Summary:
· Like other highly liquid assets before it, the British pound experienced its first flash crash on October 7, 2016, with the currency falling by more than 12% against the US dollar on some platforms in just two minutes.
· Some traders lost heavily that day, as did some companies in the real economy. Sports Direct, the UK’s leading sports retailer, employing 18,000 people, lost £15 million in the crash, or around 5% of its net profit.
· High-frequency traders (HFTs) probably played a significant role in this flash crash, although the role of each participant has yet to be determined. In response to this event, an investigation was launched by the Bank for International Settlements (BIS) at the request of Mark Carney, Governor of the BoE.
On October 7, 2016, for no apparent reason, the pound sterling (the world’sthird reserve currency) suffered a « flash crash, « a large-scale crash characterized by an unusual liquidity shock resulting in two sudden and extreme movements, one following the other: 1) a dramatic fall in the value of the British currency against other currencies, particularly the US dollar, and 2) a rapid rise in the value of the pound and a return to a level close to its initial level, all within a relatively short period of a few minutes.
On that day, the pound fell by more than 12% against the US dollar (9% according to Thomson Reuters) in just two minutes, between 8:07 and 8:09 a.m. (Tokyo time), taking many traders by surprise, particularly in Asia, where the day had just begun. The cable—the GBP/USD pair (the third most traded currency pair in the world)— – fell from $1.26 to $1.10 on some platforms ($1.1491 according to Thomson Reuters[1]), with a 250-fold increase in the spread, i.e., the difference between the best bid and ask prices[2], before the pound recovered and rose to $1.24 19 minutes later. Although this crash may seem insignificant at first glance—after all, the value of the pound eventually rose back to a level close to its initial level a few minutes later—some traders lost heavily that day, as did some companies in the real economy. Sports Direct, the UK’s leading sports retailer, which employs 18,000 people and is listed on the London Stock Exchange in the FTSE 100, announced that it had lost £15 million during this flash crash, or 5% of its net income, causing its share price to fall by 12.5% immediately after the announcement. Flash crashes, although seemingly harmless, can therefore have a detrimental effect on the real economy.
1 – The umpteenth flash crash in a long series
The first flash crash of its kind, known as the » U.S. Flash Crash » due to its suddenness and brevity, hit the US market on May 6, 2010. On that day, $1 trillion in market capitalization evaporated in a matter of minutes, and the Dow Jones Industrial Average, the historic index of the New York Stock Exchange and one of the most liquid indices in the world, fell 10% in a few minutes before recovering and returning to a level close to its equilibrium. Several large-scale aftershocks have occurred since then, such as on October 15, 2014, on the 10-year US bond, another extremely liquid asset, with a movement of 7 standard deviations, which theoretically should only occur once every 1.6 billion years, or on August 24, 2015, at the opening of the US market. On that day, the leading US index, the S&P 500, fell 7.8% in a matter of minutes, triggering circuit breakers (trading halt for a few minutes) on 1,278 stocks, before, once again, prices recovered and returned to their initial equilibrium level a few minutes later.
In addition, there were numerous » mini flash crashes, » i.e., mini lightning crashes characterized by a sudden and extreme drop or rise in prices for no apparent reason within a very short time frame (a few seconds this time), including a V-shaped or inverted V-shaped reversal process (in the case of a bullish crash) bringing the price back to a level close to its initial level, have been observed on a daily basis in various markets for several years now. In 2010, for example, the share price of the American company Progress Energy (PGN) fell by around 90% in less than a second, again for no apparent reason, before the share price recovered and returned just as quickly to its initial price after a circuit breaker was activated. In this specific case, some transactions took place at $4.57 for an initial price of around $44.57, representing a potential gain of around 1000% for the most responsive traders.
2 – What we know about the events of October 7
No one knows the reason for the flash crash that occurred on the pound sterling to date, but many hypotheses can be put forward or ruled out.
Firstly, the flash crash of October 7, 2016, occurred at the opening of the Asian markets, in other words at a time when the imbalances between supply and demand in the order books are at their strongest, reminiscent of the flash crash of August 24, 2015, which occurred at the opening of the US markets. The few empirical studies that focus on mini flash crashes also show that most mini flash crashes occur at market open and close, when imbalances and volatility are potentially at their highest. It would therefore appear that these characteristics are shared by flash crashes.
Secondly, the flash crash of October 7, 2016, took place on a Friday (i.e., just before the weekend), at a time when the foreign exchange market is relatively illiquid. It was 7 p.m. in New York, midnight in London, 1 a.m. in continental Europe, and only 7 a.m. in Singapore, Hong Kong, and Shanghai at the time of the crash, a time when most traders are not active. Only Japanese and Australian traders were really on deck, along with some high-frequency traders (HFTs) who were still active.
Thirdly, it would appear that, as is usually the case with flash crashes, the initial downward movement was the strongest and most sudden, with the GBP/USD pair falling from $1.26 to $1.25 in just 20 seconds, before the start of a chain reaction that sent the cable into a kind of nosedive. This strangely echoes the flash crash of May 6, 2010, when a sell order for 75,000 E-Mini index futures contracts, valued at $4.1 billion, was placed by a Kansas-based hedge fund using an automated execution algorithm calibrated to execute the sell order at 9% of the volume calculated in the previous minute, in order to allow the fund to hedge its long positions in the US stock market, which was experiencing high volatility at the time. It is therefore entirely possible that a large sell order on the pound triggered the initial downward movement, very quickly leading to an inability for market makers (THFs and others), in a market that was then very illiquid, to absorb the entire sell order. These orders, executed using execution algorithms, differ from THFs in that the algorithms used only serve to execute buy or sell orders at the best price and therefore do not respond to any particular strategy other than their execution strategy.
Fourthly, while some may have suspected that a trader made a mistake when placing their sell order on the GBP/USD pair or on an asset strongly correlated with this pair, an error commonly referred to as a » fat finger » (a Japanese trader placed an erroneous order for $617 billion in October 2014, before it was canceled by the broker in charge of the order), the fact that no such error has been reported to date indicates that this hypothesis can be ruled out definitively.
Fifth, options on the GBP/USD pair expired on that day. It is therefore entirely possible that some option sellers, most likely banks, had to hedge their positions below a certain threshold (by selling the pound against the dollar, but also the pound against the yen and other currencies due to correlations), thus dragging the cable into a downward spiral and triggering stop orders placed in the various order books (the sharp increase in volumes below $1.25 seems to support this theory). Hedgers therefore most likely contributed to the flash crash that day (with the triggering of automatic hedges below a certain threshold), as the fall in the pound led to an additional need for hedging and resulted in a de facto accumulation of short positions on the pound, which could only push the pound to ever lower levels.
Sixth, the publication at 8:07 a.m. (Tokyo time) of a Financial Times article entitled » Hollande demands tough Brexit negotiations » was suspected for a time of having triggered the flash crash on the pound sterling. Upon verification, it appears that the article appeared 10 seconds after the crash began. That said, it is entirely possible that the sharp rise in volatility on the GBP/USD pair, combined with the publication of the FT article a few seconds later, may have triggered the activation of certain THFs specializing in decrypting information flows (see below).
3 – A flash crash that appears to bear the hallmarks of high-frequency trading
First appearing in the early 2000s, HFT has grown significantly over the last ten years, now accounting for around 50% of transactions in the United States, 40% of transactions in Europe, and just over 30% of transactions in Asia. HFT is characterized by a fully automated, algorithm-driven trading process that accelerates time and profoundly changes the microstructure of financial markets.
High-frequency traders (HFTs) use five main types of strategies:
(1) Market making
(2) Arbitrage
(3) Directional trading
(4) Structural trading
(5) Manipulation strategies
All five may have played a role, to varying degrees, in the flash crash of the British pound.
· Potential role played by high-frequency market makers
Traditional market makers, typically well-capitalized financial institutions, have gradually been replaced by high-frequency market makers, which are generally much less capitalized than their predecessors. Thanks to their greater speed (able to buy and sell in milliseconds), high-frequency market makers are able to reduce their exposure to inventory risk. In exchange for this reduction in their exposure to inventory risk, they are able to offer competitive spreads , thereby reducing transaction costs for end investors. Under normal market conditions, their role is therefore highly beneficial. However, the ability of high-frequency market makers to reduce their exposure to inventory risk also means that they are able to cut their exposure very quickly in the event of a sharp increase in volatility, which potentially leads to an increase in risk (on an occasional basis) for end investors, as the exit of high-frequency market makers has the ability to dry up liquidity in order books very quickly.
On October 7, 2016, thousands of transactions took place during the pound’s dramatic fall, 40% of which were aggressive buy orders[5] according to data collected by Nanex LLC, characterized by price increases over several tiny time periods, This suggests that high-frequency market makers played a significant role in this flash crash, « passing the hot potato » until it had cooled down enough, in other words after a large discount, to be able to return to it with more conviction (and less risk), as was the case with the flash crash of May 6, 2010. At the same time, liquidity dried up very quickly, as evidenced by trading volumes in the two minutes following the start of the crash, suggesting that high-frequency market makers did indeed withdraw from the market. The BIS’s ongoing investigation will likely shed more light on this issue in the coming months.
· Potential role played by high-frequency arbitrageurs
High-frequency arbitrageurs seek to take advantage of small price differences between two or more assets, as is the case, for example, with triangular arbitrage in the foreign exchange market, by simultaneously taking long (buy) and short (sell) positions on these assets via market orders. It is therefore highly likely that some high-frequency arbitrageurs took advantage of the extreme weakness of the pound to take action, exploiting the imbalances between the pound and the dollar and the yen, for example, or even between the pound and other more exotic currencies. This could partly explain why 40% of orders were market orders to buy (in a very rapid process of buying and reselling between the pound and other currencies).
· Potential role played by high-frequency directional traders
High-frequency directional traders focus on one side of the order book by analyzing changes in the short-term balance between supply and demand and using signals based on an analysis of flows and updates to the order book, the arrival of new information (whether true or not), such as a press release, or an analysis of the footprints left by other traders engaged in a large-scale acquisition or liquidation process (algorithmic trading). At the time of the pound’s fall, all the warning lights were flashing red (massive shift in the balance between supply and demand, drying up of liquidity in order books, arrival of new information 10 seconds after the start of the crash (the Financial Times article), and potentially massive liquidation triggering the initial downward movement). The asset in question, the pound, could therefore only be sold by high-frequency directional traders. If this is indeed the case, then the BIS investigation should reveal significant use of market sell orders during the downward phase of the crash.
· Potential role played by high-frequency structural traders
High-frequency structural traders focus on structural market inefficiencies by taking advantage of cross quotes or exploiting, thanks to their speed of execution, the slow adjustment of certain limit orders. It is clear that structural trading plays a role in any flash crash, with short limit orders in the order book being the first victims of the fastest traders who are able to trade on multiple platforms at once.
· Potential role played by high-frequency manipulators
Finally, high-frequency manipulators use so-called « breakout » manipulation strategies, a non-exhaustive list of which is provided below:
1) Momentum ignition
2) Spoofing
3) Order book flooding (quote stuffing)
4) Quote dangling
5) Pack hunting
6)Stop loss hunting
Spoofing appears to be the most widely used manipulation strategy in recent years, despite being officially banned in the United States since 2010 by the Dodd-Frank Act. This strategy involves placing a large number of fictitious limit orders (which cannot be executed due to the ability of THFs to cancel their orders in an infinitesimally short time) on only one side of the order book in order to create the illusion of an imbalance between supply and demand, and thus encourage other traders (particularly high-frequency directional traders) to favor the side of the order book where orders are accumulating, thereby shifting the market in the desired direction. The trader then takes a position in the opposite direction in order to catch the market off guard before liquidating their position at a profit. The arrest in 2015 of British trader Navinder Sarao, who apparently specialized in this strategy and was accused by the US authorities of being partly responsible for the flash crash of May 6, 2010 (Sarao is expected to be extradited to the US in the coming weeks, with a trial scheduled for 2017), invites us not to rule out this possibility, although it is far from being the most likely. Finally, a strategy of hunting for stop orders cannot be ruled out either, given the very large number of stop orders placed at $1.25.
Conclusion
Like other highly liquid assets before it, the pound sterling has experienced its first flash crash. Such a price gap in a market as liquid as Forex naturally raises many questions, particularly about the state of liquidity in the markets in general, but also about the role played by THFs in such price movements. As highlighted in a note from Bank of America Merrill Lynch, reported by the Financial Times, the speed and magnitude of the decline during the flash crash of the pound sterling underscores the dangers of potentially phantom liquidity, characterized by an almost immediate drying up of liquidity when volatility increases sharply and extremely.
The investigation launched by Mark Carney, Governor of the BoE, at the Bank for International Settlements (BIS) should provide us with some initial answers to the questions raised in this analysis.
References
Biais Bruno and Foucault Thierry, HFT and Market Quality, Bankers, Markets & Investors, No. 128 January-February 2014
Martin Katie, Sterling flash crash highlights warnings on FX liquidity, Financial Times, October 11, 2016
Blitz Robert and Stafford Philip, Pound plummet blamed on ‘liquidity holes’, Financial Times, October 10, 2016
Hughes Jennifer and Lewis Leo, How ‘all hell broke loose’ on flash crash Friday, Financial Times, October 7, 2016
Notes:
[1]Currency pairs are traded on different platforms, with the low point displayed on a platform corresponding to the last transaction carried out on that platform, which explains the differences in low points reported in the media.
[2]The spread corresponds to the compensation required by a market maker to cover the risk of dealing with traders who are better informed than they are.
[3]Bruno Biais and Thierry Foucault (2014)
[4]Inventory risk is the possibility that a price change will negatively affect the value of assets held in stock by a market maker. In the event of low market maker capitalization and a sharp drop in the price of assets held in stock, the market maker will have no choice but to liquidate its stock quickly to limit its losses, thereby accentuating the decline in asset prices.
[5]Aggressive orders are market orders, as opposed to limit orders.
