I remember when Harry Kat was getting a lot of play in 2006/7 for his idea you can replicate various hedge funds via futures. I was skeptical because while any fund with 5 years of monthly returns will generate a nice in-sample fit with something, such data are not really 60 independent observations because their will be a handful of big bets that were correlated with some sector. In my experience working at hedge funds you could point to a few key strategic moves that explained most of the returns, and as people changed, or the strategy died, it was hardly an internal factor loading, just a coincidence. I got a couple calls from industry magazines doing pieces on Kat, and both declined to incorporate any of my criticisms.
I haven't heard much about it since. Some algorithms seemed to do very poorly in 2008 with most strategies, but presumably Kat says not his. This highlight a major problem in finance, that so many fund managers are encouraged to keep their records private, because then they can use the fawning press to imply huge returns while never actually saying so, and so everyone else seems to be beating the market (just as most people who go to Las Vegas say they were either 'flat' or 'up big'). So how have they done? In Antti Ilmanen's Expected Returns, he has a couple cryptic pages (238-9), where he states 'the in-sample fit to HF index and HF sector index returns can often be surprisingly good, but out-of-sample results less so.'
I can say from my knowledge of BNP and the old group that spun out of BSC to do HF replication that they have failed miserably. There has been a consistent performance lag to invest-able (SMA platform) benchmarks which have in turn had an additional lag to non-invest-able quoted HF benchmarks.
I'm sure there is some spin available about how they've done exactly what they were constructed to do but the anemic growth of AUM for any replicator speaks for itself.
As a side note, in a prior incarnation I was at a fund of funds that was pitched by a french group with a replication model that they were looking to sell for $300,000 a year as a risk management tool (i.e. it would tell you what the real risks of your portfolio of fairly opaque managers were). They finally rolled out a demo for us (after we provided return streams for our underlying funds sans-names) in our offices and promptly told us that our FIG-dedicated fund was most correlated to a negative bet on agricultural commodities and our high yield credit guy was most correlated to a long volatility factor. It did make more sense in some areas, it could identify our short dedicated/biased guys as being essentially short. I don't think that was much of an insight for us. I think you're spot on that the "factors" driving most HF returns and risk aren't stable or fundamentally consistent enough to be modeled and replicated reliably.
What happened to the hedge fund industry? During my time as a hedge fund analyst (Sep 2007 - Dec 2008) most managers where just long a single risk factor (nat resources, EM, FX carry, HY, ...). So many institutional investors rightly thought: Why should I pay 2-20 for exposure I could replicate (directly, not via a replication product) pretty cheaply?
And currently investable HF indices (e.g. HFRXGL) which probably do not include HFoF fees look like a deleverd version of the SPX. So what are the arguments in favour of hedgies without relying too much on picking skills?
Harry Kat... in one article he actually said he can replicate the return distribution of hedge funds pretty well except for the mean! Yeah, and I can replicate the payoff of lottery ticket except for the case where you got the right numbers!
Amazing the idea of hedge fund clones ever saw any traction. Back in 2006 when replication was getting attention, the counterargument then and now won.
Post a Comment