A Conditional Approach to Hedge Fund Risk

Abstract

This article applies a two-step conditional Bayesian approach to hedge fund risk. First, a mixture of two normal distributions is estimated for a core asset; one distribution being identified as linked to a “quiet” regime and the other to a “hectic” regime. The conditional probabilities of each regime are then inferred and a mixture of distributions is deduced for peripheral assets. The core asset is alternatively chosen as the S&P index or the Baa/Treasuries yield spread whereas the peripheral assets are chosen to be the major hedge funds strategies over the period 1990–2004. This methodology has several advantages given specific features of hedge funds returns, notably non-linear exposure to standard assets returns and short sample history. Significant changes in the distribution (mean and standard deviation) of hedge fund returns are identified across regimes. Results are less clear-cut for the correlation with standard assets, as modifications can be imputed to a certain extent to a form of selection bias. An application of this methodology to stress tests on hedge funds portfolios is presented.

TOPICS: Real assets/alternative investments/private equity, big data/machine learning, risk management, portfolio construction