What are 'Dark Pools' for?

Depending on who you ask, you will get different answers to the question. As ever, some context is helpful.

First, it’s important to recognise what has happened to our ‘lit’ markets. Competition has driven trading and clearing fees lower (including rebates for firms adding liquidity to order books), allowing traders to profit from smaller bets on smaller price movements – so the prevalence of higher-frequency proprietary-trading has increased (and will continue to do so as costs fall further). At the same time, partly in the pursuit of best execution in light of the first phenomenon, and partly to reduce their own costs, brokers have increased their reliance on algorithmic trading. Both of these factors have lead to an ‘atomisation’ of liquidity in the market – with smaller, faster orders and trades – an environment which poses challenges for those trying to execute institutional-sized orders without being discovered or taken advantage of by short-term/momentum traders. Dark pools, with no pre-trade transparency, offer an alternative in which market impact should be lower, although certainty and immediacy of execution are also sacrificed.

Second, as buyside institutions have increasingly used ‘low cost execution channels’, brokers have become ever more sensitive to the proportion of the commissions they earn which are paid away to exchanges, MTFs and CCPs. Broker internalisation has evolved from (expensive) sales-traders matching blocks telephonically towards (inexpensive) automated matching in broker-operated crossing networks which are free (for the broker at least) and which (by virtue of not using a CCP) don’t incur clearing charges either.

For brokers executing institutional orders, two plausible answers to our opening question could be;

• “The pursuit of best execution, as an alternative to ‘atomised’ lit order books”
  This rationale would work for both broker crossing networks and MTF/exchange venues
• “Cost-avoidance and margin-preservation”
  This rationale is applicable to broker internal crossing networks only, since MTF/exchange dark pools are actually more expensive that lit books. So brokers using our midpoint book are putting the pursuit of best execution above their own margins.

It is the first of any of these answers that resonates with institutional investors, many of whom equate dark pools with less information leakage, lower market impact and better execution quality. Precisely because dark pools do not offer the same certainty or immediacy of execution offered by lit markets, institutions come to meet other price-sensitive (rather than time-sensitive) participants who care more about mitigating market impact than they do about immediacy. But institutions (and others) question whether dark pools are delivering on those promises.

And they have a point, because as soon as there’s some liquidity resting in midpoint dark pools, other participants or flows (with different objectives) are attracted to participate. Anyone about to lift the offer or hit the bid in a lit market should be tempted to ‘pass through’ the midpoint dark pool first for ‘price improvement’ (especially if it’s fast enough). So both for firms trading on behalf of clients and on their own behalf, there is another answer to our opening question;

• “For price improvement against the lit market quote for aggressive orders”
  

But – there’s an asymmetry here – with some participants resting passive orders in dark pools to avoid ‘atomised’ lit markets (hoping to meet counterparties of similar patient profile), and others using them for aggressive order flow destined for those same lit markets. This asymmetry leads to the passive participants being ‘adversely selected’ – trading against the same aggressive/immediacy-seeking flow that they wanted to avoid, and giving up half the spread unnecessarily (as this flow might otherwise have hit their bid or lifted their offer).

Some execution venues exacerbate this asymmetry by offering aggressive routing strategies that ‘bundle’ access to the dark and lit order books into a single high-speed order type – effectively baking this adverse selection into their market structure. They do so because it earns them much more money (they charge both sides of the trade in the midpoint book, rather than rebating one party in the lit book), encourages the use of their (higher priced) midpoint book as a way to ‘intercept’ aggressive liquidity before it reaches their lit book, and reduces the average time between submission of a passive hidden order and receiving a first execution (a factor to which many broker routing algorithms are extremely sensitive).

Unlike our competitors, and despite some market demand, Turquoise does not offer such strategies. Even though it would attract more flow and increase match rates, we believe that exacerbating this asymmetry would lead to measurably poorer execution quality for those posting passively in our midpoint book, and would ultimately undermine its value proposition.
So our own answer to the opening question is;

• “For meeting the legitimate market need for an alternative to atomised lit books, with a service that offers measurably different trading characteristics.”

It’s not helping anyone to create a service which mimics lit order book characteristics with the lights turned out, and so we’re prepared to forgo some short-term growth to build a differentiated market which offers a demonstrably safer place to trade institutional client orders. We invite brokers and their clients to measure the execution quality they receive in our midpoint book and to discuss the results with us.

We will return to this topic in a future post and explain some of the other ways in which we are helping brokers achieve superior execution quality in our midpoint book.

Where there’s fire there’s smoke....

A recent white paper by Themis Trading, titled “Data Theft on Wall Street” has stirred up a lot of debate. The authors allege that exchanges, ECNs and MTFs are deliberately disseminating data that allows sophisticated high-frequency trading firms to take advantage of less sophisticated market participants. Is this true?

They cite two particular problems;

• When matching non-displayed orders, exchanges/ECNs/MTFs such as Nasdaq, BATS, Turquoise and Chi-x reveal information about which side of the match was passive and which was aggressive – effectively alerting participants to the possible presence of hidden resting orders on one side.
  
• By reporting a consistent OrderID for multiple executions and amendments of non-displayed, iceberg & pegged orders, they allow participants to identify such orders and infer information about their pricing strategy or non-displayed size.

Since its acquisition by the London Stock Exchange Group, Turquoise has sought to differentiate its pan-European Midpoint Dark Book from competitors’ offerings by making it a ‘safer’ place for brokers to trade institutional client orders. Accordingly, Turquoise’s matching systems (both existing and new) do not reveal information about which side of a non-displayed execution is passive or aggressive respectively, and nor do they reveal if the same non-displayed passive order is executed against more than once. For our new matching system, these were deliberate design decisions which actually required specific software development – because the standard behaviour of other markets is to release such information. Why would they do that?

It’s not as sinister as the authors of the Themis Trading paper would have you believe.

When exchanges first started offering electronic order entry disseminating a book feed, participants wanted to be able to identify their own orders and executions in the public data feed. This allowed them to know their queue position in the order book, and to display this on a client front-end. It allowed them to perform better transaction cost analytics – by identifying which executions on the ‘tape’ were theirs. It allowed them to measure the latency of the public market data against their own Execution Reports. And it allowed multiple OMS and EMS systems within the firm to identify their own orders & executions without having to feed each system with drop-copies of the order entry/execution feed.

So participants initially welcomed these new enriched market data specs because they allowed for more sophisticated order & data management, and helped drive the evolution of sophisticated execution algorithms that need information on queue position to estimate execution probability.

But – the authors do have a point. The specs make it relatively easy to identify iceberg/reserve orders as soon as the visible peak is first refreshed, and also to identify pegged orders as soon as they are modified by the market. And whist Turquoise has unilaterally addressed the issues in relation to non-displayed orders (driven by the quest for a competitive advantage), a solution to improve the ‘anonymity’ of iceberg or pegged orders would require non-trivial development by vendors and brokers.

The authors call for immediate regulatory intervention. In my opinion that’s unnecessary - the exchange business is incredibly competitive, and if market participants truly want a ‘safer’ venue there will always be one or more to choose from.

We invite feedback from brokers, competitors, regulators and investors on our approach and our views.

Update – May 21st 2010

On May 19th & 20th, European buyside investors reacted to the Themis Trading white paper on by asking brokers to exclude the Chi-x and BATS dark books from their routing strategies. Many brokers did so, resulting in a significant drop in volumes in Chi-x and BATS midpoint order books. In contrast, the Turquoise midpoint book saw strong volume growth, as brokers directed their flow to venues trusted by their clients.

Chi-x and BATS were both quick to acknowledge the problem and to confirm to their market participants that they would mimic the Turquoise approach by the end of the week. This is a healthy evolution of market structure, and should in time increase confidence in the use of MTF dark books.

We believe that this episode demonstrates clearly that the buyside do care about the choice of venues brokers make, and that a singular focus on dark book match rates might be superseded by a more nuanced appreciation for the quality of liquidity to be found.
We also believe that our soon-to-be-introduced functionality, designed to make our dark pool safer for institutional order flow, will help differentiate the brokers that choose to promote and use these capabilities.