Publication Abstract

Abstract 

This paper assesses the forecasting performance for the highly volatile and non-storable two commodity prices, electricity and gas with the volatility models- Autoregressive-Conditional-Heteroscedasticity types under alternative error distributions. The forecasting performances of models are compared with proposed specifications. The results suggest that the proposed an augmented ARCH model -the EGARCH_2KNT for in-sample estimation, and for the out-of sample, the GARCH_2KNT and GARCH_KPT models fit well when evaluated with the RMSE. The generalised error distribution can improve the out-of-sample accuracy in particular for the dynamic forecasting while the Student-t distribution improved the accuracy of both in- and out-of-sample estimation. Overall the static forecasting shows lower forecasting errors for the 30-day forecasting in gas prices, and the 10-day and 90-day in the electricity prices for the sample period between 2003 and 2008.  Application is made for the UK electricity and gas markets that have recently structurally reformed and liberalised, and the energy derivatives on both prices are traded therefore the accuracy of the out-of-sample forecasting for the behaviour of price volatility is essential for derivatives valuation for trading and utilizing a risk management tool for a power plant portfolio as an example.