Robust portfolio selection based on asymmetric measures of variability of stock returns (Q843134): Difference between revisions

Since asset expected returns and their covariances are often measured with errors, and because of this the resulting solutions to the Markowitz model of portfolio selection are very sensitive to these measurement perturbations, there has been a need to device a new approach which will take into account the estimation errors. This paper proposes a new uncertainty set with randomly fluctuating bounds \(\xi_i= [u_i- \theta^1_i, u_i+\theta^2_i]\) instead of earlier studied ``unknown-but-bounded'' uncertain sets. This new form of the uncertainty set is able to capture asymmetric downside and upside deviations of real world data. The authors apply their interval chance-constrained programming to find robust solution to the mean-variance portfolio selection. A hybrid-intelligent algorithm is employed for this purpose. The numerical experiments suggest that the worst-case behavior of portfolios can be improved using the proposed methodology. The robustness is achieved at low cost.