Hedge Funds and Systemic Risk - HF Profile (II)

In installment II of our look at Hedge Funds and Systemic Risk, (remember, that none of this is my own work, but rather that of the referenced paper). All page numbers refer to the numbered pages. If you are looking at the PDF doc, the PDF page will be higher than the numbered page.

I wanted to provide a bit of a backdrop of hedge funds (HF's). You may not be a qualified investor and may never will be, so HF's funds may never be in your investment horizon. Nevertheless, they are in the same financial arena as you competing for returns. And the whole point of my writing about this is that if they screw up, it can cost you money.

Let's reflect a moment on the basic tenets of the paper.

• HF's have proliferated;
• They have a high attrition rates;
• Risk profiles (due to leverage and investment styles) are unavailable;
• Operations are not transparent; and
  
• Non-correlated dynamic strategies can correlate into phase locking during periods of market stress.

(I hope not annoyingly, but I'm purposefully repeating some of the core concepts of the paper to solidify them and make them relevant to the current installment without your having to keep going back.)


In their paper, the authors started with a population of 4,781 funds using information from 02/1977 - 08/2004 using the TASS data base. This database does NOT include all hedge funds, but rather is the data base that included all of the information they needed. As these guys appear to be professionals, we can trust that they designed their study to minimize underlying bias in the data. In fact, they go to some pains to explain that, but I'll not do it here.

They parsed this population into HF's that were Live as of 08/04, and those that were in the Graveyard as of 08/04. That split was 2920 and 1861, respectively. Due to the data vagaries, one cannot equate that with a failure rate. See attrition rates below. The authors had to go through and design there study in a way that required them to tweak these numbers so that there was appropriate homogeneity in terms of reported returns etc in the study population. Accordingly, the Live list was reduced and the Graveyard was increased.

Here is a graphic (p. 17) of the final study population by investment style. I would urge you to look at the Appendix A of (p. 84) the paper to read about these styles.





The authors note: "it is apparent from these figures that the representation of
investment styles is not evenly distributed, but is concentrated among
four categories: Long/Short Equity (1,415), Fund of Funds (952),
Managed Futures (511), and Event Driven (384). Together, these four
categories account for 71.9% of the funds in the Combined database."

There's another graphic that is telling. Below (Figure 1) are two pie charts--one showing the composition of the Live Funds and Graveyard funds. Click to make larger or refer to the original study on p. 19.

The authors note "databases are roughly comparable, with the exception of two categories: Funds of Funds (24% in the Live and 15% in the Graveyard database), and Managed Futures (7% in the Live and 18% in the Graveyard database). This reflects the current trend in the industry towards funds of funds, and the somewhat slower growth of managed futures funds."

I'm going to gloss over the section that talks about the correlation matrices. There are three points that are worth noting:

1. In general hedge fund index is not well correlated to the S&P 500 (p. 19)
2. Correlations among/between investment styles can vary widely (p. 23)
   
3. Events such as LTCM can increase correlation--in 1998 10 out of 13 syle-category indexes yielded negative returns. (p. 24)
   

Remember that non-correlation moving to correlation is a key part of systemic risk. To show just how pervasive and lingering affect an event such as LTCM can have on correlations, I'm going to include the author's table (CTML) here (oval and highlighting are my own):



The authors report some excellent detailed data regarding HF's using the TASS database. To give you a sense of the increase in HF's over the period, I want to share the authors' table with you. Three things to keep in mind while reviewing this table:

1. these are additions, only.
2. 2004 is reflective of the study cut off, so I would not use that
3. Funds exiting the database were not tracked.
   

Attrition Rates:

It's important to note that the database used (TASS) by the authors does not included all hedge funds--but rather it is a database that contained all of the information that they needed for their study. The point in considering failure rates is that a hedge fund failure could be one of those "events" like LTCM. In fact, as I wrote this and as I considered the blow up in sub-prime mortgage I wondered if that would pose problems for fixed income arbitrage hedge funds. I asked this of R. Nusbaum on his website.

Here's a screen capture of the ANNUAL average attrition rates in the TASS database:


The authors then go on to build a logit model to estimate liquidations based on the mix of investment styles. This analysis is beyond my ability to convey, but the takeaway is that their estimate is over 11% for the 2004 data base which is 25% higher than the average of 8.8% (p. 63). Here's some information from p 15 cited in their literature review.



I hope that you found this installment interesting. In this installment, I hope that I conveyed:

• The number of hedge funds and the predominant types of investment styles
  
• The increase of hedge funds since inception
• The attrition rates, and how those attrition rates can be affected by events (LTCM and tech bubble), as well as fund performance to include both funds and volatility.
  
• How events can affect the correlation of non-correlative styles (though admittedly, I gave just a small slice and did not produce any of the wonderful tables and work that the authors did. If you are statistically inclined, I hope that you will review their work.)