Weibo 微博 has made news headlines in recent years for exposing the wrongdoings of officials and reporting protest actions in China. The Chinese authorities have responded with a series of crackdowns on Weibo users and usage, while at the same time stepping up their own use of online media. This paper seeks to understand the implications of Weibo for the landscape of public opinion in China, in particular with regard to opportunities afforded to Chinese citizens to express in public their concerns and thoughts and the impact this has on the functioning of the party-state apparatus.
This paper is motivated by a bigger concern: the implications of the internet for a more open and democratic China. Scholars have arrived at three schools of thought in this regard. One sees the internet as empowering citizens and civil society organizations and thus has the potential to increase freedom and facilitate democracy.Footnote 1 Another sees the authoritarian party-state using the internet to control society further and strengthen the regime as part of its adaptation.Footnote 2 A middle-ground position sees the future as indeterminate, with civil society engaged in a continual battle of “resistance” against and “negotiation” with the party-state, resulting in the two sides becoming mutually transformative.Footnote 3 This paper contributes to this discussion by examining the usage of one particular internet technology, Weibo, in the exposure of 29 officials accused of corruption in the two months following the confirmation of the new Xi Jinping 习近平/Li Keqiang 李克强 administration in November 2012. Similar studies on China have mostly employed a descriptive-analytical case study approach, favouring cases that involve resistance and push-back from civil society. This study is complementary by studying a large number of cases that were not selected for their outcome, so there is no initial bias of selection. The online exposures of corrupt officials at the end of the 18th Party Congress marked the beginning of many more cases being revealed the following year during an official anti-corruption campaign, and foreshadowed a subsequent crackdown on Weibo users. As such, the exposures can be considered indicative of a newly emerging pattern of ideological formation of continuing significance. The concepts of “opinion leaders” and “agenda setting” guide the empirical study.
Opinion Leaders and Agenda Setting
The “opinion leaders” concept was put forth in the two-step flow theory of persuasion in communication studies in the 1940s and 1950s.Footnote 4 Lazarsfeld, Katz and his colleagues suggested that opinion leaders acted as intermediaries between mass media and society to pass on ideas that they derive from the mass media to their peers. In its original conceptualization, opinion leadership is “related (1) to the personification of certain values (who one is); (2) to competence (what one knows); and (3) to strategic social location (whom one knows).”Footnote 5 Opinion leaders are not leaders in the common sense of the term, but rather people found in every level of society, who influence their family, their friends and their peer group. They have higher exposure to the mass media than those they influence. Although subsequent studies have contested the original conceptualization of opinion leaders and the two-step flow theory,Footnote 6 by the 1970s, the two-step flow that recognizes the personal influence of opinion leaders/influentials had become the single most popular theory of the dominant paradigm in communications.Footnote 7 To date, the role of interpersonal interaction in mediating mass media effects has become accepted in other disciplines.Footnote 8 Through this time, the conceptualization of an opinion leader has been modified, remodified and extended.
Agenda setting is a hypothesis proposed by McCombs and Shaw to refer to the process in which what was prominently reported in news came to be perceived as important by the audience.Footnote 9 Influentials have been identified as playing a role in setting the public agenda as “early recognizers” who identify emerging issues in the media and diffuse them to the public.Footnote 10
Identifying Influentials Online
In the online world, influence is commonly measured either by the position of a user in the social network or the extent of information adoption by others of a user's message. The page-rank and number of followers are popular measures of a user's position in social networks,Footnote 11 whereas the number of mentions and retweets on TwitterFootnote 12 are measures of information adoption.
On Twitter, Cha, Haddadi, Benevenuto and Gummadi found – contrary to what is commonly assumed – that the most followed users did not necessarily generate the most retweets or mentions with their messages. The most followed accounts were news sources and public figures; the most retweeted accounts were content aggregation services, businessmen and news sites; and the most mentioned users were mainly celebrities.Footnote 13 Lee, Kwak, Park and Moon suggested that information adoption was a more relevant indicator of influence since for 80 per cent of users only 20 per cent of their followers were first exposed to the thread of a topic through a tweet by their followed writer. Using the number of “potential readers” (meaning those first exposed to a piece of information) as the measure of influence, they found that news media made up most of the influential users on Twitter.Footnote 14 Meraz and Papacharissi also found that on Twitter the most retweeted messages came from news media accounts.Footnote 15
One might ask if it is the author or the content of the message that spurs retweets. Counting the total number of reposts of a message in the entire diffusion tree, Bakshy, Hofman, Mason and Watts found that most events did not spread at all, and even moderately sized cascades were extremely rare. Content-based attributes could not predict the scale of diffusion, but past success in being reposted and the number of followers could.Footnote 16 Their last result seems to differ from the findings of Cha and colleagues. This might be explained by the fact that Cha et al. used the number of direct explicit retweets (and not total diffusion) of the message as the retweet count.
The Roles of Online Influentials
Online, influentials are identifiable by various roles. Java, Song, Finin and Tseng found three main types of roles in social networks on Twitter: information source, friends and information seekers.Footnote 17 Information sources have large numbers of followers and they post updates. Tinati, Carr, Hall and Brentwood differentiated five roles among users in the Twitter retweet networks: idea starters (“who start a conversation meme by generating a certain minimum number of retweets”); curators (“who connect two idea starters together”); amplifiers (who are the first “to retweet a chain of retweets”); commentators (“who do not fit into any other group”); and viewers (“who take passive interest in the conversation”).Footnote 18 The roles were identified based on retweet counts, providing further support to previous studies that suggest retweet counts can be used as a measure of influence.
The influence of idea starters and amplifiers can be explained by findings that the earlier posts in the timespan of a topic are more likely to produce mentions (although exceptions were found).Footnote 19 The number of potential readers increases rapidly in the early stages, but this growth slows down over time.Footnote 20 The term “information source” or “idea starter” could be misleading as ideas that appear first in an online space might not have come from the users who first post them. A study on political Usenet newsgroups found that 95 per cent of messages posted by “discussion catalysts” – who received a disproportionate number of replies – contained content taken from elsewhere on the web, with two-thirds coming from traditional news sources.Footnote 21 This means online influentials often filter and amplify other media content for their audience, exerting personal influence just as in the mass communication process.
In this study, the term “initiator” is used instead to refer to the user who publishes the first post about a case and who draws at least one retweet. Retweeting is set at the lowest level to avoid conflation with the other opinion leader roles. “Amplifiers” in the scheme of Tinati et al. bear resemblance to “early recognizers” in agenda setting research in that they bring emerging issues to the attention of their audience (their followers in the case of social media).Footnote 22 Here, these early posters are called “agenda setters” to highlight their role in spreading the meme at an early stage. Although in most cases, messages published at the early stage of a meme draw more responses, this is not so in the dissemination of all topics.Footnote 23 To identify influentials in the overall diffusion of an issue, this study further identifies those posters – called “disseminators” – who spread the meme at a later stage of the issue. Following Bakshy and his colleagues, this study takes the total repost and comment count of the message in the diffusion tree as the measure of influence.
Opinion Leaders on China's Weibo
With the launch of Sina Weibo in 2009, the significance of public opinion formation online reached new heights,Footnote 24 contemporaneous with rising public outcry over incidents such as the crash of the new high-speed train in Wenzhou 温州 in 2011. On Weibo, users with large followings are commonly taken to be opinion leaders. At the time of writing, four of the top five users with the largest number of followers on Sina Weibo are entertainers; the fifth user is a writer of romantic fiction and prose. However, little is known about the opinion leaders who actually influence the spread of news events and issues on Weibo. What is known is that more than half of Sina Weibo accounts have published nothing, and less than 5 per cent of those who have published account for more than 80 per cent of original posts.Footnote 25 Verified accountsFootnote 26 are far more likely to post or repost, and more inclined to post (nearly 30 per cent) than repost (17.5 per cent).Footnote 27
Li Biao found that opinion leaders on Sina Weibo were mainly holders of verified accounts (68.6 per cent), overwhelmingly male (92.7 per cent), middle-aged (with an average age of 43.6 years old), and had tertiary-or-above education.Footnote 28 Almost 30 per cent of opinion leaders were media workers, close to 20 per cent were entertainment celebrities, and a few were professionals. Yet, Li did not indicate how the source post he studied was identified, as there are usually many source posts for each case, or what level of repost was used to define an opinion leader. Li's definition of opinion leaders also excluded news media and official agencies, thus failing to provide an overview of the relative prominence of various user categories. Gillian Bolsover found that the Weibo service provider itself was the top opinion leader, with news organizations ranked as second and bloggers as the third most influential sources.Footnote 29 She also found that government officials and departments, with a few exceptions, generally had low levels of influence. Bolsover's study, focusing on the dissemination of stories published by news organizations, did not provide an understanding of the relative influence of different user types in initiating topics. Tong and Zuo found that local residents and ordinary people, although being the most active users on Weibo, tended to repost rather than post original messages.Footnote 30
Censorship, Control and Propaganda on Weibo
In post-Mao China, the mass media play the key role in official propaganda. Hand in glove with propaganda is the censorship and control of the media. Broadcasting is run by the government, and newspapers above the county level must have an official sponsor.Footnote 31 The internet is the only public medium through which alternative messages can be circulated.Footnote 32
Sina Weibo comes under the same system of technical control and human scrutiny as other internet service providers. In addition, since the end of 2011, Weibo users have been required to register with their real names.Footnote 33 Regulations became law at the end of 2012, when real-name registration was required of all internet users.Footnote 34 A judicial interpretation in September 2013 made people who publish slanderous comments online liable to defamation charges if their posts are read by 5,000 or more users, or reposted 500 times or more.Footnote 35 Soon after the appointment of the Xi–Li administration, a number of critical Weibo writers with large followings were arrested in what has been described as China's biggest crackdown in years.Footnote 36 Moreover, Weibo's “big Vs” (verified users with huge followings) were invited to publicize the official line of propaganda.Footnote 37 Online channels, including Weibo, are now considered to be the “main battle ground” in the struggle for the ideological security of the country.Footnote 38
In 2013, governments of all levels were called upon to be more open with government information and interact more with the public by making use of channels such as government websites and government Weibos.Footnote 39 The number of Weibo accounts registered to public agencies and officials rose by almost 250 per cent in December 2012 over the previous year, totalling 176,700 accounts on four major Weibo service providers (Sina, Tencent, People's Daily Online and Xinhuanet).Footnote 40 Some localities have incorporated performance on Weibo as part of the routine evaluation criteria of officials. At the same time, the number of news media accounts on Weibo also jumped exponentially.Footnote 41
A survey in 2009 found that corruption ranked as the top concern of Chinese citizens, with 82.9 per cent of respondents considering it as having exceeded tolerance levels.Footnote 42 The new Xi–Li leadership has made combating corruption a top priority. Weibo is one of the key online spaces used for what has come to be known as the “Weibo anti-corruption” campaign. This means that official authorities are in competition with other online users in the quest for public opinion formation, and makes the exposure of corruption cases an appropriate issue for studying the relative prominence of various user types on Sina Weibo.
Sina Weibo in China
Since the second half of 2012, Sina Weibo has captured more mainland Chinese internet users (28.4 per cent) than any other Weibo site. It is worth nothing that nearly half (45.5 per cent) of the entire population now use the internet.Footnote 43 Compared to Tencent, which draws users from the mass market, Sina Weibo – supported by its fleet of Chinese celebrities including movie, singing and sports stars, business and media figures – is closely identified with the urban elite.Footnote 44
Sina Weibo is often described as the Chinese equivalent of Twitter, although it also includes Facebook features. It offers ordinary citizens a channel to post short messages of 140 characters (or more if the “long Weibo,” photo or video functions are used) to their networks of “followers.” Users bring their offline social networks to the platform, and can also acquire new connections by “following” and being “followed” by other users, resulting in a social network that connects the 500 million Sina Weibo user accounts. At the time of writing, the number of followers of the top user account on Sina Weibo was in excess of 76 million, while new anonymous ordinary users have hardly any followers. Building on its interconnected social network, Sina Weibo has become a platform for the formation of public opinion, with 80.3 per cent of its users (more so than with any other Weibo platform) following news and “hot” social events on the platform.Footnote 45
Methodology
We identified 31 corruption cases from news searches conducted during the two months following the date of exposure of the first case after the 18th Party Congress, between 20 November 2012 and 19 January 2013.Footnote 46 We excluded two cases: one because its exposure came on two distinct dates more than two weeks apart, which made it difficult to study, and another because there were hardly any posts about it on Sina Weibo. A total of 29 cases were included. Searches on the Sina Weibo user interface were conducted to identify the time and date of the first post published about each of the cases. Where the first post was deleted, cues from reposts of the deleted post were taken.
We consider that diffusion of the cases consists of three stages of communication: case initiation, agenda setting and dissemination. Different diffusion patterns have been found for different types of events on Twitter: 73 per cent of trending topics on Twitter have a single active period, 31 per cent of which last one day, and only 7 per cent last longer than 10 days.Footnote 47 Half of retweeting occurs within one hour, and 75 per cent within one day. In the absence of similar studies on Sina Weibo, the grounded-theory approach was taken to determine the duration of the periods of agenda-setting and dissemination.Footnote 48 Based on observations of changes in patterns of the number of retweets and comments of the source post, the agenda-setting period was initially set at three hours following the publication of the first post about a case. The dissemination period, which is meant to cover the entire diffusion of the case, was set as one month following the agenda-setting period.
We conducted a search of the Weiboscope database using the names of the exposed officials as the keywords for the two periods.Footnote 49 In cases where the character combination of the name was common, we supplemented the keywords with job titles. For cases with the first post deleted, the first period was extended backwards to cater for possible unknown earlier posts. To compensate for the absence of accounts with less than 1,000 followers in the database, advanced hour-by-hour keyword name/title searches for the agenda-setting period were conducted for each of the corruption cases on the Sina Weibo user interface. In cases where few reposts occurred during the first three hours, further searches were conducted on the database and on the user interface with the first period extended by three hours up to 24 hours. The agenda-setting period of a case was then set at when the total reposts reached 100 and comments reached 40. On this basis, the agenda-setting period for four cases was extended to six hours, one case to 12 hours, and another case to 21 hours. Three other cases, probably owing to the deletion of posts, did not return any search result in the first 24 hours.
This study defines public opinion leaders on Sina Weibo as one of three types of users: initiators of a case (who are the first to publish and who get retweeted at least once); agenda setters (whose message gets heavily reposted or commented upon during the agenda-setting period of the case); and disseminators (whose message gets heavily reposted or commented upon during the dissemination period). Four measures were used to identify agenda-setting opinion leaders:
• Accounts that published the three most reposted messages in period one;
• Accounts that published the three most commented upon messages in period one;
• Accounts that published messages with a repost count of 25 per cent or more of the most reposted message in period one;
• Accounts that published messages with a comment count of 25 per cent or more of the most commented-upon message in period one.
Dissemination opinion leaders were operationalized as:
• Accounts that published the eight most reposted messages in period two;
• Accounts that published the eight most commented upon messages in period two;
• Accounts that published messages with a repost count of 25 per cent or more of the most reposted message in period two;
• Accounts that published messages with a comment count of 25 per cent or more of the most commented-upon message in period two.
Referencing and adapting categorizations used in previous studies, the identity of public opinion leaders was coded according to a 17-category scheme: business organization, business person, cultural/media worker, entertainment worker, expert, government/Party body, government/Party personnel, independent media worker, information source, news organization, news worker, online media organization, online media worker, professional, NPO organization, NPO worker, and citizens (see Appendix).Footnote 50 Accounts that were not retrievable on the Sina Weibo user interface for checking were excluded. The quantitative analysis was supplemented by the qualitative reading of posts and reposts.
Results and Discussion
The vast majority of source posts about the cases did not get reposted at all. This means that those who did get reposted hold influence over other users to different degrees.
Citizens: witnessing and activism
Using the definition of initiators as first posters who elicit any reposts, citizens made up almost a quarter of all initiators (24.14%, n = 7) (see Table 1). Four of the citizen initiators drew a substantial number of reposts (over 1,500 in one case), while the other three drew very few. Three of the four substantially reposted posts were authored apparently by people involved in the cases, making the posts first-hand citizen witness reports. Two of the posts, seemingly published by the former mistresses of the officials Du Zeyong 杜泽勇 and Chan Zengde 单增德, gave details of their extra-marital sexual affairs and included photographs. These two first posts were deleted at an early stage but were still accessible as reposts as they were copied and pasted in subsequent posts or simply referred to. The initiator post in the third case was supposedly published by someone who lived in the village of the accused corrupt officials. These first posters did not have large followings. The user who exposed the village officials, for example, only had 108 followers on 21 January 2013. Nevertheless, these first posts succeeded in spreading news of the cases. For example, the most popular message about the Du Zeyong scandal (authored by a news organization and which referred to “an exposé by an internet user” and reported an official announcement) drew over 8,000 reposts. This suggests that citizens, despite their weak position in the social network on Weibo, are able to initiate exposés of cases in which they are involved. Given the advantage of citizens as affected parties, it is, however, surprising that not more of the cases were initiated by citizens.
Table 1: Initiators of the Cases
The fourth case that drew substantial reposts was initiated by a clearly politically aware citizen, who stated in her personal information section that: “dissenting opinion about current affairs is the highest form of patriotism.” Her post, although the first post about the case on Sina Weibo, seemed to be a copy-and-paste of a post published on another Weibo platform, apparently by what this study would categorize as an information source. The content was about an official's sexual assaults of a number of secondary school pupils and his nine mistresses. Her account had 61,041 followers on 22 December 2013. She published posts concerning at least four of the 29 cases under study, and was among the top eight disseminators of one other case.
Media organizations: the most prominent opinion leaders
News organizations and online media (overwhelmingly the Headline News account of Sina News, a news aggregator that republishes news headlines provided by news organizations) together were responsible for initiating ten cases (almost 35 per cent) (Table 1). In some of the cases, the news aggregator unexpectedly trumped the news organization that supplied the story as initiator, with some of the posts published by news organizations including a URL that pointed to a Sina News page and not their own news website. This phenomenon, which might have come about through some commercial agreement between news organizations and Sina News, makes the differentiation between the two user categories in this case less significant.
Although citizens were prominent initiators, they relied on other users to focus attention on and spread news of the scandals. News organizations came top in the user category of agenda setter and disseminator, followed by online media. This is a consistent pattern across reposted (Table 2) and commented-upon measures (Table 3), except when disseminator opinion leaders were operationalized as those who were reposted/commented-upon at more than 25 per cent of the level of the top message of the case; then online media surpassed news organizations as the top category. Both categories have transferred their credibility as offline public communicators to their verified accounts on Sina Weibo. As can be seen from Tables 1, 2 and 3, a high percentage of those who initiate, set the agenda for and disseminate cases are verified accounts, which indicates that they have an advantage.
Table 2: Accounts with the Most Reposted Messages
Table 3: Accounts with the Most Commented-upon Messages
Direct and indirect influence of government/Party bodies
Government/Party bodies initiated less than one-fifth of the cases (Table 1), and they were not effective agenda setters or disseminators (Table 2 and 3). Of the five cases initiated by government/Party bodies, four of the original posts drew few reposts (from five to 27). The one post that had a reasonable reposting rate (n = 300, as on 10 January 2013) was published by the Communist Party Disciplinary Committee in Ningbo 宁波 concerning the Party secretary, Jin Junjie 金俊杰, who had a much higher ranking than the other officials exposed by government/Party accounts. Although the first official posts about these cases only contained formulaic official statements, the relative success of the initial post about Jin suggests that good information disseminated by official accounts still draws a response.
Three of the five cases initiated by official accounts drew only lukewarm responses from Sina Weibo users (the most popular post was reposted only 620 times). The case of Lu Yingming 吕英明, however, was a major exception. A post published by a columnist that contained information, sourced from “the web,” that Lu had affairs with many female performers, owned 63 houses and 47 mistresses, was reposted more than 20,000 times. Sex and the network status of the columnist, who had over half a million followers at the time of writing, are likely factors that helped dissemination of Lu's case.
However, the real influence of government and Party accounts is more profound than the numbers indicate. Examining posts published by Sina Headline News reveals that they were mainly copied and pasted from the introductions of news reports by news organizations, which, in turn, had published identical or slightly abbreviated versions of the same posts published by government/Party bodies. The following message, published by the Nanfang Daily at 17:27 on 26 November 2012, is typical:
[Rush: Deputy head of Department of Land and Resources of Guangdong province under investigation for suspected violation of discipline] The reporter has got confirmation from responsible personnel of the Disciplinary Committee of Guangdong province that the deputy head of the Department of Land and Resources (previously head of Department of Water Resources), comrade Lu Yingming, is under investigation by the organization for a suspected serious violation of discipline. Reporter Zhao Yang of Nangfang Daily http://t.cn/zjGhazv Footnote 51
The same message was published by Sina News' Headline News at 17:38 the same day; this was a slightly shorter version of an almost identical message published at 17:22 on 26 November 2012 by the official account of the Disciplinary Committee, Bureau of Supervision and Bureau of Corruption Prevention of Zhongshan 中山 city, Guangdong province. This shows how news organizations and Sina News worked in unison to propagate the messages published by official agencies. All initial postings by news organizations and Sina Headline News regarding the cases in the study sourced information from official agencies and followed the official format for information announcements.Footnote 52
The sources of information for more than half (51.72%, n = 15) of the 29 cases were government and Party bodies. Nearly 90 per cent (89.74%) of messages published by news and online media that fell within the most reposted three in the agenda setting period were official announcements. The percentage was lower among the top eight in the disseminator period but still remained at over 60 per cent (60.65%). This is evidence of the continued dominance of the authorities as sources of key information about public life in China. Yet, the first posts published by news organizations and online media, although providing virtually the same information as government and Party bodies, were reposted more than those published by government/Party bodies. This suggests that users prefer to interact with media organizations than with official agencies on Sina Weibo. It could also explain why some official bodies release news of cases to news media instead of making the announcements themselves. Official content also forms an important source of information for ordinary citizens, including the politically aware citizen initiator, some of whose messages were copy-and-pastes of official announcements.
The possibility of challenging official propaganda
In general, news organizations and Sina Headline News posted only information released by official agencies about the cases. This is owing to the media's role as a primary propaganda organ. However, they did publicize cases initiated by other user types, which indicates that media organizations play an additional, secondary role as the voice of the people. This role no doubt helps them to maintain their opinion leadership over official agencies on Sina Weibo. The second most reposted message (33,431) was about the widely disseminated Gong Xianxia 龚仙霞 case and came from Sina Headline News. It was published as a summary of netizens' posts of the day:
[Today's web messages] Anti-corruption, anti-corruption. The more anti-, the more corruption! Local internal investigation is only a show! Ultimately fighting corruption and championing clean governance depends on: first, media follow up, second, beauties taking off their pants, third, uncles parading their wealth, fourth, house sisters changing houses, fifth, officials' sex scandals, sixth, the helplessness of unpaid workers, seventh, deserted orphans in Lan Kao, eighth, cries of turned-away petitioners, ninth, arrogance of forced demolitions … Netizens lamented the present state of anti-corruption. Click to read a summary of today's web messages: http://t.cn/zYAYKZp.
This post, written as a piece of parody, certainly serves little propaganda purpose. The post could have been written to appeal to Weibo users, who generally take an anti-authoritarian stance.Footnote 53 Despite being a commercial company, Sina has to comply with censorship and the propaganda requirements of the authorities and, similar to news organizations, must serve two masters: the market and Party.Footnote 54 Besides publishing the official announcements about the corruption cases, People's Daily online published this message, which became the sixth most reposted message (14,509) about Lei Zhengfu 雷政富:
[Hello, tomorrow] From Lei Zhengfu to Liang Daohang 梁道行, within the short period of several weeks, many officials have fallen. The determination of the [Party] central to fight corruption is clearly seen. Applause has to be accompanied by alertness. Determination can achieve instant effects but only perfecting the system can produce lasting governance and stability. Governing its officials is the first step in governing a country. Governance of officials must be severe. Let us hope that every case becomes an opportunity to improve the system. “Only when people are allowed to rise to supervise government would the government not slacken. Only when everyone rises to take responsibility would people live and the regime survives.” Good night.
The positive sentiment about the anti-corruption measures and emphasis on improving the system are consistent with the Party line; however, the suggestion of “people … allowed to rise to supervise government” comes close to crossing official ideological boundaries.
News workers initiated only 13.79 per cent (n = 4) of the cases (Table 1), but they all drew a heated response. Sex apparently explained the wide interest in two cases. In one case, a news anchor exposed an official who had forced himself sexually upon her. Her first post was deleted but one of her subsequent posts elicited over 4,000 reposts. In another case, a news reporter posted photographs of an official visiting his mistress who happened to live in the same building as the reporter. This initial posting was reposted nearly 3,000 times. The position of the official involved and the network status of the initiator explain the heated response in the third case, which was an exposé of Liu Tienan 刘铁男, deputy director of the National Development and Reform Commission, by the-then deputy editor of the respected Caijing magazine, Luo Changping 罗昌平. Luo had over half a million followers at the time of writing. Luo's three messages, published within one minute of 11:01 on 6 December 2012, drew a total of over 50,000 reposts in the first three hours. His third message was also the most reposted one about the case.
Unlike the posts published by official agencies or media organizations, posts published by news workers often contained details about officials' alleged wrongdoings. An example is Luo Changping's second message exposing Liu Tienan, posted at 11:01 6 December 2012:
[Reporting to the Central Disciplinary Committee in real name: Two] Liu Tienan and businessman Ni Ritao 倪日涛 formed a merchant-official alliance. His wife at chu 处 rank, Guo Jinghua 郭静华, and son, Liu Decheng 刘德成, hold shares of Ni's company, and cheated loans from domestic banks in foreign takeovers. Liu Decheng's HSBC bank account in Canadian dollars, 10112-376762-150, in US dollars, 10112-376762-250, and US dollars, 028- 490415-833, received multiple huge remittances from Ni Ritao's company. Read Caijing's report http://t.cn/zjME1YO.
It is well documented that news workers do not necessarily agree with the ideology imposed on news organizations, and there are numerous cases in which news workers have tried to present events and issues differently, within given constraints. Weibo offers them a valuable space to do just that. As their public profile helps them to gather a following, they can exert influence relatively independently as initiators, agenda setters and disseminators of cases. News workers are connected to the news source networks, recognize important stories, and possess the skills to gather information and write up the story. The fourth case initiated by news workers, about Gong Xianxia, was not an original exposé but was sourced from one of the Sina Weibo forums, presumably with information provided by citizens. News workers are likely to be among the most motivated users to source information provided by civilians elsewhere on the web to initiate stories on Weibo. The combination of their expertise and drive seems to make news workers the most promising user category to challenge the official propaganda machine built by official agencies and media organizations.
Conclusion
With the rise in popularity of social media, much hope has been expressed about citizens being empowered to form public opinion. This study investigates who indeed initiated, set the agenda of, and disseminated messages about 29 corruption cases on Sina Weibo after the 18th Party Congress. It finds that ordinary citizens made good use of the social media platform to initiate cases, but that they relied on media organizations to spread information. As such, their role as opinion leaders is limited. News organizations and Sina Headline News were the major initiators of cases, as well as the most prominent agenda setters and disseminators. This suggests that their importance as public communicators has been transferred online to a large extent. This finding is consistent with previous studies. The party-state, generally considered a weak player on Weibo, was found to be an active initiator of cases, some of which drew heated responses from Weibo users, but neither government nor Party bodies were strong agenda setters or disseminators, and their messages elicited few reposts. However, the real influence of the authorities is better gauged by considering news organizations and the Sina Headline News since they replicated official messages which were then heavily reposted. Over half of the cases were exposed by the authorities through government/Party bodies directly, or media organizations indirectly.
Yet, news organizations and the Sina Headline News were not merely official propaganda organs. They also represented the voice of the people by publishing a small number of messages that were not official announcements while staying within official boundaries. This role likely helps to reinforce their dominance as opinion leaders and hence also helps them play an effective role as propaganda organs.
Potential challengers to the official propaganda are most likely to be news workers. News workers initiated close to 14 per cent of the cases and their posts all drew robust responses from Weibo users. This might be owing to their expertise in identifying and then pursuing good stories; it might also be owing to their status in the social network on Weibo. Although they trailed behind news organizations and the online media as agenda setters or disseminators, they were stronger than other user categories. Through active and quality posting, individual users from other categories have played prominent roles in a minority of cases. Verified independent journalist, Ji Xuguang 纪许光, who has earned himself a reputation by exposing scandals, initiated the Lei Zhengfu case on Sina Weibo. His initial post was based on information published by another independent journalist who first exposed the case elsewhere online. His posts, which included photographs of sexual acts, were heavily reposted. “Famous grassroot” Zuo Yeben 作业本, renowned for his humorous and insightful comments, published the most reposted messages of all the cases. The success of these individuals seems better explained by what they do as individuals rather than their social positions as a particular user category.
Unlike “idea starters” on Twitter, “initiators” in this study did not necessarily exert strong influence in the retweet chain. Identifying them, however, provides an understanding of how different user categories tried to use the Sina Weibo platform to express their concerns about certain issues. This differentiates initiators from users who exert influence at other stages of information dissemination. In the absence of previous relevant studies, this study used the grounded theory approach to define the agenda setting and disseminator periods of the cases. The results show that the pattern of dominance is consistent across the two periods. This gives confidence that the boundary drawn between the two periods has not affected the results. The consistent pattern found from using two different measures (reposts and comments), as well as two different indicators (top users by rank versus top users by percentage of reposts/comments), further supports the validity of the results.
This study relies on real data collected from the Sina Weibo API and user interface. The impact of censorship on public opinion formation is demonstrated concretely. In at least seven of the cases, the initial post was not accessible on the user interface. In many cases, the number of reposts shown decreased over time, suggesting that some of the reposts might have been deleted as time passed. Censorship poses a culturally specific challenge to the study of Chinese media. A usual practice of scholarship is to discard incomplete data. This study took a different decision and included censored cases on the basis that censorship forms part of the user experience of Weibo subscribers, and public opinion formation can only be studied if these practices are included in the study. Through various searches on the user interface at different times, it was discovered that some messages that were not accessible at one point could become visible at a later time. So were certain accounts. Given the erratic pattern of censorship, what is accessible varies at different times for different users. It is assumed that the data accessed by this study on the user interface is typical of the experience of an average user. As such, results presented here should be taken as a good indication of the landscape of public opinion formation on Sina Weibo over an issue of great concern to both Chinese citizens and the authorities.
Despite the recent decline in the usage of Weibo, it remains a significant public platform where Chinese citizens can express and discuss their views on public issues. With the cutback in service provision by other companies, Sina Weibo stands out as an even more significant Weibo platform than before. The corruption exposés studied in this project are considered indicative of a new trend in which the party-state regains its dominance in public communication. Whether influential users concerning the issue of corruption also hold influence over other topics, as found in studies on Twitter, will need to be tested in studies on other topics on Weibo. What is certain is that the influence of the party-state as an online opinion leader cannot be adequately gauged without taking into account its structural control over the news media, and of course, the entire mechanism of internet censorship and control.
Appendix
Coding of Poster/Reposter Identity
Development of the categorization scheme and the coding was done by the first author. Categories from Bolsover and Xia were integrated into a preliminary coding scheme and applied to the identified opinion leaders. When the existing categories were felt to be inadequate, new categories were added. After coding all the identified opinion leaders, the entire classification scheme was reviewed and all the opinion leaders re-coded using the evolved coding scheme.
Categorization was made on the basis of the user's identity description on the user's home page on Sina Weibo, where available. Where the user's description suggested more than one identity, the account was categorized according to the user's source of income, as far as possible. Where identity description was unavailable or insufficient, searches were conducted on the internet of the account name for additional information.
Appendix Table 1: Categorization of Opinion Leaders
