TRADETECH WEST: Unstructured Data=Alpha
September 11, 2012
SAN FRANCISCO -- Even with every scrap of news being turned into values that can be acted upon by quantitative traders, there’s plenty of public and private information that can be turned into above-average returns, says a Harvard Business School associate professor.
Such returns, called alpha by traders and firms that want to find it, don’t have to be generated by insider information, luck, or well-known or hard-to-find patterns in market activity.
“Unstructured data is a great way to get it,''' said Chris Malloy, associate professor in finance at Harvard Business School.
The key sources are public information that has not yet been incorporated into stock prices and legally obtained private information that also has not been incorporated into prices, he told traders and technologists gathered at TradeTech West here.
One mechanism is to map out how information flows between local networks of interested parties, business networks, political networks and social networks.
This could produce what the Securities and Exchange Commission could consider insider information, but does not have to.
"There is a fine line, but I don't think it's necessarily the case that information gleaned from a social network has to be considered inside information,'' Malloy said.
Statistics show, he said, that, over a 25-year period, that members of the investment community who have direct connections with executives who run companies get better results.
In effect, they get better results on the five stocks where they have access to senior managers, perhaps through school ties, than the 20 stocks where they don’t have similar access.
The same phenomenon occurs with “geographic proximity.’’ Investors get better returns if they are located near firms they invest in. They tend to be able to get more, usable information and better results. They can conduct their own interviews, make their own site visits and such.
"It's not about the size of network. It's about you using your particular network,'' he said.
Because of the huge amounts of unstructured data available now on the Web, there is also ample opportunity to take information in the public domain and “connect the dots in different ways.’’
Securities filings, lobbying data, campaign contribution reports and the like all provide fodder for making fresh connections about what companies and their stocks might outperform their peers.
In that vein, he said, a company called Recorded Future has been backed by Google and the Central Intelligence Agency to try and predict the future by capturing in great detail the present information available on the Web.
Recorded Future harvests discussion across more than 150,000 blogs, social netowkrs, government filings, and niche publications. Then it analyzes its findings to try and identify hidden relationships.
It also collects and organizes discussions of future events, scores what if finds, and then makes a prediction about an event, such as the possibility that Steve Jobs would step down from his post as chief executive of Apple, for health reasons.
The trick, said Malloy, is to interpret information “already out there” in novel ways.