The greatest source of market-driven hedge fund blow-ups is alpha, or manager skill risk. Leslie Rahl, Richard Horwitz and Erin Simpson of Capital Market Risk Advisors suggest this is due to an inadequate management of alpha risk
First, some definitions. Beta is the return driven by exposure to the underlying markets. Over a long period, markets reward investors for accepting the inherent risk.
Alpha is the additional return generated in excess of market betas. It is broadly generated by:Security selection Beta timing Complex, illiquid, opaque (CIO) premium
The initial two drivers are based on skill and ultimately represent a zero sum game. The last is rewarded by the market to compensate investors for taking the risk of holding more complex, less liquid and more opaque securities.
As with betas, CIO securities are not a zero sum game. These are not favoured by the typical, risk-averse investor. They therefore generally trade at a discount and generate a premium return. A significant amount of hedge fund alpha is driven by exposure to CIO risk.
Therefore, hedge funds, in general, go long more complex, less liquid and more opaque securities and hedge by shorting simpler and liquid common securities.
Although, hedge funds present themselves as alpha generators, this comes with a significant amount of beta. The correlation of hedge fund indices and the equity markets is 0.86. This implies that the majority of systematic hedge fund behaviour is driven by market betas.
However, the HFR Composite Index, since the autumn of 1998 demonstrates a worst month of 3.5% and the largest drawdown of 6.4%. This level of risk is inconsistent with the hedge-fund blow-ups that repeatedly occur.
The primary cause of the biggest market-driven hedge fund blow-ups that have occurred since 1994 (see box) was alpha risk. In all of the blow-ups, a primary source of the alpha risk was CIO risk.
In taking exposure to CIO risks, many hedge funds routinely earn a small premium in exchange for the risk of a low probability adverse event. These hedge funds are effectively selling insurance, alternatively, shorting volatility.
This is the cause of their ‘left-tail' tail behaviour, where the distribution of returns is a large number of relatively small gains intermixed with a small number of relatively large losses. Since the gains happen significantly more frequently than the losses, it can take some time, often years, for an investor to see evidence of this extreme left-tail tendency.
As the alpha of these CIO exposures is generally very small, leverage is often used to amplify them to generate satisfactory returns. This leads to a potentially dangerous combination of illiquidity and leverage and the cause of most of the biggest market-driven blow-ups.
Valuation risk is highly inter-related with alpha risk. The fundamental problem is that valuation of CIO securities is difficult and imprecise, as they are, by definition complex, liquid and opaque.
Therefore many blow-ups, where the root cause was alpha risk, have resulted in subsequent valuation issues, covering up the initial losses.
So how well do hedge funds manage alpha risk? The answer is, not very well. The fundamental problem is that risk management techniques have been developed by the sell-side to focus on structural risk exposures or betas.
The fundamental approach is statistical risk management. Statistical risk management applies historical return behaviour to the portfolio to determine what the potential return of today's construction would have been over recent history.
The Basel Accord mandates that banks and broker/dealers use value at risk (VaR) to calculate their capital requirements to cover risks undertaken. VaR is excellent at permitting an institution to aggregate risk across asset classes.
The vast majority of risk of the securities issued by these institutions is beta (including interest rate duration). Financial institutions, in general, do not have significant alpha risk exposures (excepting recent losses resulting from sub-prime mortgage holdings). The primary objective of applying VaR is to aggregate the risk of the portfolio of outstanding positions and to permit the institution to hedge residual risk. Therefore the methodology is primarily focused on beta risk management.
The problem of applying VaR methodology to hedge funds is that statistical risk management is backward looking. Were history to repeat itself statistical risk management would be adequate. However almost every financial crisis has resulted from some market dislocation or a discontinuity of historical relationships.
While some crisis has occurred almost every year in recent decades, each specific event is idiosyncratic. Shocking a portfolio based on prior extreme events is unlikely to be informative about future discontinuity. Therefore, statistical risk management does not handle the alpha, or left-tail event risk to which hedge funds are exposed.
VaR has not proved to be a effective hedge fund strategy. It works well in global macro (directional traders that are similar to the prop desks of banks). Yet equity strategies (long/short, market neutral, activist) have not embraced VaR, as they generally develop their own spreadsheets to identify the style/industry alpha exposures.
Credit (distressed, high yield, structured, convertible arb) and event (capital structure arbitrage, merger arbitrage) strategies also do not use VaR because many of their holdings are relatively illiquid and because the idiosyncratic risks of the lower quality credits in which they participate are similair to equities.
Given that statistical risk management does not work for many of these strategies, how does one manage alpha risk? The simple answer is diversification. As alpha risk is by definition idiosyncratic, it cannot be predicted. Investors are seductively drawn to CIO exposures, as large losses are generally preceded by consistent gains. Therefore you cannot inoculate yourself from it, as it is inextricably linked to the alpha generated by hedge funds.
Instead you can diversify your portfolio and keep your concentration to each specific exposure to an acceptable level. Furthermore, while one can easily hedge beta risks, alpha risks cannot be hedged.
Diversification is the primary lever of risk management in hedge funds. Risk is neither good nor bad, as long as it has a prospect of being well rewarded. However, a concentrated exposure to risk is bad. This is consistent with the concept that there is no risk that you should be unwilling to accept to some level of exposure as long as the risk is attractively enough compensated.
If the way to manage alpha risk is through diversification, how should diversification be managed? The key to managing the diversification of alpha risk is to understand all of the basis exposures one is taking. There are many sources of basis risk: instrument (for example, cash versus derivative), asset class (for example, equity, interest rate, commodities), position type, liquidity group, geography, industry, market cap, P/E group, maturity, credit quality, debt seniority and capital structure, structural versus synthetic versus primary.
The market disruptions that caused the biggest blow-ups were historically unprecedented. One must fundamentally understand a fund's basis exposures and be comfortable that the fund could tolerate a significant discontinuity in each. Furthermore, an investor in multiple funds must be able to aggregate these exposures across his or her entire portfolio to understand the aggregate exposure to thee alpha risks.
Finally, as many of these exposures are driven by CIO premia and there is always a risk of a broad flight to quality as occurred in the autumn of 1998, investors must similarly feel comfortable that they could tolerate a general market flight from CIO securities.
A good proxy for CIO risk is bid-offer spreads. In general, the more complex, illiquid and opaque a security is the larger the bid-offer spread. A systematic way to test for potential flight to quality risk is to stress the bid-offer spreads. While doing so does not provide an exact measure of an individual portfolio, consistently doing so across portfolios provides a very relevant measure of exposure to CIO risk.
This is an edited version of an article that first appeared in the AIMA Journal, winter 2007 issue, and is reprinted with permission