Byline: John Lanchester
ON MAY 6 2010, out of a clear blue sky, the US stock market suddenly dropped by 10 per cent. A trillion dollars was wiped off the value of US equities. The precipitate dive took place over 20 minutes, an astonishingly brief time span which earned it the name "flash crash". Over the next few hours, stock prices gradually recovered, making it clear that there was no real systemic crash, and leaving investors with one simple question: "What the hell was that?" When things like this happen, the first scapegoat is the "fat finger". Fat finger mistakes are caused by traders making keyboard entry errors in the quantities or prices of stock they are trying to trade. The commonest is putting the decimal point in the wrong place. These boo-boos come along at fairly regular intervals. My personal favourite was in 2001: a trader at UBS Warburg who meant to sell 16 shares of a company at 600,000 yen, and ended up selling 610,000 shares at 6 yen. Doh! Cost of mistake: about $100 million. Difficulty of explaining what had happened to boss: considerable.
But these mistakes can usually be absorbed pretty quickly by the market, since it is clear after some scrutiny that the underlying trade makes no sense at all -- so there is no real issue for the investment community to worry about.
There is, however, a second potential scapegoat for a sudden market movement like the flash crash. The biggest one-day stock market crash in history was caused not by people but by computers. It happened on October 19, 1987, a day which became known as "Black Monday", and was the direct result of an innovation in the stock market. The new thing was "portfolio insurance" -- a way of automatically setting a selling price for a basket of stocks. If the stocks fell below that price, the computer monitoring the portfolio would execute a sell order, and the investor's loss would be limited. It was supposed to be a foolproof way of protecting people's money.
On Black Monday, however, the markets had gone into the weekend feeling a little twitchy, and a backlog of sell orders had built up. When those orders were executed first thing on the Monday morning, prices started to drop, and then the portfolio insurance kicked in. The trouble is that all the computers involved were using broadly similar mathematical algorithms -- so they all started to do the same thing at the same time. What in theory should have been an orderly, safety-first sell-off instead became a stampede for the exit. The computers were all selling into a market with no buyers -- and that is a recipe for prices to freefall. In the event they lost 22.6 per cent of their value in one day, much worse than any equivalent period since the great crash of 1929.
One of the morals of the Black Monday story is that computer trading has big risks. Unfortunately, because there have been no equivalent computer-induced spasms inside the big exchanges since, complacency has built up. But conditions are changing in a way that makes another computer meltdown more likely. It was this possibility that caused markets to be so spooked by the flash crash this year: the phenomenon looked like the forewarning of a disaster which is beginning to seem thinkable.
The phenomenon responsible for the new uncertainty is high frequency trading. This is a technique used by the big boys: investment banks, hedge funds and large institutional investors. The trading platforms involved use extraordinary computer power and super-sophisticated computer algorithms. If Gordon Gekko were to make a list of all the new things that had happened in the stock market since he came out of jail, high frequency trading would be right at the top.
The computers are sited as close to the stock exchanges as possible, and use the fastest data connections that money can buy. All this is to gain an advantage of thousandths of a second.
The computers analyse the flow of data through the market and study patterns of movements in stocks. …