PORTFOLIO OPTIMISATION USING RISK CLUSTERING AND CORNISH-FISHER VALUE-AT-RISK: AN APPLICATION TO SINGAPORE STOCKS

Abstract

The classical mean-variance portfolio strategy is often criticized for penalising positive deviations in returns when investors mainly care about downside risk. Moreover, it fails to account for the skewness and kurtosis of portfolio returns. Portfolio optimisation using the Cornish-Fisher Value-at-Risk (CF-VaR) addresses these weaknesses by choosing the asset weights to induce positive skewness while reducing kurtosis so as to minimise the adverse tail risk of the portfolio. However, CF-VaR optimisation can be computationally intractable for large number of assets. By clustering stocks based on their risk profile into a handful of groups and selecting one asset from each group, the number of assets is reduced thus making the CF-VaR optimisation strategy computationally feasible. When applied to the Singapore stock market, the combined strategy of clustering and CF-VaR optimisation produces strong synergies that appears to beat classical portfolio strategies and the popular passive strategy of investing in the Straits Times Index.