Abstract

An accurate wholesale electricity market forecast has become an essential tool in bidding and hedging strategies in competitive electricity markets. This paper provides a dynamic asymmetric long memory heteroscedastic model to account the high volatile daily wholesale electricity markets in New England and Louisiana. This model implemented power Cox-Box transformation (Tse, 1998) under the Chung’s (1999) model specification to the time-varying volatility. The model is able to capture various empirical stylized facts that commonly observed in electricity markets including clustering volatility, news impact, heavy-tailed and long memory volatility. Under the forecast evaluations, the long memory model outperformed the traditional model in all the forecast time-horizons. Finally, the outcome of the analysis is further applied in quantifying the market risk in term of value-at-risk.

Key words: Electricity markets, long memory generalized autoregressive conditional heteroskedasticity (GARCH), value-at-risk, time series analysis.