Handling Parameter Uncertainty in Portfolio Risk Minimization

Abstract

Portfolio allocation decisions are frequently made according to optimization algorithms that treat parameters such as means, variances, and covariances of returns as given. These parameters, however, are estimated through error-prone procedures like statistical modeling or subjective evaluation. Robust optimization has become a way to incorporate uncertainty directly into the formulation of optimization problems. The technique can be applied to modeling uncertainty in the expected returns in portfolio shortfall minimization. Tests using simulated and real market data suggest that portfolio allocation strategies resulting from robust optimization formulations outperform strategies obtained using classic optimization methods.