High Frequency Trading in Commodity Markets
Book: Commodities: Markets, Performance, and Strategies
Editors: H. Kent Baker, Greg Filbeck, Jeffrey H. Harris
Publisher: Oxford University Press, 2018
ISBN: 9780190656010
From Pit Traders to Algorithms
Chapter 23, written by Raymond Fishe and Aaron Smith, covers the rise of high frequency trading (HFT) in commodity markets. And if you have ever watched old footage of traders screaming on the floor of the Chicago Board of Trade, this chapter explains what replaced all that chaos.
Before electronic trading platforms existed, “scalpers” or floor traders acted as middlemen. They stood in pits, shouting and waving hand signals, providing temporary liquidity until a hedger or speculator came along to hold the other side. Being loud and fast was the key to winning trades.
Then electronic platforms showed up. The CME launched Globex in 1992 for currency futures. The CBOT introduced “Project A” in 1996 for interest rate products. But the real wake-up call came in 1998 when Eurex, an all-electronic European exchange, opened and within a year surpassed the CBOT in contract volume. The London International Financial Futures Exchange (LIFFE) saw the writing on the wall after losing its Bund futures business to the electronic Deutsche Terminborse. LIFFE went from positive profits in 1997 to losing 64 million pounds in 1998. It closed all its open outcry pits by November 2000.
By 2015, the CME Group had closed almost all of its pits, including the historic agricultural pits that traced their origins back to 1865.
Who Are the High Frequency Traders?
HFTs are a subset of algorithmic traders who profit by sending and executing orders faster than their competitors. They are not all the same. The chapter distinguishes between several types of electronic traders:
HFTs seek speed above everything. They update quotes and monitor risk exposure in milliseconds. When new information hits the market, they react before anyone else can.
Smart order routing (SOR) algorithms are used by large institutions to break up big orders into smaller pieces. The goal is not speed for its own sake but rather to reduce market impact costs. These algorithms pace their activity to match the natural flow of liquidity.
Proprietary algorithmic traders use various strategies based on information or technical analysis, but speed is not necessarily their main advantage.
The data from Haynes and Roberts (2015) shows that automated trading systems (ATS) are dominant across most commodity futures. For E-mini S&P 500 futures, an ATS was on one side of 86 percent of trades. For WTI crude oil, the figure was about 76 percent. For corn and soybeans, it was lower at around 58 percent. The agricultural contracts still had more manual trading, possibly because relevant information in those markets still arrives through human networks.
What Do HFTs Actually Do in the Market?
Most research finds that HFTs primarily act as market makers. About 70 percent of HFTs appear to supply liquidity by posting passive limit orders on both sides of the bid-ask spread. But their market making is not purely passive. They tend to supply liquidity when spreads are wide and take liquidity when spreads are narrow. In other words, they discipline the market toward fair pricing.
One key detail: algorithms react to quote changes in 2 to 3 milliseconds. The fastest human reaction time, based on a study of 37 million mouse clicks, ranges between 200 and 250 milliseconds. Humans simply cannot compete with algorithms when it comes to quote matching.
The Good and the Bad
The research uses various exogenous events to isolate HFT effects from broader market trends.
The good stuff. When co-location facilities became available (allowing traders to put their servers physically close to the exchange), researchers found improved liquidity and tighter bid-ask spreads in both equity and futures markets. On the Australian Securities Exchange, the four most active futures contracts showed lower bid-ask spreads after co-location was introduced in 2012.
The not-so-good stuff. Weather events that disrupted microwave communication networks between Chicago and New York effectively neutralized the speed advantage of HFTs. During those disruptions, price impacts, effective spreads, and realized spreads all decreased. This suggests that in normal times, fast traders actually increase some trading costs through adverse selection.
During the Lehman Brothers crisis in 2008, HFTs actually increased their volume. But during the Cushing, Oklahoma oil storage shortage in 2009, HFTs pulled back. The difference? Cushing involved site-specific information that locals knew better. Lehman was a common shock with less asymmetric information.
The Flash Crash
The chapter spends significant time on the May 6, 2010 flash crash. Here is what happened: prices of E-mini S&P 500 futures dropped about 5 percent in four and a half minutes, starting at 2:41 p.m. They bounced back to near pre-crash levels within 23 minutes.
The trigger? A mutual fund (later identified as Waddell & Reed) launched a sell program for 75,000 E-mini contracts using an algorithm that targeted 9 percent of recent trading volume. The algorithm was set to ignore both price and time. It just kept selling.
HFTs were initially net buyers for several minutes before switching to net sellers. They traded more than 1.45 million contracts that day but rarely held more than 3,000 contracts at any time. They did not change their basic strategy. They just traded with each other.
A London trader named Navinder Singh Sarao was later charged with spoofing, which means placing orders with the intent to cancel them before execution. He pled guilty in 2016. But a deeper study by Aldrich, Grundfest, and Laughlin (2016) found that the flash crash was better explained by the combination of bad market conditions, a poorly executed large sell order, and timing problems in the equity data feed.
The Arms Race Problem
Perhaps the most thought-provoking finding is the “arms race” phenomenon. Millions of dollars are spent building faster communication networks for only milliseconds of improved speed. Budish, Cramton, and Shim (2015) argue that corporate finance decisions are unlikely to change based on these tiny speed improvements. The resources would probably be better spent elsewhere in the economy.
The optimal trading rate, according to Du and Zhu (2016), varies from 0.63 milliseconds for E-mini S&P 500 to 15.2 milliseconds for 10-year Treasury futures. Those numbers suggest that avoiding adverse selection effects on manual traders is basically impossible in the current environment.
My Take
This chapter is a solid academic survey, but it raises an uncomfortable question that it does not fully answer: who benefits from all this speed? The evidence suggests that HFTs have lowered transaction costs during normal market conditions. But during stress events, they may withdraw liquidity exactly when it is most needed. That is a real problem for hedgers who use commodity futures to manage actual business risks.
The flash crash section is especially interesting because it shows how a human decision (ignoring price in an execution algorithm) combined with electronic speed created a disaster. The machines did not go rogue. A person made a bad choice about how to use the technology.
Previous: Food Price Volatility: Causes and Solutions
Next: Energy Risk Management: Protecting Against Price Swings