Predictability of Stock Price Behaviour in South Africa: A Non-Parametric Approach

This paper investigates the forecasting power of stock prices using two methods, namely, the random walk and the non-parametric methods. Using daily prices of the FTSE/JSE All Share index it is found that non-parametric methodology reveals distributional behaviour in the time series that is not captured by the random walk model. Based on the out-of-sample predicted mean square error, the F-test for two variances (those of both the observed series and the predicted one) and the bootstrap confidence interval and volatility, this method predicts the future behaviour of stock prices more accurately than the traditional random walk model has done.