Master's Thesis Defense: "Measuring the Political Ideology of Social Media Posts"
Abstract: Though political speech is increasingly expressed on social media, this type of communication data is often difficult to quantify. A common measurement challenge in political science involves placing political speech along a one-dimensional scale of ideology. However, extant methods have been limited to utilizing the textual features of social media content. In this manuscript I develop a latent space measurement technique to represent the ideological content expressed in social media posts based on the network of social engagement meta-data attached to each post. The advantage of this approach is that it incorporates more than just textual content alone in the measurement process, including image, audio-visual, hyperlinked web data, and other non-text contextual features. I compare multiple data reduction techniques in terms of measurement accuracy and computation time in a simulation environment to determine the optimal data reduction method for this measurement technique. I apply the optimal method, singular value decomposition, to a set of Facebook posts authored by legislators elected to the 115th U.S. Congress to determine their ideological content. The application highlights that the measurement technique uniquely identifies the ideological content contained in legislators' Facebook communications, and enables future empirical scholarship on questions surrounding politicians' social media behavior.
Host: Jimin Ding