On October 7, Cable flash crashed in early Asian trading, leading to chaos in the market and an official investigation into events surrounding the move. Colin Lambert takes a look at what happened.
A few minutes into October 7 UK time, at 12.07.03am to where there are grounds to believe that the transaction is be precise, Cable traded through 1.2600 having fallen 30 points in the previous minute. Just 23 seconds later it traded below 1.2200 and 45 seconds later it had traded at 1.1378 on one platform. Just two minutes later the market was trading back above 1.2100 and just 10 minutes after the initial move, Cable was trading above 1.2400. The market had “flash crashed”.
Reasons for the move remain unclear, however Bank of England governor Mark Carney has asked Guy Debelle, chair of the BIS’s FX Working Group responsible for creating the Global Code of Conduct, to investigate events.
Market speculation has focused on two main theories, one that a participant sold aggressively with the intention of hitting option barriers or stop-losses, and the other that news- reading algos picked up on a Financial Times story in which French president Francoise Hollande discusses a so-called “hard” Brexit.
If the move was indeed started by someone seeking to trigger certain levels there are questions for the Code of Conduct writers in that Principle Four of the Execution section states “Market Participants should not request FX transactions with the intent to disrupt the market and Market Participants handling Client orders may decline a transaction meant to disrupt the market.”
The Code then goes on to further state, “There are certain transactions that may be required in the course of business, for example transactions related to merger and acquisition activity, which could have a sizable impact on the market. These transactions should be appropriately monitored and executed by the relevant Market Participants.”
As noted by several sources, FX liquidity is at its thinnest between the New York close and Tokyo open and even though this move happened after most Tokyo dealers were at their desks, markets were still very thin. “If you want to have an impact that’s the time to hit the market,” observes an Asia- based trader. “It doesn’t take a lot to get things moving, although I think on this occasion nobody would have thought it would have gone so crazy.”
Although it cannot be confirmed, multiple sources suggest that option barriers under 1.2500 with large payouts attached were the target of the selling and as liquidity dried up and stop- losses were hit, further exacerbating the move, the market gapped sharply.
An interesting aspect of the move that tends to argue against the news-reading algos theory was how focused trading was on Cable. Sources say activity in the crosses was minimal and that everything went through the headline pair. If sterling had generally been the target of speculators reacting to the Hollande headline it would be fair to assume that activity levels would have spiked in EUR/GBP and other sterling crosses. This was not the case, according to dealing and broking sources.
It is unlikely that any barrier being targeted was a long way from the market when the order started so such action would appear to be only the trigger for events, and not the driver. Dealers reported increased interest from the ubiquitous Japanese retail sector below 1.2500, but again little in GBP/JPY.
One characteristic of the move was how, in spite of its speed, a lot of levels “printed”. “We saw a lot of trading all the way down to 1.2050, but then it disappeared,” says a dealing source in London. “Below 1.2000 things got really bad, we printed once at 1.1940, then saw a few deals at 1.1835-40, then nothing to 1.1500, where we saw another flurry of trades. Apparently it traded lower away, but we didn’t see that. From 1.1500 we saw a sharp bounce driven by a lot of manual buying interest between 1.1900 and 1.2150.”
As was the case on January 15, 2015 when EUR/CHF flash crashed following the Swiss National Bank’s abandonment of the currency peg, there was a great deal of confusion over that actual or “official” low.
A great deal of legal documentation in FX markets uses Thomson Reuters Matching as the platform of record for sterling trading and that platform has rules that it shares with users. Several participants on the Matching platform tell Profit & Loss that the firm set the official low at 1.1491 after one trade at 1.1378 was “re-papered” at a higher rate.
Even 1.1491 created some confusion, however, for the same sources say that only one trade executed there according to their market data, and under Thomson Reuters’ (TR) rules for Matching, an official high/low can only be set if five trades are executed at a level within a three minute window. If this is indeed the case, the low should, these sources state, be 1.1500 where “several trades were executed.
A spokesperson for TR clears the matter up, telling Profit & Loss that the firm reports the absolute low as well as an official low. Apparently in this instance the absolute low was the one trade at 1.1491; however, according to its rules, the “official” low set was 1.1500.
Notwithstanding the initial confusion, the establishment of the level by TR has helped solve some disputes that are inevitable in such circumstances. Profit & Loss understands that several clients were pressuring banks to re-paper trades below 1.1840 and that the banks were resisting. “Unlike the Swiss move when EBS and a few banks got together to apparently arbitrarily set the low at 0.85, this seems like a genuine low,” says the head of FX trading at a bank in the US. “It wasn’t pretty but there were clearly multiple trades at 1.1500 on multiple venues and that means we have a genuine low print.
“Some customers are inevitably unhappy, but would they be complaining if the market had carried on to 1.10?” the trading head asks. “More to the point we had several customers buy from us below 1.1840, should we re-paper those trades? It’s unfortunate that some people have lost money on this, but in terms of following the rules everything was done by the book. There’s no asymmetric treatment of clients, everyone gets the appropriate fill whether it’s in our favour or not.”
The fair treatment of clients certainly seems to be one of the lessons learned from the SNB event, as well as an impact of the Global Code of Conduct. “There were delays but we made sure we knew what the low was, as set by a trusted and independent third party, before confirming fills,” an Asia-based e-trading head says. “It probably took too long, but luckily it was only a few trades and most customers were happy with the outcome.”
The confusion below 1.2000 was further compounded by a lack of consistency among the multi-participant platforms over how they treat extreme market conditions. Some, such as LMAX Exchange, use volatility bands under which the platform “greys out” once thresholds are hit. According to sources familiar with the matter, LMAX, which saw 2,500 trades in one minute during the flash crash, hit its Vol Band at 1.1989 and when the market had normalised it reset, at which time Cable was trading at 1.2050.
Circuit breakers continue to be a thorny issue for the FX industry, for while they undoubtedly take some of the heat out of a flash crash situation and stop participants trading at extreme levels, there are questions as to what would happen if the market has a large, one way, move.
“FX is a continuous and fragmented market so there’s no real value in circuit breakers unless you have a uniform approach, across all venues,” says the head of FX trading at the US bank. “There is also the danger of a large directional move, in which the circuit breakers kick in and customers can’t exit risk until they are reset, with the market another 500 points away. This is a subject that needs more discussion, preferably at industry level.”
Something that does argue in favour of more circuit breakers on public platforms is their use by market makers. Market sources tell Profit & Loss that automated market makers started stepping away from the market once 1.1900 traded because of their own risk management functions. “Price engines have fail safes that cause them to stop pricing when spreads go beyond a certain threshold and/or the market moves at a certain speed,” explains an e-trading head in Asia. “I suspect a few have 51 point spreads in Cable as their threshold so they shut down temporarily while the algos were reset. It was probably only a matter of seconds, but that was enough for another 400 points.”
“Price engines have fail safes that cause them to stop pricing when spreads go beyond
a certain threshold and/or the market moves at a certain speed.”
Again though, there seems to be little uniformity when it comes to kill switches and the like among market makers, with some liquidity providers continuing throughout, albeit with very wide spreads, and others pulling out. “Unless we can get everyone on the same page events like this will always be chaotic,” the Asian e-trading head observes.
One aspect that did appear to work well was the resolution process over what is now agreed was an off-market trade at 1.1378, with both parties agreeing to a re-papering. That this was apparently done quickly and easily helped ease some of the confusion, although questions need to be asked how the trade happened in the first place.
Anecdotal evidence from the industry indicates that, with very few exceptions, most of the selling below 1.1850 was machine related, while the buying accounts were generally resting orders or manual trades. If this is indeed the case and at least one side of the trade at 1.1378 was machine- generated – most probably the selling interest – then questions need to be asked over the quality of the programming. If a price engine is behind the sale at an off- market rate it suggests that either it is badly programmed or the participant’s risk management criteria needs tweaking.
There is one other possible protective measure against flash crashes in FX, but it is not one likely to garner favour among some central bankers. Given the speed of the moves it is likely that much of the damage is done before a central bank notices; however, if the bank places bids and offers in a “band” around the current spot rate they would at least be alert to a sharp move by the ticket notifications.
This type of policy has been used before in “dirty” floats where central banks limit the distance their currency can move within a day. The Bank of Canada historically used such a policy, being prepared to enter the market if USD/CAD moved more than 75 points at times during the 1990s. If central banks used a wide band, within pre-determined time windows, they would be able to be alert to potential flash crashes and, if the banded orders are sufficiently large, even stop the flash move dead in its tracks.
“There’s no doubt the machines were spooked on [October 7], and they only really recovered when significant buying interest, from manual traders, came into the market,” says the head of FX trading at the US bank. “Historically there would have been reasonable resting order interest in the market ,which would have slowed or maybe even stopped the move, and until we get that sort of interest back in the market we will run the risk of these types of move.”
The events of October 7 have served to frame what is an ongoing debate over liquidity in foreign exchange markets. While there are still many who believe there is no problem in the level of liquidity in FX markets, others are less convinced, pointing out that the truest measure is depth of book, as well as liquidity levels when volatility spikes.
If nothing else the Cable flash crash has given lie to the naysayers who refuse to accept there are liquidity issues. Equally, notwithstanding whether the flash crash was triggered by an aggressive sell order or news-reading algos, there can be little doubt that the event provides more evidence that there are market structure issues in foreign exchange.
Partly this is due to the thinning out of voice traders following the chat room scandals and partly it is regulatory driven due to the increased capital costs of running a risk-taking foreign exchange business. The move on the part of some banks towards offering agency services at the expense of principal risk warehousing services appears to have exacerbated liquidity conditions further.
Also into the mix can be thrown changing investor attitudes, as highlighted by a note from Citi’s global head of G10 FX strategy Steven Englander on October 2. In the note, Englander highlights the small number of clients who are putting on hedges or taking positions on the US election in spite of it being a true risk event. “The absence of pre- positioning has been expressed in other trades as well,” Englander writes. “When I was in the UK a couple of weeks ago, almost all clients were bearish sterling over the medium term, but few were willing to position short sterling until they had a concrete signal that either UK economic data were softening or there was a signal that Article 50 would be triggered sooner rather than later.”
“We saw a lot of trading all the way down to 1.2050, but then it disappeared. Below
1.2000 things got really bad.”
Observing that many investors may have been stopped out before the big move on Brexit, Englander suggests that “decent money was made by being patient on Brexit” and that investors generally seem content to miss the first part of the move in favour of awaiting a stronger signal. “It is hard to argue with this approach, except that it seems to violate one of the laws of thermodynamics (the one that says you can’t get something for nothing),” Englander writes.
Looking at where the risk in this approach lies, Englander hits the nail on the head by writing, “My conjecture is that investors are assuming that it will be much easier to be the second one in the door of a soon-to-be-very-crowded theater than it would be to be the first one out of a burning one. This may not be true to the degree that they expect.
“If everyone is positioning to pull the trigger on positioning as soon as they know the outcome, the repressed volatility may emerge in a very sharp burst,” he continues. “When investors realise that this is the strategy that investors as a group is playing, the incentive is to pull the trigger early and often when the news comes out. The outcome would likely be that prices move much more quickly than expected with much less opportunity to get the position on than investors would like and some possibility that one-way markets lead to overshooting.”
Citi was not the only institution looking at liquidity issues either. In Bank of America Merrill Lynch’s Global FX Weekly, analysts noted, “Market fragility is increasing as phantom liquidity creates the illusion of stability.”
The report further observes, “While on the surface, traditional measures such as bid-ask spreads in FX have indeed narrowed in the past couple of years, our volume-based analysis shows market liquidity has materially worsened. We find the market impact of a given volume is now 60% greater than in 2014. Notably, the frequency and amplitude of outsized volatility events has also increased.”
“FX is a continuous and fragmented market so there’s no real value in circuit breakers
unless you have a uniform approach, across all venues.”
The BAML note further states, “We define phantom liquidity as the appearance of liquidity that may not be present when investors need to transact with urgency. For example, during Brexit we saw record levels of bid-ask spreads and volatility, despite only somewhat elevated volumes. This suggests massive slippage over periods of volatility and gaps in the price action.”
BAML’s analysts further say they believe that phantom liquidity fuels complacency among investors who, believing liquidity is good during calm periods, may not see the associated asymmetric risks for higher volatility.
In spite of the Code of Conduct’s role in restoring confidence to the foreign exchange industry by providing a framework within which people can act comfortably and in the full knowledge they are within the bounds of acceptability, it is hard to see how the FX industry can “put the horse back in the stable” regarding restoring the balance between manual and automated traders.
There are some who believe events such as the Cable flash crash highlight the need for FX to move onto totally transparent, highly regulated exchange-type venues and indeed it is hard to escape that conclusion. A word of warning on this, however, for whilst, as was shown in the wake of the US equities flash crash, it is possible to discern what happened, it is still very much a post-facto explanation and does nothing to stop the flash event happening in the first place.
There are concerns that making the FX market totally transparent would negatively impact end users because their orders would become visible to one and all – and the resulting signalling risk will translate into slippage, more expensive hedging and, potentially, even more flash events.
There is little doubt that a degree of standardisation around certain procedures may help reduce the confusion going forward, however, and a good starting point would be how the market establishes the high/low. Profit & Loss understands that the BIS FX Working Group is looking at recommendations for a framework for this as part of its work on the second stage of the Code of Conduct.
All that is likely to achieve, however, is a clearer picture of what happened. As far as the big issue is concerned, avoiding further flash crashes, it is hard to escape the conclusion that they can’t be averted, unless more players are willing to risk warehouse for longer periods, customers can be convinced to leave resting interest orders, or, most controversially, central banks play the role of arbiter and establish parameters at which they are willing to step in to stem the worst of the damage.