What is the FX Market Going to Do About Last Look?

The last week has provided no little insight into how the FX market views the still tricky subject of last look. The Global Foreign Exchange Committee started the ball rolling with a draft guidance paper on the issue – it is seeking feedback from market participants up until the end of May – and then XTX Markets published its own paper on the practice, seeking to debunk some “myths” around last look.

Together the two papers provide some insight into what have become entrenched positions on the GFXC over last look in recent years. On one hand the sense is the committee generally thinks the subject is closed and only needs a few tweaks, XTX on the other hand, clearly feels more needs to be done.

Something that is, to an extent, thwarting progress in this matter, is the GFXC’s understandable desire not to become mired in a re-write of the Code. If such a thing were to take place, there seems little doubt that Principle 17 would be the focal point, however, as noted, such an occurrence is off the table, at least for now.

It would be helpful if a consensus could be reached, that more specific guidance could be provided in the form of an annex. The latest guidance paper from the GFXC is a nod to this approach, however the sense reading the document is that it is an opportunity missed.

Aspects of the guidance paper without doubt represent progress on the issue – not least the recommended use of more uniform disclosures, however overall it is hard to avoid the observation that vested interests have diluted its impact to that of a twig rather than, if not a club, the required cane.

High level principles remain the ideal framework for guiding behaviour, but it should be acknowledged – and is not in some quarters – that in areas of contention such as this, more specific guidance is necessary. It was disappointing to read in the guidance paper, therefore, that it was not intended to be prescriptive. It could be argued this is exactly what the FX industry needs on the issue of last look.

Of course, the GFXC paper is requesting feedback on its proposed draft, personally I think it would send a strong message if firms – especially on the buy side – provided that feedback, namely that they want stronger guidance around what is, and is not, acceptable.

Perhaps the missed opportunity of the guidance paper is not so much that it doesn’t prescribe conduct around last look, but that it is not stronger in having LPs empirically disclose and explain their activities. It is absolutely correct that high level principles are provided, and that LPs should, as the paper states, “strive for fairness and predictability”.

It would send a strong message if firms – especially on the buy side – provided feedback that they want stronger guidance around what is, and is not, acceptable.

Equally, the statement that “LPs should aim to minimise this period of uncertainty for the LC and therefore LPs should promptly make their decision to accept or reject a trade,” is undoubtedly right, but could we not provide some parameters around what is, or is not, an acceptable time horizon? Again, this could be part of peoples’ feedback.

The guidance paper does say that Disclosures should include the expected length of the last look window, but if an LP can state that the window is between 10 and 200 (or more) milliseconds, then what real use is that to a client? It also stresses that LPs should inform clients if they use last look asymmetrically – again though, the sense is that, if desired, the LP could hide behind a very generic disclosure.

Issues

The XTX paper, while seeking to debunk some “myths”, also raises some interesting questions and the key one is, should last look be used for commercial purposes? Although it really is a question of language, in any guidance document language is important, and the Code is perhaps not clear enough. Specifically, at no time does it attempt to deal with additional hold times or latency buffering and these, it could be argued, have become the big issues around last look rather than the use of information in the window (which is now well covered in the Code, but, according to the disclosure cited in the XTX paper from a non-bank market maker, not taken seriously by all players).

At the heart of the problem is the foreign exchange market’s obsession with spread.

It is hard not to agree with the sentiment of the XTX paper, although it also has to be acknowledged that the firm has an interest in abolishing latency buffers as it would, in all probability, improve its performance as an LP and, aside from a small group, widen the gap to other LPs. Equally, it acknowledges the existence and practicality of last look, (charts in the paper highlight how XTX can have a 10ms round trip time with some clients, but has zero hold time before checks are complete), but argues, convincingly in this writer’s opinion, for increased detail in how it is used.

At the heart of the problem is the foreign exchange market’s obsession with spread. Liquidity is often measured by this parameter, which can often be ridiculous, especially with so much “liquidity” subject to last look. Across the two papers, there is an unwritten consensus that clients need to take some responsibility for how their flow is handled.

Buy side firms have previously be quick to criticise the use of last look, but as the XTX analysis shows, that doesn’t stop them dealing with the LP with the tightest price and highest reject rates. The GFXC paper also argues that clients should monitor this issue, and the feeling is that the only driving force capable of effecting real change is the client base (or a lawsuit).

To a degree, some firms’ best execution policies, especially their lack of sophistication, work against clients measuring cost of rejects – too many of these policies revolve around hitting best bid and offer, and if the trade is rejected, they just hit the next. If, however, you are on the buy side and don’t like being last looked, then this is your opportunity to have your voice heard – you can’t complain if you don’t raise an objection when prompted.

What both papers call for, XTX’s more stridently, is for clients to analyse the cost of rejects when assessing their LPs. It is hard to find a client (especially in the oversight department) who accepts the premise, but tighter is not always better.

The other issue dominating this debate is LPs’ use of latency buffers as a protection against toxic flow. This also speaks to the obsession with spread, only this time on the part of the LPs, many of whom seem to think (because their clients often tell them to be fair) that all those clients want is a tight spread. There is perhaps, an irony in a firm using such advanced techniques as XTX does, supporting one of the oldest maxims of the FX market, but it remains true to this day as it always has been: the best way to deal with difficult counterparties is to widen the spread. Yes, the ego takes a hit, and you will probably hear from the client, but is it that hard?

Sadly the ego also plays a role in this, because too many Alphas don’t want to be accused of being anything other than top dog. Being told by a client (from whom you are probably not making any money incidentally) that others are better just doesn’t fit the mental make up of too many in this industry. Rather than tell the client it is actually their execution style, they prefer to come up with an obscure and opaque way to defend against it. All very well, one might say, but what happens when the practice becomes too widespread? Last time it was a $300 million fine or two.

One final issue is how this conduct can be viewed. With apologies for going “old school” here, imagine this scenario in a voice-dominated world.

LP to voice broker 1: I am 50-55

[Broker quotes what is effectively a streaming price until the LP says “off”]

Broker: Yours at 50

LP: Hang on a sec

LP to Broker 2: What’s the market?

Broker 2: No price

LP to Broker 3: What’s the market?

Broker 3: 48-53

LP to Broker 4: What’s the market?

Broker 4: Last given at 49, it’s 46-51.

LP to Broker 1: Sorry, nothing done.

In the voice era this would (rightly) have been seen as scandalous behaviour and what was a truly self-regulated market would have dealt with the issue and ensured it didn’t happen again. Why is an artificial hold time not seen as scandalous behaviour now?

Pathways

The GFXC paper represents a start on the journey towards more transparency and granularity in how last look is both used and notified to clients. The paper itself, acknowledges that as each LP has its own last look policy, “without clear explanation the nature of its last look process may be difficult for other market participants to understand”.

The GFXC’s paper hints that there is an understanding that some LPs are still not acting fairly, when it states, “Unpredictable or unusually long last look windows may indicate inappropriate use of last look.”

This is a statement which points to a lack of transparency of action, something the FX Global Code is designed to eradicate. Equally, the GFXC’s paper hints that there is an understanding that some LPs are still not acting fairly, when it states, “Unpredictable or unusually long last look windows may indicate inappropriate use of last look.”

If LPs were behaving themselves and totally transparent in how they use last look, why would this statement be necessary?

Ultimately, the gap between the two papers can be highlighted in one area. XTX’s paper on one hand, calls for the use and sharing of data to manage last look’s use. The GFXC’s paper, in keeping with the general tone of the Code it has to be said, still relies upon “internal controls” and “management oversight” to supervise last look’s use.

The way forward is likely to be dictated by the GFXC rather than an individual firm – even if that firm does have support elsewhere in the industry. This means that the FX industry will continue to exist on a knife edge and in the hope that generic disclosures will be adequate defence if a client decides to take the legal route (and there are plenty of law firms likely to be keen to take on the banks again especially).

It is hard to argue against the use and sharing of data to explain how last look is being used – equally it is hard to understand why some LPs take so long to start their risk checks. After all, if the client really is at the centre of everything they do – as they insist on telling us every available opportunity – then what they are effectively saying by having a hold time of, say 200ms, is that they don’t trust the client. That or they want to ensure they make money out of every trade – and if it’s that then what does that say about the relationship?

Last look will, therefore, continue to be a thorn in the side of the FX industry, unless the GFXC can further strengthen up its guidance. That, thanks to the schedule and the likelihood that it will be another three years until the Code is reviewed, is very unlikely, which is a shame, because the FX industry could do with a more stringent approach to last look.