Since I suffer from a case of chronic, debilitating financial nerdliness, imagine my overwhelming joy when I discovered Google Lab’s new* goodie: Google Correlate. That’s right, the same people who brought you such wonders as Google Docs (providers of the spreadsheets I post on this blog) and Google Flu Trends have brought us a simple tool that lets you find Google search terms that correlate with your time series or state location. They even put together a cute comic that explains how Google Correlate works.
So what do I do when I find something like this? Play with it! Check this out:
First, I go and get a time series of data. In this case I will model the S&P 500** by using the price data for SPDR S&P 500 (Ticker: SPY), an ETF*** that is designed to closely follow the S&P 500 Index. I mosey on over to the finance.yahoo.com for SPY and get myself a time series of data by fiddling with the historical prices for SPY. I make the start date 1/1/2003 and the end date 5/23/2011. Then I set the interval for weekly. When I do this Yahoo! gives me this page:
As you can see from my circle, I can’t just copy-paste the data that comes from finance.yahoo.com into my time series. Its got all the dividend brakes that screw up the data. So I scroll to the bottom of the page, and lo, behold, I can output to a spreadsheet. I do so.
And this is what I get:
Boom, done. Now I just have to clean up the data. (You can see that the volume is listed in scientific notation. Excel does this when the cells are too small to display the whole number. If you make the column wider, the notation goes away.) I make another sheet and copy-paste the dates and prices such thusly:
And now we’re ready to shove this into Google Correlate. I go over to Google Correlate, sign in with my Google User ID and choose “Enter your own data,” then choose “Time Series.” I get an input page that looks like a spreadsheet and simply copy-paste my data into the Google Docs spreadsheet. I name my data set “SPY Price.”
I then choose “Search Correlations,” and, in a few moments, this is what I get. First, a list of search terms that are most correlated with my data set.
And, below that, a chart showing the correlation of the most highly correlated search term.
Isn’t this delightful! Before I ran this, I had no idea what “stockhouse” could possibly mean. Well, as it turns out, Stockhouse.com is a website that features scads of conversation between active investors. Isn’t it interesting that the frequency of the search term “stockhouse” follows along so well with the price of the SPDR S&P 500? Some of the correlations also seem to make superficial sense, like “401k calculator” or “investment calculator.” But, “dhl track?” “Biagio’s restaurant?” I couldn’t speculate wildly as to what association, if any, the terms would have with SPDR’s price. Maybe the correlation is just luck.
* It appears to have been released just yesterday!
** An index of the 500 largest stocks on the U.S. market. Google it.
*** Exchange Traded Fund. Google it.