Follow Datanami:
April 26, 2013

Predicting Financial Markets with a Google Search Counter

Ian Armas Foster

Tracking and predicting human behavior is the goal of many a big data initiative, as companies look to judge the market climate in determining the next set of stock market transactions or the effectiveness of a proposed marketing campaign. New research out of London corroborates in a simple manner the ability of big data to predict behavior, as it found that if one were to place investments solely based on change of Google search terms over time, one would have made a profit of 326 percent between the years of 2004 and 2011.

“We were intrigued by the idea that stock market data serves as a really large record of all the actions people take in the stock market, but don’t necessarily tell us much about how people decided to take those actions,” said Suzy Moat of University College London, co-author of the paper. As a result, they looked at the volume of Google searches for terms like “stocks” and “revenue.”

“We wondered whether by looking at Google, we could get some insight into some early information-gathering stages of how people make decisions,” Moat continued. As it turned out, almost like clockwork, as the number of total searches for the word ‘stocks’ rose, the actual stock market would decline. If one derived a simple stock-buying strategy based on that information, the profits would represent 326 percent of one’s initial investment, as opposed to 16 percent if one simply bought a stock in 2004 and sold it in 2011.

“It makes sense and is in line with the scientific concept that there are more efforts to collect information before we see subsequent negative moves on an aggregated scale,” said Tobias Preis of Warwick University, another co-author of the paper, noting how the increase in results leading to a flooding of the market makes logical sense.

In the end, it boils down to collecting the Google metadata, an important but rather data-intensive task. “From a scientific point of view, it’s really excellent that… we have really got the technology – or the data based on technology – which makes it possible to look to some extent into early decision-making processes,” Preis said.

As a result of the research, the team has received a grant from the Engineering and Physical Sciences Research Council to cull this information into a software package that would take advantage of this information.

It would be interesting to see if the recognition of this trend across the financial services would render it insignificant, as sometimes when market quirks are found and exploited by enough people, they cease to exist. Either way, the ability to predict behavior by aggregating Google searches is a simple example of big data fueled predictive analytics.

Related Articles

When Big Data Wakes the Sleeping Giant

Giving Big Data a Sporting Chance

UCSD Peers Through Big Data Prism