Sammendrag
Jørund Ystrøm Haartveit
Power prices in Germany. Price formation and fundamental factors. Theoretical and empirical evidence
As a result of the deregulation of the different electricity markets around the world, electricity price forecasting has in the last years received increased attention. The UK and Scandinavia were the first ones in Europe to deregulate its electricity markets; the rest of Europe is now following by implementing EU s (European Union) decision to deregulate its electricity market. As Europe s largest electricity market Germany plays a crucial role in the deregulation process.
One central element in the deregulation has been the creation of wholesale markets or pools where markets participants bid to sell or buy electricity. Because electricity is a commodity that must be consumed at the time of generation, the pools have a day-ahead structure where electricity for the following day is traded.
A consequence of the deregulation is a more volatile electricity price and a higher price uncertainty for the different market participants. The importance of proper risk management provides a need for accurate price forecasts and to understand the electricity price variation.
Producers and consumers rely on forecasted information when they make up their corresponding bidding strategies and accurate price forecast can contain crucial information for both groups when they decide how to bid in to the pool to maximize their own benefit. On the medium time horizon suppliers must make decisions on how much to bid into the pool and how much they will sell through bilateral contracts. Consumers must make similar decisions how to cover their electricity need.
Standard financial models of asset prices have been applied to describe and forecast electricity prices, but because electricity is an instantaneous commodity and must be consumed at the time of generation, the arbitrage argument that these models emphasise do not hold, and they fail to capture the nature of electricity prices .
The aims of this thesis is to establish an econometric model that incorporates important supply and demand variables to describe the price generating process and then discuss in what degree it is able to explain and forecast short term (day-ahead) electricity prices at the EEX . The thesis is structured as follows: In Chapter 2 a descriptive analysis and an overview of the German power market are given. A description of the different market places where electricity is traded in Germany and especially the EEX is given here. In the theoretical assessment of the electricity market in Chapter 3 I use a simplified version of a model developed by von der Fehr and Harbord.(1993) This model emphasises that the level of demand relative to the given capacities/ available production is important to explain how high prices on the electricity spot market can occur. The main conclusion here is that in the low and the high demand period, given the generators capacities, the price is determined by different factors, in the low demand case the price is cost driven, but in high periods the price is influenced by generators strategic bidding. In Chapter 4 I discussed the fundamental factors that influence and affect the overall supply and demand in the price determination process. I point out that demand in the short run is driven by changes in weather, time of day and day of the week and that there consist considerable uncertainty of electricity supply in the short run. Both these factors are reflected in the erratic and volatile price behaviour. Chapter 5 contains the empirical investigation in this thesis. Here I discussed the data and checked the time series properties of the variables. A more detailed investigation of the relationship between electricity price and the demand and supply variables in the data set was also given here. On the basis on the previous discussions I created an econometric model and I examined its explanatory power in the sample period. I found that is was hard to isolate the individual contributions of the exogenous explanatory variables. Several restrictions were imposed in order to see if a restricted model could be accepted. However, when tested, the restrictions imposed were not consistent with the data. Despite the difficulties in assessing the individual exogenous explanatory variables contribution to explain the price behaviour, they did have explanatory power. The forecasting ability of the model was assessed and carried out in the last section of Chapter 5. Their predictive accuracy was compared to a simple Autoregressive (AR) model. Even if the created models outperformed the simple AR model, their forecast ability was relatively poor. In the two remaining Chapters, limitations of the thesis are discussed and conclusion is given.
The data used in the estimations of the models was provided by Natsource-Tullett Scandinavia (NTS). The data was prepared and organized in a Microsoft Excel spreadsheet, and the calculation of the average hourly prices at the different days of the week was carried out using Excel. The generation of binary variables, the analysis of the data and all estimations of the models were conducted by using the PcGive software package.