Analysis of the cross-section of equity returns on the JSE Securities Exchange based on linear and nonlinear modeling techniques

Abstract

This research investigates the relationship between firm-specific style attributes and the cross-section of equity returns on the JSE Securities Exchange (JSE) over the period from 1 January 1997 to 31 December 2007. Both linear and nonlinear expected returns forecasting models are constructed based on the cross-section of equity returns. A blended approach combining a linear modeling technique with a nonlinear artificial neural network technique is developed to identify future potential top performing shares on the JSE. 1. Both linear and nonlinear models identify book-value-to-price and cash flow-to-price as significant styles attributes that distinguish near-term future share returns on the JSE. 2. This thesis found updating the identity of attributes is equally important as updating the factor payoffs of attributes in applying the stepwise regression approach. 3. Nonlinearity on the JSE equity returns is found to complement the forecasting power of linear factor models. 4. In terms of artificial neural network modeling, the extended Kalman filter learning rule introduced in the thesis is found to outperform the traditional back-propagation approach. 5. This thesis found that updating the identity of attributes via a genetic algorithm in the nonlinear forecasting models is superior to the static nonlinear forecasting models. 6. Both linear and nonlinear models are found to be more adequate in identifying future outperformers than identifying future underperformers on the JSE. The results of the research provide for potential alpha generating stock selection techniques for active portfolio managers in the South African equity market using the blended linear-nonlinear approach.