Power exponential price returns in day-ahead power exchanges

This paper uses the Subbotin power exponential family of probability densities to statistically characterize the distribution of price returns in some European day-ahead electricity markets (NordPool, APX, Powernext). We implement a generic non-parametric method, known as Cholesky factor algorithm, in order to remove the strong seasonality and the linear autocorrelation structure observed in power prices. The filtered NordPool and Powernext data are characterized by an inverse relationship between the returns volatility and the price level - approximately a linear functional dependence in log-log space, which properly applied to the Cholesky residuals yields a homoskedastic sample. Finally, we use Maximum Likelihood estimation of the Subbotin family on the rescaled residuals and compare the results obtained for different markets. All empirical densities, irrespectively of the time of the day and of the market considered, are well described by a heavy-tailed member of the Subbotin family, the Laplace distribution.