I know I have floated ideas around this issue before, but do we need to do more about that hour after the New York close than just talk about it? There was quite a lot of chatter around a so-called flash crash in the Swiss franc this week, when EUR/CHF jumped just over 100 points (as did USD/CHF) and although my sources tell me it was a fat finger error, it comes on the back of a bigger event just the month before. These events are starting to occur a little too frequently for some peoples’ liking, so what can we do about it? Actually quite a lot I think – or at least it will be a lot if all the noise around data capabilities isn’t just that – noise.
Someone claiming to have knowledge of the Swiss matter tells me that a price was entered in EUR/CHF with the wrong big figure and whereas the amount involved would normally have seen the order exhausted within a few points, because it was in the “witching hour” there was no liquidity to fill it, hence at least part of the incorrect bid appeared in the market and everyone reacted.
Other sources are guessing – and this often happens when we get a move like this – that someone was going after stop losses at 1.01 in USD/CHF, however I tend to feel that chatter is written by the numbers – the story is fitted to the price action.
Either way, we had yet another liquidity event in this window, ironically just after the Reserve Bank of Australia issued a brief report on the January 3 incident, which of course happened around the same time of day.
The RBA report offered little new in the way of analysis, but reading it only reinforces my original belief from that day – a confluence of events came together to create that move. The potential issue for the FX market is that I can see these confluences happening more and more.
My reading of the events of January 3 was that the Apple announcement triggered risk off; in that scenario high yielders are sold and yen is bought; and Japanese retail is pretty much always short yen thanks to their thirst for the carry trade – which itself pretty much outdates e-FX, so is unlikely to ever go away (unless something truly amazing happens and the BoJ gets rates to 3-4%).
The issue with the Swiss move this week, and the yen move of January 3, is that amounts that would barely be noticed for 21 or 22 hours of the day have had an out-sized impact because of when they were transacted. I don’t think we can move the planet and rearrange the clocks so the issue isn’t going to go away, so perhaps we need to do something about it beyond what was another worthy, but ultimately vanilla attempt by a central bank to find out what went on?
I think we need some forensic analysis of the FX market during one or more of these events. I have yet to be convinced that the real story in the sterling flash crash of 2016 wasn’t what happened beforethe crash, when the market moved down to 1.2600, and I suspect in-depth analysis of the January 3 event would unveil some interesting findings as well.
Every time I raise this issue with a variety of people I am told it’s “too hard”. Those who might conduct the analysis point to the fragmented and relatively opaque nature of the FX market – and platforms say pulling the is too difficult and that it wouldn’t be worth anything on its own. I know it won’t be easy, but will it really be that difficult? After all, I am besieged with platforms telling me how great their data is and how much easier it is to price and trade markets with this data – why not put it to a different use?
Equally, for all the chatter about fragmentation, the FX market revolves around three primary venues – CME Globex, EBS Market and Matching. Of course there is significant ECN volume executed over CboeFX, Currenex and Fastmatch, but as a recent report from the BIS noted, price discovery is still the domain of the primary venues. What does this mean? Well in simple terms it means that a lot of so-called liquidity providers recycle the price from one or more of the primary venues onto others – most of the liquidity in the public market starts with a bid or offer on one of those three venues.
I fully accept that some market makers price to multiple venues and fill the first hit and cancel the others, but the important fact to point out here is that none of the primary venues operate last look – these LPs can’t reject trades on these platforms, they can only cancel before being hit (and to highlight how complex this issue is, many LPs are having problems being swept and can do little about it, as I wrote about on Monday). Equally, however, any forensic analysis can include the secondary venues – everyone now time stamps to the nth degree so it should be possible to see where the action started.
There is a challenge in deciphering how far back to go, and absolutely the weight of data involved will be immense, but the fact is we knowwhen the sterling, yen and even the relatively minor swissy moves started. The RBA report this week noted how in the hour or so after the Apple news markets were orderly. Another common refrain of today’s markets is that things happen much quicker than they did before – it may be simplistic to say this but, good, it means we only need to analyse a minute or so before the actual move! We don’t need to look at most of the time when things were orderly, only from just before it started to unravel.
Given the data sets available, this would be a huge task – it means reconstructing a window within the foreign exchange markets, but we are actually only talking about something like five minutes (if we need to go further back we can of course) and, given the relative lack of activity, and relatively small amount of trades, it wouldn’t be as hard as it could be. Building an accurate picture of that five minute window, using data from a host of platforms allows the investigator to take advantage of yet another recent improvement that we shout from the rooftops of the FX market – it would enable conversations.
Let’s assume a sell event, an ID may or IDs may be prevalent during the event on, for example, the aggressive side of the price. It may be that a customer has gone out and hit multiple LPs direct at the same time, if so, those LPs are quick every other time of day to talk to that customer about their execution style, so you can’t tell me they wouldn’t know what has gone on. The investigating body can ask the question of the LPs “who sold to you?” and get a pretty limited number of answers.
Equally, the platform data and time stamps may reveal persistent sellers either on the primary venue or across multiple venues – if so, ask the question of that seller, “why?” Because here is the crux of the issue – these markets do move that far without someone triggering it – in my above example, someone has to start the selling off and weith enough hard work it can be found who. By asking these questions that arise from the data, we can talk to those who, wilfully or not, kicked the whole thing off.
It could be something totally innocent – as I am sure the Swiss event of this week was – people make mistakes and, hopefully, they learn from them. Just as likely, however, is that someone is gunning for stops, triggers or margin calls – if that is the case then they can be dealt with appropriately and by doing so a powerful lesson could be sent to other would-be miscreants.
More importantly, this could be a structural issue that has grown over the past couple of years as markets have become more inter-locked across asset classes. It could be a totally innocent event in that a bunch of investors are locked into risk on-risk off strategies, or are following risk neutral strategies, and they need to execute a trade at the end of the New York day when a lot of portfolio revaluations take place. If FX is now part of that execution – and after all we have risk on and off currencies – then it could be helpful to these investors if their poor timing can be explained to them.
If nothing else, such a discovery could see them re-think their strategies, or see liquidity providers or central banks take appropriate measures to help out. As far as the former is concerned if they know there is going to be decent business going through that is (in timing terms) pretty unsophisticated, they may want to increase the resources they allocate to that time? At the moment I don’t see why someone would because they are getting free money by letting the event happen and then taking advantage of the reversion after the exit trades are done, but going forward?
If nothing else, a forensic examination of what happens during specific events would allow intelligent people to draw intelligent conclusions and, into the bargain, it should dampen the temptation in someone having a bad month to run some stops at an illiquid time of day. It may be, that like circuit breakers on exchanges, various authorities seek to encourage a shut down in the market for one hour? I don’t like that idea because trading halts worry me if something fundamental happens to push a currency pair, but again, it would be a decision brought about by considered analysis.
The power and enlightenment brought about by the data revolution is rammed down out throats almost daily – why then, can’t we use all this power to truly enlighten ourselves to reinforce one of the weakest links in the FX chain? This issue should no longer sit in the “too hard” tray.