The Last Look…
Posted by Colin Lambert. Last updated: September 21, 2021
We are going to learn a lot about how seriously FX market participants take the FX Global Code over the coming year.
There has been a buzz around the FX industry following a comment from Global Foreign Exchange Committee chair Guy Debelle last week, who noted at a conference that additional hold time, especially when it is just to see the direction of the market, is contrary to the Code.
I ought to say first up that I am not sure this is a huge shift in policy, take the following statement, which was in the GFXC’s guidance paper on last look (with my italics for emphasis):
“Where used, the last look process is intended to be used for the price and validity check only, and for no other purpose…In the interest of fairness, LPs should aim to minimise this period of uncertainty for the LC and therefore LPs should apply the price and validity check without delay and promptly make their decision to accept or reject a trade. Incorporating any delay that is additional to what is required to complete the price and validity check is considered contrary to Principle 17.”
I read the guidance paper as saying latency buffering was in breach of the Code back then (I feel it does the same about significant asymmetric round-trip times by the way), others, as I noted in a column at the time, probably did not share that view. They can be in no doubt now.
The problem is, as I noted after the paper was published – and this is an issue for any best practice document – will the guidance, even reinforced by Guy Debelle’s comments, lead to adherence, or will “clever” legal minds effectively endorse business as usual via a series of loopholes?
We should not be surprised, given its historic scene as a battlefield between vested FX interests, that in last look, the FX Global Code may face one of the biggest challenges to its credibility, and last week’s comments may have, perhaps inadvertently, upped the ante (although I still tend to see them as being a reiteration of what the guidance paper said).
To continue to operate with additional hold times or latency buffering after the GFXC’s guidance and a reiteration/clarification by the committee chair, will involve a systematic and deliberate decision to ignore the Code.
We have seen how in a court of law, a bank’s legal team have apparently referred to the Code as “guidance” as part of a defence of their actions (in this case on SNB-Day). That is, no matter what your thoughts on the tactic, effectively a discussion about a one-off circumstance in extreme conditions.
To continue to operate with additional hold times or latency buffering after the GFXC’s guidance and a reiteration/clarification by the committee chair, will involve a systematic and deliberate decision to ignore the Code. That is much more serious in terms of its credibility. There is without doubt, serious dissent on the GFXC over last look, but how will it look if some of the organisations backing the Code, and who are on the committee, decide to ignore one critical aspect? That would not be a good look for the Code and it most certainly would not be a good look for the market participant(s) concerned.
As a matter of interest, following the publication of the guidance paper on last look I downloaded as many LP disclosures as I could find (and some are easier to find than others). I don’t expect there to be changes this quickly, the legal professional can move at a glacial pace, but over time, if these disclosures are not changed to reflect no latency buffering then one has to question the LP’s commitment to the Code – in spite of what will no doubt be profuse public statements to the contrary. This is public information, so we will know, in the course of time, which LPs have decided that on last look, they stand with the FX Global Code and those who stand against (and I don’t think the concept of proportionality has any place here).
As I have noted in a separate article, the GFXC has probably gone as far as it can in providing best practice for last look – the fact is that apart from certain circumstances, no customer should be held for longer than one primary market update. This is 5ms at headline level, but for those LPs and markets where the update is 25ms, it sits there. Everyone knows these parameters so we should have no more 100ms+ round trip times unless the technology has gone wrong.
If the GFXC has gone as far as it can then who will police this? It has to be the customers with the help of the trading venues. Banks operating proprietary platforms should have last look checks hard-coded in, and provide reports to clients of any deviation away from best practice times. The fact is on the SDPs this really isn’t that much of an issue, but in the multi-dealer world? There, a lot more can be done.
Platforms should be providing customers with average response times as well as, I would argue, a cost of rejects analysis. A shout out to CboeFX, which I know provides these data, and that surely is the way forward and other platforms should follow. If a customer is given the information to help them make a rational decision, then as I argued the other week, if they choose to carry on and get ripped off, whose fault is that?
If it starts costing serious money of course, then customers will probably wake up. Last look has been around in volatile markets – and anecdotally what I would consider “good” customers tell me reject rates went to 20%+ last February and March – but not in sustained volatility. A lot of customers say they’re not bothered by last look, fair enough, but they haven’t really experienced it in extended volatile conditions – and I’m not talking crazy vol here, just what a certain generation would term “busy days” where the market just keeps on moving.
If customers take some responsibility – assuming the platforms provide them with the required level of data – then bad actors can be ring-fenced and either forced to change or exit the market.
In those conditions – and watch out for interest rate divergence coming to a market near you soon – the cost of rejects soon adds up. Not much can be done about rejects in the 5/25ms range, but a lot can be done about those holding longer – if the clients have the courage and foresight to actually do something.
If customers take some responsibility – assuming the platforms provide them with the required level of data – then bad actors can be ring-fenced and either forced to change or exit the market. It worked with latency arbitrageurs back in the day, it can work with those who abuse last look. Conversely, when dealing with those clients whose flow is “tricky” or “toxic” (and there are very different definitions of that around) the solution remains what it always has been – quote them wider, not at all or on a venue that means you, the LP, only deal when it suits (i.e. a firm CLOB where you post interest).
There will still be rejects, I reluctantly have to accept that, but they will be the result of a customer actually sniping an LP and being caught out, rather than an LP being given extra free insurance by the client for what is effectively an indicative quote dressed up as “dealable”.
There is an irony in how many LPs, who extend the hold time, run to the firm CLOBs when things get busy. They do this because they know it’s the best environment in which to get business done in busy markets – it’s a shame they won’t extend the same benefit to their clients.
Hopefully Guy Debelle’s comments last week have clarified the thinking at some LPs and prompted a few meetings (and a decision to change to no latency buffering). A game-changer? Probably not. A clear and unmistakable message? Most certainly. More importantly, LPs, platforms and customers are on notice – you know what should be happening – make sure it does.