Market Efficiency and Volatility in Stock Market of Bangladesh

Abstract

This thesis attempts firstly to seek evidence of the weak form efficiency of Dhaka Stock Exchange (DSE) by hypothesizing random walk assumption. In this case, both parametric tests (unit root test, variance ratio test, autocorrelation test and ARIMA model) and non-parametric test (run test) have been employed. Secondly, this study examines the volatility pattern of daily return, volatility-return relationship and contemporaneous trading volume-volatility relationship. Volatility models like GARCH (1,1), GARCH (1,1)-M, EGARCH (1,1) and GJR-GARCH (1,1) have been used to capture volatility dynamics in return series. In addition, the causal relationship between contemporaneous trading volume and volatility has been studied under VAR modeling framework. Now a day, forecasting stock price and return volatility have been considered as prime issues in finance. Theoretical basis of weak form efficient market hypothesis is that the successive stock price/return is independently and identically distributed and past prices/returns have no predictive content to forecast future trend. On the other hand, volatility clustering, leptokurtosis and asymmetric impact of news (leverage effect) are very peculiar characteristics of stock return. To examine and capture such types of phenomenon, this study uses daily closing value of two main indices (DGEN and DS20) of DSE for the period of 2001 to 2012. The both return series of DSE show positive skewness, excess kurtosis and deviation from normality. Results of unit root tests, run test, autocorrelation test and variance ratio test provide evidence that the return series do not follow random walk model. In addition, the coefficients of ARIMA are significant at various lags of autoregressive and moving average terms and using best fitted ARIMA (3,0,1) model for DGEN return series and ARIMA (2,0,2) model for DS20 return series, future return can be predicted lucratively. On the other hand, ARCH-LM test shows significance presence of heteroscedasticity in return series and GARCH family models capture the phenomenon effectively. Results of volatility models exhibit the presence of volatility clustering (i.e., large change follow a large change and small change follow a small change) in return series. In DSE, impacts of shocks to volatility are highly persistent and old news is as much important as new news. Findings of GARCH-M model indicate the relationship between volatility (time-varying risk) and return is positive and significant. GJR-GARCH model ensures the existence of leverage effect i.e., the bad news have more impact on volatility than the good news of equal magnitude. Here, we also measure the impact of trading volume on volatility using best fitted GJR-GARCH model and found that there is significant and positive relationship between trading volume and volatility. The asymmetric impact of news on volatility becomes higher when contemporaneous trading volume is added as an additional explanatory variable in volatility model but it reduces volatility persistence. VAR and Granger causality test indicate that trading volume influences volatility at earlier and later both lags but volatility influence volume after 6 lags. Overall, the findings indicate that the Dhaka Stock Exchange is not efficient in weak form and highly volatile which is one of the main barriers to investing in this market. Findings obtained in this study have significant implications to the investors, security analysts, policy makers and regulatory authorities and these findings can be used as important guiding rules to enhance the investors’ confidence and efficiency level in stock market of Bangladesh.