That’s what the folks over at Engage DC are saying. After analyzing Facebook likes from the “thousands of users of Trendsetter”, Engage DC used “predictive modeling” to tie together web browsing habits and political engagement.
At face value, the analysis is extremely interesting. The graphic matches our intuitive sense of what these websites and platforms represent. For example, Engage DC suggests that if you are a user of Spotify or Tumblr, you almost certainly will support President Obama. Conversely, if you play FarmVille or search the web with Bing, you’re more likely to support Mitt Romney.
I’m not even sure what Trendsetter is. But in the blog post posted on their website, Engage DC says that their users have been responding to polls about politics in the past few months. They use these polls in conjunction with Facebook data to come up with the above graph. Unfortunately, there’s nothing in that same blog post that describes that “predictive model” they use.
I suspect what’s happening here is a confounding relationship. I contemplated drawing a similar diagram that I show students in introductory statistics courses, but I realized that Engaged’s own graphic tells the story pretty well:
Both of these relationships (Internet activity & political preference; Internet activity & engagement) can probably be explained by age. Age can cause direct differences in Internet activity, political preferences, and political engagement. Without considering age, it seems like Internet activity directly affects political preferences and engagement.
And looking at the graphic, it’s immediately evident that age tells a huge part of this story. Younger people are more likely be play Angry Birds or Farmville, or use the blogging platform Tumblr, or the social networking site Reddit. On the other hand, older people are more likely to be users of Paypal or Ebay. The age distinction I drew out is not 100% perfect, but it calls for a more sophisticated analysis than the correlation they present here.