CFM Seeks to Multiply Capacity with ISD

Research, university collaborations, and innovation

Originally published in the February | March 2015 issue

CFM is developing in multiple directions: expanding its research collaboration with leading universities; sponsoring scientists to pursue theoretical research; publishing more and more of its own applied research; and pulling these strands together by carving out lower-fee, less capacity-constrained strategies designed to appeal to the largest institutional investors and the smallest retail investors alike, including via UCITS funds.

Applied and academic research
CFM’s latest research paper, released in September 2014, is entitled “Risk Premia: Asymmetric Tail Risk and Excess Returns”. This substantial survey of risk premia confirms that investors are rewarded for taking on negative skewness, or negative tail risk, and not for symmetrical volatility per se, contrary to traditional theories of asset pricing. CFM makes no claim to have pioneered the tail risk concept, but wanted to test the robustness of the concept more thoroughly than had been done before.

CFM’s extensive research suggests that the risk premium is “universal and exhaustive, applying across asset classes and strategies,” says CFM co-founder, chairman and management board member, Jean-Philippe Bouchaud (JPB). So, the tail risk premium has existed amongst the 27 equity markets reviewed dating back to 1870 for some; it has applied within equity markets to classic Fama-French factors such as small caps versus large caps, value as measured by price to book value ratios, and to momentum; it is relevant to the superior returns delivered from high-yield and distressed credit investing; to the currency carry trade; and to strategies such as selling volatility in commodities, currencies and equities. The wide applicability is an important test of soundness, as models that only work for specific assets, strategies or time periods can be dubbed “over-fitted” and less likely to be repeatable phenomena. So CFM’s firm conclusion here is that risk premiums are nearly always negative skewness premiums. These tail risk premiums provide the empirical rationale for many strategies CFM pursues, mostly in its new ISD (Institutional Systematic Diversified) product.

But there is one strategy where a negatively skewed return profile cannot explain the returns: “Trend following is almost unique in displaying a positive risk premium as well as positive skewness,” says JPB. Granted, the “value equity” strategy, of shorting stocks with a high price to book value ratios and going long of those with low price to book value ratios, does also show some positive skew, but this is not as statistically significant as that applying to trend following. As with the tail risk premium, CFM has sought to quantify the trend-following return premium more rigorously. Their April 2014 “Two Centuries of Trend Following” research paper on the topic stretches back 200 years, and finds that all 10-year periods since 1800 have shown positive returns for trend following. The explanation here may lie in behavioural finance: “even though markets have massively evolved over that 200-year period, a persistent human trait explains the success of the strategy,” argues JPB.

That trend following in general struggled to produce positive returns post-crisis, between 2009 and 2013, is perfectly consistent with its historical pattern. “If a strategy has a Sharpe ratio of around 0.5 nobody should be surprised by some multi-year periods of negative performance,” explains JPB. The real surprise for CFM is that “people turned so negative on trend following, as phases of poor performance are all part of the contract,” he adds. Evidence of persistence over decades – and sometimes centuries – is CFM’s over-riding criterion for existing, and any future, strategies.

CFM maintains stable allocations to all sub-strategies, as they do not think they have any edge at timing the alphas. “We think it is very hard to predict the opportunity set, so we maintain very static allocations to each sub-strategy,” says JPB, who contends that the spectacular resurgence of trend following in 2014 shows the dangers of looking in the rear-view mirror and cutting recent underperformers. JPB is prepared to admit that “maybe we haven’t yet found the magic way to time strategy allocations, so for now we prefer to be agnostic.”

University collaboration
JPB has for many years been challenging the efficient markets paradigm, and CFM has always kept abreast of the leading edge of academic research in multiple disciplines, not just finance, since the Paris-headquartered firm was founded in 1991. CFM has been collaborating with universities since the 1990s in various ways: hiring interns and post-doctoral students, offering academics opportunities for one-year secondments, and working together with universities on specific projects. JPB also teaches at France’s elite Ecole Polytechnique, and although there is no formal collaboration, his engagement there sowed the seeds of the recently announced alliance with Imperial College: some 20 years ago Professor Rama Cont of Imperial was one of JPB’s PhD students.

“The Imperial deal is a natural extension of our long history of working with universities, opening up collaboration horizons to non-French universities and in particular UK ones,” explains JPB. CFM was particularly keen to join forces with a UK institution to raise their profile in the UK, which is home to Europe’s largest pool of pension fund assets. CFM offers grants to PhD students in France to study pure maths and pure physics, but the CFM-Imperial Institute of Quantitative Finance is sponsoring PhD students, post-doctoral fellows, and visiting international scholars to work on finance.

This marks a new departure, as CFM has not until now recruited from economics or finance. Their traditional recruiting grounds included hard sciences such as maths and physics, and also theoretical biology and theoretical computer science. But there are certain common qualities running through these disciplines. “We need people who are good at handling concepts, data, and computers, and who are fluent with concepts and methods so they can apply them to research and finance,” elaborates JPB. CFM will be offering grants to one PhD per year at Imperial, and to two post-doctoral students – on top of the two or three PhDs working in the Paris office. Brevan Howard’s research alliance with Imperial, The Brevan Howard Centre For Financial Analysis, was launched at the same time as CFM’s joint venture, but is quite separate: Brevan Howard is tying up with the Business School whereas CFM is joining forces with the Mathematical Finance Department. The CFM venture need not be completely siloed from other departments at Imperial, however: Professor Cont says, “We see the CFM-Imperial Institute as a platform for interdisciplinary research and exchange of ideas, open to academics from all departments as well as our industry partners.”

This open-source ethos means that CFM will not dictate the research agenda. Indeed, “researchers have complete freedom, and even with employees good research requires freedom,” reflects JPB. CFM has always employed post-docs who are straddling finance and academia, and who want to keep a foot in each camp in case they want to return to an academic career. CFM respects the need of such researchers to continue publishing more abstract research. In any case, while CFM will provide guidance and ideas, “the best students come up with their own subjects,” recalls JPB. CFM sponsored the first two post-doctoral students at Imperial in October 2014, and they are completely free to research whatever they wish.

Moreover, the researchers are free to publish their academic research. “The lesson of the last 20 years is that we would have lost a lot of human talent if we had kept everything proprietary,” reflects JPB, and staff turnover at CFM is indeed low. CFM has been reaping the rewards of publishing research since the 1990s. JPB harks back to 1997, when the book he co-authored with CFM’s co-CEO, Dr Marc Potters, on financial risk and derivative pricing, was first published, in French; the second edition in 2004 was also available in English. “As data became more widely available people understood the theory better,” he says, “and this helped CFM to become more widely known in public, raising awareness of us.”

Yet the nuance here is that CFM does have to clearly delineate what is publishable and what is not. Although the PhD students and others may be spending some time in CFM offices working on more practical and applied topics, they do not have open-source access to the entire CFM network, and they have no exposure to some data that needs to be kept confidential. The ultimate reason why CFM is relaxed about students publishing the more theoretical research is the challenge of translating theory into practice. “There is a huge amount of work between research and implementation, and the devil is in the detail,” says JPB. Half of CFM’s costs are research in one way or another, which includes 38 people on the research side, data, computers and technology. The firm boasts straight-through-processing infrastructure across seven data centres on three continents, and 64 staff are employed in IT. CFM handles 2.5 terabytes of data per day, and its library contains 3 petabytes of data (note that one terabyte is 1,024 gigabytes, and one petabyte is 1,024 terabytes, so one petabyte is over one million gigabytes!).

The Imperial alliance may foster both theoretical and practical research. CFM is actively supporting more applied research that is intended to provide “insights into the problems that confound traders, risk managers and regulators.” One example here is market impact, or how making a trade can adversely change prices. “Transactions costs are key as you grow bigger,” recalls JPB, and market impact is also central to the whole debate over high-frequency trading (HFT), an activity that CFM is not engaged in. CFM takes pride in the Market Microstructure conferences held in Paris every three years, in association with the Institute Louis Bachelier. The last one in December 2014 attracted both industry practitioners and some of the world’s most eminent academic experts on HFT and transactions costs. And an acute awareness of how transactions costs can increase with assets is one driver behind the creation of the ISD strategies.

Introducing ISD
Stratus remains CFM’s flagship product, but its assets of c.$4 billion could also be close to full capacity. When it comes to predicting capacity for either Stratus or ISD, JPB has some humility in admitting that the true capacity for a strategy can only be discovered through experience. Here there are two issues. So-called alpha decay is the constant bane of the quant’s life, because “most models slowly degrade over time, and so we need to move on and deal with degradation of signals.”

Therefore, future capacity partly depends on the extent to which CFM can find and implement new strategies, and JPB ventures that new models and ideas might allow for capacity of $6 billion, but is doubtful about growing beyond that. Capacity could also depend on locating new pockets of liquidity, as the market impact of trading grows with the size of assets. JPB has somewhat more confidence in the capacity potential of ISD partly because “the slower strategies have more capacity” – simply because a longer time window increases the volume, depth and breadth of market liquidity that can be accessed. Right now in early 2015, JPB suggests ISD might have capacity of $10 billion, but he admits “it might be double or half this amount, because you only know after you start trading.”

As well as (probably) offering more capacity, ISD is differentiated from Stratus in several ways that may appeal to certain categories of investors. Given that CFM is a signatory of the Hedge Fund Standards Board (HFSB) and is overseen by five regulators (the CFTC, the SEC, the AMF in France, the Japanese FSA, and theUK FCA) the manager already ticks many boxes that may be sought by institutional investors. But Stratus cannot satisfy the criteria of some investors who seek more transparency or lower fees.

ISD’s greater transparency is attractive to those investors who cannot accept the “black box” nature of some parts of Stratus. ISD is transparent about the types of betas or alphas it is seeking to capture, and it could come under the umbrella of “alternative beta” – though this is, of course, a very broad label. Long-term trend following, based on a 150-day look-back, is one. Risk premia, including the currency carry trade and selling volatility on a controlled risk basis, are the second bucket. The third approach, quant equity market-neutral, seeks to capture returns from three factors: momentum, value and quality.

Currently each of these three buckets gets an equal, one-third, risk allocation. The three strategy groups have historically shown very low correlation to one another, and ISD’s correlation to equities should be low but variable. The equity market-neutral part is, by definition, market-neutral. The trend-following sleeve can go either way, but has no structural bias to being long or short of equities. One of the sub-strategies within risk premia is, arguably, implicitly long of equity: the systematic short volatility component will usually have some positive correlation to equities because volatility tends to move in the opposite direction to equity markets. But the trend-following and short volatility strategies are both trading other asset classes besides equities, so overall equity exposure should never be more than marginally long.

The strategy mix in ISD may change when and if new strategies are added. Envisions JPB, “If we find another slow strategy with good performance and transparency we may add another layer.” CFM is exploring the possibility of adding credit strategies, merger arbitrage, and also “betting against beta” modules that exploit the outperformance of low-volatility, low-beta stocks. If interest rates on short-dated government debt stay negative for extended periods, CFM might research whether it makes sense to take some credit risk with cash to earn a positive yield, but this has not been investigated yet.

ISD has made a strong start and has already demonstrated that it is not simply a generic trend-following system. ISD’s trend module did well in 2013, despite the general headwind for trend following, while the 33% return for the trend part in 2014 was well above average for managed futures or CTAs. The trend part is already available on a standalone basis as ISF. The overall ISD strategy returned 13.76% in 2014.

ISD’s fee structure of 1% and 10% is palatable to some types of institutions that cannot sign off on traditional hedge fund fees of 2% and 20%. Some of the simpler ISD sub-strategies might even be accessible with different fee structures, and CFM is open to devising customised solutions for specific investors. The ISD strategy could be offered via managed accounts with a minimum size of $50 million, while the LP structure has a minimum subscription of €1 million. Historically CFM’s investor base has been comprised of pension funds, endowments, funds of funds, insurance companies, and family offices, but CFM has just launched a UCITS version on the Rothschild platform with a minimum investment of just €5,000. CFM chose to partner with Rothschild as they are well equipped to deal directly with a large spectrum of investors. A ’40 Act product could be launched for the US market as well. Investors should watch this space, as CFM keeps researching, publishing, evolving strategies and developing new products.