Does High-Frequency Trading Cause Wild Stock Swings?

A new study claims that robot traders make the market less volatile, but the evidence disagrees

600 computer trader2 REUTERS Christian Hartmann.jpg

The crazy market swings we saw last week were largely blamed on so-called high frequency trading. That's when computer bots are programmed to automatically buy or sell stocks based on a set of criteria that quantitative traders establish. Critics of this practice say that volatility is worsened by robot traders, which cause momentum to build quicker through lightening-speed trading and cannot as easily exercise restraint. But a new study finds that these bots actually make the market less volatile. Is this plausible, given what actually occurred last week?

So how is high frequency trading supposed to have exacerbated volatility last week? The idea, simply, is that positive or negative news caused stocks to rise or fall more than usual, and as momentum built, triggers within trading algorithms were hit, causing bots to push price movements even farther. This, in turn, led to even more momentum and bigger increases or declines in prices. This sounds reasonable enough, but maybe it has the role played by bots all wrong.

In the Wall Street Journal, Michelle Price reports on a new study that presents an alternate theory of how robot traders affect volatility. The study is by Australia-based Capital Markets Cooperative Research Centre. Alex Frino, professor of finance and chief executive of the group explains:

Part of this function is the way (high-frequency trading) algorithms identify trading opportunities--they're built to recognize when prices are abnormally high or low, and their response to this naturally pushes prices back towards equilibrium.

Certainly, well-designed algorithms would lead bots to take advantage of under- and over-priced stocks by buying or selling when broader market conditions also dictate doing so. So this assertion makes sense. But as Frino says, this is only "part" of an algorithm's function. Any algorithm that senses momentum building in one direction or another may also respond accordingly by making sure that it doesn't endure big losses or miss prime buying opportunities. This would increase volatility.

So which is it? We would need to examine the algorithms to know for sure. But it might help to look at how the algorithms actually performed during last week's wild trading. A separate article by the WSJ's Jenny Strasburg says that the volatility produced "record profits for high frequency traders following months of disappointing returns."

Now let's think about Frino's claim. If these bots did better than slower traders, as this second article states, then they must have led the market's movement. For example, anyone selling too late would have lost more money than a robot selling sooner as prices decline. Similarly, anyone buying after prices had already begun rising wouldn't do as well as a quicker robot that bought first.

Since high frequency trading does better than the market's slower movers when volatility is high, it's hard to swallow the claim that it actually reduces volatility. Instead, its trading creates more momentum that slower traders are forced to respond to, which is precisely what leads to bigger price swings.

With that said, Frino's claim makes sense in a market that isn't being hit by big news or unforeseen shocks. As uneventful market conditions persist, algorithms would likely prevent volatility for precisely the reason that Frino states. They would be written to capitalize on quick arbitrage opportunities for mispriced stocks, which would cause price equilibrium faster. But when the volume is soaring, those same algorithms likely move based on other triggers being pulled, likely making volatility even more severe.