A Critical Evaluation of Tracking Public Opinion with Social Media: A Case Study in Presidential Approval

Abstract

There has been much interest in using social media to track public opinion. We introduce a higher level of scrutiny to these types of analyses, specifically looking at the relationship between presidential approval and "Trump" tweets and developing a framework to interpret its strength. We use placebo analyses, performing the same analysis but with tweets assumed to be unrelated to presidential approval, to assess the relationship and conclude that the relationship is less strong than it might otherwise seem. Secondly, we suggest following users longitudinally, which enables us to find evidence of a political signal around the 2016 presidential election. For the goal of supplementing traditional surveys with social media data, our results are encouraging, but cautionary.