WHAT WE KNOW (AND DON’T KNOW) ABOUT HIGH-FREQUENCY TRADING
by Charles M. Jones
High-frequency trading (HFT) has attracted considerable negative press coverage recently. Is any of it warranted? In a recent paper,1 I review the existing academic research on HFT so that researchers, practitioners, policymakers, and other interested parties can become familiar with the current state of knowledge and some of the outstanding economic issues. Rather than relying on emotional appeals, regulators need to consider the evidence on HFT and automated markets.
HFT firms can trade thousands of times per day for their own account, with typical holding periods measured in seconds or minutes. Many HFT strategies are not new. They are familiar trading strategies updated for an automated environment. For example, many HFTs stand ready to buy or sell like traditional human market-makers, but with lower costs due to automation. As a result, HFT market-makers have mostly replaced the human variety. Other HFT strategies conduct cross-market arbitrage, such as ensuring that prices of the same share trading in both New York and London are the same. This trading strategy can be implemented faster and at lower cost with computers.
Liquidity – the ability to trade a substantial amount at close to current market prices – is an important, desirable feature of financial markets. The key question is whether HFT improves liquidity and reduces transaction costs, and economic theory identifies several ways that HFT could affect liquidity. The main positive is that HFT can intermediate trades at lower cost due to automation. These can be passed on to investors in the form of narrower bid-ask spreads and smaller commissions. The biggest potential negative is that the speed of HFT could disadvantage other market participants. The resulting adverse selection could reduce market quality. There is also the potential for an unproductive arms race among HFT firms racing to be fastest.
Over the past ten years, HFT has increased sharply, and liquidity has improved markedly. But correlation is not necessarily causation. Empirically, the challenge is to measure the incremental effect of HFT on top of other changes in equity markets. The best papers for this purpose identify market structure changes that either facilitate or discourage HFT. There have been several such changes, and the results in these papers are consistent. When a market structure change leads to more HFT, liquidity and overall market quality have improved. It appears that market quality improves because automated market-makers and other liquidity suppliers are better able to adjust their quotes in response to new information.
A remaining concern is that HFT could make markets more fragile, increasing the possibility of extreme market moves and episodes of extreme illiquidity. During the May 6, 2010 Flash Crash, for example, S&P futures fell almost 10% in 15 minutes before rebounding. Some individual stocks moved far more. During the Flash Crash, the CFTC and SEC find that HFT firms initially stabilized prices but were eventually overwhelmed, and in liquidating their positions, HFT exacerbated the downturn. This appears to be a common response by intermediaries, as it also occurred in less automated times during the stock market crash of October 1987 and a similar flash crash in 1962. Thus, there does not seem to be anything unusually destabilizing about HFT, even in extreme market conditions. Short-term individual stock price limits and trading halts have been introduced since. This appears to be a well-crafted regulatory measure that should prevent a recurrence. A trading pause should give market participants a chance to re-evaluate and stabilize prices if the price moves appear unwarranted. More recently, stocks fell by about 1% in less than one minute after false rumors about explosions in Washington circulated on a hacked Associated Press Twitter feed. Trading pauses could also be useful in cases like this.
Regulators in the US and abroad are considering other initiatives related to HFT. Many issues associated with HFT are the same issues that arose in more manual markets. For example, there is concern about the effects of a two-tiered market. Today, the concern is that trading speed sorts market participants into different tiers. In the floor-based era, the concern was access to the trading floor. Many of the abuses in the floor-based era were due to a lack of competition. Now, regulators are appropriately relying on competition to minimize abuses. If there is some sort of market failure, however, then robust competition may not always be the solution, and regulation may be in order. In evaluating any regulatory initiative, it is important to identify the market failure and to ensure the cure isn’t worse than the disease. Proposed regulatory initiatives include:
Consolidated order-level audit trails. Audit trails have always been needed for market surveillance, and robust enforcement is important to ensure investor confidence in markets. With HFT, malfeasance is possible in order submission strategies, and it may be possible to hide by scattering trades across different exchanges, so regulators need ready access to order-level data from each trading venue.
Order cancellation or excess message fees. If bandwidth and data processing requirements are overwhelming some trading venue customers, it may be appropriate for trading venues or regulators to set prices accordingly and charge the participants who are imposing those costs on others. Some markets around the world have imposed these fees. There could be unmeasured benefits, but the early evidence suggests that market quality worsens, as liquidity providers widen their spreads and reduce depths to avoid the fees or recover their costs.
Minimum order exposure times. Under these proposals, submitted orders could not be cancelled for at least some period of time, perhaps 50 milliseconds. This would force large changes in equity markets and could severely discourage liquidity provision. The economic rationale here is particularly suspect, as the overriding goal in market design should be to encourage liquidity provision. But this hasn’t really been tried yet, so there is no empirical evidence one way or the other.
Securities transaction taxes. Based on jurisdictions where transaction taxes have been imposed, removed, or changed, it is clear that these taxes reduce share prices, increase volatility, reduce price efficiency, worsen liquidity, increase trading costs, and cause trading to move offshore.
Restrictions on order types. Exchanges and trading venues have introduced a variety of new order types in the past few years. HFT firms are the main adopters, and there is concern that these could disadvantage other traders in some way. Studying the empirical effects would be worthwhile.
Overall, the vast majority of the empirical work indicates that HFT and automated, competing markets improve market liquidity, reduce trading costs, and make stock prices more efficient. Better liquidity lowers the cost of equity capital for firms, which is an important positive for the real economy. Minor regulatory tweaks may be in order, but those formulating policy should be especially careful not to reverse the liquidity improvements of the last twenty years.
1) Charles M. Jones (2013), “What do we know about high-frequency trading?”, working paper, available at ssrn.com. The paper is based on a number of previous lectures and talks on HFT with the same title; Citadel (an HFT firm) provided financial support to turn the slides into a paper.