Lars Jaeger’s recent commentary on “alternative beta” (see posting) raised the ire of Professor Harry Kat (see posting). Today, Dr. Jaeger responds to Kat’s protests by highlighting the “inconsistencies” in his arguments.
Special to AllAboutAlpha.com by: Lars Jaeger, Ph.D., CFA, FRM, Partner, Partners Group
In his reply to my recent contribution at AllAboutAlpha.com, Harry Kat says that he agrees with “several points” in my argument (I only made a few key points anyway). Specifically, Kat seemed to agree that factor models capture mostly the “traditional beta” in hedge funds. Further, he seemed to agree with my argument that “hedge fund replication isn’t really about replicating hedge funds. It is about replicating hedge fund indices.” And he goes on to re-state many of my original points.
We seem to be in agreement on some very fundamental points. That is good news to me, as Harry has not always agreed with much what I have said in the past. However, he then suggests that there is a need to “fill in the picture” as my comments were only “part of the hedge fund replication story.”
I surely never claimed to know the entire hedge fund replication “story”. But what he actually provides us with – in order to “fill in the picture” – is merely a performance comparison between the ART Index (Goldman Sachs’ replication product) and the PG ABS Index (the Partners Group Alternative Beta Strategies).
In doing so, Harry is inconsistent in at least three ways. Firstly, he is inconsistent in the way he applies fees in his analysis. While he compares the PG ABS net of fees, he chooses to report the performance of the ART index gross of fees, a rather important difference as hedge fund investors surely understand.
Furthermore, he neglects to state that the performance of the ART Index from 1996 until 2007 is almost entirely backtested (with the exception of last year), while – as he correctly states â€“ the PG ABS Index is live and based on real trading. This amounts to an even more important inconsistency as most financial engineers know all too well (miraculously, backtests always seems to perform be better then real trading).
And finally, he takes the replication product that (by whatever coincidence) performed best last year – a decision which I shall refer to as the “Kat selection bias”.
Then, based on only one year of real data, Kat concludes that the ART Index kept up well with the HFR FoF Index over twelve years – a rather remarkable statement for an academic who knows about the concept of “statistical significance”.
Even more remarkable is his ultimate conclusion based on this small observation: that the “backward-looking nature of factor models might actually be less of a problem” (which sharply contrasts with his previous statements on the matter, by the way).
Plotting his case further based on this rather unbalanced comparison, he states that the PG ABS Index did not keep up with the FoF index. I note that the picture changes somewhat if we account correctly for fees. No doubt, in the last eight months the PG ABS Index has underperformed the FoF Index. In fact, I shed some light in my article on the recent performance deviation of alternative beta-based approaches in general. But as his graph shows, the PG ABS Index was very well able to keep up with the FoF Index and actually outperformed it prior to that period (a consistent accounting for fees makes this even clearer).
If I had included the PG ABS’s backtest data, the outperformance would have been even more significant. However, we are humble enough to admit that no backtesting is perfect. So we ultimately rely on the actual performance of our models to judge their quality.
In summary, it seems to me that Harry has joined the – by now rather crowded – club of believers in linear factor models for hedge fund replication (which is astonishing given his past statements â€“ some readers might also recall his previous emphasis on 12,000+ lines of C++ code required to run his own hedge fund replication approach). I continue to believe that this method works only when investors need it least to work. This approach assumes a constant equity exposure of hedge funds – which is not necessarily what investors want in the first place.
Maybe Harry should have been patient enough to read Einstein’s quote a little further: “Models should be as simple as possible, but not simpler than that“ (which actually did come from Einstein himself. He separated from Mileva before he had become famous for quotes like this). In that sense I cannot see how rigorous economic arguments can be taken “just one step too far”. We do indeed use a lot of the mathematics these days, and this has given modern finance its very shape. But I also contend that we should continue to believe in the soundness of economic arguments rather than on imperfect statistical analysis.
– Lars Jaeger, February 23, 2008
The opinions expressed in this guest posting are those of the author and not necessarily those of AllAboutAlpha.com.