How high-frequency algorithmic trading programs can make bad stock market days even worse
The stock market has had one of its most tumultuous months on record, with the S&P 500, Dow and Nasdaq all soaring to new highs in mid-February only to crash to within a whisker of bear territory on Monday. Market watchers once again are casting a suspicious eye on the role of high-frequency algorithmic trading in exacerbating the slide.
Algorithmic trading, where a computer automatically executes trades based on pre-programmed instructions, has been around for a long time and is now a big factor in the daily ups and downs of the stock market. The days when you would call your broker to instruct a human to place the trade are mostly gone.
A computer makes “buy” or “sell” decisions quickly and dispassionately, unburdened by emotion or instinct that might cloud human judgement.
But, when the markets fall dramatically, as they did on Monday, algorithmic trading is often accused of magnifying the market slump, and fueling investor panic. When stop-loss limits are triggered en masse, for example, it can lead to a snowball selling effect, sending a market into a downward spiral.
In the midst of Monday’s historic sell-off, some markets observers were pointing a finger at algorithmic trading as a possible cause, calling it a “dangerous accelerant of volatility.”
It’s notable, however, that you rarely hear criticism of computerized trading when stocks are booming.
Guy De Blonay, a fund manager at Jupiter Asset Management, told CNBC in 2018 that 80% of daily moves in U.S. stocks were machine-led, while Marko Kolanovic, global head of quantitative and derivatives research at J.P. Morgan, said in 2017 that “fundamental discretionary traders” accounted for only about 10% of trading volume in stocks, compared with 60% for passive and quantitative investing, which uses algorithms to make investment decisions.
The advantages of algorithmic trading, typically used by institutional investors and hedge funds, are speed of execution, lower trading costs and sticking to a consistent strategy.
An algorithm might be designed to momentum strategy—that is buy stocks that are rising, or sell shares that are falling. Or, the software is programmed to buy or sell shares that have broken above or below their recent trading range.
And so when the markets lurch significantly lower, the software will quickly execute big sell orders. The U.S. markets have so-called staged circuit breakers to halt trading when shares surge down 7%—and later by 13% and 20%. That’s precisely what happened a few minutes into the trading session on Monday.
…or markets shock-absorber?
Algorithmic trading also enables investors to make lucrative arbitrage trades by identifying tiny differences in the price of assets, perhaps profiting from exchange rates fluctuations.
Computers can also be programmed to react instantaneously, and in a set way, to timed releases of economic data—think jobs numbers or Fed interest rates moves—which can magnify their impact on markets.
The software can also be used to avoid big hiccups during the trading day. For example, big institutional investors, such as pension funds, making a large stock purchase, may break up their order into a lot of smaller orders, using automated trading software to avoid driving up the price of the shares.
Algorithmic trading is increasingly being coupled with machine learning to create ever more sophisticated automated investing.
Investment bank J.P. Morgan said last year it was applying machine learning to provide competitive pricing, and optimize execution in the $6.6 trillion-a-day foreign exchange market with its Deep Neural Network for Algo Execution (DNA).
Meanwhile, investors are increasingly turning to robo-advisors, which use algorithms to create and manage online investment strategies tailored to an individual’s goals.
A sub-set of algorithmic trading is high-frequency trading, where investors buy or sell shares in a fraction of a second, seeking to profit from tiny fluctuations in prices. High-frequency trading brings liquidity to markets but the practice is controversial as it’s blamed for contributing to a number of unexplained market crashes.
On May 6, 2010, around $1 trillion was wiped off U.S. stocks as the Dow Jones Industrial Average plunged by nearly 1,000 points in a bizarre “flash crash” before recovering most of the losses.
An official 2010 report said the crash happened on a nervous day in the markets when a mutual fund sold $4.1 billion of EMini S&P 500 futures contracts via an automated execution algorithm in 20 minutes, precipitating a liquidity crisis in the EMini market.
Since then, algorithmic trades are thought to have magnified a number of flash crashes in the foreign exchange market.
Michael Lewis, in his 2014 book “Flash Boys”, alleged that high-frequency traders used split-second advantages provided by ultra-high-speed communications to make billions at the expense of other market players.
Since then, a growing number of exchanges have created “speed bumps”, tiny delays in executing trades, to blunt the advantage of the high-frequency traders, as The Wall Street Journal recently reported.
More must-read stories from Fortune:
—Why plunging Treasury yields are so alarming
—Furious Robinhood customers want payback following two day outage
—Why investors suddenly turned on pot stocks
—With stocks down sharply are we approaching”buy” territory? Not by a longshot.
—Why it’s so hard to find the next Warby Parker
Subscribe to Fortune’s Bull Sheet for no-nonsense finance news and analysis daily.