In this paper, we have estimated the memory of theTehran stock exchange indices. The estimation of fractional differencing parameter is carried out by various methods such as MLE, NLS, Hurst Exponent, GPH, Lo, whittle and wavelet.
The estimation results of whittle, wavelet, Hurst, and Lo methods allow us to conclude that the returns on stock indices (TEPIX, TEDPIX, TEDIX, Financial index and Industrial index) have long memory. The results obtained from GPH method show the existence of long memory in all indices of the Tehran stock market with the exception of TEDIX.
Since estimations which obtained from ML and NLS methods are not significant in almost all intervals, we do not use them for studying the trend of the market’s memory.
The results also show the memory of the Tehran stock exchange does not have an important trend . In the other words, in our study period the efficiency of the market does not show any significant changes
JEL Classifications: C14, C32, D53