Berkeley study finds scarce evidence of market 'front-running'

NEW YORK (Reuters) - Two University of California, Berkeley professors have taken a deep dive into new data and found that claims in Michael Lewis’ bestseller “Flash Boys” that retail investors are being gouged, or “front-run,” by high-speed traders not to be true - at least not now.

Author Michael Lewis smiles during an interview at Reuters regarding his book about high-frequency trading (HFT) named "Flash Boys: A Wall Street Revolt," in New York in this file photo dated April 3, 2014. REUTERS/Lucas Jackson

Professors Robert Bartlett and Justin McCrary said their findings contradict the common belief that fast traders systematically exploit others who rely on public data feeds, which in the past were notoriously slow.

“Flash Boys: A Wall Street Revolt” touted two theories of market abuse that the study, “How Rigged Are Stock Markets? Evidence from Microsecond Timestamps,” disproves, said Bartlett, a securities lawyer.

One theory is that market-makers such as Citadel LLC cheat customers by not giving them the best price available, he said. “It turns out that’s just not right,” he said.

The Justice Department has subpoenaed information from Citadel and rival market maker KCG Holdings Inc KCG.N related to their execution of stock trades, Reuters reported in May, citing people familiar with the investigation.

The other theory asserts that traders using faster data know when a stock quote becomes “stale,” run ahead to buy the security and immediately sell it back when the pubic feed updates. This practice allegedly allowed high-frequency traders “risk-free” opportunities to pick off orders.

“That’s not happening either,” Bartlett said.

The study found little evidence that users of a slower feed of quotes and trade prices transmitted by Securities Information Processors, or SIPs, were disadvantaged. In fact, traders pricing off the SIP gained on average of 3 cents per 100 shares.

The study examined nine months of quote and trade data on the 30 stocks that comprise the Dow Jones industrial average.

The Securities and Exchange Commission last year ordered the SIP data be time-stamped to the microsecond and carry the time when the exchanges transmit to their direct feeds. The two data sets can now be aligned and the latency - the time it takes for the data to travel - can be compared for the first time.

A prior inability to fully assess the data had hampered understanding the extent to which front-running, or “latency arbitrage,” actually occurred, the professors said.

About 97 percent of trades occur at a time when data from both feeds match. For the remaining 3 percent of trades, less than 1/10th of that left a liquidity taker in a worse position, they said.

“It’s a clean and crisp study on the subject of are liquidity takers being harmed,” said David Weisberger, head of trading analytics at IHS-Markit. “Their methodology is quite good and uses conservative assumptions, which means their numbers, if anything, overstate the problem.”

Reporting by Herbert Lash; Editing by Leslie Adler