In March, we learned that a reliance on monthly data to assess hedge funds might be leading us astray. Thomas Schneeweis, Editor of the Journal of Alternative Investments, argued that daily, or even weekly, data was far superior than monthly information.
While Schneeweis was referring to the relative advantages of managed accounts at the time, the same general theme has now made its way into the debate on hedge fund replication. Alpha Male shared a cold one with one of the luminaries of hedge fund academia a couple of months ago (a Hall of Fame member) when said guru expressed concern over the so-called “distributional replication” approach to hedge fund cloning. In his opinion, Kat’s reliance on monthly data to pump out daily trading instructions was a source of potentially considerable error.
Now another group of academics, backed by one of the world’s largest hedge fund investors, has attempted to address this concern. Nicolas Papageorgiou, Bruno Remillard, and Alexandre Hocquard of Montreal’s HEC Business School have just released the first version of a paper that aims to improve on the Kat-Palaro method (available here at AllAboutAlpha.com). Papageorgiou tells AllAboutAlpha.com that Canada’s Desjardins Global Asset Management has been quietly managing a “beta” fund since late 2006 and will be adopting this approach for an official launch in September. So expect to see Desjardins and Papageorgiou on red carpets around the hedge fund conference circuit this Fall (including this one featuring a panel of both Papageorgiou and Kat for – with apologies to Seinfeld’s Mr. Peterman – “a good old fashioned Kat fight“).
According to Papageorgiou:
“The efficiency measure as presented by Kat and Palaro (2005) is…subject to several shortcomings and inconsistencies. The most significant of these relates to the way that the daily trading strategies are derived from the distribution of monthly returns. The properties of the estimated monthly distributions and copula functions proposed by the authors are not infinitely divisible and therefore the true properties of the daily returns are not known. As a result, the replicating strategy will not be precise.”
He goes on to argue that a more “precise” model decreases the so-called “time to convergence” required before the replication’s statistical properties match those being targeted. Furthermore, he says that the precision of the replication strategy must be known before any conclusions can be drawn about the replicatability of various hedge fund indices:
“[Kat and Palaro’s] analysis can…be misleading if we do not also examine the precision of the replication strategy. Before dismissing the hedge fund indices as poor-performers, we need to properly evaluate whether the properties of the replication strategies and hedge fund indices are truly the same. A proper examination of both the cost and the precision of the replication strategy is fundamental before any strong conclusion can be drawn about the model’s ability to replicate hedge fund indices.”
Papageorgiou asks a question similar to the one posed by Northwater in a recent paper, namely: does the choice of reserve asset (the asset or assets being dynamically traded) impact the precision of the outcome? In particular, he concludes that while the choice of reserve asset might affect the mean return of the replication, it does not affect the precision (a.k.a. “success“) of the replication model. In fact, Papageorgiou doesn’t mince words as he wonders why previous studies have neglected this finer point:
“Contrary to the conclusions put forth by recent studies at EDHEC and Northwater (2007), the choice of reserve asset does not impact the model’s ability to replicate the statistical properties of the indices. The choice of reserve asset only impacts the initial cost of investing in the replicating portfolio (and hence only impacts the return of the replicating strategy). This is not to say that the return generated by the model is not important, however it is not a measure of the model’s success. One must dissociate the technical issues of the replicating methodology (i.e. how to best model the returns and solve for the optimal trading strategy) from the choice of the reserve asset. Our contribution is to provide a robust framework for the replication methodology, and address the technical shortcomings of the much publicized research of Kat and Palaro. It is a little surprising that the two aforementioned reports, who have clearly spent considerable time studying the Kat and Palaro (2005) approach, do not address the technical shortcomings of the proposed approach, and focus instead on non-model related issues.”
While he acknowledges that the mean return of a distributional replication depends on the mean return of its inputs (“Garbage in, garbage out”), he finds that the bar is actually set pretty low when it comes to beating hedge fund indices with distributional replications:
“As is the case with any investment strategy, the returns depend on the choice of assets. The results in this paper indicate, however, that it is not necessary to select the best performing assets over the sample period in order to replicate and outperform the hedge fund indices. In fact, we show that by using run-of-the-mill exposures in our reserve asset we can nonetheless outperform the majority of hedge fund indices. We purposely selected two reserve assets that have exposures to different yet common market premia over the sample period, and we find that both reserve assets outperform a large percentage of the indices. We also find that the EDHEC indices, which are subject to less significant biases, are more easily to replicate than the HFRI indices.”
This paper is quite technical even though Papageorgiou tells us he stuck most of the technical issues in the appendix in order to make it more accessible. But for any non-rocket-scientists out there, the introduction and conclusion make all the salient points.
While the paper does indeed propose specific improvements to the Kat-Palaro method, some might describe these as incremental, rather than dramatic. For his part, Harry Kat tells AllAboutAlpha.com that he sees the proposed changes to his procedure as “fairly trivial”.
Regardless, what’s probably more noteworthy is the simple fact that the distributional replication camp will soon be welcoming a second competitor. To date, Kat’s “Fund Creator” has been the only game in town (although it provides trading instructions and does not actually manage assets itself). Ironically, the entry of Desjardins into this space will likely be a good thing for Fund Creator since it goes a long way toward legitimizing a method that has so far been met with considerable skepticism from many corners.