Do quant strategies work, or are their proponents simply slaves to the algorithm? Nick Fitzpatrick investigates.
The advantage of being an investment management business built mostly on computing power is space. You shouldn’t need a lot of it.
Such is the case at Kestrel Investment Partners. Its premises close to Trafalgar Square in London are comparatively small, with barely room for two people in the groundfloor kitchen.
However, what Kestrel – with its £324 million (€378 million) of assets under management and 18 staff – saves on space, it makes up for in computing power. When I visit the firm’s office in June 2018, its chief executive, John Ricciardi, is sitting in front of a projection of what could be a lunar mountain range but is in fact a surface chart designed to predict where returns will be generated.
The ‘mountains’ are built from data – mountains of it, if you’ll forgive the pun – crunched by Kestrel’s computer. “When we founded the firm, we started with 50 million time series, but narrowed it down to 2 million that we find useful,” Ricciardi says.
Assets in computer-powered funds have grown in recent years. Quants are the best-known type, and more hedge funds have been observed to employ data-crunching quant techniques.
Simultaneously the term ‘quanta-mental’ has been coined to express a convergence of quantitative and fundamental investing, where the manager may have more discretion over the portfolio than is usually the case in pure quant funds.
Smart beta and factor investing strategies, meanwhile, are often described as a form of quant management.
Whatever the best-fitting term may be, the commonality is that these tech-powered strategies rely on identifying statistical patterns drawn from large amounts of data that, crucially, should support an investment hypothesis created by the manager or by academics and which is coded into computer algorithms.
The amount of human input in the final investment decision might vary, but the preference is for the hyper-rational computer to make investment decisions in place of an emotional, human fund manager who might act irrationally. As one manager puts it: “The computer doesn’t come into work unsettled after an argument with a spouse that morning.”
Greater data range
Quants and other investors that rely heavily on data and statistical interpretation have been criticised in the recent past for drawing similar conclusions from the same data. It meant strategies could be imitated, crowded or arbitraged until there is no return left. However, the investors Funds Europe consulted for this article say they are now drawing from a greater range of data than quants were a decade ago. They argue Big Data and the means of extracting it enables them to better distinguish themselves.
At Kestrel, Ricciardi says the data is interrogated so much that “you wouldn’t recognise it” from its original form.
He is explaining one of its strategies – the Global Portfolio, which is available in a Ucits vehicle and allocates to a global multi-asset portfolio, including commodities, cash and property. It’s a macro fund using quantitative and fundamental analysis; Ricciardi describes it as a “data-driven discretionary fund” rather than a quant, but like macro quants, it measure millions of economic data points for macro-economic effects. The ultimate objective is to see how the data will influence which investment factors are likely to drive markets in the next quarter.
Investment factors, made up of macro and style factors, are well-established drivers of returns. Kestrel employs macro factors: economic growth, inflation, interest rates and liquidity.
The style factors that some other strategies reference include momentum, company size, value and volatility.
So much data is crunched that the job of gathering macro numbers originally took 16 days of computer time per country. This was eventually reduced to 47 minutes with the help of “multinational matrices”. Implementing a “relational data base” helps the team identify economic trends from the data, which are then filtered for logical sense and mapped on to 80 markets.
“We map growth to equities, inflation to bonds, liquidity to currencies and real rates to commodities. If something changes in line with changes in the factors, we ask ourselves how the authorities are going to react. We look for surprises and consider how the authorities would likely deal with them,” says Ricciardi.
This is where the discretional aspect of the strategy comes in – that portion where the human intervenes. And from this, long-only positions are taken in sectors rather than individual securities.
It was finding a commonality between value stocks at certain points in the investment cycle that persuaded Harindra de Silva that investment factors are a greater driver of performance than stock-picking. De Silva is president and portfolio manager of quant house Analytic Investors and believes the 2004 launch of low-volatility investment strategies made him one of the first factor investors. He sold Analytic to Wells Fargo Asset Management, where he now works, in 2016 and believes fundamental managers still do not recognise the role of factors in their returns.
The Global Long/Short Equity Fund he co-manages is a quant-based fund that aims for 8%-12% returns but with much lower volatility than the MSCI World Index. De Silva describes the data that Analytics Investors collects as voluminous. It includes standard data about company fundamentals and GDP, but also includes information about the economic sensitivity of assets.
“Some of the more obscure data we subscribe to is ESG data on emerging market companies,” he adds. The algo cleans the collected data and tries to predict appropriate weights for assets at a given point in the business cycle.
In the last quarter of 2018, the volatility factor came into focus, says de Silva. “In Q4, the only thing that worked was buying companies with low beta. Volatility reduction did really well and subsequently outperformed the index, especially strategies that shorted high-risk stocks.”
Active, fundamental managers are under more pressure to be transparent about what role investment factors play in their returns. But similarly, systematic investors are under pressure to be more open about their methods. “Investors don’t want to see the code, but they do want to know where the value is and that it’s sustainable,” says de Silva.
Mistrust of quants
Research by Cambridge Associates last year showed institutional investors were uncomfortable about quant strategies, mainly due to a perceived lack of transparency. Cambridge itself defended quants: its co-leader of European endowments and foundations, Simon Hallett, said it “may be true in a few cases” that some lacked transparency, but many could write down exactly how decisions were made and on what data.
He also said that quant strategies offered comparable returns and lower costs when compared with traditional active management strategies. (Discretionary-managed funds were best placed for “longer term and more idiosyncratic investments”.)
Median global equity fund fees for quant strategies were shown to be 55 basis points, compared to 75 basis points for fundamental investment strategies.
Investors traditionally perceive quants as being at risk of crowding into the same investment positions and crashing. However, Cambridge Associates said quants use algorithms that can express the typical strategies used by active fund managers “very efficiently” without suffering from subconscious biases, as they will not invest in anything that does not fall within preset parameters.
This point chimes well with Asif Noor, a portfolio manager at Aspect Capital, who says “systematic managers have fewer degrees of freedom. Our positions are on continuously.”
Noor runs global macro and foreign exchange portfolios at Aspect Capital. He joined Aspect in 2016 after he sold his own quant business - Auriel Capital Management, which he had co-founded with Anoosh Lachin – to the firm.
Fewer degrees of freedom for fund managers to make portfolio changes are the hallmark of rules-based, systematic strategies. With hyper-rational computers, managers have less leeway to tinker with portfolios because of the mood they are in.
But this lack of freedom requires a great deal of confidence in the investment hypothesis that underlies the rules. “Everything we do has to have a hypothesis and a rationale, not just to us but to our clients,” says Noor. “It is not good enough to just perform strongly in a given year. The client wants to know where you were making and where you were losing money. They want the rationale for it.”
An example of a hypothesis would be that the equity market with the highest earnings yield will outperform ones with lower earnings yields.
But what about the decision to abandon an approach like this when it becomes crowded? In the 1990s, strategies that compared earnings yields on a monthly basis worked really well until being arbitraged away in the 2000s. This strategy was replaced by algorithms looking at the earnings yields on sectors, which also got arbitraged away, Noor says. Next came a model based on analyst forecasts for individual stocks, but even that is arbitraged away now, he adds.
“When to run out of patience with a model is one of the hardest decisions to make. We look at several factors including the draw-down profile of the strategy versus its holding period.”
A model could be valid for three to five years and have 20 years of back-testing, says Noor. “We have to keep doing research, we have to keep evolving … the code we write helps move us from point A to point B in parameters that are adjusted once a decade.”
Using more sources of macro data means Noor’s portfolios now have shorter holding periods. Such periods could be 20-25 days long.
An example of alternative information sources is shipping data. Available from specialist vendors, this can be used as a faster predictor for trade flows in place of official import/export figures.
But as well as a wider range of data sources presented by Big Data, data-gathering tools are proliferating with the help of artificial intelligence and machine learning. De Silva at Wells Fargo says an innovation in the past ten years was for quants to use textual and audio recognition tools to automate the reading and analysis of corporate disclosures and to gauge sentiment from analyst phone calls. “We have the ability to take data feeds and link them to our processes and although active fund managers could do this, it will take them a while to incorporate this technology due to nervousness around the technology,” he adds.
But can’t investors be said to possess this same nervousness too? Surely this isn’t good for quants and similar computer-driven investors?
“There’s been a massive shift in what we call ‘algo-aversion’,” says Noor at Aspect Capital. “People are becoming more comfortable with the idea of driverless cars and they are becoming more trusting of data scientists. Investors also appreciate the greater level of transparency – they like that we can explain why alternative data sets can be better for our strategy.”
The proof of the pudding where algo is concerned is, as the saying goes, in the eating. Back at Kestrel, Ricciardi says the portfolio’s computer algorithm is 80%-90% accurate over the three months that are forecasted for. “Right now [June 2018], it’s telling us that US inflation will go down in August.”
Two months later, the annual inflation rate in the US indeed fell, to 2.7% in August from 2.9 percent in July. The wider market had also expected this, but in the latter half of 2018, US core inflation took an unexpected turn when it stopped rising toward the Fed’s 2% target and started falling toward 1.8%.
Computer-led investing is once more on the rise, aided in part by the wider awareness of investment factors and the colossal increase in data. Lower headcounts, pressure on active fees, even a reduced need for expensive big-city office space could lift it further.
The popularity of ESG (environmental, social and governance) will help. Wells Fargo’s Analytic Investors is not alone in ESG screening; Kestrel also says its 180-strong stock universe is ESG-checked, and recently Style Analytics – a firm founded by quants that helps asset owners identify what role investment factors play in their managers’ strategies – introduced a screen for ESG, showing that ESG could be an investment factor in its own right.
But for all the emphasis on computer-driven techniques, the industry would be nothing without its people – smart minds to create hypotheses; coders to give a hypothesis a digital existence.
The start-up frenzy means the pool of coders available to the mainstream financial world could be diminished. De Silva says finance houses are competing for talent with Silicon Valley - and it’s just a fact that financial firms are not going to offer their recruits stock ownership.
©2019 funds europe