Has the FX Market Finally Solved Last Look?
Posted by Colin Lambert. Last updated: August 20, 2021
The FX Global Code, as its founders were keen to stress, is a living, breathing document and the most recent three-year review of the Code exemplifies the Global Foreign Exchange Committee’s approach. Last look has been something of a perennial in reviews of the Code such is the level of controversy around the practice, and repeated attempts to clarify matters have had, to be kind, mixed results. So, has the GFXC finally “solved” last look with the latest guidance paper?
In reality it is hard to say because of the different interpretations that will inevitably be used by different market participants. On one hand, by reiterating that last look is only to be used for price and validity checks, the guidance hints at a distaste for using the practice for commercial purposes, but it stops short of actually saying so – hence there is a fear over what “clever” legal minds will make of it. Is adding a latency buffer a risk and validity check? If I am concerned about the risk of losing money from a trade, then is that sufficient reason to have that additional hold time?
Reasonable people would probably respond with an unequivocal ‘no’, but last look does strange things to people – not least those liquidity providers who rely so much on the churn method, have under-invested in their pricing and risk technology, and seek to hold as little risk as possible for the shortest time possible. Minimising losing trades is crucial to some LPs business models, hence longer, and asymmetric, hold times – are these participants expressly going to change their model? It’s doubtful because they probably see a loophole in the guidance, or if they don’t, they will find one.
The guidance paper makes a point of noting that Principle 17 of the Code “explicitly disallows” the use of last look for purposes of information gathering and characterises unacceptable behaviour as potentially including high reject rates and, perhaps significantly, “differences in the time take to accept or reject and/or significant market movement against the LC when a request-to-trade is rejected.”
Although there is no direct link, this guidance would appear to suggest that asymmetric round trip times potentially mask misconduct on the part of an LP. Whilst there are inevitably small differences in round trip times due to an LP’s internal last look processes, this calls into question large differences in accept/reject times as highlighted by The Full FX in a recent article.
The crux of the GFXC guidance paper consists of three recommendations aimed at reinforcing the guidance on Principle 17 and ensuring a fair and effective last look process, the enhancement of ex-ante disclosures, and the information available to enable participants to evaluate how an LP handles trade requests.
The first recommendation highlights the need for LPs to act in a fair and predictable manner in designing their last look processes. It reiterates that LPs should minimise the period of uncertainty for the LC while the trade request is being checked, and stresses how “LPs should apply the price and validity check without delay and promptly make their decision to accept or reject a trade.” (underlining in the original)
More importantly, perhaps, the guidance adds, “Incorporating any delay that is additional to what is required to complete the price and validity check is considered contrary to Principle 17.”
Again, the reasonable assumption is that this statement rules out the use of additional hold times or latency buffering, but, once again, how it is read by certain LPs, or more importantly, their legal teams, will determine its effectiveness.
The latest guidance reinforces the message over the inappropriate use of information as well as the use of the word “clients” actually refers to all counterparties, disclosed or anonymous. It also recommends that LPs ensure there is sufficient internal controls and management oversight in place around last look procedures. The GFXC suggests such a move would “provide LCs with confidence that they are being treated in a manner which is consistent with the Code’s principles”, however it should also be noted that the same managerial oversight was in existence when asymmetric hold times were extreme. If the participant firm does not buy into the argument that asymmetricity and latency buffering are unfair, management oversight will change nothing.
The second recommendation calls for greater use of, and more detailed, disclosures. For critics of the Code’s approach to last look, there will be concerns that one suggested disclosure is whether the LP uses asymmetric hold times – given how this is seen as evidence of potential misconduct, the very act of allowing it to be disclosed gives it, in some eyes, a legitimacy it doesn’t deserve.
That said, the guidance document does state that the choice of price check logic could materially impact the predictability of the LP’s last look processes and therefore the outcomes for the client. “Asymmetric price checks introduce greater complexity for LCs in monitoring the effectiveness of their execution and so it is particularly important for LPs who use asymmetric price check to disclose the background, and for LCs to understand the circumstances in which they are used,” the paper states.
For those worried about additional hold times, there is good news in that the disclosures are now recommended to include minimum and maximum last look times, however again, there is room for an LP to disclose reasons why they may differ. It may have been more helpful if the paper expanded this by advising that if a reason is anything other than technical problems, the LC would be wise to investigate further.
Something that may be music to the ears of the budding trade analysis industry can be found in the third recommendation, which calls for information to be made more available on reasons for trade requests, especially around rejects. It says that LPs should be accurately recording trade rejection information for orders and transactions and be able to disclose it at a trade‐by‐trade level at the time of the rejection where requested by LCs.
“With appropriate transparency from LPs, LCs should be able to determine whether their methods of execution continue to meet their needs over time, including whether to trade with LPs that are using last look,” the paper states. “A trade rejection due to a price check should be an indication that the market has moved between the time the quote was provided by the LP and the time the LP received the request to trade.”
The paper adds that while it is difficult to definitively attribute the precise reason for rejects occurring, some potential areas for concern include, but are not limited to, longer reject times than in the disclosure document; high reject rates, combined with long and asymmetric response times; unpredictable responses; and market impact.
The paper also notes, in the interests of fairness, the LC behaviour could be a part of any problem, specifically, if an LC is seeing high reject rates or market impact of their own trades, they may want to re-evaluate their execution strategy.
A New World?
Will, therefore, this guidance paper change matters? Will behaviour around last look improve? Sadly, the answer is probably not, because there is simply too much ambiguity, which some would say is inevitable from a committee with such diverse interests.
for all the hard work that has gone into this work, driven by demands from participants for greater clarity, the reality is for those firms whose activities drove the debate on last look in the first place, little has changed
That said, a straightforward reading of the paper would lead reasonable people to believe that latency buffering and excessive asymmetricity is frowned upon by the FX Global Code. The fact is, it is, but the paper stops short of adding the additional guidance desired by some that would unequivocally rule out such practices.
Without such strong guidance, it is inevitable that some LPs will continue to routinely hold customer orders for an inordinately longer time before rejecting than accepting, and, more importantly, will add latency to allow them to quote a tighter spread for potentially toxic flow.
If this occurs the landscape is slightly skewed, especially if customer continue to be obsessed by spread. And it is there that perhaps the best course of action lies – with the customers. There is no doubt that the GFXC and Code are strongly encouraging, some would say demanding, that the buy side accepts some responsibility for what goes on – if they don’t analyse, assess and complain then it is hard to have sympathy for them if they suddenly have a large loss due to a reject.
Independent players are springing up to provide this level of detail around the cost of rejects and LP behaviour in general, and perhaps a future iteration of the Code could strengthen guidance around the use of analytics by LCs to better assess their and their providers’ performance.
Until then, however, the sense is that the industry has moved a fraction towards a stronger stance on last look, but this in in inference rather than action. This means that for all the hard work that has gone into this work, driven by demands from participants for greater clarity, the reality is for those firms whose activities drove the debate on last look in the first place, little has changed.