How Google can tell you when to buy or sell stocks

Researchers from the U.S. and Britain suggest keeping track of what people are searching for on Google and investing accordingly can significantly increase returns.

In recently published research in Scientific Reports, Tobias Preis of the University of Warwick and Helen Susannah Most and H. Eugene Stanley of the University of Boston retroactively looked at the correlation between what terms people were looking up on search engines like Google, and the performance of major stock markets following spikes and troughs.

The base of the research was compiled from Google Trends, an analytical tool that tracks what search terms are most popular over time. The trio then cross-referenced that data with stock market performance, and found some interesting relationships.

Key words matter

Broadly, when searches for terms like "debt," "inflation," "unemployment," "crisis" and "short sell" peak, stock markets tend to do poorly in the immediate aftermath. When search volumes for those terms decline, stock markets tend to rally.

"We can learn from Google that, at least historically, search behaviour of people can be linked to subsequent stock market moves," Preis said in an interview with CBC News.

Indeed, the word "debt" proved to be one of the best predictors of the stock market's direction. The researchers chose the Dow Jones Industrial Average as their proxy for the stock market overall. And they took the Dow's performance between January 2004 and February 2011 as the base.

Not including dividends, a simple buy-and-hold strategy over that time frame would have yielded a total return of 16 per cent. But by retroactively watching what people were searching for at different stages and investing accordingly, the research team was able to more than quadruple their hypothetical money.

Whenever searches for the term "debt" spike, the researchers went back and sold or shorted the Dow for the next week of trading. "If in this week, there are more surges for the term debt compared to previous weeks, then we sell the market in the coming week," as Preis puts it. Whenever there was a decline in the number of "debt" searches, they went back and bought.

Their research suggests that trading strategy would have been enormously successful. Instead of a 16 per cent gain, selling on "debt" spikes and buying on troughs yielded a return of 326 per cent over the test period — although those figures don't include trading fees, and the impact of dividends.

'The challenge for an investor is to identify terms that are important for the market' —Tobias Preis

Other terms also hinted at strong correlation to the market, but none were quite as pronounced as "debt" searches.

The study is similar to other data, also based on Google searches, that's helping medical researchers predict influenza outbreaks by monitoring where and when people are searching online for flu-related terms. And other studies that suggest the number of clicks on search results stemming from any given country correlates with the amount of investment in that country.

While the results are eye-opening, he acknowledges that the conclusions are far from fool-proof. Even with a tool as broad as Google, the system can still be gamed to pre-select certain results — which is why the researchers looked at 98 terms overall.

Even the mere publishing of the report may encourage copycat strategies, Preis notes, which means gauging the real sentiment of debt-related searches will be more and more diluted over time

The tricky bit, Preis says, is figuring out what words best have their finger on the pulse of investor sentiment at any given time. As he puts it, "The challenge for an investor is to identify terms that are important for the market."