Long Memory in Stock Returns: Evidence from The Dhaka Stock Exchange

This study examines the long memory property in the weekly return series and its certain transformations of the Dhaka Stock Exchange over the period of January 1989 to January 2004. Well-known methods for detecting the long memory property of a time series such as the classical rescaled range (originally developed by Hurst, 1951) and its modified version propounded by Lo (1991) are used. Empirical results obtained in this study suggest statistically significant but weak evidence of long memory for weekly stock returns at levels. But for nonlinear transformations of return, such as the absolute and squared returns, the series show strong and significant long memory. The finding that above mentioned transformations of return series contain long memory supports the claim made by Taylor (1986) and Ding et al. (1993).