The dramatic growth in hedge fund assets under management, largely driven by institutional investors, has been supported by active marketing to differentiate hedge funds from their long-only cousins.
Traditional managers market their funds by describing them in reference to a benchmark. The beta versus this benchmark defines the level of defensiveness or aggressiveness and outperformance against the benchmark characterises the value added by the manager. In such an MPT paradigm, fund return = beta x benchmark return + alpha + specific risk.
Many institutional investors and consultants are attracted to hedge funds because they are uncorrelated with the market, thus assuming (a) there is no benchmark or beta = 0, and (b) that they are, therefore, pure 'alpha generators' meaning: fund return = alpha + specific risk.
This method of promoting hedge funds raises a number of interesting questions.
Firstly, anecdotal evidence does not support such an assertion, since one can relatively easily mimic such 'alpha generation'.
One example is to use a classical 'short put out of the money'. In such a case, the fund return equals the option theta plus a low probability risk of being in the money with associate high losses.
Other dynamic strategies can replicate a short put pattern without using derivatives.
For example, a manager could apply mean-reversion approaches based on systematic trading rules; such as doubling exposures in the event of losses and reducing them in the event of gains.
These strategies don't eliminate market risk; just shift the exposure to the tails of the distribution.
Other issues, such as illiquidity and thinly traded markets, give rise to practices such as return-smoothing.
Under such circumstances, time lag can be introduced, causing a decoupling of market events from recognition of returns. Consequently, more advanced statistical measures, such as co-integration, must be used to help identify systematic risks.
The most compelling evidence contradicting the 'pure alpha' bandwagon was observed throughout 2004 and 2005, during which period, most hedge funds pursuing 'market-neutral' strategies underperformed and then outperformed in lock step. Sceptics or statisticians would find it hard to believe such events were due to pure chance.
Riskdata's research indicates many of these 'market-neutral' strategies share a systematic exposure to the levels of volatility and correlation in their respective markets; underperforming when volatility is low and markets are highly correlated and outperforming when volatility is higher and markets de-correlated.
A more sophisticated analysis shows that hedge funds do have a proportion of systematic risk, but that the systematic factors necessary to explain hedge fund risk are unique to this asset class.
Riskdata recently presented out-of-sample back-test results, which showed that 75% of significant loses can be explained on an 'ex-ante basis' by systematic factor exposures.
The key challenge is that the factors necessary to explain systematic hedge fund risk require us to search beyond the traditional benchmark-relative view; beyond assumptions of linearity and especially beyond Gaussian 'normal market' conditions and the sole volatility/correlation analysis.
It becomes increasingly clear that the traditional beta/alpha approach is not appropriate for hedge funds.
Standard benchmarks do not make sense precisely because hedge funds do not belong to traditional asset classes. Thus, to assert that they have no beta relative to a traditional benchmark is irrelevant at best, and highly misleading at worst.
Using hedge fund indices or 'blind-factor models' does not address the fundamental needs of institutional investors. It does not help them to understand the systematic drivers of performance or the exposure to risk in extreme market events. Institutional investors need some way to distinguish between luck and skill within the hedge fund arena. As such, they must identify systematic performance drivers to determine if outperformance was by chance or due to skilful detection of mispricing opportunities.
The questions institutional investors should really seek answers to are:
What is likely to happen to my fund under extreme market conditions (such as the S&P 500 falling by by 25% or credit spreads widening by 100bps)?
What is likely to endanger my investment?
What would be the impact of another LTCM crisis on my fund?
Can unwanted or unacceptable levels of extreme market risk be diversified through a combination of managers?
Risk transparency on the factors that impact fund returns systematically is undoubtedly the best way to address such investor demand. Yet risk transparency is achievable without the empty promises of full positional transparency, which poisoned chalice of investors offers a nightmare of data overload and negligence claims in the event of severe losses.
It does offer some type of risk warning - provided positions are given at least daily and in a timely manner. However, because of lock-up periods, the investor is in a similar position to a plane passenger being warned that the aircraft is going to crash, with no real possibility of action.
A non-linear factor-based approach, involving a powerful factor-selection tool, allows investors to see which market scenarios can impact the fund's return; thus enabling them to assess objectively the extent to which the fund provides diversification benefits or excess marginal return. It enables investors to rationalise periods of underperformance caused by adverse market conditions and thus to help differentiate between luck and skill. Most importantly, it isolates factor sensitivities in the tails of the distribution, allowing investors to predict more accurately how managers or combinations thereof are likely to behave in extreme market events.
We can see the concept of "pure alpha" is about as useful as a chocolate tea pot. It perpetuates the illusion that hedge fund returns are independent of systematic market forces when we know those relationships can be controlled. We need only look beyond the traditional paradigm for the less obvious drivers and use more advanced techniques to understand the relationships.
The week on Risk.net, August 4–10Receive this by email