Though part of the new iteration of the Markets in Financial Instruments Directive (Mifid II) and its associated regulation, Mifir, which became law on January 3, Esma had delayed the implementation of the DVC until March because of “data quality and completeness issues,” but published its first set of data for January and February on March 7—albeit for significantly fewer instruments than expected.

Dark pools have never liked the rules, which can suspend dark trading of any stock that exceeds a percentage of the market-wide volume—4 percent for a single pool and 8 percent for all pools—over the past 12 months. But they also fear the mysterious lack of data that prompted Esma to postpone the DVC—requiring an extra two months of data—could yet resurface.

If dark trading in a stock breaches these levels, local regulators can suspend trading waivers for the dark pools to trade the instruments for six months.

But even temporarily removing specific players from the market for a given instrument—either all dark pools, or an individual dark pool that triggers the 4 percent cap—has competition implications. “Regulators do not want, and are not mandated, to get in the middle of that,” says Juan Pablo Urrutia, European general counsel at ITG, which runs the Posit dark pool.

Other pool operators note that, despite the delay, the rules have already affected the way investors trade, steering them toward auction trading, block trades that benefit from the large-in-scale waiver, or to so-called systematic internalizers (SIs)—dealers that have a big share of a given market and are therefore subject to additional transparency requirements.

If dark trading shrinks voluntarily, of course, then the caps are less likely to be triggered. But some market participants still have reservations about whether trades are being classified correctly, and the response of regulators and policymakers if the caps do not function as expected.

Four major dark book operators tell Inside Data Management stablemate Risk.net that they submitted all data as requested. Two sources suggest problems with the UK Financial Conduct Authority’s (FCA’s) market data processor reporting system in the early days of Mifid II may have had some impact on Esma’s receipt of data from UK venues. Unlike most other jurisdictions that asked their venues to report directly to Esma, the FCA collected the double-volume cap data from UK venues and forwarded it to Esma.

The data reported by dark pools itself depends on reference data on individual equity instruments drawn from primary listings on national incumbent exchanges. If there were any delays or gaps in that reference data between the Mifid II go-live and January 9, it would stymie the reporting process for the dark pools.

“As the golden source reference data has to come from the primary market—that base record that Esma relies on—that puts a sequencing into the whole equation that means nobody else can submit, potentially, until the primary market has submitted its data,” says one industry source.

However, there are indications the problem was broader and more profound. Esma’s statement announcing the postponement said: “While Esma’s systems are functioning and ready to receive data, a large proportion of trading venues have yet to provide complete data.” The regulator indicated that only 75 percent of venues had submitted, and complete data was available for just 2 percent of the approximately 30,000 European equity instruments.

“If you get data from 75 percent of venues, unless the 25 percent of them that didn’t report represent 98 percent of the data and failed completely, you have to draw the conclusion that some of the 75 percent didn’t get through, for whatever reason. That could be because of erroneous submission, erroneous processing, or something in between,” says David Howson, COO at Cboe Europe, which operates both dark books and a periodic auction facility.

The additional two months were intended to provide time for Esma to tackle specific queries triggered by the initial data reporting, and validate the data it has. Sources suggest part of the challenge was that venues needed to fill the reports with data on the use of Mifid II waivers from lit trading, but the period covered—January 2017 to January 2018—was before the new rules came into force.

As a result, the waivers used at the time were actually those under the Mifid I criteria. Under Mifid II, only waivers for large-in-scale trades will be fully exempt from the double-volume cap, whereas trades using the reference price or negotiated (off-book, on-exchange) trade waivers will be subject to the caps, and therefore need to be reported in greater detail than under Mifid I.

“The requirement is for trading venues to collate and prepare data retrospectively that wasn’t required to be collected—to the granular level at which you need to give [it to] Esma—at the point when the trade happened. Obviously in the future, the problem will go away, because now everyone knows what is expected of them, so they can collect the right data; but there is an issue in the transition from one to the other,” says Christian Voigt, senior regulatory adviser at trading technology vendor Fidessa.

Individual Exemptions

Even the criteria for individual waiver types have been modified in Mifid II. Although negotiated trades in general count toward the cap, some are still exempt and can be executed in the dark without limitations. Mifir Regulatory Technical Standard 1 specifies certain types of negotiated trade, such as trades based on a benchmark, transactions that are part of a portfolio trade, or trades contingent on the execution of a derivatives contract that must be executed as part of a single lot, do not contribute to price discovery. They are therefore not eligible to be counted toward the caps.

“Getting all the right nuances and flagging all of the trades to make sure ‘Does this trade count or not, does it go in the numerator and the denominator, or just the denominator?’—that kind of granularity was not mandated in Mifid I,” says Cboe’s Howson.

He points out some of these flags were available under FIX Trading Community’s Market Model Typology, which Cboe’s platforms have supported since 2013. That typology is more likely to be used now that Mifid II is in effect, because it maps all of the Mifid II flags

“As we go forward, the data is going to get cleaner and easier to determine. But [given] the combination of the fidelity of the trade flagging under Mifid I and the amount of data—it was a full year’s data we had to submit—as well as the ingestion process by the regulators, it was not too surprising in the end that it was a challenging thing to do in that time,” Howson adds.

ITG’s Urrutia believes the policy decision to leave Esma with responsibility for all the data collection process was flawed. Instead, the regulator should have a consolidated tape provider—either a commercial entity or a public utility—perform the task, he says.

Mifid II has created a licensing regime for consolidated tape providers, but no company has so far applied to offer this service. Article 90 of Mifid II leaves open the possibility of a public procurement process if no firm steps up to provide a consolidated tape, instructing the European Commission and Esma to report on the matter by September 2019. Therefore, any public utility is unlikely to emerge before 2020.

Consensus on Reporting

But there is a more immediate issue to resolve in reaching industry consensus around how trades are reported under Mifid II, says Rebecca Healey, head of European market structure and strategy at global institutional trading network Liquidnet, citing the perceived substantial increase in trading volumes on SIs. Since these dealers must provide pre-trade transparency in equities via firm quotes, they are not subject to the double-volume cap.

According to data compiled by Liquidnet, average daily volumes for equities trading on SIs in January peaked at approximately €6.6 billion ($8.2 billion), around double the pre-Mifid II levels, once volumes have been filtered to exclude non-interactable liquidity.

Excluding the out-of-hours trades, average trade sizes are around €6,000, which may suggest SIs are not necessarily being used for the large, principal trades that policymakers had in mind when they allowed continued SI trading in equities, but instead may be serving as a substitute for the old broker crossing networks that are forbidden under Mifid II. Lack of industry consensus on what can be reported as an SI trade means some firms are including on-exchange, off-book negotiated trades within SI totals, whereas others are not.

“We as industry participants need to be clear on what we think is legitimate SI activity, versus what is activity that should really be reported on exchange—not necessarily as an SI-negotiated trade, but potentially as an off-book, on-venue transaction that should be under the trading venue. So there is an education process that needs to go on between industry participants in just trying to clarify what is truly addressable liquidity, versus that which is not really representative of trades a market participant could interact with,” Healey says.

Failing that clarification work by the industry, she is concerned that Esma and policymakers may be dissatisfied with the lack of transparency and potential confusion in understanding how trades have been executed. Alongside the increase in the use of systematic internalizers, there has also been a jump in large-in-scale trading. Fidessa data shows large-in-scale trades have risen from 12 percent of dark trades at the start of 2017 to more than 22 percent on February 9, 2018.

The other development has been the rise of Cboe’s periodic auction facility, which reported €296 million average daily notional value traded in January 2018—a rise of more than 885 percent compared with the fourth quarter of 2017. Although they do not provide continuous trading opportunities, periodic auctions allow the execution of large trades in the lit market with minimal price impact. Posit has now launched a periodic auction facility as well, while the London Stock Exchange’s Turquoise platform is expected to debut one this month.

The combination of periodic auctions, large-in-scale trades and systematic internalizer use could substantially limit the numerator for the double-volume caps. Liquidnet estimated around half of equity instruments could be capped, but of those, around 30 percent are very close to the boundary, so the universe of names subject to the double-volume cap could turn out to be substantially lower than previously thought. For example, 17 instruments in January and 10 instruments in February saw trading on a single dark pool exceed 4 percent of total volume across all EU trading venues over the past year. However, for the same months, 727 and 633 instruments, respectively, saw their percentage of trading across all EU trading venues exceed 8 percent of their total volume for the past year.

“[There is a need] to create industry guidelines that give the regulator the transparency they are looking for, provide politicians with the information they need that dark pools are not being misused and are actually there to provide price or size improvement, and give market participants the color they need so they can actually decide what are the most appropriate venues they can trade on, so they get the best execution they need to deliver to the end-investor,” Healey says.

Dates Distort Data

Before Esma published the DVC data earlier this month, market participants were unsure whether the regulator would use the 12 months to March, or the original 12 months to January that should have been used if the caps had been implemented on schedule. Some pointed out that it would make little sense to use January 2018 data to trigger caps in March. And Esma confirmed that it planned to implement the caps based on the data for the year to March 2019. But some venue operators say there are drawbacks to this approach.

Cboe’s Howson says it effectively extends the risk of a cap being triggered further into the future. The rolling averages for January and February will be beyond what they would otherwise have been if caps had been triggered on schedule, because dark pools continued eating into their 4 percent solo and 8 percent all-venues entitlements.

However, even after the postponement of the caps was announced, market participants continued to adjust their behavior in expectation of their eventual implementation. That means more use of large-in-scale, systematic internalizer and periodic auction trades, shrinking the size of the numerator for the first round of caps in March. That could mean fewer initial dark pool caps than policymakers had expected Mifid II to deliver.

“If you are looking at volumes March-to-March, the risk is there will be an insufficient number of instruments breaching the double-volume cap to deliver the anticipated reduction on trading in the dark. While that may be politically difficult to follow through, it is important to understand the differences in dark trading [between] that which can legitimately go back to lit venues versus wholesale activity, which needs the protection of the dark to deliver best execution to end investors,” says Liquidnet’s Healey.

Voigt says this is exactly the purpose of the caps—to change and curb the use of dark pools. He points to Article 5, paragraph 7b of Mifir, which explicitly states that “Operators of trading venues shall be obligated to have in place systems and procedures to ensure it does not exceed the permitted percentage of trading allowed under those waivers… under any circumstances.”

In other words, dark pools are supposed to reduce their footprint precisely to avoid triggering the double-volume caps. While this is difficult to do on the aggregate 8 percent level prior to Esma releasing the public data on dark pools, Cboe has already published data on its individual compliance with the 4 percent cap, and Voigt expects other venues will also manage their compliance with the solo cap.

“I could easily see a mechanism where venues voluntarily, if they get close [to the cap], interrupt trading—not for a pre-defined six-month period, but maybe for a month, waiting until their moving average reduces to a level that allows them to reopen again. In my mind, the benefit of that mechanism is that dark pools comply with their requirement not to breach, but more importantly, it gives them some flexibility about the length of the interruption. Once they breach the 4 percent, there is no discretion; it is a six-month interruption,” Voigt says.