Last week’s publication of Michael Lewis’ Flash Boys has focused new attention on high-frequency trading (HFT). Fortunately, economic research has illuminated many issues associated with HFT, and recent work has proposed remedies that could limit the costs while retaining key benefits of HFT.
What is HFT? It refers to the rapid electronic bids, offers, and executions – where the time interval is often measured in nanoseconds (billionths of a second) – that have come to dominate secondary market trading in U.S. stocks over the past decade. On many occasions in history, astute traders have used new technologies to gain an edge in asset markets. Easley and others identify a variety of such innovations, ranging from the telegraph and telephone to the computer monitor. HFT can be viewed as another step in this progression.
Advantages of high-speed electronic trading are well known. In a fully efficient market, customers can see orders and watch as they are executed at low cost. Decentralization of electronic networks seems like a pretty good thing. It both increases competition, which should lower costs, and reduces some operational risks, like the shutdown of the New York Stock Exchange that lasted for several days after the terrorist attack of September 11, 2001.
However, there are problems. First, electronic operations are prone to errors that can threaten the existence of brokers who participate in them. Early examples included “fat-finger incidents” that triggered large losses – like selling 610,000 shares of a stock at ¥1 instead of 1 share at ¥610,000. (In December 2005, Mizuho Securities lost the equivalent of $340 million on that one.)
Of course, you can get computers to watch humans and keep them from pushing the wrong buttons. What’s harder is to find a way to watch the computers. With the advent of HFT, we are now in a world where there can be wayward computer algorithms (rule-based programs for automatic order submission). When such programs go awry, it can undermine the safeguards people have put in place: the classic example is that of Knight Trading, which lost an estimated three-fourths of its market value on August 1, 2012. These sorts of problems pose a challenge both for the firm and for the authorities watching over them. Supervisors need to exercise special caution when a systemic intermediary, whose failure can threaten the financial system as a whole, engages in HFT.
A second problem comes from the fact that HFT algorithms can overwhelm the normal supply of liquidity from equity broker-dealers and other investors, leading to market disruptions. That appears to be what happened on May 6, 2010, during the “Flash Crash.” More generally, the prevalence of HFT encourages the usual suppliers of market liquidity to charge a premium to compensate for the risk that a speedy algorithm will pick off their bids and offers before they can be altered in response to market-moving news. If the risk were to become sufficiently large, liquidity could wither. Sudden increases in perceived risk have often caused markets to dry up; with HFT it just happens much faster and with more dramatic consequences.
Third, when market makers worry about risks from HFT, they seek “safer” venues to trade, with the unintended consequence of balkanizing markets. This is a particular problem for the U.S. equity market, where trading has shifted to electronic networks that need not report their off-market bids and offers to public exchanges. Instead of one efficient market where the order book is transparent to all buyers and sellers, the bids and offers on such networks may be opaque to typical investors. As a result, there is no assurance that a bid or offer brought to an exchange results in the best possible price execution.
Fourth, recent research has revealed a level of volatility in high-frequency quotes that markedly exceeds the level of volatility arising from fundamental factors. Such excess volatility results in execution risk and makes quotes less reliable as pricing benchmarks.
Finally, the competition to implement costly HFT technologies can trigger a socially unproductive arms race among traders. It’s hard to see why using billions of dollars of computing and communications equipment to shave a few millionths of a second off the equity market’s response to economic news adds meaningfully to market efficiency or to national welfare. What it can do is create a temptation for front-running that may be difficult to resist. Because of the damage front-running does to investor confidence and the willingness to participate in the market, many people, including Michael Lewis, view this as the worst aspect of HFT.
Fortunately, economists are thinking about remedies for the existing and potential ill effects of HFT. One recent example calls for the use of frequent batch auctions (say, at one-second intervals) to “transform competition on speed into competition on price.” More generally, mechanisms that place a grain of sand in the HFT wheels may be able to preserve the liquidity benefits of improved technology while reducing its costly side effects.