Major financial markets have seen three serious “flash crashes” in the last few years. None of the flash crashes were triggered by significant external factors; they were all completely caused by the reaction of market participants to relatively ordinary behaviors by other market participants. Although regulators investigated the incidents, and some systemic changes made to prevent similar occurrences in the future, flash crashes remain a topic of great interest to economists.
In a blog post on Liberty Street Economics, the Federal Reserve Bank of New York analysts “consider a few of the important similarities and differences among three major flash events in the U.S. equities, euro–dollar foreign exchange (FX), and U.S. Treasury markets that occurred between May 2010 and March 2015. All three flash events involved high trading volumes and long-term impacts on depth, but the U.S. Treasury event stands out in terms of both price volatility and price continuity.”
Volume of trading
Note that the chart above plots the volume of contracts traded per second during each event, along with the price. You can see that the volume per second moved up notably in all three events, particularly during the initial legs up, with the rebound occurring with much lower volume in every case.
Flash crashes had long-lasting impact on liquidity
Schaumburg and Yang also point out all three events flash crashes “had persistent, negative impacts on market liquidity, as measured by order book depth in the days following each event.” The graphs above illustrate the amount of standing limit orders at the top ten levels of the order book, with buy and sell limit order depth on the positive and negative axes, respectively, together with intraday price. You can see that order book depth was decimated in the days after each flash crash. Liquidity typically took up to a week to return to pre-crash levels.
The charts here illustrate the intraday volatility during a three-hour window surrounding each flash crash, together with intraday price. Not surprisingly, volatility skyrocketed during each window. Schaumburg and Yang also highlight the brief spikes in volatility at significant news announcements prior to the Treasury and FX events (marked by the vertical dashed lines). But volatility in both events stayed very high leading into the crashes, particularly during the morning of the Treasury market crash. On the other hand, there was no news announcement before the equity flash crash, and volatility remained relatively low until the crash actually began.
Finally, the fed analysts point out that all three events involved very large price movements that were over relatively quickly. To measure price continuity, the analysts looked at the number of trades occurring at each price during each leg of the crash. Panels A, B, and C in the chart above illustrate price continuity during each event window, with the first leg the positive axis and the second leg the negative axis. Panels D, E, and F are exploded views of prices near the peak and trough of each flash crash event. Schaumburg and Yang note: “In general, price continuity was relatively lower (fewer trades per price level) during the initial legs of the events, followed by higher price continuity during the second legs.”