|
|
Rethinking Risk Controls for Quants
Eva Xu of of Bayswater reviews the implications of the recent experience for
quantitative trading. For more on her and Bayswater, see the previous section,
Founders.
Let’s be honest about it: we quantitative managers as a group did not pay enough
attention to downside risk. This is a mentality that probably reflected our
generation’s experience. After all, financial markets had behaved nicely during
much of our careers.
Even when you’re back-testing the model with historic data, often there aren’t
enough stress points. There may be just the one major incident of Long-Term
Capital Management. As a result, most of the models indicated very low frequency
for the type of meltdown we experienced in August 2007—less than 5%. So that’s
how you calibrated the model.
We did not reduce risk positions too much because the probability was low. In
2007 we collectively paid a price. The points below reflect what I learnt from
my experience at Bayswater, but the issues are common to all quantitative
managers.
Nowadays I spend 98% of my time thinking about the 2% probability of downside
events. That was not the case previously. But quants have diverged in their
reaction to the past two-and-a-half years’ experience. Some have learnt their
lesson and made risk management a major component of their program. Others think
we’ve gone back to the normal state of markets and don’t need to change how we
do things. They’re more likely to just pay lip service to risk management!
My main takeaway is that we need to have our risk management programmed in
dynamically. Consider that by the time the VIX indicates something unusual is
happening, you have already lost money. You have to prepare beforehand to
respond to fast changes. Unfortunately, many quants don’t do this.
Risky Scenarios
In August 2007, one of major hazards faced by quantitative strategies surfaced
without warning. Suddenly, markets and investments became highly correlated.
This did not happen because quant managers all had the same trades—for instance,
systematic macro funds did not have similar positions to equity market neutral
funds, but about a week after equity funds lost, macro funds also made heavy
losses.
What caused this contagion was margin calls from prime brokers. Any similarities
between strategies was less important. To meet margin calls, multi-strategy
funds sold not only equities but also, for example, currency holdings. So our
currency books went down. It did not matter what your positions were, sales to
cover margin calls spread across markets and drove down prices. What that shows
is that having what looks like an uncorrelated strategy won’t protect the
portfolio when a liquidation crisis hits.
It turned out that quants were the canary in the mine. We got a taste of the
crisis early on. Later, in 2008, all strategies failed except short selling and
commodity trading advisors. There is a silver lining for being early: we had a
little bit more time to prepare for 2008 than everybody else.
First of all, let’s distinguish between two situations:
1. Crises, which are short and fast.
2. Recession and recovery, a more long-term process.
These are different types of risky scenarios, so you need different responses.
By definition, crises are shocks. We simply don’t know where the next one will
come from. It won’t be subprime, it’s going to be something else. So you have to
pay attention to different markets, not just the one you’re trading in right
now. Something happening in other markets can come and hit your position.
You can no longer assume that the long-term risk of a certain asset is X and be
done with it. You have to watch whether markets are functioning normally. Most
of the time they are, of course. But there are occasional big shifts in market
behavior.
Take the equity bond risk premium. The expected premium is around 3%, but over
the past 20 years equity risk relative to bonds has varied from about 2% to
nearly 5%.
In short, our models need to adapt to changing risks. We can’t rely on static
risk measures; we need a dynamic approach.
Crises
At Bayswater, we went into high drive to understand and meet the challenge. It
took us several months to track down the causes and figure out what we could do
about them.
By early 2008, we had developed a new risk management regimen.
We found that instead of cutting positions or getting out of the market in
crisis situations, we’re better off hedging. When you get out completely, you
have no way to capitalize on the upside. In August 2007 the market recovered
quickly by month end and funds that did not get out did better in the quant
quake. By hedging, you make sure you don’t get hurt in the crisis while
preserving the upside.
We developed a crisis response that relies on investor sentiment indicators to
hedge our positions. We collect 12 market-based investor sentiment series that
capture equity, liquidity and credit risk. We take care to normalize these
series so that we can add them together.
This is an ongoing project. We add new series as we notice changes—for instance,
originally we did not have sovereign debt data, but once we saw how much
countries are going into debt, we added a sovereign credit default swap index.
There are arguments that hedging does not work. Say you’re long the S&P 500,
short Treasuries and face a 2007 type market dislocation. You may not be able to
get out of stocks fast enough. People say you can’t really hedge because markets
rebound so fast you’ll get whipsawed.
However, equity and bond markets behave differently. Equity markets move fast,
but bonds have a lot of inertia. With bonds you have time to get out, so you
have opportunity to unwind the hedge and hold on to the equity investment. These
are the types of market behaviors that we have to use to our advantage.
We added dynamic hedging in February 2008 and straightaway caught the Bear
Stearns collapse in March. If we had not added the hedge, we would have had a
big drawdown. We were protected and our currency and directional strategy made
money.
We think the hedge will make a big difference in the next crisis. It will happen
again, whether soon or in the longer term.
Recession and Recovery
To address changes in markets during recession and recovery, we added a risk
overlay that changes the model’s inputs. Basically, you need to have different
models for the two types of market behavior and switch when there is an
indication of problems. We use the difference between macro and micro analyst
forecasts as the indicator that the cycle is turning.
Looking at the economic situation now, it is striking that government policies
drove all markets in 2009. What happens from here on depends especially on
central bank policies. Usually recessions are not synchronized globally, but
this time they were. So we’re in a situation that four major central banks – US
Federal Reserve, Bank of England, European Central Bank and Canadian Central
Bank – will start tightening at around the same time.
There are two dangers:
1. Central Banks may exit easy money too early, causing another crisis, though
likely a small one, and then a double-dip recession. We plan to hedge against
the crisis and then switch models to prepare for equities coming back.
2. Central Banks exit easy money too late, or markets perceive it as too late.
If inflationary expectations take hold, if people believe we have inflation,
then we will have inflation!
Gasoline prices make it look like inflation is imminent but there is huge excess
capacity in the economy and there is no way wages will go up. Inflation is not
really a big risk. However, the perception of inflation may force the Fed’s
hand.
A replay of 1994 can happen. That year the Fed came up with a sudden rate raise
and bond markets followed. That is a big risk and we’re preparing for the
challenging transitions ahead.
|
RSS
