As the Dow Jones industrial average collapsed 700 points in 20 minutes Monday afternoon and the stock market jerked from bad to cataclysmic, traders and analysts coalesced around an increasingly routine explanation: Blame the machines.
Lightning-fast trading models, automated sell orders and an arsenal of sophisticated algorithms may not have driven every bit of misery in what became the biggest stock-market point drop in U.S. history.
But investors and market experts say they are likely to have made a crazy trading day that much crazier, spooking anyone with a retirement fund and sparking an outburst of panic selling — and making Tuesday's rebound seem even more baffling.
The market is always “just one step away from massive volatility because of programmed trading,” said Michael Yoshikami, the chief executive of Destination Wealth Management, an investment-management firm in Walnut Creek, Calif. “There’s no way that investors can compete with a computer making 1,000 trades a second. What it does is it ramps up the psychology of fear and greed for individual investors.”
Algorithmic trading — proprietary computer programs that can perform thousands of trades a second — has become a reality for global markets. Marko Kolanovic, JPMorgan Chase’s global head of macro quantitative and derivatives strategy, wrote in a report last year that passive and algorithm-driven “quantitative investing” accounted for about 60 percent of stock trades, compared with 10 percent from more traditional trades.
Treasury Secretary Steven Mnuchin said Tuesday that algorithmic trading “definitely had an impact” in Monday’s 1,175-point Dow drop. But Tuesday’s trading also showed just how quickly the algorithms can jostle the markets and disappear: The Dow closed up 567 points.
It is impossible to know yet the size of the effect that algorithms had in Monday’s sell-off and in the rebound. “The only way to know is with audit trail data,” said Eric Scott Hunsader, a software developer with the market-data firm Nanex, and “they keep that under lock and key.”
Market analysts said the market swings were not the fault of computers alone: Jittery trading, sky-high stock values and other unnerving market indicators had set the stage for the broader market’s sell-off. Kolanovic wrote in a note to clients that “trend-following strategies” had likely turned negative becaue of last week’s stock slowdown and a quick increase in the “fear gauge” that measures the market’s predicted volatility.
But Monday’s abrupt fall — which followed months of rising markets and engineered at superhuman speeds — had many analysts remembering the 2010 “flash crash,” another breakneck fall and rebound blamed on the chaos of unchecked automated trades.
Most trading algorithms respond to the same evidence people do, selling, for instance, if they recognize certain triggers. Analysts hypothesized that algorithms may have responded to Monday’s trades falling below the market’s 50-day moving average or as measures of market volatility grew.
If an algorithm appears to sell rashly or emotionally, analysts said, it is because a human taught it to do so. Some use their mathematical speed to exploit tiny fluctuations in price, buying and reselling within a fraction of a second. Others attempt to predict how markets will move by analyzing a torrent of data points, including indicators in labor-market news releases and the wording of politicians’ speeches.
Yoshikami, who works near Silicon Valley, said he knows designers who are building algorithms to gauge the speaking tone and inflection of Federal Reserve officials, looking for signs of how people will trade.
Algorithmic trading does not just ding the broader indexes, but it also affects individual companies. One analyst pointed to the stock prices for McDonald’s and Johnson & Johnson — two very different companies that both nevertheless showed the same sudden, violent dip.
The computers react to evidence exponentially faster than any human — think millionths of a second, instead of minutes — and can move en masse, trading at high volumes around the world. That makes them potentially more dangerous: A group of financial chiefs from seven countries wrote in 2015 that algorithmic trading has “caused significant volatility and market disruption” and that its growing complexity has increased “the potential for systemic risk ... over very short periods.”
Investment managers say the algorithms’ cold calculations end up sparking hysterical sales among the humans, undermining confidence and feeding a vicious cycle that leads more and more algorithms to do their thing.
“The machines started the sell-off yesterday, but now retail investors are calling up their advisers and saying, ‘Let’s make some profits or let’s get out,’” said Heather Zumarraga, a vice president of Vision 4 Fund Distributors, a distributor of mutual funds. She advises people to do what the algorithms are not doing: Don’t react.
“The retail investor can’t act so quickly, so drops like this make you feel really nauseous,” she added. “To see these kinds of volatile, wild swings, it just makes you think: ‘I don’t know if that’s the safest place to put your money if you’re a long-term investor, and this is what happens in one hour or one day.’”
Investment advisers say the algorithms can contribute to a climate of fear when they are involved in helping to drive the market down. Jon Ulin, the managing principal of Ulin & Co. Wealth Management in Boca Raton, Fla., said his phone has been constantly ringing with clients unnerved by such a sudden drop. He told them a lot of the trades came from algorithms, but insisted they use it as a wake-up call for their own investments.
“People were panicking and not recognizing it wasn’t just people calling up their brokers and yelling, ‘John, sell everything,’” Ulin said. (The “bullpens” of shouting floor traders, once the predominant image of Wall Street, have devolved largely into the set dressing for cable TV.)
Few analysts expect this new reality of high-speed, high-data trades will change, save for a crackdown from government regulators or the stock exchanges themselves, which make money from fast-paced trading by selling access to by-the-microsecond data feeds of market activity. The Nasdaq exchange made $156 million, or a quarter of its revenue, in the most recent quarter from “information services” including selling trading data.
But critics — including those named in “Flash Boys,” Michael Lewis’s 2014 book on high-frequency trades — say that computerized trading can end up rigging the markets in favor of super-fast trading firms at the expense of everyone else.
“It is not your father’s stock market anymore,” said Sal Arnuk, a co-founder of the New Jersey equity brokerage Themis Trading. “The exchanges don’t care about those price swings at that speed … but investors do care. Investors see that action and want to exit.”
“The market structure is responsible for exacerbating scary, jaw-dropping price moves,” Arnuk added. “We are used to it, but perhaps Main Street is not.”