Slow it down, folks.

Photographer: Matt Cardy/Getty Images

High-Frequency Traders Need a Speed Limit

a | A

In today's electronic financial markets, a single investor can execute more than 10,000 trades a second, meaning that more than 1,000 trades can happen in the blink of an eye. Electronic trading firms are willing to spend hundreds of millions of dollars to shave off even millionths of a second. This raises an important question for human welfare: If real economic activity depends on people doing things, and people can't possibly think or react this fast, who benefits from all the speed?

Judging from new research, hardly anybody does. The markets may be running too fast for their own good.

Trading on Speed

Today's exchanges operate in a way that lends itself to speed: Computers place orders to buy and sell, then other computers execute the orders as soon as a match can be found. Such “continuous” trading, though, isn’t the only way to organize a market. An alternative is to use “batch auctions”: An exchange would accumulate buy and sell orders over a discrete period of time, then match them all at a price that balances supply and demand.

Some earlier research has suggested that batch auctions might eliminate the fleeting arbitrage opportunities that drive the current arms race in trading speed. Now, two researchers -- Austin Gerig of the U.S. Securities and Exchange Commission and Daniel Fricke of Oxford University -- have shown in a detailed analysis of batch execution that the optimal time period between trades should be something close to half a second. In other words, there's a speed -- and it's not terribly fast -- beyond which more is harmful, not beneficial.

Compared with the markets of two decades ago, when human brokers could take half a minute to execute an order, faster trading has some obvious advantages. If you put in an order buy when Apple stock is trading at $106.47, you don’t want it to sit around too long, lest the price jump above what you think the stock is worth. The longer the trading interval, the greater the chance that investors, even if they have good information, will get burned by random fluctuations.

On the other hand, too much speed isn't good, either. For a batch market to work well, the exchange needs time to accumulate enough orders to get a good picture of supply and demand. Otherwise, a few unusual orders can distort the price, again exposing investors to undesirable risks.

Somewhere in between is the sweet spot, the optimal trading interval that would make markets work best for investors. Looking at historical data for U.S. stocks, Gerig and Fricke find it to be somewhere between 0.2 and 0.9 seconds -- that is, no faster than about 5 trades a second. The exact number would vary by security, depending on characteristics such as volatility, trading volume and the extent to which the price moves in sync with the market as a whole. (Here's a more technical discussion.)

As the researchers put it, their arguments suggest that “speed is important in U.S. markets and that time delays of even a fraction of a second can harm market quality,” but millisecond and microsecond speeds are “unnecessary.” Gerig and Fricke's result isn't definitive, as real markets differ from the simple batch execution scenario they examine. Hence, the optimal speed for real markets might differ a little. Still, their work offers good evidence that trading has gotten too fast, and that slowing down would be desirable.

So here's one idea for ending the wasteful arms race among high-frequency traders to achieve ever faster execution: Implement a kind of speed limit, perhaps at 0.1 second, and require exchanges to run a batch clearing system. The time interval might even be adjustable, rising and falling with measures of market volatility.

A speed limit on trading would probably be bad news for companies currently profiting from the technology-fueled microsecond scrum over information. But markets are supposed to serve investors and the companies they support, not financial intermediaries.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

To contact the editor on this story:
Mark Whitehouse at mwhitehouse1@bloomberg.net