Korean Energy Economic Review ISSN:1599-7057(Print) 에너지경제연구

This paper uses a Markov Switching Multifractal (MSM) volatility model to forecast short-, medium-, and long-term volatilities of system marginal price (SMP) and compares forecasting performances with several competing models. MSM model is known to describe well- known stylized facts of the volatility process such as volatility clustering, regime change, long memory, and substantial outlier with relatively small number of parameters within a multi-frequency framework. Using the weekday sample covering from 2003 to 2015 this paper finds empirical evidence that, in terms of the in-sample goodness of fit, MSM model shows a substantial improvement compared to competing models and that, except for QLIKE, MSM model improves significantly the out-of-sample forecasting performance. This result can be confirmed from the formal test such as the Diebold-Mariano test. We also find that main conclusion holds for 2nd and 3rd sub out-of-samples except 1st. As such the results strongly suggest that considering both multi-frequency factors and Markov regime switching in modeling a rather complex process of the SMP volatility make a critical contribution towards improving the forecasting performance.