“Super Tuesday” has come and gone with no clear winner on the Democratic side of the ledger. Hillary Clinton apparently won more delegates, but Barack Obama won more states and stayed very close to Clinton in the total delegate count. On the Republican side, John McCain scored a big, but nowhere near decisive, victory.
Our question for today is how predictive the GoogleTrends database was of political success on the biggest day in the history of the American presidential nominating process?
In the Democratic race, the GoogleTrends data show evidence for the “surge” in interest in Obama that the American media reported heavily before Tuesday, but the search request data certainly were not predictive. If the national graph were correct, then Obama would have emerged as a clear winner. If we look at particular states like Georgia, New York, or Massachusetts, the search requests more closely (but definitely not exactly) mirror the results of the local primaries. It could also be, of course, that Clinton is searched fewer times because voters in those states already feel as though they know her and her record, while Obama remains more of an unknown quantity–and so needs to be “Googled” more often.
On the Republican side, the national data seem to mirror the results of the primaries somewhat better if Ron Paul is abstracted from the graph. Paul’s candidacy lives much more on the Internet than in the analog world and so if he is added in, then it appears that he is a much more viable candidate than he is in reality. Looking at the same states–Georgia, New York, and Massachusetts–we again see that the search results more closely mirror the outcomes of the voting…but still not completely. Mitt Romney won the “search primary” in Georgia, but Mike Huckabee won the actual primary. And John McCain won the “search primary” in Massachusetts while Romney won the actual one.
So now we know that the search primary doesn’t predict actual voting. But as an educator, I intend to use these data to get my students thinking about ways that digital media might be subverting conventional thinking about things as important as the presidential nominating process. How can data such as these help us to understand voter choice? How can they mislead us?
Very interesting concept.
One variable that you don’t mention, and could potentially make a difference, is how many of those people who did a search actually voted. Also, not taken into account is how many times the same person did the search. (I’m not sure if Google takes that into account or not.)
As you say, it’s not a viable predictor for presidential elections, but is still an interesting comparison to make. I don’t doubt that the future will push this concept a little closer to reality. As more and more of the younger generation come into the voting age, technology will surely affect their views.