A Powerful Competition Policy - 5. Market modelling

5. Market modelling

5.1 Why modelling

Due to various analytical challenges in competition analysis as such and competition analysis in the electricity sector in particular, modelling of markets and firm behaviour can be a useful tool. A market simulation model can be utilised to study the effects of for instance a merger on the Nordic market.

The purpose of the present chapter is to illustrate how competition authorities may use market models in the analysis of the electricity market as well as other markets. The Eltra market model is used to study exertion of market power. Further work needs to be done in order to establish the full welfare effects.

The Nordic market for wholesale electricity is inherently prone to market manipulation. One special characteristic of the electricity market is that even firms with relatively small market shares may exert market power unilaterally in certain periods. This is due to the mix of the non-storability of electricity and capacity constraints in the production and transmission of electricity, cf. chapter 4.

The unilateral or multilateral exertion of market power may have widespread effects in the Nordic market. Higher prices due to the exercise of market power in one area may cause higher prices in other areas, and expensive generation may be substituted for cheaper resulting in a real economic loss and not just a reallocation of wealth. Inter-Nordic mergers can influence the overall flow and pricing of electricity.

Note however, that despite of the many valuable insights that can be gained from the use of simulation models in competition policy the models are still only one of a number of tools that can be employed in order to reach decisions in competition cases. Furthermore, the competition authorities need to gain more experience from the application of models in competition policy analysis.

What is studied in the model is the "exertion of market power" not to be confused with the legal term "abuse of market power". By definition a firms has market power if it profitably can raise its prices above marginal costs.

This chapter presents results stemming from two model simulations done by Eltra [27] on request of the Nordic competition authorities. Eltra has developed a model of the wholesale power market in the Nordic area including the northern part of Germany. This chapter starts with a short description of the model and the theory behind it. Afterwards two model simulations are analysed. The first model simulation studies the incentives of Nordic generators to exercise market power and the effects on the Nordic market. The second simulation studies the effects on the Nordic market of an inter-Nordic merger. In the simulation a Norwegian and a Finnish generator is merged.

5.2 The Eltra Model

The model used for analysing market power in this chapter is Eltra's market simulation model, MARS. The development of the model started in 2000 and is now (May 2003) in its final phase. This section is based on Eltra (2003a).

The model is a supply function equilibrium model. The theory of supply function equilibrium mixes Cournot and Bertrand competition. Klemperer and Meyer (1989) argued that under uncertainty firms would adopt supply functions as strategic variable instead of quantities (capacities) or prices. In this way firms form price quantity pairs stating exactly what quantity can be supplied at a given price or vice versa. This is what happens in the Nord Pool market.

A supply function equilibrium model has Bertrand competition (perfect competition) as one extreme and Cournot competition as the other. Under perfect competition a firm will bid in production at it's true marginal costs. If they did not they would loose sales. If the firm, however, possesses market power it would not maximise profit by bidding in true marginal costs. Due to the market power it can profitably raise the price and/or withhold capacity, cf. chapter 4. In the model, firms can choose to add a mark-up to their marginal costs. A more detailed description is made below. Reference is also made to the description of the model and the simulations done be Eltra enclosed at the end of this report.

The purpose of the model is to analyse the effects on prices, production, demand and exchanges in the wholesale market for electricity of market power. All outputs are calculated on an hourly basis. The model uses the principles of the Nord Pool market mechanism including the division of the Nordic countries into price areas with price-dependent bids. Particular focus has been given to the use of game theory in analysing the producer's behaviour, i.e. the incentives to exercise market power. It should be noted that the model ignores the existence of bilateral contracts. The model assumes that all electricity on the wholesale level is traded at Nord Pool prices.

As a supplement to the analyses presented in the previous sections the model can provide valuable insights into the consequences of changes in transmission or generation capacities, market design, demand or further tightening of the capacity balance.

Model input is based on data reported to Nordel by transmission system operators. The data is converted into hourly basis using the available information of for instance the distribution of consumption. Data of cost structure of production plants are one of the model inputs.

The model simulates the Nordic area containing 6 different potential price areas and 9 interconnectors between price areas. Price areas and interconnectors are shown in figure 5.1. [28] All interconnectors are modelled as market-controlled (as in Nord Pool). However, point-of-access tariffs are included in the allocation of capacity of the 3 interconnectors to Germany. This is a short cut in order to model the fact that the German interconnectors are not fully integrated in the Nordic market.

Figure 5.1 The model area in MARS (as of April 2003)

Figure 5.1 The model area in MARS (as of April 2003)

Source: Eltra (2003)

Modelling of demand
The demand is modelled as price dependent with constant elasticity. The inverse demand function is a Cobb-Douglas function

p=kq^1/β

where p is price, q is quantity, k is a calibration constant and β is the elasticity of demand.

Supply

A supply function is made every hour for each production plant in the model. Some supply is price dependent and some is not. The short-term variable cost of the individual plants forms the basis of the estimation of the supply functions.

Hydropower production from plants with reservoirs that can be regulated is also modelled as price-dependent production. In estimation of the supply functions water values from the Nordic EMPS model are used. The water values in the EMPS model are calculated from the date used in MARS.

The water values express the marginal value of the water in the reservoirs and are used in MARS to determine market equilibrium in the same manner as the marginal costs of thermal generation. This means that the supply function of a hydropower producer in a given hour is affected by the volume of water left in the reservoir, the time of year and the inflow into the reservoir.

Optimisation and equilibrium

The market price calculation in the model is done on an hourly basis as a maximisation of the socio-economic surplus in the model areas consisting of consumer and producer surplus and congestion rents. The optimisation is restricted by the capacity constraints in generation and transmission. After the calculation of production the content of water reservoir is updated using the calculated production and inflow data before calculating the equilibrium price in the next hour.

In the Nordic market model, an exporting producer obtains the price applicable in the producer's price area. Similarly, the importer (the consumers in the import area) pays the price applicable in the price area where he is situated. In cases where transmission capacity limits the exchange between price areas, the price in the import area will be higher than the price in the export area. This creates a positive difference (congestion rent) between the payment from the consumers in the import area and the payment to the exporting generator. This congestion rent accrues to the system operators involved and is a part of the socio-economic surplus. It equals transmission capacity multiplied by the price difference.

Simulation of market power
It is assumed in the model that all generation maximises company profit. The Nordic power market is an oligopolistic market. Players in an oligopoly expect other players to react to different choices of strategic variables such as price and quantity. This means that a player's profit depends not only on the player's own activities but also on the activities of other players in the market and visa versa. The objective of the model simulations is to find Nash equilibria in the market.

In the model the opportunity to exercise market power is given specific players in the market by letting them choose to add a mark-up to the marginal costs. This may affect the resultant supply function. The (inverse) resultant supply function with market power is modelled as p=µq+mc, where p is the price, q is the supply at p, µ is a mark-up coefficient and mc is the marginal cost at q. The mark-up is then equal to µq and µ=0 corresponds to price taking behaviour. The model calculates Nash-equilibria in µ. This means that no player has the incentive to change bidding strategy (her choice of µ) given the strategies (µ) of all the other players.

The search for the equilibrium strategies is time-consuming, as it is necessary to calculate a sufficient number of price equilibria in each hour in order to determine the optimal strategy. Furthermore, the more generators that are able to exercise market power (µ>0), the more steps the procedure will comprise, which makes it even more difficult to achieve convergence. Thus, in order to facilitate calculations in the model the players can choose strategy from a final discrete strategy space. The discrete strategy space means that the individual generator has a predefined final number of mark-up coefficients to choose from when maximising profit.

The fact that a set of possible strategies is predefined (and exogenous) for each individual generator makes it possible to find a solution. However, it introduces requirements to the selection of the strategy space. If the intervals between mark-up coefficients are large in order to test different values there is a risk of missing Nash-equlibria. However, if the intervals are too small in order to make a more precise estimation of µ Nash-equilibria outside the set might be ignored.

5.3 Nordic snapshots of the exertion of market power

This section presents the results of a simulation of market outcomes in a high demand winter week (week 3) in 2005. The effects of introducing imperfect competition are analysed. That is, the benchmark scenario is one of perfect competition where all generators bid in price quantity pairs in accordance with true marginal costs.

The figures shown are price curves and flow maps. Price curves show the development in prices during a week and flow maps show a snapshot of the Nordic system in one particular hour. First, price curves are presented for selected Nordic areas showing the simulated price development of week 3 in 2005. From these curves the exertion of market power can be spotted. Second, three individual hours are selected for further analysis by means of flow maps. The flow maps also indicate welfare effects.

Week 3 is characterised by a relatively high demand, especially in Norway. In this situation the two primary thermal systems – Denmark and Finland – are net exporters.

Simulations of a summer week has been done but is not presented here. However, some of the results are alike. Generally, the incentives to exercise market power in peak load hours are intact. During low load hours incentives differ due to the lower water value, which makes import into the thermal areas possible. Reference is made to Eltra (2003b).

In the simulation presented in this section and section 5.4, seven producers are price setters while the rest are price takers. The distribution of price setters is: One in DK1, one in DK2, one in Norway, two in Sweden, one in Finland and one in Germany North.

Figures 5.2, 5.3 and 5.4 show the price pattern in the simulated winter week in DK1, DK2 and Finland. The simulated Norwegian and Swedish price patterns are not shown in this section since they follow the Danish price patterns except during nights and weekend (the price patterns are shown in section 5.4 below). [29] The figures show the price in all hours of the week (1 to 168) starting Monday hour 1, that is the hour between 24 and 1 the night between Sunday and Monday. Two price curves are shown: "PC" for Perfect Competition and "MP, no merger" for Market Power without merger, that is with seven independent price setters. (In section 5.4 two of the price setters are merged).

Figure 5.2 Prices in DK1 – week 3, 2005 – introducing market power

Figure 5.3 Prices in DK2 – week 3, 2005 – introducing market power

Figure 5.4 Prices in Finland – week 3, 2005 – introducing market power

Figures 5.2 to 5.4 show at least two different examples of generators exercising market power. The first episode reflecting the exertion of market power appears during early morning Monday (day 1) in Denmark and to a certain degree also in Finland. Similar patterns occur during the morning of all work days.

In week 3 in a normal year with the expected power balance in the Nordic countries the marginal cost of thermal generation is below the marginal cost of hydro generation (the water value). Therefore the power will flow in direction of the large hydro areas. This leads to surplus generation in Finland – where thermal capacity accounts for approx. half of total capacity – and in the two Danish areas. To equalise supply and demand under free competition the Finnish and Danish area prices fall below the Swedish and Norwegian prices.

Figures 5.5 and 5.6 show the flow maps corresponding to the hour 3 Tuesday in week 3, 2005. That is the hour between 2 and 3 in the morning. The flow maps show prices, generation and consumption in all price areas and exchanges between price areas. It can be seen that when no generator is exercising market power electricity fills all transmission lines from Denmark and Finland to Sweden and Norway and from Sweden to Norway. Thus, Sweden is used for transit to transport Danish and Finnish power to Norway. Norway and Sweden are high price areas. The price in Sweden is lower than the price in Norway due to the lower cost of nuclear power generation compared to the water values assumed in this simulation.

In the perfect competition scenario, the generator in DK2 (Denmark East) bid in capacity at its true marginal costs. This result in an area price of DKK 150 per MWh. Granted the possibility to add a mark-up the generator would profit from raising its price to just below the Swedish price still exporting 1.700 MW to Sweden.

A Finnish generator adapts differently. To avoid congestions in the transmission line to Sweden – and hence raise the Finnish price to the Swedish level – the price setting Finnish generator reduces production by approx. 500 MWh and thereby obtains a DKK 10 price increase on all infra marginal production.

Figure 5.5 Flow map – perfect competition

Figure 5.6 Flow map – introducing market power

The manipulative behaviour illustrated in the figures results in higher prices in Denmark and Finland. Furthermore, the behaviour leads to an efficiency loss since more expensive generation in Sweden is substituted for cheaper generation in Finland. Whereas the increase in prices alters the allocation of wealth, the increase in Swedish generation reflects a real economic loss: Relatively inefficient production plants produce electricity when more efficient plants still have spare capacity. Calculating the immediate reallocation of wealth if this particular form of market manipulation happens five nights a week for five hours in three winter months the Danish and Finnish consumers loose DKK 56 Mio. each year.

Note that the price increase in Western Denmark does not reflect the exertion of market power by the local generator. The loss of import from Germany reduces supply and causes an increase in local production and, hence, in the marginal costs. This can be seen from studying mark-ups (not shown).

A second episode of the exertion of market power appears during peak load hours on Tuesday evening shown in figures 5.7 and 5.8 below. In this hour all price-setting generators manipulate the price upward resulting in a decrease in consumption (a so-called dead weight loss).

Figure 5.7 Flow map – perfect competition

Figure 5.8 Flow map – introducing market power

In the perfect competition scenario shown in figure 5.7 a Swedish generator sets the prices in the region with exception of the Finnish price. The capacity of the transmission line between Sweden and Finland is too small to equal supply and demand resulting in a (slightly) lower Finnish price. Figure 5.8 shows the model simulation where the price-setting firms are allowed to add a mark-up to their price. First of all, the general price level in the Nordic countries is increased by approx. DKK 100 per MWh due to market power lowering the overall consumption. Second, the Finnish generator reduces generation in order to avoid filling the cable to Sweden and in turn keep the Swedish price level.

One immediate consequence of this upward manipulation of the price level is that the Nordic consumers in this particular hour pay approx. DKK 6 Mio. more for what is consumed, and they consume 1.945 MWh less electricity. If this kind of market manipulation happens twice a week every second week the total consumer loss would amount to DKK 330 Mio. a year.

Figures 5.9 and 5.10 show a snapshot of the Nordic system in a weekend hour, namely the hour between 3 and 4 on Sunday morning. Figure 5.9 shows the well-known picture of exports out of the thermal areas Finland and Denmark resulting in lower prices. Introducing market power reveals that the incentives of the thermal generators are very alike. Both the Finnish and the Danish generator (DK2) reduce production to avoid congesting the cable to Sweden. It follows directly from the flow maps that this behaviour has rather dramatic welfare consequences. Besides the price increase in Finland and in Denmark East more efficient production units are substituted for less efficient ones. The generation in Sweden is increased by approx. 700 MWh while generation in both Denmark East and Finland is reduced.

Figure 5.9 Flow map – perfect competition

Figure 5.10 Flow map – introducing market power

Calculating the immediate consumer loss reveals a loss of DKK 55 Mio. in DK2 and Finland. Consumption and price in the other areas are unaltered by the market manipulation in this particular hour.

5.4 Effects of an inter-Nordic merger

In a market as integrated as the Nordic wholesale market for electricity the effects of a merger among Nordic generators are difficult to establish. In this section a model simulation attempts to pinpoint effects of the hypothetical merger of a large Norwegian and Finnish generator. The simulation is done for a high and a low demand week as in the previous section.

The simulation of area prices is shown below. Figure 5.11, 5.12 and 5.15 are similar to figure 5.2, 5.3 and 5.4 except from the inclusion of the simulation of the price after the above merger.

From the price patterns shown in figures 5.11 to 5.15 one important insight emerge: A Nordic merger among generators can have effects on the entire Nordic system. Due to the merger of a Finnish and a Norwegian generator the price level in peak load hours is increased in all Nordic areas. In low demand hours there seems to be no effect on the market. This fits the theory of flexible production, cf. Chapter 4. In peak load hours competing generators are less able to respond to price increases due to capacity constraints.

Figure 5.11 Prices in DK1 – week 3, 2005 – Nordic merger

Figure 5.12 Prices in DK2 – week 3, 2005 – Nordic merger

Figure 5.13 Prices in Norway – week 3, 2005 – Nordic merger

Figure 5.14 Prices in Sweden – week 3, 2005 – Nordic merger

Figure 5.15 Prices in Finland – week, 3, 2005 – Nordic merger

Since the effect of the merger is in peak load hours only the flow maps corresponding to day 2 hour 18 is shown, figure 5.17. This figure 5.17 can be compared to figure 5.7 and 5.8 in section 5.2. These figures show the perfect competition and market power, no merger simulations for that hour.

Comparing figure 5.17 below with 5.7 and 5.8 it is obvious that the merger results in a reallocation of wealth from consumers to producers. The merger leads to a price increase in this particular hour in all areas of approx. DKK 150 per MWh compared to the perfect competition scenario and DKK 50 per MWh compared to the market power scenario. In hours with a low demand the merger has little effect on the market.

The immediate transfer of wealth from consumers to producers in the Nordic area in this particular hour amounts to approx. DKK 10 Mio. and DKK 4 Mio. if compared to the perfect competition and market power scenario respectively. If this price manipulation due to the merger happens five times a week for seven hours every forth week the yearly transfer of wealth amounts to approx. DKK 4,5 bio. and DKK 1,6 bio. respectively.

Figure 5.17 Flow map – Nordic merger

The merger also affects the distribution of generation. As can be expected the merged firm reduces output – production is reduced from 22.498 MWh in Norway and 12.932 MWh in Finland to 20.782 and 11.238 respectively – in order to increase prices. Production in the other areas remains unaltered compared to the perfect competition scenario. Compared to the market power scenario production is decreased in Norway and Finland and increased in Sweden.

It should be noted that welfare effects indicated by the model are only static welfare effects. The model does not capture dynamic welfare effects.

The merger alters the profits of all price-setting firms – especially the ones not participating in the merger. The changes in profits of the price-setting firms in this simulation are shown in figure 5.18 below. Three different changes in profits are shown for each generator. "MP, no merger – PC" shows the relative increase in profits from introducing market power in the model compared to the perfect competition scenario. "MP, merger - PC" shows the relative increase in profits from merging two generators with market power compared to the perfect competition scenario. "MP, Merger – MP, no merger" shows the relative increase in profits from merging two generators with market power compared to the scenario where two generators with market power are not merged.

For instance the generator in DK2 sees a 25% increase in profits from exercising market power relative to not exercising market power. If the Norwegian and Finnish generator are merged the generator in DK2 sees an increase in profits of almost 70% compared to the scenario where no generators are merged and no generators have market power. Compared to the scenario where generators are exercising market power the generator in DK2 sees a 35% increase in profits due to the merger if it is given the opportunity to exercise market power.

It is shown that all firms gain from exercising market power. Without the merger the two Danish generators gain the most relatively. The simulation of the merger shows that all the non-merging firms gains relatively more than do the merging firms.

The fact that the non-participating firms receive the largest increases in profits due to the merger is not surprising. As can be seen form comparing figure 5.7 (or 5.8) and 5.17 the merging firms reduce output in order to increase prices. The other firms profit from the higher price and do not have to reduce output.

Figure 5.18 Change in profits – week no. 3, 2005

Elasticity of demand
To describe the flexibility of demand an elasticity of demand is used in this model. In the two model simulations presented above a wholesale elasticity of –0.1 is used. This is a quite large (numerically) elasticity. A lower elasticity (numerically) would increase the price setting firms' incentives to exercise market power because the decrease in consumption due to a price increase would be smaller. This would increase the price obtained in the two market power scenarios and increase the transfer of wealth from consumers to producers.

Cross-ownership
The model treats all firms as independent. This is a problematic assumption. Cross-ownership can in a number of ways inflict production decisions and alter the incentives to exert market power, cf. Chapter 3. Introducing cross-ownership into the model would most likely sharpen the results presented in this chapter.