Hedge Fund Replication: Old theory, new interest, inconclusive results
Consultant and journalist Pierre Saint-Laurent* continues his coverage of EDHEC’s Asset Management Days in Geneva this week for All About Alpha. On Tuesday, he attended a much anticipated presentation on hedge fund replication featuring researchers NoÃ«l Amenc, Jean-Christophe Meyfredi, FranÃ§ois-Serge Lhabitant and Walter GÃ©hin. Other industry leaders** joined the EDHEC group on a subsequent panel discussion. Here again is Saint-Laurent’s dispatch from Geneva.
Hedge fund replication is controversial. For some, it’s a gallant effort to lower fees, increase transparency and boost liquidity. For others, it’s a work in progress whose promoters may be getting ahead of themselves.
But one thing is for sure: Hedge fund replication is not a new subject according to EDHEC researchers. The general concept of factor modeling dates back decades, and one of its most famous applications to hedge fund returns was completed by Fung and Hsieh in their 2002 style-based research.
Factor-based modeling, however, raises four concerns for the EDHEC researchers:
- “specification risk” (are the factors well specified?),
- “model risk” (is the model linear or non-linear?),
- “sample risk” (is the model sensitive to outliers?),
- “stationarity risk” (is the factor exposure of hedge fund managers time-varying, requiring a conditional modeling approach?).
Factor models have shown a relatively good fit with in-sample data but not necessarily with out-of-sample data. The challenge, according to the researchers, is to specify factors that are not subject to one of the risks mentioned above. For this reason, EDHEC is not ready to replicate hedge fund returns just yet.
The second approach (the â€˜payoff distribution approach’) is a well-known one advocated in Amin and Kat’s 2003 Journal of Financial and Quantitative Analysis article. This approach aims to copy the distributional properties of hedge fund returns over a specific time horizon. (Think of it as a European option strategy with a payoff at some point in future). It aims to replicate a distribution by making multiple â€˜drawings’ from that distribution over the defined time horizon. It’s like flipping a coin: you could flip 10 heads in a row, but you know that the limiting value, with sufficient flipping, will be 0.5.
According to the researchers, this approach involves 3-steps:
- define a reserve asset and estimate the payoff function that maps the asset return to a hedge fund return,
- price the payoff function, and
- derive the replication strategy.
According to EDHEC, the synthetic creation of a payoff function makes this approach a novel one.
Cautioning that this approach is highly sensitive to the selection of the reserve asset, EDHEC attempted to reproduce the results using to Amin and Kat’s original methodology (as gleaned from their articles). Researchers used 13 EDHEC hedge fund indices, nearby futures on the S&P500 and the 3-month Eurodollar as a proxy for the risk-free rate. Out-of-sample data was applied to the analysis using 100,000 geometric Brownian simulations (to price the payoffs).
The result: highly variable goodness-of-fit measures depending on the hedge fund strategy (the best: convertible arbitrage. The worst: fixed income arbitrage.) But all-in-all, the average performance of clones fell far short of the indices’. According to the researchers, this could be due to the fact that – contrary to Kat and Palaro – EDHEC applied bear-market out-of-sample data, not only bull-market data.
Average performance of the equity market neutral clone was very good, while the averages on CTA Global, Emerging Markets and Short Selling were not (they were rejected by F-tests). With shorter timespans, clones for other strategies such as Merger Arbitrage and Relative Value also dropped out of contention.
EDHEC concludes that while factor-based models require a conditional and nonlinear approach, the “payoff distribution” approach requires a patient investor who doesn’t care as much about portfolio performance and cares more about the higher moments of the return distribution.
As Martellini told AllAboutAlpha yesterday, the EDHEC researchers suggest caution and insist they do not have a definitive view on the “pay-off distribution” approach. They say it is difficult to test the published methodologies and that the required modeling is very complex. Nevertheless, FranÃ§ois-Serge Lhabitant indicated in his presentation that, as it stands today, hedge fund replication may be a tough proposition. He questioned an investment for which an investor needs to wait 10 years to confirm success and which may show negative returns in the meantime. He also cautioned that this approach may behave poorly in â€˜tail events’ such as a crash – precisely when accurate cloning is required most. Lhabitant also posited that replication essentially replaces manager risk with model risk.
But the industry panelists were more divided. Oliveros** indicated that some existing approaches actually just replicate traditional underlying assets. Still, he lauded hedge fund replication as a way to help investors better understand the sources of hedge fund value-added. Woods** pointed out that hedge fund replication may actually be more robust than currently reported given new research on the topic. He added that, although hedge fund replication may not be possible for all strategies (especially highly idiosyncratic ones), it still has the potential to reduce fees while increasing transparency and liquidity. Van Santen** indicated that the replication of highly illiquid and opaque strategies with liquid and transparent instruments seems somewhat contradictory. After all, hedge fund returns are derived from proprietary, idiosyncratic, and non-transparent investment strategies.
But in the end, most panelists agreed that powerful marketing will quickly drive adoption, irrespective of academic trepidation. (ed: a view shared by Prof. Kat in an op/ed for All About Alpha).
* Pierre Saint-Laurent, MSc, CFA, CAIA is president of AssetCounsel Inc., one of Canada’s top alternative investments consultancies with offices in Toronto and Montreal.
**Gumersindo Oliveros, the World Bank; Christopher Woods, State Street Global Advisors; Pim van Santen, Shell Pension Fund.