Sage Journals: Discover world-class research

Abstract

The rapid onset of COVID-19 created a desire for information about the virus and government-directed response efforts. This study explores the sharing of Twitter posts from U.S. government officials and organizations during the onset of the COVID-19 pandemic. This study followed two paths of investigation. First, significant differences in the retweet frequency of government sources were determined based on the source (public officials and agencies, Republican- or Democrat-led states) and differentiated by equal time periods (40 days pre- or post-pandemic declaration). The study findings show that the frequency at which government accounts were retweeted in the U.S. was significantly impacted by the pandemic period, with the post-pandemic period significantly more active as governments wrestled with unprecedented economic, policy, and societal concerns related to the crisis. A greater reliance on public officials as opposed to public agencies or departments was seen, as was a greater tendency to retweet the posts of government officials from Democrat-led states. The second path examined the content of government tweets to determine the linguistic factors which predict retweet frequency. Study results found that both positive and negative emotional language in tweets predicted the frequency of retweeting government sources, though the results varied based on the source. The study results provide insight into the government-affiliated information sources users chose to share with their networks as well as the types of language that appeal to citizens during times of crisis.

Keywords

government social media Twitter COVID-19

Introduction

The onset of the COVID-19 pandemic resulted in widespread societal changes, including social (distancing and restricted activities), economic (increased need for online commerce), professional (remote work), and educational (remote learning), among others. The rapid onset of societal change, coupled with the reactive nature of U.S. government responses to the pandemic, shook an already tenuous confidence in the ability of governments to ensure public safety. The uncertainty of the COVID-19 virus during the early stages of the pandemic resulted in a hunger for information, and social media was an easily accessible outlet for citizens to obtain and share pandemic-related information. As information consumers, social media users were faced with a highly polarized environment rife with politicized, questionable, or false information. In times of crisis, it is not uncommon for citizens to seek out government sources for information. Choosing which government social media sources to share with others may be driven by diverse motivations, including but not limited to familiarity, political leanings, perceived trust or credibility, or even the sentiment or linguistic content of the message. Examining posts from one example social media platform – the microblog Twitter (now X) – can provide insight into the government-affiliated information sources users chose to share (“retweet”) with their networks. Additionally, assessing the linguistic content of retweets can provide insight into types of language that appeal to citizens during times of crisis.

The following paper explores the sharing of Twitter posts from U.S. government officials and organizations during the onset of the COVID-19 pandemic. The paper proceeds as follows. First, a brief overview of trust in public sector institutions and its relationship to e-government and public sector social media is provided. This overview leads to a review of public sector use of social media during crisis events, including citizens’ retweeting of public sector sources during these events. The literature review ends with a set of research questions and hypotheses which explore the public sector sources and linguistic factors which may impact retweet frequencies during the early days of the COVID-19 pandemic. Finally, the research study used to investigate these hypotheses is discussed.

The research study described by this paper follows two paths. First, a corpus of 998,482 U.S.- and COVID-specific retweets were investigated to uncover significant differences in the retweet frequency of U.S. government sources based on the type of source (e.g. public officials, public agencies, state-level sources with Republican leadership, state-level sources with Democratic leadership) and differentiated by the time period (40 days prior to and 40 days immediately after the World Health Organization pandemic declaration). Assessing the relative frequency at which official, pandemic-related information was shared in the early crisis stages can provide insight into which public sources citizens are inclined to share within their networks during public health crises. Second, I examined the content of these government tweets to determine the linguistic factors which predict retweet frequency. Examining the content of tweets can help to explain what linguistic factors may appeal to citizens and, therefore, encourage them to share tweets with their networks.

I find that the frequency at which government accounts were retweeted in the U.S. was significantly impacted by the pandemic period, with the post-pandemic period (March 11-April 20, 2020) being significantly more active as governments wrestled with unprecedented economic, policy, and societal concerns related to the crisis. A greater reliance on public officials as opposed to public agencies or departments was seen, as was a greater tendency to retweet the posts of government officials from Democrat-led states. These results suggest that personalities rather than organizations were more common information sources, and that the prominent role of Democrat-led states in the early response efforts may have had an impact. In terms of tweet content, the results show that the use of both positive and negative emotional language predicted the frequency of retweeting government sources, though the results varied based on the source category.

Background

Trust and E-Government

The concepts of trust and confidence in public sector institutions are often intertwined. For purposes of this study, trust – loosely defined as placing confidence in another person, organization, or other entity, based on positive assessment of intentions or behavior – can be used as a synonym for the concept of political trust. Political trust – i.e. trust in both political actors as well as political institutions to effectively govern for the benefit of its citizens – has long been a subject of political science and social science research (e.g. Abramson et al., 1999; Hamm et al., 2019; Hetherington & Rudolph, 2020; Pressgrove & Kim, 2018; Rudolph & Evans, 2005). Trust in democratic government is considered a leap of faith of sorts, in which citizens commit to being governed, with the expectation that the government will follow through on its stated mission and elected officials will be accountable to the people. Such assumptions underpin the notion of governmental legitimacy (Hardin, 2002).

The recent divisions and populist resurgence in U.S. politics may be based on a lack of trust in institutions of governance; though most visible in the post-2016 environment, trust in democratic institutions has been said to be on the wane since the 1960s (Dalton, 2017; Foster & Frieden, 2017; Hetherington, 2005). This decline in trust has been attributed to a perception that public institutions are often non-responsive to the needs and desires of the citizenry (Torcal, 2014) and is thought to impact civic participation, including voting (Abramson et al., 1999). Hetherington and Rudolph (2020) have argued that the increasing polarization of political trust along party lines has supported to political gridlock in the U.S. and may be due to increasingly diverse worldviews, a lack of warmth for political opponents, and selective evaluations of government performance. As democracy represents a social contract between government and its citizens, a decline in citizens’ trust with both politicians and political institutions is thought to threaten the long-term viability of democracy (Dalton, 2017). Others have argued that the declines in trust may be the result of an increasingly skeptical rather than cynical public, and that concerns over democratic viability are premature (Cook & Gronke, 2005).

The building and maintenance of citizens’ trust in public sector institutions offers several benefits for governance. Along with the enactment of laws, trust can provide public sector institutions with social-political legitimacy for the implementation of policy and can promote stability and reliability in operations (Park et al., 2016). Trust is essential to pushing forward policy; a lack of trust can discourage citizens from complying with laws, promote civil unrest, and discourage citizen-government cooperation (Kearns, 2004; Tolbert & Mossberger, 2006). Government-citizen cooperation can be especially important during times of crisis; for example, contact tracing efforts during the COVID-19 pandemic relied upon this cooperative relationship. Trust in government during crises is crucial to gaining citizen buy-in and ensuring policy compliance (Kritzinger et al., 2021), and research into compliance with government-related COVID-19 policies and regulations has supported this link (Devine et al., 2021). Trust in government has also previously been shown to correlate with attitudes towards vaccinations (Baumgaertner et al., 2018) as well as perceptions of the threat posed by COVID-19 (Jennings et al., 2021).

The concept of citizens’ satisfaction with the performance of public sector institutions is linked to their trust in those institutions, such that increases in satisfaction result in increased trust (Keele, 2007). The use of diverse forms of information technology and e-government is thought to result from an intentional effort to increase citizen satisfaction and trust in public sector institutions (Bannister & Connolly, 2011). E-government can be defined as the use of ICTs in government to enhance transactions, increase information dissemination and communication, and provide the potential for greater public participation (Bonson et al., 2015; Li & Feeney, 2014). By providing a more efficient, accessible means of citizen-government interaction, e-government efforts hope to build citizens’ trust in the ability of their governments to fulfill their mandates. While increased trust is a desired outcome of e-government, research has shown that mutual trust is also considered a success factor for e-government implementation (Meijer et al., 2012), though overall research into the relationship between e-government and citizen trust and satisfaction are mixed (Porumbescu, 2016). This may be due to a perceived decline in social capital due to modernization, which is said to erode both participation and trust in political institutions (Putnam, 2000). While research into the relationship between civic engagement and social capital accrued through social media is mixed, recent work by Lee (2022) suggests that virtual social ties may work to facilitate civic engagement, both individually and collectively.

Early e-government implementations have adapted over time to include more direct communication channels through interactive, Web 2.0-style tools. Social media outlets such as Twitter, Facebook, and Instagram act as an accessible and interactive form of government-citizen communication, overcoming traditional communication barriers such as time and distance. As communication is considered a key element of trust between social groups, public sector organizations can leverage social media communication channels to reinforce and foster trust by sharing relevant information. The general credibility of public sector organizations gives these organizations an advantage in trust-building; trust is said to develop through the perceived credibility of shared information, based on such things as expertise, evidence, benevolence, reputation, honesty, and predictability (Corritore et al., 2003; Levi & Stoker, 2000). As social media becomes an increasingly popular choice for real-time information, public sector organizations must utilize these channels effectively and leverage whatever trust capital they have accrued.

The Role of Social Media

The number of U.S. citizens accessing news through online channels has been increasing, especially through mobile devices (Mitchell et al., 2016). Social media platforms are at the forefront of this move away from traditional information sources. Social media has been defined as the digital space that facilitates the “diffusion of compelling content, dialogue creation, and communication to a broader audience” (Kapoor et al., 2018, p. 536). Social media has been seen as an accessible, engaging tool for improving information transparency, trust, legitimacy, and confidence in government (Bonson et al., 2017; Park et al., 2016; Silva et al., 2019; Song & Lee, 2016; Warren et al., 2014). Social media provides organizations with access to large numbers of geographically dispersed information consumers, providing efficient communication channels for information dissemination, collaborative problem-solving, and interactive, insightful dialogue, and is seen as an essential component of e-government (Bertot et al., 2012). Public sector organizations can leverage social media to encourage civic participation and interactive dialogue, social media can also help these organizations increase government responsiveness through commonly used information channels (Bryer, 2011).

Research into the reasons for public sector social media use, as well as realization of the participatory promise of social media is mixed. Research suggests that public sector social media use may be more often a function of perceived competition and institutional obligation rather than a desire to engage the public in collaborative policymaking or gather public feedback (Manetti et al., 2017; Reddick et al., 2017), and scant evidence exists suggesting that citizens use social media for interactive participation with their government (Haro-de-Rosario et al., 2018). The dissemination of information has therefore become the primary use of public sector social media.

Consumption of public sector social media has also been shown to influence citizen trust and satisfaction. Porumbescu (2016) found that citizens’ use of public sector social media channels was significantly related to satisfaction with and trust in public sector organizations, suggesting that communication channels with less information (e.g. social media) may offer more effectiveness in this area than channels offering more information (e.g. traditional news outlets). Social media provides far more rapid dissemination of information than traditional media channels. As a result, the real-time and interactive nature of social media allows public leaders to influence the agenda independent of traditional “mass media” such as television or newspapers (Aharony, 2012; Park et al., 2016). The perceived credibility or “ethos” of the message source – i.e. the authority or believability of the communication source, coupled with perceptions of expertise and trustworthiness – can amplify the effect of agenda-setting, framing, and priming. Evaluation of source credibility impacts perceptions of trustworthiness as well as potentially influencing behavior (Housholder & LaMarre, 2014) and is especially challenged in the digital arena, as tools like social media level the proverbial playing field for many different sources of information (Flanagin & Metzger, 2017).

Social media allows consumers of information to be selective in their sources; consequently, many users will gravitate towards information sources which are thought to be credible and align with their political or ethical concerns – so called “political consumerism” (De Zuniga et al., 2014). Prior research has shown that such consumerism can limit exposure to information from non-preferred sources, resulting in the potential for confirmation bias (Thorson, 2016). Political information is often shared on social media with people of similar political beliefs, and it is these beliefs which often determine which information is consumed or ignored (Klein & Robison, 2020). In terms of source credibility, official outlets, such as the feeds of public officials and established news organizations, are more likely to be perceived as credible (David et al., 2016; Stieglitz & Dang-Xuan, 2013). This perception of credibility among official outlets is especially important, given that the credibility of the channel is also an evaluative factor (Pressgrove & Kim, 2018); research shows that much of the public does not view social media as a credible source of information, as least for news. A 2016 study by the Pew Research Center found that only 7% of surveyed adults trusted the information they obtained from social media (Mitchell et al., 2016).

By providing a more direct contact – perceived or actual – between citizens and their government, social media channels can foster trust in ways that traditional communication channels cannot. This reduced psychological distance (i.e. how far away the stimulus appears to be) can result in a more positive perception of the message sender (Porumbescu, 2016; Trope et al., 2007). Prior research has suggested that perceived distance can be a factor in a reduction of citizens’ trust (Morgeson et al., 2011; Park et al., 2016). The influence of perceived trustworthiness and credibility can be especially important during disaster events, including pandemics, when the hunger for information – and the desire to share information – may be increased. Research has suggested that official sources are often considered more credible for disaster-related information, though concerns about timeliness lead consumers to consider both official and unofficial information sources (Liu et al., 2016).

Public Sector Social Media Use and Disaster Events

Crisis/emergency events such as pandemics and natural disasters cause severe interruptions in economic, social, and political life, and as such they facilitate increased levels of uncertainty and urgency. This context facilitates increased communication, creating knowledge gaps which must be filled with credible and timely information (Lee & Yu, 2020; Liu et al., 2016; Martinez-Rojas et al., 2018; Shklovski et al., 2008). According to the World Health Organization (2020), social media is considered an important vehicle for information transfer during times of pandemic. This importance has aligned with the increasing perception of social media as a credible information source during crisis events (Lee & Yu, 2020; Lin et al., 2016; Martinez-Rojas et al., 2018). However, perceived credibility and trust in government was further called into question by the COVID-19 pandemic, especially in the United States. Social media – with the ability to increase political polarization, and to foster false and conflicting narratives – has been called a contributing factor to declining trust in government, though research remains mixed (Klein & Robison, 2020).

Evidence of social media’s importance during crisis events was seen during the COVID-19 pandemic; Basch et al. (2020) demonstrated the ability of social media outlets such as YouTube to disseminate information necessary for mobilization and mitigation efforts, and Goldberg et al. (2020) confirmed that government response and recommendations can have a measurable impact on public health behaviors during a pandemic. However, the COVID-19 pandemic also highlighted how social media can impede public health efforts. Vaccine hesitancy and conspiratorial views of the pandemic response were said to be exacerbated by social media, especially through the hostility voiced towards the CDC, WHO, and other health organizations by President Donald Trump and other figures (Albrecht, 2022). The politicization of the pandemic was fueled in part by social media, which amplified the conservative backlash associated with some mitigation measures. This backlash and minimizing of the pandemic’s risks and effects, led by the hostile comments of President Trump, is said to have weakened the effectiveness of CDC and WHO recommendations in the U.S. (Hamilton & Safford, 2021; Stroebe et al., 2021) through both official rebukes and accusations of CDC and WHO politicization. The perceived decline in government trust and credibility is not universal, however; other areas (e.g. Western Europe) experienced increases in government trust in popularity during the pandemic (Bol et al., 2021; Oude Groeniger et al., 2021).

Research has shown the wide acceptance of social media for emergency management purposes, including but not limited to the popular microblog Twitter, now known as X (Martinez-Rojas et al., 2018). Prior studies have examined how Twitter was used to disseminate information during public health crises, such as the 2014 Ebola outbreak in West Africa (Liang et al., 2019) and natural disasters like the 2013 Typhoon Haiyan in the Philippines (David et al., 2016), as well as floods, earthquakes, and other events (e.g. Palen et al., 2010; Doan et al., 2011; Murthy & Longwell, 2013; Sutton et al., 2014; Kitazawa & Hale, 2021). User activity on Twitter is limited to 280 characters of text, though the inclusion of images, links, and short videos is common; consequently, users must choose their messages carefully to maximize impact and reach. Among public sector organizations, Twitter is most often used for the broadcasting of public information messages while other outlets (e.g. Facebook) are more often used for more interactive dialogue (Manetti et al., 2017). Besides the dissemination of information during disaster events, public sector organizations can consume the information provided by other users to enhance situational awareness. Tweets can communicate real-time reports of conditions and experiences, providing valuable information to both emergency managers and first responders (David et al., 2016; Hughes & Palen, 2009).

Twitter/X is recognized as one of the more popular social media outlets; as of 2021, approximately one-quarter of American adults used the service (Auxier & Anderson, 2021). Prior estimates suggested that there are over 150 million active daily users of Twitter and that 500 million “tweets” are posted daily (Aslam, 2020). In terms of demography, Twitter users in the U.S. tend to be younger, more educated, have higher incomes, and lean more towards the liberal end of the political spectrum (Wojcik & Hughes, 2019), though these demographics may change over time due to recent leadership and policy changes. These factors limit generalizability across the population, but Twitter provides an example of an information source commonly used by political actors to share crisis information. As a virtual public square, political discussion on Twitter by both political figures and citizens is common. Approximately one-third of tweets involve political content, and older adults (50 years or more) are more likely to post about politics than younger users (Bestvater et al., 2022). Twitter use also involves a significant amount of information re-sharing; recent research has suggested that 35% of tweets involve retweets (Chapekis & Smith, 2023) and this activity has been called the primary means of information dissemination on the platform (Lee & Yu, 2020).

Retweeting – the sharing of a tweet by one user to their audience – can help to generate awareness of information the user perceives as “important” (David et al., 2016). Reasons generally accepted for retweeting include the desire to share entertaining or useful information, to persuade others, to acknowledge agreement with tweet content, and to archive the information (Thelwall & Thelwall, 2020). Retweeting is also essential part of building a large audience, as such activity attracts other users to “follow” the source (Steele & Dumbrell, 2012). During times of crisis, original tweets of real-time conditions are common, and retweeting can help spread information to the masses. It is also common for users to retweet to share anxiety or warnings about current or future conditions. Since social media has been shown to exert influence on behavior (e.g. Margetts et al., 2015), retweeting posts about the actions of some individuals could encourage others to take similar action (Kitazawa & Hale, 2021). Prior research has shown that examining retweet patterns can be used to assess communication effectiveness and/or provide evidence of the perceived importance of content (David et al., 2016; Wang & Zhuang, 2017).

As the COVID-19 crisis formed, citizens had a growing number of information sources from which they could learn about the virus, the impacts, and mitigation efforts. Given the nature of federalism in the U.S., federal as well as state governments played a prominent role in the initial response to the onset of the COVID-19 pandemic. At the federal level, official information sources included the President and Vice President, the Centers for Disease Control and Prevention (CDC), and the National Institutes for Health (NIH), among others. At the state level, Governors and state-level health departments became critical sources for up-to-date information about response and mitigation efforts, especially on social media.

Prior research suggests that governments play a dominant role in crisis-related social media conversations (Kitazawa & Hale, 2021). Assessing the relative frequency at which official, pandemic-related information was shared in the early crisis stages can provide insight into which sources citizens may gravitate towards in public health crises, including specific U.S. public officials (e.g. Governors, the President) or specific public agencies (e.g. the CDC, state-level health departments). The increasing political division in the U.S. in the post-2016 era, coupled with the tendency of consumers to share information from those with whom they share beliefs, means that the party affiliation of leadership may play a role in patterns of sharing. To investigate the public sector social media sources utilized during the early onset of the pandemic, I explore the following research question:

RQ₁

Do significant differences in the retweet frequency exist based on the source of the tweet during the onset of the COVID-19 pandemic in the U.S.?

and I propose the following hypotheses:

H₁

There are significant differences in retweet frequency based on the source category (public officials or public agencies/departments) and the time period (pre- and post-pandemic).

H₂

There are significant differences in retweet frequency for state accounts based on the political party of the governor (Republican or Democrat) and the time period (pre- and post-pandemic).

The content of tweets has also been shown to impact the likelihood of retweeting. Prior research into sentiment/linguistic factors and retweeting tendencies did not focus on public sector tweets, but the findings may be informative in this context. Lee and Yu (2020), examining Twitter data during the 2013 Colorado floods, suggested that the likelihood a tweet will be retweeted is unrelated to the use of emotional language (positive or negative) while Miura et al. (2016), examining tweets during the Great East Japan Earthquake, found that tweets with higher levels of anxiety language were more likely to be retweeted than those with angry or positive emotion. As this study explores the retweeting of public sector sources, exploration of the content of retweets can provide insight into the linguistic factors which may appeal to citizens in times of crisis. Based on the prior literature review, I consider the following research question:

RQ₂

Does the level of emotional sentiment predict the retweet frequency of tweets from government accounts during the onset of the COVID-19 pandemic in the U.S.?

and I propose the following hypotheses:

H₃

The level of emotional sentiment present in tweets from government sources significantly predicts the retweet frequencies for government sources, controlling for time period (pre-pandemic or post-pandemic).

H₄

The level of emotional sentiment present in tweets significantly predicts the retweet frequencies for public official accounts, controlling for time period (pre-pandemic or post-pandemic).

H₅

The level of emotional sentiment present in tweets significantly predicts the retweet frequencies for public agency accounts, controlling for time period (pre-pandemic or post-pandemic).

H₆

The level of emotional sentiment present in tweets significantly predicts the retweet frequencies for state government sources in states with Republican governors, controlling for time period (pre-pandemic or post-pandemic).

H₇

The level of emotional sentiment present in tweets significantly predicts the retweet frequencies for state government sources in states with Democratic governors, controlling for time period (pre-pandemic or post-pandemic).

Methodology

Tweets were collected for the period February 01, 2020, through April 20, 2020. As March 11, 2020, marks the date of the WHO pandemic declaration, this dataset provides insight into retweeting patterns before and after that declaration in equal proportions. I used custom Python scripts to collect tweets based on tweet IDs obtained from the GeoCoV19 dataset from CrisisNLP (https://crisisnlp.qcri.org/covid19). The GeoCoV19 dataset (Qazi et al., 2020) contains over 500 million IDs of tweets containing one or more of 800 possible COVID-19 related hashtags and keywords. Custom scripts were used to filter the tweets for English-language and U.S.-specific tweets, where U.S.-specific is defined as source location as found in the GeoCov19 metadata for each tweet. This filtering resulted in a reduced dataset of 63,036,772 tweets.

I then filtered the set of U.S.-specific tweets based on whether the tweets involved a retweet (excluding quoted tweets), resulting in a set of 47,922,526 tweets, or 76.02% of the English-language, U.S.-specific tweets. This retweets dataset was further filtered based on whether the origin of the retweet was any one of 118 different Twitter accounts. These accounts (see Table 1) included eight (8) federal accounts and 110 state-level accounts, including accounts for all 50 state governors, all 50 state health departments, as well as the only 10 accounts specific to state health secretaries found during preliminary investigation. The federal accounts included accounts for the President, Vice President, Surgeon General, CDC Director, Health and Human Services (HHS) Secretary, and the official accounts for NIH, CDC, and HHS. The selection of public sector accounts was meant to be as inclusive as possible at the state level (all governors, all health departments, and available state health department secretaries) and representative at the federal level, including the major public officials in the crisis as well as the major public health agencies. Seven accounts – three from state health secretaries, three from state health departments, and one federal-level account – were removed from further analysis due to a lack of shared tweets in the final dataset. The final dataset included 998,482 retweets shared from 111 public sector accounts (2.08% of the retweets dataset).

Table 1.

List of Twitter Handles for Data Collection.

Governor Accounts	GovBillLee	HealthNYGov
GovernorKayIvey	GregAbbott_TX	ncdhhs
GovDunleavy	GovHerbert	NDDOH
Dougducey	GovPhilScott	OHdeptofhealth
AsaHutchinson	GovernorVA	HealthyOklahoma
CAgovernor	GovInslee	Healthoregon^a
GovofCO	WVGovernor	PAHealthDept
GovNedLamont	GovEvers	RIHEALTH
JohnCarneyDE	GovernorGordon	scdhec
GovRonDeSantis		SDDOH
GovKemp	State health departments	TNDeptofHealth
GovHawaii	ALPublicHealth	TexasDSHS
GovernorLittle	Alaska_DHSS	UtahDepOfHealth
GovPritzker	AZDHS	Healthvermont
GovHolcomb	ADHPIO	VDHgov
IAGovernor	CAPublicHealth	WADeptHealth
GovLauraKelly	CDPHE	WV_DHHR
GovAndyBeshear	CTDPH	DHSWI
LouisianaGov	Delaware_DHSS	health_wyoming^a
GovJanetMills	HealthyFla
GovLarryHogan	GaDPH	State health secretaries
MassGovernor	HIgov_Health	Drcarachrist
GovWhitmer	IDHW	DrKaraWalker
GovTimWalz	IDPH	CHFS_CaresforKy^a
Tatereeves	StateHealthIN	Robertmgordon
GovParsonMO	IAPublicHealth	MDHCommMalcolm^a
GovernorBullock	KDHE	TCBPubHealth
GovRicketts	KYHealthAlerts	SecMandyCohen
GovSisolak	LADeptHealth	SecretaryLevine
GovChrisSununu	MaineDHHS^a	Sddohkmr
GovMurphy	MDHealthDept	WaHealthSec^a
GovMLG	MassDPH
NYGovCuomo	MichiganHHS	Federal Officials/Agencies
NC_Governor	Mnhealth	realDonaldTrump
DougBurgum	msdh	VP^a
GovMikeDeWine	HealthyLivingMo	SecAzar
GovStitt	DPHHSMT	Surgeon_General
OregonGovBrown	NEDHHS	CDCDirector
GovernorTomWolf	DhhsNevada	CDCgov
GovRaimondo	NHDHHSPIO	NIH
Henrymcmaster	NJDeptofHealth	HHSGov
Govkristinoem	NMDOH

^aAccount was excluded from analysis due to a lack of shared tweets.

As March 11, 2020, marks the date of the WHO pandemic declaration, this dataset provides insight into U.S. public retweeting patterns immediately before and after that declaration in equal proportions. The final dataset was organized by day, and a dichotomous variable was used to indicate the time period from February 01 through March 11 (pre-pandemic) and the time period from March 12 through April 20 (post-pandemic). Total retweet counts were organized as overall counts per day, by level (federal-level, state-level), and by role (accounts of public officials, and accounts of public agencies/departments). Total retweet counts for state-level sources (state-level accounts of public officials, and accounts of public agencies/departments) were also filtered based on the party affiliation of the governor (Republican or Democrat) during this time period. Among the 50 states, there were 26 Republican and 24 Democrat governors in 2020.

Data analysis followed three paths. Preliminary t-test analyses were used to identify significant differences in overall retweet frequencies as well as for the five source categories (combined federal and state sources, public officials, public agencies, states with Republican governors, states with Democrat governors). Next, I conducted a set of 2 (time period, pre-pandemic vs. post-pandemic) x 2 (source of tweet) mixed Analysis of Variance (ANOVA) procedures to test hypotheses H1 and H2. Finally, I performed a series of multiple regressions to test hypotheses H3-H7. The independent variables for hypotheses H3-H7 were operationalized from measures produced by the Linguistic Inquiry and Word Count (LIWC) software. LIWC provides automated analysis of text-based data for persistent themes and constructs, including pre-defined affective, cognitive, structural, and behavioral indicators. LIWC generates a numeric measure for over 80 linguistic categories (for a full description of the categories, see Pennebaker et al., 2015). Prior research has shown the validity and reliability of LIWC measures across different contexts and languages (e.g. Alpers et al., 2005; Bantum & Owen, 2009; Tausczik & Pennebaker, 2010) and for correlating linguistic style with psychological, emotional, and other factors (e.g. Wang et al., 2016; Chandra Guntuku et al., 2019).

For this study, I analyzed the retweets on each day using the LIWC software. Emotional sentiment was operationalized as four separate LIWC categories/measures: positive emotion, anxiety, anger, and sadness. Each of these words has an associated dictionary of words; for example, the dictionary for “anger” includes words such as “hate” or “annoyed”, while the dictionary for “sadness” includes words like “crying” or “grief” (Pennebaker et al., 2015). LIWC reports scores as a percentage of the total words in the text that fall in the specific dictionary. The choice of the positive emotion, anxiety, anger, and sadness measures for this study was influenced by prior uses of this construct in the literature (e.g. Lee & Yu, 2020). I also included a dichotomous variable for time period (pre-pandemic vs. post-pandemic) as a controlling factor in the regression analyses.

Results

This study found that the pandemic period was an important differentiator for and predictor of the retweet frequencies of government Twitter accounts. Preliminary t-test analyses uncovered significant differences based on the time period (pre-pandemic vs. post-pandemic) for the overall retweet frequencies of government sources (t = −5.710, df = 70.563, p < .001, Cohen’s d = 1.277), public officials (t = −4.891, df = 71.934, p < .001, Cohen’s d = 1.904), public agencies/departments (t = −6.771, df = 78, p < .001, Cohen’s d = 1.514), state-level accounts for states with Republican governors (t = −8.703, df = 47.728, p < .001, Cohen’s d = 1.946) and state-level accounts with Democrat governors (t = −11.737, df = 68.313, p < .001, Cohen’s d = 2.624). Descriptive statistics are shown in Table 2.

Table 2.

Descriptive Statistics.

	Retweets	Mean	SD	Min	Max
Public sector retweets	998482	12481.025	9821.513	62	35207
Federal-level	506906	6336.325	6855.767	51	29579
State-level	491576	6144.700	5946.637	3	25884
Governor (R)	186670	2333.375	2688.233	1	13214
Governor (D)	304906	3811.325	3568.621	0	12670
Public officials	819729	10246.613	8520.210	17	33617
Public agencies	178753	2234.413	1979.175	34	8911

Pre-pandemic	287288	7182.200	9553.486	62	33269
Federal-level	229845	5746.125	8283.892	51	29579
State-level	57443	1436.075	2424.015	3	8367
Governor (R)	18335	458.375	869.504	1	3004
Governor (D)	39108	977.700	1705.026	0	6908
Public officials	245779	6144.475	8520.859	17	30630
Public agencies	41509	1037.725	1469.705	34	6423

Post-pandemic	711194	17779.850	6821.552	3952	35207
Federal-level	277061	6926.525	5086.338	1373	25627
State-level	434133	9223.372	4502.951	1943	25884
Governor (R)	168335	4208.375	2582.811	918	13214
Governor (D)	265798	6644.950	2533.555	1025	12760
Public officials	573950	14348.750	6318.725	3387	33617
Public agencies	137244	3431.100	1684.374	565	8911

As could be expected, the preliminary t-test results show that the frequencies at which COVID-related tweets were shared from government accounts significantly increased in the post-pandemic period. Social media has become an important tool for government organizations to collect and disseminate information during a crisis, providing a real-time mechanism for brief, focused government-to-citizen communication. The increase in retweets from these accounts can be expected to increase in the post-pandemic period, as prior research has shown that the psychological impact of such crises – e.g. anger or nervousness – leads people to share crisis information (Chen & Sakamoto, 2013). This increase is reasonably thought to be a function of uncertainty and the resultant thirst for information in the presence of an unprecedented public health crisis. The significant increase in post-pandemic sharing of government tweets was seen for state-level accounts (regardless of the Governor’s political affiliation) as well as the accounts of public officials and public agencies. The effect sizes in all cases were very large, suggesting a concerted effort at spreading government-sourced information during the early days of the pandemic.

ANOVA Results: Group Differences in Retweet Frequency

Examining differences between various public sector groups, including groups based on roles (public officials or public agencies) and the political affiliation of state-level leadership can shed light on which sources citizens gravitate towards during times of crisis.

Hypothesis H1: Public officials and Public Agencies/Departments

I conducted a 2 (time period, pre-pandemic vs. post-pandemic) x 2 (source of tweet, public officials vs. public agencies) mixed ANOVA to determine whether retweet frequencies significantly differed between tweets from public officials and public agencies, considering differences in the time period. The analysis was also used to determine if the interaction between the time period and the source of the original tweet (public officials or public agencies) had a significant impact on retweet frequencies. The results showed a significant main effect of tweet source (public officials vs. public agencies) on retweet frequencies, F(1, 78) = 105.612, p < .001, partial η² = 0.575. There was also a significant main effect of time period (pre-pandemic vs. post-pandemic) on retweet frequencies, F(1, 78) = 32.600, p < .001, partial η² = 0.295. The interaction between time period and source type was significant, F(1, 78) = 13.888, p < .001, partial η² = 0.151. So, ignoring other variables, the daily retweet frequencies differed based on both the time period and the source of the original tweet. Hypothesis H1 is supported; see Table 3.

Table 3.

Estimated Marginal Means for Time and Source (Officials or Agencies).

Time Period	Source	Mean	SE	95% Confidence interval
Time Period	Source	Mean	SE	Lower bound	Upper bound
Pre-pandemic	Officials	6144.475	1186.019	3783.292	8505.658
Pre-pandemic	Agencies	1037.725	249.929	540.155	1535.295
Post-pandemic	Officials	14348.750	1186.019	11987.567	16709.933
Post-pandemic	Agencies	3431.100	249.929	2933.530	3928.670

Figure 1 describes the significant interaction results. During both time periods, Twitter users retweeted the posts of public officials in greater numbers than posts from public agencies. However, this difference was much more pronounced in the post-pandemic period, suggesting that users looked more towards public officials (e.g. the President, Governors, and State Health Secretaries) for COVID-related information rather than public agencies (e.g. the CDC, state health departments). The differential between retweets from public official and public agency accounts was much more pronounced in the post-pandemic period, despite a more than threefold increase in the number of retweets from agency accounts. This finding suggests that Twitter users felt more compelled to share information from the accounts of public leaders, despite the content-expert status often afforded to federal- and state-level health agencies (e.g. CDC, NIH, state-level health departments). This may be due to the front-and-center nature of public officials; in the absence of other compelling factors, people will gravitate towards information sources with which they are familiar. In seeking out information, preference appears to have been placed towards public officials rather than public agencies. The politicization of the pandemic – and the associated distrust in scientific experts and related government agencies – are factors to consider in this result. Future research should explore in more depth the role that public personality and position play in citizens’ decisions to retweet information.

Figure 1.

Interaction between time period and source type (Officials/Agencies).

Hypothesis H2: State-Level Accounts and Gubernatorial Political Affiliation

I conducted a similar 2 × 2 mixed ANOVA to determine whether retweet frequencies significantly differed between tweets from those state-level accounts originating from states with Republican governors or those with Democrat governors, considering differences in the time period. The analysis was also used to determine if the interaction between the time period and the source of the original tweet (Republican-led states or Democrat-led states) had a significant impact on retweet frequencies. The results showed a significant main effect of the state-level tweet source (Republican-led states vs. Democrat-led states) on retweet frequencies, F(1, 78) = 47.527, p < .001, partial η² = 0.379. There was also a significant main effect of time period (pre-pandemic vs. post-pandemic) on retweet frequencies, F(1, 78) = 135.643, p < .001, partial η² = 0.635. The interaction between time period and state-level tweet source type was significant, F(1, 78) = 19.995, p < .001, partial η² = 0.204. So, ignoring other variables, the daily retweet frequencies differed based on both the period and the state-level tweet source (officials and agencies of Republican-led states vs. Democrat-led states). Hypothesis H2 is supported; see Table 4. Significant interaction results are shown in Figure 2. During both time periods, Twitter users retweeted the posts of public officials and agencies from Democrat-led states in greater numbers than posts from public officials and agencies from Republican-led states. However, this difference was much more pronounced in the post-pandemic period, suggesting that users looked more towards public officials and agencies in Democrat-led states for COVID-related information once the pandemic was declared.

Table 4.

Estimated Marginal Means for Time and State-Level Source (Affiliation).

Time Period	Governor affiliation	Mean	SE	95% Confidence interval
Time Period	Governor affiliation	Mean	SE	Lower bound	Upper bound
Pre-pandemic	Republican	458.375	304.692	−148.219	1064.969
Pre-pandemic	Democrat	977.70	341.431	297.963	1657.437
Post-pandemic	Republican	4208.375	304.692	3601.781	4814.969
Post-pandemic	Democrat	6644.950	341.431	5965.213	7324.687

Figure 2.

Interaction between time period and state-level source type (Affiliation).

The significant differences between Republican-led and Democratic-led states may be a function of the more media-prominent role that Democratic Governors like Andrew Cuomo (NY) and Gavin Newsome (CA) held in the early stages of the pandemic, as well as the potential hesitancy of Republican governors considering President Trump’s slow response strategy. The increasing conservative hostility towards the CDC and pandemic mitigation measures may have also been a factor, coupled with a Twitter population which tends to skew younger, more educated, and more liberal. Future research should consider how such things as the size of the Governor’s social media following and the strength of opposition within the state impact the frequency of retweeting state-level sources, as well as whether the demography of the platform impacts perceptions of government sources as credible or trustworthy.

However, it is important to note that users retweeted substantially more tweets from all state-level sources than from federal-level sources once the pandemic was declared. To be exact, the number of retweets from state-level sources post-pandemic increased by 7.56 times from pre-pandemic levels. While this overall increase is at least partially due to an increased number of COVID-related tweets coming from these state-level sources, the growth of state-level retweets is indicative of the nature of the early pandemic response effort. States were widely seen as taking the lead in the pandemic response through economic shutdowns and orders for masking, sheltering in place, and other mitigation efforts. This state-level leadership, coupled with the diversity of state response strategies, would necessarily influence the decision for Twitter users to share these tweets to inform members of their network.

Regression Results: Predicting Retweet Frequencies

The group differences point out which groups are more likely to be retweeted (and when) but examining the content of tweets can help to explain what linguistic factors may appeal to citizens and, therefore, encourage them to share tweets with their networks. To test hypotheses H3-H7, I conducted five multiple regression analyses. The regression results found that both the pandemic period and the use of emotion-related language predicted the frequency of retweets from public sector accounts, though the effects differed between the different groups.

Hypothesis H3: Emotional Sentiment and Government Sources

The first regression analysis was conducted to determine which of the four LIWC factors predicts the retweet frequencies for government sources, controlling for the time period (pre-pandemic vs. post-pandemic). Preliminary data screening for outliers led to the removal of four cases. Pearson correlations were examined to tested to assess any significant correlations between the four LIWC factors. The results found only small to moderate correlations between factors, so a decision was made to retain all factors for the regression model. The results found that the model significantly predicted the retweet frequencies for government sources, R² = 0.434, R²_adj = 0.394, F(5, 70) = 10.750, p < .001, supporting Hypothesis H3. The model accounted for 39.4% of the outcome variance.

For all government sources, the use of positive emotion-related language (e.g. nice, sweet) and sadness-related language (e.g. sad, grief) increased the likelihood of retweeting, as did the post-pandemic period. The frequency of retweets for government sources increased by 0.229 for each unit increase in the positive emotion factor (p < .05), by 0.225 for each unit increase in the sadness factor (p < .05), and 0.529 units post-pandemic (p < .001). A summary of the regression coefficients is presented in Table 5. Higher amounts of positive emotion words have previously been tied to efforts to promote leadership (Slatcher et al., 2007) and may be intended to lower anxiety and fear among citizens. The predictive power of sadness-related language may also be attributed to the tweet topics, as the isolation and uncertainty brought on by initial response measures (e.g. stay at home orders, economic shutdowns) may have increased the prevalence of messages related to mental health resources and concerns, in addition to information on death totals and illness symptoms. Future research would be useful to drill down into specific tweet topics which comprise government-sourced retweets.

Table 5.

Regression Coefficients (Retweets of Public Sector Sources).

	B	Beta	t	p	Bivariate r	Partial r
Positive emotion**	2152.133	0.229	2.340	0.022	0.404	0.269
Anxiety	−1256.887	−0.037	−0.324	0.747	−0.203	−0.039
Anger	350.327	0.007	0.071	0.944	0.022	0.008
Sadness**	8358.599	0.225	2.290	0.025	0.100	0.264
Pandemic (0/1)***	10363.734	0.529	5.152	0.000	0.589	0.524

R² = 0.434, R²_adj = 0.394, F_{(5 70)} = 10.750, p < .001.

***p < .01; **p < .05; *p < .10.

Hypothesis H4: Emotional Sentiment and Public Official Accounts

The second regression analysis was performed to determine which of the four LIWC factors predicts the retweet frequencies for public officials, controlling for the time period. Preliminary data screening for outliers and normality led to the removal of two cases. The results found only weak Pearson correlations between variables, so all factors were retained. The regression results found that the model significantly predicted the retweet frequencies for the accounts of public officials, R² = 0.329, R²_adj = 0.283, F(5, 72) = 7.076, p < .001, supporting Hypothesis H4. The model accounted for 28.3% of the outcome variance. For public officials, increased use of language related to anger (e.g. hate, annoyed) and sadness were significant predictors of retweet frequencies, as was the post-pandemic period. The frequency of retweets for public officials increased by 0.221 for each unit increase in the anger factor (p < .05) and by 0.250 for each unit increase in the sadness factor (p < .05). The frequency of retweets also increased by 0.448 in the post-pandemic period (p < .01). A summary of the regression coefficients is presented in Table 6. Given the political nature of many of these accounts, along with the high level of acrimony towards the Trump administration and the onset of the 2020 election cycle, it is possible that the tendency towards negative affect is at least partially a result of the perceived lack of a coordinated federal response. It is also possible that the negative affect is simply a function of the subject matter. Prior research suggests that negative emotion words are used in writing about negative events (Tausczik & Pennebaker, 2010) where uncertainty and fear are not uncommon.

Table 6.

Regression Coefficients (Retweets of Public Officials).

	B	Beta	t	p	Bivariate r	Partial r
Positive emotion	247.578	0.037	0.342	0.733	0.024	0.040
Anxiety	−4553.698	−0.216	−1.638	0.106	−0.242	−0.190
Anger**	6530.999	0.221	2.188	0.032	0.188	0.250
Sadness**	5742.017	0.250	2.170	0.033	0.039	0.248
Pandemic (0/1)***	7659.676	0.448	4.124	0.000	0.502	0.437

R² = 0.329, R²_adj = 0.283, F_{(5, 72)} = 7.076, p < .001.

***p < .01; **p < .05; *p < .10.

Hypothesis H5: Emotional Sentiment and Public Agency Accounts

The third regression analysis focused on predicting the retweet frequencies for public agencies/departments, controlling for the period. Preliminary data screening for outliers and normality led to the removal of four cases. The Pearson correlation results found only weak correlations between variables, so all factors were retained. The results found that the model significantly predicted the retweet frequencies for public agency sources, R² = 0.473, R²_adj = 0.436, F(5, 70) = 12.574, p < .001, supporting Hypothesis H5. The model accounted for 43.6% of the outcome variance. The use of both anger-related and sadness-related language was found to significantly decrease the retweet frequency, while the use of anxiety-related language (e.g. fear, worry) was a significant predictor of higher retweet frequencies, as was the post-pandemic period. The frequency of retweets for public agencies/departments increased by 0.291 for each unit increase in the anxiety factor (p < .01) and by 0.454 in the post-pandemic period (p < .01). The frequency of retweets decreased by 0.216 for each unit increase in the anger factor (p < .05) and by 0.329 for each unit increase in the sadness factor. A summary of the regression coefficients is presented in Table 7. The use of anxious language may appeal to the existing anxiety of users during times of crisis; prior research has suggested it is common for users to retweet in order to share anxiety or warnings about current or future conditions (Kitazawa & Hale, 2021) or perhaps to increase their own following or retweet count (Recuero et al., 2011). It should be noted that the use of negative language, including angry or anxious language – despite potentially having positive effects during times of crisis, for promoting cautious behavior – may also be used to promote unnecessary or negative outcomes, such as scapegoating for public health crises. As mentioned previously, retweeting posts can impact the behavior of others (Kitazawa & Hale, 2021). Future research should explore the use of negative language in government-initiated social media messages, especially in combination with examinations of message topics.

Table 7.

Regression Coefficients (Retweets of Public Agency/Department Sources).

	B	Beta	t	p	Bivariate r	Partial r
Positive emotion	−0.317	−0.035	−0.317	0.752	0.367	−0.038
Anxiety***	2.830	0.291	2.830	0.006	0.113	0.320
Anger**	−2.254	−0.216	−2.254	0.027	−0.221	−0.260
Sadness***	−3.192	−0.329	−3.192	0.002	−0.422	−0.356
Pandemic (0/1)***	3.985	0.454	3.985	0.000	0.586	0.430

R² = 0.473, R²_adj = 0.436, F_{(5, 70)} = 12.574, p < .001.

***p < .01; **p < .05; *p < .10.

Hypothesis H6: Emotional Sentiment and Republican-Led State Accounts

Examining the predictive factors by state-level political leadership provided a diverse picture. The fourth regression analysis focused on predicting the retweet frequencies for state government sources in states with Republican governors, controlling for period. Preliminary data screening for outliers and normality led to the removal of 10 cases, all from the pre-pandemic period. The Pearson correlation results found only weak correlations between variables, so all factors were retained. The results found that the model significantly predicted the retweet frequencies for state government sources in states with Republican governors, R² = 0.545, R²_adj = 0.510, F(5, 64) = 15.362, p < .001, supporting Hypothesis H6. The model accounted for 51.0% of the outcome variance. Two of the five factors were found to be significant predictors. The frequency of retweets for Republican-led, state-level accounts decreased by 0.293 for each unit increase in the anger factor (p < .05) and increased by 0.936 in the post-pandemic period (p < .01). A summary of the regression coefficients is presented in Table 8. The results show that, for state-level-accounts in Republican-led states, the retweet frequencies could be predicted by the level of anger in the tweets; in short, more anger-oriented language meant less retweets.

Table 8.

Regression Coefficients (Retweets of State-Level Sources for Republican-Governor States).

	B	Beta	t	p	Bivariate r	Partial r
Positive emotion	−310.196	−0.132	−1.255	0.214	0.294	−0.155
Anxiety	2515.014	0.139	1.444	0.154	0.138	0.178
Anger**	−5232.621	−0.293	−2.630	0.011	−0.271	−0.312
Sadness	−594.531	−0.032	−0.334	0.740	−0.084	−0.042
Pandemic (0/1)***	5181.647	0.936	7.452	<0.001	0.686	0.682

R² = 0.545, R²_adj = 0.510, F_{(5, 64)} = 15.362, p < .001.

p < .01; **p < .05; *p < .10.

Hypothesis H7: Emotional Sentiment and Democratic-Led State Accounts

The fifth and final regression analysis was used to predict the retweet frequencies for state government sources in states with Democratic governors, controlling for period. Preliminary data screening for outliers and normality led to the removal of four cases. Pearson correlation analysis found a strong correlation between the anxiety factor and sadness (r = 0.792, p < .001), so the decision was made to remove the anxiety factor from the model. The results found that the model significantly predicted the retweet frequencies for state government sources in states with Democratic governors, R² = 0.632, R²_adj = 0.611, F(4, 71) = 30.476, p < .001. The model accounted for 61.1% of the outcome variance. The frequency of retweets for Democrat-led, state-level accounts increased by 0.719 in the post-pandemic period (p < .01). None of the emotion-related factors were significant predictors, so Hypothesis H7 is not supported. A summary of the regression coefficients is presented in Table 9. The differences in predictive factors between Republican-led and Democrat-led sources may be a matter of the early pandemic response communication, as well as a tendency for scientific mistrust to be more common among conservatives (Hamilton & Safford, 2021).

Table 9.

Regression Coefficients (Retweets of State-Level Sources for Democratic-Governor States).

	B	Beta	t	p	Bivariate r	Partial r
Positive emotion	377.329	0.116	1.278	0.205	0.524	0.150
Anger	202.578	0.014	0.177	0.860	0.316	0.021
Sadness	−85.298	−0.008	−0.114	0.910	−0.016	−0.014
Pandemic (0/1)***	4622.151	0.719	8.104	<0.001	0.788	0.693

R² = 0.632, R²_adj = 0.611, F_{(4, 71)} = 30.476, p < .001.

p < .01; **p < .05; *p < .10.

Conclusion

The uncertainty brought about by the COVID-19 pandemic resulted in a hunger for information among citizens. Social media is an easily accessible outlet for citizens to obtain and share (sometimes questionable) information, especially from information sources thought to be trustworthy or credible – including but not limited to public sector sources. This study examining the retweeting of posts from Twitter provided insight into the public sector sources users chose to share during the early stages of the crisis. As could be expected, the post-pandemic period (March 11-April 20, 2020) was especially active as federal and state governments wrestled with their responses to the emerging crisis and citizens struggled to adapt to the rapid and unprecedented changes to their daily routines. The significant group differences reflected a greater tendency to share COVID-related content from public officials as opposed to public agencies or departments. This result reinforces the idea that personalities rather than organizations were the more sought-after information sources, perhaps due to the widespread media coverage of certain public actors involved in the response (e.g. New York Governor Cuomo, President Trump). A similar effect can be seen in that Twitter users were significantly more likely to retweet public officials and agencies from Democrat-led states; as many of these states were at the forefront of the response efforts, it makes sense that these information sources would be shared. The proactive responses, combined with the demographics of Twitter users and the response hesitancy and contempt shown by some Republican politicians, promoted an environment whereby more Democrat-leaning sources would be more attractive for COVID-related content.

Analysis of the tweet content provided a diverse picture of the relationship between emotional language and retweet frequency. The regression results for the use of emotion-related language suggest that while positive language use predicts retweet frequency for all government sources, sadness, anxiety, and anger were also significant predictors for some groups. While at least some language falling in these categories is inevitable when talking about a public health crisis, the potential for using this language to promote negative outcomes (e.g. scapegoating, division) exists. Future research should explore the topics and linguistic features of government social media posts more in-depth, especially in combination with explorations of contextual (e.g. level of political opposition) and temporal factors which may impact use of the medium.

Additional research beyond the focus on U.S.- and English-specific tweets – i.e. a more inclusive and international sample, including from other social media platforms – may also allow for more generalizable findings. Future research may also seek to determine whether the findings of this study extend into other platforms. As previously mentioned, Twitter is commonly used by political actors to share crisis information, and retweets – information re-sharing within networks – is one of the more common platform activities. Twitter is not unique, as other social media platforms allow users to share (and re-share) information. Given the diversity of platforms used by governments, public officials, and citizens during this period, information from one social media platform commonly finds its way onto others. Exploring the retweeting patterns of government accounts on Twitter (now X) may therefore be representative of the information shared on social media during the onset of COVID-19; however, additional research would be necessary to uncover whether the findings of this study extend beyond a single platform.

Footnotes

Declaration of Conflicting Interests

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

Funding

The author(s) received no financial support for the research,authorship,and/or publication of this article.

ORCID iD

Jeffrey A. Stone

References

Abramson

Aldrich

Rohde

(1999). Change and continuity in the 1996 and 1998 elections. CQ Press.

Aharony

(2012). Twitter use by three political leaders: An exploratory analysis. Online Information Review, 36(4), 587–603. https://doi.org/10.1108/14684521211254086

Albrecht

(2022). Vaccination, politics and COVID-19 impacts. BMC Public Health, 22(1), 96. https://doi.org/10.1186/s12889-021-12432-x

Alpers

G. W.

Winzelberg

A. J.

Classen

Roberts

Dev

Koopman

Barr Taylor

(2005). Evaluation of computerized text analysis in an Internet breast cancer support group. Computers in Human Behavior, 21(2), 361–376. https://doi.org/10.1016/j.chb.2004.02.008

Aslam

(2020). Twitter by the numbers: Stats, demographics & fun facts. https://www.omnicoreagency.com/twitter-statistics/. Accessed 30 July 2020.

Auxier

Anderson

(2021). Social media use in 2021. Pew Research Center; Pew Research Center. https://www.pewresearch.org/internet/2021/04/07/social-media-use-in-2021/

Bannister

Connolly

(2011). The trouble with transparency: A critical review of openness in e‐government. Policy & Internet, 3(1), 1–30. https://doi.org/10.2202/1944-2866.1076

Bantum

E. O. C.

Owen

J. E.

(2009). Evaluating the validity of computerized content analysis programs for identification of emotional expression in cancer narratives. Psychological Assessment, 21(1), 79–88. https://doi.org/10.1037/a0014643

Basch

C. E.

Basch

C. H.

Hillyer

G. C.

Jaime

(2020). The role of YouTube and the entertainment industry in saving lives by educating and mobilizing the public to adopt behaviors for community mitigation of COVID-19: Successive sampling design study. JMIR Public Health and Surveillance, 6(2), Article e19145. https://doi.org/10.2196/19145

10.

Baumgaertner

Carlisle

J. E.

Justwan

(2018). The influence of political ideology and trust on willingness to vaccinate. PLoS One, 13(1), Article e0191728. https://doi.org/10.1371/journal.pone.0191728

11.

Bertot

J. C.

Jaeger

P. T.

Hansen

(2012). The impact of polices on government social media usage: Issues, challenges, and recommendations. Government Information Quarterly, 29(1), 30–40. https://doi.org/10.1016/j.giq.2011.04.004

12.

Bestvater

Shah

Rivero

Smith

(2022). Politics on Twitter: One-third of tweets from U.S. Adults are political. Pew Research Center - U.S. Politics & Policy. https://www.pewresearch.org/politics/2022/06/16/politics-on-twitter-one-third-of-tweets-from-u-s-adults-are-political/

13.

Bol

Giani

Blais

Loewen

P. J.

(2021). The effect of COVID‐19 lockdowns on political support: Some good news for democracy? European Journal of Political Research, 60(2), 497–505. https://doi.org/10.1111/1475-6765.12401

14.

Bonsón

Royo

Ratkai

(2015). Citizens' engagement on local governments' Facebook sites. An empirical analysis: The impact of different media and content types in Western Europe. Government Information Quarterly, 32(1), 52–62. https://doi.org/10.1016/j.giq.2014.11.001

15.

Bonsón

Royo

Ratkai

(2017). Facebook practices in Western European municipalities: An empirical analysis of activity and citizens’ engagement. Administration & Society, 49(3), 320–347. https://doi.org/10.1177/0095399714544945

16.

Bryer

T. A.

(2011). The costs of democratization: Social media adaptation challenges within government agencies. Administrative Theory and Praxis, 33(3), 341–361. https://doi.org/10.2753/atp1084-1806330302

17.

Chandra Guntuku

Buffone

Jaidka

Eichstaedt

Ungar

(2019). Understanding and measuring psychological stress using social media. Proceedings of the International AAAI Conference on Web and Social Media, 13, 214–225. https://doi.org/10.1609/icwsm.v13i01.3223

18.

Chapekis

Smith

(2023). How U.S. adults on Twitter use the site in the Elon Musk era. Pew Research Center. https://www.pewresearch.org/short-reads/2023/05/17/how-us-adults-on-twitter-use-the-site-in-the-elon-musk-era/

19.

Chen

Sakamoto

(2013). Perspective matters: Sharing of crisis information in social media. In 2013 46th Hawaii international conference on system sciences, Grand Wailea, Maui, Hawaii (pp. 2033–2041). IEEE.

20.

Cook

T. E.

Gronke

(2005). The skeptical American: Revisiting the meanings of trust in government and confidence in institutions. The Journal of Politics, 67(3), 784–803. https://doi.org/10.1111/j.1468-2508.2005.00339.x

21.

Corritore

C. L.

Kracher

Wiedenbeck

(2003). On-line trust: Concepts, evolving themes, a model. International Journal of Human-Computer Studies, 58(6), 737–758. https://doi.org/10.1016/s1071-5819(03)00041-7

22.

Dalton

R. J.

(2017). 23. Political trust in North America. In Zmerli

Van der Meer

(Eds.), Handbook on political trust (pp. 375–394). Edward Elgar.

23.

David

C. C.

Ong

J. C.

Legara

E. F. T.

(2016). Tweeting supertyphoon Haiyan: Evolving functions of Twitter during and after a disaster event. PLoS One, 11(3), Article e0150190. https://doi.org/10.1371/journal.pone.0150190

24.

Devine

Gaskell

Jennings

Stoker

(2021). Trust and the coronavirus pandemic: What are the consequences of and for trust? An early review of the literature. Political Studies Review, 19(2), 274–285. https://doi.org/10.1177/1478929920948684

25.

De Zúñiga

H. G.

Copeland

Bimber

(2014). Political consumerism: Civic engagement and the social media connection. New Media & Society, 16(3), 488–506. https://doi.org/10.1177/1461444813487960

26.

Doan

B. K. H.

Collier

(2011). An analysis of Twitter messages in the 2011 Tohoku earthquake. In International conference on electronic healthcare, Málaga, Spain, November 21-23, 2011 (pp. 58–66). Springer.

27.

Flanagin

Metzger

M. J.

(2017). Digital media and perceptions of source credibility in political communication. The Oxford Handbook of Political Communication, 417(2017), 417–436. https://doi.org/10.1093/oxfordhb/9780199793471.013.65

28.

Foster

Frieden

(2017). Crisis of trust: Socio-economic determinants of Europeans’ confidence in government. European Union Politics, 18(4), 511–535. https://doi.org/10.1177/1465116517723499

29.

Goldberg

M. H.

Gustafson

Maibach

E. W.

Ballew

M. T.

Bergquist

Kotcher

J. E.

Marlon

J. R.

Rosenthal

S. A.

Leiserowitz

(2020). Mask-wearing increased after a government recommendation: A natural experiment in the U.S. During the COVID-19 pandemic. Frontiers in Communication, 5(44). https://doi.org/10.3389/fcomm.2020.00044

30.

Hamilton

L. C.

Safford

T. G.

(2021). Elite cues and the rapid decline in trust in science agencies on COVID-19. Sociological Perspectives, 64(5), 988–1011. https://doi.org/10.1177/07311214211022391

31.

Hamm

J. A.

Smidt

Mayer

R. C.

(2019). Understanding the psychological nature and mechanisms of political trust. PLoS One, 14(5), Article e0215835. https://doi.org/10.1371/journal.pone.0215835

32.

Hardin

(2002). Trust and trustworthiness. Russell Sage Foundation.

33.

Haro-de-Rosario

Sáez-Martín

del Carmen Caba-Pérez

(2018). Using social media to enhance citizen engagement with local government: Twitter or Facebook? New Media & Society, 20(1), 29–49. https://doi.org/10.1177/1461444816645652

34.

Hetherington

M. J.

(2005). Why trust matters: Declining political trust and the demise of American liberalism. Princeton University Press.

35.

Hetherington

M. J.

Rudolph

T. J.

(2020). Why Washington won't work: Polarization, political trust, and the governing crisis. University of Chicago Press.

36.

Housholder

E. E.

LaMarre

H. L.

(2014). Facebook politics: Toward a process model for achieving political source credibility through social media. Journal of Information Technology & Politics, 11(4), 368–382. https://doi.org/10.1080/19331681.2014.951753

37.

Hughes

A. L.

Palen

(2009). Twitter adoption and use in mass convergence and emergency events. International Journal of Emergency Management, 6(3/4), 248–260. https://doi.org/10.1504/ijem.2009.031564

38.

Jennings

Stoker

Valgarðsson

Devine

Gaskell

(2021). How trust, mistrust and distrust shape the governance of the COVID-19 crisis. Journal of European Public Policy, 28(8), 1174–1196. https://doi.org/10.1080/13501763.2021.1942151

39.

Kapoor

K. K.

Tamilmani

Rana

N. P.

Patil

Dwivedi

Y. K.

Nerur

(2018). Advances in social media research: Past, present and future. Information Systems Frontiers, 20(3), 531–558. https://doi.org/10.1007/s10796-017-9810-y

40.

Kearns

(2004). Public value and E-government. Institute for Public Policy Research.

41.

Keele

(2007). Social capital and the dynamics of trust in government. American Journal of Political Science, 51(2), 241–254. https://doi.org/10.1111/j.1540-5907.2007.00248.x

42.

Kitazawa

Hale

S. A.

(2021). Social media and early warning systems for natural disasters: A case study of Typhoon Etau in Japan. International Journal of Disaster Risk Reduction, 52(2021), Article 101926. https://doi.org/10.1016/j.ijdrr.2020.101926

43.

Klein

Robison

(2020). Like, post, and distrust? How social media use affects trust in government. Political Communication, 37(1), 46–64. https://doi.org/10.1080/10584609.2019.1661891

44.

Kritzinger

Foucault

Lachat

Partheymüller

Plescia

Brouard

(2021). ‘Rally round the flag’: The COVID-19 crisis and trust in the national government. West European Politics, 44(5-6), 1205–1231. https://doi.org/10.1080/01402382.2021.1925017

45.

Lee

C. H.

(2020). The impact of language on retweeting during acute natural disasters: Uncertainty reduction and language expectancy perspectives. Industrial Management & Data Systems.

46.

Lee

Y. J.

(2022). Social media capital and civic engagement: Does type of connection matter? International Review on Public and Nonprofit Marketing, 19(1), 167–189. https://doi.org/10.1007/s12208-021-00300-8

47.

Levi

Stoker

(2000). Political trust and trustworthiness. Annual Review of Political Science, 3(1), 475–507. https://doi.org/10.1146/annurev.polisci.3.1.475

48.

M. H.

Feeney

M. K.

(2014). Adoption of electronic technologies in local US governments: Distinguishing between e-services and communication technologies. The American Review of Public Administration, 44(1), 75–91. https://doi.org/10.1177/0275074012460910

49.

Liang

Fung

I. C. H.

Tse

Z. T. H.

Yin

Chan

C. H.

Pechta

L. E.

Smith

B. J.

Marquez-Lameda

R. D.

Meltzer

M. I.

Lubell

K. M.

K. W.

(2019). How did Ebola information spread on Twitter: Broadcasting or viral spreading? BMC Public Health, 19(1), 438. https://doi.org/10.1186/s12889-019-6747-8

50.

Lin

Lachlan

K. A.

Spence

P. R.

(2016). Exploring extreme events on social media: A comparison of user reposting/retweeting behaviors on Twitter and Weibo. Computers in Human Behavior, 65, 576–581. https://doi.org/10.1016/j.chb.2016.04.032

51.

Liu

B. F.

Fraustino

J. D.

Jin

(2016). Social media use during disasters: How information form and source influence intended behavioral responses. Communication Research, 43(5), 626–646. https://doi.org/10.1177/0093650214565917

52.

Manetti

Bellucci

Bagnoli

(2017). Stakeholder engagement and public information through social media: A study of Canadian and American public transportation agencies. The American Review of Public Administration, 47(8), 991–1009. https://doi.org/10.1177/0275074016649260

53.

Margetts

John

Hale

Yasseri

(2015). Political turbulence: How social media shape collective action. Princeton University Press.

54.

Martinez-Rojas

Pardo-Ferreira

M. d. C.

Rubio-Romero

J. C.

(2018). Twitter as a tool for the management and analysis of emergency situations: A systematic literature review. International Journal of Information Management, 43, 196–208. https://doi.org/10.1016/j.ijinfomgt.2018.07.008

55.

Meijer

Koops

B.-J.

Pieterson

Overman

ten Tije

(2012). Government 2.0: Key challenges to its realization (10, pp. 59–69). Electronic Journal of e-Government.

56.

Mitchell

Gottfried

Barthel

Shearer

(2016). The modern news consumer. Pew Research Center. https://www.journalism.org/2016/07/07/the-modern-news-consumer/

57.

Miura

Toriumi

Komori

Matsumura

Hiraishi

(2016). Relationship between emotion and diffusion of disaster information on social media: Case study on 2011 Tohoku earthquake. Transactions of the Japanese Society for Artificial Intelligence, 31(1). https://doi.org/10.1527/tjsai.nfc-ec1

58.

Morgeson

F. V.

VanAmburg

Mithas

(2011). Misplaced trust? Exploring the structure of the e-government-citizen trust relationship. Journal of Public Administration Research and Theory, 21(2), 257–283. https://doi.org/10.1093/jopart/muq006

59.

Murthy

Longwell

S. A. T.

(2013). Disasters: The uses of Twitter during the 2010 Pakistan floods. Information, Communication & Society, 16(6), 837–855. https://doi.org/10.1080/1369118x.2012.696123

60.

Oude Groeniger

Noordzij

Van Der Waal

De Koster

(2021). Dutch COVID-19 lockdown measures increased trust in government and trust in science: A difference-in-differences analysis. Social Science & Medicine, 275(2021), Article 113819. https://doi.org/10.1016/j.socscimed.2021.113819

61.

Palen

Starbird

Vieweg

Hughes

(2010). Twitter‐based information distribution during the 2009 Red River Valley flood threat. Bulletin of the American Society for Information Science and Technology, 36(5), 13–17. https://doi.org/10.1002/bult.2010.1720360505

62.

Park

M. J.

Kang

Rho

J. J.

Lee

D. H.

(2016). Policy role of social media in developing public trust: Twitter communication with government leaders. Public Management Review, 18(9), 1265–1288. https://doi.org/10.1080/14719037.2015.1066418

63.

Pennebaker

J. W.

Boyd

R. L.

Jordan

Blackburn

(2015). The development and psychometric properties of LIWC2015. University of Texas at Austin.

64.

Porumbescu

G. A.

(2016). Linking public sector social media and e-government website use to trust in government. Government Information Quarterly, 33(2), 291–304. https://doi.org/10.1016/j.giq.2016.04.006

65.

Pressgrove

Kim

(2018). Stewardship, credibility and political communications: A content analysis of the 2016 election. Public Relations Review, 44(2), 247–255. https://doi.org/10.1016/j.pubrev.2018.01.003

66.

Putnam

R. D.

(2000). Bowling alone: The collapse and revival of American community. Simon & Schuster.

67.

Qazi

Imran

Ofli

(2020). GeoCoV19: A dataset of hundreds of millions of multilingual COVID-19 tweets with location information. SIGSPATIAL Special, 12(1), 6–15. https://doi.org/10.1145/3404820.3404823

68.

Recuero

Araujo

Zago

(2011). How does social capital affect retweets? Proceedings of the International AAAI Conference on Web and Social Media, Barcelona, Catalonia, Spain, July 17-21 5(1), 2011.

69.

Reddick

C. G.

Chatfield

A. T.

Ojo

(2017). A social media text analytics framework for double-loop learning for citizen-centric public services: A case study of a local government Facebook use. Government Information Quarterly, 34(1), 110–125. https://doi.org/10.1016/j.giq.2016.11.001

70.

Rudolph

T. J.

Evans

(2005). Political trust, ideology, and public support for government spending. American Journal of Political Science, 49(3), 660–671. https://doi.org/10.2307/3647738

71.

Shklovski

Palen

Sutton

(2008). Finding community through information and communication technology in disaster response. In Proceedings of the 2008 ACM Conference on Computer Supported Cooperative Work, San Diego, CA, November 8–12, 2008 (pp. 127–136).

72.

Silva

Tavares

A. F.

Silva

Lameiras

(2019). The good, the bad and the ugly: Three faces of social media usage by local governments. Government Information Quarterly, 36(3), 469–479. https://doi.org/10.1016/j.giq.2019.05.006

73.

Slatcher

Chung

Pennebaker

Stone

(2007). Winning words: Individual differences in linguistic style among US presidential and vice presidential candidates. Journal of Research in Personality, 41(1), 63–75. https://doi.org/10.1016/j.jrp.2006.01.006

74.

Song

Lee

(2016). Citizens’ use of social media in government, perceived transparency, and trust in government. Public Performance and Management Review, 39(2), 430–453. https://doi.org/10.1080/15309576.2015.1108798

75.

Steele

Dumbrell

(2012). Putting the public into public health information dissemination: Social media and health-related web pages. In Proceedings of the Seventeenth Australasian Document Computing Symposium, Dunedin, New Zealand, December 5–6, 2012 (pp. 135–138).

76.

Stieglitz

Dang-Xuan

(2013). Emotions and information diffusion in social media—sentiment of microblogs and sharing behavior. Journal of Management Information Systems, 29(4), 217–248. https://doi.org/10.2753/mis0742-1222290408

77.

Stroebe

vanDellen

M. R.

Abakoumkin

Lemay

E. P.

Jr. Schiavone

W. M.

Agostini

Bélanger

J. J.

Gützkow

Kreienkamp

Reitsema

A. M.

Abdul Khaiyom

J. H.

Ahmedi

Akkas

Almenara

C. A.

Atta

Bagci

S. C.

Basel

Berisha Kida

Bernardo

A. B. I.

Leander

N. P.

(2021). Politicization of COVID-19 health-protective behaviors in the United States: Longitudinal and cross-national evidence. PLoS One, 16(10), Article e0256740. https://doi.org/10.1371/journal.pone.0256740

78.

Sutton

Spiro

E. S.

Johnson

Fitzhugh

Gibson

Butts

C. T.

(2014). Warning tweets: Serial transmission of messages during the warning phase of a disaster event. Information, Communication & Society, 17(6), 765–787. https://doi.org/10.1080/1369118x.2013.862561

79.

Tausczik

Pennebaker

(2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29(1), 24–54. https://doi.org/10.1177/0261927x09351676

80.

Thelwall

(2020). Retweeting for COVID-19: Consensus building, information sharing, dissent, and lockdown life. arXiv preprint arXiv:2004.02793.

81.

Thorson

(2016). Belief echoes: The persistent effects of corrected misinformation. Political Communication, 33(3), 460–480. https://doi.org/10.1080/10584609.2015.1102187

82.

Tolbert

C. J.

Mossberger

(2006). The effects of e‐government on trust and confidence in government. Public Administration Review, 66(3), 354–369. https://doi.org/10.1111/j.1540-6210.2006.00594.x

83.

Torcal

(2014). The decline of political trust in Spain and Portugal: Economic performance or political responsiveness? American Behavioral Scientist, 58(12), 1542–1567. https://doi.org/10.1177/0002764214534662

84.

Trope

Liberman

Wakslak

(2007). Construal levels and psychological distance: Effects on representation, prediction, evaluation, and behavior. Journal of Consumer Psychology: The Official Journal of the Society for Consumer Psychology, 17(2), 83–95. https://doi.org/10.1016/S1057-7408(07)70013-X

85.

Wang

Zhuang

(2017). Crisis information distribution on Twitter: A content analysis of tweets during Hurricane Sandy. Natural Hazards, 89(1), 161–181.

86.

Wang

Hernandez

Newman

Bian

(2016). Twitter analysis: Studying US weekly trends in work stress and emotion. Applied Psychology, 65(2), 355–378. https://doi.org/10.1111/apps.12065

87.

Warren

A. M.

Sulaiman

Jaafar

N. I.

(2014). Social media effects on fostering online civic engagement and building citizen trust and trust in institutions. Government Information Quarterly, 31(2), 291–301. https://doi.org/10.1016/j.giq.2013.11.007

88.

Wojcik

Hughes

(2019). Sizing up Twitter users. Pew Research Center: Internet, Science & Tech; Pew Research Center: Internet, Science & Tech. https://www.pewresearch.org/internet/2019/04/24/sizing-up-twitter-users/

89.

World Health Organization . (2020). Communicating risk in public health emergencies: A WHO guideline for emergency risk communication (ERC) policy and practice. World Health Organization.