Extreme value theory for heavy tails in electricity prices

Florentina Paraschiv, Risto Hadzi-Mishev, Dogan Keles
2016 Journal of Energy Markets  
Typical characteristics of electricity day-ahead prices at EPEX are the very high volatility and a large number of extreme price changes. In this paper, we look at hourly spot prices at the German electricity market and apply extreme value theory (EVT) to investigate the tails of the price change distribution. Our results show the importance of delimiting price spikes and modeling them separately from the core of the price distribution. In particular, we get a realistic fit of the generalized
more » ... f the generalized Pareto distribution (GPD) to AR-GARCH filtered price change series, and based on this model accurate forecasts of extreme price quantiles are obtained. Generally, our results suggest EVT to be of interest for both risk managers and portfolio managers in the highly volatile electricity market. 2 large price increases. However, from a methodological point of view, modeling the negative tail of the distribution is completely analogous to modeling the positive tail. A comprehensive review of the existing literature on modeling electricity prices is given in [26] . Studies which apply time series models for conducting forecasts of the extreme quantiles for energy markets can be found in [6], [1], [11], and [15]. Most of these authors employ GARCH models with various specifications for the innovations. [21] applied a quantile regression model and found that it performs better in out-of-sample forecasts of the electricity price distribution than pure time-series models like CaViaR and GARCH models. [7] and [24] show that EVT-based models are well suited for modeling the extreme tails of electricity prices. In this paper, we extend the existing literature on EVT-based models by a rigorous examination of the location of the threshold at which the tail becomes extreme. We found realistic forecasts of tail quantiles in electricity prices, which is of great importance for risk management purposes. The rest of the paper is organized as follows: Section 2 briefly introduces the main characteristics of electricity prices. Section 3 focuses on the theoretical background of the EVT (POT) approach. Section 4 shows the application of EVT in estimating and forecasting the extreme tail quantiles of the electricity price change distribution. Section 5 concludes.
doi:10.21314/jem.2016.141 fatcat:cmwlyjp5uvfvbcn23wip52fcaq