Robust optimization by constructing near-optimal portfolios
Many investors use optimization to determine their optimal investment portfolio. Unfortunately, optimal portfolios are sensitive to changing input parameters, i.e., they are not robust. Traditional robust optimization approaches aim for an optimal and robust portfolio which, ideally, is the final investment decision. In practice, however, portfolio optimization supports but seldomly replaces the investment decision process.
In this paper, we present an approach that both solves the robustness problem and aims to support rather than replace the investment decision process. The method determines a region with near-optimal portfolios which, especially in light of the robustness problem, are all good allocation decisions. Then, as is already common practice, an investor can bring in expert opinion or additional information to select a preferred near-optimal portfolio. We will show that the region of near-optimal portfolios is significantly more robust than the optimal portfolio itself.