The Citadel Settlement, Off-Exchange Market Makers, and Giant Brokerages

The recent settlement between the U.S. Securities and Exchange Commission (“SEC”) and Citadel Securities[1] is a landmark in the market structure enforcement program. In a nutshell, the regulators targeted high-speed algorithms that opportunistically used different market data benchmarks and involved undisclosed order handling mechanics. Importantly, this settlement may have revealed more than just the SEC’s skillful exposure of yet another hidden wrinkle in the modern electronic marketplace or a description of discontinued practices at one firm. This settlement could shed light on the opaque overlap between high-frequency trading (“HFT”) and off-exchange market making, thus subjecting to scrutiny—whether regulatory, legal, or competitive—key players known as “wholesalers” or “internalizers.” It is also impossible to ignore that the settlement’s key points are already spilling over to other segments of the securities industry, as shown by several important lawsuits aimed at large retail brokerage firms. More broadly, this enforcement action may be shaping up as an assault on the stronghold of existing—and potentially widespread—practices relating to order handling, best execution, and payment for order flow (“PFOF”).

The settlement focused on two algorithms—or, in the industry lingo, “algos”—with the monikers FastFill and SmartProvide, which were run by Citadel’s wholesale market making unit. Both algos were triggered by price discrepancies between the consolidated and private data feeds, i.e., the “official” market data distributed by a designated Security Information Processor (“SIP”) and more detailed and inherently faster market data products offered by exchanges themselves.[2] The very existence of such trading strategies, which may be classified under the umbrella of “latency arbitrage,” and a specific damages estimate are contrary to “myth-busting” efforts practiced in some industry circles. Likewise, this settlement undermines the so-called “Berkeley Study,” which concluded that off-exchange market makers can neither profitably engage in data feed arbitrage by “filling marketable orders at (or within) the SIP-generated NBBO [National Best Bid and Offer] . . . at stale prices to the disadvantage of retail investors” nor “choose as their pricing benchmark the slower SIP-generated NBBO to boost their performance metrics.”[3] Ultimately, this study’s assumptions skirted the fact that relevant strategies do not rely on choosing either the SIP or direct data feeds but on cherry-picking individual trades, as illustrated by the Citadel settlement. More to the point, using a slower version of market data consistently for marking up customer trades or for regulatory reporting is not necessarily more advantageous for the off-exchange market maker in question. After all, any given discrepancy is not automatically in that firm’s favor.

In some circumstances, FastFill presented a perfect illustration of latency arbitrage: “[C]ontemporaneous with determining to internalize the order at the SIP NBB [National Best Bid] or NBO [National Best Offer], as applicable, FastFill sent a proprietary order to the market in an effort to execute for itself at a price better than the SIP NBB or NBO, as relevant.”[4] However, given the price guarantee feature for orders exceeding the available displayed liquidity, “FastFill improved the overall execution price for a substantial number of (predominantly larger) orders.”[5] On the other hand, certain orders were disadvantaged: “[A] substantial number of smaller orders fared worse because of FastFill in that there was sufficient liquidity displayed in the market to fill all or most of such orders at a price better than the SIP NBB or NBO, as applicable.”[6] The regulators also observed that “the order often received price improvement, but this amount often was not sufficient to equal the price difference that had triggered the [underlying] strategy.”[7] In turn, this observation casts doubt on the accuracy of price improvement statistics reported by off-exchange market makers. Interestingly, some years ago, IEX also questioned seemingly favorable price improvement statistics: “[W]hen asked about the 90% of price improvement realized on . . . retail trades [that some brokerage firms’ Rule] 606 reports support, IEX [stated that such statistics] could be from ‘stale pricing’ from the SIP as well.”[8] Ultimately, the respective magnitudes of price improvement and losses from transactions amounting to trade-throughs need to be compared. In the extreme scenario, all gains from price improvement could be wiped out if enough exploitable price jumps measured in pennies outweigh more frequent and yet inherently smaller subpenny gains.

Turning to SmartProvide, this algo had a number of interesting twists. As its pivotal feature, SmartProvide converted marketable orders into nonmarketable orders, which could have been motivated by Citadel’s deliberate decision to capture liquidity rebates offered by exchanges. Moreover, this algo introduced significant time delays. More specifically, an order could end up being “displayed for up to one to five seconds, depending on the size of the order,”[9] and this timeframe is much longer than a typical delay of less than one millisecond, i.e., one thousandth of a second, for the consolidated data feed compared with faster private data feeds.[10] In other words, this algo went beyond a simple data feed arbitrage, and, as one might speculate, it probably involved additional predictive number-crunching and HFT-style market structure shortcuts. Ultimately, despite being advantageous to some orders, SmartProvide led to a subset of orders “receiv[ing] a price that was worse than they would have received” in the scenario of immediate execution.[11]

Starting well before the Citadel settlement, several commentators, including the author, have expressed doubts about the lucrativeness of pure latency arbitrage between the consolidated and private data feeds conducted by off-exchange market makers. For one thing, these two benchmarks are rarely misaligned, which should also be considered in tandem with the arrival rate of directed orders. According to an industry study, “For most stocks, [the consolidated and private data feeds] are identical 99.9% of the time—that’s all but 23 seconds of the day.”[12] The facts in the settlement lend some support to this observation: “FastFill executed approximately 2.7 million retail orders between June 2008 and January 2010, which amounted to approximately 0.4% of CES’s [Citadel Execution Services’] overall order flow during that period.”[13] During the same time period, “SmartProvide handled approximately 690,000 marketable orders. Of those orders, approximately 490,000 orders were internalized following handling by SmartProvide.”[14] The settlement’s disgorgement amount was set at $5.2 million, which is a significant but not exorbitant figure as a proxy for damages, although the specifics of these calculations have remained undisclosed and thus are subject to reevaluation. Once again, as maintained by the regulators, some customer orders had realized tangible gains from these undisclosed order handling mechanics, although the aggregate magnitude of such gains is unclear. On the other hand, as analyzed above, SmartProvide had a very different timeframe compared with data feed discrepancies, which suggests that at least some order handling delays by off-exchange market makers may be problematic for a different reason. Importantly, recent voluntary disclosure by several off-exchange market makers indicates that the speed of execution of market orders is still low in relative terms,[15] which routing brokerage firms seem to tolerate. In fact, this speed is much lower than the famous—and much-criticized—350-microsecond bump pioneered by IEX, which has had quite a bit of influence on other speed bumps, such as the Chicago Stock Exchange’s proposed Liquidity Enhancing Access Delay and NYSE MKT’s proposed Delay Mechanism.

Importantly, the enforcement action against Citadel was based on the existence of affirmative misrepresentations made by the firm as a de facto stand-alone trading venue. As noted in the settlement, “During the relevant period, CES [Citadel Execution Services] provided a written disclosure to certain retail broker-dealer clients that described a market order as an ‘[o]rder to buy (sell) at the best offer (bid) price currently available in the marketplace,’ and made other, similar representations to its clients [which] suggested that CES would either internalize the marketable order at, or seek to obtain through routing, the best bid or offer from the various market data feeds CES referenced.”[16] Perhaps the most intriguing implication is that the regulators decided not to pursue the approach based on breaches of the duty of best execution. In fact, this duty may apply to entities other than customer-facing brokerage firms, and it may cover off-exchange market makers as a result of “a voluntary assumption of this duty, the existence of a broad agency relationship, or the reach of FINRA’s regulatory requirements.”[17] In Citadel’s case, at least one of its order handling agreements during the relevant period contained a provision mandating the best execution standard.[18] Moreover, the SEC itself recently penalized another major player, KCG, for its off-exchange market making activities in “pink sheets” stocks based on breaches of the duty of best execution, which KCG voluntarily assumed.[19] Looking forward, off-exchange market makers may become a direct target of private lawsuits on the basis of breaches of this multifaceted duty—despite occasional efforts to subtly disclaim it. For instance, promptness is one of the dimensions of the duty of best execution, and any deliberate—and potentially gameable—bumps in either executing customer orders or marking them up would be contrary to this stringent standard. Subsequent routing of customer orders by off-exchange market makers may also become controversial, by analogy to the famous study of routing practices of retail brokerage firms.[20] Likewise, order handling functionalities offered by off-exchange market makers, such as not-held orders, may come under scrutiny.

The Citadel settlement has done more than just re-energize the market structure debate. Any discussion of this settlement would be incomplete without mentioning its effect on several recent lawsuits. The class actions in question are directed against E*Trade and Charles Schwab, two giants among brokerage firms, in connection with alleged breaches of the duty of best execution.[21] Among other things, these lawsuits attack captive order flow arrangements and PFOF practices, which constitute the backbone of the business model of off-exchange market making. While the phenomenon of PFOF has never been declared illegal per se, its long history is a continuous battle between lower execution costs and distortionary conflicts of interest detrimental to investors. Naturally, an extrapolation of the practices described in the Citadel settlement would give more ammunition to any current or future plaintiffs. While not mentioning the settlement by name, the Charles Schwab complaint contains a detailed description of data feed arbitrage by off-exchange market makers.[22] The E*Trade complaint specifically mentions the Citadel settlement, while stressing that the best execution standard had not been met in those circumstances.[23] Moreover, this complaint provides a sample of retail orders that appear to have been abused, and of one such transactions took place on a dark trading venue and with a significant delay, suggesting the circumstances “when latency arbitrage opportunities existed for sophisticated off-exchange market makers.”[24]

It remains to be seen whether these lawsuits will shed more light on arrangements between off-exchange market makers and brokerage firms, given that the relevant terms are typically undisclosed and, in some cases, poorly documented. In any event, the ongoing courtroom battles are likely to strain, if not disrupt, PFOF arrangements, which are still seen as a key competitive tool. Unsurprisingly, Virtu’s recent bid for KCG was apparently motivated in large part by KCG’s captive order flow arrangements.[25] While the PFOF model as such will probably survive, certain practices of off-exchange market makers and brokerage firms are ripe for scrutiny and eventual elimination. Furthermore, the overlap between HFT and off-exchange market making is important because it typically involves agency-based functions and thus the necessity of complying with the duty of best execution, despite the common perception that HFT firms have no customers. Overall, despite the long-term trend of diminishing transaction costs for retail investors, their order flow is still vulnerable to abuse. The resulting tangible harm, which may upend the picture-perfect execution quality statistics, is a problem that calls for vigilance from the regulators and the plaintiffs’ bar.

ENDNOTES

[1] Citadel Sec. LLC, Securities Act Release No. 10,280, Exchange Act Release No. 79,790 (Jan. 13, 2017) (settled proceeding), https://www.sec.gov/litigation/admin/2017/33-10280.pdf.

[2] For the author’s earlier analysis of the phenomenon of data feed arbitrage by off-exchange market makers, see Stanislav Dolgopolov, Wholesaling Best Execution: How Entangled Are Off-Exchange Market Makers?, 11 Va. L. & Bus. Rev. 149, 185–96 (2016), https://ssrn.com/abstract=2744904.

[3] Robert P. Bartlett, III & Justin McCrary, How Rigged Are Stock Markets?: Evidence From Microsecond Timestamps 2–4 (Nat’l Bureau of Econ. Research, Working Paper No. 22,551, 2016), http://www.nber.org/papers/w22551.pdf.

[4] Citadel Sec. at 8.

[5] Id. at 9.

[6] Id.

[7] Id. at 8.

[8] Richard Repetto & Mike Adams, Sandler O’Neill + Partners, A View of Market Structure from IEX 3 (Apr. 9, 2014), https://www.thefinancialengineer.net/wp-content/uploads/2014/04/IEX_ViewMarketStructure0414.pdf.

[9] Citadel Sec. at 4.

[10] This statistic was discussed in a recent industry study, which also cautioned that “[a]round 30% of all trading happens as quotes change.” Phil Mackintosh & Ka Wo Chen, KCG Holdings, Inc., The Need for Speed V: How Important Is 1 ms? 2 (May 2016), https://www.kcg.com/news-perspectives/article/the-need-for-speed-v-how-important-is-ms.

[11] Citadel Sec. at 9–10.

[12] Mackintosh & Chen, supra note 9, at 2.

[13] Citadel Sec. at 8.

[14] Id. at 9.

[15] For some summary statistics, see Dolgopolov, supra note 2, at 182 n.105.

[16] Citadel Sec. at 10.

[17] Dolgopolov, supra note 2, at 197.

[18] Id. at 169 & n.61

[19] KCG Ams. LLC, Securities Act Release No. 9996, Exchange Act Release No. 76,705 passim (Dec. 21, 2015) (settled proceeding), https://www.sec.gov/litigation/admin/2015/33-9996.pdf.

[20] Robert Battalio et al., Can Brokers Have It All? On the Relation Between Make-Take Fees and Limit Order Execution Quality, 71 J. Fin. 2193 (2016).

[21] For the most recent complaints, see Second Amended Complaint for Violations of Federal Securities Laws, Schwab v. E*TRADE Fin. Corp., No. 1:16-cv-05891-JGK (S.D.N.Y. Feb. 10, 2017); Amended Class Action Complaint for Violations of Federal Securities Laws, Crago v. Charles Schwab & Co., No. 3:16-cv-03938-RS (N.D. Cal. Jan. 20, 2017). A similar class action against TD Ameritrade, another large retail brokerage firm, has already passed the motion to dismiss hurdle. Zola v. TD Ameritrade, Inc., 172 F. Supp. 3d 1055 (D. Neb. 2016).

[22] Amended Class Action Complaint for Violations of Federal Securities Laws, Crago v. Charles Schwab & Co., supra note 21, at 23–27.

[23] Second Amended Complaint for Violations of Federal Securities Laws, Schwab v. E*TRADE Fin. Corp., supra note 21, at 31–32.

[24] Id. at 48–49.

[25] For a description of this view, see John D’Antona Jr., KCG Retail Order Flow at Stake in Virtu Grab, Traders Mag. (Mar. 16, 2017), http://www.tradersmagazine.com/news/hft/kcg-retail-order-flow-at-stake-in-virtu-grab-116026-1.html.

This post comes to us from Stanislav Dolgopolov, a regulatory consultant with Decimus Capital Markets, LLC. His work has involved some of the issues and legal actions discussed in this post.

9 Comments

  1. Curious George

    Question on Smart Provide:

    Smart Provide says it could be invoked if the referenced Direct Feed(s) was worse than the SIP. Does that mean it could apply if *any* of the referenced Direct Feeds were worse than *any* SIP quote? Or only when a referenced direct feed was worse than its *corresponding* SIP quote?

    For example, assume the SIP NBO in GOOGL is:
    100 NSDQ @ $929
    100 EDGA @ $930
    100 ARCA @ $931

    Say I place marketable buy limit order at $930.50. If Citadel referenced the EDGA Direct Feed and saw a price of $930 (matching the SIP EDGA quote), could they still apply Smart Provide because that price is worse than the NSDQ quote @ $929 (SIP NBO)?

    Thanks.

  2. Stanislav Dolgopolov

    As stated in the settlement, “The second strategy, known as SmartProvide, was triggered when the SIP NBB or NBO, as applicable, was better than the best price from at least one of the depth of book feeds.” (p. 4, paragraph 12)

    • Curious George

      Okay assuming if *any* referenced DF was worse than the single best SIP NBBO, Smart Provide could have basically been implemented at will, since there is almost always a venue away from the best price.

      Hypothetical: I’m handling client orders and passing back the fills but keeping the rebates, and I also use SmartProvide. Whenever I get a marketable order, I check to see if an “inverted” market is available at the NBBO. If so, I route to them immediately. If not, I post the shares on a traditional maker/taker market and try to obtain a rebate (or “SmartProvide”).

      Now to maximize my rebate and probability of being filled in the 5 seconds, naturally I’ll post visibly and at the minimum tick away from the NBBO. On occasion, upon my order entry the opposite side NBB or NBO moves (is “scared”) away from my order by several cents. At this point, I recheck to see if an inverted exchange is available and if so, I route immediately. If not, I cancel the order and internalize the shares $.0001 inside the new NBBO (that I probably altered) after 5 seconds.

      Could Citadehave been doing this?

  3. Robert Bartlett

    Stanislav,

    You state that the Citadel settlement “undermines the so-called ‘Berkeley Study’ which concluded that off-exchange market makers can neither profitably engage in data feed arbitrage by “filling marketable orders at (or within) the SIP-generated NBBO [National Best Bid and Offer] . . . at stale prices to the disadvantage of retail investors” nor “choose as their pricing benchmark the slower SIP-generated NBBO to boost their performance metrics.” You further state that our study “skirted the fact that relevant strategies do not rely on choosing either the SIP or direct data feeds but on cherry-picking individual trades.”

    We take strong issue with these statements. We believe them to be plainly inaccurate and to misrepresent what our study concludes. Please allow us to explain.

    First, we fully account for cherry picking. Panel C of Table 5 of our paper displays an upper bound to the profits available to liquidity providers seeking to take advantage of SIP latency. Across more than $3 trillion of SIP-priced trades in the Dow Jones, the total profits are about $500,000. For trades in non-exchange venues, the amount is $239,241. Both figures are for 10 months of trading. We further estimate these same figures for the entire market over the course of a year. Our estimate for all non-exchange trades is between $2.9 million and $3.5 million. Again, these are gross profits to non-exchange liquidity providers assuming perfect cherry-picking of all marketable orders.

    Second, the Citadel settlement covered the period between June 2008 and January 2010. We state clearly “Nor do our results rule out the possibility that latency arbitrage arising from stale SIP quotes might have been prevalent in the quite recent past (e.g., 2014), for the simple reason that our data are not available until mid-2015.” We additionally state in our conclusion that “Because our data commence in August 2015, we emphasize that these findings may very well reflect a new market environment in which the HFT strategies depicted in Flash Boys are less prevalent than in the past.”

    Our paper is available at the following link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2812123

    Our claims are narrow and carefully reached, as a thorough reading of our paper should make clear. We thank you for bringing attention to the SEC enforcement action against Citadel, which we agree is an important development in the law.

    Robert Bartlett & Justin McCrary

  4. Stanislav Dolgopolov

    As described (or rather sketched) in the settlement, SmartProvide was a much more complex algo compared to FastFill. One may only speculate how SmartProvide actually worked. In any instance, both algos were triggered by occasional discrepancies between the SIP NBBO and private data feeds (essentially, a synthetic NBBO with some depth-of-book issues probably coming into play). In the case of SmartProvide, such discrepancies served as a triggering event, but the actual time horizon was much longer. As I mentioned in my post, collecting rebates could’ve been a key motivation behind this algo, but retaining rebates from customer order routing isn’t necessarily wrongful per se. What matters is whether a rebate-maximizing strategy interferes with best execution.

    • Curious George

      I think it’s quite obvious that best execution was violated here (presumably for rebate-maximization, but the motivation shouldn’t matter).

      When Citadel thinks they can get a better price than the current SIP, they fill the customer immediately. Fine. On the flip side, SmartProvide was invoked when a discrepancy indicated the current quote might widen (crumble).

      So we know that Citadel has:
      1) Determined that a fill at a better price not likely (otherwise they’d internalize the current price and try to obtain it themselves); and
      2) Determined the quote is likely to move away soon; yet
      3) Posted the customer’s marketable order at a non-marketable price.

      Maybe Citadel was avoiding paying a rebate. Maybe they were trying accelerate the crumbling of the quote to open up a new price level where they could take from an inverted exchange or internalize (i.e. Spoofing?) — It doesn’t really matter. Whatever their motivation, Citadel is posting a client’s *marketable order* a price where Citadel doesn’t believe it will fill. How is that not a best execution violation?

  5. Stanislav Dolgopolov

    As a recap of my subsequent debate with Professors Bartlett and McCrary (from my point of view) . . . In my post, I described and cited the NBER version of their study, but the most recent revision was posted on SSRN on May 2, which was well after my post’s submission for publication (sounds like latency arbitrage, doesn’t it?). As opposed to the NBER version, the SSRN version explicitly recognizes and discusses the cherry-picking argument. I certainly applaud this change, as it helps conceptualize Citadel-like abuses. Moreover, I never wanted to suggest that the relevant time period for the Citadel settlement (2007-10) overlaps with the data sample for the Berkley Study (2015-16). In addition to being a historical data point contrary to the conclusions derived from the data sample from a *later* time period, the Citadel settlement may also have some relevance going beyond 2010. The very feasibility of latency arbitrage (despite its relatively small scale, as implied by the settlement) suggests that other players in this segment of securities markets may have used it later. Moreover, a complex algo like SmartProvide would be hard to detect in a later data set. Once again, SmartProvide was not a pure data feed arbitrage strategy, but it was still triggered by SIP-direct feed discrepancies.

Comments are closed.