Information Control and Echo Chamber: Evidence from Millions of Book Copurchases

• In Event: Media Strategies and Online Censorship in China and Russia

Abstract

China observers have long noticed that the Chinese state keeps a tight grip on academia and the press. Publication censorship is pervasive. Scholars and journalists are deferred from producing unwelcome content by self-censorship. Meanwhile, institutional incentives are given to induce research aligned with the state’s agenda and interests, such as the ‘Belt and Road Initiative’, or Xi Jinping’s thoughts. Some ‘disciplines’ are severely ‘flooded’ by propaganda cloaked in ‘research’, such as Scientific Socialism, Party History and Party Building, which are independent disciplines in mainland China but not the case in most other countries.

The real-world effects of such a strategic combination of censorship and propaganda, however, are inadequately investigated. Previous works mostly interrogate censorship and propaganda separately, and empirical tests mainly rely on experiments, resulting in the following research gaps: 1) external validity could be problematic, especially when subjects are aware that treatments are about propaganda or censorship; 2) censorship and propaganda take place every day in the real world, yet the existing literature does not explore diverse data sources from information control in practice; 3) censorship and propaganda are two sides of one coin, and their effects are interwoven instead of independent, therefore should be examined together.

Draw on an original dataset that contains millions of book co-purchase records, this paper examines whether information control, i.e., censorship and propaganda, leads to an “echo chamber” effect among ordinary Chinese Internet users. I hypothesize that, because of variation in information control, publications of different disciplines carry different amounts of politically sensitive information and propaganda. If citizens consume and are persuaded by these books, they are more likely to purchase books with similar tendencies. I would argue that the manipulation on the “upstream” of information production, by coercing or co-opting a relatively small group of intellectual elites, affect the “downstream” of information consumption of the general public, thereby boosts the state’s resilience. Starting from two seed books representing contrary political preferences on Dangdang.net, a Chinese book-selling site, I download all the books on the list of “customers who purchase this book also purchase”, iterating this process until all related books are contained. This ends up with the biggest strongly connected component in Dangdang’s co-purchase network. I use a Bayesian model to estimate political alignment and polarization scores of different disciplines. Leveraging on cutting-edge techniques and real-world behavior data, this paper contributes to the long-lasting puzzle that why the Chinese authoritarian state goes hand in hand with a booming Internet.

Author

• Yan Gu, Airbnb