MIT Prof.: Flash Crash jitters--HFTs were not the cause

MIT Sloan professor finds latency affects price formation process in financial markets

CAMBRIDGE, Mass., Feb. 4, 2015 – In the last decade, trading floors have been supplanted by server farms and human traders have been replaced by anonymous algorithms. This transition came with the promise of using faster and cheaper technology to drastically lower execution costs and improve price discovery for all market participants. However, this transition also led to the emergence a new breed of secretive, hyperactive trading algorithms called high frequency traders (HFTs). In the latest of a series of studies on HFTs, MIT Sloan School of Management Prof. Andrei Kirilenko discovers that latency – the delay in signal processing by automated trading platforms – can offer important insights into the price formation process in modern automated markets.

Kirilenko began studying HFTs after the Flash Crash of May 6, 2010. “That was an eye-opening systemic event in which markets in the U.S. moved dramatically down before bouncing back over about 30 minutes,” he says, noting that he led the investigation of the Flash Crash for the Commodity Futures Trading Commission (CFTC). The crash was initially blamed on the HFTs’ fast and destructive release of energy into the market. However, Kirilenko and his colleagues found that HFTs did not cause the Flash Crash. They did, however, accelerate the price move and help make the event systemic.

In a subsequent paper on the systemic role of HFTs, Kirilenko found that the HFT industry was dominated by an oligopoly of fast and aggressive incumbents who earned high and persistent profits while taking little risk. For some reason, competitive market forces were unable to break up the oligopoly, and benefits of automated markets were not being fully realized by all market participants. Instead of competing to provide best execution to customers, he says that incumbent HFTs seemed to be engaged in a “winner-takes-all arms race” for small reductions in latency.

“The lack of competitive market dynamics made me concerned about the price discovery process in markets with HFTs,” he says. “I realized that it’s not the HFTs per se, but the responses of slow traders to their presence as they try to compete for their survival that affect prices and liquidity in the market. The mere presence of HFTs can make transaction prices more volatile even in the absence of any fundamental information.”

In another paper, he and coauthors theoretically showed how market quality can be systematically affected by the presence of traders operating on different time scales or latency. “These two studies made me realize that latency, the delay between a signal and a response, plays an important role in the price discovery process of automated markets,” he says.

A recently completed study on trading platform latency co-authored by Kirilenko, offers a number of insights. It turns out that trading platform latency is a highly variable stochastic process rather than a constant number. In principle, this can make some algorithms overreact or underreact to changes in market conditions.

“Effects of this sort, if any, should show up in market prices, especially in price volatility. Indeed, latency, and especially the dispersion of latency -- which proxies for something called jitter -- does turn out to have a predictive power over both volatility and the volatility of a liquid asset,” he observes. In other words, studying latency can provide the clues to a better understanding of the price formation process in modern automated markets.

He points out that this finding explains why many regulators and policymakers are focusing on latency measures to slow things down so as to remove this advantage of HFTs.  “But despite their good intentions, if applied without a solid understanding of the entire ecosystem of market participants and their trading strategies, these proposals could possibly result in extra costs and risks to the very participants they are designed to protect,” says Kirilenko, who testified on automated and high frequency trading before the U.S. Senate.

“Instead of new regulations, we recommend increasing latency transparency,” says Kirilenko. “Trading platforms should report characteristics of latency to market participants on an ongoing basis so that any valuable nontrade information contained in latency can be discovered along with asset prices.”

He adds, “We also need additional studies to better comprehend the price formation process in our new automated financial markets. Without a new level of understanding, we won’t be able to build safeguards to protect against the type of energy release that led to the Flash Crash much less keep market participants on an even playing field.”

Andrei Kirilenko is co-director of MIT Sloan’s Center for Finance and Policy, and coauthor of “Latency and Asset Prices.” https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2546567