|Stock trading 'fractures' may warn of next crash
| February 9, 2012
Might a fleeting and little understood aspect of stock market dynamics hold the key to warding off financial crashes? That is the tantalising suggestion to emerge from a group of physicists who have been studying stock movements.
The study is the first to focus on an ultra-fast feature of market dynamics that the team, led by Neil Johnson at the University of Miami in Coral Gables, calls a "fracture". Fractures happen when the price of a stock briefly shoots up or down, often before returning to its original level. They take place so quickly, sometimes lasting less than half a second, that they can be invisible to any human following the price. "If you blink you miss it," says Johnson. His research shows that there seems to be a link between these fractures and sudden stock market crashes, known as "black swans".
If fractures are a source of instability, computerised trading algorithms may be to blame. Use of these algorithms, which make automated trades in milliseconds, has mushroomed in recent years. Competition between rival algorithms is so great that one company is spending $300 million to build a transatlantic cable that will shave 6 milliseconds off the time it takes to exchange signals between the financial hubs of London and New York. But finance experts fear that one or more out-of-control algorithms could cause a crash, as may have happened in the so-called Flash Crash of May 2010.
Johnson's research, which was posted online on 8 February, is based on price logs from over 60 different markets collected by Eric Hunsader of Nanex, a Chicago-based company that sells streaming market data. Johnson and Hunsader trawled through the data and found that fractures are remarkably common – 18,520 took place between 2006 and 2010.
Intriguingly, the number of daily fractures increased about a week before the stock market crash of September 2008, and also before a sudden but smaller crash in May 2010, which is still not fully understood. Johnson thinks that the build-up of fractures can in some cases destablise the entire market, much as the accumulation of tiny cracks can lead to catastrophic failures in structures such as aircraft wings. "You're seeing something starting to break open," says Johnson.
The link between fractures and stock market crashes requires further investigation, but suggests it might be possible to build an early-warning system based on the rate at which fractures occur. "Johnson makes a compelling case," says Dave Cliff, an expert in complex systems at the University of Bristol, UK. "Developing [his] analysis techniques into methods or tools for reliably predicting crashes is a very appealing prospect." Forecasting every crash is probably impossible, adds Cliff, but the research could lead to a system that at least alerts regulators to incipient instability. "This work could turn out to be a major first step in that direction."
"There is a huge amount of interconnected algorithms that cannot be controlled by regulators," says Tobias Preis, a researcher at Boston University in Massachusetts and founder of Artemis Capital Asset Management in Holzheim, Germany.
When Johnson analysed the frequency at which fractures of different size and duration occur, he found that fractures lasting one second or less follow a different statistical pattern to longer ones. The one-second cut-off is significant, says Johnson, since it is about the time it takes a human trader to weigh up a piece of evidence and make a decision. On shorter timescales computers control market dynamics since they can make decisions in mere milliseconds.
This ultra-fast computing-controlled trading regime is barely recognised, let alone controlled by any government regulator with an interest in preventing crashes. "Nobody really knows what's going on down there," says Johnson. "It's like the wild west."