Self-Censorship under Social Pressure: Evidence from an Online Q&A Forum in Authoritarian China

In this paper, I theorize and test with original online discourse data how self-censorship is incentivized by social pressure in authoritarian China. Self-censorship is the behavior of withholding expressions of genuine political views from others. It is considered a prevalent political behavior with implications for authoritarian survival. Existing studies primarily treat self-censorship as a precaution against sanctions by the state. I contend that this conventional wisdom has overlooked the fact that everyday political conversations are mostly social activities in authoritarian China, where the intended audience at stake is more likely fellow citizens than the state. I theorize that citizens under authoritarian rule self-censor when their fear of reputational sanctions outweigh intrinsic and expressive utilities of expression. I test the theory with original political discourse posted from 2010 to 2016 on Zhihu, a popular question-and-answer forum in China with an active community for political discussion. Zhihu provides a unique “anonymity” option that hides authors’ virtual identities from other users but not from the state. Using anonymous answers as proxies for unobservable uncensored political expressions, I find users self-censor close to 20% of their political discussion. Notably, they self-censor discussion about sensitive domestic political issues and that with a negative sentiment. Additionally, users self-censor to avoid explicitly taking a position, engaging in unfriendly debates, and exposing personal information. The empirical study employs unsupervised and supervised text-as-data methods, including embedded topic modeling, sentiment analysis, manual text content analysis, and variable selection methods. This paper contributes to our understanding of authoritarian mass political behavior and methods for opinion research under authoritarian rule.


Haohan Chen is a computational social scientist specializing in political behavior and methods for text and network data. He is currently a postdoctoral fellow at the Center for Social Media and Politics at New York University. He obtained his Bachelor of Social Sciences with first-class honors from the University of Hong Kong in 2013. He earned his MS in Statistical Science and Ph.D. in Political Science from Duke University in 2019. His substantive research focuses on political behavior and political communication in both authoritarian and democratic contexts. His methodological research focuses on machine learning methods, particularly Bayesian statistical learning, for text and network data from social media. His dissertation project studies preference falsification and self-censorship under social pressure in authoritarian China using original online discourse data from Zhihu, one of China’s largest social networking sites. His thesis in statistical science develops a Bayesian dynamic model for streaming network data. In his current position at NYU, he studies the dynamics of political polarization and misinformation in the United States by applying deep neural network models to rich discourse data from large streaming collections of tweets. His research has appeared in the Proceedings of the National Academy of Sciences. He won the APSA Political Communication Section Paul Lazarsfeld Best Paper Award in 2019 and the BEST Award for Master’s Research at Duke Statistical Science in 2020.

Speaker(s) Dr Haohan CHEN
Postdoc fellow, Center for Social Media and Politics, New York University
Date 29 Jan 2021 (Friday)
Time 11:00 am
Venue Online Via Zoom (link will be sent via email)

Please use your HKUST login to join the zoom meeting.
This meeting is being recorded. By joining, you are giving consent for this meeting to be recorded.

Hong Kong University of Science and Technology
Home  |  HKUST  |  The School  |  Jobs  |  Site Map  |  Contact Us