Evidence of Long Memory in the Indian Stock Market

Abstract

This article tests the small sample properties of three approaches in estimating the long-range dependence measure (the R/S approach, the R/S-AL approach and the Detrending Moving Average (DMA) approach) using Monte Carlo simulation experiments. We use simulated series of Gaussian noise and fractional Gaussian noise of varying lengths (252, 504, 1008, 2016 and 4032) which approximately represents daily data series of 1 year, 2 years, 4 years, 8 years and 16 years period. We observe a pronounced reduction in the variance and the mean absolute error under the DMA approach when the sample size increases, which indicates that the DMA approach performs better than the R/S and the R/S-AL approaches. In addition, we test for the presence of long memory in daily data from the Indian stock market. In particular, we examine the dynamic nature of market efficiency in the Indian stock market by using a moving sub-sample window of 1,008 observations (approximately 4 years of daily data). Our findings support the presence of long-range dependence in the Indian stock market. Furthermore, S&P CNX Nifty shows a highly dynamic behaviour in its efficiency characteristics.