At his coming-out hearing as chairman of the Federal Reserve on Feb. 27, Jay Powell made all sorts of news in finance-land, including a suggestion that the bank saw potentially faster inflation ahead. Also notable was his assessment of the causes for the volatility that roiled Wall Street and saw trillions of dollars lost, gained, lost, and then regained in a matter of days in early February. In wonk speak, Powell remarked that he didn’t think that ETFs—exchange-traded funds—were a particular culprit, though he conceded that the issue deserves further study.
Powell’s reassurance notwithstanding, it is at the least too soon to draw a conclusion. Too much has changed too quickly in the past few years to say with any confidence that we understand the interplay of humans and machines as applied to markets. The trading of ETFs—particularly when it is rapid and automated—is but one of many concerns about the functioning of stock and bond markets. Put simply, transactions are now governed less by people shouting orders and pushing paper and more by software and computers. That shift, affecting tens of trillions (yes, trillions) of dollars globally, merits more attention than it currently receives.
The rise of ETFs highlights the increasing role of technology in markets. ETFs are low-cost baskets of stocks or bonds that mirror indexes or reflect investing themes, such as semiconductor makers or global consumer stocks. They can be bought and sold like stocks, and now account for as much as 30 percent of all US stock trades. But ETFs trade as units. If I own an ETF that mirrors all large US companies and I decide to sell, a tiny piece of every company in that basket gets sold. The same is true for traditional mutual funds, which have been around for decades, but mutual funds can only be traded once daily. ETFs are traded in fractions of a second, which means that every company with listed shares or bonds can also be traded in fractions of a second, as quickly as a computer program can process the data. Those programs, and the algorithms that drive them, are beginning to upend and distort the multi-trillion-dollar business of buying and selling stocks and bonds.
For nearly two years, global stock markets were calm. Eerily calm. Between February 2016 and February 2018, US stocks climbed steadily and never suffered a drop of more than a few percent. US politics were dramatic, as were global crises, but financial markets, after years of turmoil following the financial implosion of 2008, were placid.
In early February, the calm ended, in a spectacular fashion. In rolling waves of frenetic selling and buying, the Dow Jones Industrial Average moved up and down hundreds of points within hours, as did other major global indexes. For now, the frenzy seems to have subsided. But those weeks raised concerns that have been building for some time and are not yet understood.
Markets go up and markets go down; so it has ever been and likely will be. What’s new is the peculiar and still unclear role of trading done not by people but by algorithms and programs, executed in milliseconds with no short-term human agency.
Before ETFs, which have only become a substantial portion of the market in the past few years, there were certainly market panics and collapses. But the recent frenzy should be a wake-up call that technology is altering financial markets as dramatically as it has other segments of society, and we’d best figure how to understand and control it.
ETFs alone wouldn’t change the equation absent the rise of software programs and algorithms that trigger trades under various conditions. The recent bout of selling and then buying happened so fast in part because a few programs generated automatic buy and sell orders at the rate of thousands per minute based on the algorithms that control them. Because ETFs are baskets, and there are now thousands of ETFs, the turnover in shares triggered by algorithms can be exponentially larger than when each trade required a human-generated physical ticket. The combination of ETFs and algorithms means that markets can turn over in minutes, rather than days. Tech here, as elsewhere, accelerates everything.
That’s why, for instance, prices could swing by 10 percent on trillions of dollars of stocks in the space of a few hours. Yes, there have been past crashes when everyone sold at once. But the difference last month was that there was massive selling and then buying in a few hours without people making decisions. Human herds can stampede for the exits and cause a stock market crash or a bond market freeze, but human herds of traders do not reverse course in the midst of a stampede and then turn around and start buying again in the space of minutes. Algorithms do.
Computer-generated trading programs are, of course, created by humans who write their instructions. These are often not on company fundamentals but on the basis of market movements. To put it differently, many programs are designed to sell when prices start going down and then buy when prices fall to a specified level. The reality is massively more complex, sweeping in structured derivatives that might be designed to generate twice or three times the returns the return of an index or sector, or generate the inverse, as well as an increasingly robust option market of future obligations to buy or sell that are themselves bought and sold in rapid fire.
The result is that at any given time, a majority of the market is now determined not by humans making decisions but by computers trading with one another based on programs. There are no hard and fast numbers, though JP Morgan recently estimated that only 10 percent of trading now consists of people trading with people based on fundamental decisions about company A or company B.
With more money pouring into ETFs and more trading dominated by algorithms, the essential nature of stock and bond markets is morphing. For now, it isn’t clear into what. So far, the effect of the machines has been to speed up cycles of selling and buying, so that you can now have a stock market sell-off and recovery in days instead of weeks and months. That in itself is no big deal assuming that everything returns to some level of stability once those computer-generated storms have passed.
But what if they don’t pass? What if the programs break down or run amok? No regulatory agency in the world has figured out how to contain the risks of the new mix of algorithmic trading, high-frequency trading, and investment vehicles like ETFs. For the much of the past decade, those agencies have been focused on preventing a repeat of the last financial crisis, and like all regulatory bodies, they tend to be more reactive than proactive. The same is largely true for the world’s largest banks.
The prospects of programs trading with programs should alert us to a world where there are very few breaks on collapses and surges, and where current circuit breakers are wholly inadequate. Here as elsewhere, humans tend to need a crisis to spur action, but here unlike elsewhere, we have ample tremors warning that the big one might be coming. The rise of the machines need not be a threat, but unless we plan for it, the entire financial system is at risk.