Executive Vice President, Market Regulation
New York, NY
Good afternoon. Thanks John [D’Antona], for that introduction, and thanks to MarketsMedia for the invitation to talk with you today.
It’s always a pleasure to be at a forum like this that’s focused on pressing equity market structure issues. And it’s particularly rewarding to be here for this unique conference within a conference format that brings together senior traders, technology experts, regulators and business executives to talk in small groups about issues like best execution, trading technology, the future of exchange-traded funds and the Consolidated Audit Trail. It’s also fitting that we are meeting today at the very moment the SEC is gathered to consider whether to approve the exchanges’ and FINRA’s CAT NMS Plan—which will fundamentally change how regulators surveil the markets.
As the Flash Crash and the Flash Boys book are now further behind us in the rear-view mirror, I would like to posit that, while the markets are still not perfect, they are much more resilient to withstand the velocity that can occur in our modern market environment where microseconds and nanoseconds really matter. This is largely due to a series of initiatives that the SEC, the exchanges and FINRA have put in place over the last few years.
These initiatives created a multi-faceted safety net for the markets and are designed to promote investor confidence. Among the changes, regulators adjusted the marketwide circuit breakers, which give market participants an opportunity to assess their positions, valuation models and operational capabilities when extreme periods of volatility occur. On top of that, we implemented a limit up/limit down regime, which addresses the type of sudden individual stock-price movements that the market experienced during the May 2010 flash crash. Under the plan, a limit up and limit down mechanism prevents trades in national market system stocks from occurring at prices outside of certain ranges. And if the changes in price are more significant and prolonged, the limit up/limit down rules would trigger a trading pause in that security.
The SEC also passed the Market Access Rule, which requires firms entering orders into the market, or allowing their customers to enter orders into the market, to have pre-trade controls to avoid erroneous and duplicative orders and to establish pre-trade capital and credit controls on orders entered into the market, among other things. More recently, the SEC implemented Regulation SCI to strengthen the technology infrastructure of the U.S. securities markets. Specifically, the Regulation SCI rules are designed to reduce the occurrence of systems issues, improve resiliency when systems problems occur, and enhance SEC oversight of the technology infrastructure of markets.
With this stronger foundation of resiliency, my sense is that the market structure debate has shifted from questions such as “Should there be HFT?”, “Is HFT bad for the markets?”, and “Are the markets rigged?”, to issues such as “How can technology improve access to liquidity?”, “Have regulators come around to accept HFT?”, “How do you define best execution in an increasingly complex electronic marketplace?”, and “Have top buy-side shops surpassed the sell-side in savvy and sophistication?”. In full disclosure, the questions in the last set are just some of the many provocative questions that were prepared by the planners of this conference that you will be discussing during your roundtable discussions.
In sum, I think we are rightly focusing on the evolution of the market more than whether there is something seriously wrong with the market. So in that vein, I would like to focus on how FINRA is working to stay ahead of issues through our focus on transparency and by making use of innovative technology in our surveillance programs.
Transparency Initiatives
Market transparency is a central part of any market regulator’s job and it is a significant portion of what we do at FINRA every day, in both the stock and bond markets.
FINRA has a long history of bringing transparency to the equity and bond markets. When we launched the Trade Reporting and Compliance Engine, or TRACE, in 2002, information about most trading of corporate bonds was not immediately available to brokers or investors. To find out where a given bond was trading, a broker would place a series of telephone calls to dealers and hope he or she was getting valid and recent information as to what price a bond had recently traded. Today, TRACE provides wide access to trade data for corporate bond transactions, including the price and size. Since we introduced TRACE, we have steadily expanded it to make trading in bonds less opaque. Investors now have access to information about transactions not just in corporate bonds but also in asset-backed securities, mortgage-backed securities and Small Business Administration-backed securities.
Consolidated information on equities and options trading has been crucial for investors and regulators, and has made it possible for regulators to readily access transaction and quote data to pinpoint the cause of a market event or better detect potential rule violations. That data is a necessary step in promoting the integrity of our markets.
While regulators of equities, options, and futures markets can readily access transaction and quote data, it’s not the same for the $13 trillion Treasury market. In October 2014, when yields during a single trading session in the 10-year note plummeted, then rebounded without a clear reason for the highly unusual volatility, regulators didn’t have real-time access to trading data to determine the cause. There is currently no centralized trade reporting system for Treasuries. Regulators, including FINRA, the SEC, the Treasury Department and the Federal Reserve Board, have taken steps to implement a transaction-reporting regime for Treasuries. Starting next July, firms will have to report certain transactions in Treasury securities to TRACE. At this time, FINRA will not disseminate information on transactions in Treasuries. This new requirement will significantly enhance the ability of FINRA and other regulators to understand trading activity in Treasury securities.
On the equity side, we have also taken steps to increase market transparency of alternative trading systems, including dark pools. In June 2014, we began publishing on our website volume and trade-count information for equity securities executed on an ATS. And in April 2016, we began posting the remaining, non-ATS OTC equity volume by member firm and security. And just last month, FINRA began publishing monthly ATS block-size trading statistics in all National Market System stocks.
Surveillance
These initiatives to improve transparency, along with the safety nets to manage periods of extreme volatility, have shored up our equity market and helped rebuild investor confidence. But the industry has to continue to do more to shift from a reactive approach to a preventive stance.
At FINRA, we’re taking steps in our market surveillance program to prevent problems before they occur. For example, we have enhanced our ability to pull together data across exchanges and alternative trading systems to see one virtual market instead of a disjointed patchwork of individual markets.
In today’s increasingly fragmented and interconnected market, bad actors can consciously disperse their trading activity across markets, asset classes and broker-dealers in an attempt to hide their fingerprints to avoid detection. So having the ability to review trading across markets, rather than at a single market at a time, is essential to any market surveillance program.
Through our agreements with exchange clients, FINRA monitors 99 percent of the listed equity market and 65 percent of the listed options market. And our cross-market surveillance program allows us to run dozens of surveillance patterns and threat scenarios across the data we gather to look for manipulative activities, such as layering, spoofing, algorithmic gaming and other abusive conduct. This sophisticated surveillance allows us to detect activities that were previously more difficult to see. For example, 47 percent of our cross-market equity alerts identify potential manipulative activity by two or more market participants acting in concert, and 65 percent of our cross-market equity alerts identify potential manipulation by a market participant across multiple markets.
Last year, FINRA’s Market Regulation department made 142 referrals to the SEC related to our cross-market surveillance program, and we have made 171 referrals so far in 2016.
We continue to refine existing patterns and develop new surveillance patterns to address new threat scenarios. For example, we have recently launched a new cross-market surveillance pattern to detect ramping at the open and close; we have launched a new surveillance pattern that looks for layering in the equity market to create favorable options prices; and we are developing additional cross-market surveillance patterns to more closely monitor trading in ETPs.
Currently, FINRA’s cross-market surveillance patterns include ETPs, and our cross-market surveillance program has several threat scenarios directly intended to detect potential problematic activity related to the unique attributes of ETPs. In addition, FINRA performs a number of other regular reviews focused on specific aspects of ETP trading.
However, given the prominence of ETP trading, FINRA believes that additional cross-market surveillance patterns and threat scenarios are necessary to provide both broader surveillance coverage of all ETPs and to detect potential problematic behaviors unique to specific subcategories of ETPs.
So we plan to introduce prototypes of a number of these surveillance patterns in the coming months. Among other things, the patterns will include scenarios designed to detect potential gaming activity among ETPs, their components, related ETPs and options on ETPs; potential manipulation of the ETP creation and redemption process; and a scenario specific to actively-managed ETPs.
In addition to FINRA’s groundbreaking work with cross-market and cross-product surveillance, the SEC’s proposed Consolidated Audit Trail will also help regulators spot problems earlier. Today, regulators don’t have a single database with comprehensive and readily accessible order and execution data across equities and options. FINRA and every other regulator use their own separate audit trail systems to track information relating to orders in their respective markets. While FINRA is grateful to receive audit trail data from our client exchanges that enable us to view 99 percent of the listed equity market and 65 percent of the listed option market, this data set is not as complete or as granular as the data set that will be available through the CAT.
The CAT will significantly enhance regulators’ ability to monitor and analyze trading activity across markets and securities. Among other things, the CAT will require unique identifiers for accounts, provide for better order audit trail information for options, require more detailed information about each trade, and provide for better linkages between related equities and options trades. Together, these pieces of data will allow regulators’ surveillance systems to more effectively zero in on problematic trading, with the expected result being surveillance reviews and investigations that will better serve investors.
Regardless of which entity is ultimately chosen to operate the CAT—and FINRA is one of the finalists—once it is in place, it will greatly enhance FINRA’s ability to look at data across exchanges and alternative trading systems. When combined with FINRA’s unique equity and options cross-market surveillance program, FINRA will be even better positioned to detect problematic trading across markets and products.
At the industry level, FINRA is using the information we gather from our cross-market and cross-product surveillance programs to give firms information that can help them prevent and stop layering activity. In April, we began offering monthly cross-market equities supervision report cards to firms.
FINRA sends the report cards to firms when we identify potential spoofing or layering by the firm or entities to which the firm is providing market access. The reports provide a summary of the identified market activity, detailed information about the exceptions, and trends in such trading over the preceding six months. The report cards do not reflect conclusions that violations have occurred; rather they indicate potential problems that the firm needs to review.
While most firms attempt to surveil and review for manipulation, it is difficult for an individual firm to detect that type of behavior across markets and to determine what portion of the activity is conducted through that firm.
So FINRA is leveraging our cross-market data and using sophisticated automated surveillance technology to provide firms with knowledge that potentially toxic order flow is emanating from or passing through their firm. Doing so helps firms add that data to their own surveillance and supervisory processes, and take appropriate action to address the activity even before FINRA can complete a formal investigation.
Since we began issuing the report cards, we’ve seen a 75 percent decline in layering activity.
FINRA will continue to investigate suspected manipulation and, where appropriate, take enforcement action or refer the activity to the Securities and Exchange Commission if the market participants are outside of FINRA’s jurisdiction. But FINRA also believes it is crucial that we empower firms with data generated by our sophisticated surveillance systems so that they can take proactive steps to mitigate, if not eradicate, bad behavior.
We’re also tackling layering and spoofing activity by broker-dealers or their customers with new disruptive trading rules.
FINRA’s Board recently approved rule proposals to expressly identify layering and spoofing as disruptive trading activity and to establish an expedited process for issuing cease-and-desist orders to prevent firms from engaging in the activity or providing access to a customer that engages in the activity. The proposed rule changes are similar to the rules the Bats Exchange, Inc. and Nasdaq recently adopted, but builds on FINRA’s existing process for temporary cease-and-desist orders.
When we believe disruptive and manipulative quoting and trading activity is occurring, we open an investigation to determine whether to bring a formal disciplinary action against those responsible for the activity.
While we aim to act as quickly as possible, the time from alert generation and review to a final resolution of the matter can often take up to several years. However, there are certain clear instances of disruptive and manipulative behavior, or cases where the potential harm to investors is so large, that we believe we should have the authority to initiate an expedited proceeding to stop the behavior from continuing.
Once the rules are approved, FINRA will have the authority to initiate an expedited proceeding to stop the behavior from continuing in those limited instances in which a firm is engaging in or facilitating a frequent pattern or practice of layering or spoofing activity, the firm has received sufficient notice with an opportunity to respond, but such activity has not ceased.
Technology
In addition to the disruptive trading rule and our cross-market and cross-product surveillance tools, FINRA continues to look for new ways to find and stop manipulative behavior—and uncover manipulative schemes that we don’t yet know about.
Specifically, given that FINRA processes, on average, 50 billion events a day, we’re exploring how we can apply data science tools like machine learning in our surveillance development and ongoing parameter adjustment processes.
Today, our surveillance programs operate by selecting from billions of market events only those events that are relevant to a specific situation. The programs then classify the behavior of market participants, and flag potential rule violations or patterns of behavior. Building a surveillance program requires a significant amount of interaction between the surveillance analysts and data scientists to identify the right conduct and to set the appropriate parameters for the surveillance program going forward.
Machine learning can help us accelerate the process of surveillance development by having the surveillance analysts identify specific instances of behavior of interest and allow algorithms to “train” themselves to identify additional instances in new data sets. This self-learning process would improve the efficiency of the process to determine appropriate parameters for the surveillance program.
The technology would also allow us to take the next step and enable the machine-learning algorithms to make modifications to the parameters as market behavior changes. Machine learning also could provide us with tools to detect manipulative activity that we have not seen before.
Machine learning is not wholly new to us. We used machine learning to develop the initial parameters for a number of our surveillance programs and are looking to engage in a pilot in the coming year to run a “self-adjusting” machine-learning version of an existing pattern to assess the benefits of such an approach to our current automated surveillance program.
Conclusion
Tools like machine learning have the potential to give regulators an even better view of market activity, and possibility root out some manipulative schemes that we’re not aware of. It is imperative that regulators and the industry continue to work together to strengthen surveillance programs and to stay ahead of problems that affect the quality of our markets.
Thanks for listening. I’d be happy to take any questions you have.
