Is There Long Memory in Indian Stock Market Returns? An Empirical Search

Abstract

The issue of long memory, though has important theoretical and practical implications, has not received much attention in India. This article examines the issue of long memory in mean of the stock returns by employing a set of sophisticated time-series tests including a bias reduced log periodogram test of Andrews and Guggenberger. The study used daily values of 29 major indices including sectoral indices traded on the National Stock Exchange and Bombay Stock Exchange from April 2003 to March 2012, which provide insights into relation between composition of indices and long memory. The findings of the study suggest significant presence of long memory in mean returns of the medium- and small-sized indices and weaker evidences for large cap indices. Further, the study identifies a relationship between presence of long memory and market structure variables. The use of linear models in the presence of long memory would result in incorrect inferences, and this calls for investigation of appropriate long memory model to generate profits in Indian stock market.

KEYWORDS:

• long memory
• fractional integration
• random walk
• market efficiency
• Indian stock market
• NSE
• BSE

Notes

1. For a further discussion on these conditions, see, Palma (2007).

2. If the estimated value of d is greater than 0.5, it is called as nonstationary long memory.

3. Brazil introduced structural reforms, known as Real Stabilization Plan in 1994 keeping the objective to stabilize macroeconomic uncertainties.

4. When r = 0 the test becomes GPH.

5. The statistics are not provided to save the space. They are available with the authors.