Portfolio Allocation of Hedge Funds

Research in hedge fund investing proposes different solutions to build optimal hedge fund portfolios. However, these solutions are direct extensions of the usual meanvariance framework, and still suffer from model risks. More complex approaches start to be used but are related to numerous estimation risks. We compare in this paper the out-sample properties of different allocation models through a dynamic investment exercise using hedge fund indices. We show that the best out-of-sample properties are obtained by allocation models that take into account the specific statistical properties of hedge fund returns.