| Print |
Man versus machine
Like a runaway train, a berserk computer nearly sent a leading US stockbroker into bankruptcy last week. It could happen again.
The number one retail market maker in US listed securities, Knight Capital, teetered on the brink of destruction last Wednesday morning when its trading systems spewed out thousands of false orders to buy and sell US traded stocks.
As company executives tallied the cost of this latest financial institution misadventure, the credit lines of New Jersey based Knight Capital dried up and the online brokers using the Knight Capital digital pipes to execute large volumes of trades pulled their business.
Early estimates put the cost of unwinding the false trades at US$440 million, more than twice the firm’s profit over the 12 months to March 2012 and nearly 30 percent of its net equity. Confronted by its own mortality after an 80 percent fall in its share price, Knight Capital opened itself to bidders for all or a part of its assets.
New York stock exchange listed Knight Capital has recently accounted for one-third of the 14.6 billion shares traded for US retail investors in an average month and 15 percent of the market making activities in US stocks. It was a critical link in U.S. equity markets adding liquidity and facilitating executions. But on Wednesday morning it was haemorrhaging at a rate of $10 million a minute.
Technology now permits firms like Knight Capital to relay millions of stock market orders through one or more of the 18 stock exchanges in the USA in a matter of seconds. Being able to execute an order a single millisecond before someone else gives a firm such as Knight Capital a competitive advantage. Being able to do that millions and millions of times a day means it has a business.
Orders can be initiated through trading algorithms in the absence of any human intervention. Human beings slow things down. Machines are much faster. Unimaginably large numbers of trades, each with microscopically fine profit margins, can make for a meaningfully profitable business. And, the speedier the machines become, the more they are asked to do and the greater their distance from any human oversight.
Unfortunately, the machines lack good, old-fashioned human common sense. As the Knight Capital computers went berserk on Wednesday morning, human traders on the floor of the New York stock exchange knew something was going awry. They expressed their concerns to the business news reporters on the floor of the exchange who reported a succession of strange trades in some 140 different stocks in the early hours of morning trading.
As the humans watched, the trading debacle continued to unfold for three-quarters of an hour as the computers unthinkingly did what they had been instructed to do, probably by someone who did not fully understand the instruction he was giving the machines and possibly horrified by the thought that he had created the equivalent of a high tech runaway train.
One solution to this problem is now obvious. There is apparently no requirement for the equivalent of the “dead man’s switch” on the runaway train. Parallel systems are needed so that as soon as a dodgy computer shows signs of going off the rails, another recognises what is happening and closes the first one down perhaps within just a few minutes if not seconds.
Of course, another layer of technology would add costs and make an already complex business even more difficult to understand.
As complexity rises, fewer and fewer people are able to make sense of what is happening. Sure, many will grasp the general idea of buying or selling fast (ie, high velocity trading) but few will comprehend the mechanics of how the digital technology signals the machines to buy or sell and then reports back to clients. Fewer still could explain why one machine can execute an order in one millisecond while another compromises business success because it takes three.
In this new age of trading, the chief executive looks on just as perplexed, bamboozled and impotent as anyone else as his machines unleash their destructive fury. In fact, in this instance, the idea of a “chief” executive does not exist.
In days gone by, chief executives could call the shots. They were in charge. If a company was going to be ruined, they would be the ones to do it. No longer.
At Knight Capital, the chief executive is in charge of cleaning up the mess. The identity of the most powerful person at the firm is a mystery. He is the person who knows where to find the off switch to start and stop the computers. But, wait, in a world of 24/7 global trading, there may not be any off switch.
The most powerful person might not even be in the building. He or she might be someone who wrote the software or engineered the system and is, now, long gone. For all we know, they could have deliberately or inadvertently left a bug that activates itself some day in the future to create enough havoc to close global markets for days.
The chief executive and members of the board might have no inkling of the dangers they face. Or, if they do, they come to work each day with fingers crossed hoping that the computer systems will not blow up on their watch. That is the extent of their power.
For investors, the Knight Capital trading failure is a frightening reminder that financial institutions relying on sophisticated trading systems will be prone forever to these types of events. Now add Tom Joyce to the names of Jon Corzine and Jamie Dimon on the list of ashen faced chief executives scratching their heads and asking “how did that happen?”
John Robertson is an economist and corporate investment adviser. He has had over 25 years of experience in international financial and commodity markets, corporate strategy, financial and business evaluation and government policy in Australia, London and New York.
Want to read more articles by John Robertson Click on the links below
Copyright © John Robertson . Published by MercatorNet.com. You may download and print extracts from this article for your own personal and non-commercial use only. Contact us if you wish to discuss republication.