The Last Look…
Posted by Colin Lambert. Last updated: September 12, 2023
Last week’s column, not unexpectedly I suppose, elicited quite a bit of feedback and comment and it was interesting that some correspondents thought the data might have been an outlier, so egregious were some of the examples. I bring bad news for those people – but into the bargain have bad news for myself.
Whilst some of the numbers cited last week were the worst I have seen, it was not so much the actual time taken that bothered me (unlike many of my correspondents), but the inequity of the whole thing. Aside from the old “we’re giving the client trade more time to get back in court” excuse, I just don’t see why an LP would be multiple times longer to reject a trade than accept it. I hear that excuse quite regularly, and while I haven’t asked all who use it the question, those I have asked, “so what percentage of trades are actually completed thanks to the longer hold time?” are beyond reluctant to provide an answer!
I do believe that several LPs use the additional hold time on particularly toxic flow, but given I quoted a monthly average across all flow, it doesn’t say much for the quality of counterparties!
As noted, though, my big problem is asymmetry – and while the odd millisecond here and there can be glossed over, reluctantly, the sheer scale of the additional hold time deployed meant, in my eyes at least, that either the platform has a real problem with its participants or the LPs are getting away with their wilful ignorance of the FX Global Code and its best practices.
Amongst the feedback I received was data from another venue which highlighted the problem is not unique to one channel. I would note that the slowest response time was under 50ms, unlike some of the absolute crackers from last week, but the table of LPs was riddled with asymmetric response times.
Participants can’t happily accept one breach of best practices while complaining about another…You either adhere to these codes or you don’t
Collectively, the top 10 on this channel held for an average 3.3 times longer for a reject than an accept, the worst case being 16 times longer, the second worst 12 times longer. No LP in the top 15 had a symmetric response time, although in fairness to a couple of providers they were within a couple of milliseconds.
So for those who think this is something that happens in isolation, the argument against is building, however there is also bad news in this data for me, and my arguments.
You have to get to LP number 14 before you have an LP within 2ms asymmetry and numbers 19 and 20 before you get to anyone operating symmetrically – and, incidentally, both are at zero response time.
This tells me one thing – the consumers don’t care, because the use of longer reject hold times is helping some LPs quote tighter. As such, my chagrin at such behaviour is meaningless because the clients are clearly unconcerned about this practice. That, of course, is their right, but it needs to be made clear that they can’t happily accept one breach of best practices while complaining about another. You either adhere to these codes or you don’t – and before people throw proportionality at me; yes, that is a crucial concept of the Code, but symmetric response times and handling requests without additional hold time are core tenets of an LPs adherence, and therefore absolutely proportional to their Statement of Commitments.
What bothers me most about the latest table, and to a degree the first, is that LPs who are trying to do the right thing are winning less business – that is hardly an incentive to adherence, and several people have expressed their frustration to me over that. All data such as that written about here for the past two weeks does is incentivise a deterioration in behaviour – not only that, but to the lowest common denominator.
Last look was, and remains, a controversial aspect of the FX market, but maybe now for different reasons. We do need to re-examine the industry’s attitude to it, however, for behaviour seems to be getting worse, not better, when it comes to last look’s use. The relative lack of buy side signing up to the Code has been a public concern for some time now (although I would note that many of the largest asset managers are signatories), so while we question that, perhaps we also need to discuss their views on the use of last look?
Are these customers happy with asymmetric hold times or are they ignorant as to their potential cost? If, for example, they have all the data on costs of rejects to hand and decide they’re OK with asymmetry, fine. If, however, they aren’t looking at the issue, they need to be told, in the clearest terms (i.e. in writing!), what is going on to remove any legal risk at a later date.
If a broad consensus among liquidity consumers can be achieved, then perhaps the Code can be adjusted accordingly, if it cannot, then surely we need to err on the side of caution and strengthen the Code by formally including the recommendations of the 2021 guidance paper into the main document?
As a digression, there was an interesting theme emerge from last week’s feedback, something said to me by a couple of liquidity consumers who were repeating something I have heard increasingly over the past few months – the growing instances of trades being rejected, followed by the price refreshing at the same level.
I managed to ascertain that in some circumstances the reject/refresh was the same LP and my correspondents questioned the ethics of this, however I would argue this is what happens with a zero additional hold time. One of my correspondents is very much on top of this whole issue and informed me that two LPs were at zero additional response times, but were in the habit on some channels of rejected then replacing at the same level.
We should not forget how important the FX Global Code was in repairing the reputation of our industry, this is not the time to start eroding its effectiveness
Effectively, this is what happens on a no-last look CLOB – the price is there or it isn’t. If it changes or is off, the client cannot hit it – back in the voice days it was called hitting on a change, and led to a lot more arguments! In this case, the LP is probably trying to do the right thing, although the hope is they are not flooding the platform with quotes.
In my view, this is actually not a bad way of dealing with toxic flow, reject it immediately if the market has moved, but if supply is there, re-submit and accept – it’s a cleaner version of the additional hold time (and of course I still think the best way is to quote toxic flow wider or not at all!)
Overall though, there are causes for concern in how the use of last look is evolving, and this needs discussing. I am sure that other channels will have the same data as I have written about and that makes it an industry-wide issue.
We should not forget how important the FX Global Code was in repairing the reputation of our industry (and yes there is work still to do), so this is not the time to start eroding its effectiveness in such a contentious area.