Nerds on Wall Street: Math, Machines, and Wired Markets

3.14. Agents for News and Pre-News

A salient feature of the relationship between news and markets is that many news events lag the market, but some lead it. Examples of lagging news would include any story that says, "XYZ was up 30 percent today because.... " In the old, strong-form efficient market hypothesis days, academics would argue that there was no news worth trading on since all information, public and private, was immediately reflected in prices. The new so-called adaptive market hypothesis and a certain degree of common sense allow that some news (but not all) is news to everyone at the same time, and that someone can be the first to profit from it. This opens yet another front in the algo wars.

In the past year, we have seen the major news providers, Dow Jones^[] and Reuters,^[] offering costly high-end, low-latency news feeds designed for machines. In addition to being faster, they include extensive XML tagging for a variety of stories. These semantic Web approaches allow clever algo warriors to extract the salient facts with much greater accuracy than they could achieve writing code to parse plaintext feeds designed for human readers.

What kind of tags are they talking about? The Dow Jones product is described as over 150 macroeconomic indicators, in developed markets, and a wide range of news on publicly traded U.S. and Canadian firms, as well as some in the United Kingdom. Sample tag categories include earnings, mergers, analyst reports and opinions, executive changes, ...