The ‘Flash Crash’ of May 6th – Yes, it was a Single Algorithmic Sale
In my book for Kazys Varnelis’ Network Architecture course at Columbia GSAPP, I focused on the landscape of data centers, especially those with ties to high-frequency trading and stock exchanges, in the New York City area. One catalyst for this research was the ‘Flash Crash’ of the afternoon of May 6th where the Dow lost 600 points in minutes, and then regained it nearly as quickly. This was due to a huge volume of algorithmic trading, redoubling the demand for regulation of high-frequency traders, and undoubtedly catching an unaware public off-guard with a troubling display of uncertainty. Normally high-frequency traders are interested in more volatile, cheaper stocks, but this sale had two really unusual aspects which caused the crash – first it was selling futures related to the S&P 500 index itself, and it also executed its huge sale – $4.1 billion – very quickly. Even this huge sale might not have affected the market itself, but as all the other algorithmic traders (‘black boxes’) saw this sale, a cascade of automated sell orders started.
To call these agents ‘traders’ is a stretch – these are not people but computers, all connected through fiber-optic data lines directly to the exchanges – this network is what I spent time mapping for the Fiber Finance book. These transactions are processed in microseconds without any real human involvement. Interestingly, once the Dow started in a downward spiral, electronic trading was stopped. Suddenly the shouting traders at the Stock Exchange floor on Wall Street took back the exchange, to slow it down. The NYSE building in Lower Manhattan is always represented by frantic men shouting buy and sell orders but Fort Knox might be a more appropriate image to its function as a human slowdown machine. The fast-acting trades are happening silently in Northern New Jersey.
Read the New York Times Article: A Single Sale Worth $4.1 Billion Led to the ‘Flash Crash’