Return predictability in South Asian stock markets

Research Doctorate - Doctor of Philosophy (PhD)Stock return predictability has been a subject of considerable interest due to its implications for market efficiency and devising investment strategies. The extant literature on stock return predictability is overrepresented by studies providing spurious evidence of predictability. Some of the key sources of spurious findings are the use of a biased estimator, data snooping, and statistical inference based on test statistics with poor size and power properties. The literature on return predictability in emerging markets is still in its infancy and routinely ignores these potential sources of spurious results. The central aim of this thesis is to investigate time-varying return predictability in the emerging stock markets of South Asia by mitigating spurious predictability. This thesis comprises three studies, a descriptive study and two empirical studies. The descriptive study provides a comparative analysis of characteristics of the four South Asian stock markets – Bangladesh, India, Pakistan and Sri Lanka. These markets have exhibited substantial growth in size and liquidity over the last two decades, which may have been attributed to the opening of capital markets to foreign investors, automation of the stock exchanges, and adoption of international standard corporate governance. The India and Pakistan markets are more developed and integrated with global markets than the Bangladesh and Sri Lanka markets. The latter two markets lack adequate liquidity, an efficient trading mechanism and appropriate security regulation. The first empirical study investigates return predictability from the pattern of past price changes. While past evidence of return predictability on emerging markets may be imprecise due to the use of efficiency tests with low power, this study uses the wild bootstrap automatic variance ratio test and price delay that possess superior statistical properties. In line with the partial price adjustment hypothesis, this study shows significant autocorrelation in stock returns, which is inversely related to firm size and trading frequency. Greater delay in incorporating global information compared to local information into stock prices may reflect limited participation of foreign investors. Time-varying return predictability is found to be negatively related to market development, automated trading and volatility regime, however, positively associated with financial crisis and liquidity. This finding is consistent with the adaptive market hypothesis (Lo, 2004). The second empirical study evaluates dividend yield and short-term interest rate’s ability to predict stock returns. Unlike previous studies of return predictability in emerging markets, this thesis mitigates the risk of finding out biased predictive relationships, by employing a restricted vector autoregressive model and estimating model parameters with the generalised least squares estimator. Dividend yield emerges as a significant predictor of stock returns in India and Pakistan while short-term interest rate is a significant return predictor in all the four markets. After controlling for model uncertainty and parameter instability by imposing theoretically motivated parameter restrictions, the predictive models generate superior out-of-sample forecasts both in statistical and economic terms (utility gains). The predictive ability of dividend yield and short-term interest rate is found to vary over time and is moderated by the degree of persistence of the predictor variable and business cycle fluctuations. Given that the extent of return predictability depends on economic states, and the sensitivity of return predictability to macro factors varies across the countries, investors may be able to devise market timing and country rotation investment strategies