The 2016 US presidential election and its Chinese audience

Abstract

Motivated by the question of how the public in an authoritarian political environment may perceive democratic elections, we analyze the underlying interests and sentiments of the Chinese audience regarding the 2016 US Presidential Election with the social media data collected from a large Chinese online community. We extract several latent topics of interest to the community from the text corpus by applying the unsupervised learning method of Latent Dirichlet Allocation (LDA), and explore the amount of interests received by each topic by applying the supervised learning methods, including the Bayesian Additive Regression Trees and the Bayesian LASSO model. Results reveal much more attentions paid by the audience to the sensational news, especially the controversies related to Hillary Clinton’s email leakage and Donald Trump’s anti-political-correctness and anti-globalization remarks, than to the substantive issues, e.g., regarding the candidates’ policy agendas or the democratic process.