Can aggregated data
from Google predict the stock market? Research from three economists says yes to Big Data. The three researchers have published a paper indicating that data from Google Trends can predict up and down price moves on the Dow Jones Industrial Average.
The researchers had previously discovered that the number of searches on a company's name, and the number of times that company's stock was traded were related, but it couldn't determine pricing -- meaning it couldn't help investors.
Frequency of Search Terms
So the team took a broader view of the stock market. They tracked 98 finance-related search words or phrases, such as "debt" or "derivatives," using data from 2004 to 2011. The frequency of those searches was compared with the DJIA's closing price, which is based on 30 stocks.
The researchers employed an investing game, using imaginary money, to test the concept. If financial searches went down, thus apparently indicating less interest in the market and lower prices, they "bought" stocks to hold them for the long term. If searches rose, possibly indicating an increase in market interest, they shorted the market. In short sales, borrowed stocks are sold and bought back later at what is hoped will be a lower price because the market has gone down.
Shorting assumes the market will fall, and the researchers' speculated that Google searches increased because stock owners were getting nervous about the market, and getting ready to sell their stock. In fact, that seemed to be the case -- an increase in searching about financial terms preceded a fall in prices.
'Debt' Is the Key
The key term: "debt." When searches for that term dropped, they went long; when "debt" searches increased, they shorted. Using this strategy, they increased the value of their imaginary holdings by 326 percent, compared with a 16 percent return for just buying-and-holding stocks.
The economists are Tobias Preis of Warwick Business School in the U.K., Helen Susannah Moat of University College London, and H. Eugene Stanley of Boston University. Their research is being reported in the journal Scientific Reports.
What if everyone is reading this and other articles on their research? The correlation could evaporate, or perhaps some market conditions respond more exactly to other search terms. But expect more of this kind of finding-the-real-meaning in Big Data for investing, as researchers did deeper into other Google Trends patterns, or large numbers of Wikipedia searches, or social media chatter.
Searching trends in Google's search engine are also being examined for other possible correlations.
When 6 million Net-surfing volunteers agreed to have their searching patterns tracked, for instance, an increase in searches for hyperglycemia (high blood sugar) by those who took both the antidepressant Paxil and the statin Pravachol led medical researchers to the discovery there was such a side effect. In separate research, flu-related searches, in conjunction with data from the Center for Disease Control and other sources, are similarly being investigated for their ability to predict flu outbreaks.