The Journal of Applied Business Research – November/December 2014 Volume 30, Number 6 Copyright by author(s); CC-BY 1587 The Clute Institute A Comparison Of Mean-Variance And Mean-Semivariance Optimisation On The JSE Jiten Vasant, University of Cape Town, South Africa Laurent Irgolic, University of Cape Town, South Africa Kanshukan Rajaratnam, University of Cape Town, South Africa Ryan Kruger, University of Cape Town, South Africa ABSTRACT This study investigates the effectiveness of semivariance versus mean-variance optimisation on a risk-adjusted basis on the JSE. We compare semivariance and mean-variance optimisation prior to, during and after the recent financial crisis period. Additionally, we investigate the inclusion of a fixed-income asset in the optimal portfolio. The results suggest that semivariance optimisation on the JSE in a pure equity case produces lower absolute returns, yet superior risk-adjusted returns. Further investigation suggests that semivariance metrics are effective within a certain range of portfolio sizes and diminishes in benefit once portfolios become larger. A fixed income asset scenario tested under the hypothesis of semivariance optimisation favoured greater bond weightings in optimal portfolios. Keywords: Semivariance; Johannesburg Stock Exchange, Optimisation, Mean-Variance 1. INTRODUCTION Portfolio theory has been evolving ever since its inception by Harry Markowitz and his pioneering research shaped the portfolio risk-return model known as mean-variance optimisation. A newer concept of post-modern portfolio theory is that of downside risk metrics for risk measurement in the context of portfolio optimisation. Previously the convention was to use a mean-variance strategy to determine the efficient frontier for a portfolio. The foundations of this theory however, are based on a set of strict assumptions with the result that the majority of models fail to capture reality perfectly and exhibit significant model risk. Downside risk optimisation, which models the efficient frontier using semivariance, has exhibited potential for providing better risk metrics. This paper aims to test the use of semivariance as a more realistic method of portfolio optimisation in a South African context. While empirical testing has been investigated on foreign markets, studies relating to South Africa in particular are limited. Furthermore, the use of downside risk optimisation has implications for the management of pension funds, thus providing cause for research. In addition, portfolio size will be examined in relation to both optimisation models. Prior research in other financial markets suggests that larger-sized portfolios are indifferent to the method of optimisation used. The effects of including a fixed-income asset will also be tested against both models. The remainder of this paper is structured as follows. Section 2 reviews prior literature on the development of mean-variance and semivariance. Section 3 presents our data and discusses the methodology adopted in testing the hypotheses. The results are presented in Section 4 while Section 5 concludes.