Rethinking risk in a crisis
Between the Global Financial Crisis and the latest COVID-19 induced market corrections, managing risk through extreme volatility is quickly becoming a feature of 21st century risk management.
Speaking at the 2nd Risk and Performance Summit earlier this month, a panel of risk experts, led by Viktor Svatek, Head of Investment Risk at the Future Fund, discussed how these “fat tail” events have shaped risk-modelling.
Darrel Yawitch, Chief Risk Officer at Man Group in London said focusing on building portfolios that can withstand a variety of shocks, rather than trying to predict the market with 100% accuracy, is key to managing risk through cycles.
“You have to try and be prepared for a shock in the system, but it is very difficult to predict exactly what that shock is going to be, so it’s more about resilience than about being 100% accurate in what you’re doing. It’s about trying to prepare for every eventuality, but thinking through different scenarios,” he said.
Yawitch said that he had not expected to see a crisis of the same magnitude and scale as the global financial crisis so soon after 2007/2008, and that - particularly in the early days of the crisis - there had been a widespread underestimation of just how significant the impact of COVID-19 would be globally. However, he said that from a risk management point of view it has proven a good test of the systems that have been put in place.
Greg Hall, Executive Director at AQR Capital Management in Sydney said that particularly in equities markets, sharp corrections are part and parcel of investment management. Taking a conservative approach to liquidity risk management is key to navigating through these crises.
“We invest in equities [and] we know that there is going to be a minus 10 per cent or a minus 20 per cent week somewhere, so there will be a tail in those returns, we know it can happen and we plan for it,” Hall said.
Hall said that within long-short equity strategies, AQR manages these risks by seeking to ensure there is enough cash to last out an event horizon. “Historically, these events have been very short-lived. We don’t want to be forced sellers, so we look to use cash as a cushion,” Hall said. “We think about [our cash levels] as providing a long runway, so that we have an opportunity to make decisions through that event,” he said.
Svatek asked the panel what factors had driven decisions to either hold on to risk, or to cut it, during the during the risk sell-off in March. Yawitch said that while cutting risk is generally the easier decision to make – holding on to positions may work out well – though it requires having the flexibility to keep these positions. At Man Group, such decisions are made on a case by case basis.
“The first question is what is our mandate, what are our clients expecting us to do and what is in the best interest of our clients?” he said.
On the systematic side of the business, Man Group employs automated risk management controls that limit the total amount of risk - in many cases using volatility scaling - so that as volatility spikes, positions are naturally reduced. “That makes my job as Chief Risk Officer a lot easier, knowing that many of our signals do this automatically,” Yawitch said.
Where discretionary long-only mandates were concerned, being true to mandate in some cases meant letting risk run, a difficult decision, but the right one if the alternative meant deviating from the style and mandate agreed beforehand. “We approved limit waivers for a certain period of time, after which point we would review and the trick is many of those we are still reviewing. Risk has spiked, it has remained elevated and only now - six months later - are we starting to see risks, depending on what speed and frequency you use, return to normal levels,” he said.
In other strategies, Yawitch said, the problem wasn’t so much that risk was elevated, but rather that there was a specific stress that caused abnormal movements. “We saw that in the treasury bond versus futures basis trade, we saw it in risk arbitrage, we saw it in market neutral equities.[ In those situations] there was a more nuanced discussion to be had. This wasn’t a function of an increased risk, it was a function of ‘the markets not behaving in a normal way’. When normality is restored, everything will revert and you have to be quite careful not to cut at the bottom, because these things do bounce back,” Yawitch said.
Managing model limitations
Associate Professor Geoff Warren of the Australian National University stressed the importance of tailoring risk models to their objectives and – in particular – the time horizon over which these objectives are to be achieved. “The risk metrics you’d apply to short term horizons are quite different to what you’d use for longer term investment horizons. If you are worried about fluctuations in your portfolio value over shorter horizons, volatility and factor models will do the trick. When you lengthen your time horizon, the key definition of risk isn’t volatility, but permanent loss of value. In that case you need a different set of toolkits, for example measures that focus on cashflow delivery, rather than repricing risk,” Warren said.
Most vendor risk models are relatively short term- focused, using daily data with relatively short horizons of historical data sets. Not surprisingly, the margin of error in risk models increases with large unexpected events and particularly in crisis situations, need to be accompanied by stress testing and scenario analysis.
“Risk models will provide you with a framework and the data to give you an understanding of the way that the market works and behaves in the short term, but they definitely ignore structural breaks [and] you need to go off piste, outside of a vendor model, to be able to account for that in a sensible way,” Ian Hissey, Vice President Risk & Quant Solutions at FactSet in Tokyo said.
He said that compared to 2007/2008, market understanding of model limitations has increased significantly. “We got less ‘throw your hands up in the air and the model is stupid’ compared to 2008, which is nice, so I think people have definitely got a more nuanced view and understanding of how risk models work,” Hissey said, adding that FactSet has seen a significant increase in client demand for stress testing and scenario analyses.
Man Group’s Yawitch said his firm relied on a full gambit of different risk models and stress tests, and his “overriding message” was to not rely on any one model. “If you went to the cockpit of an aeroplane and the pilot says “let me show you how I fly this thing. You see that one number there? that is all I need,” I don’t think you’re going to have a lot of confidence that the plane is going to arrive at its destination. You really do need your dashboard using different metrics,” he said.