마코프전환 멀티프랙탈(Markov Switching Multifractal) 모형을 이용한 KOSPI200 수익률의 장기변동성 예측성과 비교

초록

Volatility clustering, long memory, regime change, and the presence of outliers are wellknown stylized facts of volatility of financial time series. This paper estimates the volatility model of KOSPI200 index returns using a Markov switching multifractal (MSM) model that is known to capture these characteristics with relatively small number of parameters and compares its multi-horizon forecasting performance with that of popular competing models such as a standard GARCH model and Markov switching GARCH model. Using the sample period covering from January 2, 2003 to December 30, 2014, this paper finds that the MSM model outperforms competing models in terms of both in-sample goodness of fit and out-of-sample forecasting performance. Particularly, the forecasting performance tends to become more prominent as the forecasting horizon increases. As a byproduct the MSM model allows KOSPI200 return volatility to be decomposed into short-term and medium- to long-term volatility components. The examination of computed volatility components suggests that the nature and effect of volatilities due to the 2008 global financial crisis and 2011 European sovereign debt crisis are quite different: in addition to the increase in short- and medium-term volatilities, the low-frequency longterm volatility is also increased during the former crisis period resulting in Korean stock market shifting into the higher volatility state, whereas the shift in the volatility state is not observed during the latter crisis period implying that the effect of European debt crisis to Korean stock market was big, but only transitory.

키워드