Back to Portfolio for the Future™

Will HFTs force a market redesign?

November 12, 2014

alexskopjeBy Ginger Szala

Notwithstanding the evident benefits of HFT in electronic markets, many market participants have argued that some HFT practitioners utilize trading techniques that are detrimental to the well-functioning of financial markets. Some of the trading techniques are spoofing, layering, and quote stuffing.

Chicago Federal Reserve paper: Recommendations for Equitable Allocation of Trades in High Frequency Trading Environments, July 10, 2014

In late August 2014, the CME Group issued a regulatory notice defining and providing examples of “disruptive practices” prohibited on the exchange. These infractions were noted in the Chicago Federal Reserve paper as well as prohibited by the Dodd-Frank Act. However, what’s key about Rule 575, is it doesn’t so much outline new rules as finally give market participants seeking to comply with all market regulations more guidance about what constitutes bad faith trades.

This articulation has been lacking for years, causing headaches, confusion and in some instances, uneven penalties by regulators and exchanges. The fact that the Chicago Fed wrote a paper outlining recommendations before the exchange specified rules illustrates the debate the derivatives industry has grappled with during the growth of high frequency trading. And though the CME Group’s rule may better define infractions, it won’t mitigate them. But solutions exist today to leveling the playing field between HFT and non-HFT traders. These include frequently batched auctions, random-length frequently batched auctions, and minimum time requirements before permitting order cancellation. Implementing any of these changes requires a modest redesign of the current futures market infrastructure. But change is hard, and to use a Chicago adage, is the market ready for reform?

Industry conundrum

Just as high speed trading surged in recent years, so have regulatory penalties for improper trading practices. For example, in 2013, Newedge USA was fined $9.5 million by the Financial Industry Regulatory Authority (FINRA) and several stock exchanges for a number of violations, including allowing wash trading and “spoofing” to take place from early 2008 to late 2011. Spoofing is defined as placing orders designed to trick other firms into buying or selling stocks and marking the close by pushing the stock or commodity price just before the close.

In July 2013, the Commodity Futures Trading Commission (CFTC) and Britain’s Financial Conduct Authority (FCA) brought a successful case using the Dodd-Frank Act’s anti-spoofing rule. Panther Energy Trading and its principal Michael J. Coscia were found to use a “layering” technique in which they used a computer algorithm that basically put in a small order to sell futures immediately followed by several large orders to buy at successively higher prices that according to the regulators, Panther intended to cancel. The trades were made across 18 futures contracts. The CFTC set a $1.4 million civil monetary penalty and forced Panther to disgorge $1.4 million in trading profits for the strategy. The FCA fined Panther just over $1 million for deliberately manipulating the markets while ICE Futures Europe penalized the firm $900,000 and the CME Group hit them with an $800,000 fine. In October this year, the U.S. Attorney in Chicago filed suit against Panther and Coscia — further upping the ante. Regulators say this is the first time federal charges have been filed against abusive trading.

The penalty list goes on, many of them levied before Michael Lewis’ book, Flash Boys: A Wall Street Revolt, was released. The book increased public pressure on regulators by shining a bright light on high frequency trading. But the book did more as it went into the granular level of trading, and how either a millisecond or three or a fraction of a point or two can mean millions of dollars for and against firms and traders. What Flash Boys didn’t do was describe how brokerage firms execute the business and how exchanges and regulators have increasingly penalized brokers for not knowing the “intention” of their customer’s order (a key term used in Rule 575). In addition, exchanges will question a trade strategy, and take months to get back to the trader or broker informing them whether it’s permissible or not. In the meantime, if the broker allows the client to continue trading the strategy, and the exchange eventually finds it to be unacceptable activity, not only can the client get sanctioned, the clearing firm can also be penalized for allowing the client to continue to trade a strategy that was determined to be improper after months of study.

If this seems like Dr. Seuss nonsense, it does as well to the traders and brokers who grow increasingly frustrated with arbitrary rules coupled with a new heavy-handed approach by regulators who, they claim, aren’t familiar with how the markets or brokerage industry work. The CME’s Rule 575 may be a partial step to define bad behavior, but as one compliance officer notes, “To prove ‘intent’ still seems a bit difficult. And nowhere in the rule (or Q&A) does it mention the responsibility of the clearing firm to police this type of activity and have the ability to identify it.”

Exchange defense

The CFTC’s Technology Advisory Committee (TAC), formed in 1999, meets regularly to advise the regulator about technology and related issues in the industry. The committee met in early June to respond to perceptions of unfair HFT advantages in the futures industry and share ideas on how to develop a surveillance program for the 21st century. The group heard testimony from the CME Group, Intercontinental Exchange (ICE), data wonks, regulators and traders, from both the futures and securities markets.

In testimony, exchange representatives described their market structures and methods for handling order traffic and trade matching while also outlining key surveillance software. The CME Group was particularly detailed about the process as they have been criticized by the non-HFT sector as beholden to HFT firms that make up so much of CME Group liquidity. In addition, the CFTC came down on the CME Group last year for not being strict enough in policing its electronic trading. Since that report it appears to traders as if the CME has been more aggressive in questioning trades and opening investigations.

Further, CME Group was sued earlier this year by a group of traders stating, among other things, that the CME allowed certain HFT firms to see orders before they hit the central order book. The CME responded that the suit is without merit, and is “devoid of any facts supporting the allegations and, even worse, demonstrates a fundamental misunderstanding of how our markets operate.”

The exchange, sensitive to the swirl of regulators focusing on it, would not comment on compliance issues except to point to testimony or presentations already in the public domain.

During the June TAC meeting, Bryan Durkin, chief commercial officer of the CME Group, reported in depth the steps the exchange takes to process orders, inform traders and keep track of bad behavior. “We use a central limit order book. It’s a single integrated marketplace, allowing for concentrated liquidity in one transparent location,” he said. “The identity of traders and firms is protected from disclosure on all bids, offers and execution reports. Bids and offers are available to all market participants and are matched according to transparent exchange matching algorithms. Our market data is sent to everyone at once. There are no preferential data feeds that are provided exclusively to a particular segment of user base…No one can see orders prior to them hitting our match engine and being made available to the order book. We maintain a complete and a comprehensive audit trail of every single message, order and trade down to the message ID of every participant in our markets. This information is fed into a sophisticated surveillance system that allows us to identify and to prosecute prohibited trading practices that violate our rules.”

Durkin went on to address the criticism of co-location facilities, oft mentioned in the lawsuit against the exchange. “Although co-location requires an investment, it is open to all market participants and serves to level the playing field unlike several years ago when a firm could gain an advantage by buying real estate near an exchange or where a server was thought to be in a data center. With the development of our co-location facilities, no one trader may gain an advantage over another due to proximity to that match engine. All customers in our co-location facility are treated equally.”

So if the exchange’s facilities seem to be working fairly, at least according to the exchange, then what is causing some of the anger against HFTs in the industry, especially as they provide so much market liquidity?

Industry complaints

Some complaints focus on the telecom arms race. Although customers may not be treated differently inside the data centers, much of the technology money being spent relates to data movement between data centers. For example, consider data transmission between CME’s DC3 and the ICE matching engine on Cermak Road in Chicago. While elite trading firms employ microwave towers blasting beams of data to race each other to the finish line, the less well-heeled utilize land lines of varying bandwidth and speed.

And though many market participants believe the complaints of spoofing, etc., aren’t anything different than what has been happening since the inception of markets, others, traders, fund managers and brokers, say the bad faith trades have hurt them. In addition, associated compliance issues post-Dodd-Frank have cost them dearly, and now to be under a “gotcha” watchdog is ruining the business. Obviously HFT liquidity is critical, but what is its cost?

Both the equities and derivatives markets have been dealing with this topic. The equities market is perhaps under greater pressure due to Lewis’ book. But the futures markets, which are much different than the equities’ fragmented mess, still have issues. For example, the exchanges all have different rules, and even within the CME Group, rules vary between the CME, CBOT and NYMEX markets. So what may be permitted on one exchange may not be on another. Greater clarity and harmonization of the rules certainly seems logically in order.

Compliance experts agree the exchanges have been more aggressive, especially with electronic trading issues. Many of these were due to malfunctioning software that affected market pricing. For example, in 2011, the CME fined Infinium, a former HFT shop, $850,000 for three separate computer mishaps that affected the markets in 2009 and 2010. More recently, Credit Suisse was fined $25,000 by ICE when the firm’s Russell 2000 index futures automated trading system broke down in 2012 resulting in the entry of “thousands of orders for multiple Credit Suisse proprietary accounts.”

Technology glitches happen, but in today’s high-speed world, a couple seconds can cost more in the market place than any disciplinary fine. Still, the exchanges, despite beefing up their compliance and surveillance teams, are severely backlogged, hence taking up to a year or even more to complete investigations. Further, many of the exchanges’ compliance staffs are not familiar with the nuances of trading, which cause problems for brokers and their clients. One trading firm was tagged with an infraction that didn’t make sense. When the firm appealed to the exchange’s upper echelon to complain, the investigation was pulled, largely because the investigating staffer was wrong. Brokers say the same about the CFTC compliance staff. One FCM compliance officer noted “Some of the things they come up with or question just shows an absolute ignorance of how the markets work.”

Potential solutions

It’s no secret that spoofing and similar behavior has been under study by many groups. Two have studied trading behavior closely and made recommendations to alter market design.

In July 2014, the Chicago Federal Reserve Bank released an updated working paper, “Recommendations for Equitable Allocation of Trades in High Frequency Trading Environments,” written by John McPartland, Senior Policy Advisor in the Financial Markets group of the Chicago Fed. The paper acknowledges the benefit of HFTs in providing liquidity to the marketplace (in this case, both equities and derivatives), but also states that institutional investors have argued that “HFT places them at a competitive disadvantage.” The paper reviews findings of multiple studies, which found HFTs increased liquidity, narrowed bid/ask spreads and reduced trading costs. However, two studies found something else. Elaine Wah and Michael P. Wellman’s “Latency Arbitrage, Market Fragmentation, and Efficiency: A Two-Market Model” stated that market fragmentation and the presence of a latency arbitrageur reduced the “total surplus and impact[ed] liquidity negatively,” and went on to recommend periodic batch auctions or call markets.

Similarly, University of Chicago Associate Professor Eric Budish et al’s “The High-Frequency Arms Race: Frequent Batch Auctions as a Market Design Response,” concluded that frequent batch auctions can lead to “narrower bid/ask spreads, deeper markets, and greater social welfare.”

In the Fed paper, McPartland outlines the problems caused by some HFTs, such as spoofing, layering and quote stuffing, writing that “rather than propose solutions that might preclude specific HFT strategies, we propose to simply change the economics of the trading environment by modifying the criteria of order allocation priority and by discouraging certain questionable industry practices to strike a more equitable balance between the high frequency trading community and the investment management community.” To that end, the paper lays out nine recommendations. One of the recommendations is changing markets from continuous trade matching to dividing trading sessions into discrete periods of one half second.

Budish notes in his paper that the HFT chase for speed has become an arms race that is hurting the market overall and basically is a symptom of “a simple flaw in poor market design.” He recommends that time should be discrete, like tick sizes, and should be at frequent, distinct intervals, such as 100 milliseconds. In a video presentation, he notes that stale quote sniping takes place regularly in a continuous market with today’s high speeds. This cost of being picked off is passed on to the investors.

Frequent batch auctions eliminate this, he says, adding that it reduces the value of a tiny speed advantage so firms have plenty of time to react (thus cancelling stale quotes) and it transforms competition on speed to competition on price. He adds this really isn’t anything new, that Milton Friedman originally proposed uniform-price auctions in the 1960s for Treasury auctions.

Budish also presented his paper at the 2013 CME Group-MSRI Prize in Innovation Quantitative Applications forum, which had a panel made up of academics including Myron Scholes, of the Black/Scholes formula, as well as the CME’s Durkin. When asked if the exchanges would be the ones to implement such a market design change, Durkin stated: “One of the things I’ve asked Eric to consider in future analysis is the practicalities of integrating this concept with the fact that our markets globally are so interconnected.”

A half-second — 500 milliseconds or 500,000 microseconds — is a lifetime to a high frequency trader, and to many non-HFTs. The reaction to this idea from the industry ranges from “that’s goofy” to the reasoned recommendation that if the exchanges implement frequent batch auctions, it should be in time intervals of 100 milliseconds or less. One HFT, who agrees too much money is being spent on “arms,” recommends 20 microsecond intervals, otherwise, firms would be exposed to too much risk, and eventually, they would move out of the market and take their liquidity with them.

Budish is not convinced of the risk management danger if the intervals were small enough. “I agree on the sign of this effect, but am skeptical about the magnitude,” he says. “Remember that we are talking about batch auctions conducted extremely frequently throughout the day, such as every tenth of a second, and that risk scales with time. So a very small amount of time translates to a very small amount of risk. On the other hand, the liquidity benefit from frequent batch auctions—eliminating sniping and its harm to liquidity—is realized even for extremely short batch intervals, and our empirical results suggest that the magnitude of this effect is large,” he says. Budish willingly admits that he does not know the precise interval that will formulate the best batch. He allows that it might be a smaller interval than 100 milliseconds.

Some industry participants contend that one problem with simply adjusting the markets to a batch processing matching engine as described above is it simply alters the “arms race.” If market participants know that precisely every 100 milliseconds or any other exact and pre-specified period of time will result in a batch process, the fastest trader with the most expensive infrastructure will be able to be the last one to send in a trade before the batch closes and so they get to keep their trade edge. Some suggest this edge is best negated by having a batch processing methodology with randomness as to the exact length of the batch. One batch might be after 20 microseconds and the following batch might be after 35 microseconds. It would likely make sense to have some structure to the random batching. For example, they may be no shorter than 20 microseconds and no longer than 40 microseconds. Adding the combination of batching and randomness to the matching methodology likely produces a marketplace fairer to more participants with less economic waste on the technological arms race. The very small batch period formulated in microseconds still offers reward to participants heavily invested in technology. It also rewards responding rapidly to changing information in the marketplace. At the same time, it negates some of the advantage of being the absolute fastest.

Another adjustment to electronic trading suggested by regulators could require orders to rest for some period of time. Both HFTs and non-HFTs contend this would disrupt their strategies as traders have multiple product positions on at once, and having to hold trades for even a few milliseconds would increase their risk. However, others contend that unless those trades are put in bad faith, a resting period shouldn’t be an issue.

One market maker and former floor broker says for energy traders and their clients who use the trade-at-settlement (TAS) tool, HFTs have become disruptive because they are pushing players, especially commercials, to take the settlement plus 1 or even higher because HFTs jam the market mechanism with bogus orders making it impossible for real participants to be involved. The client, seeing the flat price overly bid and not wanting to take a chance on a worse price, takes the TAS+1, and meantime the disrupters cancel large numbers of trades just before the settlement. “The exchange says these guys provide liquidity, but if they always get flat, what liquidity do they provide?” he asks. He warns that the HFTs are chasing out true market makers and liquidity providers and commercials are passing on their extra tick(s) expense to the consumer. Rule 575 does apply to pre-opening and closing periods as well as all trading sessions.

What is the next step?

Scholes was not convinced on the discrete time/batched order idea, noting during the panel that changing the market set up to accommodate “stale quotes” wasn’t worth it. With due respect to Scholes, there’s more to it. Millions of dollars are being spent to keep up with the speed race — by firms, exchanges and regulators. Durkin noted the CME spends $40 million a year on monitoring and surveillance, not to mention updating trade mechanisms. The CFTC budgeted more than $50 million in 2013 to operate and maintain IT systems and infrastructure for surveillance and enforcement, and spent another $7 million in developing, modernizing, and enhancing surveillance and enforcement systems and IT infrastructure. And in their quest to achieve even faster speeds, HFTs are spending millions of dollars. Spread Networks famously built the fiber optic cable from Chicago to New York for $300 million to shave three milliseconds off the time to transmit data between the cities, in which firms paid $300,000 a month to “rent” bandwidth. Jump Trading, a Chicago-based HFT, and others are spending millions to buy microwave towers around the world. By slowing down the market, Budish hopes to achieve better market pricing as well as have less money spent on “arms.”

There also is the cost to brokers, who not only have to keep up with new Dodd-Frank rules, but also spend to improve software systems to keep in line with compliance requests as well as service clients.

Some of this spending could be reduced or eliminated with batched order auctions. When asked how hard it would be to implement the idea, Budish says, “Broadly, there are three obstacles to practical implementation. First, there are regulatory ambiguities, as SEC Chair White mentioned in her June speech. Second, there are vested interests in the status quo. Third, there are the coordination problems associated with getting a new marketplace off the ground – ensuring that trading firms and investors show up. These three obstacles apply equally well to both stock and futures markets. My sense is that the programming and technology costs associated with launching a new market are tiny relative to these three bigger issues.”

Change always presents challenges, but perhaps implementing these changes would serve the overall market. Would the CME lose some of its volume by implementing batched order auctions? (An idea under “intense discussion” now in Europe.) Could it be applied to a single (less busy) contract at one of the exchanges to see if the market approves? (ICE has hinted it might test some of its reduced order types and maker-taker changes on the old Amex or another small market.) Would it be wise to start an exchange with that model to see if it’s accepted? (Such as what IEX has done with its 350 microsecond speed bump.)

It seems two things can happen: the speed race, as some believe, will burn out on its own. There’s only so fast traders can go without bumping into the laws of physics. Or, changes will be made to the market structure because traders and brokers grow weary of getting spoofed, sniped, mugged or fined by high speed traders and heavy-handed regulators. One audience participant pointed out at the CME/MSRI forum that back in the floor days if someone was doing something wrong, there were hundreds to witness it. Today, although many players see what’s happening on the screen, it’s only the exchanges that can truly monitor the bad behavior. This makes participants restless, because they believe the exchanges may give too much latitude to their largest customers: HFTs.

As one fund manager who has been around the industry for decades, notes, “What was going on in the pits pales in comparison to today. No scalper or market maker ever pulled the kind of money out of the pit that [HFTs] are pulling out of the electronic market. You can draw your own conclusion.”

Surely futures market players don’t begrudge traders making money, after all, that is the name of the game. But there has been a shift in what many view as fair play. With overseas regulators already implementing new rules, such as in Germany with the HFT Act, U.S. industry participants and exchanges might be wise to build a consensus and determine a coordinated action before U.S. regulators step in and force their hands.

(Note: The original version of this piece appeared in Inside Advantage, Fall 2014 issue.)

__________________________________

 

Ginger Szala is the former editor-in-chief and publisher of Futures Magazine Group. She has reported on and written about the global derivatives and managed funds business for the past 30 years. Today she is a freelance journalist, business writer and media consultant. You can follow her on Twitter @gingerszalaink or contact her at: gszala@gingerszalaink.com