Who turned out the lights?
Dark liquidity, which started as a way to hide big trades,now mostly offers liquidity in bitty, small packages. But Martin Steward finds signs that the pendulum is swinging back again
Dark pools of liquidity now account for about one-tenth of European equity trading volumes, split equally between private internal broker crossing networks (BCN) and public ‘multi-lateral trading facilities' (MTF). Trading in the dark allows large orders to be filled while minimising market impact, which saps $50-100m from investors' returns in the US alone according to Liquidnet - a dark venue whose members' trades sometimes account for 25% of a stock's average daily volume. It also saves on exchange fees and is usually executed at the mid-point of the on-exchange bid-offer spread.
Dark trading originated with stockbrokers' upstairs trading desks. In this form, the problem was that brokers would have to make calls to establish interest in liquidity, or worse, blab over lunch with clients or prop-desk colleagues. Just as ‘parent' orders are split into ‘child' orders on-exchange, traders would therefore split large orders between brokers. During the 1980s, the process was automated and in Europe MiFID led to an explosion of electronic dark pools.
"It's gone from ‘Sir Bufton Tufton' who's very well-connected in the City to anonymous electronic trading," says Peter Randall, CEO at Equiduct and co-founder of Chi-X, the first pan-European MTF.
That should have removed much of the threat of information leakage, opening the door for re-consolidating those ‘child' orders. However, LiquidMetrix data suggests that, aside from large-fill specialists like Liquidnet and ITG POSIT, average order sizes on European dark pools are about €10,000 - not much different from what is possible on lit exchanges. Why? As post-MiFID competition erupted, technology sped up order processing. This created a self-feeding cycle of lots and lots of very small bargains, greater activity by high-frequency traders (HFT) and narrowing bid-offer spreads.
"We've done this to ourselves with greater electronification and the tighter spreads," says Paul Daley, head of capital markets product development at SunGard, which operates as an algo-trading broker-dealer and also maintains its own dark pool. "There is less available liquidity at any one price point because the price points are closer together."
Larger asset managers had to adapt. They used to wait for a big match and ignore little orders as noise, says Allianz Global Investors Capital's head of equity trading, Kevin Chapman. "But once the market fragmented, all of those 100-shares-at-a-time trades started adding up to serious liquidity."
Émile Marshall, equities dealer at Danish pensions giant ATP, which executes about 15-20% of its orders in dark pools, says he has lost count of how many dark pools there are. "It means you really have to split your order up and shop around for your liquidity," he says. "And because everyone is shopping around, orders don't sit in the venues for a day anymore, but for five minutes, which means that it's much more difficult to find the liquidity I need. There are simply too many venues."
Large fills can still get matched. Compared with the €10,000 average execution on most European dark pools, Liquidnet Europe clocks in at more than €800,000, according to LiquidMetrix. It does this by closing out HFTs while welcoming more than 700 of the world's largest asset managers and owners, but also by deploying technology that can aggregate and call upon all of the liquidity sitting as ‘parent' orders on members' books.
"This is the most valuable information on Wall Street," says Liquidnet CEO Seth Merrin. "Our members would never give us access to that if we did not provide them with a guarantee on the level of safety."
First-mover advantage goes a long way to explain why all dark pools don't emulate Liquidnet. But there are also confidentiality concerns - the SEC recently looked into its use of members' data for marketing. Competitors describe the model as ‘blotter scraping', which says something about the battle over perceptions in this market.
Can traders get size elsewhere? Fabien Orève, global head of the dealing desk at Dexia Asset Management, questions how much added-value Liquidnet's automation of buy side-to-buy side matching brings to a desk, like his, which maintains substantial telephone-based trading with its brokers. This is where he goes for ‘natural' matches to his large blocks, before prioritising some favoured BCNs - MS Pool, PIN, CITI Match, MLXN, Cross Finder and SuperX - where he says he can match average orders around 20% bigger than most public MTFs.
But this just shows that Orève and other traders expect to run out of ‘natural' liquidity and move into pools where volumes are higher but orders smaller. For orders below 5% of average daily volume, Orève deploys algorithms across both lit and dark venues; above 5% he will get as much as possible done in ‘natural' matches before spilling over into BCNs and, finally, public MTFs.
"The dark pools where you can trade large size have traditionally been buy side-only," as Mark Hemsley, CEO of BATS Chi-X Europe, puts it. "But unless you can hit a very high volume of activity through those, you don't get executed."
Even at Liquidnet, which Merrin describes as uniquely focused on execution size while others focus on volume, the fact is that for three years its ‘natural' buy-side liquidity has been supplemented with ‘supernatural' liquidity from the sell-side and other dark pools. It seems that the choice is not between one size and another size, but between size and speed of execution.
"If there's an arbitrage trader who is ready to trade right now at a good price, it might suit you best to trade one-tenth of your order with them," suggests Turquoise CEO Adrian Farnham. "It could be the difference between implementing your portfolio idea or watching it go stale."
It is a delicate balance. Fewer transactions means lower transaction costs, but a bigger gamble that you miss the ‘best execution': the more price points that are available across more venues, the higher the probability that the one price you opt for will be off-the-pace for that day. And that balance changes with market conditions: Orève says that his desk has been prioritising liquidity-seeking - "buy as much as you can around here light or dark" - over minimising implementation shortfall during the more volatile recent months. Chapman at AGIC is outspoken about implementation shortfall being the most objective measure of best execution, and argues that it would make dark pool traders more patient again. "But you need fund managers to be onboard, and in this volatile environment the fund managers tend to want to get the portfolio changes done as soon as possible."
That can upset the other important balance, between speed and market impact. "There is information leakage from dark pools as well as lit," warns Rhodri Preece, director of capital markets policy at the CFA Institute. "There are still algorithms ‘pinging' all of these different pools to sniff out orders."
The more a trader hits the market to fill a big order, the more signals he sends about the presence of that big order; and the smaller the hits, the cheaper it is for smart algos to pick up those signals by providing liquidity.
This is the reason able price of HFT liquidity - "A lot of people want information without giving any information away, but you do need to participate," says Chapman - but large institutions probably interact with them more often, slicing orders into smaller sizes, than is strictly necessary. Chapman, for one, argues that falling volumes since 2008 have reduced the need to track the noise of every tiny packet of liquidity, making a case for more patient trading.
And MTFs are responding. In mid-June Turquoise changed its dark pool's liquidity prioritisation from the traditional first-come, first-served model to one that matches the largest orders first. It also offers a unique function that interrupts continuous liquidity-matching with a randomly-scheduled mini auction designed to match ‘patient flow' and deter smart algos, and now allows members to stipulate minimum execution sizes - if you stipulate blocks of 1,000 the system won't try to match you up with five buyers of 200.
"Since implementing these changes, our average execution size has increased by about 25%," says Farnham. "Generally, as a venue, we are one step removed from the buy-side, but we have taken our ‘random periodic uncross' function to them and they like the idea, especially combined with size priority and minimum execution size. These customers want access to dark liquidity but in a way that gives them some control."
The demand from the buy-side is clear. Orève, for example, insists that its brokers apply a minimum fill of $15,000 on any flows routed to external dark pools. Marshall at ATP prioritises BCNs because he is more likely to be able to get minimum fills there than in the MTFs, and he remains sceptical of MTFs' attempts to provide similar solutions. It is arguable that the Liquidnet model evolved thanks to its founders' relationships with the buy-side community as broker-dealers - they were simply better attuned to the buy-side's needs. The rest of the dark liquidity community may be catching up. We won't see every MTF matching an average €800,000 per trade; but we should start to see more activity between the extremes of €10,000 and €800,000 - making best execution much less of a headache for larger institutions.