Quant: An alternative Aspect
Martin Steward talks to Martin Lueck about why systematic managed futures shone in 2008 while other quantitative strategies sank
As a co-founder in 1987 of Adam, Harding & Lueck (AHL) alongside Michael
Adam and David Harding, Martin Lueck can lay claim to being a founding father of European managed futures. Early practitioners evolved out of the trading pits of Chicago and New York, applying simple systematic trading strategies, usually based on technical analysis, mainly to the commodity futures markets. AHL was one of the first to bring a natural-sciences philosophy and serious computing power to bear on the project - Harding was a Cambridge natural sciences graduate, Lueck holds a physics MA from Oxford and Adam was a gifted programmer - and in the process they played a big role in developing what we now know as systematic managed futures: technology-driven trading of price trends across an array of commodity and financial futures and currency markets.
“The entrepreneurial beginnings of AHL was really the physics nerds trying something that seemed to work - and wondering why everyone wasn’t doing it,” says Lueck, now director of research at Aspect Capital, which he co-founded with Adam, Anthony Todd and Eugene Lambert in 1997. (Harding went on to found Winton Capital after AHL was sold to Man Group in the early 90s). “I’ve had a 20-year career asking myself that question. It takes a 2008 for people to notice us and ask, ‘Hold on, how does this work again?’”
In 2008 equity markets crashed by nearly 40%, the broad hedge fund universe lost 20%, and HFR’s Quantitative Directional index was down 23%. The press described how ‘computer-driven quants’ were devastated as their models struggled to negotiate strange market patterns and wild volatility. And yet trend-following managed futures sailed through it. HFR’s Macro Systematic Diversified index finished 2008 up 18%. Aspect Capital’s flagship Diversified Program returned 25%.
So why did this species of ‘quants’ outperform so spectacularly, just as all the others sank? The answer has to do with the strategy’s exposure to two related factors - liquidity and volatility - that loom large in any financial storm. Managed futures tends to be on the other side of the liquidity and volatility bets of most other quants strategies, as well as the average pension fund portfolio. Interestingly, this differentiation seems to be related to one of the key criticisms levelled at quants: that they build naïve models of how the world should be from a short period of in-sample data and then expect them to work when fed with new, out-of-sample data.
“Traditional quantitative is about fair-value models applied in the belief that if the market doesn’t agree with you then it must be wrong and eventually it will figure it out,” says Lueck, taking arbitrage strategies as an example. “Pairs of securities are ‘out of line’ so I’ll buy that one, sell the other and neutralise all other risk.”
Theoretically, this position has to make money because it is not a directional bet - the only question is how much time it will take. That bet - ‘Can I hold this position long enough to make money?’ - is essentially about liquidity. And that liquidity risk is intensified by another characteristic of this kind of trade.
“You are already in a situation where you have susceptibility to crowding,” Lueck observes. “The more people look for the same ‘mis-pricing’, the more likely the majority of that mis-pricing will be arbitraged away. To get any juice out of what’s left you have to apply leverage. By contrast, trends actually benefit from more market participants.”
Without leverage, if prices move a bit further ‘out of line’ an arbitrageur could hang on for as long as it takes for prices to ‘come back in-line’ again - or at least until clients start banging on the door. With leverage, margin is called, which forces the arbitrageur (and all the other arbitrageurs) to sell positions, and this pushes prices even further ‘out of line’. These ‘tail events’ occur precisely because quants leverage their bets on the middle part of their models’ probability distributions. The higher the volatility, the bigger the losses.
One can see that a lot of traditional quants involves building models that describe long-term averaged features of markets in order to facilitate trading positions that fight against what shorter-term price data is telling us. By contrast, trend-following managed futures is less theoretical and more empirical in its reliance upon data signals - as the name suggests, it follows them. And this leads us to how trend-followers are exposed to rising volatility, but also to the ways in which rising volatility and rising demand for liquidity are related.
“When people say that managed futures is a long-volatility strategy, that’s not necessarily true,” Lueck observes. “But it is true that, once a market begins to move in a particular direction, more and more people join the trend: they don’t want to miss out, their stops get triggered, they begin to re-evaluate existing positions, they are forced to sell. That all generally leads to higher volatility, albeit ancillary to the directional movement that we are really trying to capture. In addition, most managers will try to control for volatility: all other things being equal, if the market becomes twice as volatile I will scale my position down by half to maintain an equivalent value-at-risk.”
This last point is fundamental. The more people get involved in quants-driven arbitrage, the more leverage is required to maintain a given level of return, volatility of returns, or VaR; but the more people get involved in trend-following, the more volatile markets become, and the less risk a trend-follower needs to apply to maintain targets. So while the trend-follower will remain long a rising price trend (or short a falling price trend), it will also reduce its leverage to that trend as it becomes more pronounced, effectively taking profits. If a trend follower had 300% exposure to a short S&P500 index futures position going into 2008, for example, by the peak of the crisis when S&P500 volatility had increased by a factor of three, it would have reduced that position by a factor of three - effectively buying 200% of S&P500 exposure back from all those frantic sellers in October.
“That was the classic case of months of white water before Niagara Falls,” says Lueck. “The models began to establish a short position at the start of 2008, but started to buy back a lot once the market was in freefall because it was scaling back from massively increased volatility. That was liquidity provision for a market that was furiously demanding liquidity: the trimming we were doing was against the direction of the market.”
It is important to remember that this apparent long-volatility, provider-of-liquidity profile is incidental to the trend-following model, however. There are situations in which the strategy can find itself negatively-correlated with higher volatility and higher demand for liquidity.
The first of these would be an environment of high volatility but little directionality - ‘choppy markets’. The effect is compounded if that volatility is high in an absolute sense but low relative to recent experience, precisely because trend-following models reduce risk when volatility spikes. March 2009 was a classic case: markets bounced violently as central banks flooded the world with liquidity, but then drifted for the rest of the year. Equity market volatility was at 20% through 2009 - 50% higher than its medium-term average, but 100% lower than its 2008 peak. No directional trend lasted long enough for models still calibrated for super-high volatility to pick up: the HFR Macro Systematic Diversified index finished 2009 down 1.7%, and Aspect’s flagship programme lost 11.2%.
The second situation is when markets suddenly reverse. A managed futures programme is no better placed than any other strategy if it happens to be long on the day the market crashes (Japanese equities in March and the brutal correction for silver in May will have hurt), or indeed short on the day the Fed launches quantitative easing. “Sometimes we’ll be a taker of liquidity, trying to get through the same revolving door as everyone else,” as Lueck puts it.
In these environments, everything depends on how many of the trends the programme is following get reversed. The ‘Black Monday’ crash of October 1987 appeared to come out of nowhere, but while AHL’s long-equity position was hammered, the overall effect was balanced somewhat by its long-bonds position. The hit from May 2006 was much more painful. In that instance, a range of rising markets suddenly reversed on a bout of anxiety about stalling US house prices, the US current account deficit, weakening dollar, rising inflation and mixed messages from newly-installed Fed President Ben Bernanke. The 3.7% loss posted by the HFR index that month was worse than any single month in 2009.
“That was the first real thump we got from a correlated-risk shock,” Lueck recalls. “All assets completely reversed very quickly. That spurred us to manage risk in a different way from what the data was suggesting to us - because the data was telling us this was a once-every-250-years event.”
This leads us to consider how managed futures differ again from traditional quants, which measure diversification using long-term correlations and use it to manage risk via mean-variance optimisation. This approach tends to concentrate positions in a small number of (negatively-correlated) risks; and fails to account for the risk of correlations rising sharply in stressed markets.
In classical ‘naïve’ managed futures, diversification is used, not to manage risk but to maximise the number of exploitable trends: it doesn’t matter so much if a programme has to sit out of eight markets that are not trending, as long as it can trade two that are. Risk is managed, after all, by the natural calibration of each position’s exposure to momentum.
“Our first portfolios just relied on the fact that the yen is different from wheat, which is different from gold,” says Lueck. “Like everyone else we had a very laissez-faire approach to portfolio risk management, believing that our individual position risks were controlled and the correlation risk would take care of itself.”
That is a dangerous assumption, because correlation tends to increase as volatility rises: the ‘naïve’ trend-follower might assume that, because each individual position is being scaled back as volatility rises, aggregate portfolio risk must also have been scaled back - when, in fact, the reduced risk of individual positions will be counter-balanced by rising correlation. Those that are aware of the rising risk might take a discretionary decision to de-risk.
“We saw some of our worthy peers reaching in to dial down the risk in late 2008,” recalls Lueck, “which undermines the whole merit of having a systematic approach in the first place. So, in striving for a complete methodology, we implement a systematic risk management policy that uses correlation data to calculate our VaR, and at a certain point of rising correlation the model will constrain the amount of additional risk that gets put on.”
Not every managed futures firm goes for this kind of thing. Some prefer to keep this risk-management overlay discretionary; others prefer to keep the trend-following as ‘pure’ as possible. But the important thing to note is that Aspect’s decision to enhance its programme by taking correlations into account was a move away from the assumption that long-term average correlations are basically stable, towards acknowledging that correlations are unstable and must be managed dynamically. Lueck regards that decision as a very important enhancement to Aspect’s programme, and thinks that a failure to take dynamic and rising market correlations into account can be seen in the deteriorating risk-adjusted returns of some competitors.
“The group of managers that grew up in the US in the 1970s had 10 years of zero competition, without the same kind of dynamic, scientific research we have now,” he observes. “Through the 1980s and 90s the sales pitch for managed futures was, ‘We have developed this model and we never change it because it is a true manifestation of investor psychology’. But the markets have moved on. Once every few years there will be a 2008 when the stars align and off those original models go; but you have to suffer almost intolerable volatility and 40% drawdowns waiting for it to happen.
“The band of managers who grew out of AHL, and others who share that philosophy of scientific curiosity, can, I think, show more consistent risk-adjusted returns because we are always asking whether the facts of past data still apply. Ultimately, even a naïve trend-following strategy is remarkably robust and will get through most things - but there is a big difference between doing it well and doing it poorly.”