Yves here. This post, in analytical value, falls into the “History never repeats itself but it rhymes” category. It looks at a book ban campaign in China to see its impact on publication and self-censorship, and looks at what happened when the ban was lifted. It does offer the hope that things can revert to an old normal. Readers can hopefully look past the dig at Chinese censorship, which seems gratuitous in light of the Twitter files, the current Trump efforts to denigrate DEI even if it falls short of formal censorship, and efforts in the US and EU to stomp out anti-Zionist speech.
By Ying Bai, Ruixue Jia, Associate Professor at the School of Global Policy and Strategy, Co-Director of the China Data Lab University of California, San Diego, and Jiaojiao Yang, PhD candidate in Economics Chinese University Of Hong Kong. Originally published at VoxEU
Censorship, a widespread practice with a centuries-long history, has left a lasting imprint on languages around the world. In China, the “burning of books and burying of scholars” in 200 BCE marked an early instance of suppressing knowledge. In Europe, the Index Librorum Prohibitorum, which was in effect from 1560 to 1966, served as a systematic effort to control what could be read and disseminated. Similarly, in the Soviet Union, the phenomenon of samizdat – self-publishing under censorship – became a means of resistance against state-imposed restrictions.
Understanding how censorship shapes knowledge production is critical for examining its broader political and economic implications. A bourgeoning literature on religious censorship in Europe highlights censorship’s detrimental effects – such as the setbacks experienced by banned authors and publishers (Becker et al. 2021, Blasutto and De la Croix 2023, Comino et al. 2024) – but important questions remain to be answered. How does censorship influence knowledge creation in general? What role does self-censorship play? If the power of censors wanes, could there be a permanent loss of knowledge due to decreased interest and availability in censored subjects, or might there be a revival?
In Bai et al. (2024), we explore these questions by studying the largest book-banning campaign in Chinese history, during the Qing dynasty’s compilation of the Siku Quanshu (Complete Library in Four Sections) from 1772 to 1783. Analysing over 161,000 book records spanning three centuries (1660s–1940s), we investigate the short-, medium-, and long-run effects of censorship on knowledge production and content. The significant political shifts over this period, including China’s forced opening to foreign powers in the 1840s, a major civil war from the 1850s–1860s, and the collapse of the Qing dynasty in 1911 provide a useful lens to investigate how knowledge producers – publishers and authors – adapted to shifting state power and control.
Historical Context and Research Methodology
The Complete Library, the most extensive book collection in Chinese history, catalogued over 13,000 books, including approximately 3,000 banned works. Heavily censored categories included history, imperial decrees and memorials, military strategy and conflicts, and various religions – topics seen as threatening to the Qing regime’s legitimacy. Unlike prior episodic censorship efforts, the Complete Library campaign institutionalised systematic, idea-based censorship, with ambiguous enforcement creating an atmosphere of fear and self-censorship.
This censorship practice shares notable parallels with censorship and policy implementation in contemporary China and other contexts. First, delegation played a key role: local officials – many of them motivated by career incentives – were tasked with confiscating banned books. Second, ambiguity was central to the process: vague guidelines created uncertainty while reporting mechanisms encouraged mistrust and fear. Finally, probabilistic but severe punishments amplified the chilling effects. While most banned books did not result in individual punishments, the cases of punishment enforcement were exceedingly harsh, including the execution of authors, publishers, and their family members. Together, these characteristics suggest significant chilling effects on knowledge production.
In Bai et al. (2024), we compile a comprehensive dataset of book publications drawn from the General Catalog of Pre-modern Chinese Books, which includes detailed information on publication years, authors, and publishers. We integrate records of banned books from official archives and historians’ research, leveraging the long-standing classification system that divided books into four sections (hence the name Complete Library in Four Sections) and 50 categories. To measure censorship, we calculate the share of banned books within each of the 50 categories relative to the total number of collected books (see category-level censorship in Figure 1). Employing a difference-in-differences approach, we analyse publication patterns and book content before, during, and after the Complete Library campaign, with yearly data between 1662 and 1949.
Figure 1 Censorship degree by categories
Note: Figure plots the level of censorship across 50 categories. For each category, the level is measured by the share of banned books among total collected books.
Suppression and Resilience in Knowledge Production
Figure 2 presents the event-study estimates on the impact of censorship, highlighting both suppression and resilience. From the 1770s to the 1830s, categories subjected to higher levels of censorship experienced a significant decline in publications, with a one standard deviation increase in censorship associated with an 18% reduction in book production.
However, the 1840s marked a turning point. Political upheavals, including the Opium Wars and the subsequent Taiping Rebellion, significantly weakened state control. We observe a resurgence in book production within previously censored categories followed, likely reflecting the diminished state control over society. We also find that the revival began earlier in treaty ports (which were forced to open) and their neighbouring regions compared to inland China.
Figure 2 Impact of censorship on books logged
Note: Figure plots the estimates of the effect of censorship on book publication every five years, using 1765–1772 as the reference period.
Book Contents and Chilling Effects
To explore how censorship influences book content, we analyse keywords in book titles. Figure 3 displays keywords from banned books alongside those from the Complete Library collection. The analysis reveals that the number of unique keywords within a category followed a pattern similar to the decline and revival observed in the number of book titles, indicating that these dynamics affected not only publication volumes but also the diversity of ideas.
By measuring a book title’s similarity to the two sets of keywords, we construct a proxy for the sensitivity of a book’s topic. Our findings indicate that both sensitive and less-sensitive books within restrictive categories experienced a decline and subsequent revival. However, since less-sensitive books constituted the majority of publications, chilling effects played a critical role in shaping the overall dynamics. Furthermore, the suppression and revival patterns extended to new keywords, highlighting the chilling effects on the generation of new ideas during the suppression period.
Figure 3 Keywords in banned books and Complete Library full-text books
Note: Left panel displays keywords found in banned books; right panel displays keywords from the Complete Library full-text books. Font size indicates word frequency. Red colours indicate words used in the section ‘histories’; blue colours indicate words used in the section ‘classics’.
Responses from Publishers and Authors in the Decline and Revival
By examining publishers and authors active across different periods, our research reveals that both groups adapted to censorship. Specifically, authors who died before 1772 (and could not respond to censorship), with books in more censored categories were less likely to be published during 1773–1839, reflecting publishers’ adjustments to censorship. Similarly, for publishers active after the 1840s (i.e. after the suppression effect dissipated), authors alive during the period 1773–1839 were less likely to have work in more restrictive categories, indicating authors’ adaptive responses.
Notably, the decline and revival in book publications can be attributed largely to the exit and entry of publishers. During the 1770s–1830s, publishers were more likely to exit from highly restricted categories. However, after the 1840s, new publishers began entering the market, revitalising previously suppressed fields. Using a Bartik-style instrument – which predicts the number of publishers in a category based on aggregate changes and the initial distribution of publishers across categories – our study confirms that these publisher dynamics can explain the observed decline and revival in book production.
Implications
Despite seven decades of suppression following a systematic censorship of ideas, this study reveals a remarkable revival in knowledge production following the loosening of state control. This finding both complements and diverges from the existing literature, which highlights the long-term detrimental effects of censorship through channels such as social and human capital (Xue 2021, Drelichman et al. 2021, Dewitte et al. 2024). While these channels may have been active during the suppression decades, our study suggests that publishers, authors, and readers adapt their behaviours in response to significant shifts in the political climate. This aligns with the broader notion that environmental changes play a critical role in shaping the balance between persistence and change (Giuliano and Nunn 2021).
While this resurgence is notable, it does not diminish the significant negative effects of censorship. From the 1770s to the 1830s – a period often characterised as one of intellectual stagnation – China experienced significant population growth but saw limited innovation and cultural development. This intellectual stagnation contrasts sharply with the contemporaneous flourishing of progressive ideas and technological advancements in Europe during the Industrial Revolution (Mokyr 2016, Almelhem et al. 2023). The suppression of knowledge in China during this critical period may have had important political and economic consequences, hindering the country’s ability to adapt to and engage with the transformative global changes of the era.
Furthermore, the importance of publishers in driving these patterns is noteworthy. It suggests the influence of intermediaries – such as publishers – in implementing and responding to censorship. This insight is still relevant in the digital age, when platforms and intermediaries continue to shape the flow of information and knowledge production.
Censorship has left a lasting imprint on languages and publishing practices around the world. This column analyses the impact of state censorship on knowledge production during the largest book-banning campaign in Chinese history, from 1772 to 1783. Categories subjected to stricter censorship – including history, conflicts, and religious studies – saw significant declines in publication following the bans, but political upheavals and the erosion of state control after 1840 triggered a resurgence. Publishers played a crucial role in both the suppression and subsequent revival of knowledge production.
Originally published at VoxEU
“Poem Theory”. No kidding. Not going to find that in the American version of that word salad. We get “Extra Ranch”.