The Google Trends Primary

I am always fascinated by the results at Google Trends, Google’s own tracker of relative search activity. Entering in our top three candidates at the moment, we see that Barack Obama would win the primary in Googlestan hands down:


The letters on this graph correspond to news stories that came out the same day of the spike, much in the same way that Wall St. analysts attempt to ascribe troughs and valleys in stock prices to external news events. I’m not sure these spikes correspond to anything more than increased search activity caused by increased news coverage in general, and not to any specific news article, but the fact that the spikes occur on some very prominent days (Super Tuesday, for one) provides empirical evidence that the graphs are not a “random walk” but do in fact provide a mirror of the interest of some subset of the population.

What subset is, of course, an open question. I’d love to dig deeper, but Google frustratingly/wisely does not give us datapoints, or even the y-axis, so we can only look at these relative indicators of search activity and wonder. Recent data released by Hitwise suggests that Google users are more well-heeled than users of other search engines, an income group that has been a strength for Obama. Yet the same data also reveals that the groups that index highly for Google tend to be 55+, and definitely older than groups that index highly for Yahoo!, which suggests a pro-Clinton lean. Certainly, whatever the biases of the population of Google search users, Obama outperforms the real vote here in Googlestan (at least, by my guesstimation of the relative lift of these lines, anyway) and, in fact, did extraordinarily well in states that he actually won. Here are the top ten states for Obama search activity in February:


Does the fact that Obama “outsearches” Clinton in Texas and Ohio portend victory? Clearly we can’t tell from this–consider the margin of search activity in Missouri and the actual margin of the Missouri Democratic Primary, which was one percentage point. Consider also that even when we rank the states in order of search activity for Clinton, the past 30 days would indicate that the Senator from New York didn’t even come close in her home state, which she actually won quite handily:


So what conclusions can we draw from this? Is it a comment on the online population? Doubtful–since online access is nearly ubiquitous. Perhaps, given the inverse relationship between the “Search Performance” of Obama, Clinton and McCain it is merely a function of how much available information voters have about each candidate–in other words, an inverse function of how long each candidate has spent in the public eye. In any case, Google–in case you, well, Google this and read my question–how about giving us a y-axis for this chart!