Using Constraints to Improve the Robustness of Asset Allocation

10036). nyone famhar with an asset allocation software package has faced the problem that the composition of a portfolio that lies on the frontier just doesn’t seem like a reasonable mix to hold to acheve the inlcated risk and return. As a result, most users end up constraining the allocation of the asset classes that they are “uncomfortable” with until the portfolios that fall on the frontier align with their perceptions. Our primary objective is to establish that investor intuition with respect to asset allocation (i.e., the revealed preference for certain asset classes) can be mapped into investor confidence pertaining to risk-return estimates of lfferent asset classes and be implemented as constraints during portfolio optimization. We achieve this objective by developing a rigorous framework for the interpretation of the constraints in a mean-variance asset allocation framework. We first examine the cost associated with constrained optimization in terms of forgone expected returns. In particular, we seek constraints that sacrifice the smallest possible expected return for a significant change in the portfolio allocation. Next, we use an dustration to demonstrate two benefits of constraining: 1) the abhty to include objective or subjective beliefs of the risk, return, and correlation estimates, and 2) the inclusion of other investor goals such as minimizing downside risk or the probability of drastic shortfalls.