Estimating Long Memory Volatility Using High-Frequency Data of Asian Stock Markets

Abstract

This article analyzed the presence of long memory in volatility in 5 Asian equity indices namely SENSEX, CNIA, NIKKEI225, KO11 and FTSTI, using 5 minutes intraday return series ranging from 05-jan-2015 to 06-Aug-2015. The study employed ARFIMA-FIGARCH model and ARFIMA-APARCH model and compared them with GARCH (1,1) model and APARACH(1,1) in terms of in-sample forecast accuracy. The results confirmed the presence of long memory in both the return and volatility series for all the five markets under study. Among the group, CNIA and STI showed most persistence in both the return and conditional volatility. In terms of forecast measures, the long-memory GARCH models were found to be performing better compared to the short-memory GARCH models.