Long Memory Behavior in the Chinese Stock Market Based on Semiparametric Estimation Method

Abstract

In this paper long memory property is examined in the Chinese stock market by means of high frequency data. Two semiparametric methods in frequency domain, local Whittle (LW) estimation and log periodogram (LP) regression, are used to analysis the fractional integration order d. The results show that LW estimation can solve choice of the parameter m compared to LP regression, and neglect intraday effect of high frequency data, which can prove the scale invariability of long memory. Thus, LW estimation is applied to find the relationship between long memory and the intrusive events, which provide that exogenous shock induced by the events appear to have more intense long memory behavior