Multilevel government crisis communication on social media: A comparative and longitudinal approach

Abstract

Agenda-setting is an important way for the government to capture public attention. Different government branches, with their specific target audiences, may have different agenda prioritization strategies. In the current study, we compared the agenda-setting process across seven countries and three levels of government branches during the COVID-19 pandemic. We employed a multilingual topic model to extract the agendas from social media data. Our results showed that countries with similar media and political systems are consistent in terms of their agenda patterns. Executive actors prioritized health issues more than legislative actors. Over time, there has been a noticeable increase in the attention given to health information, whereas interest in political issues has declined across all levels of government in most countries. Our findings illustrate an overarching landscape of government agenda-setting during a prolonged crisis, highlighting the importance of incorporating crisis research and a longitudinal perspective into comparative political communication frameworks.

Keywords

Government crisis communication pandemic agenda-setting multilingual topic modeling social media

Introduction

Online communication is an important way in which governments maintain public relations and drive policy initiatives (Sanders et al., 2011). During crisis events, governments aim to minimize public risk, making it essential to allocate public attention and guide public behavior (Sellnow and Sellnow, 2023). To achieve this, government communication often emphasizes specific topics to address public concerns (Sjölander-Lindqvist et al., 2020). This strategic selection and prioritization of issues, aimed at shaping public focus and perception, is known as government agenda-setting (Jones and Baumgartner, 2005; Kingdon and Stano, 1984). Unlike agenda-building, which emphasizes the influence of key actors on shaping communication priorities (Strömbäck and Kiousis, 2019), agenda-setting reflects government's efforts to coordinate responses and messaging during emergent situations.

Managing crisis events typically involves communication from different perspectives. Both crisis-specific and political perspectives are crucial (Johansson et al., 2023; Wang et al., 2021). On the one hand, crisis-specific information, for example, safety instructions, is important so that citizens can better protect themselves (Sellnow et al., 2017). On the other hand, communicating political and societal decisions, such as financial support and relevant restrictions, has been crucial to the daily lives of citizens. In emergent situations, the urgency of the public's demand for information increases, requiring the government to provide timely and accurate updates to mitigate the impact of the crisis (Gruber et al., 2015).

Social media (SM) has emerged as an important tool to meet the growing demand for timely government communication (DePaula et al., 2018). Its capacity for rapid information dissemination, wide reach, and interactive features enables governments not only to communicate their priorities effectively but also to facilitate public engagements (DePaula et al., 2018; Graham et al., 2015). Over the past decade, platforms such as X (formerly Twitter) and YouTube have been integrated into government crisis communication strategies across a variety of contexts, including public health crisis (e.g., COVID-19 pandemic; Hyland-Wood et al., 2021), natural disasters (e.g., earthquakes; Cho et al., 2013) and public safety threats (e.g., terrorist attacks; Kimotho and Nyarang’o, 2019). While challenges such as the spread of misinformation (Yu et al., 2025a) and the politicization of crisis messaging (Zhou et al., 2023) persist, these issues do not undermine the critical role that social media plays in supporting effective crisis response (Tsao et al., 2021).

Even though government crisis communication and agenda-setting have been the subjects of numerous studies (Tsao et al., 2021), the literature has revealed a research gap. First, most of the publications were limited to single case studies (e.g., Loiti-Rodríguez et al., 2021; Wang et al., 2021) or examined only a limited number of political or cultural contexts (e.g., Rivas-de-Roca et al., 2021; Sjölander-Lindqvist et al., 2020). Because of the varying impacts of crisis events across different countries, a broader comparative perspective is needed. Second, most of the publications have primarily focused on executive actors at the national level (e.g., Reyes-Bernard et al., 2021; Sjölander-Lindqvist et al., 2020), whereas legislative and regional (i.e., subnational) actors have received less scholarly attention (Bergström et al., 2022). Because different government branches have distinct roles and responsibilities in crisis management, the complexity of the government hierarchy should be taken into consideration. Third, analyzing the dynamics of government agendas during crises can provide valuable insights into how government priorities evolve in response to changing circumstances (Drerup and Birkner, 2022). Therefore, an extended longitudinal perspective is needed to capture the evolving government agenda throughout the crisis.

In this article, we address these research gaps by longitudinally analyzing and comparing multilevel government crisis communication in seven countries, namely Germany (DE), Italy (IT), the Netherlands (NL), Spain (ES), Sweden (SV), the United Kingdom (UK), and the United States (US). This selection of countries reflects different media and political system models (Hallin and Mancini, 2004), which serve as the primary comparative framework for our study. A detailed justification for the country selection is provided in the online appendix¹ (Chapter 1). We use the COVID-19 pandemic during 2020 and 2021 as a reference crisis case because it is a large-scale, long-term crisis that occurred simultaneously across the world, hence offering a unique opportunity to compare government communication in the same context (Hyland-Wood et al., 2021). We use two SM platforms, X (formerly Twitter) and YouTube, as our data sources because of their high relevance for government communication (Hodson et al., 2022; Tsao et al., 2021).

Government agenda-setting: A comparative perspective

Government agenda-setting in various countries

Analyzing global crises requires comparative studies because countries face different challenges and may develop distinct crisis management strategies (Sjölander-Lindqvist et al., 2020). Accordingly, strategically selecting and prioritizing certain issues enable a more targeted response to each country's specific circumstance (Xu and Löffelholz, 2024). Therefore, understanding how governments in different countries set their agendas to address unique challenges is a relevant question for comparative research. In this context, Hallin and Mancini (2004) theorized three models that can systematize comparisons of countries with different media and political systems: the North Atlantic liberal model (e.g., UK and US), the Northern European democratic corporatist model (e.g., DE, NL, and SV), and the Mediterranean polarized pluralist model (e.g., ES and IT).

With the rapid advancement of information and communication technologies, questions have emerged regarding the continued relevance of traditional media and political models in the digital age (Esser and Pfetsch, 2020). However, both quantitative (Humprecht et al., 2022) and theoretical (Mattoni and Ceccobelli, 2018) studies in comparative research have alleviated these concerns. While digital media, particularly SM, has undeniably transformed the modern communication landscape (Mattoni and Ceccobelli, 2018), cross-national and cross-cultural differences remain deeply embedded in established media and political systems (Esser and Pfetsch, 2020; Humprecht et al., 2022). In the context of government agenda-setting, while communication platforms have evolved, the formulation of agendas remains rooted in the institutional and political context of the individual governments operating within specific political systems. Accordingly, existing political structures are more likely to shape how governments communicate on SM (Xu and Löffelholz, 2024), rather than SM transforming those structures.

Hallin and Mancini's models provide a valuable framework for comparing government crisis communication, as each of the models reflects distinct characteristics in both the form of government and the role of the state (Hallin and Mancini, 2004). These institutional differences may shape government crisis management strategies, and subsequently, influence how agenda priorities are set, even in the digital age. In particular, the North Atlantic liberal model is characterized by a predominantly majoritarian government and a liberal role of the state (Hallin and Mancini, 2004). In this model, government power is often concentrated in a single party, and government decisions are not necessarily dependent on the consensus among various political fractions or actors. During crisis events, decisions must frequently be made under time constraints (Hyland-Wood et al., 2021), which may result in limited emphasis on broad political deliberation. At the same time, the liberal role of the state typically implies a limited capacity for direct crisis intervention, instead placing greater emphasis on individual responsibility (Cornia et al., 2016; Hallin and Mancini, 2004). Consequently, government communication agendas are more likely to prioritize crisis-specific instructions that empower individuals to take action for self-protection (Sellnow et al., 2019). This assumption is supported by empirical findings, showing that, during the pandemic, governments in the US and UK emphasized personal protective measures and hygiene practices in their SM communication (Rivas-de-Roca et al., 2021; Xu and Löffelholz, 2024).

In contrast, the Northern European democratic corporatist model is characterized by a predominantly consensus government and a strong welfare state (Hallin and Mancini, 2004). Because of this consensus system, government decision often rely on agreement among multiple political factions (Hallin and Mancini, 2004; Sanders et al., 2011). Therefore, extensive deliberation may play a more central role in Northern European countries than in North Atlantic countries. In parallel, the strong welfare state characteristic of these systems (Hallin and Mancini, 2004) reflects a socially oriented role of the state (Cornia et al., 2016), placing less emphasis on individual responsibility and more on collective public support. Accordingly, government communication agendas are expected to prioritize overarching measures of public assistance rather than individual-level instructions. This assumption is supported by empirical studies on German government SM messaging, where the emphasis was placed on legislative processes and political decisions aimed at delivering broad public assistance (Xu and Löffelholz, 2024).

Mediterranean polarized pluralist model countries are characterized by a mixture of both majoritarian and consensus forms of government (Hallin and Mancini, 2004) and rather moderate welfare states. These institutional features place them in the middle ground between North Atlantic and Northern European models. Accordingly, Mediterranean countries are expected to have a more balanced agenda distribution between political discussion and crisis-specific instructions. Empirical research on government SM communication supports this notion, revealing a relatively balanced distribution between political messaging and individual-focused crisis guidance, particularly in comparison to the other two models (Rivas-de-Roca et al., 2021).

The assumptions derived from the three models also reflect public expectations for crisis information, which are shaped by broader cultural norms related to risk (Cornia et al., 2016). In North Atlantic countries, an individual-oriented risk culture supports a public preference for actionable, personal guidance, reinforcing the emphasis on self-directed crisis measures (Cornia et al., 2016; Xu and Löffelholz, 2024). In Northern European countries, a state-oriented risk culture corresponds with public expectations for coordinated governmental responses and comprehensive support (Cornia et al., 2016; Xu and Löffelholz, 2024). Meanwhile, in Mediterranean countries, a more fatalistic risk culture, characterized by skepticism and limited trust in government competence, encourages governments to keep a balance between individual-focused and collective-focused messaging in order to maintain public trust (Xu and Löffelholz, 2024).

However, these assumptions can be challenged. First, the media and politics models, initially proposed to compare political communication (Esser and Pfetsch, 2020; Hallin and Mancini, 2004), did not explicitly address crisis communication, and empirical studies on SM crisis communication remain limited in scope and quantity (i.e., focusing on one single government SM account; Rivas-de-Roca et al., 2021; Xu and Löffelholz, 2024). To date, it remains unclear whether the assumptions derived from these models are applicable in the context of long-term crises. Analyzing government agenda-setting during the COVID-19 pandemic not only clarifies government communication priorities during a prolonged crisis event but also serves as a way to examine the validity of the media and politics models in the crisis context.

Second, although previous studies have demonstrated clear distinctions in agendas among the three models from an overarching perspective (e.g., Sanders et al., 2011; Sjölander-Lindqvist et al., 2020), the COVID-19 pandemic also provides the opportunity to longitudinally compare agenda consistency in different countries, which has rarely been explored before (Boulianne and Larsson, 2024; Sutton et al., 2021). Therefore, we ask the following question:

RQ1: How did the crisis communication agendas of countries in the three media politics models differ during the COVID-19 pandemic from a longitudinal perspective?

Government agenda-setting at various levels

Government agenda-setting involves the identification, selection, and framing of key issues that governments intend to emphasize (Kingdon and Stano, 1984). In the context of crisis communication, governments should select and prioritize agendas based on the severity of the crisis (Reynolds and Seeger, 2005), the target audience (Graham et al., 2015), and the specific roles of different government branches (Bergström et al., 2022). This could optimize the distribution of information (Sellnow and Sellnow, 2019), along with several other benefits that can lead to successful crisis management. For example, prioritizing regional agendas to local residents would meet people's specific information needs (Graham et al., 2015). In addition, this agenda-setting strategy would avoid confusion caused by the blurred boundaries between certain government branches (Baumgartner et al., 2011; Sellnow and Sellnow, 2023) and minimizes conflicts that may arise in intergovernmental cooperation (Bergström et al., 2022).

A comparison of political crisis communication at different levels could provide valuable insights into how crisis communication agendas are negotiated, coordinated, or contested between various government branches, hence shedding light on the complexity of multilevel governance and its impact on policy outcomes (Bonafont et al., 2020). In addition, this comparison contributes to the refinement of crisis communication theories (Esser and Hanitzsch, 2012) because current frameworks mainly provide overarching guidelines for crisis communication strategies (e.g., Sellnow et al., 2017) while specific government roles or responsibilities are not taken into account. In the current study, we focus on two dimensions of government actors who are important in government crisis communication: executive-legislative actors and national–regional actors (Bergström et al., 2022; Cormack and Meidlinger, 2022).

The executive branch has the primary responsibility for implementing laws and exercises management through leadership in decision-making and strategic coordination (Turnbull and Bernier, 2022). During public crises, the executive branch often takes the lead in public messaging, providing regular updates, reassurance, and guidance to the public (Graham et al., 2015). Nevertheless, the decisions made by the executive branch often depend on support from the legislative branch (Petrov, 2020) because legislators have the main responsibility for lawmaking and the oversight of executive decisions (Petrov, 2020). Members of the legislative branch set agendas to address pressing issues and concerns affecting their constituents (Barberá et al., 2019; Cormack and Meidlinger, 2022), this includes interpretation and explanation of political decisions (Cormack and Meidlinger, 2022).

When comparing crisis communication agendas during COVID-19, the executive branch tends to prioritize actionable information and instructions (Sellnow-Richmond et al., 2023), such as sanitary guidelines, treatments, and medical advancement (Rivas-de-Roca et al., 2021; Zhou and Zheng, 2022). In contrast, the legislative branch does not necessarily prioritize crisis-specific instructions. Instead, it often addresses broader issues surrounding the consequences and side-effects of the crisis (Box-Steffensmeier and Moses, 2021; Zhou et al., 2023), which typically require more deliberation and public debate. Vaccination-related communication offers a clear example of this distinction: while executive actors primarily promoted vaccination campaigns and encouraged public participation (Gherheș et al., 2023; Maarek, 2022), legislative actors were more likely to politicize the issue (Zhou et al., 2023), fostering debate over whether vaccination should be mandated or left to individual choice (Zhou et al., 2023).

However, while our overall comparison framework the three models of media and politics typically take countries as the primary units of analysis (Esser and Pfetsch, 2020; Hallin and Mancini, 2004), the framework does not consider distinctions between government branches as a factor that differentiates the models. In addition to examining whether the role of branches is predominant in the prioritization of communication agendas at their respective levels, another important question is whether government branches within the same model follow the same agenda patterns across time. On the other hand, although countries from the same model share similar political systems (Hallin and Mancini, 2004), differences in the agenda-setting process between government branches have not been empirically examined in the literature. Following this doubt, we pose the next research question:

RQ2: How did government executive and legislative branches in different countries differ in their crisis communication agendas during the COVID-19 pandemic from a longitudinal perspective?

In the context of the national–regional dimension, the roles of and relationships between national and regional governments vary across political and administrative systems (Bergström et al., 2022). However, in general, national governments are primarily responsible for overarching coordination and policy formulation (Berman, 2020), whereas regional governments are primarily responsible for policy implementation, adapting upstream policies to the circumstances of their respective administrative areas (Bergström et al., 2022; Berman, 2020). Therefore, the communication agendas of national governments tend to be broader in scope, prioritizing more general and fewer specific issues (Adeel et al., 2020). During the COVID-19 pandemic, the national government agenda included issues such as the coordination of health resources (e.g., Moreno et al., 2021). In contrast, regional government agendas have a narrower focus, addressing more specific and localized challenges (Drerup and Birkner, 2022), for example, the allocation of health resources.

In addition to examining national and regional government agendas during prolonged crisis, similar to the scenario of executive and legislative agenda comparisons, national and regional government distinctions were not considered as main factors that differentiated the three media and politics models. The interplay between the government hierarchy and three models has not been empirically examined. A similar question needs to be answered: Do national and regional governments within the same model follow the same agenda patterns over time? Together with executive- and legislative-level comparisons, these two pairs of comparisons provide a more nuanced understanding of how media and politics models operate across various governmental structures. We pose the following research question:

RQ3: How did national and regional executive actors in different countries differ in their crisis communication agendas during the COVID-19 pandemic from a longitudinal perspective?

Methods

Data collection and processing

Following a deductive approach (Yu et al., 2025b), we sampled government actors from the seven selected countries, focusing on national executive, national legislative, and regional executive branches. We included key government agencies (e.g., ministries), politicians (e.g., ministers and parliamentarians), and relevant health institutions (e.g., Centers for Disease Control and Prevention). The respective codebook and sampling results are available in the online appendix (Chapter 2). Next, we identified the X and YouTube accounts associated with the sampled actors, and we collected the all posts made by these government accounts during 2020 and 2021 using Twitter Academic API and YouTube Data API. This data collection was conducted in 2022, before the termination of Twitter Academic API.

For all the countries, we used SM posts that were written in their respective official or predominant languages. To ensure comparability and maintain a focus on government communication itself, for data from X, we only used the original tweets in our analysis. Retweets, quoted tweets, and replies were excluded, as they often originate from external sources and may not represent government's own messaging. For YouTube, we concatenated each video's title and description into a single text unit for our analysis. Given that both the title and description tend to be concise and typically highlight the core content of a video (Lai et al., 2024; Lu et al., 2024), this concatenation enhances comparability with the short-text posts on X while preserving essential information from the videos.

The last step of data processing was classifying SM posts as either pandemic-related or non-pandemic-related. This binary classification combined official annotations² from the platforms and a curated list of qualitatively evaluated keywords (Yu et al., 2025b). Specifically, we first created a series of manually coded ground truth data sets (i.e., training sets) to assess the performance of official COVID-19 annotations on X and YouTube in each of the countries. We then identified posts that were pandemic-related but not captured by X/YouTube annotations. From this subset, we extracted the most frequent³ unigrams, bigrams and trigrams, and qualitatively selected the most relevant terms in identifying pandemic-related posts. These terms were later applied as filter words in addition to the official annotations.

We validated the classification results via manual coding (i.e., test sets) for each of the countries and platforms. The details of the classification as well as the precision and recall metrics used for validation are available in the online appendix (Chapters 3, 4, and 5). Ultimately, a total of 858,629 pandemic-related SM posts were analyzed for this research (N_X = 814,786, N_YouTube = 43,483). A summary of the collected data per country, dataset, and platform is available in the online appendix (Chapter 6).

Multilingual topic modeling

Following a common approach in communication and political science research (e.g., Barberá et al., 2019; Gilardi et al., 2022), we used topic modeling to identify government agendas. We employed a zero-shot cross-lingual topic model (Bianchi et al., 2021) to extract the topics from the multilingual SM datasets. In the context of multilingual research, a zero-shot model is trained on one language and can subsequently be generalized to multiple other languages for which it was not explicitly trained (Bianchi et al., 2021). Unlike the traditional bag-of-words approach, this model uses pretrained (multilingual) embedding models to generate contextualized representations of documents (Bianchi et al., 2021).

To train and test the model, we randomly extracted a sample that represented 5% and 20% of the X and YouTube data, respectively (N = 49,437, N_X = 40,740, N_YouTube = 8697). The details regarding the distribution of the sample per country and platform are available in the online appendix (Chapter 7). Because the model requires monolingual data for training, we leveraged the DeepL API to translate all the training data into English. We used 98% of the translated sample as the training set (N_X = 39,932, N_YouTube = 8520) and the remainder as the test set (N_X = 808, N_YouTube = 177).

Following the recommendation of the topic model developers,⁴ we used paraphrase-multilingual-mpnet-base-v2 as the embedding model. We used two metrics to estimate the number of topics (K): normalized pointwise mutual information (NPMI, Bouma, 2009) to measure topic coherence and inversed rank-biased overlap (RBO, Terragni et al., 2021) to measure topic diversity. As a result, three promising models with 10, 15, and 20 topics were selected and qualitatively annotated based on the top 10 highest-probability terms and documents per topic.

We subsequently measured the stability of the models in a multilingual context for non-English texts. Using 100 randomly sampled non-English texts, two sets of topics were computed for each model in both the original non-English text and its English translation. The results showed that the K10 and K20 models slightly outperforming the K15 model.

The topic distribution of the test data was predicted via the three promising models. Following common practices (e.g., Bianchi et al., 2021), an expert-level researcher evaluated the agreement with the predicted topics via a 5-point Likert scale (1 = totally disagree, 5 = totally agree), which yielded average scores of 2.98 (SD = 1.79) for the K10 model, 3.40 (SD = 1.69) for the K15 model, and 3.68 (SD = 1.61) for the K20 model. Considering the model stability and evaluation results, we selected K20 as the final model for our research. We discarded one topic because of its low interpretability. Details regarding the model estimation, selection, and manual validation are available in the online appendix (Chapter 8).

We additionally discarded three other topics (press conferences, press briefing, and government information update) because they were mostly promoting messages with links to, for example, press conferences, and were not directly related to crisis management and communication. Next, we grouped the remaining topics into two domains: one explicitly focusing on health and disease information (HDI) and the other on political and societal information (PSI).

Longitudinal analysis

We aggregated the topic modeling results by the respective publication date. Because of the very different numbers of data entries in different countries, instead of the raw numbers of documents, we used percentages to measure the prevalence of domains. To obtain the overall landscape of topic evolution over the 2-year period, we extracted the trends from our data and analyzed them with a series of linear regression models. We used the slopes of these regression models to show the tendency of the evolution of government agendas over time.

Results

Domain composition and comparison across countries

Overall, the HDI included seven topics that were explicitly related to health information (e.g., scientific explanation and uncertainty reduction), disease prevention (e.g., health suggestions), and health resource availability (e.g., covid test availability). Together, they made up 45% of pandemic-related government communication. The PSI included nine topics related to general political and parliamentary debates (e.g., covid policy and debate), pandemic measures (e.g., lockdown and restrictions), and crisis support (e.g., business and economic support). Together, they made up 55% of pandemic-related government communication. Figure 1 visualizes the overall prevalence of the topics, along with their respective domains. More detailed topic descriptions and example documents are available in the online appendix (Chapter 9).

Figure 1.

Prevalence of the topics. Note: The lengths of the bars (x-axis) represent the relative percentages of the topics in the corpus.

The North Atlantic countries (UK: 48%, US: 52%) emphasized HDI the most. The Mediterranean countries (ES: 41%, IT: 36%) paid the second most attention to the HDI, and the Northern European countries (DE: 26%, NL: 27, SV: 24%) paid the least attention to the HDI. The attention given to PSIs was distributed in the opposite way. The variation in domain distribution was not large between the countries in the same model. The US was the only country for which more than half of the SM content was centered on HDI (Figure 2).

Figure 2.

Comparison of domains across the seven countries (overall and overtime perspective). Note: On the left side, an overall comparison of communication domains across the seven countries is depicted. On the right side, the comparison over time is faceted by country. The x-axis represents the date, and the y-axis represents the percentage of the domain per day.

We describe the main findings of the longitudinal analyses here; however, because of the large number of comparison patterns, details regarding the linear regression models are attached in the online appendix (Chapter 10). In all the studied countries, at the beginning of the pandemic, PSIs were emphasized over HDIs. However, the attention given to the HDI maintained an increasing trend over the 2-year period, whereas the PSI showed a contrasting trajectory. In North Atlantic countries, the attention given to HDIs had already surpassed that given to PSIs by 2020. In the Mediterranean countries, this transition did not occur until 2021. Conversely, in Northern European countries, PSI has always been a greater focus than HDI, despite its declining trend in this regard. Among all the countries, the US had the highest rate of HDI increase (b = 0.05, 95% CI [0.05, 0.05], t(710) = 59.82, P < .001; b* = 0.004, 95% CI [0.004, 0.005]) and the highest rate of PSI decrease (b = −0.05, 95% CI [−0.05, −0.05], t(710) = −61.54, P < .001; b* = −0.004, 95% CI [−0.005, −0.004]).

Comparison between government branches

At the level of national executive actors (red bar), in North Atlantic countries, more than half of pandemic communication focused on HDIs (UK: 62%, US: 53%), whereas Northern European and Mediterranean countries presented the opposite results. At the level of national legislative actors, in all the studied countries, the HDI received very little attention (M = 0.18, SD = 0.05), with actors in North Atlantic countries paying slightly more attention than those in the other countries (the UK and US: 25%). In all the studied countries, executive actors (M = 0.43, SD = 0.13) had a much greater focus on HDI than legislative actors, and for PSI, the opposite was true (Figure 3).

Figure 3.

Overall comparison of the three levels of government communication. Note: The columns represent the seven countries, and the rows represent the two domains.

At the level of national executive actors (red line), in all the studied countries, the attention given to the HDI tended to increase over the 2-year period, whereas the PSI showed the opposite trend. At the level of national legislative actors, the increasing attention given to the HDI showed an increasing pattern similar to that of national executive actors. Although a decreasing focus on PSI was present for most of the countries, in cases such as Mediterranean countries, the communication of PSI was relatively stable over the 2-year period (ES: b = 0.008, 95% CI [0.007, 0.010], t(707) = 9.77, P < .001; b* = 0.002, 95% CI [0.001, 0.002]). IT: b = 0.004, 95% CI [0.002, 0.006], t(718) = 4.10, P < .001; b* = 0.001, 95% CI [0.001, 0.001]). In all the studied countries, the HDI was consistently more emphasized by executive actors, and in most cases (except SV), the PSI was consistently more focused on by legislative actors, despite an overall decreasing trend in this regard (Figure 4).

Figure 4.

Comparison of the three levels of government communication over time. Note. The x-axis represents the date, and the y-axis represents the percentage of the domain per day.

Overall (Figure 3), at the level of regional executive actors, with the exception of DE and NL, more than half of the regional communications were dedicated to HDI, with PSI showing the opposite landscape. The North Atlantic countries had the highest relative proportion (UK: 71%; US: 72%). The number of Mediterranean countries was the second highest. Among the national and regional executive actors, with the exception of those in NL, regional executive actors (M = 0.58, SD = 0.13) paid more attention to the HDI than their national counterparts. For PSI, the reverse was true.

Over time (Figure 4), at the level of regional executive actors, all studied countries paid increasing attention to the HDI. Moreover, the attention given to PSIs continued to decrease in most of the cases, except in NL. Among national and regional executive actors, in North Atlantic countries and Mediterranean countries, during the entire or most of the 2-year period, regional actors prioritized the HDI more than national actors. Among Northern European countries, the attention given to the HDI over time had different patterns: SV regional actors had consistently greater HDI priority than their national counterparts do. DE regional actors had a similar HDI development pattern as national actors, and NL regional actors (b = 0.07, 95% CI [0.07, 0.08], t(99) = 41.30, P < .001; b* = 0.005, 95% CI [0.005, 0.006]) had a faster rate of increase in the attention given to the HDI than national actors (b = 0.04, 95% CI [0.04, 0.04], t(582) = 26.84, p < .001; b* = 0.004, 95% CI [0.003, 0.004]). However, it typically remains a low priority. In the case of PSI, except for NL, the emphasis at the national level was consistently greater than that at the regional level.

Discussion

Our results have illustrated a multifaceted picture of how different government branches responded to a prolonged crisis event on SM. We expand our discussion in three aspects that correspond to our outlined research gaps: multicountry comparison, multilevel government comparison, and longitudinal comparison.

Comparison across countries

Our research results confirm assumptions for government communication derived from the media and political systems models. Countries with similar media and political characteristics exhibited similar agenda prioritization patterns. The differences observed across the three models can be attributed not only to institutional factors, such as the role of the state, but also to varying public expectations for information during crisis, which are shaped by the respective risk cultures (Cornia et al., 2016; Xu and Löffelholz, 2024).

The relatively high prevalence of HDI in North Atlantic countries can be linked to their individual-oriented risk culture (Cornia et al., 2016). This cultural orientation aligns with the liberal role of the state, in which individuals expect specific guidance to ensure their own safety (Sellnow et al., 2019). In this context, the government's primary role is to provide actionable information rather than to coordinate broad societal responses (Cornia et al., 2016). Therefore, the emphasis on HDI not only reflects public expectations rooted in values of autonomy and personal responsibility, but also institutional norms that favor limited state intervention (Hallin and Mancini, 2004; Xu and Löffelholz, 2024).

In contrast, the relatively high prevalence of PSI in Northern European countries can be attributed to their state-oriented risk culture (Cornia et al., 2016). This risk culture aligns with the strong welfare state, in which the public expects robust institutional involvement and comprehensive protective measures (Hallin and Mancini, 2004). As a result, government communication in these contexts tends to emphasize collective coordination, reassurance, political and economic accountability (Xu and Löffelholz, 2024). The focus on issues tied to political decision-making and legislative processes further reflects not only public expectations but also the government's commitment to a broad, socially oriented crisis response (Sellnow et al., 2019).

In Mediterranean countries, the relatively balanced distribution between HDI and PSI in agenda setting can be linked to public skepticism toward government competence (Cornia et al., 2016). This distrust creates a dual communication challenge: authorities must deliver both clear, actionable guidance and visible signals of political support to maintain credibility and foster public trust (Cornia et al., 2016). The observed balance in agenda priorities suggests an effort to address these simultaneous demands. This communication pattern also reflects the moderate welfare state characteristic of the region, where the state plays a less extensive role as in Northern European countries, yet is more engaged as in North Atlantic countries (Hallin and Mancini, 2004).

Beyond this reasoning, our results underscore the importance of considering more crisis research in the comparative analysis of political communication. This is because political communication examines “the flow of information and the exchange of messages among political actors, citizens, and the media” (Esser and Pfetsch, 2020: 337). When a crisis event occurs, the dynamics among the politician actors, citizens and the media might be intensified. This creates a unique context where countries around the world are submerged in highly similar crisis environments (Xu and Löffelholz, 2024), contributing to a more equivalent comparison when studying cross-national political communication (Esser and Hanitzsch, 2012). On the other hand, crisis events, particularly long-term crises, usually pose threats that extend beyond their time frame. Its side effects, such as the proliferation of misinformation (Yu et al., 2025a), remain and become even stronger in the current political environment. Therefore, taking crisis research into political communication frameworks would allow researchers to examine the effects or even the causes, of certain political phenomena, for example, the inseparable links between crisis-related conspiracies and the fast-growing far-right political powers in Europe (e.g., Zehring and Domahidi, 2023).

Comparison between government branches

The main responsibilities of the different levels of government are well reflected by our results because regional and national executive actors placed a stronger emphasis on HDI than legislative actors did and, for PSIs, the reverse was true. Compared with national executive actors, regional executive actors generally place greater emphasis on the HDI. Importantly, in most of the countries, regional executive actors’ agenda prioritization patterns followed the three media and politics models. The only exception was the Netherlands, where regional executive actors paid the least attention to the HDI. This is because, in the Netherlands, health emergencies are dealt with mainly by municipalities and safety regions (in Dutch: Veiligheidsregio, Szmulewicz Ramírez, 2022).

Two points can be summarized from these findings. On the one hand, the commonalities of multilevel government communication across our selected countries show that the varying role of government branches is an important factor in determining their respective agendas. This highlights the importance of considering the government hierarchy when analyzing crisis and political communication (Bergström et al., 2022). On the other hand, because regional executive agenda patterns are mostly consistent with media and politics models, we suggest that, for future comparative studies that involve multilevel government hierarchies, media and politics models would also be valuable frameworks for analysis.

Comparison from longitudinal perspective

Our research results also demonstrated high consistency in terms of agenda prioritization among countries within the same media and politics models and across the three levels of government. Specifically, North Atlantic countries had the earliest transitions of the primary agenda from PSI to HDI. This early shift highlights the proactive approach taken by these governments, in which specific measures and fast responses were seen as the central strategy for crisis management (Sellnow et al., 2019). In contrast, the Mediterranean countries experienced a late transition in the primary agenda. One of the possible reasons is because of the complex sociopolitical landscapes and existing challenges these countries faced (e.g., political turbulence, Casula and Pazos-Vidal, 2021) because governments would ensure political and social stability in the early stages of the severe crisis. This late transition of the communication agenda reveals a rather reactive approach for crisis management, possibly influenced by the deep impact of the crisis on these countries (Sjölander-Lindqvist et al., 2020) and the subsequent need to focus more on specific measures as the crisis unfolded. However, for Northern European countries, no agenda transition was observed over time. This consistent focus on PSIs may be because of the robust health systems in these countries (Blümel et al., 2020), which allows these governments to solely focus on policy deliberation and overarching social support.

Across the three levels of government and the 2-year timespan, the HDI tended to be prioritized more strongly by all levels of government communication over time. However, at the beginning of the pandemic, political decisions and deliberation were the focus of most government communication. Policies such as lockdowns, school suspensions, and business support were widely emphasized by government actors (Wang et al., 2021). In contrast, attention to HDIs remained relatively low during this period, possibly because of the lack of practical medical resources, such as vaccines. As these resources became available, the prioritization of HDIs in government communication gradually increased.

Involving a longitudinal perspective in comparative communication research has been largely encouraged by scholars (Boulianne and Larsson, 2024; Esser and Vliegenthart, 2017) because the media and political systems and the cultures under study are consistently changing over transformation processes (e.g., Americanization, globalization, etc., Esser and Vliegenthart, 2017), and potentially influenced by other factors, such as crisis events, economic growth, or recessions (Esser and Pfetsch, 2020). Employing a longitudinal perspective helps in understanding whether these systems or cultures have been affected in different ways (Esser and Vliegenthart, 2017). Our results do show cross-country differences in terms of agenda prioritization over time; however, the consistency observed in the longitudinal evolution of the primary agenda patterns suggests that the COVID-19 pandemic did not overtake the role of media and political systems in shaping government agendas. This observation implies that a crisis alone cannot fundamentally alter the established patterns of media and political influence; instead, it highlights the resilience of these systems.

Limitations and future research

The limitations of this article encompass potential directions for future research. First, our primary focus has been the comparative analysis of government communication across two domains. A nuanced analysis of government agenda-setting on specific issues is omitted. A promising approach for future research would be to analyze these issues and analyze government agenda-setting in specific contexts (e.g., lockdown and restrictions and the severity of the pandemic). Second, investigating the dynamics of intergovernmental agenda-setting has been outlined as an important research direction (Baumgartner et al., 2006, 2011) in determining whether regional governments follow national government agendas, which would enrich our understanding of government decision-making processes and intergovernmental relations (Baumgartner et al., 2011). Because our study aims to provide an overview of government communication agendas during the COVID-19 pandemic, one path for future research is the exploration of intergovernmental cooperation in the context of setting the government agenda during this global health crisis. Third, even though our results clearly distinguish agenda prioritization patterns in different government branches, because of the limited extension of this article, other aspects that could matter for distinguishing multilevel governance, such as the government system (i.e., presidential or parliamentarian) and system of autonomy (i.e., unitary or federal), were not included in the discussion of the current work. Future studies may consider involving more perspectives to consolidate our results. Finally, the present study focuses on the analysis of high-income countries, future studies may consider to include a broader global perspective to enhance the generality of our findings (Barreto de Souza Martins et al., 2023).

Conclusion

Drawing on a comparative analysis of multi-country and multi-level government agenda setting on SM, this study addresses three essential gaps in the existing literature. First, it affirms the continued relevance of Hallin and Mancini's framework for cross-national comparisons, particularly in the context of prolonged crisis. This provides a valuable reference point for future research focusing on ongoing global challenges, such as climate change. Second, our results highlight the critical role of government hierarchy in crisis and political communication studies. Focusing on a narrow set of actors may overlook the complexity and diversity of government communication practices. Third, while crisis events may temporarily shift agenda priorities, our findings suggest that media and political systems continue to exert a dominant influence over time.

Together, these insights underscore the study's contribution: leveraging large-scale data to systematically investigate government agenda-setting dynamics under crisis conditions. By comparing government agenda-setting patterns across multiple political systems and levels of government, our analysis revisits and tests the applicability of Hallin and Mancini's framework in contemporary digital and political contexts. Our findings further demonstrate how big data can support broader empirical generalizations and help identifying structural differences in agenda-setting processes. Building on these findings, future research could expand the comparative scope to include more countries, exploring how media systems, political structures, and levels of government shape public agendas in an increasingly interconnected and challenging world.

Footnotes

Acknowledgments

Not applicable.

ORCID iDs

Jingyuan Yu

Emese Domahidi

Bengt Johansson

Funding

The research reported in this article was funded by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under project number 458225198.

Declaration of conflicting interests

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

Data availability statement

Data and additional online materials are openly available at the project's Open Science Framework page ().

Notes

References

Adeel

Catalano

, et al. (2020) COVID-19 Policy response and the rise of the sub-national governments. Canadian Public Policy 46(4): 565–584.

Barberá

Casas

Nagler

, et al. (2019) Who leads? Who follows? Measuring issue attention and agenda setting by legislators and the mass public using social media data. American Political Science Review 113(4): 883–901.

Barreto de Souza Martins

Domahidi

(2023) A global health crisis with divided research traditions? A comparative review of Brazilian and international research in communication on the COVID-19 pandemic. Annals of the International Communication Association 47(4): 1–18.

Baumgartner

Green-Pedersen

Jones

(2006) Comparative studies of policy agendas. Journal of European Public Policy 13(7): 959–974.

Baumgartner

Jones

Wilkerson

(2011) Comparative studies of policy dynamics. Comparative Political Studies 44(8): 947–972.

Bergström

Kuhlmann

Laffin

, et al. (2022) Special issue on comparative intergovernmental relations and the pandemic: How European devolved governments responded to a public health crisis. Local Government Studies 48(2): 179–190.

Berman

(2020) The roles of the state and federal government in a pandemic. Journal of Nation Security Law and Policy 11(61): 61–82.

Bianchi

Terragni

Hovy

, et al. (2021) Cross-lingual contextualized topic models with zero-shot learning. In: Proceedings of the 16th conference of the European chapter of the association for computational linguistics: main volume, Online, April 2021, pp. 1676–1683. Association for Computational Linguistics. Available at: https://aclanthology.org/2021.eacl-main.143 (accessed 22 February 2023).

Blümel

Spranger

Achstetter

, et al. (2020) Germany: Health system review. Health Systems in Transition 22(6): 1–272.

10.

Bonafont

Green-Pedersen

Seeberg

(2020) Comparing agenda-settings: The comparative agendas project. In: Handbook of Research Methods and Applications in Comparative Policy Analysis. Cheltenham: Edward Elgar Publishing, pp. 90–112. Available at: https://www.elgaronline.com/display/edcoll/9781788111188/9781788111188.00012.xml (accessed 13 July 2023).

11.

Boulianne

Larsson

(2024) Comparative digital political communication: Comparisons across countries, platforms, and time. Social Science Computer Review 42(5): 1092–1100.

12.

Bouma

(2009) Normalized (pointwise) mutual information in collocation extraction. In: Proceedings of the biennial GSCL conference, 2009, pp. 31–40.

13.

Box-Steffensmeier

Moses

(2021) Meaningful messaging: Sentiment in elite social media communication with the public on the COVID-19 pandemic. Science Advances 7(29): eabg2898.

14.

Casula

Pazos-Vidal

(2021) Assessing the multi-level government response to the COVID-19 crisis: Italy and Spain compared. International Journal of Public Administration 44(11–12): 994–1005.

15.

Cho

Jung

Park

(2013) Social media use during Japan’s 2011 earthquake: How twitter transforms the locus of crisis communication. Media International Australia 149(1): 28–40.

16.

Cormack

Meidlinger

(2022) Congressional communication in a pandemic: “follow the leader” politics and responsive representation. Congress & the Presidency 49(3): 273–298.

17.

Cornia

Dressel

Pfeil

(2016) Risk cultures and dominant approaches towards disasters in seven European countries. Journal of Risk Research 19(3): 288–304.

18.

DePaula

Dincelli

Harrison

(2018) Toward a typology of government social media communication: Democratic goals, symbolic acts and self-presentation. Government Information Quarterly 35(1): 98–108.

19.

Drerup

Birkner

(2022) Themen & Timing. Die Krisenkommunikation der deutschen Landesregierungen in der Corona-Pandemie – eine (teil-)automatisierte Inhaltsanalyse zentraler Kommunikationskanäle. M&K Medien & Kommunikationswissenschaft 70(3): 256–285.

20.

Esser

Hanitzsch

(2012) On the why and how of comparative inquiry in communication studies. In: The Handbook of Comparative Communication Research. New York: Routledge, pp. 3–22. Available at: https://www.taylorfrancis.com/chapters/edit/https://doi.org/10.4324/9780203149102-8/comparative-inquiry-communication-studies-frank-esser-thomas-hanitzsch.

21.

Esser

Pfetsch

(2020) Political communication. In: Comparative Politics. Fifth. Oxford: Oxford University Press, pp. 336–358. Available at: https://www.zora.uzh.ch/id/eprint/186963/ (accessed 28 August 2023).

22.

Esser

Vliegenthart

(2017) Comparative research methods. In: The International Encyclopedia of Communication Research Methods. John Wiley & Sons, Ltd, pp. 1–22. Available at: https://onlinelibrary.wiley.com/doi/abs/10.1002/9781118901731.iecrm0035 (accessed 4 January 2022).

23.

Gherheș

Cernicova-Buca

Fărcașiu

(2023) Public engagement with Romanian government social media accounts during the COVID-19 pandemic. International Journal of Environmental Research and Public Health 20(3): 2372.

24.

Gilardi

Gessler

Kubli

, et al. (2022) Social media and political agenda setting. Political Communication 39(1): 39–60.

25.

Graham

Avery

Park

(2015) The role of social media in local government crisis communications. Public Relations Review 41(3): 386–394.

26.

Gruber

Smerek

Thomas-Hunt

, et al. (2015) The real-time power of Twitter: Crisis management and leadership in an age of social media. Business Horizons 58(2): 163–172.

27.

Hallin

Mancini

(2004) Comparing Media Systems: Three Models of Media and Politics. Communication, Society and Politics. Cambridge: Cambridge University Press. Available at: https://www.cambridge.org/core/books/comparing-media-systems/B7A12371782B7A1D62BA1A72C1395E43 (accessed 14 November 2023).

28.

Hodson

Veletsianos

Houlden

(2022) Public responses to COVID-19 information from the public health office on Twitter and YouTube: Implications for research practice. Journal of Information Technology & Politics 19(2): 156–164.

29.

Humprecht

Castro Herrero

Blassnig

, et al. (2022) Media systems in the digital age: An empirical comparison of 30 countries. Journal of Communication 72(2): 145–164.

30.

Hyland-Wood

Gardner

Leask

, et al. (2021) Toward effective government communication strategies in the era of COVID-19. Humanities and Social Sciences Communications 8(1): 1–11.

31.

Johansson

Ihlen

Lindholm

, et al. (2023) Introduction: Communicating a pandemic in the Nordic countries. In: Communicating a Pandemic: Crisis Management and Covid-19 in the Nordic Countries. Nordicom: University of Gothenburg, 11–30. Available at: https://urn.kb.se/resolve?urn=urn:nbn:se:norden:org:diva-12699 (accessed 10 April 2024).

32.

Jones

Baumgartner

(2005) How government processes information and prioritizes problems. In: The Politics of Attention: How Government Prioritizes Problems. Chicago: University of Chicago Press, pp. 1–34. Available at: https://press.uchicago.edu/ucp/books/book/chicago/P/bo3644482.html.

33.

Kimotho

Nyarang’o

(2019) Role of social media in terrorism crisis communication: A case of westgate mall terror attack in Nairobi. International Journal of Information Systems for Crisis Response and Management 11(1): 65–82.

34.

Kingdon

Stano

(1984) Agendas, Alternatives, and Public Policies. Boston: Little, Brown.

35.

Lai

Brown

Bisbee

, et al. (2024) Estimating the ideology of political YouTube videos. Political Analysis 32(3): 345–360.

36.

Loiti-Rodríguez

Genaut-Arratibel

González

M-JC

(2021) Crisis communication in audiovisual format: information from Spain’s National Health System on YouTube in 2020. Profesional de la información 30(4): 1–19.

37.

Song

Kwon

, et al. (2024) Investigating the coverage of China’s vaccine crisis on YouTube: Networked framing, grassroots activism, and homophily. Journal of Broadcasting & Electronic Media 68(2): 176–197.

38.

Maarek

(2022) Introduction: Similar and dissimilar patterns of government communication on COVID-19. In: Maarek

(eds) Manufacturing Government Communication on Covid-19  : A Comparative Perspective. Cham: Springer International Publishing, 1–12. Available at: https://doi.org/10.1007/978-3-031-09230-5_1 (accessed 23 January 2025).

39.

Mattoni

Ceccobelli

(2018) Comparing hybrid media systems in the digital age: A theoretical framework for analysis. European Journal of Communication 33(5): 540–557.

40.

Moreno

JEC

Bahena

Cruz

(2021) Coordination and crisis management. The case of the federal Government and two sub-national governments in Mexico. Gestión y Análisis de Políticas Públicas 27: 101–114.

41.

Petrov

(2020) The COVID-19 emergency in the age of executive aggrandizement: What role for legislative and judicial checks? The Theory and Practice of Legislation 8(1–2): 71–92.

42.

Reyes-Bernard

Basit

Sofija

, et al. (2021) Analysis of crisis communication by the Prime Minister of Australia during the COVID-19 pandemic. International Journal of Disaster Risk Reduction 62: 102375.

43.

Reynolds

Seeger

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

44.

Rivas-de-Roca

García-Gordillo

Rojas-Torrijos

(2021) Estrategias comunicativas en Twitter y portales institucionales durante la segunda ola de Covid-19: Análisis de los gobiernos de Alemania, España, Portugal y Reino Unido. Revista Latina de Comunicación Social 79: 27.

45.

Sanders

Crespo

MJC

Holtz-Bacha

(2011) Communicating governments: A three-country comparison of how governments communicate with citizens. The International Journal of Press/Politics 16(4): 523–547.

46.

Sellnow

Johansson

Sellnow

, et al. (2019) Toward a global understanding of the effects of the IDEA model for designing instructional risk and crisis messages: A food contamination experiment in Sweden. Journal of Contingencies and Crisis Management 27(2): 102–115.

47.

Sellnow

Lane

Sellnow

, et al. (2017) The IDEA model as a best practice for effective instructional risk and crisis communication. Communication Studies 68(5): 552–567.

48.

Sellnow

(2019) The IDEA model for effective instructional risk and crisis communication by emergency managers and other key spokespersons. Journal of Emergency Management 17(1): 67–78.

49.

Sellnow

(2023) Before Crisis: The Practice of Effective Risk Communication. San Diego: Cognella Academic Publishing. Available at: https://titles.cognella.com/before-crisis-9781793549068 (accessed 25 January 2024).

50.

Sellnow-Richmond

George

Lukacovic

, et al. (2023) A comparative analysis of U.S. state government communication and resident compliance to CDC COVID-19 guidelines. Journal of Applied Communication Research 51(4): 380–403.

51.

Sjölander-Lindqvist

Larsson

Fava

, et al. (2020) Communicating about COVID-19 in four European countries: Similarities and differences in national discourses in Germany, Italy, Spain, and Sweden. Frontiers in Communication 5: 97.

52.

Strömbäck

Kiousis

(2019) Political Public Relations: Concepts, Principles, and Applications. 2nd ed New York: Routledge. Available at: https://www.routledge.com/Political-Public-Relations-Concepts-Principles-and-Applications/Stromback-Kiousis/p/book/9781138484054 (accessed 8 October 2025).

53.

Sutton

Rivera

Sell

, et al. (2021) Longitudinal risk communication: A research agenda for communicating in a pandemic. Health Security 19(4): 370–378.

54.

Szmulewicz Ramírez

(2022) Governing decentralised States under emergency situations: Italy and The Netherlands during the COVID-19 crisis. DPCE Online 54(S.I.): 521–536.

55.

Terragni

Fersini

Messina

(2021) Word embedding-based topic similarity measures. In: Natural language processing and information systems: 26th international conference on applications of natural language to information systems, NLDB 2021, Saarbrücken, Germany, June 23–25, 2021, proceedings, Berlin, Heidelberg, 23 June 2021, pp. 33–45. Springer-Verlag. Available at: https://doi.org/10.1007/978-3-030-80599-9_4 (accessed 8 March 2023).

56.

Tsao

S-F

Chen

Tisseverasinghe

, et al. (2021) What social media told us in the time of COVID-19: A scoping review. The Lancet Digital Health 3(3): e175–e194.

57.

Turnbull

Bernier

(2022) Executive decision-making during the COVID-19 emergency period. Canadian Public Administration 65(3): 538–546.

58.

Wang

Hao

Platt

(2021) Examining risk and crisis communications of government agencies and stakeholders during early-stages of COVID-19 on Twitter. Computers in Human Behavior 114: 1–13.

59.

Löffelholz

(2024) Multimodal crisis messaging in times of pandemic: Comparing Instagram posts published by governments and public health institutions in Germany, Türkiye, the UK, and the USA. International Journal of Communication 18(0): 28.

60.

Domahidi

Benmaarouf

, et al. (2025b) Quantity or quality? Comparing social media data sampling strategies for government crisis communication research. International Journal of Disaster Risk Reduction 124: 105531.

61.

Domahidi

Gamannossi degl’Innocenti

, et al. (2025a) Crisis, country, and party lines: politicians’ misinformation behavior and public engagement. arXiv:2502.15321. arXiv. Available at: http://arxiv.org/abs/2502.15321 (accessed 1 July 2025).

62.

Zehring

Domahidi

(2023) German Corona protest mobilizers on telegram and their relations to the far right: A network and topic analysis. Social Media + Society 9(1): 20563051231155106.

63.

Zhou

Liu

Yang

(2023) Politicization of science in COVID-19 vaccine communication: Comparing US politicians, medical experts, and government agencies. Political Communication 41(4): 649–671.

64.

Zhou

Zheng

(2022) Agenda dynamics on social media during COVID-19 pandemic: Interactions between public, media, and government agendas. Communication Studies 73(3): 211–228.