Deciphering Global Hedge Fund Statistics

"The secret of staying young is to live honestly, eat slowly and to lie about your age."

Lucille Ball
Actress (1911-1989)

One of the biggest problems in the hedge fund space is data mining. Aside from the crazy assumption that historical data and returns can extrapolate into the future, many users of commerically available databases somehow think that they have discovered or uncovered rare secrets from within a mystical hedge fund black box. This is hubris.

Take the latest example unleashed on the world by Pertrac. There are a vast litany of negative concerns related to hedge fund data including self-selection bias, sample selection bias, survivorship bias , backfill bias, and infrequent pricing bias.

Added to these, there are additional very serious classification issues that exist when you have two or more totally different databases with their own labels and you try to fit them together. For instance, where does one put an ABS fund in the HFR database, MSCI database and the Lipper TASS Hedge Fund Database? Is it a fixed income arbitrage fund, a multi-strategy fund or a relative value fund? No-one really knows. But where you put it will certainly impact the data results of risk and return across a number of strategies.

Perhaps the biggest "problem" leveled at databases is that of self-selection bias. While critics point out that only the better funds tend to provide their data to the outside world a reverse argument (and equally legitimate one) is that the best funds do not report to the same databases and so perhaps there are under-reporting biases that should be considered.

Of course, you don't hear much about these funds whenever a news releases comes out!

Which brings me to the case in point. Pertrac, recently analyzed hedge fund data reported to a large number of databases which are consolidated in their analytics package. They claimed that hedge funds report their best performance in their first 2 years. They claimed that a track record of less than two years produced, on average, 11.7% while performers over 2 years produced 10.2% return - a difference of 1.5% on average annual performance.

What the study did not point out was quite a lot. Namely what about the risk-adjusted performance? What about the fact that the smaller funds probably had significantly lower AUM and so in simple numerical terms starting from a lower base their performance was over emphasized. And what about all those closed big funds out there which are not in the databases? The Kensington Global and Medallian Funds of the world...what happens if you include this data into the results?

In fact in an alternative view of the data world a recent comment at a forum in Hawai'i by a academic pointed out just how much of a gap in our understanding of these big closed funds might be out there.

In the presentation covering 20 of these large single manager funds with AUM of over US$6.4 billion there was a definitive performance advantage favoring larger funds. How much? Apparently, an average of over 2.50% per year and for the top performing single managers (top quartile) the number was a staggering 6.00% outperformance per year since 2003!

The reference point for these returns was over the Credit Suisse-Tremont Hedge Fund broad hedge fund index, and, if true, it points to a very significant advantage of big over medium and presumably small too.

Data mining too? Could be. But it might at least shed some light on why the biggest institutional investors in the world are willing to pay higher fees and take on longer lock-ups in order to get access to such funds. It might also explain once and for all the key competitive advantage of the best performing fund of hedge funds - real estate- or simply being invested in the right funds.

And if one wants to focus on age then the average age of those best performing big funds was 13 years. So maybe as Lucille Ball and Marlene Dietrich would have said size and experience matters too. Mahalo.