Bounds for portfolio weights in decentralized asset allocation

ABSTRACT We consider an asset allocation problem for a multi-asset fund where multiple decentralized managers implement investment strategies in separate asset classes that include alternative investments. In this setting it is common practice to use traditional asset allocation methodology (e.g. mean-variance optimization) to obtain optimal target weights and to compel investment managers to stay close to their targets. However, the sophistication and high level of specialization involved in alternative investments often result in deviations from target weights. Misallocations may require ad hoc investment adjustments to maintain feasibility of the fund with respect to overall risk and return profiles. We develop an asset allocation model that provides a range of portfolio allocations (i.e. bounds) for individual managers. These bounds represent the maximum flexibility managers have in their investments without affecting the overall portfolio characteristics. Based on out-of-sample testing, our model is more robust than traditional asset allocation methodology with respect to misallocations.