'Big Data' Researchers Turn to Google to Beat the Markets

The New York Stock Exchange Photograph by Scott Eells/Bloomberg

Time to fire your portfolio manager? “Big data” researchers have found that mining Google search terms related to finance and plugging that data into an investment strategy would have outperformed all but the world’s greatest stock pickers over an eight-year stretch ending in 2011.

The findings of the research, “Quantifying Trading Behavior in Financial Markets,” were published today in the journal Nature by a team of big data academics, Professor Tobias Preis of Warwick Business School in the U.K., Helen Susannah Moat of University College London, and H. Eugene Stanley of Boston University. Freis and Moat have already created a stir by using similar data-crunching approaches to quantify and model stock price fluctuations of companies on the Standard & Poor’s 500 index and gross domestic product growth among many of the world’s biggest countries.

The latest effort may be their most controversial yet. The team analyzed changes in the frequency of 98 general finance terms such as revenue, unemployment, credit, and Nasdaq in Google searches from 2004 to 2011. The team then used the search volume data to determine an investment strategy that it called the “Google Trends Strategy.” It generated a market-beating return over the eight-year span.

Here’s how it works: The team analyzed on a week-by-week basis the ups and downs of Google search volume for the 98 finance-related terms. Based on the results, it would sell or buy into its theoretical Dow Jones industrial average index-weighted portfolio. “If the volume of search terms went up in the previous week, we would open up a hypothetical short position and sell in the following week,” explains Preis. “If there was a decrease in volume, then we’d buy.”

Using this strategy, the team scored a 326 percent return when analyzing the keyword “debt.” (Other terms scored lesser returns). In contrast, a conservative buy-and-hold strategy that ignored Google searches and followed a more traditional investment strategy netted a return of 16 percent over the same period, the researchers found.

Preis says the findings—that aggregate search data could be used to determine early warning signs of subsequent stock market movements—have caught the finance industry’s attention. “It’s been very exciting for some people in finance,” he says. “It may lead to change in their investment algorithms or strategies.”

Adds Moat, “it’s exciting to see that online search data may give us new insight into how humans gather information before making decisions—a process which was previously very difficult to measure.”

Moat and Preis are among the big data researchers whose work has been funded in part by the federally funded Open Source Indicators program, which is housed within the Intelligence Advanced Research Projects Activity, or IARPA. The U.S. intelligence community has become enamored with big data and how it could be used to forecast a deadly disease outbreak, a sudden currency collapse, or even war. This is the first piece of predictive research published by Moat and Preis that has come out of that program.