Back to all announcements

Riskdata research establishes that treating hedge funds as a distinct asset class dramatically improves quality of risk prediction

Research by Riskdata's quantitative team was unveiled by Raphaël Douady, Riskdata's Director of Research, in a speech to the Hedge Fund Analytics Conference in New York last week. The aim of the research was to quantify the benefits of Risk Profiling, a breakthrough risk management technique which treats hedge funds as a distinct asset class, capturing their dynamic and accounting for their statistical specificities. The results show that the predictive power of Risk Profiling is above 50% - thus reducing by more than half the uncertainty on future fund returns - compared to a zero or even negative predictive power when using traditional factor analysis.

Raphaël Douady commented: 'Our study is a quantitative confirmation of what all professional investors in hedge funds know: they must be treated as "non linear" stock, ie as a distinct asset class. Unlike equity stocks, a hedge fund does not react in a linear way to market shocks - it can absorb negative shocks and amplify positive shocks, or the reverse - and its behaviour will change in extreme market conditions. On the other hand, as with an equity stock, the interaction between the manager and the market are rather systematic and can therefore be predicted, simply because they are implementing systematic strategies, credos and guidelines. As with equity, this systematic dynamic pattern is critical in predicting the risk. Nobody would imagine that predicting the risk of Microsoft can be done by considering only a snap-shot view of its balance sheet: capturing the systematic patterns behind its past performance is much more relevant. It's not different for hedge funds.'

- To evaluate the efficiency of this approach, Riskdata's team consistently back test the results, both globally on a set of more than 1,000 hedge funds and fund of funds, and in depth on a representative sub sample of 240 funds. The back testing is run and compared with traditional methods, such as stepwise linear regression. The current results of the study on the test panel subset, analysing the period from January 1999 to December 2003, are as follows. The average explanatory power of risk profiling (the % of explained risk by the selected factors) is in a range of 55% to 75%, in line with the stepwise regression.
- Unsurprisingly, the highest explanatory power is for directional strategies (80%), with the lowest results being for fund of funds (50%), due to the highest alpha, which results from diversification.
- The main difference, however, is in the average predictive power when factors are selected (predicted risk versus fund risk). For Risk Profiling it is above 50%, meaning that risk profiling can reduce by more than 50% the uncertainty on the fund returns, while it is nil or even negative for the stepwise replication (due to the spurious correlation effect).

The main reasons explaining the efficiency of Risk Profiling compared to the traditional regression methods are as follows:
- Risk Profiling captures the asymmetrical patterns in hedge fund distribution: a significant proportion of hedge funds have asymmetrical exposures, with for example returns that are boosted when the market goes up and when the market goes down, or on the contrary have a profile whereby returns are boosted when market is stable. This is not only the case for hedge funds using derivatives: a significant proportion of traditional long-short equity managers demonstrate this asymmetry in their returns. Capturing this asymmetry is critical for making good predictions.
- Risk Profiling can distinguish between normal market moves and extreme market moves: the pattern of a significant proportion of hedge funds changes in times of crisis.

The test panel of 240 hedge funds included, among others, 75 directional funds, 64 non-directional funds, 32 arbitrage funds, and 24 special/event funds.

The Risk Profiling technique was applied on a set of 200 factors organised in 12 types: nine market movement types (geographic, sector, style, market capitalisation indices, interest rate, curve slope, credit, currencies, commodities), together with three market dynamic types (market volatility, correlation within the markets and their level of convergence).