The Micro-foundations of Authoritarian Rule Unveiled by Digital Traces: New Theories and Methods with Applications to Chinese Social Media

How do citizens in authoritarian China talk about politics with one another in the social media era? Political talk is a social activity in authoritarian regimes as much as it is in democracies. However, the scholarship has so far dominantly focused on state-citizen interactions in political communication in authoritarian regimes, overlooking the dynamics of citizen-citizen interaction. In this dissertation, I present a new theory, original data, and two novel methods to understand social media political talk in authoritarian China. I argue that citizens in the social media era are engaged in a new form of preference falsification: expressing truthful political opinions to strangers outside their network while lying to “friends” in their network. I theorize that the behavior is attributable to a combination of psychological rewards for being truthful and social punishment for being a dissident. A consequence of the behavior, I posit, is discouragement of collective action, which stabilizes authoritarian rules. I test the theory with an original dataset I collected from Chinese social media. In addition, I develop two novel methods to analyze big social media data of political communication in authoritarian China and beyond. I develop ATIOS, a system based on distributed semantics that generates valid and replicable text-as-data measurement. I introduce Bayesian Dynamic Network Modeling, a method that efficiently models time series of social media networks. With this dissertation, I contribute new theories and methods for the study of contemporary Chinese politics and comparative political behavior.