From denial to accountability: Co-evolution of government communication and public sentiment in a social-mediated crisis in China

Abstract

This study examines the interaction and co-evolution of government crisis communication and public sentiment in China, using the 2022 chained woman incident as a case. Analyzing five government statements and 1375 Weibo comments, it integrates quantitative techniques (correlation analyses, log-ratio transformation, and Grey Relational Analysis) with qualitative interpretation. Results show that government strategies shifted from defensive to accommodative, and public sentiment remained largely negative but significantly improved. While early denial and inconsistency undermined credibility, later accountability contributed to sentiment recovery. The study highlights the government–public interplay and the role of public expectation, offering a process-oriented perspective for social-mediated crisis communication.

Keywords

government crisis communication social-mediated crisis SCCT (Situational Crisis Communication Theory)public sentiment Weibo

Introduction

A crisis is an unexpected and non-routine event or series of events that generate high uncertainty and threaten an organization’s goals or reputation (Seeger et al., 2003). Coombs (2007) conceptualized two broad areas of crisis management: government and corporate, with this study focusing on the latter. Government-related crises have become increasingly visible and consequential in recent years, as allegations of misconduct and governance failures frequently emerge in both traditional and social media (Wu, 2018). Domestically, these crises undermine public trust and tarnish government image (Cai et al., 2009). Internationally, local whistle-blowing cases circulate further and damage national image. Effective communication is thus crucial to manage uncertainty and restore credibility.

The rise of social media has amplified both the visibility and impact of government-related crises. Traditional top-down information flow is increasingly replaced by new modes of multi-center and multi-directional radiation (Lin et al., 2016). The circulation of user-generated content on social media can make it a breeding ground for crises (Cheng, 2016). It enables interaction between governments and publics, and reshapes the landscape of public governance (Cui and Tong, 2022). In China, governments primarily engage with publics through WeChat and Weibo. WeChat is the world’s largest standalone mobile APP with 1.414 billion monthly active users as of December 31, 2025, and Weibo 588 million.

In early 2022, two Chinese women made headlines in both domestic and international media. The New York Times (2022) reported: “Who is the Real China? Eileen Gu or the Chained Woman?”. The former is the two-time Olympic gold medalist, while the latter, surnamed Yang and married Dong, is a mother-of-eight with mental illness. Ironically, the incident originated from a government-organized charity activity to help Dong’s financially disadvantaged family. The family was then visited by a vlogger who posted a video showing Yang standing near a rundown shack with a chain running through her neck. The two-minute video quickly went viral, garnering 1.92 billion views (Cao and Feng, 2022). Public speculation about human trafficking arose due to Yang’s non-local accent and unusually large family. In response, the local county government issued two statements to explain. However, these mishandled statements, which denied trafficking, triggered a subsequent online storm. City and provincial governments then intervened with another three statements to update information; however, the apparent inconsistency between these statements and the previous two further intensified public skepticism. The evolving communication makes the case well suited for examining the dynamic government–public interaction.

This case embodies features of a typical crisis: unexpectedness, time pressure, and threat to established values (Barton, 1993). Yang’s sudden public exposure and the subsequent online criticism caught the local government off guard, as reflected by the hasty first two statements. The later-proven human trafficking and dereliction of duty intensified the moral and legal stakes, further amplified by rapid social media dissemination (Zhang et al., 2021). Governments had to mobilize resources to confront the counter-narrative online.

Crisis communication research largely falls into two groups: the organizational perspective and the public perspective. The former has been considered the dominant one (Fraustino and Liu, 2022), focusing on how organizations navigate crises using frameworks like Image Repair Theory (IRT) (Benoit, 1997), Crisis and Emergency Risk Communication (CERC) (Reynolds and Seeger, 2005), and Situational Crisis Communication Theory (SCCT; Coombs, 2007). However, this organization-oriented approach has been criticized for a “managerial bias,” as it prioritizes reputational protection while underestimating the agency of publics in shaping crisis trajectories, particularly in the social media era (Heath, 2010). Accordingly, recent scholarship has increasingly shifted toward the public perspective. Jin and Liu (2010) proposed the blog/social-mediated crisis communication model (BMCC/SMCC) based on SCCT. With the availability of large-scale social media data, researchers explore public sentiment and behavioral response using computational methods. These studies offer valuable insights into public engagement during crises.

However, the two strands have largely progressed in parallel: organization-oriented studies prioritize strategy effectiveness and reputation outcome, while public-oriented research emphasizes public response. However, the rise of social media has transformed crisis communication from a largely one-way, organization-centered model into a more interactive and decentralized process. This shift necessitates integrating two perspectives, particularly in prolonged social-mediated crises where communication unfolds as a dynamic and adaptive process. Thus, the limited use of processual and relational approaches linking the two perspectives constitutes an important gap in current crisis communication scholarship.

This gap is particularly pronounced in government-related crises, where legitimacy and public trust are central concerns. Admittedly, there are few studies on macro-level government-public communication. For example, Weng and Ye (2013) outlined four communication patterns, with public-driven government response being the most dominant (83.52%), and close two-way interaction rare (1.1%). However, micro-level analyses, such as how government messages and public responses interact, and how such interplay shapes crisis trajectories, remain limited.

In view of the above-mentioned gap, this study adopts a process-based interactional perspective to examine the co-evolution of government communication and public sentiment, taking Yang’s case as an example. It draws on the SCCT framework to unravel how heterogeneous voices from governments and netizens contested and negotiated, and how governments at different levels responded to public questioning, appeased anger, and restored image. Specifically, three questions are answered:

RQ1. Government strategies: What crisis communication strategies did governments adopt? How did strategies evolve as the crisis unfolded? How did later statements respond to earlier ones?

RQ2. Public sentiment: What public sentiment did government statements elicit, and how did sentiment evolve over time?

RQ3. Government–public interplay: How did netizens perceive government responses, and how did strategy evolution correlate with public sentiment change? How did the interaction shape the crisis communication trajectories?

Literature review

Two perspectives of crisis communication studies

As mentioned in the “Introduction,” the organizational perspective has long dominated crisis communication research, with the SCCT being among the most widely applied frameworks. Building on Benoit’s IRT (1997) and Weiner’s (2006) attribution theory, Coombs (2007) proposed SCCT, linking strategy choice with crisis clusters (victim, accident, and preventable). Identification of clusters proffers a preliminary assessment of responsibility attribution. Coombs (2023) further outlined four strategic postures: denial (e.g., attacking the accuser, denial, and scapegoating), diminishment (excusing and justification), rebuilding (compensation, apology), and bolstering (reminding, ingratiation, and victimage). SCCT offers guidance for choosing strategies according to specific crisis types. For instance, compensation is suitable when a visible victim exists, denial is limited to rumor crises, and scapegoating should be avoided in preventable crises.

With the rise of social media, publics can witness, perceive, and respond to crises more readily (Ji and Kim, 2021). Social-mediated crises feature rapid information diffusion, wide public mobilization, and high emotional intensity (Cheng, 2016). Crisis communication in such contexts is marked by continuous interaction and renegotiation between organizations and stakeholders/publics. Moreover, in the temporal dimension, public perceptions and expectations are fluid in online crises. As such, research has expanded from the organizational to the public perspective, examining the emotional contagion, opinion polarization, and collective sense-making processes in online environments (Zhang et al., 2020). In particular, macro-level quantitative methods like computer modeling and simulation have enabled scholars to track large-scale emotional dynamics and public reactions (Zhao et al., 2022). For example, Topic modeling has been used to explore comparative agendas of public interest (Pinto et al., 2019). There are also public sentiment analyses based on sentiment dictionaries and deep learning probing into netizens’ communicative behavior (Yang et al., 2025).

These studies are meaningful for predicting public reactions, improving early warning systems, and informing future crisis management (Liu et al., 2024). However, they often examine public reactions without systematically linking them to organizations’ communication strategies. In social-mediated crises, public attributions and expectations change dynamically, shaped by both official information and online discourse. Meanwhile, governments recalibrate strategies in response to emerging evidence and public sentiments. Neglecting this interplay risks a fragmented understanding of crisis communication in the social media era. Together, these call for more relational and process-oriented approaches. In this vein, the present study conceptualizes crisis communication as an evolving and interactive process in which government communication and public responses co-construct each other. By doing so, it moves beyond traditional strategy-centered studies and foregrounds the co-evolution of government communication and public sentiment in social media environments.

Government crisis communication in the Chinese context

Compared with corporate crises, government-related crises involve broader concerns like public accountability, governance legitimacy, social justice, and moral responsibility, making them highly sensitive and politically consequential. Governments are held to a higher standard, and failures may generate stronger emotional reactions and moral implications (Chon, 2019). Government crisis management is also subjected to closer public scrutiny (Liu et al., 2010).

Research on government crisis communication in China has examined response patterns, strategy selection, and communication effects. For patterns, Li and Han (2014) outlined government response as discursive (like explanation, and acknowledgment of transgression), behavioral (like launching investigations, administrative intervention, and punishment), and institutional (like amendment of relevant law provisions). Similarly, Sharma et al. (2023) classified trust repair strategies into verbal and behavioral. Regarding strategy selection, meta-analytic evidence suggests that corrective action is the most effective response, whereas denial is the least, although the latter remains most widely used and the former ranks third (Arendt et al., 2017). Empirical studies in China report similar patterns: during the early COVID-19 outbreak, governments often employed bolstering strategies to mitigate blame (Li et al., 2021), while case studies of three emergencies highlight a preference for informing and corrective action (Zhang et al., 2020). Instances of cover-up, including restricting information and avoiding direct response, have also been documented (Huang et al., 2016).

Studies of communication effects show mixed outcomes. In some cases, the communication is well received with negative sentiment receding without escalation, while in others not (Fang and Wang, 2012). Effective approaches typically involve sincere concerns and proactive attitudes, apologies and compensation, thorough investigation and timely disclosure, and remedial actions like punishing perpetrators and holding officials accountable. For example, although preventable crises usually provoke intense public anger, the 2023 Qiqihar gymnasium collapse incident illustrates that reputational damage can be mitigated through prompt acceptance of responsibility, transparent investigation, and visible corrective action. In contrast, ineffective communication often involves delayed responses, defensive rhetoric (denial or cover-up), scapegoating, or shirking responsibility (Peng and Yao, 2014). Communication effects are influenced by event types (Yang et al., 2017), strategies used (Fang and Wang, 2012), government levels (Zhang and Sun, 2025), and the timeliness and quality of responses (Wang et al., 2023). For example, a timely response (within 1 day or even 1 hour) may fall short due to compromises in thorough investigation (Wang et al., 2023), as observed in Yang’s case. The number of responses ranges from 2 to 8, with an average of 4.6 (Peng and Yao, 2014), although correlation analyses indicate no significant relation between response frequency and communication effects (Yang et al., 2017).

Overall, these studies provide valuable insights into government management of public emergencies in China but tend to exhibit a managerial bias. Most research foregrounds the government perspective, focusing on how authorities manage information to protect institutional reputation and leaving the public underrepresented. When the public is involved, it is often treated as a variable influencing strategic optimization or a benchmark against which communication effectiveness is measured, rather than as an active agent shaping crisis trajectories. Current social-mediated crisis communication inherently involves continuous contestation and mutual adjustment among stakeholders. However, empirical studies capturing such dynamic processes remain limited in the Chinese context, where distinct institutional structures, political norms, and cultural expectations shape communication strategies and public reactions. Compared with Western settings, crisis communication in China is further conditioned by more centralized governance and heightened expectations of governmental responsibility, producing distinct patterns of authority–public interaction (Huang et al., 2016).

Furthermore, prevalent studies focus on disasters, corporate misconducts, and public health emergencies with high injuries (Chen, 2009; Tian and Yang, 2022). The scope of crisis events awaits expansion. Crises involving governance issues, law issues, and moral violations, such as human trafficking, have received less attention. These crises often generate increased expectations for justice and accountability, posing unique challenges for government communication in a context that values social stability and harmony (Cui and Tong, 2022). These crises thus offer valuable opportunities to examine how public expectations reshape communication strategies and how government responses adapt to public pressure.

Against this background, the present study examines the dynamic government–public interplay in the highly controversial 2022 chained woman incident. The case not only involves crimes like human trafficking and unlawful detention but also exposes governance issues such as defective oversight and bureaucratic inertia. It also reflects the deficiency and confusion in grassroots governance in China’s early years, systemic maladies like gender imbalance and involuntary bachelors in rural areas. As a preventable crisis with strong responsibility attribution, Yang’s case triggered intense outrage and prolonged scrutiny online. By tracing the co-evolution of government statements and public sentiment over time, this study seeks to address the research gaps and questions mentioned in the “Introduction” and provide a more process-oriented and interactional understanding of how the Chinese government responds to online controversies.

Methodology

Data collection

Both government statements and their comments were collected. To avoid possible tampering caused by sharing and reposting, all five statements were collected from their original platforms: the first two from the official WeChat of Feng County Publicity Department, the third and fourth from the official Weibo of Xuzhou People’s Government Information Office, and the fifth from the official Weibo of China Central Television News Center. The original plan was to collect the comments from the same source, yet the official WeChat allowed no comments. Thus, comments on the first two statements were collected from Weibo reposts with the most accessible comments: the first one from Xinjingbao (the official Weibo of Beijing News, China’s first cross-regional newspaper keen on reporting hot-spot issues), and the second from the official Weibo of Global Times, a state media outlet with international recognition. Accessible comments on each statement numbered 295, 573, 477, 384, and 279. Comments consisting solely of punctuation or ambiguous emoji were excluded, leaving the smallest group with 275 counts. To ensure equal sample size, the most “liked” 275 comments on other statements were collected. All 1375 comments were collected and coded manually.

Coding

Coding of government statements

The five statements and their Weibo comments were coded separately using different procedures. To ensure rigor and intercoder reliability, a professor specialized in public relations studies was invited to join in. Before coding, all statements were carefully read to extract the main points and strategies used. Sentences were taken as the coding unit, and coding was mutually exclusive across categories. Statement coding involved identifying sub-strategies and then grouping them.

In this process, the sub-strategy of future commitment emerged, which was not covered by SCCT but observed in Statements 2, 3, and 5. As the event unfolded, the expression of future commitment became lengthier, more specific, and sincere. For example, a vague commitment to further investigation and accountability was made in Statement 2. The names of the suspects and government agencies involved were specified in Statement 3. These commitments serve to reassure publics of improving governance.

With agreement on the main features of each strategy, we randomly sampled 30% of sentences in each statement (n = 48) for pre-coding and calculated Krippendorff’s alpha to test the intercoder reliability. With another three rounds of revisions and reconciliations, Krippendorff’s alpha ranged from 0.85 to 1 with an average of 0.96, acceptable for content analysis (Neuendorf, 2002). The second step was to categorize the strategies. For this purpose, we referred to SCCT (Coombs, 2023) and concurred upon four major categories: denying, diminishing, corrective actions, and rebuilding. Then, for a broader perspective, illuminated by Coombs’ (2007) categorization of primary (stressing denying) and secondary (emphasizing bolstering) strategies, we grouped all strategies into defensive and accommodative categories, as presented in Table 1.

Table 1.

Strategies observed in five statements (with examples).

Strategies	Features	Quotes from statements
Defensive strategies
Denying
Dismissing speculation	Rejecting netizen speculation	There was no human trafficking. (Sta. 1)
Shifting blame	Scapegoating or finding excuses for misbehavior	Neighbors say that Yang often beats children and elders for no apparent reason. (Sta. 2)
Diminishing
Bolstering	Active response, like launching investigation	The joint investigation team went to Yunnan and other provinces, visited over 4600 people, and read over 1000 archives. (Sta. 5)
Acknowledging transgression	Admitting dereliction of duty and malpractices	Government workers for marriage registration did not closely examine Yang’s identity. (Sta. 2)
Accommodative strategies
Corrective actions
Compensation	Assisting victims financially or in other ways	The Civil Affairs Department has offered a subsistence allowance to Yang and her family. (Sta. 4)
Accountability	Holding traffickers and officials responsible	Feng County People’s Procuratorate has approved the arrest of Dong for maltreating family members. (Sta. 5)
Rebuilding
Clarifying doubts	Clarifying netizens’ doubts	As to the query why the birth of Yang’s eldest son predates her marriage, in fact it was another man with the same name as the son. (Sta. 5)
Information disclosure	Detailed fact disclosure	The investigator compared the DNA of Yang and that of Li’s mother, and found no biological parentage. (Sta. 3)
Expert endorsement	Citing results by authorities	The public security institutions compared the DNA of Yang, her half-sister, and her deceased mother, confirming her identity. (Sta. 4)
Future commitment	Commitment to future efforts	Future investigation findings will be announced in due course. (Sta. 3)

Coding of Weibo comments

For coding the comments, we referred to Ki and Nekmat’s (2014) categorization of public sentiment: negative, positive, and neutral. After reviewing all 1375 posts, we observed a preponderance of negativity, accompanied by occasional neutrality and limited positivity. Negative mood includes questioning, skepticism, anxiety, anger, and cynicism (Ki and Nekmat, 2014). The spectrum of neutrality ranges from expressing patience for further investigation to advocating for more protection of women. Positivity found only in comments on the last two statements indicates acceptance and trust. Examples of comments are included in the qualitative analysis in section “Results and Discussion.”

Results and discussion

The following section presents both quantitative and qualitative analyses. The quantitative analysis visualizes the dynamics of strategy use and public sentiment across five statements, employing stacked bar and line chart, Spearman correlations, and Grey Relational Analysis (GRA) to identify co-evolutionary patterns and enhance reliability through triangulation. The qualitative analysis interprets the content of government statements and netizen posts, linking public expectations to government communication strategies and sentiment shifts. These approaches substantiate each other, highlighting both the temporal dynamics and the underlying mechanisms of government adjustments and public perceptions. To begin with, the main points of all statements are summarized in Figure 1.

Figure 1.

Main points in five statements (“Sta.” for “statement”).

Quantitative analysis

Statistical display

The strategy use in each statement is presented in Table 2, followed by the strategy evolution in Figure 2. Table 3 presents the Weibo sentiment across statements, including pairwise comparisons between adjacent sentiment groups.

Table 2.

Strategy proportion.

Strategies	Sta. 1% (n)	Sta. 2% (n)	Sta. 3% (n)	Sta. 4% (n)	Sta. 5% (n)
Defensive strategies
Denying
Dismissing speculations	20% (1)	8.3% (2)
Shifting blame	20% (1)	16.7% (4)
Diminishing
Bolstering	40% (2)	8.3% (2)	11.8% (2)
Acknowledging transgression		8.3% (2)			3.4% (3)
Accommodative strategies
Corrective action
Compensation	20% (1)	29.3% (7)		66.6% (4)	4.5% (4)
Accountability				16.7% (1)	25.8% (23)
Rebuilding
Clarification					6.7% (6)
Information disclosure		8.3% (2)	41.2% (7)		38.2% (34)
Quoting experts		8.3% (2)	23.5% (4)	16.7% (1)	18.0% (16)
Future commitment		12.5% (3)	23.5% (4)		3.4% (3)
Total	100% (5)	100% (24)	100% (17)	100% (6)	100% (89)

Figure 2.

Strategy evolution.

Table 3.

Pairwise comparison of five sentiment groups.

Sentiment	Sta. 1 % (n)	Sta. 2 % (n)	Sta. 3 % (n)	Sta. 4 % (n)	Sta. 5 % (n)
Positive	0% (0)	0% (0)	0% (0)	5.1% (14)	3.6% (10)
Neutral	4% (11)	6.5% (18)	8.7% (24)	24% (66)	33.8% (93)
Negative	96% (264)	93.5% (257)	91.3% (251)	70.9% (195)	62.6% (172)
	Sta. 1 & 2: χ² = .834, p>0.05
		Sta. 2 &3: χ² = .294, p>0.05
			Sta. 3 & 4: χ² = 14.287, p<0.01
				Sta. 4 & 5: χ² = 2.308, p>0.05
	Sta. 1 to Sta. 5: χ² = 66.868, p<0.001

As shown in Figures 1 and 2 and Table 2, the local county government initially adopted defensive strategies of denial (40%) and diminishing (40%). Despite the use of corrective action (29.3%) and rebuilding (29.1%), Statement 2 dismissed human trafficking again and continued scapegoating. The city government then intervened with Statement 3, shifting toward rebuilding (88.2%) mainly through information disclosure regarding Yang’s identity and early experience. Statement 4 reiterated this information while emphasizing corrective action (83.3%), including compensation and accountability. Eventually, 25 days after the first statement, Statement 5 was issued, underlining rebuilding (66.3%, mainly information disclosure) and corrective actions (30.3%, mainly accountability) with a minor proportion of admitting past mistakes (3.4%).

Table 3 presents the temporal distribution of Weibo sentiments across five statements. Overall difference using Fisher’s exact test was to examine whether sentiments differed significantly across statements, while pairwise comparisons using chi-square or Fisher’s exact tests were conducted to explore public perception in a staged manner. Results show that although negativity dominated throughout, there was a significant overall shift (χ² = 66.868, p < 0.001). In particular, positive sentiment appeared with Statement 4, and the change between Statements 3 and 4 was significant (χ² = 14.287, p < 0.01).

Correlation analysis

Given that the core of this study is to examine the government–public interaction across statements, a multi-step analytical procedure was designed (Yin, 2018). (1) A stacked bar and line chart (Figure 3) visualizes the temporal dynamics of strategies and sentiments. (2) Spearman’s rank correlation analyses were performed to assess the strategy–sentiment associations, including contemporaneous (Figure 4) and lagged correlation (Figure 5). The former was to examine the immediate relationship between statements and sentiments (Statement t ↔ Sentiment t, n = 5). Lagged analyses were to examine how sentiments related to subsequent strategies (Sentiment t ↔ Statement t + 1, n = 4), and also how strategies affected sentiments in the next round (Statement t ↔ Sentiment t + 1, n = 4). Together, these correlation analyses enable a comprehensive bidirectional relationship. (3) Given the compositional nature of the data (i.e., proportions summing to 100%), log-ratio transformations were applied to mitigate spurious correlations, followed by a robustness check using Spearman correlation to see if the results are consistent (Aitchison, 1986) (Figure 6). (4) GRA was performed as a complement to further assess the strength and consistency of associations (Figure 7), thereby enhancing reliability through triangulation (Deng, 1989). Together, these methods provide a triangulation that enhances reliability.

Figure 3.

Strategy and sentiment dynamics.

Figure 4.

Contemporaneous Spearman correlation.

Figure 5.

Lagged Spearman correlation.

Figure 6.

Robustness check with log-ratio transformation.

Figure 7.

Grey relational analysis (GRA).

It should be noted that, given the small sample size (n = 5 or 4), this study did not employ regression analysis, VAR models, Granger causality tests, SEM, or CCA. Instead, Spearman’s rank correlation was adopted, as a non-parametric method suitable for small samples that does not rely on strict distributional assumptions and is appropriate for detecting monotonic relationships in exploratory settings. The very small sample also means that emphasis is placed on the direction and magnitude of effect sizes (r_s), rather than statistical significance (p) (Cohen, 1988).

Figure 3 visualizes the correspondence between strategy shifts and sentiment changes. As defensive strategies (denial and diminishing) declined, negative sentiment decreased monotonically, while neutral sentiment increased. The co-evolution suggests that the shift from defensive to accommodative strategies concurs with overall sentiment improvement—primarily from negative to neutral rather than to substantive positivity. Specifically, the high ratio of defensive strategies in the first two statements (80% and 41.6%) coincided with persistently high negative sentiment (96% and 93.5%). The shift toward rebuilding in Statement 3 did not yield significant sentiment improvement, which will be further examined in section “Qualitative analysis.” The most pronounced change occurred in Statement 4, where defensive strategies were fully replaced by accommodative approaches, particularly corrective action, accompanied by the only pairwise significant improvement in sentiment as seen in Table 3. The change indicates the potential effectiveness of corrective action, which will be furthered in qualitative analysis. In Statement 5, strategies reverted to the mixed pattern; although sentiment continued to improve, the change was not significant, and positive sentiment slightly declined. These nuances are also explored qualitatively.

Figure 4 illustrates the direction and the extent to which strategies correlate with sentiments. It shows that defensive strategies are strongly related to all three sentiment types. For example, denial exhibits a near-perfect positive correlation with negative sentiment (r_s = 0.9000, p = 0.037), and diminishing shows a very strong negative correlation with neutral sentiment (r_s = −0.894, p = 0.041). In contrast, corrective action is very strongly and positively correlated with positive sentiment (r_s = 0.894, p = 0.041). No strong correlation was observed for rebuilding (|r_s| ⩽ 0.50, p ⩾ 0.05).

In Figure 5, panel (a) examines how sentiment at time t relates to strategy use at time t + 1, whereas panel (b) assesses how strategy at time t associates with sentiment at time t + 1. Given the very small sample size (n = 4), most p values are statistically insignificant; thus, the interpretation focuses on effect size direction and magnitude. Panel (a) indicates that higher levels of negative sentiment are associated with increased subsequent use of denial (r_s = 0.800) and diminishing (r_s = 0.755), but decreased use of corrective action (r_s = −0.600), suggesting that governments tend to defend themselves when negative sentiment intensifies. Panel (b) shows that denial perfectly positively correlates with subsequent negative sentiment (r_s = 1.000), and diminishing is also very strongly positively correlated (r_s = 0.949). Such correlation indicates a lagged amplifying effect of defensive strategies: their use in one stage is associated with more negative sentiment in the next.

Figure 6 presents a robustness check using the centered log-ratio (CLR) transformation (Aitchison, 1986). As both strategy and sentiment are compositional data, the CLR transformation can mitigate the inherent interdependence issue by converting proportions into log-ratios. Given the very small sample size, combining Spearman’s correlation with CLR provides a more reliable assessment. Zero values are replaced with a small constant (0.5%) following standard practice. Panel (a) reports the original contemporaneous Spearman correlation as in Figure 4, while panel (b) shows the CLR-transformed results. The direction of all 12 correlations remains unchanged, indicating that the findings are not driven by compositional bias. Moreover, some associations become stronger after transformation: the correlation between denial and negativity increases from 0.900 to 1.000, and that between corrective action and positive sentiment rises from 0.894 to 1.000, demonstrating robustness.

Figure 7 presents the GRA heatmap with the distinguishing coefficient set at ρ = 0.5, the most commonly used default value. Unlike Spearman’s correlation, which measures monotonic relationships, GRA assesses the similarity in the overall trajectories of sequences. The two methods are thus complementary, and their convergence enhances the credibility of findings. In GRA, values closer to 1 indicate greater similarity in trends. Sensitivity analyses using ρ = 0.3, 0.5, and 0.7 were also conducted. Results show that rebuilding is most strongly associated with neutral sentiment (0.805), corrective action with positive sentiment (0.797), and denial with negative sentiment (0.678). The ranking of associations remains unchanged across different values of ρ, indicating robust results. The patterns are broadly consistent with the Spearman findings (e.g., denial/diminishing ↔ negative, corrective action ↔ positive), with differences expected given that the two methods reflect distinct dimensions.

The above quantitative analyses capture the government–public interaction and co-evolution. Across five statements, strategies shifted from predominantly defensive to increasingly accommodative, accompanied by declining negative sentiment and rising neutrality. Correlation results show that defensive strategies, particularly denial, are strongly associated with negative sentiment, whereas corrective action is linked to positive sentiment. These findings are largely consistent with previous research (Arendt et al., 2017). Lagged correlation analysis further suggests a bidirectional dynamic: heightened negative sentiment tends to elicit more defensive responses, while the use of defensive strategies reinforces subsequent negativity. These patterns are validated by robustness checks using CLR transformation and GRA. The following qualitative analysis further investigates the underlying mechanisms and contextual factors driving these patterns.

Qualitative analysis

For closer analysis, the communication process was divided into four periods: defensive stage (Statements 1–2), contestation stage (Statement 3), remedial stage (Statement 4), and receding stage (Statement 5). The qualitative analysis complements the quantitative findings by examining (1) the evolution of key messages and strategies in statements, particularly how later statements responded to earlier ones, (2) corresponding public sentiments and their evolution, and (3) possible reasons for evolution in statements and sentiments, particularly the significance of public expectation in the government–public interplay.

Defensive stage (Sta. 1–2): Denial—negatively responded

In response to the viral video of Yang in chain, the local county publicity department rushed into the first statement dominated by denial (40%) and diminishing (40%), which is a common initial reaction possibility to shield oneself from possible accusations (Cai et al., 2009). Yet blunt dismissals without compelling evidence mostly turn out ineffective or even counterproductive (Arendt et al., 2017). It is found that the tactics of denial and diminishing are often used in concert with mortification and corrective actions (Frederick et al., 2024). However, Cai et al. (2009) argue that pledging remedial actions amid the distancing efforts of denial adds to the existing suspicion of culpability. In this study, corrective actions of helping Yang and her family were taken, but Weibo comments suggest that these actions are generally perceived as quick fixes.

Weibo comments suggests that public expectations centered on thorough investigation, particularly of Yang’s identity and the much-likely trafficking, full information disclosure, and accountability. Netizens raised questions against the “no trafficking” claim:

Where is Yang from? What about her parents? How did she lose her teeth? Was there domestic violence? Why did she become mentally disturbed? Has she ever been treated? Who gave Dong the right to chain her? Marriage registration requires legal documents of one’s identity, so how did they get married? Was the pregnancy voluntary? Was it not marital rape? How come there are eight children with the birth control policy? Does the local Women’s Federation know about all these?

In addition, blame-shifting narratives, such as “Yang often beats the elders and minors out of no reason,” further intensified public dissatisfaction. Thus, negative sentiment reached 96%, marking the outbreak of the crisis. The correlation between denial and overwhelming negativity was also demonstrated in the quantitative analyses.

In preventable crises, publics tend to vent emotion and assign blame, which increases their information-gathering and truth-seeking behaviors, and heightens the expectations for accommodative responses like apology, admitting wrongdoings, compensation, and accountability (Ji and Kim, 2021). However, Statement 1 offered only compensation, falling short of other expectations. Worse, subsequent developments indicate that the primacy effects of this initial negative sentiment persisted throughout the entire communication, as initial responses tend to color later evaluations (Coombs, 2023). This delayed effect has also been proven in the lagged correlation analyses, which found a close link between previous defensive strategies and subsequent negative sentiment (Figure 5).

To pacify online indignation, Statement 2 was posted 2 days later. Contrary to public expectation, denying and diminishing were not just sustained but also substantiated, despite more use of corrective actions (e.g., helping Yang and her family) and rebuilding efforts (e.g., future commitment). Denial was supported by the negative results from the national DNA database of trafficked/missing population, ignoring the possibility of unreported cases. As for other crucial issues like Yang’s identity and her pre-marriage experience, the team claimed that “Yang was begging on the street and was taken in by Dong’s late father, and then married Dong,” which fueled skepticism and ridicule as found in posts like “What a coincidence! A poor man cannot find a wife, and his father picked up one from the street for him!” Moreover, the team repeatedly linked Yang’s mental illness to the chaining, reinforcing the blame-shifting and further enraging netizens. This pattern is also consistent with the lagged correlation results, which suggest that earlier negative sentiment may reinforce subsequent defensive responses (Figure 5).

Meanwhile, a positive move was acknowledging transgression: the local government admitted that they failed to carefully check Yang’s identity in her marriage registration with Dong and did not fully impose birth control, which was then a national policy. Another silver lining was that the compensation improved from general to specific. However, these efforts were overshadowed by repeated denials, resulting in an insignificant change in public sentiment. Furthermore, the timing further limited the overall effectiveness: released on the 3rd day when public attention typically peaks (Fang and Wang, 2012), the statement failed to meet the heightened expectations for more accommodative responses.

Contestation stage (Sta. 3): Adjustment without accountability—negativity sustained

One week after the second one, Statement 3 was released amid mounting online denunciation and eroding credibility. This time the baton was passed from county to city-level joint investigation team. While this higher-level involvement enhanced perceived credibility, it also raised public expectations for full disclosure and accountability. As for strategies, denial gave way to rebuilding, including information disclosure of Yang’s identity (41.2%), expert endorsement to prove her health condition (23.5%), and commitment to future efforts (23.5%). In Statement 3, the missing jigsaw piece of Yang’s pre-marriage life was located: Yang, born and bred in a remote village in Yunnan (a southwestern province) under the name Xiaohuamei, was brought to Jiangsu (an eastern province 1000 miles away from Yunnan) by a town fellow (proved to be a trafficker in later statements) under the pretext of medical treatment and good marriage. The disclosure intensified public expectations to verify trafficking and ensure accountability.

The shift from denial to rebuilding (mainly fact disclosure) reflects an adaptive response: sustained online pressure compelled governments to move beyond defensive postures, and the need to restore credibility also motivated more accommodative approaches. Meanwhile, updated cross-regional investigations provided factual grounds for narrative adjustment. Against this backdrop, the apparent inconsistency across statements was implicitly rationalized as an outcome of progressive clarification, in which earlier statements were reframed as provisional assessments based on incomplete information.

Successive statements updated information on one hand, while creating inconsistency and trying netizens’ patience on the other. As one post reads: “If this were the first statement, I would have believed it.” From netizens’ perspective, although Statement 3 abandoned denial and disclosed details of Yang’s identity, it still failed to address the core public concern of confirming trafficking and punishing those responsible. Moreover, given the preventive nature of this case, rushing into rebuilding efforts without the much-expected corrective action like accountability proved weak in actual effect. The gap between the heightened expectation following higher-level intervention and the actual limited official response further sustained the negative sentiment. The persistent negativity aligns with the correlation finding that rebuilding alone shows a limited correlation with sentiment improvement (Figure 4). It also echoes the lagged correlation result that the negative carryover effects of earlier defensive strategies extend into the next stage (Figure 5).

The continuous negativity brings us to blame attribution, a typical initial public response in preventable crises. It shapes publics’ attitudinal, emotional, and behavioral reactions in crises (Coombs, 2023). Blame attribution in preventive crises is typically more intense than in victim and accident clusters (Coombs, 2023). In Yang’s case, Weibo comments suggest that the public assigns dual blame: to traffickers and former negligent government staff for causing the tragedy, and to current officials for releasing misinformation and failing to protect vulnerable groups like Yang. Once blame is assigned, accountability is expected. Yet, it remained absent in Statement 3. Such repeated insufficient responses limited the effectiveness of government communication in improving public sentiment.

Remedial stage (Sta. 4): Highlighting correction action—negativity tempered

The turning point came with the fourth statement, released 3 days after the third by the same organization. Unlike its three predecessors, Statement 4 highlights corrective actions (83.3%), which were absent in Statement 3 and used sparingly in the previous two, where actions were limited to compensation. Corrective action has been regarded as the most effective one (Arendt et al., 2017). It avoids explicit expression of guilt while communicating genuine concerns for those affected, and helps mitigate negative sentiment and expedite image repair (Sellnow et al., 1998). Benoit and Pang (2008) also argued for its effectiveness for image repair, a finding supported by other empirical studies, for example, Wu and Xu (2020).

Specifically, a major step forward in Statement 4 was the confirmation of trafficking and the accountability of traffickers. Besides, the endorsement of the Ministry of Public Security on the DNA tests about Yang’s identity reinforced the credibility of the official narrative. Previous studies indicate that in China the trust of government increases as the administrative level rises, that is, national and provincial governments are generally perceived as more trustworthy than local ones (Li, 2022). The combination of national-level expert endorsement and the presence of the much-expected accountability led to the first-time significant pairwise improvement in public sentiment, as reflected in posts like “Finally, there is good news. Justice is served—how satisfying!” This beneficial effect corresponds to the findings in quantitative analyses.

However, the improvement did not resolve the crisis, and this leads to the importance of public expectation in crisis communication. In public relations, expectations are “beliefs of what should happen,” or “reference points against which judgments are made,” or “collective values about appropriateness and acceptability” (Olkkonen and Luoma-Aho, 2015). If public expectations are repeatedly unaddressed, trust erodes, potentially culminating in a legitimacy crisis (Brändström, 2015). As noted earlier, core public expectations in preventable crises are remedial responses such as admitting wrongdoings, compensating victims, and accountability (Coombs, 2023). Admitting dereliction and compensation had been enacted with Statements 1 and 2, and accountability of traffickers with Statement 4. Yet, expectation regarding government accountability remained unmet, as reflected in posts like “Why haven’t the officials who concealed the truth been held responsible?” This gap, plus the unresolved issues, such as claims that Yang is actually Li Ying, who went missing long ago, given that they look much alike in photos, prompted the last statement by the provincial government.

Receding stage (Sta. 5): Fact disclosure and accountability—mixed reaction

The final statement, roughly 30 times longer than the shortest one, covers seven aspects: Yang’s identity (DNA results confirming that she is Xiaohuamei, not Li Ying), her pre-marriage experience, her health condition, the eight children and the birth-control dereliction, conviction of traffickers, medical treatment and assistance, and punishment of 17 officials. This statement employed rebuilding strategies (66.3%) of information disclosure, and corrective actions of accountability (30.3%), alongside minimal use of admitting dereliction (3.4%). Two points stand out: the detailed disclosure addressed netizens’ expectations for thorough investigation and information transparency, and the accountability of negligent officials aligned with the previously unmet expectation. These fulfilled expectations likely contributed to the significant sentiment improvement from Statements 1 to 5 (χ² = 66.868, p < 0.001). In addition, prior research suggests that intense online sentiment often subsides over time in the absence of new triggers (Yang et al., 2017).

Nevertheless, pairwise comparison shows an insignificant change relative to the fourth statement. One possible reason is that, despite the detailed disclosure, Statement 5 is largely a meticulous reiteration and elaboration of Statements 3 and 4. More importantly, accumulated negativity and undermined credibility take time to diffuse and rebuild, even after public expectations are addressed. Weibo comments indicate that remaining negativity stemmed mainly from: (1) the dissatisfaction with Dong’s conviction for unlawful detention and abuse of family members rather than trafficking and raping, which are more severe and carry longer prison terms, as reflected in comments like: “It’s not family member. Dong is no husband, but a rapist!”; (2) the perceived unfairness in contrasting the victim’s irreversible harm with traffickers’ few-year imprisonment, together with the lack of official response to the advocacy for harsher penalties for traffickers, and equal legal responsibility for both buyers and sellers; and (3) the frustration over the absence of information regarding similar cases of suspected trafficking exposed by netizens. The latter two concerns extend beyond Yang’s case to broader issues of the protection of women and legal reforms.

According to Li and Han (2014), government response can be classified into discursive, behavioral, and institutional actions. In Yang’s case, the first two were in place eventually, for example, issuing statements and admitting dereliction (discursive), conducting investigations and taking corrective actions (behavioral). However, with the case concluding and the provincial and national authorities intervening, public expectations extended to institutional change to ward off similar future tragedies, such as stricter anti-trafficking legislation and systemic protection of vulnerable groups like women. While discursive and behavioral responses are retrospective in nature and address immediate concerns, institutional responses are prospective and preventive. Despite their rarity, precedents show that single crises can trigger systemic change—the 2010 wrongful conviction case led to nationwide judicial reform in China. Given such precedents, the eventual absence of institutional changes contributed to the stubbornly high negativity.

Conclusion and implications

This study examines how the Chinese government handled the online public opinion crisis over a trafficked woman. The crisis unfolded with the county government repeatedly denying netizens’ trafficking argument. Crisis management began to go back on track with the city government issuing a new statement to rectify early misinformation, yet the absence of accountability toned down the accommodative effects of this statement and sustained the negative sentiment. Afterwards, Statement 4 came with the highly anticipated confirmation of trafficking and conviction of traffickers, marking a transition and prompting significant sentiment improvement. The final statement disclosed further information and ensured full accountability of both traffickers and officials.

Throughout the case, Yang was reframed from a lawfully married woman to a trafficking victim, and public sentiment improved significantly, though negativity persisted. As noted in the literature review, government crisis communication can be effective or unsatisfactory (Peng and Yao, 2014). Yang’s case aligns more with the latter with governments’ initial defensive attempts, widespread public censure, later adjustments, and ultimately partial recovery of public sentiment.

Despite being based on a single case, this study offers theoretically meaningful insights given its representativeness. First, it moves beyond the organization-oriented and strategy-centered application of SCCT by integrating it with public participation, which is increasingly critical in the digital era. Second, it proposes an interactive, process-oriented analytical framework rather than viewing crisis communication as a series of one-way static snapshots. By foregrounding temporal dynamics, this study demonstrates that communication effects emerge not only through immediate responses but also via delayed and recursive processes. Methodologically, the integration of correlation analyses and staged qualitative interpretation enables the tracing of how government strategies and public sentiment mutually shape each other over time. Finally, by situating the analysis in the Chinese sociopolitical context, the study extends SCCT beyond its predominantly Western settings. In particular, it demonstrates how hierarchical governance structure conditions crisis communication practices, and the interplay between top-down intervention and bottom-up public engagement highlights the context-specific dynamics that call for more nuanced, culturally and institutionally grounded applications of SCCT.

The study also sheds light on how Chinese netizens pursue truth and advocate for social justice within censored digital environments. Different from previous understanding that criticism of the state, leaders, policies are censored online in China, studies found that critiques of governments’ failure to deliver public goods (termed “performance challenge” as opposed to criticism of the leadership of the Party and the central government known as “political challenge”) are often permitted on social media, offering channels to defend public interests (Shao, 2018).

For government crisis management practitioners, the study can also be enlightening, particularly to evade potential Tacitus traps. In this study, the government deserves credit for its public-oriented efforts like speedy response, ever-deepening investigation, admitting past dereliction, and eventual accountability. Yet, as an incident propagated negative sentiment domestically and internationally, lessons can still be drawn as follows.

First, a major handicap is the county government’s underperformance with hasty denials. Weibo comments suggest that public opinion would not have been so negative without them. Apparently, denial may be related to the swift yet superficial investigation. In a deeper sense, the tactic may also be driven by the face-saving culture, the uncertainty-avoidance and risk aversion mentality (Yu and Wen, 2003). Moreover, maintaining social harmony remains a key priority for the Chinese government, creating incentives to conceal possible scandals (Huang et al., 2016). Researchers also noted that a key factor underlying unsuccessful crisis communication in mainland China is that local governments tend to suppress uncertainty, particularly at the outset of crises when the truth is obscured (Cai et al., 2009). However, it is inadvisable to draw premature conclusions, especially in this age of netizen-led investigation. When self-publishing platforms have rendered netizens from onlookers to participants, governments’ pacifying practices may well fall through and agitate crises in both degree and duration (Yu et al., 2017).

Second, in crises for which governments bear responsibility, well-orchestrated accommodative strategies, particularly corrective actions, need to be well in place. Besides the sub-category of compensation, accountability, particularly of high-ranking officials, is significant in image repair. In government-related crises, publics tend to assign blame not only to officeholders who fail to detect and address issues earlier and lead to irreparable consequences, but also to those who steer and mishandle crisis management. A case in point is the vaccine scandal in 2018, in which high-ranking officials were highly blamed for formalism and lax supervision (Huang et al., 2016). Netizens pay close attention and seek to identify possible misbehavior of officials, creating high accountability pressure (Bonner, 2009). Therefore, in later stages of crisis communication, governments often resort to accountability to restore credibility, because it fosters an “us-vs.-them” distinction that distances organizations from “the few bad apples” (Huang et al., 2016).

Third, this study highlights the role of public expectation in government–public interplay in crisis communication. It shows that the extent to which governments address public concerns largely determines the course of crisis communication. If publics (trustors) feels that governments (trustees) failed to meet expected responsibilities, negative sentiments arise. Moreover, the focus of public expectation may change over time. In Yang’s case, it evolved from confirming trafficking to accountability of officeholders, and finally to institutional responses. The fulfilled expectation of accountability with Statement 4 contributed to the improved public sentiment, but the improvement was not significant when the expectation of institutional responses fell short with Statement 5. While other factors also influence public sentiment, the role of public expectation cannot be underestimated.

Fourth, the two-way communicative potential of social media remains to be tapped. Governments and publics are engaged in a co-evolving feedback system, in which each side continuously responds to and reshapes the other. Public participation in online opinion crises functions as a key driver of truth-seeking and social accountability (Cui and Tong, 2022). In this study, each new statement can be seen as an outcome of “control and resistance” (Gleiss, 2015). Governments may consider public unrest as a reputational threat, but if reputational considerations override responsibility to victims and public concerns, crisis communication risks being perceived as inadequate or evasive (Coombs, 2007).

On a final note, the intervention of higher-level governments, as occurred in this case, is commonplace in crisis communication in China. Peng and Yao (2014) found that this pattern accounts for a substantial proportion of government responses (33.3%), alongside others like same-level responses (15.2%), moving between higher and lower-level government (15.2%), and delegation to lower levels (6.1%). Top-down involvement reflects China’s hierarchical governance, in which local governments typically manage frontline communication, while higher-level authorities intervene when crises escalate or exceed jurisdictional boundaries (Ma and Christensen, 2018). Higher-level intervention often enhances credibility in crisis communication, as in Yang’s case. Taken together, this study suggests that online crisis communication in China is shaped by the interplay between hierarchical governance and public engagement, highlighting the importance of institutional coordination and responsiveness to public concerns.

Footnotes

ORCID iD

Jinfang Ma

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by “Fundamental Research Funds for the Central Universities in China” (Grant No. 24CX04007B), “Shandong Province Social Science Planning Project” (Grant No. 23CWYJ36), and “the Scientific Research Planning Project on Higher Education by China Association of Higher Education (CAHE)” (Grant No. 24WY0102).

Declaration of conflicting interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental material

For those who are interested in the data and command files employed in this study, please refer to the author’s webpage on ResearchGate at . A zip folder named “Supplemental materials for JAAS” can be openly accessed and downloaded.

Author biography

Jinfang Ma is an Associate Professor in the School of Foreign Studies at China University of Petroleum (East China). Her research interests include discourse analysis and crisis management. She has publications in SSCI indexed journals like Social Semiotics and Feminist Media Studies.

References

Aitchison

(1986) The Statistical Analysis of Compositional Data. Chapman and Hall.

Arendt

LaFleche

Limperopulos

(2017) A qualitative meta-analysis of apologia, image repair, and crisis communication: Implications for theory and practice. Public Relations Review 43(3): 517–526.

Barton

(1993) Crisis in Organizations. Southwestern Publishing Company.

Benoit

(1997) Image repair discourse and crisis communication. Public Relations Review 23(2): 177–186.

Benoit

Pang

(2008) Crisis communication and image repair discourse. In: Hansen-Horn

Neff

(eds) Public Relations: From Theory to Practice. Pearson, pp.244–261.

Bonner

(2009) Media as social accountability: The case of police violence in Argentina. International Journal of Press-Politics 14(3): 296–312.

Brändström

(2015) Crisis accountability: Ministerial resignations in Sweden. Scandinavian Political Studies 38(3): 301–320.

Cai

Lee

Pang

(2009) Managing a nation’s image during crisis: A study of the Chinese government’s image repair efforts in the “Made in China” controversy. Public Relations Review 35: 213–218.

Cao

Feng

(2022) The mystery of the chained woman in China. Available at: https://kuow.org/stories/the-mystery-of-the-chained-woman-in-china

10.

Chen

(2009) Institutionalizing public relations: A case study of Chinese government crisis communication on the 2008 Sichuan earthquake. Public Relations Review 35: 187–198.

11.

Cheng

(2016) How social media is changing crisis communication strategies: Evidence from the updated literature. Journal of Contingencies and Crisis Management 26(1): 58–68.

12.

Chon

(2019) Government public relations when trouble hits: Exploring political dispositions, situational variables, and government-public relationships to predict communicative action of publics. Asian Journal of Communication 29(5): 424–440.

13.

Cohen

(1988) Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Routledge.

14.

Coombs

(2007) Protecting organization reputations during a crisis: The development and application of situational crisis communication theory. Corporate Reputation Review 10(3): 163–176.

15.

Coombs

(2023) Ongoing Crisis Communication: Planning, Managing, and Responding. 6th ed. Sage.

16.

Cui

Tong

(2022) How does the Chinese government conduct emotional governance over COVID-19? Content analysis of video blogs. Chinese Journal of Communication 16(2): 150–167.

17.

Deng

(1989) Introduction to Grey System. The Journal of Grey System 1(1): 1–24.

18.

Fang

Wang

(2012) Study on the response of government to sudden network opinion crisis events. Journal of Beijing Institute of Technology (Social Science Edition) 14(3): 137–141 (in Chinese).

19.

Fraustino

Liu

(2022) Toward more audience-oriented approaches to crisis communication and social media research. In: Austin

Jin

(eds) Social Media and Crisis Communication. 2nd ed. Routledge, pp.129–140.

20.

Frederick

Schmidt

Pegoraro

(2024) The beginning of a reckoning: An application of situational crisis communication theory and image repair to the national women’s Soccer league. Communication & Sport 12(3): 530–555.

21.

Gleiss

(2015) Speaking up for the suffering (br)other: Weibo activism, discursive struggles, and minimal politics in China. Media, Culture and Society 37(4): 513–529.

22.

Heath

(2010) Crisis communication: Defining the beast and de-marginalizing key publics. In: Coombs

Holladay

(eds) The Handbook of Crisis Communication. Wiley-Blackwell, pp.1–13.

23.

Huang

Cheng

(2016) Crisis communication in context: Cultural and political influences underpinning Chinese public relations practice. Public Relations Review 42(1): 201–213.

24.

Kim

(2021) Public expectations of crisis outcomes in the social media era in China: A communication-mediated psychological mechanism. Public Relations Review 47(5): 1–9.

25.

Jin

Liu

(2010) The blog-mediated crisis communication model: Recommendations for responding to influential external blogs. Journal of Public Relations Research 22(4): 429–455.

26.

Nekmat

(2014) Situational crisis communication and interactivity: Usage and effectiveness of Facebook for crisis management by Fortune 500 companies. Computers in Human Behavior 35: 140–147.

27.

Han

(2014) The intensity and its analysis of the government response. Journal of Northeast Normal University (Philosophy and Social Science) 1(267): 1–8 (in Chinese).

28.

(2022) Decoding political trust in China: A machine learning analysis. The China Quarterly 249: 1–20.

29.

Chandra

Fan

(2021) Unpacking government social media messaging strategies during the COVID-19 pandemic in China. Policy and Internet 14(3): 651–672.

30.

Lin

Spence

Sellnow

, et al. (2016) Crisis communication, learning and responding: Best practices in social media. Computers in Human Behavior 65: 601–605.

31.

Liu

Horsley

Levenshus

(2010) Government and corporate communication practices: Do the differences matter? Journal of Applied Communication Research 38(2): 189–213.

32.

Liu

Zhu

Wang

, et al. (2024) The evolution of online public opinion on earthquakes: A system dynamics approach. Humanities and Social Sciences Communications 11: 1–18.

33.

Christensen

(2018) Government trust, social trust, and citizens’ risk concerns: Evidence from crisis management in China. Public Performance & Management Review 42(2): 383–404.

34.

Neuendorf

(2002) The Content Analysis Guidebook. Sage.

35.

Olkkonen

Luoma-Aho

(2015) Broadening the concept of expectations in public relations. Journal of Public Relations Research 27(1): 81–99.

36.

Peng

Yao

(2014) A study on the current situation and countermeasures of the Chinese government’s response to online public opinion. Journal of Beijing Jiaotong University(social Sciences Edition) 14(3): 102–110 (in Chinese).

37.

Pinto

Albanese

Dorso

, et al. (2019) Quantifying time-dependent media agenda and public opinion by topic modeling. Physica A-Statistical Mechanics and Its Applications 524: 614–624.

38.

Reynolds

Seeger

(2005) Crisis and emergency risk communication as an integrative model. Journal of Health Communication 10(1): 43–55.

39.

Seeger

Sellnow

Ulmer

(2003) Communication and Organizational Crisis. Praeger.

40.

Sellnow

Ulmer

Snider

(1998) The compatibility of corrective action in organizational crisis communication. Communication Quarterly 46(1): 60–74.

41.

Shao

(2018) The Dilemma of criticism: Disentangling the determinants of media censorship in China. Journal of East Asian Studies 18(3): 279–297.

42.

Sharma

Schoorman

Ballinger

(2023) How can it be made right again? A review of trust repair research. Journal of Management 49(1): 363–399.

43.

The New York Times (2022) Who is the real China? Eileen Gu or the chained woman? 15, February. Available at: https://www.nytimes.com/2022/02/15/business/china-eileen-gu-peng-shuai-chained-woman.html

44.

Tian

Yang

(2022) Deny or bolster? A comparative study of crisis communication strategies between Trump and Cuomo in COVID-19. Public Relations Review 48: 1–14.

45.

Wang

Liu

Guo

(2023) Research on the “time-degree-effect” governance of online public opinion based on evolutionary game. Management Review 35(8): 315–326 (in Chinese).

46.

Weiner

(2006) Social Motivation, Justice, and the Moral Emotions: An Attributional Approach. Psychology Press.

47.

Weng

(2013) An analysis on the government response to internet political participation for decision-making. Journal of Public Management 10(4): 23–36 (in Chinese).

48.

(2020) Making the most effective strategy more effective: Examining the situational and interaction effects of accommodative CCSs in corporate crises. International Journal of Business Communication 60(2): 1–27.

49.

(2018) Discursive strategies of resistance on Weibo: A case study of the 2015 Tianjin explosions in China. Discourse, Context and Media 26: 64–73.

50.

Yang

Cheng

Liu

(2017) The government response and the resolution of Internet mass incidents. Beijing Normal University (Social Science Edition) 260: 110–124 (in Chinese).

51.

Yang

Wang

Liu

(2025) A study on the evolution of online public opinion during major public health emergencies based on deep learning. Mathematics 13(18): 1–20.

52.

Yin

(2018) Case Study Research and Applications: Design and Methods. 6th ed. Sage.

53.

Tang

, et al. (2017) A multi-agent-based online opinion dissemination model for China’s crisis information release policy during hazardous chemical leakage emergencies into rivers. Online Information Review 41(4): 537–557.

54.

Wen

(2003) Crisis communication in Chinese culture: A case study in Taiwan. Asian Journal of Communication 13(2): 50–64.

55.

Zhang

Sun

(2025) Adept or Inept? Examining the Chinese government’s social media crisis response during the Zhengzhou rainstorm. Journal of Contingencies and Crisis 33(1): 1–14.

56.

Zhang

Nekmat

Chen

(2020) Crisis collective memory making on social media: A case study of three Chinese crises on Weibo. Public Relations Review 46: 1–9.

57.

Zhang

Hao

, et al. (2021) Dynamic assessment of Internet public opinions based on the probabilistic linguistic Bayesian network and Prospect theory. Applied Soft Computing 106: 1–1.

58.

Zhao

, et al. (2022) Modeling and simulation of microblog-based public health emergency-associated public opinion communication. Information Processing & Management 59(2): 1–12.