A Presidential election provide a ready excuse for one of my favorite “didja know” chats with my students: how a lopsided poll (e.g., Hillary 4% ahead of Trump) might even show that Trump is a bit ahead. Why? A simple reason: margin of error. It could decrease Hillary’s percentage by, say, 3 points, and increase Trump’s by 2, and the scenario is suddenly quite a bit different!
Let’s put that mini-TED talk aside for now. It shakes up our entire faith in pundits, and their predictions. What if I told you (and my students) that there’s a great way, in addition to polling, to predict an election, and it involves something we all do probably once a day – Googling.
A UCLA professor, Stuart Gabriel, and an economist, Steven Davidowitz, believe that Googling candidates may allow us to predict elections, mainly because people are loathe to reveal for whom they will vote, especially when two relatively unpopular candidates are involved (e.g., the 2016 variety of candidates).
Furthermore, although some elections could have been predicted based on the highest number of searches, the search order “could prove a meaningful supplement to polling data,” writes Gabrial and Davidowitz. “This indicator could contain information that polls miss because voters are either lying to themselves or uncomfortable revealing their true preferences to pollsters.”
It turns out that 12 percent of search queries with ”Trump” also include the word “Clinton.” More than a quarter of Clinton searches also include Trump. It is the order that appears to be most significant. The authors found that in the last three elections, search order predicted the electoral outcome state by state. And one does not need to subtract, or even add, margin of error.
Source article: Gabriel, S., & Davidowitz, S. ((2016, October 23). If They Google You, Do You Win? New York Times, Week in Review, 2.