Quants are turning out to be the false friends of human active managers.
Quants, or quantitative investment strategies (QIS), which use technology and formulas to automate the investment process, were originally seen by traditional active managers as a means of generating higher returns, and thus fending off the assault of passive investors, who have seen an inflow of $2.5trillion since 2000 – about the same amount that has been removed from active funds in the same period.
The main reason for this seismic shift of funds is the fact that passive funds are tax efficient, transparent and their fees are up to 80 per cent lower than those of active funds. Although passive funds, such as exchange-traded funds (ETFs) mirroring indexes such as the S&P500 or Russell2000, don’t promise to beat the market, they do deliver against their more modest benchmark, while the success rate of active managers – as was confirmed by the latest Morningstar Active/Passive Barometer – “was less than 25 per cent in more than half of the categories surveyed over the 10 years through June 2018”.
Until 2017 experts still expected active strategy to rebound, and maintained that traditional stock pickers need to become centaurs leveraging quant investment techniques in order to be able to reduce their fees and become more competitive – as if active managers didn’t have it hard enough.
Instead, what we see is Blackrock, the exemplar of passive investing, sack seven of its fund managers and restructure its underperforming stock-picking business through the acquisition of SAE, a $100billion computer-powered quantitative investment unit. Blackrock envisions SAE, its Systematic Active Equity (SAE) arm to operate as an incubator and develop techniques that will transform its entire investment management business. In accordance with a common quant-biased formula stating that one computer engineer can do the jobs of four traders, SAE employs 30 PhDs in computer science, physics and engineering. It is the analysts not traders that propose an investment idea, which at Blackrock is still dissected by old-guard human “referees” to test its viability.
A major milestone down the road of gradually supplanting traditional stock managers were the almost simultaneous launches of two AI-driven ETFs a year ago. The first one is AI Powered Equity ETF, which, between its inception in October 2017 and 31 August 2018, returned a promising 19.20 per cent compared to 12.12 per cent for S&P500.
AI Powered Equity ETF by EquBot uses IBM’s Watson Supercomputing to analyse the data of 6,000 publicly traded companies, ranks investments based on their “probability of benefiting from current economic conditions, trends, and world- and company-specific events”, and picks those with the best chance at outperformance in order to build the perfect portfolio of 30 to 70 stocks.
Although Toronto-based, robot-run equity ETF firm Horizon was launched a month later than AI Powered Equity ETF, it was the first to have gone global. Its AI had been back-tested over 10 years of market data and its management fee at 0.55 per cent is about half of what traditional stock pickers would charge.
How can quants gain more traction?
Fully automated quants, however, are still in their infancy and have issues of their own. With data protection making great strides, quants using AI to scour the internet for massive amounts of data to outdo humans – including satellite photos, social media accounts and data generated by credit card transactions – are an acute data protection concern. A common technique that QISes increasingly use to better comply with regulations is “differential privacy”, which entails adding noise data into the data mix in order to obscure personally identifiable information (PII) without destroying the useful features of data-sets.
“Fundamental discretionary traders account for only about 10 per cent of trading volume in stocks today.”
Another frequent charge levelled against quants is that their investment decisions are not transparent (a shortcoming they share with their human counterparts), which has led to the emergence of a new field called explainable AI, where a second layer of artificial intelligence is incorporated into the system to examine the first and work out how it came to make investments in a certain way.
Records have shown that human stock pickers came into their own and managed to consistently outperform their passive rivals only in times of big financial crises, when active managers have more wriggle room to sell and buy shares from their portfolio, while passive investors stay invested even in seriously underperforming securities.
However, they seem to be losing that edge too to the new, artificial breed of investment managers. Although some recent market volatility has already been laid at the door of AI-driven quants, and there is general anxiety about how fully automated investment funds can wreak havoc on the global financial system in the future, the first surveys seem to show that AI, due to lacking emotions and biases, will be quicker at correcting erroneous investment decision and rebalancing the market when it is getting out of kilter: a quality that has the potential of winning over sceptics.
The jury is still out
It is still early days to assess the long-term value that AI can bring to investment management. Its short, triumphant march so far is tempered by algorithmic glitches and casualties such as Sentient Investment Management, an automated hedge fund that recently closed after only two years in operation. But if JP Morgan’s estimate that “fundamental discretionary traders account for only about 10 per cent of trading volume in stocks today” is anything to go by, the odds are strongly against the long-term survival of the human active manager.
But passive trading is ready to up its game: investor guru Jim Rogers has already launched his Rogers Ai Global Macro ETF. Although in passive investment there is less room for leveraging AI ingenuity (Roger’s ETF will mostly follow US-listed single-country ETFs), it can lead to even lower margins by improving the efficiency of execution and rebalancing. So, although the antagonism between active and passive investment is here to stay, this time we shouldn’t expect a rivalry between humans and machines but rather a race between intelligent learning machines.