In this paper, we focus on the predictability of hedge fund index returns and its eventual application in a fund of hedge funds. There is now a consensus in empirical finance that expected asset returns are, to some extent, predictable, at least for traditional asset classes. However, literature on evidence on return predictability of hedge fund is still in its infancy. Amenc, El Bied and Martellini (2002) were the first to report both statistical and economic significance of predictability in hedge funds returns.
Like Amenc et al. (2002), we use factor models to find evidence of predictability in various hedge fund index returns. Given that the true set of predictive variables is virtually unknown, we extend Amenc et al. (2002) empirical analysis using various forecasting models to analyze hedge fund index returns predictability and its impact to tactical style allocation (TSA) strategies. We take into account a larger number of predictive variables reflecting the stage of the economic cycle, the interest rate environment, and the dynamic trading strategies applied by hedge funds. These variables are able to predict changes in hedge fund index returns. We finally expand the sample period until December 2003.
In order to provide some evidence of the economic significance of these predictive models, we analyze their out-of-sample performance in terms of tactical style allocation. Three portfolio construction models are performed, all based on traditional optimization (i.e. mean-variance framework).
Traditional portfolio optimization models require forecasts of the portfolio expected returns and an estimate of their covariance matrix. In this paper, we estimate the expected returns using different factor models. The difficulty arise when there is no consensus on the most appropriate factor model, that is why we attempt in this paper to compare an extensive number of different factor models beginning with the simplest form of the fund returns (linear single-factor models) and ending with more complex representation (non-linear multi-factor models).
The paper is organized as follows: the first section provides evidence of return predictability in hedge fund indexes. The second section explores the practical application of our predictive models, i.e. their uses in tactical style allocation by fund of hedge funds managers. The third section concludes.
Contents
Introduction
1 Evidence of Predictability in Hedge Fund Index Returns
1.1 Review of Literature and Background Theory
1.2 Data
1.3 Predictive Variables
1.4 Methodology
1.5 The Predictive Models
- 1.5.1 Linear Single-factor Predictive Models
1.5.2 Linear Multi-factor Predictive Models
1.5.3 Non-linear Multi-factor Predictive Models
2 A Practical Application: TSA in Fund of Hedge Funds
2.1 Operational Constraint: Redemption Notification
2.2 Investment Constraint: Turnover and Diversification
3 Conclusion
Appendix A. Hedge Fund Index Data
A.1 Hedge Fund Classification
A.2 Summary Statistics of Hedge Fund Index Returns
Appendix B. Predictive Variable Data
B.1 Summary Statistics of Predictive Variables
Appendix C. Others
C.1 Methodology Scheme
C.2 Optimal portfolio weights (without any upper weight constraint)
C.3 Optimal portfolio weights (with upper weight constraints of 20%)
References
