Abstract
The forecasting of stock returns is an area of interest that has attracted much attention, with many authors looking to find economic and financial variables that can forecast the mean of excess returns in stock markets across different countries. South Africa is not exempt to this, with authors attempting to unearth forecastability of stock returns; the experiment has produced mixed results as some works have found evidence that stock returns are forecastable while others have failed to find evidence of forecasting ability. Some reservations regarding the use of mean forecasts exist from a perspective of asset allocation decisions. The objective of this study is to assess the performance of binary probit models in South African stock market returns. The study will use variables from the literature to forecast stock market direction from February 2003 to January 2018. This study has two objectives, to use dynamic binary model to assess the in-sample forecastability of the direction of stock market returns in South Africa; and to analyse the performance of conditional country risk improving the accuracy of the binary probit models in the study. The variables employed in this study include the South Africa MSCI country index, US 3-month Treasury bill rates, US short-term and long-term government bond yields, as well as the dividend yield and earnings yield for South African stock returns. The data was obtained from Datastream, the South African reserve bank and the Federal Reserve. The in-sample results show that certain variables are significant in explain the direction of stock returns in South Africa. Dynamic extensions of the static probit model are shown to have improved performance in forecasting the direction of stock markets and produce superior returns in trading simulations. Out-of-sample results show risk measures based on conditional residual risk produce superior forecasting performance, and the risk measures produce portfolios that outperform the buy-and-hold strategy...
M.Com. (Financial Economics)