As we look back on 2013, it has certainly been a year of increasing scrutiny and criticism of capital markets trading participants. One particular area of focus is so-called "High Frequency Trading" (HFT), the practice of automated trading algorithms rapidly taking market positions. HFT is seen by many as a major cause of market crashes and volatility. Most notably, many believe the automated withdrawal of liquidity by HFT algorithms accelerated the 2010 flash crash, where the Dow-Jones Industrial Average dropped nearly 1000 points and then recovered most of the losses in only 30 minutes.
As Reuters recently reported, in Europe there is a lot of pressure from the European parliament to put measures in place to curb HFT. A proposed package of measures, including common tick sizes, synchronized exchange clocks to more easily spot abuse and more rigorous algorithm testing, are planned as part of the second Markets in Financial Instruments Directive (MiFID II). Crucially omitted was a much-debated point about minimum holding time for instruments, proposed at 500 milliseconds. Many high frequency algorithms might hold positions for a shorter time! However, this compromise might get the legislation through more easily. In addition, MiFID II will also limit the amount of liquidity traded on dark pools -- off-market venues, where transactions are not transparent.
Earlier this year, Edward J Markey, Member of Congress in the U.S. declared that High Frequency Trading (HFT) "represents a clear and present danger to stability and safety of [US] capital markets and that it should be curtailed immediately."
Congressman Markey backs his assertion with evidence drawn from market events, such as the Flash Crash and Knight Capital, as well as academic studies on the impact of HFT on markets and therefore main street. To summarize his points:
1. The speed of HFT disadvantages other styles of trading
2. HFT exacerbates volatility in the market
3. 'Other' investors are scared away
4. Volume traded by HFT is therefore a dangerously high proportion of overall traded volume
Not everyone shares Congressman Markey's view. I was involved with a UK government-sponsored expert group called Foresight, which released its findings on "The Future of Computer Trading in Financial Markets" in November 2012. And Foresight's findings were quite different! Foresight does not feel that HFT increases market volatility and raised some practical approaches to making markets safer for everyone, including circuit breakers and a consolidated tape. Foresight favored working with HFT to address perceived dangers is the committee's preferred approach over banning HFT outright.
So where does the truth lie? Is HFT bad? Before we can answer that, we need to define what exactly is meant by HFT. A huge proportion of the market is certainly electronic trading and gone are the days of open outcry trading. However, much of that is more traditional large orders, for example on behalf of a pension fund, which may use algorithms to break them down into more manageable chunks and sequence these chunks in the market. Then there may be other totally automated algorithms trading for a proprietary trading firms that might hold their positions for hours or days. The Chairman of the CFTC technology advisory committee (on which I serve) Commissioner Scott O'Malia has been trying to get a good definition of HFT. He's now got one, which is essentially "fully automated super speedy, short holding time".
Within the ranks of HFT there are almost certainly bad apples. Several incidents of alleged market manipulation have led to fines, for example, the practice of quote stuffing in which a smoke screen of orders is used to mislead other market participants. Another big concern on market stability is trading algorithms going out of control -- as we have seen many times, including at Knight Capital. This suggests a market not under our control. We must have better monitoring and control systems to prevent this from happening - or one can become sympathetic with the "speed limits" that U.S. and EU legislators might propose
However, there are many benefits to automated trading and arguably HFT is normal trading but quicker. If any market participant, high frequency or otherwise, attempts to move the price of an instrument (a.k.a. price ramping) by dominating the market (a.k.a. abusive squeeze), it is market manipulation. If said participant takes the position with the sole intention of immediately reversing it (flipping) for profit, or indeed to engineer volatility, it is market manipulation. These are well understood dangers of any style of trading and apply equally to HFT. To suggest banning HFT is to not understand that it's the behavior that is wrong not the speed at which is takes place. The question is -- do we have the ability to police trading at that speed?
What is needed is not a ban on High Frequency Trading but a mandate for "High Frequency Trading Surveillance," preferably in as near real-time a possible so humans can respond to manipulative algos in timeframes less than minutes or hours or days or never. I've been involved now for several years applying the same technology used by high frequency traders to the job of monitoring trading activity to try to spot market manipulation, market abuse, rogue traders, flash crashes and algorithms gone wild. In high frequency trading, algorithms are looking for the correlations and patterns that indicate they should place orders in the market. An example could be a break in the statistical correlation relationship between the 10-year U.S. Treasury note and the 10-year U.S. Treasury future, which could be an opportunity to buy one and sell the other. Techniques, such as Complex Event Processing (CEP) and other in-memory technologies that involve continuous analytics of real-time data and intelligent action, have been used to good effect here. Now, with the need to monitor the behavior of traders and algorithms, the same techniques can be used, for example, to spot that a trading algorithm is operating outside of its normal parameters, placing too many large trades too frequently, and should be shut down with an automated kill switch. This would deal with the Knight Capital rogue algorithm. Also a regulator should be able to detect a pattern of quote stuffing or other abuse on one of more trading venues by continuously analyzing the data, just as trading algorithms do.
In summary, let us apply the learning of advanced trading techniques to trading surveillance, monitor for any clear and present danger and stop it before it gets out of control. Not all traders should be penalized because of a few rogues. Today's CEP and allied technologies enable us to identify many impending problems and take preventative action to reduce the risk and reputational damage. A blanket ban or severe restriction of HFT through legislation is an easy way out, but one that doesn't necessarily stop the rogues.