Is Code-Only Liquidity Better? Not Always Says Euronext FX
Posted by Colin Lambert. Last updated: December 2, 2022
With The Full FX View
The latter half of 2022 has seen three major FX platforms announce that at least a portion of their business is defaulting to the provision of liquidity from FX Global Code compliers only.
As one of those three platforms, Euronext FX has also published a research note, penned by Paul Besson, head of quant research at Euronext and Mehdi-Lou Pigeard, quant research analyst at the exchange group, which provides the results of a comparative study of Code and non-Code LPs’ performance on its venues. The paper finds that while the performance of Code compliers is generally better, in almost a quarter of sessions non-Code LPs “improve the quality of liquidity on Euronext FX”.
Non-Code makers account for 32% of turnover on Euronext FX, something that suggests the presence of a group, or a dominant, non-bank market maker, given how just about every major bank in the world is signed up to the Code. The paper argues, however, that “no significant differences” are observed on takers’ spreads or markouts, observing, “This dispels the preconception that non-Code makers would display more leakage and larger markouts.”
The research uses data on Euronext FX and Euronext FX Singapore over a three-month period from 1 June to 31 August 2022, when some $1.2 trillion was executed on the venues. It seeks to ascertain LP performance based upon the aforementioned spreads and markouts, but also around cost of rejects and slippage.
The Full FX View
The results of this research have been flagged previously in these pages, but with the increased detail a clearer picture emerges – and I am less than convinced I agree with Euronext’s conclusion that there is room for non-Code compliant LPs in a pool.
To me, as is the case with an externaliser in an aggregated liquidity pool of internalisers, there are second-degree effects. In this case Code-compliant LPs are very likely to behave differently if they know they are in competition with non-compliers. Their spreads are likely to be wider, and their returns slimmer because, as the research hints at, non-Code compliant LPs are rejecting more at higher cost – in other words they are using last look to cherry-pick, and that takes potentially valuable flow off the compliers’ tables.
I still prefer the idea of a liquidity pool that offers an even playing field, and in fairness to Euronext, it is recognising this by defaulting to Code-only – it’s just that hint that it’s OK to have non-compliers in there as well that I don’t like. Of course, some clients, with certain trading styles won’t really care, and that is probably where Euronext is coming from in its conclusion that pools should be built on a case-by-case basis, but I am not sure that is place many people want to play – a pool with predatory takers and makers playing games!
I also accept that a lot of the results are not “statistically significant”, however there are differences, not least in that non-Code makers have higher reject rates and expected slippage (cost) and surely that is the biggest factor when judging the value of an LP? It could be argued that the “realised spread” analysis does not include cost of rejects, which could provide a more accurate picture and would, given the higher reject rates by non-Code makers, likely show an advantage to Code makers.
The bottom line of course, is that many, myself included, would argue that if there is no statistical difference, why run the reputational risk of having non-Code makers on the platform? Indeed, Euronext itself stresses that Code-compliant counterparties are better. Without the non-compliant players it is very unlikely that the platform would see flow disappear, it would be taken up by Code-compliant LPs with no real difference in outcomes for the client beyond they can be confident the LP is not playing games with last look, hold times and pre-hedging.
It is great that Euronext has published this analysis, just because I may not agree with the interpretation of results doesn’t mean there is anything wrong with it, it’s just I prefer the idea of a “cleaner” environment and a case-by-case framework muddies the waters. The fact that 25% of the time non-Code makers appear better shouldn’t disguise the fact that something is still likely going on and that should be unacceptable in my view.
It is good that the platform is to start publishing more data on LP performance, because this can help something that I highlighted when news of the Euronext move was first published – the ability of customers to do some of the work.
A platform can only do so much, on that I agree, which means customers have to monitor their LPs’ behavioural performance as well. I do not necessarily agree this makes the argument for a case-by-case framework, however, surely it is better to kick off the reputational risk and allow customers what would be the easier task of ensuring their Code-compliant LPs remain so?
It may be that 25% of the time non-Code players are performing (slightly) better, but in the big picture is this a good thing for the industry? After all, if you are a major LP in the FX market, I think it’s a reasonable question to ask “why have you not signed up to the FX Global Code?’, because rightly or wrongly, the optics are that you are doing something frowned upon to sustain your business.
On all currencies, taker sessions trading with Code makers display an average realised spread of 0.70 bps, the paper finds, meaning takers trade at an execution price +0.70 bps higher than the mid-price. When trading with non-Code makers, the average realised spread is 0.69 bps. “This shows that on average, when considering all currencies, we do not observe statistically significant differences between the realised spreads of Code makers and non-Code makers,” the paper authors write.
On information leakage, the paper measures markouts after 10 trades. The authors observe, again, that they do not observe statistically significant differences, with non-Code maker trades seeing a 0.43bp markout on filled trades, compared to 0.45 from Code compliers. There are some interesting divergences within this, however, for non-Code makers clearly have something going on in the, for wont of a better phrase, non-G3 G7 currencies.
In EUR/USD, non-Code maker markouts are 0.17bps, (compared to 0.11bps by Code compliers), and in USD/JPY they are 0.23bps (0.16bps). In G7, however, as a whole, non-Code maker markouts are 0.23bps, while compliers have a 0.37bps impact. In emerging markets, non-Code markouts are 0.71bps, while Code compliers are almost half that at 0.37bps.
Markouts on rejects offer the reverse in EUR/USD (non-Code 0.13bps, Code 0.24bps) and USD/JPY (0.18 and 0.26bps), while G7 (0.31 versus 0.32) and EM (0.61 versus 0.62) are close in impact terms.
Where the research does find a difference in behaviour is on reject rates, it observes that Code makers have a collective reject rate of 23% on the platform (which is, in itself, fairly high), but that non-Code makers have a 34% reject rate – the same pattern is exhibited across all currency pairs and groups. On just 14% of sessions, do non-Code makers have a lower reject rate, representing 8% of turnover.
Additionally, the research highlights how non-Code makers have a 10bps higher mid-price variation on rejects, although it should be noted that Euronext finds in 25% of sessions, non-Code makers have a lower reject cost, this represents around 27% of turnover.
To measure slippage, the research looks at the realised spread and mid-price variation before a reject and a fill, and finds that trading with non-Code makers brings higher expected slippage, by 0.11bps. Again though, the authors are keen to stress that in 25% of sessions, for 21% of turnover, these makers actually have lower expected slippage.
As noted by Euronext FX head of sales, Americas, Clinton Norton, in this video interview, this research has formed the basis for the exchange group’s consideration of changing how it operates its FX platforms and the proposed reform to default all liquidity pools to Code-only LPs – takers have, however, the right to opt out and retain their existing liquidity pools.
“We believe this new direction is in line with the preferences of our clients,” the research concludes. “Indeed, we are happy to report that more than 80% of Euronext FX’s September spot volume was transacted with Code-signatory LPs. Our goal is to provide our clients with informed choices, based on the enhanced analytical data Euronext FX provides, while promoting adherence to the best practices and principles endorsed by the FX Global Code.”