A tweet this morning caught our attention. Google's Rob Saliterman pointed out that Google search traffic for Oklahoma Republican primary candidate T.W. Shannon had risen quickly past that of his opponent in today's election, Rep. James Lankford.

Which made us wonder: Is there any predictive value in Google Trends data, which tracks that search interest? And after looking at seven close or interesting primary contests that have already taken place this year, we have an answer.

No. There isn't. At all.

Take, for example, the special election in Florida's 13th Congressional District in March. Republican David Jolly defeated Alex Sink by a narrow margin. But on Google Trends, Sink dominated. This graph will also serve as a guideline to the ones that follow. The blue bars indicate the eventual winner; the red ones, the loser, regardless of party. The first three sets of bars are Google Trends values (between 0, no significant search traffic, and 100, peak interest in the term) taken one week before the election, one day before, and on Election Day. The last set of bars is the final result, as a percentage.

So why isn't search traffic a good guide? A few reasons.

It depends on people caring about the race.

If people aren't very interested in the race, there won't be any search traffic to compare. Take the run-off in Texas between John Ratcliffe and Rep. Ralph Hall. Neither one generated much search activity, so no prediction could be made. In smaller races, this is a bigger problem.

Or, take Dave Brat's massive upset of Eric Cantor. Most people assumed Cantor would win easily, so it didn't generate any real Google activity. (Until the next day, when interest in the unknown Brat skyrocketed.)

There's no correlation between traffic and results.

This, of course, is the key factor. Here are three races, in which the Google Trends traffic favored, reading top to bottom: the eventual winner, no one, and the eventual loser.


(The use of "defeats" above is subjective. The two face off in a runoff today. But, McDaniel did get more votes than Cochran on June 3.)


Google rewards characters

You know who people like to Google? Interesting candidates. In Texas' Senate primary in March, there was no more interesting candidate than Steve Stockman who saw a surge in search interest as the election approached. This was partly because his campaign was baffling and sketchy. But Google didn't know that.

Sen. John Cornyn crushed him.

The moral of the story is this: There's simply no link just yet between search interest and actual votes. At least not yet. Caveat searcher.