Currency traders continue to be wary of the potential for flash crashes during the dreaded “witching hour” between the end of New York’s business day and the start of Tokyo’s. Flash crashes can cause massive losses to those on the wrong side of them and without central bank intervention more of these are on their way.
Early this year, on January 2nd, the infamous “witching hour” once again reared its ugly head.
During this hour of market illiquidity between 5pm and 6pm EST, which links the end of New York’s FX session with the beginning of Tokyo’s, investors watched in horror on this day as USD/JPY slumped in a matter of minutes by an amount equal to three times the pair’s average weekly range—a move of roughly 4 percent. Less than 30 minutes later, with common sense prevailing, nearly three-quarters of those losses were erased and those that remained were clawed back over the following days.
Like the yen, the Australian dollar suffered a rapid decline and recovery, this time worth 4.5 percent. Importantly, on its way down, AUD/USD crashed through a multitude of significant price levels (levels that had not traded for a decade), triggering as it went any number of “stop” and “if-touched” orders placed by hedgers and speculators alike, resulting in massive losses when the markets sprang back.
The witching hour has a steady recent history of throwing up large and unexpected currency moves, also known as “flash crashes.” In early 2016, South Africa’s rand plunged by more than 8 percent against the US dollar in a matter of minutes. Sterling did the same in October 2016, that time by 6 percent. To a lesser extent, the Swiss franc experienced the same on February 8th of this year, when within minutes it lost a percent of its value (still twice its average daily move at the time).
While the mechanics of these moves are simple, the changes needed to eradicate them are anything but.
According to data from Aite Group, foreign exchange turnover during this treacherous hour is just 2 percent of peak turnover. From an execution standpoint, this means that large sell orders entering the market at this time, or a multitude of smaller orders, can easily mop up all bids within close proximity to the last traded price. Once these bids have been “hit” and cleared, there’s nothing to stop these sell orders being matched with bids at lower and lower levels (matching will sometimes skip several prices at a time) until they’ve been filled in their entirety. Of course, the process of order matching happens extremely rapidly and in the absence of exchange circuit breakers (currencies are traded over-the-counter, not on exchanges) prices can crash.
For all their sophistication, the algorithms which are so prevalent in today’s markets make matters worse in this regard. Obviously, they facilitate speedier price changes, but in some cases algorithmic orders also trigger other algorithmic orders in a way that manually-configured orders never could, and since algorithms feel no obligation to quote a bid or offer, sometimes they back off completely just as volatility is picking up and shrink liquidity further still. Furthermore, unlike the humans that preceded them, algorithms won’t always ask if they’ve been given a duff price or order quantity—one that makes no sense given the prevailing conditions and which could send the market spiralling.
Consider also the self-fulfilling nature of this problem. If a trader lacks confidence in the market’s ability to handle their orders during an especially quiet period, or simply wants the cheapest possible transaction cost in the form of a narrow spread, he or she will simply wait until a later time to place those orders. While that’s a completely sensible strategy, it’s one that, if followed by everybody during the witching hour, would leave no market at all to speak of.
The consensus among those who care to share opinions on these matters is that structural factors creating the problem, including the “make-up and behaviour of principals, intermediating agents and trading platforms” (to quote the Reserve Bank of Australia), are irreversible.
Although witching-hour moves are yet to lead to wider market disruption, try saying that we shouldn’t worry about them to a trader who has just lost a month’s earnings, or worse, because he or she is short AUD/USD from a 10-year low; you’ll receive robust castigation.
As is their way, traders, particularly those in the vulnerable time zones of Australia and New Zealand, are attempting to predict when the next flash crash might occur, if only in an effort to save themselves. One Sydney-based senior strategist at DBS said last month that each of his team “now has Japanese holiday calendars printed in big font on their desk”—national holidays being times of even less liquidity.
It is unclear whether central banks or regulators will, or can, offer the necessary liquidity-bolstering measures to counter such moves but until they do it’s a question of when, not if, another major currency is on the wrong end of a witching-hour beatdown.