Polarization in Elite Discourses of COVID-19 on Chinese Social Media

• In Event: From Pluralism to Polarization: Public Opinion and Policy Preferences

Abstract

The COVID-19 pandemic has struck human societies by amplifying deep-seated divisions along ideologies, classes, and races. The global spread of political and social polarization poses significant obstacles for combating coronavirus, casting long-lasting impacts on the post-pandemic world. Many concerns the role social media plays in exacerbating polarization – wild conspiracy theories, propaganda messages, racist and hate speech spread at a phenomenal rate across social networking platforms. The outburst of an “infodemic” has inflamed spirals of anger, shattered trust, damaged solidarity, and fostered unprecedented intolerance toward disagreements.
The concept of political polarization originated from and exemplifies in contexts of electoral democracies. It may seem counterintuitive yet evident, as we will show, that political and social polarization manifest in various contexts far beyond electoral democracies. Under a single-party rule, despite of tight control of media and predominant propaganda, elite discourses and public opinion still diverge significantly along latent ideological differences.
We probe into polarization in elite discourses on the COVID-19 pandemic on Chinese social media. Since the initial reporting of coronavirus in Wuhan in early 2019, mounting debates have erupted on Chinese social media. We acquired a novel dataset that contained the full amount of discussion on COVID-19 between January 21 and March 23 on Sina Weibo. We zoomed in four highly controversial issues that generated nationwide large-scale debate online: Li Wenliang’s death ,Whistle Blower, Wuhan Diary, and Diplomat Zhao Lijian’s “Wolf Warriors” speech. To focus on elite discourses, we focus on the weibo posted by verified active users in the four controversial issues.
To leverage the weibo to measure ideological position of each politically active user, we first use keywords to extract the weibo posts related to four ideological controversial issues. We then conduct latent Dirichlet allocation(LDA) to categorize the weibo posts into different topics. LDA is sensitive to the parameter settings, so we use Bayesian hyperparameter optimization to find the optimal parameters, including document-topic prior, topic-word prior, trimmed word frequency, and number of topics. We also carry out sentiment analysis to classify the emotion of weibo posts into positive, negative, and neutral sentiment. And then we construct a user-cell matrix, meaning each row corresponds to a user, and each column corresponds to a topic-sentiment combination. Item response theory model is used to evaluate the ideological position of each user. As a result, the model produces two principal components to captures each users’ position on the left-right ideology scale, with the first component predicting the users’ left orientation, and the second being the users’ right orientation.
To understand how the state induces polarization in cyberspace, we examine the causal relation based on the empirical test on both individual and national levels. First, we run a regression on the individual level. The dependent variable is the users’ ideology position, and the main explanatory variable is the user’s exposure to state-owned media. To eliminate confounder, we also control for users’ gender, province, frequency of posts, and influence of posts(measured by retweeted count). Second, we measure the ideological polarization by calculating the variance of users’ ideology score, and examine the influence of government action on polarization through propaganda with time series tests like Granger test.
Our primary finding is that polarization does exist in Chinese cyberspace during COVID-19. While some people do hold a position towards the Chinese government, other people firmly stand against the government. Unlike existing studies on polarization in democracies, we find that state-owned media and government reaction to controversial issues do influence both public opinion and polarization. Exposure to state propaganda could increase support for the government while widening the gap between left and right elites.
This paper aims to contribute to the scholarship of political polarization, particularly on social media, in three folds. First, it reveals how pandemic becomes a new source of polarization, which further perpetuated political and social divergence. Second, it offers understandings of polarization beyond conventional contexts of party competition. We reveal a left-right orientation underlying elite ideology that is not often conceived in authoritarian regimes. Even under a single-party rule, elite discourses and public opinion may still diverge significantly along ideologies. Last but not least, our analyses are based on rare uncensored data directly from Sina weibo, which provides an authentic picture of Chinese elite discourses about the COVID-19.

Authors

• Tianguang Meng, Tsinghua University

• Kaiping Zhang, Tsinghua University

• Siyao ZHENG, Tsinghua University