Three essays on stock returns predictability and trading strategies to exploit it
This thesis is organized in three self-contained projects which model predictability in both advanced and emerging stock markets and attempt to exploit it via construction of appropriate trading strategies. The objectives of this research are: 1) to model mean reversion in developed stock markets and re-assess the mixed empirical findings to date; 2) to characterize the returns generating process in emerging capital markets and examine the predictive ability and profitability of technical trading rules; 3) to develop and evaluate whether trading strategies involving dividend announcements in the UK are profitable and can be classified as statistical arbitrages, with consequent implications for the market efficiency hypothesis. We investigate the existence of mean reversion in the G-7 economies using a two factor continuous time model for national stock index data. Whilst maintaining the same modeling philosophy of previous studies, we rather focus on the effects of the "intrinsic" continuous time mean reverting coefficient. Our method produces support for mean reversion, even at low frequencies, and relatively small samples. We also aim to characterize the stock return dynamics in four Latin American and four Asian emerging capital market economies and assess the profitability of popular trading rules in these markets. We find that dollar denominated returns exhibit statistically significant long memory effects in volatility but not in the mean. "wading' our findings via a number of moving average and trading range break rules, we "beat" the buy and hold benchmark strategy in all markets before transaction costs, and in Asian markets even after transaction costs. Bootstrap simulations further reinforce the choice of the modeling framework and the trading outcomes, particularly for Latin American markets. Finally, we investigate whether trading strategies designed to exploit "abnormal" price behavior following dividend initiation/resumption and omission announcements of UK firms pass the statistical arbitrage test of Hogan et al. (2004). To mitigate concerns regarding "risky" arbitrage, we also calculate the probability of making a loss for each strategy. We find that strategies involving portfolios of dividend initiating/resuming firms are profitable and converge to riskless arbitrages over time, while this is not the case for strategies with dividend omitting firms, contrary to what is suggested by US studies. In general, the robustness of our results casts doubt on the market efficiency hypothesis in both developed and emerging capital markets.