Hedge Fund Risk Disclosure and Transparency
Written Comments
by
David A. Hsieh, Duke University
I have been analyzing hedge fund strategies and risk since 1994. All of this work has been done jointly with my co-author, Dr. William Fung (Research Professor, Centre for Hedge Fund Education and Research, London Business School). The goal of this research is to try to help investors, as well as regulators, to understand the risk in hedge funds. We have published more than ten articles on hedge funds, listed on my home page at http://people.duke.edu/~dah7/vitae.htm.
Let me start by giving you the vision we have of this project—where we want to end up, and then a quick assessment of how far we have come.
A Model of Hedge Fund Risk Factors
Our vision is to create a risk model for hedge funds, similar to those for equities, such as the Capital Asset Pricing Model (CAPM) and the Arbitrage Pricing Theory (APT). In these well-known equity models, the return of each stock has two parts—the systematic and idiosyncratic components. The systematic component is the part of a stock's return related to risk factors that are common to many securities. In the Capital Asset Pricing Model, there is only one common risk factor—the market portfolio. The Arbitrage Pricing Theory allows for more than one common risk factor; empirically researchers have found 5 or fewer. The idiosyncratic component is the part of a stock's return that is unrelated to the common risk factors. This type of decomposition is very useful for investors. By diversifying over a large portfolio of stocks, an investor need not be concerned with the idiosyncratic risk of any one stock. The investor need only to be concerned with the common risk factors in constructing portfolios and managing their risk.
This type of risk modeling has been successfully applied to mutual funds. In an article published in the Journal of Portfolio Management in 1992, Professor William Sharpe showed how to model the risk of equity mutual funds using a set of common equity risk factors.
Our research goal is to do something similar for hedge funds. We want to find a set of common risk factors for hedge funds. If we know the amount of exposure of each hedge fund to this set of common risk factors, investors can understand the risk of a diversified portfolio of hedge funds, and how that hedge fund portfolio would perform with the rest of the investors’ portfolios.
Before going further, let me emphasize what this research project is NOT designed to do. We are not trying to understand the risk of any specific hedge fund, just as the CAPM and APT are not designed to model the risk of a specific stock. But, from the investor’s portfolio standpoint, the common sources of risk in hedge funds are the important risks, not the risk of a single hedge fund. Similarly, from a regulator’s standpoint, it is the amount of trading in common strategies that can be destabilizing systemically, not necessarily the strategy of a specific hedge fund or a group of hedge funds.
Next, let me turn to where we are in this project.
Extracting Common Return Components in Hedge Funds
First, we identify the common return components in hedge funds. We have done this in two ways. In an article in 1997 (published in the Review of Financial Studies), we analyzed all hedge funds and commodity funds with more than two years of monthly performance data. We used the idea that, if two funds traded similar assets in a similar manner, their returns would be highly correlated. The correlated part of their return is therefore a common return component. By grouping funds with correlated returns, we can extract their common component. In that article, we found that the five most important common components accounted for roughly 50% of the common comovement among these funds. This method is very similar to the way the CAPM and APT models were empirically implemented.
Alternatively, we have also taken subgroups of hedge funds, based on classification schemes of hedge fund data vendors. Again, we extract the common return component based on the return correlation of the funds in these subgroups.
Linking Common Return Components in Hedge Funds to Observable Market Risk Factors
Second, we relate the common return components in hedge funds to observable market risk factors, such as option prices and credit spreads. Let me give several examples of how we do this for a few subgroups of hedge funds with different trading styles.
Trend-following funds: The first example is the subgroup called trend-following funds. In an article in 2001 (also published in the Review of Financial Studies), we extracted a common return component from trend-following funds. We then linked this common return component to the observable prices of options. The model uses the following idea. Trend followers are betting on big moves. Therefore they would make money when markets are volatile, which would also be the times when option buyers would do well. In our research, we verified this theory by showing that the return of the average trend-following fund is strongly correlated to portfolios of exchange-traded options. This demonstrates that the common return components in trend-following funds can be linked to observable option prices. Subsequently, in an article in 2002 (published in the Financial Analyst Journal), we showed that this analysis continues to hold beyond the sample period of our analysis.
Fixed-income funds: The second example is fixed-income hedge funds. In an article in 2002 (published in the Journal of Fixed Income), we found that fixed-income funds are typically exposed to interest rate spreads. The reason is that many fixed-income funds buy bonds that typically have lower credit rating and/or less liquidity, and hedge the interest rate risk by shorting treasuries that have the highest credit rating and more liquidity. The difference between the yields on the two bonds is the interest rate spread. Since interest rate spreads tend to move together, especially during times of market distress, fixed-income funds can be modeled as being exposed to credit spreads.
Equity Long/Short funds: The third example is equity long/short hedge funds. In an article we are still working on, we show that equity funds tend to be slightly long the stock market (with a “beta” around 0.3), and long Small Cap/short Large Cap stocks. This is consistent with the observation that equity hedge funds tend to buy lower capitalization stocks with less market liquidity, and hedge by shorting larger capitalization stocks with more market liquidity.
Assessment of the Hedge Fund Risk-Factor Model to Date
Let me give you an quick assessment: How much of the risk of an average hedge fund portfolio we can explain, using the risk factors we have identified to date? In a working paper which will soon be available on my website, we use the average fund-of-hedge funds to proxy a typical portfolio of hedge funds. With the risk factors we have identified so far, we can explain roughly 70% of its return variation. When we analyze more hedge fund styles, we may uncover other risk factors to improve this result.
Uses of the Hedge Fund Risk-Factor Model
The type of hedge fund risk-factor model in our vision can provide a consistent framework for asset allocation, portfolio construction, manager selection, and risk management for investors.
Asset Allocation:
From an asset allocation standpoint, investors can determine which
hedge fund risk factor, and how much of it, to have in their portfolios.
Hedge Fund "Betas":
The exposure of a hedge fund to a given risk factor is quantified by
"beta". For example, in our JFI 2002 article, we estimated that the
beta of the average fixed-income arbitrage hedge fund is -5.37 with respect
to the credit spread, as measured by Moody's Baa yield minus the 10 year
constant-maturity treasury yield. That means an increase in the credit
spread by 1% will lead to a decline in return of 5.37% in these types of
funds. This is a useful piece of information. The reason is that
hedge fund history is rather short. We have reliable data only beginning
in the 1990s. This happened to be a relative benign period for the credit
spread: it had generally declined, and remained low relative to the 1970s
or 1930s. As a result, it is not possible to answer the question: "How
would fixed-income arbitrage hedge funds perform in the credit market conditions
of the 1970s?" based on their returns alone, as they did not experience
that type of market environment in our data sample. However, the beta of
these funds with respect to the credit spread will give us a good idea
of what might happen.
Armed with each hedge fund's betas with respect to the set of hedge fund risk factors, investors can construct hedge fund portfolios that will have the desired exposure to these risk factors.
Hedge Fund "Alphas":
The average return of the idiosyncratic part of a hedge fund's return
is its "alpha". It represents the average additional return the fund manager
generates, above and beyond exposure to the set of hedge fund risk factors.
This can arise from a manager's "skill", or exposure to some risk
unrelated to the common set of risk factors. Investors can use this
information in manager selection.
Implication for Hedge Fund Risk Disclosure and Transparency
What does this say about hedge fund risk disclosure and transparency? At a minimum, it would be useful to have individual fund exposures to a set of common market risk factors. That would help investors to better design hedge fund portfolios, as well as manage their risk.
Regulatory Implications
While this roundtable is mainly focused on providing information for investors, I would like to touch base on how a hedge fund risk-factor model can help regulators. In an article published in the Journal of Empirical Finance in 2000, we estimated the market impact of hedge funds in various periods of market stress. Let me use one episode to illustrate the benefits of using a hedge fund risk-factor model.
The Bond Market Decline of 1994
During 1993, the world bond markets rallied as interest rates declined.
Then in February 1994, the Federal Reserve began to raise interest rates.
This led to one of the worst bear markets in the history of the bond market.
Our article used monthly returns of various subgroups of hedge funds to
estimate their position sizes. We found that trend-following funds
and "Global/Macro" hedge funds have large long positions on European bonds
at the beginning of 1994. Trend followers were long European bonds because
they saw a prolonged increase ("trend") in their prices. Global/Macro funds
were long European bonds because they were betting on a continued decline
in European interest rates based on economic fundamentals. It turns out
that "convergence traders" were also heavily involved in the European bond
market. A study conducted by the International Monetary Fund (1998) found
that convergence traders were betting on the convergence of European interest
rates to each other, in anticipation of the launch of the single European
currency in 1999. They were mainly long Spanish and Italian bonds,
betting their yields would decline to the lower levels of German bonds.
When interest rates increased, there was a prolonged period of liquidation,
first by trend followers and convergence traders, then by Global/Macro
funds.
This episode indicates that different groups of hedge funds can "converge"
on the same trade based on completely different motivations. Presumably,
other "highly levered institutions" (a phrase coined by the Bank for International
Settlements), such as proprietary trading desks of investment banks, also
had large positions in European bonds. It is this type of situation that
is the most dangerous from a systemic standpoint. Markets become stressed,
as the bond market did in 1994, when everyone rushed for the exit.
Prices declined, triggering margin calls, forced liquidations, and further
price declines. From a regulatory standpoint, it would be useful to obtain
risk exposures of all "highly levered institutions", not only hedge funds.
In my presentation at the New York Fed in October 1998, I suggested that
it would be useful to report aggregate exposures to various risk factors.
This may forestall traders from adding on more positions, if they sense
that it may be difficult to exit from them.
Copyright (c) 2003, by William Fung and David A. Hsieh.