Google Correlate is a fascinating new web-tool that allows you not only to compare what web-search terms are correlated with each other temporally and geographically (see here for a nice example) but also to examine whether search terms correlate with your own data. The inspiration is the well-documented application created by Google that accurately forecasts flu outbreaks. The people at Google rightfully thought that the general idea ought to be applicable to other areas, so they created a new web-tool. This comic book explains the ideas well.
One obvious thought was to see what search terms correlate with public opinion trends. John Sides sent me some data on support for Obama from the 2008 presidential race and it took me about 1 minute to feed it into Google Correlate. So far so good. Now the results (note: I updated the picture):
Oh well. There are so many obvious jokes here that it is hard to know where to begin (my favorite involves, of course, “Prince Eric”). Now, it could be that you would find more reasonable things with a longer time series or that it works better with more substantive public opinion trends. For example, I tried some data I had on support for the war in Iraq (from this article) and I got “Saddam cartoons” as the first and only correlated search term (although the correlation was modest (r=0.61)). That is at least related but it does not help if we were interested in, say, predicting whether there are signs that a public is getting fed up with a war. If you have better experiences with different data, please report.