Stationary or Instationary Spreads – Impacts on Optimal Investments in Electricity Generation Portfolios

It is common practice to base investment decisions on price projections which are gained from simulations using price processes. Particularly in the electricity industry with its diverse fuels, homogenous output and long-lived assets the choice of the underlying process is crucial for the simulation outcome. At the most fundamental level stands the question of the existence of stable long-term cointe-gration relations. Since this question is very difficult to answer empirically, it is also appropriate to investigate the implications of varying assumptions. Not only the specific parameter values but also the specification of the price processes has therefore to be scrutinized. In the presence of fuel price risk, portfolio diversification will usually be drawn into consideration in investment decisions. Therefore we examine the impacts of different ways to model price movements in a portfolio selection model for the German electricity market. Three different approaches of modelling fuel prices are compared: Initially, all prices are modelled as correlated random walks. Thereafter the coal price is modelled as random walk and leading price while the other prices follow through mean-reversion processes. Last, all prices are modelled as mean reversion processes with correlated residuals. The prices of electricity base and peak futures are simulated using historical correlations with gas and coal prices. Yearly base and peak prices are transformed into an estimated price duration curve followed by the steps power plant dispatch, operational margin and NPV calculation and finally the portfolio selection. The analysis shows that the chosen price process