The FX liquidity landscape is changing. In G10 markets especially, there is a definite decade-long trend towards lower so-called primary market volumes as a fraction of total lit liquidity, to the extent that it calls the name ‘primary’ into question for a number of key currency pairs. Additionally, as these are lit venues, the signalling risk of posting on primary remains largely unchanged, as the probability of a fill goes down. This may lead to a feedback loop of decreasing usage based on markouts. With algo volumes on the increase yet again over past quarters, as evidenced by surveys to which banks contribute their own volume data in order to receive data on the market as a whole, who is supplying the missing liquidity?
Secondary markets
Could it be the ECNs or so-called secondary markets? Many are cagey about supplying volume data, but by a combination of looking at the ones that do and studying outcomes of algos that are ‘democratic’ about sourcing liquidity across ECNs, it would appear that most ECNs have seen stable volumes over the past couple of years rather than any significant increases.
Additionally, several of the ECNs are known for attracting high-frequency market participants and maintaining last-look policies that serve on the one hand to tighten orderbooks but come at the potential expense of fill ratio on the other. These characteristics don’t necessarily invite an algo SOR to send vastly greater flows.
Dark Pools
Could it be dark pools of liquidity such as midbooks, where banks can net off against each other? First, only a few of these admit algo flows, and second, overuse results in penalties applied directly to the fill. There is no evidence to suggest that algos can be significantly more liberal in their use without incurring these offsets.
Futures Markets
What about futures markets? CME volumes in G10 are public information, and aggregating volumes in the majors suggests stable volumes or even a slight downturn as a percentage of FX total in recent years. In any case, the use of futures in spot algos comes with the additional cost of managing the basis risk between the instruments, so even if the trend had been upwards, we would probably not see much greater usage in spot execution.
Two directions
In our view, the missing volume has gone in two directions. One is greater use of bespoke venues. Setting up one’s own pool of liquidity offers the ability to tailor the pool to specific liquidity needs, and comes with a satisfying level of KYC.
The algo on the taker side maintains the smart order logic used with conventional liquidity, and so continues to adapt to changing market conditions and be intelligent about when to use what liquidity.
The other direction of travel is towards more intelligent internalisation, supplementing traditional internal orderbook matching and e-book transfer with franchise skews. Triggering a skew permits matching with other institutional investors. Deploying techniques that bank e-books already use to hedge risk allows an algo to profit from a very thorough analysis on skew safety that systematic market-making desks have already performed.
The result
If true, the result of all this of course is even less market visibility on FX volumes. Having gradually moved away from marking the calendar 3 years ahead for the hotly anticipated BIS Survey towards harvesting external market data, clients may find in future that FX volumes are a function of the liquidity provider. Tools such as Goldman Sachs’ Marquee Marketview will then be essential for traders to understand liquidity conditions, because the information will simply not be available elsewhere.
