The Last Look…
Posted by Colin Lambert. Last updated: June 18, 2024
Let us return, friends, to an old favourite – the 4pm Fix – for this week there have been two stories that have highlighted the need for change, or at least a broader debate of the problem, for that is what it is.
Firstly, I was alerted to problems in the USD/JPY month-end Fix, where dealers told me their execution did not reflect the print published by WM. A quick check with sources found, as noted in our regular analysis of the month-end Fix, that there was 2.2bp difference between the Siren FX five-minute calculation, provided by mid-rates from New Change FX, and the WM print, provided by trades on EBS Market and, perhaps, Currenex and LSEG FX. I should note that my sources suggested the gap was between 1.5 and 2bp, but a gap there was.
This came one month after a similar problem in the NZD, and ironically in reporting that issue, I wrote, “In the Kiwi, this may not matter too much, but if it happens in EUR/USD or USD/JPY, or at any Fix where billions are traded, then it becomes a real worry.”
Well, it happened just a month later, and could provide, dealers tell me, a “gap” of more than $200 per million on the USD/JPY fix. We don’t know how much volume goes through the Fix or its pairs, but I think an educated guess puts it well into the billions – even on just $5 billion it’s more than a $100 million gap. Lest we forget, this is the rate that is used by these funds to calculate and move peoples’ pensions and investments around – it is not the funds’ money!
Those dealers trading at the Fix for their clients are having a really difficult time matching their execution to the WM print – and often it is to the dealer’s benefit, which means they have to go through the onerous and time-consuming process of explaining things. It’s not difficult to explain, the near-impossibility of mirroring the WM methodology solves that problem, it’s just not a good look and means more work satisfying compliance.
I have noted before that in the modern FX market, especially in a window in which so much volume is going through, one print per second is simply not granular enough – it’s prehistoric in FX market structure terms. Whatever fix is being used, the methodology needs to better reflect the actual market – and let’s not kid ourselves that these are reference rates, funds are using them as a trading benchmark – and that means more data, preferably on a five, or maybe 25-millisecond basis is needed.
This brings me to the second story we have published this week, the latest paper from Roel Oomen and Johannes Muhle-Karbe that looks at the three fixing methodologies. The paper takes an academic view of the issue and doesn’t question the amount of data points involved, but I did find it notable that the outcome of the research was that it is the length of the window that really matters, not the methodology.
I suspect this reflects the sheer size of the volume going through and, importantly, the number of traders keying on the pre-Fix window for signals. For those that weight their execution, any Fix that does so towards the back-end of the window might pick up a few performance points by benefiting from the profit taking by the speculators, but clearly, as the paper demonstrates, it is not significant.
While accepting that the paper is an academic exercise – and I applaud the authors for bringing a series of proper empirical analyses to the broader Fix issue, it is badly needed – there is one finding that I think less relevant in the real market.
As things stand, the busiest period of any month in FX, one in which billions of dollars’ worth of trades are benchmarked, is going through a flawed and out-dated process, and to make things worse it is doing so too quickly
Whichever way we cut it, especially at month-end, 4pm fix trades are rarely, if ever, small. There may be a lot of smaller trades in the mix, but thanks to the netting processes by the dealers, and the use of venues such as EBS eFix, they are swallowed up. What actually gets to the execution stage is the net, and it is nearly always large compared to the FX market’s already impressive capacity.
To me, and I should stress this is my view (and I hope I am not guilty of confirmation bias here), the paper’s analysis backs my long-term argument that a longer window is better. It notes that the Siren FX methodology is optimised for large trade execution because the longer window reduces the impact. I am no mathematician, as we all know, but I do know my way around a market, and this is intuitive – why, therefore, are we still insisting a shorter window is fit for purpose? Passive assets under management continue to grow which, again intuitively, means higher demand for a benchmark FX rate, so why are we not realising as an industry that that benchmark needs to be calculated over a longer time horizon to better reflect the actual execution?
I am not sure what real risks the industry is running – legal obviously springs to mind due to recent history – but surely whatever they are, they are unacceptable? As things stand, the busiest period of any month in FX, one in which billions of dollars’ worth of trades are benchmarked, is going through a flawed and out-dated process, and to make things worse it is doing so too quickly.
It has been well-established that Albert Einstein never uttered the words, “the definition of insanity over and over again and expecting different results”, but whoever did come up with them probably didn’t have an FX benchmark Fix in the 21st Century in mind. It accurately reflects what’s going on though, the only question is, as I asked in a post last week, “when will the madness end?”