A Comparison of Volatility Models in the Korean Stock Market

This paper attempted, on the basis of the data of KOSPI200 daily returns and 10 minute intraday returns, to identify which one of GARCH-family models is the more suitable in modelling the volatility of Korean stock market returns. We performed a comparative analysis of a set of 252 GARCH-family models. It was found, first, that the asymmetric EGARCH and GJR-GARCH models can better explain the behavior of KOSPI200 daily returns than the others. Also, the t-distribution or GED distribution were found more appropriate in accounting for the distribution of error terms. Second, it was found that the EGARCH and APARCH models are better in explaining the pattern of intraday returns of KOSPI200. Higher-order models were more suitable for intraday returns, in contrast to daily returns. As for the distribution of error term, the t-distribution was much better than normal and GED distribution. Third, asymmetric models like EGARCH and GJR-GARCH models exhibit a lower loss functions. The asymmetric GARCH models with non normal error distribution tend to improve the volatility forecast of stock returns. To sum, the asymmetric models were found more appropriate to account for both daily and intraday returns of KOSPI200. And low-order models are recommended for daily returns and high-order models for intraday returns.