Economic performance, land expropriation and bureaucrat promotion in China 

Get Complete Project Material File(s) Now! »

Types of Newspapers

China has one of the world’s largest newspaper markets, both in terms of circulation and advertising revenue. The newspaper market grew quickly until 2013 and newspapers remain a relevant source of news due to their strong online presence (Sparks et al., 2016).7 All newspapers are regulated and licensed by the State Administration of Press, Publication, Radio, Film and Television (SAPPRFT).
General-interest newspapers in China can be divided into three differ-ent types, depending on their ownership structure and revenue sources: official party newspapers, government-owned commercial newspapers and subsidiary commercial newspapers. Official party newspapers are owned and operated by the government. They are heavily subsidised and rely on subscriptions by government agencies and enterprises (see, e.g., Shirk, 2011). Qin, Str¨omberg and Wu (2018) find that party newspapers tend to be the most biased.
In contrast, commercial newspapers rely on advertising revenues. They compete for customers using more entertainment-oriented content. While some commercial newspapers are also government-owned, others are owned by other (government-owned) newspapers.8 Top personnel decisions at subsidiary newspapers are made by their parent newspapers.
The geographic distribution and hierarchy of Chinese newspapers mir-rors that of the government bureaucracy. National newspapers, such as the People’s Daily – the official newspaper of the central CCP leadership, are owned by the central government and are available in the entire country. Around 90 per cent of Chinese newspapers are circulated locally.9 These local newspapers are owned by local governments of different administra-tive levels, that is, provincial governments own provincial-level papers and prefecture governments own prefectural-level papers etc.
8No news outlet in China is truly independent of the government, as this govern-ment decree illustrates “[. . . ] no matter who its investors are, a news provider is a publicly owned resource” that has “[. . . ] just one shareholder: the Chinese Communist government” (He, 2004).
9Only the most successful provincial papers, such as the Southern Weekend, are available outside of their province.

Censorship of Newspapers

Like all media in China, newspapers are subject to strict government con-trol, both before and after publication. Ex ante, central and local propa-ganda bureaus issue reporting guidelines and hold meetings with chief ed-itors. Published content in newspapers is monitored and failure to comply with these guidelines can result in demotions, dismissals and jail sentences for journalists and editors (Shirk, 2011).
Apart from distributing propaganda and making profits, Chinese news-papers are also expected to report the performance of local officials to the central government (see, e.g. Shirk, 2011) and Qin, Str¨omberg and Wu (2018) find that central and party newspapers cover more news about offi-cial corruption in general. This role of local newspapers brings them into conflict with their local government officials. Local propaganda bureaus, which oversees newspapers, answer in principle to both the central propa-ganda bureau and the party leadership of their locality. In this paper, I find that local newspapers consistently underreport on important corrup-tion cases in their home area rather than acting as a watchdog. This is consistent with anecdotal evidence (see He, 2004) and online leaks from the Guangdong province propaganda department:
Do not independently investigate, report, or comment on the series of corruption cases in Maoming, with the exception of those which are arranged unified manner. (December 5, 2012)10 No media are to sensationalize the topic of government offi-cials making public their personal assets or related issues. Do not place reports on the front page, and do not lure readers to coverage. (January 24, 2013)11
These guidelines illustrate how local media can be forbidden from inves-tigating and reporting certain stories. During especially politically sensitive times, such as major party events, the media are generally advised not to report any negative news. The tone and framing of articles is also impor-tant and editors often copy official government announcements or articles by approved central news agencies to avoid sanctions.

Anti-Corruption Campaign

When Chinese President Xi Jinping came to power at the end of 2012, he vowed to end corruption in the CCP for fear it would otherwise “doom the party and the state”. The Central Commission for Discipline Inspection (CCDI) was tasked with investigating the over 80 million CCP members and punished 1.34 million of them for corruption in Xi’s first five years in office.12 Unlike previous anti-corruption drive, Xi promised to investigate both “Tigers” (high-level officials) and “Flies” (low-level officials).13 Contrary to press speculations, Lorentzen and Lu (2018) and Francois, Trebbi and Xiao (2016) find little evidence that the campaign is driven by factional politics.
According to the CCDI, citizen complaints are the most important source of information for the anti-corruption campaign. In 2012, 41.8 per-cent of all investigations originated in citizen complaints.14 Since then, the CCDI has streamlined complaint procedures, increased resources to process complaints and publicised the importance of complaints.15 The number of complaints have steadily increased to reach 34 million in 2018.
While the majority of corruption investigations are conducted inter-nally, the CCDI announces investigations into some officials directly on its website. These announcements include all “Tigers”, as well as a selected set of “flies”. The investigation announcements tend to be unexpected and almost always result in criminal charges or, at a minimum, the end of the official’s career. A typical example is the following CCDI post from 17 October 2013: The Deputy Secretary of the Municipal Party Committee and Mayor of Nanjing, Ji Jianye, is suspected of seriously violat-ing discipline and engaging in illegal behaviour and is currently under investigation.16
The anti-corruption campaign is covered extensively in the central state me-dia. For instance, the People’s Daily has a dedicated section on its website, where articles about new regulations, individual cases and the campaign progress are published. Online and print versions of other newspapers are also an important source of news for detailed accounts about the anti-corruption campaign. Television stations spend their limited broadcasting time on programmes about the campaign. For example, the CCDI and China’s national broadcaster CCTV produced and aired several documen-taries about the campaign, featuring on-air confessions of corrupt officials. A fictionalised anti-corruption TV drama (In the Name of the People), in part financed by the national agency responsible for prosecuting corrup-tion cases, became China’s most popular TV drama in recent years.17 This is, reportedly, the first time since 2004 that national censors allowed the broadcast of a TV show about high-level corruption and the government has since commissioned more films and TV shows on similar topics.

Data and Descriptive Statistics


The sample of newspapers consists of all Chinese-language general interest newspapers available on WiseNews, an online news archive based in Hong Kong, from the end of 2010 to the end of 2015, yielding a total of 99 newspapers. WiseNews selects influential publications from large cities and is representative of newspapers in large metropolitan areas in China (Qin, Str¨omberg and Wu, 2018).18 I collect information on newspaper ownership from the SAPPRFT website and China Journalism Yearbooks.
Panel A of Appendix Table 1.A1 presents summary statistics of the newspapers in the sample. 52 percent of the sample are subsidiary newspa-pers and 48 percent are directly government-owned newspapers, of which 21 percent are official party papers. The majority of newspapers are provincial newspapers (55 percent), central papers represent 18 percent of the sample, while sub-provincial papers represent 27 percent. Appendix Figure 1.A1 shows the 14 out of 31 province-level administrative regions in mainland China, including all four province-level municipalities, that have at least one newspaper in the sample. This represents more than half of China, in terms of population (53 percent) and GDP (57 percent).19
18Of these, 76 are available for the entire time period. WiseNews provides content for 105 newspapers, of which three are in English and another three only report about health and sports. I also consolidate local editions of the same newspaper, such as Nanfang Ribao, Nanfang Dushibao, Guangzhou Ribao and Nanguo Chengbao.


Officials under Investigation

The sample of officials under investigation contains 408 individuals. These are all the officials named on the CCDI website from the start of the anti-corruption campaign in 2012 until the end of 2014, from the 14 provinces (and central government) that are covered by the WiseNews sample.20 Ap-pendix Figure 1.A3 shows the distribution of officials and newspapers across provinces. All CCDI posts include the time of announcement and the of-ficial’s name, occupation and location. I use an officials’ occupation to determine their rank in the government hierarchy (see Appendix Figure 1.A2 for the distribution of ranks).
Panel B of Appendix Table 1.A1 shows summary statistics of the of-ficials in the sample. 9 percent of officials (34 officials) in the sample are high ranking. This represents the universe of all “Tigers” investigated over this time period from these provinces, see also 1.7.5. 78 percent of the sam-ple (320 officials) are government officials, mainly from local governments (72 percent of the sample, 295 officials), and 22 percent are officials from state-owned institutions (hospitals, universities etc.) and enterprises.


I collect all articles about these 408 officials published by the 99 newspa-pers in the WiseNews database from two years prior to one year following the CCDI announcement, resulting in 39,271 unique articles.21 By con-struction, each article can be matched to a newspaper and one or more officials.22
Summary statistics of articles are reported in Panel C of Appendix Ta-ble 1.A1 and a more detailed text analysis of the content and sentiment of the articles is shown in Appendix 1.B. The average articles contains around 1300 words, though most of the articles are shorter. More than half of the articles are on pages one to five, with a mean page number of eight.23 Among articles published following an investigation announce-ment, 22 percent of article headlines include direct references to corruption and eight percent of headlines include keywords associated with the cen-tral anti-corruption campaign.24 One percent of article headlines copy the official CCDI announcement. 27 percent of articles mention more than one official under investigation, as newspapers frequently summarise the progress of the anti-corruption campaign by listing several cases.25
In order to take into account all articles that can potentially be pub-lished, I construct a daily panel dataset for the 40,392 official-newspaper pairs over the entire sample period, resulting in around 40 million official-newspaper-day observations (summary statistics are reported in Panel D of Appendix Table 1.A1). Appendix Figure 1.A4 shows the fraction of official-newspaper pairs with at least one published article by day rela-tive to the investigation announcement at day zero. The mean probability that a newspaper publishes an article about an official on a given day 0.1 percent.

Empirical Strategy

I compare whether and how local newspapers report about officials inves-tigated during the anti-corruption campaign from their own province rel-ative to officials from other provinces. There are three potential patterns of reporting. If the central government dictates a uniform media strategy surrounding the anti-corruption campaign, newspapers might report the same stories about each official, regardless of whether they are from the newspaper’s home province. Newspapers could report more about inves-tigated officials from their own province, either because they have more information about them or because their readers are more interested in their home-province officials (see Section 1.6.1 for evidence on reader de-mand). In contrast, if newspapers report less about corrupt officials from their own province, this could be in response to pressure from their local government – a scenario I refer to as local censorship.26 There are a number of reasons for local censorship. Publicising the investigations could encour-age members of the public to come forward and implicate officials still in power. Supervisors and colleagues could be accused of condoning corrup-tion. Revealing large-scale local corruption could damage the reputation of the local government. See Section 1.6.2 for evidence on citizen complaints and Section 1.6 for evidence that local censorship is stronger when it comes to well-connected officials.

Table of contents :

1 Intergovernmental Conflict and Censorship: Evidence from China’s Anti-Corruption Campaign 
1.1 Introduction
1.2 Background
1.2.1 Types of Newspapers
1.2.2 Censorship of Newspapers
1.2.3 Anti-Corruption Campaign
1.3 Data and Descriptive Statistics
1.3.1 Newspapers
1.3.2 Officials under Investigation
1.3.3 Articles
1.4 Empirical Strategy
1.4.1 Are Local Corruption Scandals Selectively Underreported?
1.4.2 How are Local Corruption Scandals Reported?
1.5 Results
1.5.1 Underreporting of Corruption Scandals: “Tigers” versus “Flies”
1.5.2 Deemphasising Local Corruption Scandals
1.6 Mechanisms
1.6.1 Readers’ Interest in Local Corruption
1.6.2 Effect of Underreporting on Local Government Accountability
1.6.3 Conditions for Newspaper Capture
1.6.4 Importance of Local Political Connections
1.6.5 Interpretation of Results
1.7 Robustness
1.7.1 Time-Profile of Estimated Coefficients
1.7.2 Flexible official- and newspaper-province specification
1.7.3 Functional form specification
1.7.4 Investigative Reporting
1.7.5 Selection of officials and newspapers
1.8 Conclusion
1.9 Figures
1.10 Tables
1.A Detailed Sample Description
1.B Descriptive Text Analysis of Newspaper Articles
1.B.1 Sentiment Analysis
1.B.2 Content Analysis
1.C Sina Weibo
1.D Additional Figures and Tables
1.E Additional Robustness Check Tables
1.F Regression Tables for Figures in Main Text
2 Anti-Japanese Protests, Social Media Hate Speech and Television Shows in China 
2.1 Introduction
2.2 Background
2.2.1 Relationship between China and Japan
2.2.2 Media environment in China
2.3 Data
2.3.1 TV data
2.3.2 Social media data
2.3.3 Anti-Japanese protests
2.3.4 Other data
2.4 Empirical Strategy
2.4.1 Endogenous scheduling
2.4.2 Selection of TV audiences
2.5 Results
2.5.1 Anti-Japanese Protests
2.5.2 Anti-Japanese Hate Speech on Social Media
2.6 Robustness
2.6.1 Leads and Lags of Historical TV Drama Exposure
2.6.2 Excluding TV Shows By Own-Province TV Channel
2.7 Conclusion
2.8 Figures
2.9 Tables
2.A Social Media
2.B Additional Descriptive Statistics
2.C Additional Results and Robustness Tables
3 Economic performance, land expropriation and bureaucrat promotion in China 
3.1 Introduction
3.2 Background
3.2.1 Chinese bureaucratic system
3.2.2 Land markets in China
3.3 Theoretical Framework
3.4 Data and descriptive statistics
3.4.1 Prefecture CCP secretaries
3.4.2 Land expropriations
3.4.3 Macroeconomic data
3.4.4 Remote sensing data
3.5 Empirical strategy
3.5.1 Effect of competition on promotion
3.5.2 Effect of competition on policy choices
3.6 Results
3.6.1 Promotions and the number of competitors
3.6.2 Economic growth and number of competitors
3.6.3 Construction-led growth
3.6.4 Consequences of land expropriations
3.6.5 Public goods provision
3.7 Robustness checks
3.7.1 Explaining variation in starting cohort size
3.7.2 Identification checks
3.7.3 Alternative specifications
3.7.4 Data quality concerns
3.8 Conclusion
3.9 Figures
3.10 Tables
3.A Robustness Tables and Figures
3.B Model
3.C Detailed sample description
3.C.1 Bureaucrats
3.C.2 Policy outcomes
3.D Measurement of GDP growth using night light data 221


Related Posts