October 2005 saw a significant negative performance over a wide range of alternative strategies, leading some to call it "Red"October. If you believe the hype about pure and portable alpha, the lock step under-performance of such wide ranging strategies must seem like the perfect storm. But should we really be surprised by such results? Could this really be a "once ina lifetime" random event, or does it simply add to the growing body of evidence which indicates that many hedge fund strategies are indeed exposed to similar, albeit more exotic, sources of systematic risk, which are not well understood or controlled by the mainstream alternative industry?

Our research shows that a robust and well designed risk estimation process could have anticipated a significant portion of this downside event.

In October 2005, we observed the following returns:

October 2005 Performance – HFRX Global and Strategy Indices

One way to try and understand what happened during "Red" October would be to run an in-sample performance analysis to explain the underlying reasons for such losses across the board. But from the investor's perspective, it isn't sufficient to explain such events after the fact. Rather than locking the stable door once the horse has already bolted, the fundamental question investors should be asking is "should Red October really have come as a surprise?"

Moving beyond such ex-post analysis to a more forward-looking approach introduces the key question: can we anticipate which market scenarios are likely to impact our investment performance? By answering such questions and analysing how frequently such events have occurred in the past, we can achieve our ultimate goal: to take mitigating action, rebalance the portfolio of hedge funds and select new managers in which to invest. Most importantly, we can select new managers whose strategies systematically diversify unintended or unacceptable sources of risk. Such an approach debunks the myth of purealpha, and allows the investor to actively manage exposure to market factors in line with their own business objectives and expectations. In setting out to achieve this, it is important to understand the differences between Performance Attribution and Risk Management:

Performance Attribution is a backward-looking process, which seeks to explain the sources of past performance by attributing across strategies, currencies, managers and timing effects. It seeks to identify the contribution of each investment decision or market factor exposure to past returns.

Risk Management is a forward-looking process, which seeks to explain how each investment decision contributes to overall portfolio risk. It addresses the crucial questions:

• What are the market forces that are likely to drive the volatility of future returns?
• What market conditions are likely to penalise my portfolio's future returns and how often can I expect such conditions to occur?

Thus what 'Red October' can teach us is that to have an ex-post explanation is useful but insufficient. A robust risk process, however, will help the investor to anticipate such market situations before they happen; on an out of sample basis.

The principle of out of sample back-testing is to observe historical market factor shifts over a historical time window. For each window we use the observed history to calibrate our risk modelling process and to estimate future behaviour. We can then test the predictive power of our model by comparing the predicted behaviour to the actual fund return. This process is visualised in the chart opposite.

Out of sample back-testing principle means:

• Run the system on given past windows
• For each window, observe the move of factors on the NEXT 1M time step. Apply system's proxy to estimate the fund return.
• Compare to actual fund return

This 'spider graph' opposite shows the strength of the relationship between the HFR FoF index and each market source of risk, scored between 0 (no relation) and 1 (full exposure). In each market segment, the factor with the highest score is displayed: Cap Size Small Caps, Equity EMA, Style Value. Riskdata's analysis utilized HFRI Fund of Funds Diversified Index, which is published by Hedge Fund Research, Inc.

We applied this out of sample back-test methodology on a range of FOF indices up to the end of September 2005. This means we took all the available information as of the end of September, and from it, derived a directional non-linear factor model for these funds. Taking the HFR FoF Diversified Index, we found the following forward-looking risk profile. The three top factors which appear to be driving FoF risk in extreme market conditions are Equity Small Cap, Value Equity Stock and Emerging Market Asia. In normal, "business as usual" market conditions, the top three factors were Equity North America (i.e. including Canada), Value Equity Stock and Equity Small Cap. For all of these factors, the HFR FoF index exhibits a long exposure with a negative skew.
 

Riskdata's analysis utilized HFRI Fund of Funds Diversified Index, which is published by Hedge Fund Research, Inc. Note: The short exposure to the spread between large and small caps shows the long exposure to small caps, knowing the index is basically long the US equity market.

Thus, through a type of analysis known a "Bayesian Priors", and using only the information which was available at the end of September 2005, we can derive a factor profile which could predict which market forces are most likely to drive future performance. From this factor profile we then derived the stress test report above.

In order to validate our findings we then applied this forward-looking model to "Red" October in an out of sample back-test. By taking the model forecasts and simply inputting the actual factor movements observed in October on all significant factors, we achieve the following forecast behaviour:
 

The actual return for the HFR FOF index for October was a loss of 1.4%. Therefore, we observe that over the full month, our model predicted one third of the actual negative performance of the index. Over the first 22 days of October however, the model predicted two thirds of the actual loss during this period. We can therefore conclude that the "out of sample" test is quite conclusive, with the best match achieved by considering only the first 22 days of October.

In order to interpret the results we must observe how the underlying market factors actually behaved in October. We can see that all of the underlying factors (Small Cap, Value Stock, North America, Emerging Market Asia) exhibited a similar pattern; a sharp drop early in the month, followed by a small rebound.

One possible interpretation of these "out of the sample" tests is that this phenomenon was the result of a combination of a concentration of exposures plus a concentration of behaviour. The concentration of exposures means that many investors had either direct or indirect exposure to a concentrated group of "trendy" market factors such as small cap, Asia, and energy. Concentration of behaviour means that a number of managers, perhaps hoping to secure their returns for the year, liquidated their positions during the first part of the month (i.e. the drop…), resulting in widespread adverse market timing.

Thus, one important lesson to recognize is that the dynamic behaviour of hedge funds must be integrated into the risk analysis. Investors inhedge funds are not investing in static portfolios of assets, and any static snapshot approach will be insufficient to fully capture the behavioural dynamics of the manager. What happened in October was not just the result of the positions hedge funds held in their portfolios at the beginning of the month, but also of the way they changed their portfolios during the month.

So should investors press the eject button and redeem large portions of their alternative investments because of Red October? Clearly if we look at the performance for October, compared to the performance of the traditional asset management industry, the idea seems ludicrous. The level of loss observed during Red October still resulted in only half the total volatility seen in long only equity funds. Institutional investors understand that their long term investment horizons tend to mitigate the effects of short term volatility and, at least in traditional long-only asset management, have repeatedly demonstrated their ability to endure much higher and more sustained levels of loss.

And yet it remains clear that traditional asset portfolios are distinct in this regard, specifically because the investor is not surprised by negative performance when such losses are attributable to downward movements in the fund benchmark, which are expected from time to time. It is the "pure alpha" claims of the alternative industry itself which are responsible for the unease felt by many institutional investors. If "pure alpha" does exist, then how could so many of its standard bearers have delivered losses at precisely the same time?

If, on the other hand, alternative providers had clearly articulated their often asymmetricaland non-linear, yet highly systematic exposures to Asia, small cap and value, their institutional investors would not have been surprised by Red October. By presenting analysis such as that shown above, alternative investment professionals can help their clients understandthe market risks to which they are systematically exposed and, in so doing, dramatically increase the investor's comfort level and prevent redemptions. Thus risk transparency is not just a good way to attract institutional investment, it is a critical part of managing such investors' expectations and retaining their investments through periods of negative performance.

Riskdata is a leading provider of risk management solutions for asset managers, including funds of hedge funds and hedge funds