In a recent speech, SEC Chairman Mary Schapiro raised the possibility that new rules will be imposed on "high-frequency traders" in the wake of the "flash crash" of May 6.1 Specifically, Ms. Schapiro noted that high-frequency traders, who possess the most sophisticated tools to access the market and implement their strategies, are not subject to trading obligations (of the sort that governed the previous generation of such market participants, like specialists on the New York Stock Exchange). Beyond the question of whether to impose such obligations, the SEC is said to be considering imposing a minimum time-in-force for limit orders submitted by these traders. The European Union also is reported to be considering restrictions on the activity of high-frequency traders, including establishing a ratio of orders to executed transactions that they would not be allowed to exceed. While the impact of high-frequency traders on the market is debated in the media and in regulatory circles, there is surprisingly very little research on this issue.
High-frequency traders were described by the SEC in a recent concept release as "professional traders acting in a proprietary capacity that engage in strategies that generate a large number of trades on a daily basis."2 They are characterized by the use of high-speed computer algorithms that operate from servers co-located in the same facilities that host the servers of the equity exchanges. In other words, what distinguishes these traders is their need for low latency to implement various strategies that respond to the trading process itself (i.e., to "events" that are generated from within the market).
In a new research paper, Joel Hasbrouck from New York University and I attempt to shed light on the impact these "low-latency traders" have on the market.3 Like biologists looking at cells through a microscope, we examine the "millisecond environment" of NASDAQ stocks in an attempt to assess the influence of these trading algorithms. Our data suggest that the time it takes for fast algorithms to detect, analyze, and respond to a market event (like a change to the best prices in the book) is on the order of 2-3 milliseconds.4 Much activity in today's markets consists of algorithms that either "play" with one another or submit and cancel orders very rapidly in an apparent attempt to trigger an action on the part of other algorithms. These algorithms clearly do not intend to interact with human traders, as the response time of a human trader would likely exceed 200 milliseconds.
Most investors, however, do not engage the market at such speeds, and the SEC has repeatedly emphasized its duty to uphold the interests of long-term investors. A natural question is, therefore, how does the activity of low-latency traders affects the market quality that regular investors experience? To answer this question, we construct a measure of low-latency activity in 10-minute intervals throughout the day by identifying dynamic strategies in NASDAQ order-level data, and study how this low-latency activity affects measures of short-term volatility and liquidity. We use a statistical methodology that goes beyond establishing association between variables to ascertain the impact of low-latency traders on market quality and separate it from the manner in which market quality affects low-latency activity. We find that higher low-latency activity implies lower short-term volatility and better liquidity. Most importantly, these benefits appear both in a period characterized by "normal" market conditions as well as in a period of heightened economic uncertainty in 2008 where our sample stocks went down by 12 percent in one month.5 Our finding that higher low-latency activity is beneficial to market quality during a period of market stress is particularly noteworthy in light of the current regulatory emphasis on the stability of markets.
While these results convey an encouraging picture, it is also important to understand what they do not imply. In particular, it is tempting to look at our results (and those of Brogaard (2010)) and conclude that the ascent of high-frequency trading improved our market environment and hence restricting their activity is undesirable. Alas, such a conclusion does not follow from our research. First, our results do not imply that the current (post-Reg NMS) environment dominated by high-frequency traders is better than the previous market regime where human market makers operated under a set of affirmative and negative obligations (like maintaining continuous presence or stabilizing prices in the face of transient demand and supply shocks). All we can attest to is how the degree of high-frequency activity affects the trading environment in the post-Reg NMS equilibrium. Second, our finding that low-latency activity helps market quality at a time of economic uncertainty and declining prices does not rule out the scenario where a sudden shock causes large fluctuations in prices. This new breed of liquidity providers in the form of high-frequency traders has no obligations to maintain market presence at times of demand and supply shocks, and therefore it is only to be expected that they would curtail their liquidity provision activity or even demand liquidity in a destabilizing manner when such shocks materialize. The SEC/CFTC report on the events surrounding the "flash crash" of May 6 indeed documents such behavior.6
Third, the current environment of equity trading is characterized by higher degree of intermediation than in the past, despite the fact that the market structure is an electronic limit order book where investors could meet each other without the intervention of an intermediary. According to Ms. Schapiro in the aforementioned speech, high-frequency traders participate in the great majority of trades. For comparison, the participation rate of NYSE specialists a decade ago was around 25 percent. Note that the cost of trading for society is minimized when investors meet each other in the market because the price impact paid by one investor is the compensation for liquidity provision earned by another investor. Viewed from this perspective, the real cost of trading (beyond that associated with setting up an exchange) is simply the aggregated profits of intermediaries. Conceptually, if the profits made by high-frequency traders in the current environment are higher than those enjoyed by intermediaries in the pre-Reg NMS regime, then the cost of trading to our society has increased. The Securities Exchange Act of 1934 emphasized giving the opportunity for investors' orders to be executed without the participation of an intermediary. The increase in intermediation suggests that high-frequency traders effectively undercut attempts by regular investors to provide liquidity.
Therefore, current regulatory discussions should consider alongside the new evidence on how high-frequency traders improves market quality also the fragility that seems to have emerged as a result of their activity as well as the overall desirability of increased intermediation.
1. Remarks before the Security Traders Association by SEC Chairman Mary L. Schapiro, September 22, 2010.
2. "Concept Release on Equity Market Structure," Securities and Exchange Commission, 34-61358.
3. Hasbrouck, Joel, and Gideon Saar, 2010, "Low-Latency Trading," Working paper, Cornell University.
4. Since a delay of about 5-10 milliseconds is associated with gathering market data and distributing it via the "tape," co-located algorithms with dedicated data feeds can detect, analyze, and respond to a market event even before information about the market event reaches investors via the tape.
5. Brogaard (2010) analyzes the activity of 26 high-frequency traders and reaches similar conclusions as to the impact of their trading on market quality (see, Brogaard, Jonathan, "High-frequency trading and its impact on market quality," Working paper, Northwestern University).
6. "Findings Regarding the Market Events of May 6, 2010," Report of the Staffs of the CFTC and SEC to the Joint Advisory Committee on Emerging Regulatory Issues, September 30, 2010.