Key Takeaways
Treat censorship as a price on information, not just a locked door
Access is not exposure. China generates far more online material than any authority could inspect or remove. Yet censorship can still work without sealing every route. In 2010, when Google resisted filtering its search results, Chinese authorities slowed access rather than immediately banning it. People could still use Google, but the delay and inconvenience pushed many toward alternatives. The book calls this porous censorship: controls that make information harder to find or share while leaving some way around them.
Small costs redirect attention. A page that loads slowly, a result buried in search, or a post removed before it spreads can change what people actually encounter. The key distinction is between information that exists somewhere and information that reaches people who have little time or reason to search for it.
The argument shifts attention from censorship as a yes-or-no rule to the design of the information pathway. That is a useful move because access has several stages: creation, distribution, discovery, and attention. A message can survive technically while becoming irrelevant in practice. This resembles choice architecture in behavioral economics, where defaults and small frictions shape decisions without forbidding alternatives. The distinction also sharpens measurement: counting available websites says little about what people see. Still, inconvenience alone does not establish political intent. A slow service may be an ordinary technical failure, so evidence about who imposed the cost and why matters.
Diagnose censorship by asking whether it threatens, obstructs, or overwhelms
Three routes to control. Roberts defines censorship as an authority's deliberate restriction of public expression or access to information that could make it accountable. She distinguishes three mechanisms. Fear uses credible punishment to deter speech or reading. Friction raises the time, money, or effort needed to find or distribute information. Flooding promotes competing material until important information is harder to notice. The same intervention can use more than one mechanism: removing an activist's account may obstruct reporting and warn others of possible punishment.
Visibility changes the risks. Fear must be noticed to deter, but visible punishment can provoke anger or reveal that the authorities feel threatened. Friction and flooding can work without being recognized as political interventions, which makes them harder to challenge. Their weakness is that motivated people can often get around them.
The typology is valuable because it separates the state's intended effect from the tool it uses. A threat targets expected consequences, a blocked route changes the cost of access, and coordinated distraction competes for attention. Those distinctions help explain why identical outcomes, such as fewer people reading a story, can arise through different political strategies. The framework also connects censorship to organizational design: authorities weigh effectiveness against backlash, economic cost, and lost information about public opinion. Its categories are best treated as diagnostic rather than airtight. In practice, a platform block, a warning, and a burst of propaganda can operate together.
Build for tired citizens, not imaginary investigators with unlimited time
Attention is scarce. The book calls citizens rationally ignorant when they choose not to spend much time learning about politics because the personal benefit is usually small. People select information by weighing its likely usefulness or entertainment against the time and money needed to find it. They also rely on trusted sources and cues, such as prominent placement or other people discussing a story, because they cannot verify everything themselves.
Small delays matter online. Roberts notes research cited in the book finding that a one-second slowdown in search can reduce search volume by about one percent. Social media also lets people pass along information without traditional newsrooms, but much of this material is low-cost opinion, entertainment, or repetition rather than carefully reported investigation. When a story becomes harder to find, most people can simply move to easier content.
This account treats limited attention as a political resource, not merely a personal failing. The same constraint shapes shopping, health decisions, and news consumption: people use shortcuts because exhaustive evaluation is impossible. That makes interface choices and information ranking consequential even when no one is explicitly silenced. It also cautions against assuming that more content produces better-informed citizens. But low political attention is not fixed or evenly distributed. A nearby threat, a compelling personal stake, or a trusted social network can make people search much harder. The argument is strongest when it explains those shifts, not when it treats the public as uniformly inattentive.
Visible censorship can advertise the very story it aims to bury
A ban can create curiosity. In a matched comparison of 87 pairs of Weibo users who posted identical messages, those whose posts were removed were at least as likely to keep writing about the subject as their uncensored matches. Censored users were also twice as likely to complain about censorship in the days after the removal. In a survey, 36 percent of users who had experienced censorship said they would feel angry, while only 7 percent said they would feel worried or afraid.
Readers also pushed back. In an experiment, students who encountered a removed social media link were more likely to click on another post about the same subject and expressed less support for Internet regulation. The evidence suggests that obvious censorship can signal that a topic matters, undermine trust, and invite resistance rather than silence.
The results illustrate a classic signaling problem: an authority that marks a subject as dangerous may inadvertently tell people where to look. This is particularly plausible online, where links, replies, and shared posts make it easy to investigate a missing item. The combination of behavioral records, survey responses, and an experiment is a strength because each method captures a different response. Scope still matters. The experiment involved university students, while the matched users were prominent enough to appear in a monitored dataset. Neither group represents every citizen, and fear may still deter people who never post or who face greater personal risk.
China traded Mao-era total control for growth-compatible information management
Fear once reached daily life. During Mao's rule, political control shaped workplaces, schools, media, and private relationships. Citizens could face severe consequences for ordinary remarks, and the press functioned as a Party mouthpiece. The book links this atmosphere to failures in governance as well as repression: during the Great Leap Forward, local officials had incentives to report inflated grain production, helping conceal the scale of disaster.
Reform changed the calculation. After 1978, leaders loosened control over private life and commercialized parts of the media to support economic growth. The 1989 Tiananmen crackdown renewed political control, but a return to total surveillance of everyday life would have conflicted with the state's economic ambitions. Instead, authorities managed media gatekeepers, filtered online content, and promoted approved narratives while leaving many routes to information technically open.
The historical arc shows censorship as an adaptive governing strategy, not a single unchanging institution. A state that depends on growth needs information, trade, and experimentation, yet also wants to constrain challenges to its authority. Less visible controls can help balance those goals. The Great Leap Forward example also reveals a deeper danger of fear: when officials conceal bad news, leaders lose the feedback needed to correct policy. This is a general principal-agent problem, since subordinates protect themselves by distorting reports upward. The shift described here should not be mistaken for a smooth liberalization. The book presents selective control as compatible with, and useful to, continuing authoritarian rule.
A few extra steps can separate political elites from the wider public
The Firewall filters by effort. In a 2015 survey of urban residents, about 5 percent said they had used a VPN to bypass China's Great Firewall. Among people who knew bypassing was possible and used the Internet, most nonusers cited having no reason, not knowing how, or finding it bothersome. Those who did evade the Firewall were more likely to be educated, wealthier, politically engaged, and connected internationally. About 75 percent had attended college or graduated, compared with 25 percent of nonusers.
Friction also changes event coverage. Across 120 Tibetan self-immolation events, social media discussion lasted longer when an event occurred on a weekend, when fewer censors were working. The finding suggests that even a modest delay in removing posts can affect how far information spreads. Blocking or slowing a site reduces visits even when circumvention remains possible.
The political consequence is not simply that some people know more than others. The barrier can weaken the connection between informed, dissatisfied users and the broader audience needed for collective action. That turns an access gap into a coordination gap. The pattern also resembles a digital divide shaped by education, income, and technical skill, except that the divide is partly produced by policy. There are limits to the evidence: the survey covers urban residents, and weekend timing is an indirect measure of censorship speed. Nor does separation guarantee that elites cannot reach wider audiences through other channels.
Flood the feed with coordinated cheerleading, not just rebuttals
Distraction can do the work of suppression. Flooding is the coordinated promotion of information that competes with or distracts from material an authority would rather people not notice. After a deadly Yunnan earthquake in 2014, Chinese state media rapidly published coordinated coverage of an old celebrity scandal involving Guo Meimei. A leaked directive told websites to prominently display the story and organize commentary, prompting suspicions that the timing diverted attention from the disaster.
Much propaganda avoids argument. In a sample of 200 posts from a leaked archive linked to government online commentators, 85 percent were cheerleading, patriotic messages, or encouragement, rather than direct attacks on critics. Coordinated newspaper stories also spread online. Posts connecting China's Qingming remembrance holiday with revolutionary martyrs became six times more common from 2012 to 2014-2016.
Flooding highlights how repetition and convenience can shape attention without persuading anyone of a specific claim. Familiar material becomes easier to encounter, share, and recognize, while competing stories must fight for limited space. This helps explain why propaganda can function as censorship even when it adds content rather than deleting it. The Yunnan episode is vivid, but timing alone cannot prove that every reader was diverted or that the campaign caused a particular belief. Exposure, sharing, persuasion, and distraction are separate outcomes. The book's strongest contribution is showing that coordinated content can alter the information mix, not demonstrating that it reliably changes private convictions.
Expect porous controls to weaken when crises make information urgent
Urgency changes the price people will pay. When citizens have little reason to seek political news, small access barriers can be enough to redirect them. During a crisis, the information may suddenly matter to their safety, finances, or understanding of events, making them more willing to overcome those barriers.
Sudden blocks can backfire. When Instagram was blocked in mainland China during Hong Kong's 2014 protests, the VPN Express app jumped overnight from rank 1,229 among downloaded apps to number six. In Tianjin, the share of surveyed Internet users reporting that they had bypassed the Firewall rose from about 5 percent before the 2015 explosion to nearly 30 percent afterward. That survey comparison cannot by itself prove the explosion caused the increase, but it is consistent with the argument that urgent events prompt information seeking.
A crisis changes both motivation and interpretation. A blocked page is more likely to look politically meaningful when people already know something important is happening, and a disrupted service can push habitual users to learn circumvention tools. This is an example of demand becoming less sensitive to cost when information has immediate value. Yet crisis attention is not automatically liberating. Rumors can spread quickly, official messages may dominate, and fear may deter particular groups. The practical implication is not that porous controls collapse in emergencies, but that their effectiveness becomes less predictable precisely when authorities may most want to control the narrative.
Protect democratic speech by scrutinizing who controls access and ranking
Censorship extends beyond bans. Roberts argues that political control can affect not only who speaks, but also which information is collected, distributed, and made easy to find. In democracies, government agencies can make data difficult to obtain, while companies and interest groups can influence rankings or flood the public sphere with their own accounts. A costly public-records request, for example, can discourage reporting even when disclosure is legally permitted.
Look at the gatekeepers. The book points to search rankings, Internet speed, government data practices, and online propaganda as areas with political consequences. Safeguards could include stronger transparency requirements and competition among information distributors, so that no single actor can quietly determine what reaches the public. The claim is not that every editorial decision is censorship, but that distribution deserves scrutiny alongside formal speech rights.
The extension to democracies is important because formal permission to speak does not guarantee practical access to an audience. This echoes debates in media economics and platform governance: intermediaries can shape public visibility through ranking, moderation, and infrastructure. Transparency and competition may limit that power, but neither is a complete solution. Publishing every ranking rule can invite manipulation, while competition may still leave users dependent on a few dominant platforms. The concept also needs a demanding standard for political intent. Without one, ordinary editorial judgment or accidental delay could be mislabeled censorship, weakening the term's analytical usefulness.
Analysis
The book's central contribution is to treat censorship as a problem of distribution and attention, not only prohibition. That change in emphasis explains why a porous system can be politically effective: authorities do not need to prevent every motivated reader from finding a story if they can make the story less likely to reach the majority. The framework also explains why distinct tools coexist. Visible threats can target influential speakers, while less visible obstacles and coordinated distractions shape the information environment at scale.
The empirical design is unusually varied for a study of an activity meant to remain hidden. Roberts combines matched social media users, an online experiment, a survey, changes in censorship timing, website traffic, newspaper coordination, and leaked propaganda records. This triangulation makes the overall case more persuasive than any single example. Still, the evidence varies in causal strength. The experiment offers a clearer test of reactions to visible censorship, while the weekend comparison and some crisis comparisons rely on indirect indicators. The population of urban Internet users and monitored social media accounts also differs from citizens who are less connected or more vulnerable to punishment.
A deeper implication concerns political equality. Porous censorship does not simply reduce information for everyone; it sorts people by motivation, skills, resources, and social ties. Better-resourced citizens can often bypass barriers, while others encounter a narrower public agenda. The result may be a divided information sphere in which politically active groups know more but have less ability to recruit broad support. This is more subtle than a simple model of mass indoctrination, and more difficult to detect.
The framework is useful beyond China, but application requires care. Ranking, data access, propaganda, and technical failure can look alike from a user's perspective. Establishing censorship therefore requires evidence about purposeful restriction and the political accountability it is meant to avoid. The book's strongest lesson is methodological as well as political: study what people actually encounter, not just what remains technically available.
Review Summary
Censored receives mostly positive reviews, with an average rating of 4.18/5. Readers appreciate Roberts' theoretical framework for understanding internet censorship, especially in China. The book explores three censorship methods: fear, friction, and flooding. Many find the empirical evidence and insights valuable, though some note the academic tone and repetition. Reviewers highlight the book's relevance beyond China and its implications for democracies. Some criticize the dated information and lack of emphasis on similar techniques used in Western countries.
People Also Read
Glossary
Porous censorship
Circumventable information controlCensorship that makes information harder to access or distribute without making it completely unavailable. People with enough motivation, time, money, or technical skill may get around the restriction. Its political effect comes from diverting less motivated users while preserving the appearance that information remains accessible.
Fear
Deterrence through punishmentA censorship mechanism that discourages people from expressing, collecting, or accessing information by making them expect punishment or loss of reward. It works only when the target knows about the consequences and believes they are credible. Because it must be visible enough to deter, it can also provoke backlash or signal that the information matters.
Friction
Higher information access costsA censorship mechanism that increases the time, money, effort, or uncertainty involved in finding or sharing information. Examples include blocked or slowed websites, buried search results, inaccessible records, and failure to collect data. Friction can affect people who do not know censorship is occurring, but motivated users may pay the added cost.
Flooding
Coordinated competing informationA censorship mechanism in which an authority promotes abundant, convenient, or distracting information to compete with material it wants people to overlook. Flooding may include coordinated news coverage, government-supported online posts, or upbeat messages. Rather than directly removing an opposing story, it consumes attention and makes that story relatively harder to notice.
Rational ignorance
Choosing not to learnThe decision to spend little time learning about politics when the expected personal benefit of becoming informed seems smaller than the cost. Roberts uses the concept to explain why many citizens rely on familiar sources, prominent cues, and easily available information rather than searching exhaustively. It describes a calculation under limited attention, not a lack of intelligence.
Elasticity of demand for information
Sensitivity to access costsHow strongly people's information-seeking behavior changes when the time, money, or effort required to obtain information changes. Demand is highly elastic when a small increase in access costs causes people to switch to other material or stop searching. Roberts argues that this makes friction effective among users with limited interest or time.
Information pyramid
Stages of information flowRoberts's model of how information moves from potential evidence and events, through media collection and synthesis, to what the public consumes. Each stage filters out material according to production costs, audience demand, trust, and ease of access. Social media can let some information bypass traditional media, but it does not remove the public's limited attention.
Fifty Cent Party
Government-directed online commentatorsA term used for online commentators who post material at the direction of Chinese government bodies. In the evidence discussed in the book, many posts emphasize patriotism, encouragement, or positive feeling rather than directly arguing against critics. Their coordinated activity can help flood online spaces with material that distracts from sensitive events.
FAQ
What's Censored: Distraction and Diversion Inside China's Great Firewall about?
- Focus on Censorship: The book examines how the Chinese government uses various strategies to control information flow and public discourse, including fear, friction, and flooding.
- Porous Censorship: It introduces the concept of "porous censorship," where small costs are imposed on accessing information rather than outright bans, allowing some information flow while limiting politically sensitive content.
- Impact on Society: The book discusses how these methods affect citizens' access to information and political engagement, shaping public perception and discourse.
Why should I read Censored by Margaret E. Roberts?
- Insight into Authoritarian Control: It offers a deep dive into the mechanisms of control in China, essential for those interested in political science, media studies, or international relations.
- Understanding Modern Censorship: The book provides insights into how modern censorship differs from traditional forms, especially in the digital age.
- Empirical Evidence: Roberts supports her arguments with empirical research, including social media analysis and case studies, making the book a valuable resource.
What are the key takeaways of Censored?
- Three Mechanisms of Censorship: The book identifies fear, friction, and flooding as primary mechanisms, each affecting information access and public behavior differently.
- Porous Censorship Strategy: This strategy allows governments to control information while avoiding backlash, separating politically engaged elites from the general public.
- Citizen Behavior: Citizens often react to censorship with curiosity or anger, leading to increased engagement with censored topics.
What are the best quotes from Censored and what do they mean?
- "Censorship is not the correct word choice.": This suggests that traditional notions of censorship are inadequate to describe the complex ways information is managed in China.
- "Censorship acts not as a ban but as a tax on information.": It highlights that censorship imposes costs on accessing information rather than outright prohibiting it.
- "Fear is more difficult to use in the digital age.": This points to the challenges authoritarian regimes face in enforcing fear-based censorship in an era of widespread information sharing.
How does Censored define censorship?
- Formal Definition: Roberts defines censorship as the restriction of public expression or access to information by authority when it could undermine that authority.
- Types of Restrictions: The book distinguishes between restrictions on expression and access, both crucial for understanding censorship.
- Porous Nature: It acknowledges that censorship doesn't have to be complete to be effective; even small costs can impact public discourse.
What is the theory of censorship presented in Censored?
- Strategic Choice: Censorship is viewed as a strategic choice by governments to balance control and legitimacy.
- Mechanisms of Influence: Fear, friction, and flooding influence how citizens consume information, with different implications for dissemination and perception.
- Impact on Political Behavior: The theory suggests that censorship's effectiveness varies based on citizens' motivations and context.
What methods does Margaret E. Roberts use to study censorship in Censored?
- Empirical Research: The author uses a combination of qualitative and quantitative methods, including social media analysis, surveys, and case studies.
- Case Studies: Specific examples, like self-immolation protests in Tibet, illustrate the practical effects of censorship.
- Surveys and Experiments: Surveys gauge citizens' experiences with censorship, while experiments test hypotheses about behavior in response to censorship.
How does Censored explain the relationship between censorship and citizen behavior?
- Increased Curiosity: Observable censorship can lead to increased curiosity about suppressed topics, suggesting censorship can backfire.
- Empowerment to Speak Out: Censored individuals often feel more empowered to express dissent, indicating censorship can lead to greater political engagement.
- Social Media Dynamics: Social media can amplify dissenting voices, creating a complex relationship between censorship and public discourse.
What role does social media play in the context of censorship according to Censored?
- Platform for Dissent: Social media allows citizens to express opinions and organize dissent, even under censorship.
- Censorship Mechanisms: The book discusses content filtering and keyword blocking as methods to control information flow on social media.
- Backlash Against Censorship: Social media can facilitate backlash, as users share censored content to draw attention to government actions.
How does Margaret E. Roberts suggest that censorship can backfire in Censored?
- Increased Interest: Censorship can inadvertently highlight issues, leading to increased interest in suppressed topics.
- Empowerment of Citizens: Censorship can empower citizens to speak out, as they see it as a sign of weakness.
- Public Discourse Dynamics: Censorship can lead citizens to seek alternative information sources, creating a more informed and active citizenry.
What implications does Censored have for understanding authoritarian regimes?
- Tool of Control: The book shows how regimes use censorship to suppress dissent and shape public perception.
- Impact on Engagement: Censorship can influence political engagement, sometimes leading to increased activism.
- Lessons for Democracies: Insights from the book can inform discussions about censorship in democratic societies, helping protect against similar tactics.
How does Censored challenge conventional wisdom about censorship?
- Beyond Fear-Based Models: The book argues that fear is overemphasized as a mechanism, highlighting friction and flooding as impactful methods.
- Nuanced Understanding: Roberts suggests a nuanced approach to understanding censorship, focusing on strategic government choices.
- Implications for Democracies: The findings challenge the assumption that more information always leads to better outcomes, even in open societies.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.