For high-frequency traders, the speed of light may not be fast enough. Right now, technical limitations don’t allow computer systems to decrease latency, the ever-diminishing horizon between the time a trader makes the decision to trade and when the trade is made. But trading at light speed may not be far off.
The last decade has seen a big increase in the incidence of high-frequency trading, with traders buying and selling in high volume at lightening-quick speeds. Trading firms invest significant amounts of money in sophisticated computer systems in their efforts to drive latency down to millisecond or even microsecond timescales. Costly as these systems may be, they are considered necessary to maintain a competitive advantage in the fast-moving modern electronic marketplace.
Recently, latency has received attention in the context of so-called flash orders. Some financial exchanges sold milliseconds-in-advance notice to select traders, who could observe and act on certain buy and sell orders before those orders were revealed to the broader marketplace. While they were technically legal, the orders prompted calls for regulatory intervention, but NASDAQ and other major exchanges voluntarily stopped offering the services before regulators took any action. In mid-April, the SEC proposed a trader ID system to help the agency monitor high-frequency traders.
“The discussion around latency thus far has been somewhat ad hoc,” says Professor Ciamac Moallemi, pointing out that latency is poorly understood. “For example, some have asserted that only high-frequency investors are affected by latency and that, for example, pension funds, which have very long timescales, are unaffected.
“But latency is important for all investors: if you can lower latency you can lower transaction costs,” he says. “Pension funds would like to lower their transaction costs as much as high-frequency traders would.”
Moallemi worked with doctoral student Mehmet Sağlam to compare different theoretical trade scenarios, with and without latency, and used the results to create a simple quantitative model that can be used to value latency.
“The main conclusion is that we should view latency as a trading friction. So for a single individual, lower latency will reduce transaction costs. To the extent that high-frequency traders trade more often than pension funds, in dollar terms a high-frequency trader is more affected. But every trader pays transaction costs; everyone would benefit from their individual latencies being lowered.”
The implications of ever-decreasing latency for the financial system remain unclear. High-frequency traders have argued that lower latency allows them to be more active and post orders closer together, lowering bid-offer spreads, which should benefit other investors and create greater liquidity.
“To some extent, that is probably true,” Moallemi says. “But is their ability to take advantage of low latency coming at someone else’s expense? Less-sophisticated investors who can’t make the technical and R&D investments to exploit lower latency may be at a disadvantage.”
While individual investors might benefit from the ability to trade at ever-increasing speeds, many worry that this phenomenon will force all investors into a spending race to keep up, lest faster traders take advantage.
“We should take a step back and think about what financial markets are supposed to do,” Moallemi says. “Financial markets are supposed to provide a mechanism for investors who have excess capital to deliver that capacity to companies and entrepreneurs who can employ it. It is not clear that an entrepreneur can more effectively raise capital because investors can trade in microseconds.”
Ciamac Moallemi is assistant professor of decision, risk and operations at Columbia Business School.