Quant: All in the numbers
Anthony Harrington asks what went wrong for quants during the crisis, but also questions the received wisdom that ‘all quants are the same'
After a torrid time through the downturn, quantitative fund managers are doing
their best to restore investor confidence - in the teeth, quite often, of persistent myths that somehow they were responsible for the global crash of 2008.
One of the most famous attempts to pin the blame for 2008 on the quants was a 2010 article and book by former Wall Street Journal writer Scott Patterson, whose title says it all: ‘The Quants: How a New Breed of Math Whizzes Conquered Wall Street and Nearly Destroyed It'.
Patterson's WSJ article sidesteps alternative analyses of the crash - how low post-dot com interest rates fuelled a credit bubble that spread through mortgage and securitisation markets, for example, or how the emerging economies' ‘savings glut' led to huge and unmanageable inflows into the US, triggering a risk-blind chase for yield - to focus on the plausible thesis that multiple groupings of quants scattered among leading US institutions, managing vast portfolios with similar models in similar long/short statistical arbitrage strategies, will surely lead to bad things happening when the market suffers a series of shocks.
The short version of his argument is that momentum-biased quants massively exacerbated whatever perturbations the market was suffering, anyway, thanks to growing fears over US sub-prime contagion driving up stocks which should naturally have been crashing in an economic downturn, and then pushing them down when any sensible investor would have wanted to hold them through the downturn. The result was huge trading volumes generating incoherent price gyrations. Moreover, Patterson pointed out that as things got really bad, the high-frequency statistical arbitrage traders that usually provide ready liquidity for sellers instead decided to wait out the storm, leaving a market full of sellers and no buyers, and a downward spiral.
Professor Alex McNeil at Heriot-Watt University in Edinburgh is recognised as a leading expert on quantitative techniques. He is particularly interested in the complexities of valuing ‘difficult' asset classes such as insurance company liabilities, and knows the problems that quantitative techniques can run into.
"I was at quant conferences well before the crash where it was widely recognised that industry standard models for valuing credit derivatives, such as the Gauß Copula model, were not exactly brilliant mirrors of reality," he recalls. "They always required recalibration and didn't give very accurate answers."
Essentially, financial institutions failed to deal properly with the risks associated with their models. "In fair weather markets, these things would just about get you by, but you couldn't place a lot of faith in them in a crisis," he says. This was largely a corporate governance issue. The further removed management were from the originators and users of mathematical models, the greater the faith they tended to place in those models. Quants knew they were rough approximations. Management treated them like the word of God - so long as they made money for the firm. (This problem is pursued in a widely-read 2010 paper by Andrew Lo and Mark Mueller, ‘Physics Envy', and in our opening article, ‘Quant surveyors').
However, a key assumption behind the idea that a bunch of backroom geeks broke Wall Street is that all quants are the same, or at least very similar. It's an assumption that is a source of considerable irritation to quants professionals, who argue that today there are almost as many different models - some very different indeed - as there are quant funds.
Huge losses run up by major quant operations such as Morgan Stanley's Process Driven Trading unit (PDT) were indeed matched by equally massive outflows from many quant funds through to 2010. According to press reports, Barclays Capital analyst Matthew Rothman calculates that equity assets managed by US quants fell by around 60% between June 2007 and December 2009, and suffered further outflows of some $80-100bn through 2010 as high volatility and sharp trend reversals caused underperformance yet again. But, as always with broad investment trends, the closer one looks the more complex the picture becomes. There is mounting evidence that some quant funds have been generating an impressively consistent performance through 2008-10, and are attracting serious money from institutional investors.
As David Schofield, president of INTECH International notes, quant performance these days is all about consistency, rather than massive outperformance. INTECH manages $44bn, making it one of the leaders in mathematical investing, and it attracts pension
fund clients with relative return strategies that aim to beat benchmarks consistently by anything from 1.25% to 4.00%, depending on the fund composition and strategy.
INTECH is a predominantly long-only house; its strategy is very different to the dominant quant long/short statistical arbitrage approach. Its process is based on a mathematical theorem that proves that the growth rate of a portfolio exceeds that of its constituent stocks, with the excess growth rate being a function of the variance and co-variance of the different stocks - their individual volatilities and the correlations between them. Using ‘stochastic portfolio theory' developed by founder Robert Fernholz, a former Princeton maths professor, INTECH targets this specific characteristic of portfolios and, in doing so, exploits some of the inefficiencies of market cap-weighted benchmarks (thereby outperforming them).
"We focus, in particular, on the relative volatility of stocks, relative to each other and the market," Schofield explains. "If you think of a stock that goes up by 10% and down by 10% through two periods, consecutively, the average period return will be zero, but in reality you are at -1% with respect to compound returns. If you double the volatility [up 20% and then down 20%] you are at -4%. So the compound rate of return of more volatile stocks is less than that for less volatile stocks, all other things being equal, and this volatility drag on compound returns provides a mechanism for beating the market. The computer runs the optimisation, which is a set of portfolio weights to the stocks, and we have human oversight of the reasonableness of the results."
For Schofield, what INTECH does is very different from what most people would call a quant-based strategy - many of which are designed to identify securities pricing anomalies before they are ‘arbitraged away'. "We are not trying to exploit market pricing inefficiencies," he explains. "Instead we are trying to exploit volatility - a natural and permanent characteristic of any market."
As a relative return strategy you would not expect an INTECH fund to be a safe haven through a crashing market. Many of its funds beat their benchmark by a few percentage points through the dark days of 2008 but with the benchmarks losing 39% or more, that would have been cold comfort to investors. However, as Schofield points out, any fund that can consistently beat its benchmark year-in and year-out will easily outpace most traditional actively-managed funds and climb the rankings.
Janet Campagna, CEO of QS Investors, agrees with this principle. However, she adds that performance is always relative to the universe of equities. Two percentage points above benchmark will probably not constitute stellar outperformance for a small-caps quant fund, but in a large-caps fund it certainly would, over time.
QS manages $3bn and advises on a further $89bn. It, too, shoots for, and generally achieves, a few percentage points above benchmark, but it has a completely different approach from INTECH's, and different again from long/short statistical arbitrage quant funds. QS spun out of Deutsche Asset Management in August 2010 but its asset management team has been working for over 20 years.
"Quants are no more all alike than fundamental managers are all alike," Campanga says, claiming that correlation among quants is actually lower than among traditional fund managers. Nonetheless, one common feature they do share is a tendency to hold many more securities than, say, a fundamental value fund, and that is one reason why the liquidity crisis had an adverse impact on their broad portfolios particularly heavily - and blindly (without regard for idiosyncratic stock characteristics).
"We think that it is very overblown to claim that quant strategies do not work, or that they cannot handle high volatility in the markets," says Campagna. "Clearly, one of the things that happened in the crash was that some large funds got squeezed and had to pull money out, and the place to do that was equities in the large-cap space and selling there drove the market down."
Campagna points out that since enhanced index funds get lumped together with quant funds operating very different strategies, it becomes difficult to separate out manager alpha deterioration from market beta fall when the tide goes out for markets generally.
QS's Portfolio Choice approach is designed to address issues such as incomplete manager data and non-normal returns. The approach also aims to quantify portfolio efficiency, defined as the probability that any given portfolio will achieve investors' objectives. This creates a dynamic approach to portfolio selection, with investors able to see which of the constituent elements are likely to contribute best to minimising any potential shortfall in achieving their objectives.
"The most interesting thing about our strategy in, say, the US large-cap universe, is that it takes into account where in the market cycle you are at any point in time," says Campagna. "The kinds of things we look at are valuation information, sentiment information and information on growth, but the relative importance of each of these is driven by another layer, which is all about what is going on in the market. Is it dominated by greed or fear? Is it a junk rally? This is all about pinpointing where you are in the market cycle."
Of course a traditional fundamental value manager does much the same thing, using their experience. The huge difference between them and quants, however, is that for quants it's a pure numbers game. "I've never looked the CEO of a company in the eye to determine the validity of the company's growth strategy," as Campagna puts it "There's no need. It's all in the numbers."