Machine Readable Information

An article that caught my eye a few weeks back was the announced acquisition of the Reuters Group by the Thomson Corporation for over $17 billion. The combined companies would create the largest financial news provider.

More interestingly to me than the acquisition itself is the potential impact on future stock market trading. About one-third of stock market trading is currently performed through program or automatic trading. During the week of 14—18 May, for example, the New York Stock Exchange reported that "program trading amounted to 35.3 percent average of NYSE daily volume of 3,233.2 million shares, or 1,142.9 million program shares traded per day. This included program trading associated with the May 18 monthly expiration of stock-index options and futures."

Program trading is inherently "backward looking"' in the sense that the trades are automatically made based on price fluctuations that meet certain criteria. The focus in recent years has been on increasing the speed of such trades.

However, both Reuters and Thomson have been working on what is generally called machine readable news, for instance, a "Reuters system will 'read' news articles and score how positive or negative they are. The system will enable customers to analyse news across thousands of companies, far more quickly than can be done by humans. This will enable trading machines to react to market moving news in milliseconds" Not only are current news stories being made "machine readable,"but Reuters is making its archives machine readable as well.

The Financial Times reports that Thomson has developed software that can automatically "generate the stories work so fast an earnings story can be turned around within 0.3 seconds of a company making results public." In addition, as noted in the FT story, program trading, "is set to rise much further in the coming years as fund managers, along with brokers and exchanges, strive for ever-greater speed and control over the trading cycle amid heightened market competition and consolidation."

The combination of incredibly fast automatic news generation along with historical data to create predictive market responses to such news may create some interesting program market trading impacts. It will be interesting to see, as machine readable news becomes more available, whether the market becomes more volatile as a result, or whether dangerous feed-forward loops are produced during boom times, or more likely, individuals or governments will make use of this capability to deliberately hoax financial markets for either personal or strategic gain.

A government run news agency, for instance, could find it in its self-interest to plant a financial story, say involving some scarce resource, e.g., "petroleum" which could cause a panic in the market. By studying the conditions that caused market panics in the past, it might turn into a potential non-military but very effective weapon. Maybe governments (and the exchanges) may want to start thinking about how financial companies could use all this information for not only creating financial rewards, but how others could manipulate it to create major financial risks.



Risk Factor

IEEE Spectrum's risk analysis blog, featuring daily news, updates and analysis on computing and IT projects, software and systems failures, successes and innovations, security threats, and more.

Robert Charette
Spotsylvania, Va.
Willie D. Jones
New York City