At around midnight on 7 October 2016, the price of sterling dropped heavily against the US dollar - falling 9% or about £10 billion before quickly recovering most of the lost ground. This shouldn’t happen. The foreign exchange (FX) market is the biggest and most liquid of all the financial markets. It hosts trading of over five trillion US dollars’ worth of currencies every day. This partly stems from the UK and is due to trade between the pound sterling and the US Dollar. Sharp bouts of illiquidity and volatility between the two currencies are unusual.
So how did the flash crash start and why did it escalate? More than a year on, there is still no precise answer to the first question. The Bank for International Settlements (BIS) has cited a range of possible factors, including a poorly calibrated algorithm, human error and the time of day.
Price drop volatility
Source: FCA EMIR derivatives reporting data
An interesting picture is emerging about how the crash escalated and what went on between traders during 21 minutes of ‘extreme dysfunction,’ as the BIS described..
Trade data reported to the FCA from the FX OTC market – where dealers can trade directly with clients and each other – shows that the initial trigger was amplified by dealers heading for the exit doors.
In one sense this isn’t really remarkable. You expect dealers to leave the market in times of stress. Albeit they will take on more risky inventory if the price concessions are compelling enough.
What is surprising from the newly-released FCA data is just how quickly, and deeply, a risk-off sentiment swept through investment banks.
Banks and other dealers leave the market
Analysis of around 30,000 OTC derivatives transactions made during the crash shows dealers (including investment banks), were the main contributors to a drying-up of liquidity. Reducing their trading activity in OTC GBP/USD derivatives from normal levels of around £32 million per second over a whole day, to just £0.2million per second during the flash crash.
Additionally, investment banks contributed the most to increased transaction costs during the flash crash, with bid-ask spreads -a key measure of liquidity-60 times higher than normal. Inter-dealer trading also fell as a percentage of the total market from 61%, to just 2% during the flash crash period. As a result, dealers were able to hedge only 31% of their trades with clients.
Trading activity per second
Source: FCA EMIR derivatives reporting data; abbreviations: FF (financial firms other than dealers), NF (non-financial firms)
This collapse in trading was partially offset by other firms – including hedge funds, asset managers and high-frequency traders – who stepped in to provide liquidity from the buy side. Their market share increased from a typical 35% to 98% during the flash crash, albeit the absolute amount was not enough to halt the fall in price.
One other interesting pattern that emerges from the data is the interaction between the OTC derivatives markets and the underlying spot market. As expected, the flash crash starts in the spot market and quickly spreads to the OTC market. That is because OTC dealers base their prices on information gleaned from the underlying spot market.
But as the OTC market becomes more and more dysfunctional during the flash crash, this causes a negative feedback loop of illiquidity between the two markets as spot market trades cannot be hedged as easily in the OTC market.
Eventually, this feedback loop compounds the impact of the initial price shock in the underlying spot market. In fact, we estimate that illiquidity in the spot market is 12% worse than it otherwise would be due to the presence of this feedback loop.
Market share of different types of participants
Source: FCA EMIR derivatives reporting data; abbreviations: FF (financial firms other than dealers), NF (non-financial firms)
This begs the question: how do you stop a flash crash from escalating?
One step European regulators have taken already, as part of MiFID II, is to increase the transparency of trading by publishing details of certain trades in financial instruments including FX derivatives (spot FX markets are not covered by MiFID II).
Firms are now also compelled to trade some classes of derivatives (for example, key interest rate derivatives) on lit exchanges rather than OTC. Venues can adopt certain controls to restrict the occurrence of these types of rapid price movement, for example circuit breakers.
MiFID2 also seeks to strengthen electronic markets by introducing more rigorous testing requirements for algorithms. And the impact of these policies will become clearer over time.
Read the full analysis in the Occasional Paper.
