Supply Function Equilibrium in Electricity Markets using Smooth Capacity Constraints Morten Hørmann ([email protected]) and Mikkel T. Kromann ([email protected]). COWI A/S, Parallelvej 2, DK-2880 Lyngby, Denmark Abstract. In this working paper it is shown how the Supply Function Equilibrium (SFE) concept can be extended to describe smooth capacity constraints using strict convexly increasing marginal costs and supply functions. The SFE model framework with convexly increasing marginal costs and supply functions is simulated numerically. The smooth capacity constraint concept can also be used for transmission constraints, a completely novel feature for SFE models. Finally, as the simulation model has a very appealing simulation time, it has also been possible to implement the intertemporal decisions of price setting hydro producers. Market power, electricity markets, Supply Function Equilibrium, SFE, capacity constraints, transmission constraints, hydro power, intertemporal optimisation. Keywords: 1. Introduction Imperfect competition on the electricity market has been the subject of intense investigation since the 1990's, where a wave of electricity market liberalisations started. Two diering model traditions, Cournot-Nash Equilibrium and Supply Function Equilibrium have been applied to analyse the eects of potential imperfect competition in the liberalised markets. The supply function approach introduced by Grossman (1981) and Klemperer and Meyer (1989) stipulates that in some instances rms may not compete in quantities as in Cournot or in prices as in Bertrand. Rather they submit bid-like continuous functions representing their willingness to supply a given quantity at a given price. Especially when facing uncertainty of demand, it may be preferable for the rms to state a exible strategy, such that low and high demand are not met by the same supply or the same price. Unfortunately, the analytical framework of supply functions turns out to be much more complex than that of Cournot and Bertrand. The equilibrium solution to a market of N functions must then derived from a set of rms competing in supply N rst order dierential equations. c 2005 Kluwer Academic Publishers. Printed in the Netherlands. _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.1 2 M. Hørmann and M. Kromann In order to reduce the complexity of the optimization problem, a specic functional form of the supply functions is often assumed ex ante. Green and Newberry (1992), Green (1996), Halseth (1998), Baldick, Grant and Kahn (2004), Konkurrencestyrelsen (2004) and Eltra (2003) all assume that rms compete in linear supply curves. The strategies of the rms are thus reduced to deciding either the slope or the intersection of the supply curve. The merits of the assumtion of linear supply curves lie in the simplicity of the analytical framework. However, the assumption of linearity is also associated with some shortcomings, especially in relation to the modelling of markets for electricity. Because of the demand that supply functions be continuously dierentiable, it is impossible for linear supply curves to reect constraints on capacity. Furthermore, the composition of generating companies from many dierent production technologies implies an upward sloping and (not considering the discrete nature of electricity generation MC curves) strictly convex marginal cost curve. This is often assumed to imply that supply functions as well should be upwards sloping and strictly convex. Finally, the linearity of supply curves may understate the use of market power in situations of high demand. This paper departs from the assumption of linearity of supply functions and instead adopts so-called smooth capacity constraints. For the supply functions and marginal costs of generation second order polynomial curves are applied. This will allow for the modelling of capacity constraints and upwards sloping stricly convex supply curves. Other continous functional forms can also be applied. The paper also briey describes how transmission capacity and intertemporal hydro power generation constraints can be implemented using continuous constraints. This paper does not address uncertainty of demand. As Green (1996) states, the variation in demand for electricity over time is much more signicant than any random variation. Since rms submit supply functions for one hour at a time, it is assumed that any random variation within this time frame is limited. In section 2, the model framework is described including derivations of restricted polynomial marginal costs and supply functions and rst order conditions of prot maximization using supply functions. Section 3 describes the implementation of the model framework into a working numerical simulation model. 1 In section 4, results of the numerical simulation are presented and discussed, and in section 5 potential trans- 1 The model is called seems2c (Supply function Equilibrium in Electricity Markets using Smooth Capacity Constraints) and has been implemented using the simulation software package Gams. _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.2 SFE and Continuous Constraints 3 mission and hydro power extensions of the model are presented. Section 6 briey summarises the results of this paper. 2. Model Specication This section describes three models of an electricity market: A Perfect Competition model, which serves as a baseline for comparison of the two other forms of competition assumed, namely Cournot competition and Supply Function Equilibrium (SFE). In this paper we show how convexly increasing Supply Functions and marginal costs with capacity constraints can be modelled within the Supply Function Equilibrium framework. To our knowledge, this has not been demonstrated before, even though exactly this type of cost and bid structure is generally agreed to be very characteristic of electricity markets. We start by presenting our assumptions on demand and cost structure. Then we proceed to a perfect competition and Cournot competition setting before we present a single period version of our SFE model. 2.1. Demand and Cost Structure In the interest of providing an easily interpretable introduction to the dynamics of the model, the basic scenario describes a market where demand is linear. D (P ) = Qmax − sP ⇔ P (Q) = Qmax − Q s (1) Alternatively, the demand is represented by a constant elasticity demand function, for elastic demand with e < 0, and also with inelastic demand. e D (P ) = f P e ⇔ P (Q) = Q f D (P ) = Qmax for P ∈ <+ for constantly elastic demand for inelastic demand Convexly increasing marginal costs with a capacity constraint is often the better empirical approximation to electricity producers' cost structure. We apply a polynomial of second order to represent the inverse function of the marginal cost function, c.f. equation 2. Such a polynomial is described by three parameters which can be chosen freely by the modeller. However, these three parameters can be meaningfully calibrated by three important characteristica of any electricity producer: the marginal cost of the cheapest production facility _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.3 4 M. Hørmann and M. Kromann Figure 1. Marginal costs structure M Cl , the marginal cost of the most expensive production facility M Ch , and the total generating capacity available to the producer, k . This cost structure is illustrated grapically in gure 2.1. The following polynomial specication of the cost structure relates quantities to marginal costs: Q (M C) = −aM C 2 + bM C − c We can now determine the parameters where a, b, c ≥ 0 (2) a, b, c using the characteristics of the marginal costs mentioned above. The inverse of the marginal costs curve determine b Q (M C) M Ch which can be used Q0 (M C) = −2aM C + b: has its maximum at by utilising its derivative Q0 (M Ch ) = 0 ⇔ b = 2aM Ch Similarily, a (3) can be determined through above derivation of the capacity constraint k by utilising that to b and Q (M Ch ) = k : −aM Ch2 + 2aM Ch2 − c = k ⇔ a = k+c M Ch2 (4) _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.4 5 SFE and Continuous Constraints c Finally, can be determined from the marginal cost of the cheapest production unit where Q (M Cl ) = 0, c = −aM Cl2 + 2aM Ch M Cl = c = (k + c) d = MC k+c M Cl (2M Ch − M Cl ) M Ch2 kd 1−d where Rearranging the relation between for giving d= MC 2M Ch M Cl −M Cl2 M Ch2 and Q, (5) a closed expression can be obtatined, although this expression is meaningful only for the intervals Q ∈ [0; k] and M C ∈ [M Cl ; M Ch ]: −aM C 2 + bM C − c = Q −4a −aM C 2 + bM C 2 2 MC over = −4a (c + Q) ⇔ b + 4a M C − 4abM C = b2 − 4a (c + Q) ⇔ (b − 2aM C)2 = b2 − 4a (c + Q) ⇔ p b − b2 − 4a (c + Q) M C (Q) = 2a Integrating 2 Q gives the total cost function T C (Q), which can be shown to be T C (Q) = = = RQ 0 M C (Q) dQ 3 2 1 b 2 − 4a (c + Q) 2 −1 Q − b 2a 3 2a 4a 3 b 1 2 2 2a Q + 12a2 b − 4a (c + Q) (6) With this description of demand and marginal costs, we now turn to the producer behaviour. The prot maximisation problem is max πi = P (Q) Si − T Ci (Si ) where Q= PN i=1 Si . This problem can be stated as the following rst order condition P (Q) + ∂P ∂Q Si − M Ci ≤ 0 ∂Q ∂Si and noting that under linear demand, sections focus the attention on ∂Q/∂Si ∂P/∂Q = −1/s (7) the following which describes the aggregate reaction of the other rms in the market. It will be shown that Perfect, Cournot and Supply Function competition are very dierent in their specication of the reaction of the competitors. _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.5 6 M. Hørmann and M. Kromann 2.2. Perfect Competition In perfect competition, all producers assume that their choice of price or quantity does not aect the market price because the size of each producer is insignicant compared to the total size of the market. This translates into ∂Q/∂Si = 0. Thus, the rst order condition for prot maximisation is P = M Ci (8) This is also known as so-called marginal cost pricing, i.e. the producer increases his supply if the price is higher than his marginal cost, and decrease production if the price is lower than marginal costs. 2.3. Cournot-Nash Equilibrium In Cournot-Nash Equilibrium the with market power are assumed to use quantities as their strategic variable. This means that they commit to supplying a certain quantity before the market opens and the market price is realised. The quantity commitment is made simultaneously among the rms, and therefore each rm rightly believe that their choice of quantity does not aect the other rms' choice of quantities. Thus, their belief is that a change in their own quantity will be fully reected in the total market supply, i.e. ∂Q/∂Si = 1. Thus, the Cournot com- petitors rst order condition (with the linear demand described above) is 1 P (Q) − Si − M Ci ≤ 0 s 2.4. (9) Supply Function Equilibrium The electricity spot markets known today does, however, not resemble the quantity commitments of Cournot-Nash Equilibrium. Rather, the producers sell their electricity in an auction like setting, where they must submit a so-called bid schedule, specifying the quantities they are willing to produce given dierent realisations of the market price. The Supply Function Equilibrium resembles exactly that. The N rms commit to a so-called Supply Function, which species the quantity supplied by the rm at the realised market price. And not only is the rm committed to its own Supply Function, it also knows that its competitors are also committted to their Supply Functions, and thus change their supply in reaction to changing market price. Therefore, when the rm changes its supply (according to its Supply Function as a response to a market price change), it must also expect that the _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.6 SFE and Continuous Constraints 7 competing rms will change their supply in response to the changing market price. For this reason 0 < ∂Q/∂Si < 1 in SFE, depending on the particular form of the Supply Function. Thus the SFE lies in between Perfect and Cournot competition. The following structure of the rms' Supply Functions will accommodate the implied capacity constraints of the marginal cost functions: Si (P ) = −αi P 2 + βi P − γi for αi , βi , γi ≥ 0 (10) Obviously, this structure resembles the structure of the marginal costs. Contrary to the marginal costs formulation, we wish to maintain a degree of freedom for each rm to choose the shape of this supply functiton that maximises its prots. Thus, we will bind only two of the three parameters to physical characteristics of the producers generating equipment. In the previous SFE litterature, it has been observed that the exercise of market power especially takes place in situations with very high capacity utilisation, whereas the market seems rather competitive when the capacity utilisation is low. It therefore makes sense to bind the supply functions to the marginal cost of the cheapest technology and to the total capacity constraint. At the capacity constraint the slope of the supply function is vertical. P̄ , accordβi . Inserting = 0 ⇔ P̄i = 2α i bind α to β and γ through the The price chosen by the rm running at maximum capacity, ing to its supply function is thus Si0 P̄ this in the supply function allows us to capacity constraint: Si β 2α = ki ⇔ −βi2 β2 + i − γi = ki ⇔ βi2 = 4αi (ki + γi ) 4α 2αi (11) The second binding of the supply function concerns the lowest price at which the rm is willing to supply a positive quantity. As mentioned above, the literature indicates that competition is most intense when capacity utilisation is low. We thus assume that the rm will start to supply the market when the market price is equal to the cost of its cheapest production unit, M Cl : 2 Si (M Cl,i ) = 0 ⇔ γi = −αi M Cl,i + βi M Cl,i (12) The following set of equations thus dene the restricted second order polynomial supply function: Si (P ) = −αi P 2 + βi P − γi _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.7 8 M. Hørmann and M. Kromann βi2 4 (ki + γi ) 2 γi = −αi M Cl,i + βi M Cl,i αi = Green (1996) derives rst order conditions of prot maximzation for N asymmetric rms competing in supply functions. The resulting set of equations does not lend itself to an analytical solution. However, as we intend to solve the problem numerically, it is sucient to state the rst order conditions along with equilibrium conditions. Under the assumption of competition in supply functions, the strategic variable of the rms is a combination of price and quantity. The prots of rm i may be written as: πi (p) = pSi (p) + Ci (Si (p)) The supply of rm P j6=i Sj (p). i is equal to the residual demand Si (p) = D(p) − Inserting this, the prot of rm i can be rewritten. πi (p) = p D(p) − j6=i Sj (p) + Ci D(p) − P P j6=i Sj (p) The prot is dierentiated with respect to price. X X ∂Sj (p) ∂πi = D(p) − Sj (p) + p D0 (p) − ∂p ∂p j6=i j6=i −Ci0 D(p) − X Sj (p) D0 (p) − j6=i X ∂Sj (p) j6=i ∂p Setting equal to zero and rearranging yields the rst order condition of prot maximization of rm i. Si (p) = p − Ci0 (Si (p)) −D0 (p) + X ∂Sj (p) j6=i ∂p 3. A Numerical Simulation Model The model framework described above has been simulated in GAMS. This section describes how the framework is translated into a set of equations that will yield equilibrium solutions in a mixed complimentarity solver. In order to analyse the eects of uctuating demand on _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.8 9 SFE and Continuous Constraints equilibrium and to accomodate future expansions of the model in regards to hydro power and inter temporal prot maximization, an index for time has been incorporated in the model. Otherwise, the system of equations described here should be recognizeable from the previous section. 3.1. Common Equations The demand, marginal cost and market equilibrium equations are common to all three models. Demand: Dt (Pt ) = Qmax,t + sPt f Pe t Qmax,t , Linear , Constant Elasticity , Inelastic Marginal Demand: s ∂Dt (Pt ) eft P e−1 = ∂Pt 0 , Linear , Constant Elasticity , Inelastic Marginal Cost: 2 qi,t (M Ci,t ) = −ai M Ci,t + bi M Ci,t − ci If a is set to zero, then marginal costs become linear. In that case, 1 b and the intersection or c the marginal cost of the cheapest unit produced will be . b the slope of the marginal cost curve will be Market Equilibrium: Qt = Dt (Pt ) 3.2. Perfect Competition The only equation which is specic to perfect competition is the rst order condition of prot maximization. F.O.C. M Ci,t ≥ Pt ⊥ qi,t ≥ 0 _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.9 10 M. Hørmann and M. Kromann 3.3. Cournot Competition The only equation which is specic to Cournot competition is the rst order condition of prot maximization. F.O.C. M Ci,t ≥ Pt + 3.4. qi,t ⊥ qi,t ≥ 0 M Dt Suplly Function Competition The supply function problem is more complicated. First, the core of the problem consists of two sets of complimentarities. Like the perfect and Cournot competition cases, the decisisons of the rms are dictated by their rst order conditions of prot maximization. Since the supply function itself is not the strategic variable, it is necessary to further specify the relationship between the strategic variable βi,t and the supply function. Si,t (Pt ) ≥ (Pt + M Ci,t ) −M Dt + −αi,t Pt2 + βi,t Pt − γi,t ≥ Si,t (Pt ) P j6=i ∂Sj,t (Pt ) ∂Pt ⊥ Si,t (Pt ) ≥ 0 ⊥ βi,t ≥ 0 Specied along with these two sets of complimentary equations are two supporting equations. Bind Gamma Bind Alpha γi,t = −αi,t M Cl2 + βi,t M Cl αi,t = β 2 /4(ki + γi,t ) 4. Model Properties In this section the various properties of the model are illustrated using an simulations of articial market situations not based on actual data. Instead, the artical markets are constructed to illustrate the model properties in a clear way. We provide simulations of a 24 hour period where the demand uctuates, 2 and where a small number (below 5) of companies compete against each other. 2 A sinus function has been used to emulate uctuations in a linear inverse demand curve. The demand is above the base level in period 1 to 12 and below in period 13 to 24. The available total generation capacity is assumed constant over the simulated period. The MC curves each rms generation capacities are dened by their capacity, and their minimum and maximum MC. _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.10 SFE and Continuous Constraints 4.1. 11 Linear or polynomial cost and supply functions The novel feauture of the presented model is the ability to describe the market using convexly increasing supply and MC curves instead of linearly increasing ones. This has implications for the prices and markups in the simulations as well as other variables. For the interpretation of the results we rst note that the two linear MC simulations have higher prices than the simulations with polynomial MC curves. This is not at all surprising as the polynomial curves allways lie below the linear counterparts (they have the same MC at zero and maximum capacity utilisation). Figure 4.1 shows the simulated markups when using the four possible combination of linear and polynomial MC and Supply Functions (SF). It can be seen that the markups are relatively higher in periods with high demand compared to periods with low demand. Thus our model conforms to previous SFE ndings on this point. It can also be seen that the polynomial SF show distinctly higher markups, whereas linear SF gives much lower markups. The reason is that convex Supply Functions have steeper slopes (reducing competitive pressure) at lower price levels (tending not to curb consumer demand as much as is the case with linear SF's). Linear MC shows lower markups than linear ones in high demand situations, whereas the opposite is the case in low demand situations. This might partially be attributed to their higher costs associated with linear MC curves, leaving less space for higher markups. The dierences are, however, not very large. When demand is low, the simulations show that there is much less dierence between linear and polynomial marginal cost and supply functions. But contrary to the high demand situations, it is now linear MC curves that gives the highest markups. This is because the linear SF's are relatively steeper than their polynomial counterparts when demand is low. The choice of cost function seems to be of less importance. We can thus conclude that the assumptions made on the functional form of the supply curves are rather important to the markups. The interpretation of a steeper supply curve is that the rms are less sensitive to price changes. Using polynomial instead of linear supply curves thus implicitly assumes a more intense competition with low demand and less intense with high demand. The functional form of the MC curves are of less importance, and no very general impression is found in the presented simulations. _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.11 12 M. Hørmann and M. Kromann Figure 2. Markups with combinations of linear and polynomial MC and Supply Functions 4.2. Equilibrium Concept In the model three equilibrium concepts are implemented: Perfectly Competitive, Cournot, and Supply Function Equilibrium. It has also been made possible to specify a representitive competitive fringe supplier in the simulations, i.e. a supplier which always increases its supplies whenever the price is above his marginal costs. In gure 4.2 below, the markups from simulations with dierent equilibrium concepts are shown. As can be seen from the gure, the Cournot equilibrium with three rms and an additional fringe supplier tends to yield markups that are orders of magnitude higher than SFE markups. Thus, modellers choosing the Cournot Equilibrium concept for their modelling assume the least competivite equilibrium concept. 4.3. Consumer Demand Modelling The consumer demand has been implemented using two dierent demand functions: linear demand and constant elasticity demand. Con- _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.12 SFE and Continuous Constraints 13 Figure 3. Markups under low demand with combinations of linear and polynomial MC and Supply Functions 3 trary to the Cournot Equilibrium, SFE allow for inelastic demand. The simulations undertaken show that inelastic demand can lead to extremely high markups, c.f. gure 4.3. As can be seen from the gure, inelastic demand tends to give especially high markups when demand is very high compared to the available capacity. In period 6 the capacity utilisation is at 90 per cent, compared to around 25 per cent in period 18. It should be noted that the assumed elasticity of the linear demand is valid only for the base point of the linear demand. The realised demand given the market price may have dierent elasticity. 5. Perspectives The presented model is non-linear and can be implemented both as a Mixed Complementarity Problem (MCP) and as a Constrained Nonlinear System (CNS). For both implementations, the calculation time is 3 With the linear demand function D(P ) = Qmax − sP inelastic demand is represented by s = 0, which invalidates the Cournot prot maximsing rst order condition, c.f. equation 9. _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.13 14 M. Hørmann and M. Kromann Figure 4. Markups with diering equilibrium concepts very satisfying, as it is counted in milliseconds for the shown simulation examples, as well as for 168 hour simulations. Larger problems with more rms and time periods are solved in seconds or minutes. The reason for such small calculation time is that the continous SFE model explicitly describes the reaction of the dierent competitors. The small calculation time has made possible yet unnished experiments with the model formulation. These show that it apparently is possible to include two very important aspects of electricity markets in a continous polynomial Supply Function model: multiple hydro producers excercising market power and transmission constraints. 5.1. Market Power of Multiple Hydro Producers On the longer term the price setting hydro power producer has an intertemporal problem of deciding when the production should be high and low according to the prevailing electricity price in each period. This operationalises to the following limitation for each hydro producer: X StH ≤ H̄ t _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.14 SFE and Continuous Constraints 15 Figure 5. Markups with various demand modelling where H̄ is the maximum production capacity of the hydro producer over the modelled period. 4 This limitation gives rise to an endogenous shadow price on the water (the water value), λ. Alternatively, and perhaps most relevant to the modelling of relatively short time spans, the water value could be specied as exogenous (meaning that the use of water in the modelled period is quite small compared to the available amount of water). In this latter case, the hydro producer's problem is equivalent to that of the other producers'. The model eorts on the endogenous water value is still experimental. By now it is clear that the endogenous water value should not be used to x the interception of the polynomial supply curve. As the polynomial supply curve cannot be downwards sloping and it apparently is optimal to use all the water, the optimal hydro supply curves becomes very at, leading to problems of linear dependency with the marginal cost curve (dened as M C = λ). Two formulations of this problem is in the working. The rst is to assume that the long term value of water is zero and start the polynomial supply curve in (0, 0). Another solution is keep the endogenous 4 Minimum production and other hydro specic production requirements are not yet considered. _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.15 16 M. Hørmann and M. Kromann water value as the intercept but allow a linear supply function to be both upwards and downwards sloping. The latter solution still needs to be evaluated against key assumptions of the SFE framework. 5.2. Smooth Transmission Capacity Constraints A requirement of a SFE model is that the MC and Supply Functions and their rst order derivaties are continous. For this reason it is not immediately possible to include hard transmission constraints with associated shadow prices. This problem can, however, be circumvented by adding a continous transmission cost function to the MC curves of the rms. This cost function must be very low when transmission capacity is available, and prohibitively high when transmission capacity is fully utilised. A suggestion for such a function is XM C = where x¯r,s 1 x¯r,s − Xr,s Xr,s is the actual r and region s. Results for this implementa- is the transmission capacity constraint and transmission between region tion is yet too unnished to be presented here, but the calculation time are in the same scale as the simulations presented hitherto, and preliminary results of transmission cost simulations seem promising. Still work remain to adjust the capacity constraint of the supply function to also consider the transmission constraint. 6. Conclusions This working paper has demonstrated how smooth capacity constraints can be implemented in a Supply Function Equilibrium to emulate generation and transmission capacity constraints relevant for modelling of the electricity market. Furthermore, the presented model has been implemented as a numerical simulation problem. The simulation model shows to be very computationally eective. This has allowed extending the investigations into the intertemporal problem of hydro electricity producers' market power, albeit these investigations are only at a very experimental, yet promising level. The simulation model was used for comparing various properties of the Cournot and Supply Function Equilibrium concepts, as well as investigating details about the modelling of consumer demand. Of special importance, it was found that polynomial Supply Functions _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.16 SFE and Continuous Constraints 17 tend to increase markups compared to linear Supply Functions. This is especially prevalent when the capacity utilisation is high. It was also demonstrated that Cournot equilibrium tends to give much higher costs than the Supply Function Equilibrium. Acknowledgements This working paper and the construction and implementation of the model has been partly funded by the Danish Energy Research Programme. The authors wishes to acknowledge valuable comments from ... References Baldick, R., Grant, R. and Kahn, E. (2004). Theory and Application of Linear Supply Function Equilibrium in Electricity Markets. Journal of Regulatory Economics 25(2), 143-167. Eltra. (2003). Dokumentation af MARS modellen. Eltra. Green, R. J. and Newberry, D. M. (1992). Competition in the British Electricity Spot Market. The Journal of Political Economy 100(5), 929-953. Green, R. J. (1992). Increasing Competition in the British Electricity Spot Market. The Journal of Industrial Economics XLIV(2), 205-217. Grossman, S. (1981). Nash Equilibrium and the Industrial Organization of Markets with Large Fixed Costs. Econometrica 49, 1149-1172. Halseth, W. (1998). Market Power in the Nordic Electricity Market. Utility Policy 7, 259-268. Klemperer, P. D. and Meyer, M. A. (1989). Supply Function Equilibria in Oligopoly Under Uncertainty. Econometrica 57(6), 1243-1277. Konkurrencestyrelsen (2004). Fusionen mellem Elsam og NESA. Konkurrencestyrelsen, Denmark. Address for Oprints: <Address for oprints, printed at the end of the article> _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.17 _ProgramFiles_docs_seem2c.tex; 17/05/2005; 17:40; p.18
© Copyright 2026 Paperzz