Optimal Bayesian portfolios of hedge funds

Hedge fund returns are not normally distributed. Hedge fund styles related to arbitrage strategies exhibit negative skew while more directional styles, such as managed futures and global macro, are more positively skewed. We implement and test a Bayesian framework for portfolio optimisation process in order to take these characteristics, as well as the estimation risk, into account. Hedge fund returns are modelled using multivariate skew elliptical distributions. The first three predictive estimates are used in a truncated utility function to obtain sets of optimal portfolios. We show that the choice of the underlying distribution, as well as the modelling of co-skews has an important impact on the final optimal portfolios.