Last week, the Commodity Futures Trading Commission proposed a comprehensive set of new rules that, if adopted, potentially would impose many new obligations on certain CFTC registrants that use algorithmic trading systems to trade futures, options or swaps on designated contract markets (but not on swap execution facilities). Potentially impacted registrants are future commission merchants, floor brokers, swap dealers, major swap participants, commodity pool operators, commodity trading advisors, introducing brokers and certain persons proposed to be registered as floor traders for the first time because of their algorithmic trading activities.
FCMs who are clearing members of DCMs and carry accounts for covered algorithmic traders; DCMs (but not SEFs); and the National Futures Association would also be impacted by the CFTC’s proposed new requirements.
The CFTC’s proposed new rules would also mandate the registration with the CFTC as a floor trader any person who uses an algorithmic trading system that electronically and directly routes orders to a DCM other than first through a clearing-member FCM, unless such person is otherwise registered with the CFTC in another capacity. The CFTC estimates there will be a maximum of 100 potential new floor trader registrants under this proposed new requirement.
Finally, all current categories of CFTC registrants that engage in algorithmic trading and newly required to be registered floor traders would be required to maintain copies of all source code used in a production environment, including all changes, in accordance with general CFTC record-keeping requirements (e.g., retain for five years), and, upon request, make available such source code for inspection by CFTC and US Department of Justice staff without subpoena or other process of law.
Proposed Regulation AT was included in a 521-page document that provided insight into the CFTC’s thinking in proposing the new rules, as well as a cost and benefit analysis.
Previously, in 2013, the CFTC published a concept release regarding algorithmic trading to which many industry and public representatives commented. (Click here for background in the article, “CFTC Issues Concept Release on Automated Trading; Asks If More Regulation and Higher Fines Are Necessary” in the September 9, 2013 edition of Between Bridges.)
The CFTC will accept comments to its proposed rules—known as Regulation AT—for 90 days after its publication in the Federal Register. To assist its analysis, the Commission requested comment on 164 specific issues addressed by the proposed rules.
Under proposed Regulation AT, algorithmic trading is not limited to the black box derivation and electronic placement of orders on DCMs. Covered activity also may include solely automated order placement. There is no minimum number of transactions that constitutes algorithmic trading.
Algorithmic trading is defined broadly to include any trading of any future, option or swap subject to DCM (but not a SEF) rules where an order, modification, or order cancellation is electronically submitted and one or more computer algorithms or systems:
- decides whether to initiate, modify or cancel the order, or
- otherwise makes “determinations” with respect to the order including but not limited to:
- the product to be traded;
- the DCM where the order will be placed;
- the order type;
- the order’s timing;
- whether to place the order;
- the sequencing of the order (compared to other orders);
- the order price;
- the order quantity;
- the partition of the order into smaller components for submission;
- the number of orders to be placed; or
- the management of orders after submission.
Applying a literal interpretation to the CFTC’s proposed rules, a pure order entry system that permitted persons to enter orders into a DCM’s matching engine, but automatically held back some limit orders’ placement until market prices were closer to the limit levels, seemingly would constitute an algorithmic trading system as it could be deemed to make determinations regarding the sequencing of orders.
Under the CFTC’s proposal, algorithmic trading would not include the manual entry of orders into a front-end system initially derived by an automated trading system. However, while approving the release of proposed Regulation AT for comment, Commissioner Sharon Bowen suggested the CFTC should consider including certain manual entry orders within the scope of the Commission’s new rules. According to Ms. Bowen,
[t]he definition of algorithmic trading is at the heart of this proposal and we need comments on this point. If there is evidence that a form of algorithmic trading poses systemic risks but is not captured by this definition, we should expand the definition to cover that form of trading.
CFTC registrants that engage in algorithmic trading, including FCMs, floor brokers, swap dealers, major swap participants, CPOs, CTAs and IBs, as well newly required to be registered floor traders (collectively, termed “AT Persons” by proposed Regulation AT), would be required to implement certain minimum pre-trade risk controls and other measures “reasonably designed” to avoid a so-called “algorithmic trading event.” The CFTC estimates that approximately 420 persons will qualify as AT Persons and potentially be impacted by this requirement
Under proposed Regulation AT, an algorithmic trading event occurs when there is a compliance breach of any magnitude, or an operational breakdown that is disruptive at any level or that constitutes a “material” degradation. An algorithmic trading event is formally defined as an “algorithmic trading compliance issue” or an “algorithmic trading disruption.”
An algorithmic trading compliance issue would occur when the algorithmic trading of an AT Person does not comply with relevant law, CFTC regulation, or the rules of a relevant DCM or the NFA. Such an issue also would arise when the algorithmic trading of an AT Person does not comply with the AT Person’s own internal requirements, or the requirements of its clearing member.
This provision, if adopted, might encourage AT Persons to hesitate before adopting any internal requirement other than minimal requirements required by the CFTC lest they potentially be penalized later for violating a voluntarily adopted higher standard. This requirement also might place on AT Persons an obligation to comply with FCM requirements of which they might not be aware.
An algorithmic trading disruption would be an event “originating” with an AT Person that “disrupts or materially degrades:”
- the algorithmic trading of the AT Person;
- the operation of the DCM where the AT Person is trading; or
- the ability of other market participants to trade on the DCM where the AT Person is trading.
It seems odd that, in the proposed rule, there is a materiality standard for degradations but not disruptions—particularly where the impact may only be internal to the AT Person.
The risk controls and other measures AT Persons should implement for their trading systems would include, but not be limited to:
- maximum order message frequency per unit time and maximum execution frequency per unit time;
- order price parameters and maximum order size limits;
- the ability to:
- immediately disengage algorithmic trading;
- cancel selected or all resting orders; and
- prevent submission of new order messages;
- for AT Persons with direct DCM access only, systems to indicate on an ongoing basis whether there is proper connectivity to a DCM’s trading platform and market data; and
- connection with all relevant DCMs’ self-trade tools.
Prior to initial trading, an AT Person would be required to alert each DCM on which it trades whether its resting orders should be cancelled or suspended if its algorithmic trading system becomes disconnected from the DCM’s trading platform. An AT Person would also be required, prior to initial trading, to notify its clearing member(s) and each DCM where it intends to trade, that it will engage in algorithmic trading. These requirements could prove challenging when an AT Person is not a direct customer of a clearing FCM.
Under proposed Rule AT, AT Persons would also be required to implement written policies and procedures related to the development and testing of their algorithmic trading systems. AT Persons also must implement procedures to document the strategy and design of their software as well as changes that are implemented in the production environment. As mentioned, an AT Person must maintain “copies” of all source code used in a production environment in a source code repository, subject to the CFTC’s ordinary retention requirements (e.g., save for five years, two years in a readily accessible location), and potential inspection by CFTC or US Department of Justice representatives, upon request, without subpoena or other process of law.
The proposed requirements regarding source code do not appear to distinguish between AT Person-created software and third-party software.
The proposed CFTC rules also require AT Persons to have policies and procedures to ensure ongoing, real-time monitoring to detect potential algorithmic trading events, and to empower “monitoring staff” to be able to stop an algorithmic trading system from functioning (e.g., activate a kill switch) if system or market conditions warrant. Policies and procedures also must address review of algorithmic trading systems to detect potential algorithmic trading compliance issues. There must be a plan,
of internal coordination and communication between compliance staff of the AT Person and staff of the AT Person responsible for Algorithmic Trading regarding Algorithmic Trading design, testing and controls, which plan should be designed to detect and prevent Algorithmic Trading Compliance issues.
It is not clear what might be the potential liability of compliance staff, if any, if the AT Person plan to detect and prevent such compliance issues might, after the fact, be assessed inadequate.
AT Persons also must have policies and procedures to designate and train staff responsible for algorithmic training, and escalation procedures whenever an algorithmic trading event has been identified. The CFTC expects that “staff persons who are responsible for monitoring the trading of other AT Person staff should typically not be actively engaged in trading at the same time, because it would be difficult to adequately and consistently monitor trading of other AT Person staff while engaged in trading activities.”
An AT Person would be required “periodically” to review all its relevant policies and procedures, and to promptly document and remedy deficiencies. In addition, on an annual basis by June 30—for the period May 1 of the prior year to April 30 of the current year—an AT Person would be required to submit to each DCM on which it trades, a report describing its pre-trade risk controls and a certification by its chief executive officer or chief compliance officer that to the best of his or her knowledge or reasonable belief, the information in the report is accurate and complete. The annual report must include copies of written procedures addressing the development and testing of its algorithmic trading systems and its ability to detect potential algorithmic trading compliance issues.
There is no minimum threshold of activity that might exclude some potential AT Persons from registering as floor traders. However, in commenting on the Commission’s proposed rules, CFTC Chairman Timothy Massad suggested that the proposed registration requirement might not be cast in stone. According to Mr. Massad,
While we believe a registration requirement is appropriate, we have also invited market participants to comment on whether there are alternatives that can achieve the proposal’s underlying objectives. We have also asked whether the registration requirement should be limited to trading firms meeting certain volume, order or message levels – or be based on other characteristics.
All AT Persons must also be members of the NFA under the CFTC’s proposed new rules.
Clearing member FCMs would also have obligations under the CFTC proposed rules. Principal among these requirements would be to utilize pre-trade risk controls “reasonably designed to prevent or mitigate an Algorithmic Trading Disruption,” and to ensure that natural persons are promptly informed when pre-trade risk controls are breached. For direct access clients, FCMs are solely required to implement pre-trade risk controls and order cancellation systems provided by DCMs, while for non-direct market access clients, FCMs must establish and maintain their own pre-trade risk controls and order cancellation systems.
Clearing member FCMs must also "make use of" order cancellation systems required for all AT Persons, although it is not clear under what circumstances or within what time frames.
Each clearing member FCM also would be required to file with each DCM on which an AT Person customer trades an annual compliance report by June 30 of each year—also for the period May 1 of the prior year to April 30 of the current year—that describes how it complies with its maintenance of pre-trade risk control requirements. The FCM’s CEO or CCO would also have to certify to the best of his or her knowledge or reasonable belief that the information in the report is accurate and complete. This report would be in addition to compliance reports that FCMs are already required to file with CFTC annually.
DCMs and NFA:
DCMs generally would have obligations to:
- implement pre-trade and other risk controls to “address” risks from algorithmic trading. Some of these capabilities must be made available to clearing member FCMs and AT Persons;
- provide testing environments; and
- review AT Person and FCM compliance reports.
DCMs would also be obligated to implement risk control measures for orders that originate from manual or other non-algorithmic trading systems, "after making an adjustment to the mechanisms that the [DCM] determines are appropriate for such orders." It is not clear what might precisely be expected of DCMs in such circumstances. However, the Commission indicated that "certain activity" that pre-trade and other risk controls are intended to ameliorate,
...can be caused by manual order entry in addition to Algorithmic Trading. For example, fat-finger errors are a commonly-cited example of an unintentional error that can have a significant disruptive effect, which can be caused by, and may even be more likely to occur in the context of, manual order entry.
In addition, DCMs would be required to implement rules “reasonably designed” to prevent self-trading by market participants, including either applying or providing and requiring use of a self-trade prevention tool for traders. DCMs would be required to publish certain information on their websites regarding self-trades.
As part of its proposal, the CFTC also would require DCMs to:
- provide to the CFTC and publish on their website statistical and other information regarding proposed market maker and trading incentive programs, as well as the operation of their electronic matching platform or trade execution facility that materially impact the time, priority, price or quantity of execution or the ability to cancel, modify or limit display of market participant orders; and
- prevent payment of market maker or trading incentives for trades between accounts under common beneficial ownership.
The CFTC did not adopt requirements suggested by commentators to the Concept Release, that, to enhance surveillance of algorithmic traders by traders themselves and FCMs, drop copies of trade information should be made available for all DCMs and products whenever technologically practicable, and that trade reports and other information provided by drop copies should be made available in real-time or as near real-time as practicable. (Click here to access recommendations of the Futures Industry Association dated September 2013 regarding drop copies.)
NFA also would have a new obligation under the proposed Regulation AT. This would be to implement and maintain “a program for the prevention of fraudulent and manipulative acts and practices, the protection of the public interest and perfecting the mechanism of trading on designated contract markets…” The CFTC would expect NFA to accomplish this by adopting rules for each category of member that require:
- pre-trade risk controls and other measures for algorithmic trading systems;
- standards for developing, testing, and monitoring algorithmic trading systems, and compliance;
- designating algorithmic trading staff and training such persons; and
- “operational risk management standards” for FCMs for orders originating with algorithmic trading systems.”
Although the relevant rule speaks definitively in terms of the NFA “must establish and maintain a program,” the accompanying CFTC release offers that NFA’s apparent obligation is not mandatory. According to the CFTC, the relevant rule,
would not require [NFA] to issue any rules… where [NFA] believes they are unnecessary. Rather the proposed regulations leaves discretion to [NFA] to determine what rules would prevent fraudulent and manipulative acts and practices, protect the public interest, and perfect the mechanisms of trading on DCMs.
(Proposed Regulation AT addresses the requirements of registered futures associations generally. However, NFA is the only RFA at this time.)
NFA would also have to accommodate the registration and membership of all AT Persons.
SEFs are not included within the scope of Regulation AT because of the newness of those markets, according to the Commission.
Costs, Benefits and Miscellaneous:
The release accompanying proposed Regulation AT has an extensive discussion regarding the potential costs and benefits of the proposed CFTC rules. The CFTC makes reasonably clear what are its assumptions regarding its estimations. The agency invites comment on whether its assumptions are accurate.
While voting for the release of proposed Regulation AT, Commissioner J. Christopher Giancarlo expressed concerns regarding the potential costs and benefits of the proposed rules. He commented,
…after reading through the almost five hundred pages of this proposal, I am left with one major question that I still cannot answer. That question is: does this proposal sufficiently benefit the safety and soundness of America’s futures markets so as to outweigh its additional costs and burdens. I wish the answer was clearer.
Among other things, the CFTC analysis concedes that the cost of implementing the proposed rules will have a deleterious impact on at least some FCMs. According to the CFTC,
the Commission acknowledges that the compliance requirements of Regulation AT could have adverse effects on small clearing firms. [T]he rule has some potential to contribute to increased concentration among firms, i.e., fewer competing clearing firms.
Earlier this month, in a speech before the US House of Representatives Committee on Financial Services, Mary Jo White, chair of the Securities and Exchange Commission, said staff of the SEC is also currently working on a proposal to require registration “of certain active proprietary traders and improvements of firms’ risk management of trading algorithms.” (Click here for further background in the article, “SEC Chair Tells Congressional Committee Disruptive Trading Rule and Proposals Mandating Third-Party Compliance Reviews of Investment Advisers Coming Soon; CFTC Reg AT Proposal Coming Sooner;” in the November 22, 2015 edition of Bridging the Week.)
My View: According to the CFTC, “[t]hrough proposed Regulation AT, the Commission is taking its next steps in ensuring that its regulatory standards and industry practices properly address current and foreseeable risks arising from automated trading, and promote responsible innovation and fair competition among markets and market participants.” These are noble goals, and proposed Regulation AT and its accompanying release clearly reflects the thoughtful work of staff trying to implement these objectives.
However, while the CFTC acknowledges existing best industry practice and existing rules and requirements of designated contract markets and the National Futures Association related to automated trading systems, Regulation AT, if adopted, would impose additional layers of requirements that appear would add, at most, marginal benefits, while imposing substantial costs.
Moreover, in an effort to enhance compliance with what are now best practices, the CFTC potentially would cause some trading firms to consider not implementing new and innovative risk control procedures and even rolling back already installed best practices. This is because the CFTC proposes to include in the definition of an algorithmic trading compliance issue a failure of an AT Person to comply with its own compliance procedures, in addition to relevant law and rules. Under that scenario, an AT Person might decline to implement as a formal requirement any best practice above a CFTC minimum requirement, when its reward for being innovative and top in class is a potential regulatory violation and sanction.
The requirement that AT Persons make available their source code to CFTC and US Department of Justice staff for inspection—not solely pursuant, as now, to subpoena or other lawful process of law—will also rub many as substantial overreach and potentially dangerous given competition concerns and concerns about US government cybersecurity.
It also appears odd to more formally involve the National Futures Association in the oversight of AT Persons and FCMs related to algorithmic trading when, historically, it has been DCMs that exercised self-regulatory oversight regarding trading on their facilities and their clearing members.
My view is that it is appropriate for the CFTC to seek the high level objectives of proposed Regulation AT. However, the better way to achieve these objectives would be to more fully build upon approaches already implemented by DCMs and the NFA, and to encourage the development and implementation of further best practices by the algorithmic trading industry rather than construct a new regulatory infrastructure.