The New Machine Politics: Identifying Partisans Through Their Browsing History

Abstract

Political partisanship has been the driving force in American politics virtually since the birth of the nation; being able to detect partisanship has profound consequences for both academics and political professionals. I use a dataset of browsing histories from comScore in order to predict the partisan identification of a computer user. Using both random forest models and a variation on Sweeney’s exclusivity indices, I match computer users to voter files in both North Carolina and Florida, demonstrating that both the race and partisan identification of an individual can be determined from his, her, or their browsing history.