I recently wondered if Internet searches for companies correlate with the company’s stock price. A few years ago a researcher at Google was able to track flu trends by monitoring online searches for flu symptoms. Would searches translate to stock price as well? To answer the question, I used Google’s “trends” utility and Google’s finance utility to extract data on a handful of companies. It’d be nice to use a larger set, but Google shut down its financial data API late last year and disconnects any robots it detects. So here’s the results on the 12 companies for which I hand-downloaded the data.

The Good News

Some companies seem to be VERY closely tracked by Internet searches. Google’s own stock stands out as a stellar example. The x-axis shows the weeks from May 2004 until May 2013 — 9 years of results. The correlation of these two lines was 0.823!

A few other close matches:

Amazon’s searches and stock price had a correlation of 0.875 — even better than Google. It’s nifty to see the seasonal surges in the search results. The peaks of the surges also seem to show temporary dips in the stock value — perhaps investors selling off their stock when the seasons hit their peaks?

Although the stock has only been on the market since the end of 2011, “LinkedIn” had a correlation in stock price and search frequency of 0.891.

 The OK

Then there were stock prices that only sort-of followed their search results.

Apple had a correlation of 0.727. The huge surge in stock price in early 2012 threw off the correlation since the number of Internet searches was steady (minus seasonal cycles). However, look closely at the chart and you will see an interesting fact — the stock price eventually floated down from the sky and rejoined the search level. Perhaps the correct way to look at it is the stock price tracks search levels rather than the other way around.

Facebook was low — however, the data is very brief since the stock came out not long ago. Looking at the chart I see a leap in searches right around the big IPO — and a corresponding dip in the stock price. Otherwise they look like they might equalize soon.

GE had a correlation of 0.636. The big dip occurred during the 2008 recession, although searches waned much more slowly. Again, it looks like the two will be equalizing soon.

The Anti-Good

Then, there were stocks that had negative correlation. The most strongly anti-correlated stocks (of the 12 I looked at) were Yahoo! (-0.887) and Bank of America (-0.779). Interesting… Since Internet searches are still strong on these terms is it possible these stocks are undervalued and destined to recover their value?

Yahoo!’s chart:

and Bank of America:

Coming in close behind these companies is RIM with a correlation of -0.565. I actually searched for “Blackberry” since the Blackberry is the name of their flagship (and only) product.

The Boring

The remaining three companies, Microsoft, AT&T and Chase, showed a correlation close to 0.

Microsoft’s correlation of -0.0947:

AT&T’s correlation of 0.0518:

Chase’s correlation of 0.0168:

Conclusions

My optimistic conclusion on this mini-project is that the search frequency tracks what the stock price “could be”. Under that interpretation, Yahoo!, Bank of America and RIM would be definite “buys” (as of today, 5/17/2013). But I’m in no way an officially trained investment trader, so don’t take anything written in this blog post too seriously! It will be interesting to follow up on my predictions a year from now to see how well my conclusions fared.

Does anyone else have an idea for interesting stocks to compare?

UPDATE: After doing this post and then googling around to see what others have done, it turns out there’s a researcher who has used Internet searches to try to predict the whole market (the total Dow Jones index value, etc.). They’ve had some success, but were unable to predict actual stock prices using their methods. I’m guessing that’s because individual stock prices do not necessarily reflect anything rational. My approach was simpler in that I only searched for company name (or product) and nothing else and calculated the correlation with individual stocks rather than the market as a whole.

References

The website for google trends: http://www.google.com/trends/

The website for google’s stock data: https://www.google.com/finance