has an initial level of 50 V/tonCO2 in 2012 and increases constantly until reaching 100 V/tonCO2 in 2050. This value was determined by performing a sensitivity analysis by varying the initial level of the tax but leaving its terminal value of 100 V/MWh unchanged. The resulting carbon emissions under different Pigovian tax levels are presented in Fig. 3. Different carbon tax levels lead to very different outcomes. For example, fuel switching from coal to gas occurs at a CO2 price between 20 and 30 V/ton. The implementation of a carbon market is more complex. One option is to clear it iteratively together with the electricity market, changing the carbon price until the volume of CO2 emissions corresponds to the emissions cap (Chappin, 2011; Richstein et al., Table 2 Policy combinations simulated for each scenario run. Carbon reduction policy

1

Others include solar thermodynamic and other non-renewable energy sources financed under the Régimen Especial. 2 This is true for big generating companies that do not urgently require cash for running their businesses. Small players could require to sell green certificates more rapidly to maintain business operation (Swedish Energy Agency, 2012).

RES policy

None FITs TGCs

None

Pigovian tax

Carbon market

N-N F-N G-N

N-T F-T G-T

N-C F-C G-C

32

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

Fig. 2. Sensitivity to FIT markup levels.

2012). However, this method does not consider the option of banking emission allowances. To allow the modeling of banking, the model centrally calculates the fundamental value of a carbon allowance based on short-term and long-term price forecasts. The procedure used for calculating the fundamental value of a carbon emission allowance is similar to that used to calculate the fundamental value of a green certificate and is explained in the Section 3.4. In a first phase, the electricity market is cleared using the fundamental value as a carbon price. If the volume of emissions is lower than cap, the excess is banked. If the CO2 emissions exceed the cap, the market uses allowances that were banked during previous years. Only if these are insufficient, the model increases the carbon price above its fundamental value up to the point that emissions are limited to the emissions cap, or the price reaches the price cap of 200 V/ton. While the real EU ETS doesn’t have such a price cap, we consider prices above 200 V/ton to be unsustainable. Fig. 4 represents how the electricity, the green certificate and the carbon market are cleared iteratively. 3.4. Forecasting The forecasting block of the model is executed after the market clearing process. This part of the model estimates the fundamental values of carbon allowances and green certificates which are used as an input for the successive year’s market clearing process. Future price forecasts also serve as an input for the investment decisions algorithm to calculate the profitability of investment alternatives. The forecasting process is centralized in the model, implying the

Table 3 Base case and policy sensitivity analysis. Scenario

Carbon tax

Base case High carbon tax Low carbon tax High emission target High FITs Low FITs High RES Obj. Low RES Obj.

50e100 70e100 30e100 50e100 50e100 50e100 50e100 50e100

V/ton V/ton V/ton V/ton V/ton V/ton V/ton V/ton

RES Obj. CO2 emissions target FITs biasing factor 65% 65% 65% 65% 65% 65% 80% 50%

20% 20% 20% 10% 20% 20% 20% 20%

of of of of of of of of

1990 1990 1990 1990 1990 1990 1990 1990

level level level level level level level level

15% 15% 15% 15% 25% 5% 15% 15%

assumption that all the market actors have access to the same information. The model calculates electricity, green certificate and carbon prices for the next fifty years. Short-term price forecasts are made by considering the current market situation while the long-term forecasts are based on the evolution of fuel prices and on market efficiency. The long-term electricity price forecast is based on the assumption that the system evolves towards an optimal generation portfolio, which is calculated by assuming that generators exactly recover the average cost of generation, given expected future fuel prices (following Olsina et al., 2006). The long-term electricity price forecast is then obtained by clearing the electricity market using the expected future electricity demand and installed capacity. The short-term electricity price forecast is obtained by clearing the electricity market five years ahead. To do so, the model starts from current market condition and updates the system considering the development of electricity demand and fuel prices, the aging of existing power plants, the evolution of carbon and renewable energy policies, and the addition of generators that are currently under construction. For reasons of simplicity, the model assumes that no power plants are dismantled3 during that period. After the model has obtained the short-term and a long-term electricity price forecasts, it linearly interpolates from current to short-term and from short-term to long-term prices, obtaining an electricity price forecast for each of the next fifty years. We will now turn to how future carbon allowance and the green certificate prices are forecasted. With respect to the carbon market, the model initially calculates the long-term emissions level, assuming a CO2 price of 100 V/tonCO2. Then it iteratively adjusts the carbon price to converge towards the long-term emissions cap as indicated in Fig. 4, assuming no allowances are banked. For the short-term forecast, the carbon market is cleared iteratively, starting with the current spot price and adjusting it so emissions converge with the cap. The model also takes the volume of emissions allowances that are banked into account. They are distributed over a period of fifty years, proportionally to the future market volume of each year. The green certificate price is forecasted as follows. For the longterm forecast, the model calculates the optimal mix of renewable

3 How the agents take decisions regarding power plants dismantling is explained later in the paper.

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

33

Fig. 3. Sensitivity analysis of carbon tax level.

P50 F ¼

EðPiÞDi i ¼ 1 ð1þrÞi

P50

i¼1

(4)

Di

Where E(Pi) is the expected price for year i, Di is the carbon market cap/green certificate quota for year i, and r is the discount rate. Thus, market agents would be indifferent between paying the fundamental price in today’s market and waiting and buying in later years at the estimated prices. However, to reflect price uncertainty and generators’ risk aversion, the discounted price is adjusted by multiplying it by a factor which varies between zero and one depending on the number of carbon allowances/green certificates banked. See Equation (5), in which B indicates the total amount of carbon allowances/green certificates banked. Fig. 4. Representation of the clearing process of the three markets.

technologies that complies with the future quota requirement, deploying resources with a lower average generation cost first. Then, it calculates the average cost of the marginal generator and subtracts the corresponding forecasted electricity price, obtaining the long-term equilibrium price for the green certificate market. For the short-term forecast, the model calculates the equilibrium price of a green certificate as the difference between the average generation cost of the marginal plant currently present in the market and the expected average electricity price five years ahead. In this case, the demand for green certificates is reduced, assuming that banked certificates are uniformly distributed over the successive fifty years. Once the model has forecasted the prices of carbon allowance and green certificate markets for the next fifty years, it proceeds by calculating the fundamental value of these variables. These are used to clear the two markets, as explained in the previous sections. The fundamental value equals the present value of the estimated future prices, weighted by the future market volume (the carbon emissions cap or green certificate quota in each year). More specifically, the present value is calculated by summing the discounted4 cost of each year’s required carbon allowances/green certificates at the corresponding estimated price, divided by the total amount of carbon allowances/green certificates required for the next fifty years as indicated in Equation (4).

4 The discount rate is the same used to evaluate new projects and its calculation is explained in next section.

FADJ ¼ F 

B 1  P50

i¼1

! Di

(5)

In this way, generators’ willingness to pay in today’s market is lower when more carbon allowances/green certificates are banked.

3.5. Investment decision process The last block of the model is the algorithm that simulates investors’ decisions regarding the construction and dismantlement of power plants. This block is executed after the market clearing and the forecasting block of the model. Seven different agents who make investment decisions are simulated in the model. Six agents represent electricity companies, each of which can start the construction of two or three power plants per year, depending on the conditions of the electricity, carbon, and green certificate markets (i.e., when there is shortage in a market the model allows for more investments). A last agent represents financial institutions attracted by the subsidies given to RES. This agent aims to reflect the fact that renewable energy subsidies have attracted a wider public than energy utilities, including global corporations with an appetite for investments in multiple sectors (Ernst and Young, 2013). This agent behaves the same as the other agents but is only allowed to invest in RES technologies. Investments are made on the base of profit maximization, so agents invest in projects that promise the highest positive expected Net Present Value (NPV), subject to availability of funds.

34

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

Investments have a fixed capital structure of 40% equity (E) and 60% debt (D), and future cash flows are discounted using the Weighted Average Cost of Capital (WACC), supposing in the base case scenario a cost of equity (Ke) of 15%, a cost of debt (Kd) of 5% and a tax rate (Tax) of 30% (6).

WACC ¼

Ke  E Kd  D þ  ð1  TaxÞ EþD EþD

(6)

For the sake of this analysis, investment decisions are simply based on a stand-alone project evaluation and investors do not consider portfolio effects or externalities caused by lower market prices. This way, the model simulates a market without barriers to entry in which incumbents do not exercise market power and are not protected from entry by new competitors. The model considers the impact of fuel price uncertainty when calculating the profitability of alternative investment projects, reflecting how different technologies are affected by fluctuating prices. This is done by considering a risk adjusted measure of the NPV. This approach also allows for incorporating the additional price risk that is introduced by quantity-based policy mechanisms, compared to price-based mechanisms. The model uses the Conditional Value at Risk (CVaR) as a measure of risk (Rockafellar and Uryasev, 2000). The CVaR is defined as the expected loss in cases when the NPV is lower than a given Value at Risk (VaR). By definition, the probability that the NPV is lower than the VaR under a confidence level a is 1  a. See Fig. 5. The NPV is adjusted for risk by subtracting the CVaR from the expected NPV, as indicated in (7).

NPVRiskAdj ¼ EðNPVÞ  b  CVaR

(7)

The CVaR is multiplied by a risk-aversion factor b, which is an input variable of the simulation that indicates investors’ risk preference. A b of zero indicates risk-neutral investors while higher values correspond to increasing levels of risk-aversion. For the base case scenario of the simulation we assume investors to be riskaverse, applying a b of one. To calculate the profit distribution and eventually the riskadjusted NPV of each technology, the model first calculates the optimal long-term generation capacity, based on average expected future fuel prices (as discussed in the previous section). Then 3000 fuel price scenarios are randomly generated using a Weibull distribution which is obtained from the shape of monthly natural gas and coal price data from the International Monetary Fund and the World Bank covering the period October 1998eDecember 2008. The model clears the electricity, the green certificate, and the carbon allowance

markets in each fuel price scenario, obtaining 3000 long-term price forecasts. The model also generates 3000 random short-term electricity, green certificate, and carbon prices, based on past volatility and short-term forecasts, assuming un-biased, uncorrelated normal distributions. Finally, the model linearly interpolates between the current, short-term, and long-term prices, obtaining 3000 annual prices scenarios. These are then used to calculate the expected NPV of each technology, taking into account the evolution of marginal costs in each scenario and its impact on the merit order. Generation companies also make decisions regarding the dismantling of existing power plants. The model assumes that after reaching their expected service life, power plants are dismantled once they experience two or three years of consecutive losses, with half the agents dismantling after two years of losses and the other half after three. 4. Model simulations and results 4.1. Performance indicators A simulation consists of nine model runs, each one representing a different policy combination as indicated in Table 2. As a central metric to evaluate the simulation results we use the additional cost of a policy to society, normalized to one unit of electricity generation (a MWh). To evaluate how the cost of a policy is distributed between different actors in the market, we divide the effects of a policy into changes of welfare to consumers, producers, and government finances, as compared to a base case scenario where no policies are implemented (N-N). Additional revenues to manufacturers are not considered here, since this would require a distinction between imports and national value added in the production of power plants, which is outside the scope of this work. The policy cost to society per MWh of produced electricity is calculated as expressed in Equation (8).

PT PolicyCost ¼

t ¼ 1 ðDElet

 DEPt þ St  Ct Þ=ð1 þ rÞt PT t ¼ 1 Et

½V=MWh (8)

DElet represents the difference in consumers’ electricity bills in year t compared to the base case (N-N). This is calculated considering the change in the electricity price but without including the additional cost of the RES subsidy. An increase in the electricity bill corresponds to a decrease in the social welfare. Secondly, DEPt indicates the difference in producer surplus, measured as the

Fig. 5. Graphic representation of VaR and CVaR of a NPV distribution.

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

35

Fig. 6. Cost increments to society.

difference in the industry profits using the concept of Economic Profit (EP). In this case, higher industry profits increase social welfare. The Economic Profit is defined as the part of profits exceeding the required return on the book value of assets.

EP ¼ EBIT  ð1  TaxÞ  Book  WACC

(9)

Here EBIT indicates the earnings before interest and taxes, Tax represents the tax rate, and book is the balance sheet value of a power plant. Next, we calculate the cost of the renewable energy subsidy, the carbon tax revenues and the auctions of CO2 credits. While the costs of renewable policies are sometimes included in the electricity tariffs and thus affect the electricity consumer price, for the sake clarity we attribute the cost directly to the government. St represents the cost of subsidizing RES in year t, while Ct is the government income from the carbon reduction policy. Finally, the overall policy cost to society is defined as the sum of each year’s policy cost discounted to the present (using a social discount rate r of 3%), divided by the total electricity production over the simulated P period Tt¼ 1 Et . We define the efficiency and effectiveness of the investigated policies with respect to the two policy goals of CO2 emissions reduction and RES generation. We measure the effectiveness of the carbon reduction goal by dividing the accumulated CO2 emissions

by the total emissions cap (over all the years under consideration) in the corresponding ETS-only scenario (thus the lower, the better). We measure effectiveness of the renewable production goal as the percentage of renewable electricity production in the total electricity production over the entire simulation period. Efficiency in reaching the carbon reduction goal is measured as the change in the policy cost to society (in absolute values, not per MWh) divided by the tons of CO2 avoided, as compared to the scenario without any policy intervention. Similarly, the efficiency in reaching the renewable policy goals is defined as the change of the policy cost to society divided by the total renewable electricity production in MWh. 4.2. Policy effectiveness and efficiency A ‘pure’ emission market (without a renewable energy policy instrument) reaches its carbon reduction goals at a discounted cost to society of 5.15 V/MWh over the simulation period. However, it misses the carbon reduction aims in the final years of the simulation leading to high final CO2 prices (see also Fig. 8, with No RES support). This effect can be explained as the consequence of an investment cycle in low-carbon technologies. Temporarily low carbon prices and imperfect forecasting may attract insufficient investment in low-carbon technology for complying with the very

Fig. 7. Policy efficiency.

36

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

strict reduction target assumed for 2050. A tax (50e100 V/ton) was found to achieve similar or even better efficiency at a slightly higher level of carbon emissions than a carbon market (4.95 V/MWh cost to society, with a 14.9% overshoot of emissions over the entire simulation period). Like the carbon market, the green certificate market is effective in reaching its RES deployment and production targets. It does so at a total cost to society of 8.68 V/MWhRES. This is less than the cost of feed-in tariffs due to the more gradual and controlled introduction of the cheapest RES technologies to the market in a certificate market. On the other hand, the feed-in mechanism introduces renewables into the market more effectively, but the overall cost to society is higher due to its open-ended nature and more expensive resources being subsidized. Feed-in tariffs cause electricity prices to be lower, but this benefit is more than offset by the subsidy cost (See Fig. 6, which shows the terms of Equation (8)). Nonetheless, feed-in tariffs out-compete green certificate markets in terms of cost-efficiency if the tariff level is calculated correctly. See Fig. 7. This is because renewables are introduced earlier, which leads to more RES production over the simulated period. Moreover, the green certificate market introduces some inefficiency by providing the same price to all technologies. This causes windfall profits for cheaper generators, which reduces efficiency. Fig. 7 also indicates that pure carbon policies are not efficient in introducing RES to the market. Similarly, pure renewable energy policies demonstrate poor efficiency in reducing carbon emissions. When these policies are implemented together, overall policy costs increase. As a result, each policy instrument becomes less efficient on its own. See Fig. 7. However, the policy goals of carbon reduction and renewable energy development complement each other, so the costs of the policies do not simply add up: the total cost of the two policies together is far less than the sum of the individual policy costs, while reaching the same policy effectiveness, as can be seen in Fig. 6. This confirms that both policies are necessary to reach carbon reduction and renewable energy targets efficiently.

above the cap. However, assuming investors do not have such a long time horizon would lead to more investments in carbon intensive technologies, jeopardizing the long-term emission target. The combination of carbon and green certificate markets leads to more stable and lower carbon prices in the emission trading scheme. The investment cycle in low-carbon power plants, which leads to high CO2 prices at the end of the simulation in the no-RES policy case, is dampened by the investment induced by the green certificate market (cf. Fig. 8). The overall difference in the effectiveness in reducing CO2 emissions over the entire simulation period is quite small. Combining green certificate and carbon markets leads to a higher carbon reduction of about 5%, as compared to a pure carbon market. See Table 11. We would expect the effect of carbon policies on the green certificate markets to be similar to that of renewable policies on a carbon market. Both a carbon tax and an emission trading scheme are found to lower the green certificate price. However, the green certificate market is found to be more robust than the carbon market under our assumptions and prices only drop to zero towards the end of the simulation if enough certificates are available to meet renewable energy targets of 2050, as can be seen in Fig. 9. The green certificate price is more stable in combination with a carbon reduction policy, which eliminates price peaks. In contrast to the relatively stable prices, investment in renewable energy is subject to investment cycles, as becomes evident in the lower part of Fig. 9. The banking of green certificates allows for a stable price despite these investment cycles. However, our assumptions about the long-term horizon of investors may be too optimistic. If short-term considerations prevail in investments decision, the volatility of the investment cycles may increase, which could cause the green certificate price to collapse to zero periodically.

4.4. Sensitivity analyses In addition to the base case scenario, we tested the sensitivity of the results to the impact of different regulatory decisions on the simulation by changing the following input variables:

4.3. Policy interactions A feed-in tariff leads to strong initial investments in RES technologies, which in turn reduce CO2 emissions leading to lower CO2 prices. The expectation of increasing carbon prices is sufficient to achieve the CO2 reduction objectives over the modeled time horizon, although towards the end of the period, the emissions are

   

The The The The

long-term RES objective; feed-in tariff biasing factor; carbon tax level; long-term CO2 emission target for the carbon market.

Fig. 8. The impact of renewable energy policy on CO2 Emissions and the carbon market price.

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

We tested the effect of macro-economic variables by running the base case and the policy sensitivity analyses under different hypothesis about economic conditions by changing the following input variables:    

Electricity demand growth rate; Investor risk-aversion; The cost of equity; The growth rates of fuel prices.

The input variables of the base scenario and the different policy sensitivity analysis are summarized in Table 3. The CO2 tax is assumed to start from a value of 50 V/ton and constantly increases to a price of 100 V/ton in 2050. For the sensitivity analysis, the starting value of the carbon tax is changed to 30 and 70 V/ton, leaving the final price of 100 V/ton in 2050 unchanged. With respect to the RES objective, we assume a target of 65% renewable electricity production for the base case scenario, at which point the regulator stops subsidizing renewable energy. Sensitivities are run with limits of 50% and 80% market penetration. The CO2 emissions target represents the long-term objective imposed by the regulator when a carbon market is in place. At the European level, policy makers are discussing a reduction of 80e 90%, compared to 1990 levels, in 2050. We use the 80% reduction target in the base case scenario and perform a sensitivity analysis with the more stringent 90% limit. Finally, the feed-in tariff biasing factor aims to reflect the regulator’s uncertainty in the estimation of generation costs use to calculate the tariffs paid to renewable generators. In the base case scenario, the regulator adds a mark-up of 15% to the exact average generation cost to guarantee a certain return to investors so they prefer RES over conventional technologies. The sensitivity analyses are run using a biasing factor of 5% and 25% to simulate the case of the regulator under- or overestimating the exact costs. With respect to macro-economic variables, we change the electricity demand growth rate after 2020 from 1.9% to 1.7% in order to reflect a scenario in which energy efficiency measures are more effective. The demand growth rate before 2020 is not changed and assumed to be 1.5% in both scenarios. We also test the impact of investor risk aversion on the simulation results by changing the CVaR weight factor (b) from 1 to 0, which implies that investors are riskneutral. In order to study the impact of the cost of capital on investment decisions, we run sensitivity analyses with a cost of equity of 10%

37

and 20%. Finally, we test the impact of fuel prices on our analysis by changing the annual increase rate of each fuel price by 0.7%. The result of this sensitivity analysis is measured in terms of the total cost to society and of the policy effectiveness. The results are presented in the appendix. The impacts of the feed-in tariff biasing factor and of the tax level were already discussed in the previous section. The biasing factor was already optimized; a higher feed-in tariff increases the subsidy cost without significantly affecting the effectiveness of the policy, while a lower feed-in tariff performs poorly in terms of effectiveness. In case of a green certificate market, the effects of quota changes are straight-forward: a higher/ lower quota leads to higher/lower subsidy cost and policy effectiveness. Similarly, a higher/lower carbon tax leads to higher/lower carbon reduction effectiveness at a higher/lower cost to society. On the other hand, reducing the cap from 20% to 10% of 1990 emission levels only slightly affects the outcome of a carbon market. Regarding the macro-economic parameters, renewable energy sources can enter the electricity market more easily under a pure carbon policy scheme if fuel prices are high. The effect of a carbon tax depends on fuel prices, but a high enough feed-in tariff is effective in all scenarios. Higher and lower interest rates for investments (WACC) mainly affect the cost to society of both carbon and renewable policies, since capital costs are higher for non-renewable low carbon technologies, such as CCS, and renewable technologies, than for conventional generation. So the costs to society rise with rising interest rates, as expected, because financing becomes more expensive. The high efficiency scenario does not have a significant impact on the results because they are normalized to electricity production. An exception is represented by the green certificate market, the cost of which reduces because less expensive technologies are required to comply with the lower certificate demand. If investors are assumed to be risk-neutral, this has several effects. Firstly, the overall policy cost of the green certificate market alone is lower in the risk-neutral case than in the base case because generators demand a lower risk premium. Secondly, an ETS without a renewable support achieves far higher volumes of renewable electricity than the base case scenario. The explanation for this lies in the higher market risk that renewable generators face, since they are not naturally hedged against fuel-price induced electricity price movements, as for example are gas power plants (Roques et al., 2008). The effectiveness of feed-in tariffs is not affected by investor risk aversion because this mechanism

Fig. 9. The impact of carbon reduction policy on RES production and the green certificate price.

38

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41

eliminates price risk in our analysis. However, due to the lower electricity price caused by a higher level of investments, the policy cost of feed-in tariffs is higher in this case. After presenting these sensitivity analyses, we now explain the limitations of our model. First of all, we do not truly represent the intermittence of some RES in the model, so their integration cost and impact on dispatchable generators may be underestimated. In addition, we assume that investors consider a very long time horizon, while actual companies appear more oriented towards the short term. This would lead to more volatile market prices and perhaps more cyclical investment behavior than we observe. 5. Conclusions We present an agent-based model that reproduces the evolution of a power sector, loosely based on Spain, between 2012 and 2050. We simulate how the investment decisions of risk-averse, profitmaximizing generating companies are affected by carbon reduction and renewable energy policies. The goal of this paper is to investigate the dynamic interactions between these two policies. Our analysis suggests a single policy is not a cost-efficient way of achieving both a reduction of CO2 emissions and an increase in renewable electricity generation. Hence, the decision of the European Commission and of national governments to introduce renewable energy support mechanisms in addition to the EU ETS is sensible because it reduces the total cost of achieving a more sustainable electricity sector. This finding corroborates what Linares et al. (2008) found with an equilibrium model. We found that the combination of carbon reduction and renewable energy policies leads to lower and more stable costs for both policies. However, while the introduction of a carbon reduction mechanism has limited effect on a green certificate market, renewable energy policy has a stronger impact on a carbon market. A high volume of renewable electricity generation could lead to low prices in carbon markets, as has been the practical experience with the EU ETS and recognized in the literature. Our model indicates that there is also an adverse long-term dynamic effect, namely that low carbon prices may attract investment in coal-fired generators, which could lock the electricity sector into a pathway of higher future emissions (as appears to have happened in the Netherlands). To avoid periods of low carbon prices, regulators may opt for a hybrid policy instrument such as a carbon price floor. The UK introduced a national price floor for carbon, within the EU ETS framework, in April 2013. If set high enough above the market price for carbon, a price floor may effectively function as a carbon tax. The question whether such hybrid instruments work well merits further research. Our analysis demonstrates how difficult it is for policy makers to estimate the optimal level of a price-based policy. However, quantitybased mechanisms may also result in a price which does not satisfy the regulator because is too high or too low. We observe that both price and quantity policies may perform poorly when they are implemented using static parameters, as the dynamic changes to external conditions affect the prices and effectiveness of the policies. A solution may be provided by adaptive policy making. With respect to the carbon market, policy makers may intervene by lowering the emission cap faster/slower if the price is too low/high, in an attempt to keep the price within a certain accepted level. This way, an aggressive renewable energy policy would contribute to an early achievement of carbon reduction targets, rather than in low carbon costs for fossil fuel plant. An example may be found in the July 2013 decision of the European Parliament to temporarily reduce the number of allowances auctioned for the EU ETS (backloading), although this was not part of a structural policy.

Nonetheless, policy makers should avoid frequent and unpredictable policy changes as this would contribute to regulatory uncertainty, which in general discourages investment in the energy sector (Fagiani and Hakvoort, 2014). A solution is to bind policy adaptations to predictable rules, for instance to announce that backloading will take place if the carbon price is below a certain level. Another example of how regulatory uncertainty could be limited is to reduce the number of carbon allowances as a function of renewable electricity production. In practice, the regulator could, for example, calculate the reduction in carbon emissions obtained by one MWh of renewable electricity substituting conventional technologies, decreasing adequately the cap of the carbon market. The carbon equivalent of one MWh of renewable electricity would depend on the carbon intensity of the electricity sector, more precisely on which power plants the RES substitute in the merit order curve, which might be difficult to estimate. However, by doing so the regulator could offset the negative impact of renewable energy policy on the carbon market, obtaining a mechanism that would be more robust to the dynamic evolution of the electricity sector. If such an announced dynamic adjustment of emission caps is effective and feasible in practices is another topic for future research. Acknowledgments Riccardo Fagiani and Jörn C. Richstein have been awarded Erasmus Mundus Joint Doctorate Fellowships. The authors would like to express their gratitude towards all partner institutions within the program as well as the European Commission for their support. APPENDIX Additional Cost to society (V/MWh)

Table 4 Base case scenario. RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.00 e e e e e e e

8.68 e e e e e 10.84 8.02

4.95 5.36 3.81 e e e e e

5.15 e e 4.79 e e e e

12.90 e e e 16.61 5.53 15.01 10.57

13.18 13.13 13.26 e 14.70 9.76 14.37 11.55

Green certificate market

13.15 e e 13.11 14.46 8.45 14.57 11.80

9.95 10.40 9.04 e e e 11.40 8.58

10.58 e e 11.57 e e 11.72 9.10

Table 5 High efficiency scenario. RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.00 e e e e e e e

7.50 e e e e e 10.38 7.35

4.78 5.31 4.03 e e e e e

5.34 e e 5.07 e e e e

12.82 e e e 14.92 5.77 14.96 10.45

12.97 12.52 12.97 e 14.29 9.71 14.34 11.07

Green certificate market

13.08 e e 13.08 14.42 7.88 14.56 11.64

11.79 10.40 11.39 e e e 11.34 8.54

10.31 e e 11.92 e e 12.42 6.76

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41 Table 6 High fuel price scenario. RES policy

None

CO2 policy Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

39

Table 10 Low weighted average cost of capital scenario. Feed-in tariff

Green certificate market

RES policy

None

None Pig. Tax ETS None Pig. Tax ETS

None Pig. Tax ETS

CO2 policy

None Pig. Tax ETS None Pig. Tax ETS

None Pig. Tax ETS

0.00 e e e e e e e

8.53 e e e e e 10.29 5.54

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.00 e e e e e e e

6.84 e e e e e 10.44 5.52

6.01 6.14 4.70 e e e e e

4.56 e e 4.61 e e e e

12.69 e e e 16.78 6.15 14.64 10.57

12.99 12.94 13.16 e 16.91 10.11 13.95 11.77

13.00 e e 13.08 16.27 9.11 14.39 11.79

10.40 13.28 8.37 e e e 12.18 8.80

5.78 e e 5.44 e e 10.15 5.39

Feed-in tariff

4.72 4.87 3.79 e e e e e

4.84 e e 4.92 e e e e

12.05 e e e 13.90 4.70 13.46 8.87

11.30 10.98 11.26 e 14.36 8.92 11.95 10.10

Green certificate market

10.97 e e 11.11 14.08 7.69 12.24 9.85

5.85 6.06 9.32 e e e 9.50 5.23

6.89 e e 9.76 e e 12.02 9.07

Table 11 Base case scenario. Table 7 Low fuel price scenario. RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.00 e e e e e e e

11.89 e e e e e 14.99 8.25

4.06 4.47 3.40 e e e e e

4.66 e e 5.17 e e e e

13.08 e e e 17.45 5.29 15.35 10.83

13.16 12.99 13.36 e 14.81 8.29 14.72 11.33

Green certificate market

13.06 e e 13.07 14.62 8.16 14.38 11.69

11.54 10.73 10.88 e e e 11.99 10.39

12.74 e e 12.69 e e 12.92 9.08

RES policy

None

Feed-In tariff

Green certificate market

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

4.13 4.13 4.13 4.13 4.13 4.13 4.13 4.13

1.58 e e e 1.58 3.67 0.86 2.36

3.18 e e e e e 2.86 3.66

1.15 1.04 1.56 e e e e e

1.04 e e 1.05 e e e e

0.44 0.43 0.54 e 0.43 0.84 0.33 0.53

0.87 e e 0.79 0.87 1.09 0.63 0.90

0.91 0.86 1.24 e e e 0.84 0.98

0.99 e e 1.00 e e 0.94 1.02

Carbon Reduction Effectiveness (ratio of emissions compared to scenario cap).

Table 8 Risk-neutral scenario. RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.00 e e e e e e e

8.07 e e e e e 11.36 3.68

5.75 6.32 4.75 e e e e e

5.97 e e 6.26 e e e e

13.84 e e e 18.00 6.33 16.28 11.33

14.60 14.13 14.30 e 18.70 9.99 15.99 12.17

Green certificate market

14.29 e e 14.43 18.44 10.11 16.62 12.54

11.06 8.63 7.93 e e e 10.45 8.09

11.28 e e 12.21 e e 12.34 8.75

RES policy

None

CO2 policy

None Pig. Tax ETS None Pig. Tax ETS 0.00 e e e e e e e

Feed-in tariff

4.94 5.36 4.32 e e e e e

5.92 e e 6.04 e e e e

15.26 e e e 20.17 6.26 17.61 12.47

15.61 15.36 15.41 e 17.26 9.82 16.76 13.24

RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

3.94 e e e e e e e

1.64 e e e 1.62 3.45 0.97 2.36

2.98 e e e e e 2.75 3.54

1.19 1.08 1.43 e e e e e

1.00 e e 1.03 e e e e

0.42 0.39 0.52 e 0.41 0.81 0.33 0.57

Green certificate market

0.82 e e 0.80 0.80 1.08 0.65 0.83

0.89 0.84 1.19 e e e 0.80 0.96

1.02 e e 1.00 e e 0.99 1.00

Table 13 High fuel price scenario.

Table 9 High weighted average cost of capital scenario.

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

Table 12 High efficiency scenario.

Green certificate market

15.25 e e 15.25 16.84 9.41 17.05 13.57

None Pig. Tax ETS 10.78 e e e e e 13.99 5.39

11.68 12.39 11.38 e e e 13.59 8.49

11.36 e e 12.21 e e 15.01 9.44

RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

4.28 e e e e e e e

1.56 e e e 1.56 3.60 0.95 2.32

3.16 e e e e e 2.86 3.52

0.96 0.87 1.44 e e e e e

1.01 e e 1.00 e e e e

0.39 0.39 0.55 e 0.39 0.86 0.31 0.53

Green certificate market

0.89 e e 0.79 0.89 1.06 0.73 0.91

0.86 0.82 1.17 e e e 0.80 0.96

1.00 e e 1.00 e e 0.96 1.02

40

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41 Table 18 Base case scenario.

Table 14 Low fuel price scenario. RES policy

None

Feed-in tariff

Green certificate market

RES policy

None

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

3.17 e e e e e 3.06 3.49

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25

0.80 e e e 0.81 0.38 0.93 0.65

0.48 e e e e e 0.55 0.42

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

3.78 e e e e e e e

1.24 1.15 1.55 e e e e e

1.04 e e 1.03 e e e e

1.52 e e e 1.50 3.75 0.77 2.48

0.46 0.50 0.55 e 0.46 0.93 0.36 0.60

0.77 e e 0.78 0.74 1.08 0.63 0.92

1.03 0.93 1.24 e e e 0.86 1.04

0.98 e e 1.00 e e 0.95 1.00

Feed-in tariff

0.28 0.36 0.27 e e e e e

0.35 e e 0.37 e e e e

0.81 0.81 0.81 e 0.82 0.39 0.92 0.66

Green certificate market

0.80 e e 0.80 0.81 0.40 0.92 0.65

0.50 0.50 0.49 e e e 0.57 0.43

0.51 e e 0.49 e e 0.56 0.43

RES Production Effectiveness (percentage of electricity produced by RES).

Table 15 Risk-neutral scenario. RES policy

None

Feed-in tariff

Green certificate market

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

4.37 e e e e e e e

1.58 e e e 1.58 3.72 0.88 2.37

3.21 e e e e e 3.08 3.27

1.10 0.92 1.34 e e e e e

0.99 e e 0.95 e e e e

0.44 0.43 0.53 e 0.43 0.85 0.33 0.57

0.77 e e 0.77 0.77 1.09 0.59 0.89

0.89 0.72 1.10 e e e 0.79 0.90

0.99 e e 0.99 e e 0.94 1.01

Table 19 High efficiency scenario. RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.26 e e e e e e e

0.79 e e e 0.79 0.40 0.92 0.64

0.49 e e e e e 0.54 0.42

0.30 0.33 0.33 e e e e e

0.36 e e 0.37 e e e e

0.79 0.79 0.80 e 0.80 0.40 0.91 0.65

Green certificate market

0.78 e e 0.78 0.79 0.41 0.91 0.64

0.49 0.51 0.49 e e e 0.58 0.43

0.48 e e 0.49 e e 0.55 0.48

Table 16 High weighted average cost of capital scenario. RES policy

None

Feed-in tariff

Green certificate market

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

4.14 e e e e e e e

1.55 e e e 1.55 3.62 0.85 2.37

3.34 e e e e e 3.10 3.41

1.31 1.24 1.54 e e e e e

1.04 e e 1.01 e e e e

0.49 0.50 0.58 e 0.46 1.00 0.37 0.60

0.81 e e 0.83 0.78 1.13 0.63 1.01

1.03 1.01 1.29 e e e 0.96 1.09

0.90 e e 0.97 e e 0.93 1.02

Table 17 Low weighted average cost of capital scenario. RES policy

None

Table 20 High fuel price scenario. RES policy

None

Feed-in tariff

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.26 e e e e e e e

0.81 e e e 0.81 0.38 0.93 0.66

0.49 e e e e e 0.55 0.43

0.44 0.43 0.34 e e e e e

0.47 e e 0.49 e e e e

0.80 0.80 0.81 e 0.81 0.40 0.91 0.66

Green certificate market

0.80 e e 0.80 0.81 0.42 0.93 0.66

0.52 0.48 0.53 e e e 0.57 0.45

0.55 e e 0.52 e e 0.57 0.48

Table 21 Low fuel price scenario.

Feed-in tariff

Green certificate market

RES policy

None

Feed-in tariff

Green certificate market

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

4.14 e e e e e e e

1.58 e e e 1.56 3.62 0.97 2.37

3.01 e e e e e 2.83 3.45

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.26 e e e e e e e

0.80 e e e 0.81 0.40 0.93 0.65

0.47 e e e e e 0.51 0.41

0.80 0.77 1.21 e e e e e

0.97 e e 0.95 e e e e

0.39 0.36 0.49 e 0.38 0.82 0.29 0.49

0.86 e e 0.78 0.86 1.01 0.72 0.89

0.68 0.65 1.11 e e e 0.68 0.69

0.99 e e 1.01 e e 1.00 1.00

0.28 0.28 0.29 e e e e e

0.31 e e 0.31 e e e e

0.81 0.81 0.80 e 0.81 0.39 0.91 0.66

0.79 e e 0.79 0.81 0.41 0.90 0.66

0.47 0.50 0.47 e e e 0.55 0.40

0.46 e e 0.46 e e 0.55 0.42

R. Fagiani et al. / Utilities Policy 28 (2014) 28e41 Table 22 Risk-neutral scenario. RES policy

None

Feed-in tariff

Green certificate market

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.27 e e e e e e e

0.81 e e e 0.81 0.38 0.94 0.65

0.48 e e e e e 0.52 0.43

0.32 0.46 0.39 e e e e e

0.50 e e 0.45 e e e e

0.82 0.81 0.82 e 0.82 0.39 0.94 0.66

0.81 e e 0.81 0.81 0.41 0.95 0.66

0.50 0.53 0.56 e e e 0.58 0.48

0.50 e e 0.49 e e 0.57 0.47

Table 23 High weighted average cost of capital scenario. RES policy

None

Feed-in tariff

Green certificate market

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.25 e e e e e e e

0.81 e e e 0.81 0.39 0.93 0.65

0.47 e e e e e 0.50 0.42

0.28 0.28 0.32 e e e e e

0.32 e e 0.37 e e e e

0.81 0.81 0.81 e 0.82 0.39 0.91 0.66

0.80 e e 0.80 0.81 0.40 0.92 0.66

0.48 0.48 0.46 e e e 0.53 0.42

0.49 e e 0.51 e e 0.54 0.44

Table 24 Low weighted average cost of capital scenario. RES policy

None

Feed-in tariff

Green certificate market

CO2 policy

None Pig. Tax ETS

None Pig. Tax ETS

None Pig. Tax ETS

Base case High Pig. Tax Low Pig. Tax Low ETS Cap High FIT Low FIT High RES Obj. Low RES Obj.

0.25 e e e e e e e

0.79 e e e 0.80 0.40 0.91 0.65

0.50 e e e e e 0.55 0.42

0.48 0.51 0.43 e e e e e

0.44 e e 0.47 e e e e

0.81 0.80 0.81 e 0.81 0.42 0.92 0.65

0.80 e e 0.80 0.81 0.42 0.92 0.65

0.55 0.57 0.52 e e e 0.56 0.54

0.53 e e 0.48 e e 0.54 0.44

References Amundsen, E.S., Nese, G., 2009. Integration of tradable green certificate markets: what can be expected? J. Policy Model. 31, 903e922.

41

Chappin, E.J.L., 2011. Simulating Energy Transitions. PhD Thesis. Delft University of Technology, Delft. Chappin, E., Dijkema, G., de Vries, L., 2010. Carbon policies: do they Deliver in the long run? In: Sioshansi, F.P. (Ed.), Generating Electricity in a Carbon-constrained World. Academic Press, pp. 31e56 (Ch. 2). 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, 4743e4752. Del Rio, P., 2012. The dynamic efficiency of feed-in tariffs: the impact of different design elements. Energy Policy 41, 139e151. Ernst, Young, 2013. Renewable Energy Country Attractiveness Indices. European Commission, 2009. Directive 2009/28/EC (Brussels). European Commission, 2010. COM(2010) 639 def.: Energy 2020 a Strategy for Competitive, Sustainable and Secure Energy (Brussels). European Commission, 2011. COM(2011) 885 def.: Energy Road Map 2050 (Brussels). Fagiani, R., Barquin, J., Hakvoort, R., 2013. Risk-based assessment of the costefficiency and the effectivity of renewable energy support schemes: certificate markets versus feed-in tariffs. Energy Policy 55, 648e661. Fagiani, R., Hakvoort, R., 2014. The role of regulatory uncertainty in certificate markets: a case study of the Swedish/Norwegian market. Energy Policy 65, 608e618. Gross, R., Blyth, W., Heptonstall, P., 2010. Risks, revenues and investment in electricity generation: why policy needs to look beyond costs. Energy Econ. 32, 796e804. Grubb, M., Newberry, D., 2007. Pricing Carbon for Electricity Generation: National and International Dimension (Working paper). Haas, R., Panzer, C., Resch, G., Ragwitz, M., Reece, G., Held, A., 2011. A historical review of promotion strategies for electricity from renewable energy sources in EU countries. Renew. Sustain. Energy Rev. 15, 1003e1034. Hepburn, C., 2006. Regulation by prices, quantities, or both: a review of instrument choice. Oxf. Rev. Econ. Policy 22, 226e247. Jensen, S.G., Skytte, K., 2003. Simultaneous attainment of energy goals by means of green certificates and emission permits. Energy Policy 31, 63e71. Lecuyer, O., Quirion, P., 2013. Can Uncertainty Justify Overlapping Policy Instruments to Mitigate Emissions?. Linares, P., Javier Santos, F., Ventosa, M., 2008. Coordination of carbon reduction and renewable energy support policies. Clim. Policy 8, 377e394. Menanteau, P., Finon, D., Lamy, M.-L., 2003. Prices versus quantities: choosing polices for promoting the development of renewable energy. Energy Policy 31, 799e812. Olsina, F., Garces, F., Haubrich, H.J., 2006. Modeling long-term dynamics of electricity markets. Energy Policy 34, 1411e1433. Rathmann, M., 2007. Do support systems for RES-E reduce EU-ETS-driven electricity prices? Energy Policy 35, 342e349. Red Electrica de Espana, 2012. El sistema electrico espanol 2011 (Madrid). Richstein, J., Chappin, E., De Vries, L., 2012. Impacts of the introduction of CO2 price floors in a two-country electricity market model. In: 12th IAEE European PhDstudents Day. Venice, 2012. Rockafellar, R.T., Uryasev, S., 2000. Optimization of conditional value-at-risk. J. Risk 2, 21e41. Roques, F.A., Newbery, D.M., Nuttall, W.J., 2008. Fuel mix diversification incentives in liberalized electricity markets: a mean-variance portfolio theory approach. Energy Econ. 30, 1831e1849. Simon, H.A., 1957. Models of Man: Social and Rational. John Wiley and Sons, New York. Swedish Energy Agency, 2012. The Electricity Certificate System. Tinbergen, J., 1952. On the Theory of Economic Policy, second ed. North-Holland Publishing Company, Amsterdam. Weitzman, M.L., 1974. Prices vs. quantities. Rev. Econ. Stud., 477e490.