There can be little doubt that HFT represents a genuine financial innovation and is definitely here to stay. However, it is possible for this genuine breakthrough can ultimately be taken too far. HFT has certainly had a dramatic effect upon markets. Trading has become faster and in some ways more efficient. Overall, it is perhaps ultimately best to view HFT as a trade-off between enhanced price discovery at the cost of higher price volatility. There is also an ever-present danger of unintended consequences. At times HFT can simply lead to artificially inflated transaction costs [1]. Against this backdrop tales of predatory trading and Flash Crashes abound.

As the name suggests Flash Crashes involve very sudden drops in price over a matter of seconds or minutes. Worryingly these events are sufficiently commonplace to have been given the label “Flash Crashes”. They even have their own dedicated research website! [2]. Flash Rallies, where instead the price spike upwards rather than downwards, can also occur. However, the underlying market instabilities mean that Flash Rallies are unlikely to represent good news.

Flash Crashes first came to global prominence during the Flash Crash of May 6th 2010. Around 2.30-3.00pm saw dramatic upheaval on US futures and equity markets with the Dow Jones Industrial Average losing around 10% of its value before recovering. Perhaps only the timing of the event, sufficiently far away from the close of the US markets, prevented financial Armageddon [3].

It seems that Flash Crashes may represent a new financial reality. The first major Flash Crash identified in the literature dates back to a Flash Crash in the USD/JPY currency pair on August 16th 2007 [4]. In some ways Flash Crashes can also be thought of as more extreme version of the way in which early computerised Portfolio Insurance strategies, and the automated fire sales of stocks into an already downward trending market, exacerbated the 1987 stock market crash. Recent high profile Flash Crashes in the GBP/USD currency pair and in the gold futures market emphasise that Flash Crashes are far from a purely historical phenomenon.

Flash Crashes may indeed be here to stay. Flash Crashes have affected asset classes as diverse as stocks, bonds, currencies and commodities. Most worryingly, it seems that highly capitalised stocks may be particularly vulnerable to Flash Crashes. Around 200 Flash Crash events can be identified for large company stocks listed on either the New York Stock Exchange (NYSE) or the NASDAQ. Further, during the Flash Crash of 2010 the prices of major companies, such as the consultancy firm Accenture, were particularly severely affected.

Flash Crashes, HFT and the wider issue of FinTech are of course large and fast-moving subjects in their own right [5-6]. In response statistical models of Flash Crashes are still playing catch up with the new financial reality. A basic question to ask is do Flash Crashes occur simply because traders think that they can make money during such episodes? If such a self-fulfilling prophecy applies one simple implication of this would be that highly capitalised stocks are more likely to be affected by Flash Crashes.

Is there a link between market capitalisation and vulnerability to Flash Crashes? Flash Crash events for stocks listed on the NYSE and NASDAQ are shown below in Table 1 together with the category of market capitalisation. The clear implication is that more highly capitalised stocks are more likely to experience a Flash Crash event. This simple observation can be made more statistically precise using a binomial generalised linear model [7]. A probit regression model gives statistical evidence of a link between propensity of Flash Crash events and market capitalisation (p=0.000). A plot of how the probability of a Flash Crash depends upon the market capitalisation is shown below in Figure 1.

Table 1: Flash Crash events for stocks listed on the NYSE and the NASDAQ (March 2011-June 2014)

Figure 1: Estimated probability of experiencing a Flash Crash event for BYSE and NASDAQ listed stocks by market capitalisation: evidence from a fitted probit generalised linear regression model.

Allied to the above theoretical work in [8] suggests two possible approaches to try and mitigate the risk of future Flash Crashes. Firstly, as suggested by the above, it might be possible to try and limit the severity of Flash Crashes by in turn placing limits upon the profitability of HFT and predatory trading. Suggestions in the finance literature include imposing time delays on exchanges, measures limiting order cancellations and trading speeds, and transaction costs.

Secondly, one could seek to reduce the market impact of individual trades. This might be made possible by spreading HFT transactions out over either longer periods of time or across different exchanges. This reflects previous work that suggests Flash Crashes may be linked to extreme forms of market illiquidity and market concentration (whereby individual large traders may have a disproportionate impact upon the market price).

The law of unintended consequences in a complex financial system, a highly competitive trading environment, and perverse trading incentives all point to one uncomfortable truth. It seems that Flash Crashes may always be with us. In scenes reminiscent of science fiction market regulators and traders alike face a tough challenge in continuing to evolve and adapt to the new challenges posed by the rise of the machines (HFTs) [9].

References

[1] Lewis, M. (2014) Flashboys: A Wall Street Revolt. Norton and Company, New York.
[2] www.nanex.net
[3] Cliff, D. and Northrup, (2012) The global financial markets: an ultra-large-scale systems perspective. In Calinescu, R. and Garlan, D. (eds). Monterey Workshop 2012, LNCS 7539, pp. 29-70.
[4] Chaboud, A.P., Chiquoine, B., Hjalmarsson, E. and Vega, C. (2009) Rise of the machines: algorithmic trading in the foreign exchange market. Federal Reserve International Finance Discussion Paper No. 980.
[5] Aldridge, I. (2016) ETFs, high-frequency trading and Flash Crashes. Journal of Portfolio Management 43 17-28
[6] Aldridge, I. And Krawciw, S. (2017) Real-time risk – what investors should know about FinTech, High-Frequency Trading and Flash Crashes. Wiley.
[7] Bingham, N, H, and Fry, J. M. (2010) Regression: Linear models in statistics. Springer Undergraduate Mathematics Series, Springer, London.
[8] Fry, J. M. and Serbera, J-P. (2017) Modelling and mitigating Flash Crashes. Preprint.
[9] Serbera, J-P. and Paumaud, P. (2016) The fall of high-frequency trading: A survey of competition and profits. Research in International Business and Finance 36 271-87.