Relationship between spot and futures prices in electricity markets: Pitfalls of regression analysis

This work discusses potential pitfalls of applying linear regression models to explaining the relationship between spot and futures prices in electricity markets. We briefly introduce the theory for the analysis of the spot-futures price relationship and highlight selected issues of multiple regression models.We show the bias coming from the simultaneity problem, analyze the effect of correlated measurement errors in both dependent and independent variables and discuss the properties of coefficients estimated in the presence of a seasonal component and without it. We show that these problems appear in the work of Botterud, Kristiansen, and Ilic (2010), who analyze the relationship between spot and futures prices in the Nord Pool power exchange. We study a very similar dataset and show that the conclusions of Botterud et al. are partially invalid or misleading. In particular, the effect of the water reservoir level on the risk premium is likely to be positive, which is to be expected, but contradicts Botterud et al. results. In addition we show that the coeffcients obtained by Botterud et al. can be interpreted only when we consider the ex-post risk premium and not the ex-ante one.