Using constraints to improve the robustness of asset allocation

Practitioners of mean-variance optimization frequently constrain the share of asset classes within portfolios. This is achieved by imposing a floor or a ceiling on asset classes during optimization. The purpose of the constraints is to guarantee that the allocations of the portfolios that make up the efficient frontier align with investor perception. Using an alternative example, the authors provide an alternative interpretation of the constraints. They show that the preference that investors reveal for certain asset classes may be mapped into the relative confidence they place in their risk-return estimates. Further, the authors provide an intuitive explanation as to why constraints that lead to dramatic differences in the composition of the underlying portfolios may not result in a significant cost in terms of forgone expected returns, and vice versa.