Modeling the daily electricity price volatility with realized measures

• Published in: Energy economics Vol. 44; pp. 492 - 502
• Main Authors: Frömmel, Michael, Han, Xing, Kratochvil, Stepan
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01-07-2014; Elsevier; Elsevier Science Ltd
• Subjects: Applied sciences; Benchmarking; Benefits; Economic data; Economics; Electric energy; Electric power; Electricity; Electricity pricing; Energy; Energy economics; Energy industry; Energy market; Energy prices; Exact sciences and technology; Forecasting techniques; Forecasts; Gaussian; General, economic and professional studies; Intraday range; Markets; Mathematical models; Methodology. Modelling; Normal distribution; Prices; Realized GARCH; Securities markets; Specifications; Stochastic models; Studies; Volatility; Volatility forecasting; Europe; Realized GARCH; C52; C13; Electricity; Q47; G15; Volatility forecasting; G17; Q41; Intraday range; Volatility; Time series; Econometric model; Daily variation; Autoregressive model; Generalized; Modeling; ARCH model; Price
• ISSN: 0140-9883; 1873-6181
• DOI: 10.1016/j.eneco.2014.03.001

We propose using Realized GARCH-type models to estimate the daily price volatility in the EPEX power markets. The model specifications extract the volatility-related information from realized measures, which improves the in-sample fit of the data. More importantly, evidence on the out-of-sample predictability reinforces the value of the specifications, as the forecast quality is improved over the benchmark EGARCH model under eight conventional criteria. In particular, we show that the benefit of including intraday range as a realized measure is more substantial than realized variance. All the key findings are robust under rolling-window and recursive estimation schemes, Gaussian and skewed t-distribution assumptions on the innovation process, and alternative specifications on the predictable price component.
•We model and forecast daily electricity price volatility on the European Power Exchange.•We apply a realized GARCH model based on two realized volatility measures.•The range-based volatility estimator performs best, in-sample, but also out-of-sample.•The out-of-sample forecast quality is improved over the benchmark model under eight conventional criteria.