Capturing Risks of Non-Transparent Hedge Funds

Abstract

We present a model that captures risks of hedge funds only using their historical performance as input. This statistical model is a multivariate distribution where the marginals derive from an AR(1)/AGARCH(1,1) process with t_5 innovations, and the dependency is a grouped-t copula. The process captures all relevant static and dynamic characteristics of hedge fund returns, while the copula enables us to go beyond linear correlation and capture strategy-specific tail dependency. We show how to estimate parameters and then successfully backtest our model and some peer models using 600 hedge funds.

Keywords: Hedge fund, AGARCH, copula, risk

JEL Classification: C32