Managing risk more dynamically is becoming an increasingly vital function of the investment management process. Those that can demonstrate their effectiveness at reacting quickly to volatility are held in high regard. Performance is still vital, that’s why investors pay inordinate fees.
However, risk-adjusted returns can improve over time by managers taking a multi-dimensional approach to managing risk as opposed to viewing it merely as a support function.
At least that’s what Dr Matthieu Vaissie (pictured), a research associate at EDHEC-Risk Institute and senior portfolio manager at Lyxor Asset Management and co-author Serge Darolles have found in their latest research paper: The Benefits of Dynamic Risk Management: mitigating downside risk without compromising long-term growth prospects.
The paper contains two central ideas. Firstly, that in today’s complex, fast-moving environment risk management can actually become a source of added value provided relevant information is used on an ongoing basis. Secondly, that managers can use this data to construct portfolios that over time will remain well suited to the environment in which they operate.
“Risk management can be a source of added value if you properly integrate information continuously and construct your portfolio accordingly. What I mean by added value is reducing downside risk without impacting the long-term growth prospects of the portfolio. It’s essential today to think about risk management as a way to improve the risk-adjusted performance of the portfolio, not just a way to reduce risk,” Vaissie tells Hedgeweek from Paris via phone interview.
Limiting downside risk is a priority for hedge funders. Often, though, they find themselves walking a tightrope: how aggressive should they control risk? How will it impact on overall returns? Knowing when to put the brakes on and how hard is an inexact science, but Vaissie’s research at least shows that evaluating risk on an ongoing basis, and using information to modify the portfolio, can yield significant results. This is especially helpful for FoFs who construct bespoke portfolios for investors from a menu of different hedge fund strategies.
The methodology involved modelling a series of portfolios with a sample set of data based on the weekly returns of 14 Lyxor hedge fund strategy indices. The sample period chosen was from 4 January 2005 to 28 December 2010, yielding 312 weekly observations. Three portfolios were modelled: a static portfolio, a Garch portfolio, and a regime switching correlation model (SRM) portfolio.
The static model assumes that volatility and correlation are constant over time. In a mean variance framework risk dimensions are defined based on the level of risk of different assets in the allocation (standard deviation) and the degree of correlation between assets. As Vaissie explains: “The key approach in this research was: let’s differentiate the standard deviation and correlation terms, then let’s model those two components in a time-variant manner and see what the dynamics are behind them. From a technical standpoint, a Garch model accounts for time-varying volatility while a Regime Switching Dynamic Correlation model accounts for time-varying correlations.”
Two market conditions were defined for the correlation component: stressed and normal.
Vaissie says that by using this approach he found that re-correlation was driven by an increase in volatility. This has far-reaching consequences because the way someone constructs a portfolio will differ depending on what they consider to be the determinant of this re-correlation. What Vaissie’s research is essentially showing is that managers should try to think one step ahead with risk management, think about what exposures they have and the quality of diversification they have: should they maintain the same exposure, reduce it, or change the nature of risk?
Within the static portfolio, the biggest allocation was to long/short equity market neutral at 45.5 per cent, followed by convertible bond arbitrage (20.9 per cent). By contrast, when modelling for time-varying volatility over the in-sample period, the GARCH portfolio had a much lower exposure to long/short equity market neutral, dropping from 45.5 per cent to 21.4 per cent. It also contained a more diverse range of strategies including special situations, short-term CTAs, fixed income arbitrage etc. The portfolio composition yielded similar results for the SRM portfolio.
“With traditional static models what you often end up doing is discarding strategies which, on average, have a high level of volatility and focus more on low volatility strategies. Risk-adjusted returns might be good, but they’re fragile because you only need one crisis to fall below your performance expectations and this therefore makes the portfolio susceptible to fat tail risk. By introducing a time-varying notion you end up adding strategies that you would previously have disregarded. Although they might be more risky their volatility tends to be more stable relative to lower risk strategies and their contribution to the portfolio overall risk budget might be positive in times of stress,” asserts Vaissie.
This creates a portfolio that is not only more diversified but ‘better’ diversified.
The results found that while returns were slightly lower in normal conditions for the SRM portfolio it ended up with a compounded annual growth 26 per cent higher, thus confirming that the best way to make money is to avoid losing it in the first place.
“With SRM we found that when accounting for the dynamics of both the standard deviation and correlation terms downside risk was reduced by 50 to 55 per cent on average. I was expecting a strong result on downside risk but what surprised me was the performance outcome. If you reduce your portfolio’s level of risk you expect to see lower performance. However, the results show you can actually reduce risk without reducing performance to the same extent, provided you properly integrate information and dynamically adjust your portfolio overall risk budget,” explains Vaissie.
Some might find it counter-intuitive to add riskier strategies to protect downside risk. This study shows the positive diversification effects they can have on long-term risk-adjusted returns.
A simulation was also run using an out-of-sample analysis. The in-sample analysis makes the assumption that the investor has all the relevant information sets and that any future crisis will be similar to ones that have occurred in the past. The out-of-sample analysis makes the assumption that a future crisis cannot expect to be the same as all previous ones, and therefore not all information is available to the portfolio manager. This is logical enough as no two events are ever the same.
“What we found in the in-sample and out-of-sample experiments was a similar result regarding downside risk. Downside risk is much lower than the static approach. Knowing that there was no information in the initial calibration period that could have helped predict the 08 crisis our dynamic approach to portfolio construction was still able to reduce downside risk by over 50 per cent,” confirms Vaissie.
Overall, the research showed that the dynamic portfolio reduced downside risk to a similar extent in both test scenarios. Vaissie concludes that this serves to illustrate that dynamic risk management can indeed add value. In his view, if portfolio managers stop thinking about purely as risk reduction it will change the way they integrate it into their portfolio construction process: particularly useful for FoF managers: if they can clearly demonstrate the value add of risk management they’ve a good chance of standing out from the rest of the competition.
Vaissie finishes by saying: “Previously, getting more performance involved taking on more risk. We show in this study that it’s not more risk you need. Rather, what you need to do is adjust the level of risk more dynamically to the environment and also adjust the structure of risk on an ongoing basis.”