Test Of Existence Of Long Term Memory In Stock Market Returns At Nairobi Securities Exchange

Abstract

Long term memory in stock market returns has received considerable attention among academicians and finance practitioners. This paper explores the applicability of Fractal market hypothesis and Chaos theory in explaining market behaviour. The rationale of both theories that markets patterns can be studied for the possibility of predictability of returns to inform investment decisions motivated this research subject. The study embarked on the objective to test for existence of long range dependence in stock market returns in Nairobi securities exchange. Based on the fact that post automation the market is expected to have improved efficiency. The study employed a non parametric test; classical rescaled range analysis to examine long term memory which is measured by the Hurst exponent developed by Hurst (1951). The stock market returns were considered using secondary data. The daily NSE-20 share index was collected for a period of eight years from January 2010 to December 2017.A longitudinal research design was employed for the research. The data was analyzed using E-views financial software. The results show that there is long term memory in stock market returns in NSE with an H-value of 0.7 from the rescaled range analysis. It is further observed that market returns are not normally distributed for the test of normality with a negative skewness of -0.067 and the autocorrelation denoted by P-value <0.05 showing the market does not follow a random which actually invalidates the efficient market hypothesis. This indicates that there exists a chance to predict market returns and make above market profits. The research recommends factoring in long term memory properties in investment decisions.