Abstract:

The argument that hedge fund performance is primarily attributed to systematic risk-taking rather than 'skill' is lately gathering pace. Both academics and practitioners have thus been attracted by the idea of developing effective hedge fund return replication strategies with an eye to create synthetic products. The benefits from synthetic hedge funds are several, the most obvious relating to reduced cost, liquidity, transparency, and lift of barriers to entry.

Our research aims to: a) investigate the appropriateness of unexplored advanced model selection and risk attribution methodologies in mimicking hedge fund returns, and b) determine successful trading strategies that will allow the development of structured products with hedge fund-like risk/return profile.

The current literature focuses on two different approaches: moment matching and factor based replication. The latter approach is very intuitive and attractive from a conceptual standpoint. While it has gained significant interest we believe that some critical aspects have yet to be addressed. One key aspect is the determination of the relevant risk factors. Another important aspect is the methodology used to determine the weights of the factors in the replicating portfolio.

We propose to use regularization methods which allow to: a) parsimonious choice in the number of active asset positions, b) control transaction costs by promoting sparsity and c) reduce the sensitivity of the optimization to possible collinearities between assets. Our analysis is carried out with a wide variety of hedge fund indices as well as individual hedge funds.