Revolutionizing Wall Street with News Analytics

by Bob Ward

The development and adoption of news-analytics software is part of a technological revolution that is reshaping Wall Street.

Stock trading is a business in which data is one of the most valuable commodities.

Information from news reports and social media can quickly shift investor sentiment from giddy to gloomy.

Traders with the best data and the smartest, fastest computers can outmaneuver rivals.

Investors are thus increasingly using news analytics to stay on top of developments and ahead of market trends.

Proponents say the technique is effective.

"News-sentiment [analysis] is shown to outperform one-month price momentum when predicting future returns of the S&P 500," said Peter Ager Hafez, owner of Denmark's QuantView Consulting, a financial-markets research company that uses news analytics.

News Analytics: The Beginning

Before the age of electronic trading, large institutional investors often used their organizational size and connections to wring better terms from individuals who executed buy and sell orders. Capital thus typically flowed to the heavyweights who drove the hardest bargains.

Harold Bradley, former president of American Century Ventures, a division of American Century Investments, responded in the late 1990s by becoming one of the first traders to explore using computers to pick stocks.

Bradley—currently chief investment officer of the Kauffman Foundation, which studies and promotes entrepreneurship — said he explored this approach to gain a competitive advantage because smaller traders weren't getting equal access to capital.

Bradley's work laid the foundation for news analytics.

His team first created a neural network that they trained to emulate Bradley's decision-making process and to recognize the combination of factors that his instincts and experience said would indicate a significant move in a stock's price.

Eventually, the team improved the process by collecting performance data for thousands of buy-and-sell scenarios.

The resulting model performed better than random picks and compared favorably to overall market returns.

Analyzing News Analytics

Tech-savvy traders have used computers to mine data from news reports, press releases, and corporate websites since the late 1990s.
But new, linguistics-based software has improved the technology's ability to understand content and determine which items could affect investor sentiment.

News-analytics programs typically work by downloading and assembling large textual databases from many online sources, including LexisNexis (which contains legal and public-records information), news and government websites, newspaper archives, and social-media sites.
The algorithms identify important news items by parsing word definitions, grammar, context, and marker words such as "profit," "loss," "rise," and "exceeds."

A key to this process is the technology's ability to tease out the meanings of words and phrases and disambiguate those with multiple definitions.

News analytics takes several approaches to determining how data could influence investor sentiment.

For example, an application could measure positive or negative sentiments relating to a sequence of news stories about a company. When the overall evaluation moves from negative to positive or vice versa, brokers could treat it as a signal to buy or sell.

A News-Analysis Example

Los Angeles psychiatrist and investor-behavior researcher Richard L. Peterson developed news-analytics software in 2008 to improve stock-trading performance.

Following that year's stock-market slump, he also created the MarketPsy Long-Short Fund. The fund's managers bought stocks based largely on investor sentiment as determined by his software's analysis of online business news, financial social media, and corporate interviews, explained Peterson, now managing partner of investment-consultancy MarketPsych LLC.

His software quantified 400 types of investment-related sentiment and topics — including optimism, anger, and product releases — relating to 6,000 US stocks and exchange-traded funds.

Despite investing in some then-disfavored stocks, the now-closed fund outperformed Standard & Poor's 500-stock index from September 2008 through the end of 2010.

New News-Analysis Services

Business-focused news agencies such as Bloomberg, Dow Jones, and Thomson Reuters have begun offering services that promise to help their customers sift through information automatically.

For example, Bloomberg's offering includes an analysis of the most-read articles by users of its news service, which gets information from more than 30,000 global information sources. Users of the service include market participants from major global banks, investment firms, and financial institutions.

Bloomberg also analyzes the companies and topics most written about in its news service's articles and tracks 20,000 economic indicators culled from government sources, press releases, and websites.

Because of such services, even small traders are now starting to use computers to comb through and analyze news reports, editorials, company websites, blog posts, and even Twitter messages to help determine the best transactions to undertake.

Said Armando Gonzalez, CEO of news-analytics services vendor RavenPack International, "Institutional clients are starting to understand the value of news and having it faster than anybody else to detect the things you don't expect."

Gonzalez said that RavenPack's business has quadrupled in the past year. He predicts the market will grow "significantly" in the next five years.

News-analysis products work but they can take "too long for the news on earnings or other factors to move the model, so you could miss the best target price for buying and/or selling," said Loreen Washburn, a broker with Southern California-based institutional brokerage Heflin & Co. She formerly worked with Navellier & Associates, an early news-analytics adopter.

Analyst Dhuraivel Gunasekaran of stock brokerage HDFC Securities said news analysis and other quantitative investment approaches perform comparatively well but sometimes fail to adapt to changing market conditions as well as traditional schemes.

Bob Ward is a magazine business operations editor with the IEEE Computer Society. Contact him at bnward@computer.org.