Pandemics & propaganda: How Chinese state media creates and propagates CCP coronavirus narratives

Implications 

State media is a tool for public diplomacy, and affords states the power to shape narratives. Unlike publicly-funded outlets, state media lack editorial independence from government bodies and institutions. Many countries have long maintained official state media broadcast outlets across television, radio, print, and increasingly social media to communicate their points of view to the world. These media properties are sometimes called “white propaganda” entities—a reflection of the fact that the attribution of the message to the state actor behind it is fully transparent¹United States Department of Defense and United States Joint Chiefs of Staff, Dictionary of Military and Associated Terms: (Incorporating the NATO and IADB Dictionaries) (Joint Chiefs of Staff, 1987), digitized on Google Books here: https://books.google.de/books?id=Rs0auKXnjuoC..

While much of the study of state-sponsored online influence has focused on bots and subversive accounts, this essay focuses instead on the white propaganda capability of the People’s Republic of China on social media, and examines how it has been leveraged in an information conflict around the 2020 novel coronavirus pandemic. Understanding how overt online propaganda properties are developed and leveraged to shape international public opinion provides us with a more complete grasp of the narrative manipulation capabilities available to well-resourced state actors, and suggests potential gaps in tech platform misinformation policies. 

The perception of China’s handling of the novel coronavirus has been a messaging challenge for the Chinese Communist Party’s (CCP). There have been a significant number of allegations among Chinese people on China’s domestic social media platforms²Daniel Victor, “Panic and Criticism Spread on Chinese Social Media Over Coronavirus,” The New York Times, January 24, 2020, https://www.nytimes.com/2020/01/24/world/asia/china-social-media-coronavirus.html., as well as in the global press, that the CCP mishandled the crisis and covered up the outbreak’s severity. To manage the PR crisis, the CCP has attempted to control the narrative³Annabelle Timsit, “China Mobilizes against Media’s ‘Malicious’ Coronavirus Coverage,” Quartz, March 5, 2020, https://qz.com/1812162/china-mobilizes-against-medias-malicious-coronavirus-coverage. and deflect blame⁴Bethany Allen-Ebrahimian, “In Global PR Campaign, China Deflects Blame for Coronavirus,” Axios, March 11, 2020, https://www.axios.com/beijings-coronavirus-propaganda-blitz-goes-global-f2bc610c-e83f-4890-9ff8-f49521ad6a14.html. since early in the pandemic, domestically and abroad. It has done this by drawing on its substantial state media apparatus.    

China has extensive and well-resourced outwardly-focused state media capabilities (Brady, 2015), which it employs for its public diplomacy strategy (Chang and Lin, 2014). These channels, such as the CCP’s properties on Facebook (which is banned in China), relay the government’s messaging to other countries’ governments and citizens. Since 2003, building and buying media properties has been part of the CCP’s explicit effort to ensure that it has the capacity to “nudge” foreign governments and other entities into policies or stances favorable to the party⁵Louisa Lim and Julia Bergin, “Inside China’s Audacious Global Propaganda Campaign,” The Guardian 7 (2018): 12.. In periods of unrest or crisis, these properties are put to use to propagate state messaging (Shambaugh, 2017). 

Understanding the ways in which online propaganda shapes public opinion—particularly given the rising prevalence of social networks as sources of news, and the capabilities that social media offers for targeting, repetition, and audience-building—is critical to understanding how influence and manipulation play out in modern politics (Woolley & Howard, 2017). It is, however, a challenging undertaking because of the difficulty of isolating any particular account or post as the precipitating factor in shaping an opinion. A debate persists on the impact of online disinformation and misinformation even in the literature on the most widely-studied operations, such as those carried out by Russia’s Internet Research Agency (IRA)⁶Some scholars suggest the IRA’s activities had virtually no impact because trolls targeted those who were already sympathetic to the message (Bail, 2020), some suggest they may have had an impact, though a negligible one given the size of the social media campaign compared to the far-larger partisan broadcast media ecosystem (Benkler et al., 2018), and still others believe the campaign had a significant impact (Jamieson, 2018)..

Older research assessing the impact of propaganda more generally similarly acknowledges the challenge of quantitative studies. Propaganda scholar L. John Martin addressed this via a distinction between persuasive communication (communication with intent to achieve an objective) and facilitative communication. Facilitative communication “is an activity that is designed to keep lines open and to maintain contacts against the day when they will be needed for propaganda purposes.” It involves building and investing in media infrastructure that leads to the creation of a trusted relationship; successful establishment of the audience is an effect by itself. Martin notes, “[Facilitative communication] is effective if and when it opens up channels of communication with a potential audience” (Martin, 1971). The structure of social platforms lends itself to facilitation, through features designed for rapid, tailored audience-building. Social media account owners receive detailed analytics indicating the extent to which their messaging, targeting, and ads are resonating with the desired audience, which enables precisely-refined strategies. Therefore, understanding white propaganda account strategy is key to understanding the full spectrum of influence capabilities of state actors on social media.

In analyzing the extent to which the CCP’s combined strategy works, communication professor and China propaganda scholar Maria Repnikova notes, “The party’s persuasion efforts are sometimes dismissed as a rigidly ideological top-down affair or a clumsy spectacle, but the government’s handling of the Covid-19 crisis has revealed just how agile China’s propaganda operations really are: They are interactive, and they readily engage with public opinion—the better to co-opt it⁷Maria Repnikova, “Does China’s Propaganda Work?,” The New York Times, April 16, 2020, https://www.nytimes.com/2020/04/16/opinion/china-coronavirus-propaganda.html..” Our assessment of China’s activities on state social media accounts corroborates this observation using quantitative assessments. The behaviors discussed throughout our findings demonstrate ways in which the Chinese government leverages its state media apparatus to shape narratives worldwide: by spinning narratives in a way favorable to the Chinese government, engaging in revisionist history, and paying social media platforms to boost preferred narratives.

We observe long-term audience building via targeted ads, which originally relied on boosting content focused on generic, positive cultural stories. In early 2020, that activity began to involve boosting the promotion of articles related to COVID-19. The COVID-19 content evolved, moving beyond generic positive spin into misleadingly reframed events, and amplification of conspiracy theories. Conspiracy theories serve an important communication function: helping unite the audience (“the people”) against an imagined secretive, powerful elite (Fenster, 1999). By creating and spreading these theories, particularly those framing the United States as responsible for the coronavirus, the Chinese Communist Party presents itself as a defender of its people. It also perpetuates disinformation.

State-sponsored influence campaigns on social media platforms are broader than fake accounts and bots. Addressing overt state media is a policy challenge, not a detection challenge. Given evidence that the CCP has amassed over a hundred million followers on its state media accounts, our findings suggest that although platforms such as Facebook are primarily behaviorally-focused when it comes to taking down misinformation, there is still a need to consider whether affording governments the ability to use ads to push misleading content is a policy gap in counter-misinformation efforts. We believe it is, and that social media platform ad policy requires an update. Some platforms, such as Twitter, have already elected to no longer accept paid state media ads that boost highly slanted coverage; this action was taken in response to Chinese state media promoting tweets misrepresenting events during the Hong Kong protests of 2019⁸@Twitter, “Updating Our Advertising Policies on State Media,” Blog, Twitter, August 19, 2019, https://blog.twitter.com/en_us/topics/company/2019/advertising_policies_on_state_media.html.. Other platforms should adopt this position in the interests of minimizing targeted propaganda and disinformation. Additionally, more research is urgently needed to better assess how this content shapes public beliefs.

Findings

To gain insight into how Chinese state media communicates with the world beyond its borders, we analyzed a collection of posts drawn from its English-language presence on social media using Crowdtangle and the Facebook Ads API (Table 1). Simultaneously, we examined a collection of U.S. mainstream media Pages covering the coronavirus pandemic, as well as U.S. state-sponsored outlets, to assess the media coverage through a comparative lens (Table 1, discussion of selection criteria in Methodology/Appendix A).

	Data set 1: U.S. mainstream media	Data set 2: U.S. government-funded media	Data set 3: Chinese state media	Data set 4: Chinese state media advertisements
Number of Facebook pages	14	7	8	69People’s Daily, China and Global Times did not run any ads classified as political in the period we analyzed
Min – Max(Average) Page Likes10As of March 16, 2020	0.7 – 32.3 million(9.0 million)	0.2 – 11.3 million  (2.5 million)	1.2 – 97.0 million (58.9 million)	N/A
Facebook pages included	ABC News; AP; CBS News; CNBC; CNN; Fox News; Los Angeles Times, MSNBC, NBC News, NPR; POLITICO; The Atlantic; The New York Times; The Wall Street Journal; TIME; Washington Post	Voice of America – VOA; Radio Free Asia; VOA Asia; Africa 54; Radio Free Europe/Radio Liberty; VOA Studio 7; VOA StudentU	CCTV; CGTN; CGTN America; China.org.cn; China Daily; China Xinhua News; Global Times; People’s Daily, China	CCTV; CGTN; CGTN America; China.org.cn; China Daily; China Xinhua News
Date range of posts	December 31, 2019 – March 16, 2020	December 31, 2019 – March 16, 2020	December 31, 2019 – March 16, 2020	January 1, 2019 – March 29, 2020
Total number of posts	74,196 posts of which 10,728 (14%)  included “coronavirus”	5,594 posts of which 1,145 (20%) included “coronavirus”	23,011 posts of which 7,629 (33%) included “coronavirus”11Data Set 3 included 520 additional posts related to the virus’ presence and the experience of citizens in Wuhan, including terms such as #WuhanPneumonia. These were not included in the comparative set, which we limited to “coronavirus”.	146 ads (50 run after Jan 1 2020 including “coronavirus” or “covid”)

In a preliminary analysis of Facebook posts early (January to mid-March 2020) in the COVID-19 outbreak, we observed a positive skew in Chinese state media outlets, which reported positive stories about the outbreak more frequently than U.S. media. After gathering posts, we manually coded a randomly-selected subset across all three outlet types to assess how the outlets differed in their coverage. We found that Chinese outlets shared considerably more positive stories than their U.S. counterparts. 

Finding 1: Chinese outlets focus a significant share of their coverage on positive stories and emphasized Chinese Communist Party competence across a range of narratives. 

To assess differences in framing and sentiment across Chinese and U.S. media, we coded a random sample of 90 posts as positive, neutral, or negative (see Appendix B for examples)¹²See Methods section for a methodological explanation and Appendix B for examples. In the Chinese state media posts, eleven (37%) were positive, whereas only one post (3%) in both U.S. mainstream and government-funded media included positive content (Figure 1). The Chinese state media posts also included far fewer negative posts compared to U.S. outlets, with six posts in the Chinese set (20%) compared to 11 (37%) for U.S. mainstream and seven (23%) to U.S. government-funded media.

To further understand this narrative framing, we examined the use of the word “patient” across all three media environments in all posts they shared¹³In our collection of Facebook posts, “patient” appeared 1626 times in 1151 Chinese state media posts (16% of the total state media posts), 374 times in 344 posts in U.S. mainstream (3%)  and 101 times in 78 of the U.S. government-funded media posts (7%) related to coronavirus.. “Patient” appeared in 16% of Chinese state media posts, but only 3% and 7% in U.S. mainstream and government-funded media respectively. 

The terms “infected” and “coronavirus” were most frequently used consecutively with “patient” across all media outlets. Beyond these two terms, however, there were significant divergences. While the U.S. mainstream media focused reporting on patient illness (using words like  “sick” or “collapsed”), Chinese media more frequently mentioned treatment- and recovery-related terms (“treating”, “recovered”, “discharged” and “cured”) (Figure 2). Although “treat” appeared as a bigram in U.S. government-funded media, so too did “affected,” “died,” and “diagnosed” (Figure 2). Chinese state media focused on recovery whereas the U.S. articles focused on illness severity and death.

1.1 Chinese outlets praise the Chinese government response, while U.S. media highlight failures.

We observed another significant coverage divergence in descriptions of the Chinese government’s crisis response. In the English-language Chinese state media universe, the Chinese government’s response was framed as efficient, effective, and transparent, without deviation. In contrast, the U.S. media highlighted China’s failures.

To evaluate government-response narratives we randomly selected 30 posts that mention keywords related to the Chinese government¹⁴Keywords used: xi, chinese government, chinese leadership, chinese authorities, communist, local authority; Using a space before and after “Xi” to exclude posts including terms such as “taxi”. This generated subsets of 324 Chinese state media posts, 98 U.S. mainstream media posts and 35 U.S. government-funded posts for each of the three outlet types for a total sample of 90 posts. These posts were then coded as unfavorable, neutral, or favorable in their framing of the Chinese government response¹⁵See Appendix C for examples. In the Chinese state media post sample, 13 posts (43%) used a favorable framing, whereas only one post (3%) in U.S. mainstream media and two (7%) in U.S. government-funded outlets framed the Chinese government response favorably (Figure 3). Moreover, there was only one (3%) government-unfavorable post in the Chinese state media sample, compared with 12 (40%) and 11 (37%) in the U.S.mainstream media and U.S. government-funded outlets, respectively. 

The favorable-unfavorable framing dichotomy can be observed in coverage of Xi Jinping. Chinese state media consistently praised Chairman Xi’s efforts to combat the pandemic. In contrast, U.S. media noted that he had retreated from the spotlight during the early outbreak¹⁶Chris Buckley and Steven Lee Myers, “Where’s Xi? China’s Leader Commands Coronavirus Fight From Safe Heights,” The New York Times, February 8, 2020, https://www.nytimes.com/2020/02/08/world/asia/xi-coronavirus-china.html.  and reported on Xi’s lack of public appearances in the context of questions about his leadership, wondering whether he had retreated to avoid “inevitable blame” and speculating on whether prime minister Li Keqiang would take the fall¹⁷James Griffiths, “Xi Jinping Is Directing China’s Response to the Virus. But He’s Been AWOL from News Reports,” CNN, February 2, 2020, https://edition.cnn.com/asia/live-news/coronavirus-outbreak-02-03-20-intl-hnk/h_65d29eb853f1ed84b219680637fc9b44..

We did a similar comparative assessment on a third and final popular topic of coverage: news stories related to Leishenshan and Huoshenshan hospitals that were rapidly built for emergency response in late January 2020. This response effort was presented as a significant event in the coronavirus narrative from Beijing: several million viewers watched online live streams of the construction sites hosted by Chinese state broadcaster CCTV¹⁸William Zheng and Laurie Chen, “Live from Wuhan: Millions Tune in to Watch China Build Coronavirus Hospitals,” South China Morning Post, January 28, 2020, https://www.scmp.com/news/china/society/article/3047943/live-wuhan-millions-tune-watch-china-build-coronavirus-hospitals.. Chinese state media disseminated stories about how the international community was “impressed” with China’s rapid building capacity, calling the quick construction a “miracle¹⁹CGTN on Facebook: https://www.facebook.com/ChinaGlobalTVNetwork/posts/3946024018771722.”  In contrast, both U.S. mainstream and government-funded media reported on the rapid building of the hospitals as a response to overwhelmed medical facilities²⁰ Voice of America on Facebook: https://www.facebook.com/voiceofamerica/posts/10157384977428074, and noted that the temporary structures could hardly be characterized as hospitals ²¹NPR on Facebook: https://www.facebook.com/NPR/posts/10158917095001756(Figure 4).

1.2 Chinese state media maintained a favorable framing of the CCP response, adjusting the justification for lauding the CCP response over time.

Over the course of the analyzed period, Chinese state media maintained a favorable framing of the Chinese government response at all times. Articles from early in the pandemic (January 2020) claimed that China’s government had “selflessly” contained the virus through decisive action thus saving the world State media outlets declared local victory over the virus and pointed to  China’s efforts, facilitated by its superior governance system. An editorial published in the China Daily on February 20 boasted: “Were it not for the unique institutional advantages of the Chinese system, the world might be battling a devastating pandemic”²²“China’s Unprecedented Measures Prevent Novel Coronavirus Infecting the World: China Daily Editorial,” China Daily, February 20, 2020, https://www.chinadaily.com.cn/a/202002/20/WS5e4e7a7ca310128217279147.html..

As the disease spread, the narrative shifted to China had bought the world time to prepare for the pandemic. This narrative appears after Xi specifically instructed Chinese media on February 3 to make China’s response look heroic (Xi, 2020). On February 4, Xinhua ran the headline “Swift, decisive, transparent, cooperative—China buying world time in fight against coronavirus epidemic”²³Yage Guo, “Xinhua Headlines: Swift, Decisive, Transparent, Cooperative—China Buying World Time in Fight against Coronavirus Epidemic,” Xinhua, February 4, 2020, http://www.xinhuanet.com/english/2020-02/04/c_138755660.htm.. On the same date, the Facebook post with most interaction shared by U.S. mainstream media in our data set was from CNN²⁴CNN on Facebook: https://www.facebook.com/cnn/posts/10160421933651509 on whistleblower doctor Li Wenliang and his reported COVID-19 infection; we discuss Li Wenliang in section 2.1.

As deaths outside China began to outnumber those reported from within, focus shifted toward coverage highlighting China’s international shipments of personal protective equipment and delegations of Chinese medical experts.

Finding 2: Chinese state media revised, eliminated, and fabricated aspects of narratives to bolster the image of the CCP.

Our second finding is that beyond simply applying positive spin and cheering on the Chinese government, the English-language Chinese state media content revised, eliminated, and outright fabricated aspects of narratives in order to bolster the image of the CCP.  This is best exemplified in two cases: that of a whistleblower doctor, and the narrative surrounding the origin of the coronavirus.

2.1 Whistleblower doctor Li Wenliang was first silenced, then celebrated

Li Wenliang, an ophthalmologist, was one of the first medical professionals to report the existence of COVID-19. He issued a warning to fellow medics in a chat group on December 30, 2019, which was shared widely within Chinese social media. On January 3, 2020, police detained Li and forced him to sign a letter stating that he had made “false comments.” Li died of COVID-19 on February 7, 2020. The substantial public anger²⁵Verna Yu, “‘Hero Who Told the Truth’: Chinese Rage over Coronavirus Death of Whistleblower Doctor,” The Guardian, February 7, 2020, https://www.theguardian.com/global-development/2020/feb/07/coronavirus-chinese-rage-death-whistleblower-doctor-li-wenliang. that erupted over his death created a dangerous moment for the CCP regime²⁶Melinda Liu, “How Do You Keep China’s Economy Running With 750 Million in Quarantine?,” Foreign Policy (blog), February 24, 2020, https://foreignpolicy.com/2020/02/24/xi-jinping-coronavirus-china-covid-19-quarantine., and Chinese state media had to delicately cover his story without casting the government in a bad light.

In our dataset, 24 posts by Chinese media, 24 posts by U.S. mainstream media, and seven posts by U.S. government-funded media contain “Li Wenliang”. The treemaps below represent most common words used in these articles²⁷Treemaps include the most common terms after filtering out words such as “the” using the stop_words function (see “Methods” section) and excluding the terms “coronavirus”, “li”, and “wenliang”. For both types of U.S. outlets, all words shown in the treemap were used three or more times in these posts. For Chinese outlets, all words in the treemap appeared four or more times. This higher threshold is due to the length of Chinese outlet posts increasing the overall word count, resulting in 72 words used two or more times for Chinese outlets producing an illegible treemap.. 

In U.S. mainstream media, prominent terms associated with Li include “warning,” “authorities”, “silenced”, and “reprimanded”; U.S. government-funded media mentions “free” and “speech” as well as “1989”, the year of the Tiananmen square protests (Figure 5). Such terms are notably absent in Chinese media, which instead focus on him being an “ophthalmologist.” There is no mention of the whistleblower controversy or his detention early in the outbreak. 

2.2 The origin of the coronavirus: First unambiguously Wuhan, then various layers of uncertainty culminating in conspiracy theories.

Chinese state media ran a variety of stories about the origin of the pandemic over a period of six weeks (Figure 6), gradually reframing the narrative towards increasing ‘uncertainty’ about COVID-19’s origins. In its earliest coverage, the origin of the outbreak was unambiguously presented as Wuhan, and the uncertainty simply focused on what unknown animal had transmitted the virus. The uncertainty broadened to where it had originated. Finally, the media began to amplify conspiratorial speculation on whether the virus could have in fact been carried to China by outsiders – specifically, by U.S. military personnel who had taken part in the Military World Games in Wuhan in November 2019. Figure 6 documents this evolution, in which a combination of verified social government media accounts, press conferences featuring prominent doctors, and state media all amplified the conspiracy theory.

The Chinese ambassador to the United States, Cui Tiankai, refuted Zhao Lijian’s comments on March 22, calling it “crazy” to spread a conspiracy theory about a U.S. military link to the coronavirus origin²⁸Swan, Jonathan, and Bethany Allen-Ebrahimian. “Top Chinese Official Disowns U.S. Military Lab Coronavirus Conspiracy.” Axios, March 22, 2020. https://www.axios.com/china-coronavirus-ambassador-cui-tiankai-1b0404e8-026d-4b7d-8290-98076f95df14.html.. However, official Chinese state media channels such as the Global Times  continued to spread it. An April 26 headline read: “US military victim or spreader of virus?”²⁹Chinese state officials were not alone in their willingness to turn to speculation. U.S. Senator Tom Cotton, for example, claimed on January 30th that coronavirus did not originate in Wuhan’s Huanan seafood market and floated the theory that it was potentially bioweapons research tied to a nearby lab, a theory which was dismissed by experts. However, neither mainstream U.S. media nor U.S. government-funded media covered or promoted these allegations on their Facebook properties. This devolution into outright conspiracy attracted widespread public attention. China scholars saw this example as indicative of Beijing’s increasingly assertive tone on social media since 2019³⁰Zhaoyin Feng, “The Year China Got Louder on Social Media,” BBC News, December 29, 2019, sec. China, https://www.bbc.com/news/world-asia-china-50832915., and a shift from the strategy of simply disseminating positive stories about China³¹Nahal Toosi, “In Response to Trump, China Gets Mean,” POLITICO, December 8, 2019, https://www.politico.com/news/2019/12/08/china-trump-twitter-077767.. 

Finding 3: Chinese state media pays to amplify high-priority narratives. 

Chinese state media Facebook Pages used paid amplification—ads—to grow followers and increase the audience for their content. Both the volume and tone of the ads changed when the outbreak began. 

3.1 Chinese state media ads target English-speakers worldwide.

Facebook ads provide accounts with a way to push posts to people who do not already follow them, and can be targeted to viewers by country, interest, or other demographic criteria. Regional targeting data can indicate a Page owner’s geographic audience prioritization, and examining the content of the ads offers insights into prioritized topics. 

We looked at advertisements run by the English-language Chinese state media Pages in our data set from January 1, 2019 to March 29, 2020. The ads appear to be geographically targeted across a range of regions (Figure 7) . Facebook’s Ads API returned lists of cities and regions in which users were served the content. Figure 7 illustrates the regions and volume of ads; the many regions suggest that paid promotions are helping to cultivate a global audience for Chinese state media outlets.

3.2 Chinese State Media significantly increased ads after January 2020, with most new ads promoting Xi Jinping’s pandemic response.

Beginning in January 2019, our Chinese state media pages ran a total of 146 advertisements, spending between US$28,500 and US$63,054 and amassing between 80.48 million and 91.60 million impressions. ³²The number of impressions generated by an ad represents the number of times any Facebook user has seen a specific advertisement. Users may see multiple ads on their feed; this cumulative number of impressions thus does not necessarily mean that between 80.48 million and 91.60 million individual users have seen an ad by a Chinese state media page.In addition, these numbers are low estimates for two reasons: The Facebook Ad API shows impressions per ad as a range, e.g. 25,000-29,999 impressions. The lower number in our estimate equals the sum of the lower end for each advertisement. For advertisements with 1 million or more impressions however, only a lower bound is specified; in these cases, we used the lower bound (1 million) as the upper bound as well. Therefore, the total number of impressions may be even higher than calculated here. 33 of the 65 advertisements in our analysis had 1 million or more impressions. In addition, for some locations, the designation of political Facebook advertisements as “political” is optional. There is evidence of cases of such undeclared political advertisements by Chinese state media on Facebook regarding the coronavirus pandemic. Because non-designated advertisements are unavailable in the Facebook Ads API, the total number of political advertisement expenses and impressions may thus be significantly higher than our analysis shows. For more information, see the “Methods” section. Most of these ads focused on boosting general stories related to Chinese culture.

Nearly half of the advertisements in our data set (65/146 advertisements, or 45%) were created between January 1, 2020 to March 29, 2020, during the coronavirus outbreak (20% of the overall observation period; see Figure 8). 77% of these ads mention “coronavirus” or “covid.” The coronavirus-related subset of ads accounts for between US$12,100 and US$18,250 of the total ad spend, amassing between 36.72 million and 38.19 million impressions.

Stories of Xi’s leadership figured heavily in the coronavirus-related ads: His name was mentioned in 32 of 50 (64%) ads. The ads pronounced him China’s leader in the “battle” against COVID-19, quoted his speeches (“For the Chinese government, people’s safety and health always come first”)³³CGTN Advertisement in the Facebook Ad library: https://www.facebook.com/ads/archive/render_ad/?id=220154975798124&access_token=EAAHlbs2DhZCABAGvx7LmB3U4iZCsMcB8xtCqnISdZA9v38MfWdCi9WxZBN9ZAZAUsKfsYMDNXogeAIIdAdZCJ1nEyztA3m5Hf9iITny0RZADoRbEes82BblA7EosomMaC5WjKDS2qVPJZC5tKZBnbC0RL3BUfKXp9sVdWQ2zkMzZCdHC78DDPZCNeZAcB, and reported on his visits to Wuhan and Huoshenshan hospital. 

While this is not a large ad spend in the context of many types of marketing campaigns, the increase in the rate of advertising, and shift in topical focus from generic cultural stories to coronavirus and Xi-related content, suggests a push to distribute the CCP’s point of view to the broader, ad-targetable world as the information battle over the pandemic began. None of the U.S. government-sponsored media properties have run any ads during 2020; Voice of America ran 34 ads in 2019. 

Given what we observe in our findings—and given the stated commitment of social network technology companies to mitigate influence operations—clearer disclosure of state media communication, and the cessation of paid boosted posts from such entities, appears in line with platform goals. 

Methods

Because of Facebook’s international popularity and accessible data, we chose to focus our research on an English-language collection of Facebook Pages (see Appendix A for full discussion of media properties).

We created lists of Facebook pages for English-language Chinese state, U.S. mainstream, and U.S. government-funded media outlets using Crowdtangle, a social monitoring platform. The U.S. mainstream media pages included on this list were selected on the basis that they, too, are English-language media that post frequently and reach an audience of English-language audiences worldwide. To incorporate U.S. media narratives across the political spectrum, we selected major U.S. outlets and a news wire agency. We separately analyzed two U.S. government-funded outlets for an additional comparative angle.

To generate our data sets, we downloaded all Facebook posts shared by the pages on our three lists on Crowdtangle between December 31, 2019 and March 16, 2020. This resulted in 7,629 posts for our Chinese, 10,728 posts for our U.S. mainstream and  1,145 posts for our U.S. government-funded media list. Our data set included data related to user interactions with each post, as well as follower growth data for the Pages themselves. 

Sentiment analysis on posts related to the coronavirus pandemic is complicated by the fact that various terms coded as “negative” in common sentiment analysis libraries (such as the word “patient” or “hospital”) can be used in both a positive and a negative context, such as by reporting on recovered patients or on patients who have passed away. Therefore, we employed other methods such as analyzing bigrams and having two people each manually code a random selection of posts for each outlet as neutral, negative, or positive to examine a possible divergence in positivity of coronavirus-related coverage across U.S. and Chinese media outlets.  Positive, neutral, or negative indicates coverage of a positive, negative, or neutral story or framing of the pandemic. For example, a post referring to the coronavirus as “dangerous” or “deadly” was coded as negative, whereas a post reporting on a recovered patient was coded as positive. We have included examples of posts and their coding in Appendix B. 

When analyzing bigrams, we removed common words such as “a” and “the” by using the lexica included in the stop_words function from the tidytext package in R. This lexicon includes 1,149 commonly-used words.³⁴Silge, J., and D. Robinson. Tidytext: Text Mining and Analysis Using Tidy Data Principles in RJ Open Source Softw. 1 (3), 37 (2016), doi:10.21105/joss.00037 We used the same approach to remove commonly-used words for creating the treemaps for “patient” and on Li Wenliang.

To track the social media spread of the narrative involving the origin of the virus, we used CrowdTangle Search and TweetDeck. We additionally used the Facebook Ads API to analyze a set of paid advertisements run by these Chinese state media properties, to determine which coronavirus-related narratives they considered a priority to boost. The Facebook Ads API generated a data set of all the ads run by our list of Chinese media outlets that were classified as “political” from January 1, 2019 to March 29, 2020. This provided us with insights into audience growth and desired target audience for each page.