Abstract
Social media engagement—through reactions, comments, and shares—enables the public to interact dynamically with government-disseminated content. This engagement is valuable for real-time communication and for amplifying public awareness of programs, services, and rights. Despite its importance, prior research has overlooked the influence of government communication goals on such engagement. This article addresses that gap by examining objectives such as transparency, reputation management, coproduction, customer service, and citizen participation. Through an analysis of Facebook posts from 62 cities during Hurricane Florence, the article employs negative binomial regression modeling to explore how such goals impact public reactions, comments, and shares. Findings provide insights for aligning social media strategies with public administration objectives to increase engagement across various scenarios.
Key Points for Practitioners
Pursue different communication goals such as transparency, reputation management, coproduction, customer service, and citizen participation.
Recognize that each serves a distinct purpose and may influence public engagement in different ways depending on timing and context.
Ask direct questions or use conversation starters to promote dialogue.
Invest in quality visuals because photos and videos can boost post performance.
Know that reactions, comments, and shares may signify reach, but they do not always reflect meaningful dialogue or more in-depth forms of engagement.
Introduction
Social media engagement, defined as the extent to which users react to, comment on, and share posts, provides insight into how citizens interact with and perceive government communications (Bonsón & Ratkai, 2013). Such engagement is valuable because it facilitates real-time communication, measures public sentiment, and amplifies public awareness of programs, services, and rights (Bryer & Zavattaro, 2011; Mergel, 2017; Zavattaro et al., 2015). Moreover, engagement metrics partially reflect levels of public interest, the volume of community-generated content, and the degree of knowledge sharing related to government communications (Kagarise & Zavattaro, 2017). Although not without limitations, engagement metrics serve as indicators of how effectively government actors promote customized and responsive public administration.
Scholars have explored what distinguishes popular posts from less popular ones by examining factors such as agency characteristics, embedded multimedia, and the content's purpose or underlying meaning (Chen et al., 2020; DePaula & Dincelli, 2018; Goncalves et al., 2015; Lappas et al., 2018; Wang et al., 2020). Many studies narrowly define a post's underlying meaning by focusing primarily on specific service delivery areas instead of addressing the broader spectrum of government goals observed in practice (Bonsón et al., 2019; Sutton, Spiro et al., 2014). However, what are those prominent government communication goals that are observed in practice?
Government personnel promote transparency by providing updates on operations or future plans (Calhan et al., 2021; Mergel, 2012; Villodre, 2024). They enhance their agencies’ reputations through favorable presentations and other strategies (DePaula et al., 2018). Some focus on coproduction by collaborating with the public to create goods and services, while others emphasize customer service (Criado & Villodre, 2021). A smaller number solicit citizen participation to inform policymaking, although this remains an available option (Bryer, 2013). This article explores the relationship between these thematic content types and social media engagement.
Government communication goals are foundational to prescriptive theories of public administration: (a) transparency, coproduction, and citizen participation align with the principles of new public service (J. V. Denhardt & R. B. Denhardt, 2015; R. B. Denhardt & J. V. Denhardt, 2000); (b) customer service reflects the priorities of new public management (Osborne & Gaebler, 1992); and (c) reputation management underscores the importance of public perception (Carpenter & Krause, 2012). This article operationalizes such goals through specific message types disseminated during Hurricane Florence.
A disaster provides a particularly useful case study because it encompasses both routine and urgent operating conditions that demand different types of information within a short timeframe (Mileti, 1999; Tierney et al., 2001). Through an analysis of posts from 62 cities across North Carolina, South Carolina, and Virginia during Hurricane Florence, the study uses negative binomial regression models to measure the effects of these message types on per-post reactions, comments, and shares (i.e., public engagement). Findings highlight a messaging mix that reflects the communities’ changing needs and interests during the incident.
This article demonstrates the importance of integrating social media strategies with broader public administration goals. It offers practitioners a framework to enhance public engagement through targeted content. Although the findings focus on emergency management, the approach serves as a replicable model for other service delivery areas because it focuses on underlying goals rather than the specific service delivery area alone.
Messaging for Wide Audiences
Communicating effectively with wide audiences poses challenges, particularly during crises when diverse concerns and informational needs must be addressed without causing confusion or conflicting messages. The phenomenon of context collapse—when different audience groups converge on a single platform—further complicates consistency and self-presentation (Marwick & Boyd, 2011).
To navigate this complexity, government agencies often employ segmented messaging strategies. They tailor content for specific groups while maintaining a cohesive narrative. This approach enables communication professionals to pursue multiple public administration goals such as transparency, reputation management, coproduction, customer service, and citizen participation (Wukich, 2022). By integrating these themes, government agencies can address the diverse needs of their audiences while aligning their communication strategies with broader public administration objectives.
Government Goals and Their Influence on Engagement
These communication goals—transparency, reputation management, coproduction, customer service, and citizen participation—were chosen for this project because they align with prominent prescriptive theories of public administration, such as new public service (J. V. Denhardt & R. B. Denhardt, 2015; R. B. Denhardt & J. V. Denhardt, 2000), new public management (Osborne & Gaebler, 1992), and reputation management frameworks (Carpenter & Krause, 2012; Coombs & Holladay, 2023). Each goal represents a distinct yet often complementary dimension of effective governance. Together, they provide a foundation for exploring the interplay between government communication goals and social media engagement. This section lays the groundwork for the article's hypotheses by connecting these theoretical goals to their expected influence on public engagement with government social media content. The following subsections explain how these goals are applied in both routine and crisis situations and outline hypotheses regarding the anticipated effects of specific content types on reactions, comments, and shares.
Transparency
Transparency involves agencies that voluntarily share information about their operations and activities (Piotrowski, 2007; Villodre, 2024). This concept is integral to achieving accountability and trust by providing citizens with clear and factual updates about government actions. In routine contexts, transparency by cities might take the form of announcements about planned events, public works, or policy updates (DePaula & Dincelli, 2018; Lappas et al., 2018). However, during disasters, the need for transparency becomes more acute as the public seeks real-time information to understand rapidly changing circumstances (Chen et al., 2020; Sutton, Renshaw, Butts et al., 2020).
During disasters, cities promote transparency by providing clear and timely updates on their operations and activities. This includes sharing information about the efforts of city departments, such as public works crews clearing roads or restoring services. Cities also keep residents informed about executive actions, appearances, or schedule updates including, for example, announcements from mayors about emergency response plans or disaster declarations (Wukich, 2021). These updates help residents stay informed about government activities and decisions during critical moments. Given the critical role of transparency, it is expected that such communications will positively influence public engagement during disasters.
Hypothesis 1 (H1): Facebook posts containing content about city operations or executive activities will receive more reactions, comments, and shares compared to posts without such content.
Reputation Management
Government communication professionals work to promote a favorable public image through positive presentations and symbolic acts (DePaula et al., 2018; Liu & Horsley, 2007; Zavattaro, 2013). In routine contexts, this may include showcasing achievements, expressing gratitude to the community, or referencing cultural values to build goodwill. During a crisis, however, reputation management takes on a heightened importance, for agencies must address public concerns to maintain trust in their leadership (Olsson, 2014). Public relations efforts frequently attempt to deflect criticism and avoid blame (Boin et al., 2017). Coombs (2007) contends that protecting an organization's reputation during a crisis requires demonstrating competence, compassion, and alignment with community values.
Favorable presentation, as a tactic, projects likability and competence by promoting the actions and dedication of government officials, while symbolic acts, such as expressions of sympathy or solidarity, connect with community emotions and reinforce shared values (DePaula et al., 2018). In both routine and disaster situations, engagement patterns with these messages can vary based on the content type and context, for such posts often elicit reactions and comments but may be less likely to encourage sharing (DePaula & Dincelli, 2018; Lappas et al., 2018; Sutton, Gibson et al., 2015; Sutton, Spiro et al., 2014).
Hypothesis 2 (H2): Facebook posts containing content about favorable presentations or symbolic acts will receive more reactions and comments but fewer shares compared to posts without such content.
Coproduction
Coproduction involves agencies and communities jointly delivering services or creating public goods (Nabatchi et al., 2017). During disasters, coproduction can include one-way prompts for public action, such as general warnings and advisories that empower community members to reduce risks and enhance public safety (Wukich, 2022). In the context of a hurricane, these messages may include evacuation notices, landfall warnings, hazard impact updates, road closure announcements, curfew directives, resource availability information, and a range of other advisories (Jiang & Tang, 2023; Sutton, Gibson et al., 2015; Sutton, Renshaw, Butts et al., 2020; Sutton, Renshaw, Vos et al., 2019; Sutton, Spiro et al., 2014; Wukich, 2016).
Hypothesis 3 (H3): Facebook posts containing content about warnings or guidance on risk-reduction actions will receive more reactions, comments, and shares compared to posts without such content.
Authorities can also request information from the public to enhance situational awareness and operational decision-making, which typically increases engagement (Atkinson & Wald, 2007; Wukich, 2022).
Hypothesis 4 (H4): Facebook posts containing intelligence-gathering content requesting public feedback will receive more reactions, comments, and shares compared to posts without such content.
Customer Service
Customer service content emphasizes responsiveness and convenience as well as promoting municipal services and community events (Criado & Villodre, 2021; Wukich, 2021). While such content may normally see varied engagement (Lappas et al., 2018), disasters can heighten public interest in service-related information, consistent with a focus on serving the public effectively (Mileti, 1999; Sutton, Renshaw, Butts et al., 2020; Tierney et al., 2001; Waxa & Gwaka, 2021).
Hypothesis 5 (H5): Facebook posts containing content about service disruptions, event cancelations, or service fees will receive more reactions, comments, and shares compared to posts without such content.
Citizen Participation
Citizen participation involves meaningful engagement between authorities and the public about future policy directions (Bryer, 2013). In routine contexts, this may include soliciting feedback on proposed policies, inviting community input on city projects, or organizing town hall discussions (Nabatchi & Leighninger, 2015). During disasters such as hurricanes, the opportunities for citizen participation will likely shift toward recovery efforts and planning how communities can rebuild after the incident. This includes engaging the public in discussions about policy decisions related to infrastructure repair, housing redevelopment, and long-term resilience strategies (Hamideh, 2020; Pyles et al., 2018). However, authorities often hesitate to facilitate this dialogue due to concerns about blame, narrative control, or the challenges of managing diverse perspectives (Baltz, 2023; Boin et al., 2017; Bryer & Zavattaro, 2011; DePaula & Dincelli, 2018).
Hypothesis 6 (H6): Facebook posts containing content about citizen involvement efforts will receive more reactions, comments, and shares compared to posts without such content.
Goal-Spanning Message Types
Rumor management involves identifying and correcting misinformation that is contradictory to several government goals such as transparency, reputation management, coproduction, customer service, and citizen participation (Panagiotopoulos et al., 2016; Wukich, 2022, 2023).
Hypothesis 7 (H7): Facebook posts containing content about rumor management will receive more reactions, comments, and shares compared to posts without such content.
Conversation starters are used to initiate dialogue that promotes preparedness and engages the public in policy discussions, which in the context of emergency management includes conversations about rebuilding and other recovery efforts (Jackson & Foucault Welles, 2015; Wukich, 2016).
Hypothesis 8 (H8): Facebook posts containing content about conversation starters will receive more reactions, comments, and shares compared to posts without such content.
Research Design
Case Selection
The catastrophic event that is the basis for this study is Hurricane Florence in September 2018 that caused significant damage in North Carolina, South Carolina, and Virginia. Each state issued disaster declarations due to the extensive impact that included mandatory evacuations, flood damage from storm surges, wind damage, river flooding from excessive rainfall, widespread closures, power outages, and critical infrastructure failures. The estimated damages totaled $24 billion, with 52 fatalities—22 directly and 30 indirectly (Stewart & Berg, 2019).
Local governments in the United States hold primary responsibility for emergency management (Mileti, 1999). During Hurricane Florence, cities pursued various content goals over a short period. Specific message types relevant to the case (e.g., warnings, advisories, service disruption notices) were identified and linked to broader government goals to understand their effect on user engagement. A case study approach allows for in-depth investigation with the opportunity to study multiple cities, and 62 cities with populations of 30,000 or more were selected; there was the assumption that larger localities would exhibit more sophisticated social media capabilities (Mossberger et al., 2013). These 62 cities represent a complete census of all cities with populations over 30,000 in the three states.
Data Collection
The cities’ official Facebook pages provided data on engagement through their original posts and subsequent reactions, comments, and shares. Facebook, with a larger user base than Twitter in the United States, offers a comprehensive view of online activities. Official accounts were identified by reviewing city websites and confirming Facebook's blue badges for authentication. The 62 account names were used to seed Netlytic data collection software, which collected all city Facebook posts over a 21-day period—from 7 days before the hurricane's landfall to 14 days after. This period captures key activities, including the storm's forecast, initial alerts, community response, and early recovery efforts (Comfort & Haase, 2006).
In total, 1,849 original city posts were identified; shared posts from other accounts were excluded from the analysis. Data were organized chronologically in a spreadsheet, including the text for each post and the number of reactions, comments, and shares that were manually added. Netlytic provided hyperlinks for each post, which expedited the data collection process. The text enabled a deeper content analysis.
Content Coding
Four researchers collaboratively created a codebook, derived thematic categories from relevant literature, and collected information inductively from the data. For example, risk communication literature informed disaster-related coproduction message types—warnings, advisories, evacuations, and closures (Hughes et al., 2014; Sutton, Gibson et al., 2015; Sutton, Spiro et al., 2014). Customer service-related categories (e.g., service disruption, event disruption, fee reversals) emerged from an inductive reading of the data. Table 1 lists and defines each category.
Government Communication Goals, Related Message Types, and Hypotheses.
Government Communication Goals, Related Message Types, and Hypotheses.
Two coders categorized each post and noted every message type—whether a post contained a message type (=1) or not (=0). For example, Holly Springs, North Carolina advised, “Please stay off the roads today for your safety and to keep roads clear for emergency vehicles;” this post was categorized as an advisory.
Some posts contained multiple message types, which reflected the pursuit of multiple goals. For instance, Winston-Salem, North Carolina posted a photo with the caption: “Wow—Florence left behind roughly 60 tons of debris on this bridge! Our crews have been working hard to clean up after the storm.” This post was categorized into three message types: (a) hazard impact (“60 tons of debris on this bridge”), (b) favorable presentation (“working hard”), and (c) city operations (“clean up”). Coders noted these affiliations accordingly. Interrater reliability was measured using Cohen's Kappa coefficient, with scores ranging from .82 to .96, which indicated strong to almost perfect agreement (McHugh, 2012).
Three empirical models of engagement were estimated to examine the relationships between thematic content (i.e., government goals) and the number of reactions, comments, and shares per Facebook post. Negative binomial regression was used because the dependent variables constitute over-dispersed count data, allowing the conditional variance to exceed the conditional mean (Hilbe, 2011).
Each message type provided useful information that plausibly drew public attention, although some negative effects were expected. For example, favorable presentation and symbolic acts likely negatively affected share rates (DePaula & Dincelli, 2018). Table 1 lists expected effects for each independent variable.
The models included fixed effects for each city (Facebook account name) to control for city or account-level variations; there was the acknowledgment that factors exogenous to thematic content and embedded multimedia likely also affected engagement rates (Sutton et al., 2020; Wang et al., 2020). City Facebook account names were included as dummy variables, with the Virginia Beach, Virginia account as the baseline for comparison. Account names were selected as controls after exploring alternative variables independently and in combination (e.g., population, follower count, standardized follower count by population, state, geographic proximity to the coast). The selected models (with fixed effects by account name) resulted in better (i.e., lower) overall corrected Akaike information criterion scores that informed the model selection process.
Multipliers—or the expected percent change in engagement scores for posts containing a particular content category compared to those without the same content—provided an easily interpretable measure of effect size. Sutton, Gibson et al. (2015) explained: [W]e describe effect sizes … in terms of the number of additional retweets that would be gained or lost relative to the baseline upon adding or removing a message feature. Thus, a feature that multiplies the expected retweet rate by a factor of 1.5 is described as adding 50% more retweets, while a feature that multiplies the rate by a factor of 0.75 is described as resulting in 25% fewer retweets. (p. 12)
Table 2 includes the models’ coefficients, standard errors, and multipliers. Multipliers for the reaction, comment, and share rate models were calculated “by subtracting 1 from the exponentiated beta coefficient and multiplying by 100” (Renshaw et al., 2021, p. 36). This analysis underpins the findings section, which is augmented by illustrative message type examples.
Negative Binomial Models Using Communication Goals and Embedded Multimedia to Predict Number of Per-Post Reactions, Comments, and Shares.
Negative Binomial Models Using Communication Goals and Embedded Multimedia to Predict Number of Per-Post Reactions, Comments, and Shares.
*p < .05 | **p < .01 | ***p < .001.
Note: Multipler/% Change was calculated by taking Exp(B) − 1 × 100 (see Renshaw et al., 2021).
Note: Fixed effects for city Facebook account names were included in the models to control for city or account-level variation but are not shown in this table. The baseline account Virginia Beach, VA demonstrated statistically significant differences when compared against 42 of the 61 other accounts.
The goal of this article was to examine how certain Facebook content posted by city governments influenced public reactions, comments, and shares during Hurricane Florence. Results from three negative binomial regression models provide insights into which content types correlate with engagement.
Transparency
Content focused on transparency, such as updates about city operations or executive activities, demonstrated varying impacts engagement. Operational updates were associated with a 25.0% increase in reactions and a 26.9% increase in comments but did not significantly affect the extent to which the public shared content. Posts describing executive activities resulted in 131.6% more comments and 27.8% more shares but showed no significant change in reactions. These findings suggest that transparency during crises encourages public engagement, especially through comments and shares, as people seek clear and actionable information about government actions. However, the lack of increased reactions for executive activities suggests that transparency-driven content may resonate more as a discussion starter than as content that elicits immediate emotional responses. Therefore, we can only partially confirm Hypothesis 1 that transparency will positively affect all engagement metrics.
Reputation Management
Hypothesis 2 predicted that reputation management posts would increase the number of reactions and comments but decrease share rates. The results for favorable presentation posts partially support this hypothesis. These posts were associated with a 36.6% increase in reactions and a 21.7% decrease in shares, while the effect on comments was statistically insignificant. These results suggest that although audiences were more likely to react to descriptions of positive activity, they were less likely to amplify it by sharing.
Posts featuring symbolic acts, such as expressions of gratitude, sympathy, or solidarity, increased reactions and comments, which aligns with Hypothesis 2. However, symbolic acts were also associated with increased shares, which runs counter to the hypothesis that people are less likely to repost such content. This finding suggests that symbolic content may resonate more deeply during disasters, when public emotion is heightened and people are more inclined to share affirmations of community values and support.
Coproduction
Hypothesis 3 predicted that coproduction-related content, such as warnings, advisories, evacuations, hazard impacts, closures, curfews, and resource provision, would increase public engagement. The results support this hypothesis for some but not all variables. Posts about evacuations generated 67.2% more reactions, 126.6% more comments, and 114.8% more shares compared to posts without evacuation content. Similarly, posts describing hazard impacts, such as damage reports or flooding, produced 98.5% more reactions, 207.7% more comments, and 205% more shares. Curfew-related posts also generated higher levels of reactions, comments, and shares.
Other content types produced mixed or non-significant results. For example, warnings were not associated with increased reactions, and advisory posts did not significantly affect reactions or comments. Posts about closures were associated with a 20.6% decrease in reactions and showed no significant increases in comments or shares. Resource provision posts, such as announcements about sandbag availability, had no meaningful effect on reactions but did increase shares by 116.9% and comments by 93.3%. These results indicate that while certain forms of coproduction variables correlate positively with all engagement metrics, others do not. Overall, Hypothesis 3 is only partially confirmed.
Hypothesis 4 surmised that posts requesting public feedback to support intelligence-gathering efforts would increase engagement across all three metrics, but results suggest otherwise. Intelligence gathering posts were associated with a −28.3% decrease in reactions, and although comments and shares also declined, these effects were not statistically significant. Therefore, Hypothesis 4 is rejected.
Customer Service
Hypothesis 5 predicted that customer service content, such as posts about service disruptions, event cancelations, or fee exemptions, would increase reactions, comments, and shares. However, findings largely contradict this expectation. Posts about service disruptions, such as changes to trash collection schedules or utility services, were not significantly associated with increased reactions or shares; only comment rates saw a modest and statistically significant increase of 36.4%. In contrast, posts about event disruptions showed a consistent and significant decrease in engagement across all metrics: reactions (−53.9%), comments (−71.5%), and shares (−68.1%). Additionally, fee exemption posts did not produce statistically significant changes in any engagement metric. Taken together, these findings lead to a rejection of Hypothesis 5.
Citizen Participation
Citizen participation content that was aimed to involve the public in decision-making related to recovery efforts showed limited effectiveness. Thus, we can reject Hypothesis 6. Posts inviting public involvement resulted in a 68.4% decrease in reactions, indicating both a reluctance from authorities to solicit input and a lack of public interest in engaging with such content during the crisis. This category contained only eight posts, which limits confidence in the results. Still, the small sample size and lack of subsequent engagement underscore the challenges of fostering meaningful participation in high-stress scenarios.
Other Content Types
Rumor management posts were not effective in increasing engagement, which leads to a rejection of Hypothesis 7. Similarly, posts designed as conversation starters failed to drive significant interactions; therefore Hypothesis 8 is rejected as well. However, both message types were infrequently used, rumor management appeared in only 24 posts and conversation starters in just 5, which limits the strength of these findings and suggests the need for further study with larger samples.
Discussion
This article examines how government communication goals influence social media engagement during disasters. It applies concepts advanced in theoretical frameworks such as new public service, new public management, and reputation management. By explicitly grounding the analysis in public administration theories, the article provides a more precise evaluation of social media content and its impact on social media engagement. This approach contrasts with prior research that relies on vague content categorizations (Lovejoy & Saxton, 2012) or policy domain-specific content (Bonsón et al., 2019; Sutton, Spiro et al., 2014). Therefore, it offers clearer insights into what drives public engagement.
Findings highlight the contextual nature of social media engagement. Reactions, comments, and shares varied widely across communication goals and specific content types. Transparency-related content, particularly posts about executive activities and operational updates, generated increased engagement—especially through comments and shares. These results suggest that the public responds to transparency during crises, especially when the information is actionable or relevant to immediate concerns. Under routine operating conditions, transparency efforts such as policy updates or public works announcements generate less engagement (DePaula & Dincelli, 2018). The heightened public attention during a crisis likely elevates the perceived relevance and urgency of such content.
Reputation management content generally increased reactions and comments, but showed mixed results on sharing. Symbolic acts, such as expressions of gratitude, sympathy, or solidarity, boosted engagement across all three metrics, contradicting the expectation that such posts would increase reactions and comments but reduce sharing. During routine, business-as-usual conditions, DePaula and Dincelli (2018) report that symbolic content tends to generate more reactions and comments but not shares, likely because audiences view such posts as emotionally resonant but not actionable—not worth sharing. However, in the context of a disaster, symbolic messages may serve a different function. Symbolic acts may help to affirm shared values, demonstrate empathy, and reinforce community solidarity—factors that may become more salient in high-stress situations. As a result, audiences may be more inclined to share these posts as public affirmations of support. In contrast, favorable presentation posts increased reactions but decreased shares and had no significant effect on comments. These findings suggest that overt self-promotional content may have a more limited reach than symbolic acts during crises.
Coproduction effects on engagement varied. On the one hand, specific and detailed posts about evacuations, hazard impact, and curfews were strongly associated with increased engagement across metrics. On the other hand, other coproduction-related messages, such as warnings and more general advisories, did not consistently produce significant effects, and closure posts were associated with reduced engagement. These differences suggest that content perceived as more urgent, specific to people's situations, or impactful prompts stronger engagement than more general advisories.
These findings raise important questions about how the public engages with coproduction content under routine, non-disaster conditions. During emergencies, messages that clearly convey urgency and relevance prompt strong engagement. However, in everyday contexts, government efforts to involve the public in service delivery often yield limited interaction, especially when messages emphasize compliance or behavioral change without inviting meaningful feedback (Reddick et al., 2017). As Criado and Villodre (2021) note, routine use of social media by local governments tends to center on broadcasting information rather than enabling coproduction. Future research could explore whether public engagement with coproduction messages increases when messages are framed in ways that encourage ongoing participation, problem-solving, or learning.
Customer service posts, such as those addressing service disruptions or event cancelations, showed mixed impacts on engagement. While posts about service disruptions led to increased comments, event cancelation posts generally correlate with lower engagement across all metrics. These findings contrast with expectations derived from new public management principles, which emphasize residents as customers (Criado & Villodre, 2021; Osborne & Gaebler, 1992). One possible explanation is that, during disasters, audiences prioritize urgent or safety-related content over routine service updates. In non-crisis conditions, posts about events and city services may draw more attention, but in an emergency, such content may be perceived as irrelevant or even distracting.
The relatively few posts that sought feedback or conversation) whether for intelligence gathering, citizen participation, or conversation starters) either negatively affected engagement or had no effect. While dialogue aligns with democratic ideals (Nabatchi & Leighninger, 2015), its effectiveness may depend on the context in which it is presented and the audience's capacity to engage in decision-making at the time. This finding shows the importance of choosing the right time and approach when using participatory messages. It may also suggest that people are not used to being asked for feedback because public officials rarely do so (Lappas et al., 2018; Meijer & Thaens, 2013; Mergel, 2013; Sobaci & Karkin, 2013; Wells et al., 2025). In the absence of a familiar pattern or clear incentives for participation, such requests may be overlooked or dismissed, particularly during a crisis when users are focused on receiving, rather than providing, information.
The limited use of interactive strategies identifies a potential area for improvement. This reluctance may stem from resource limitations or fears of public criticism, as exemplified by the New York City Police Department's #myNYPD campaign that backfired and resulted in widespread negative feedback (Jackson & Foucault Welles, 2015). Given the spread of misinformation, abusive behavior, and political polarization on social media, meaningful conversations may be difficult to mediate (Feeney & Porumbescu, 2021).
Finally, rumor management efforts saw low levels of engagement, despite their potential value in correcting misinformation. This may reflect the difficulty of building trust and capturing attention during a crisis, when the public is more focused on actionable safety information than on evaluating the accuracy of claims. Additionally, the low volume of rumor management posts suggests that many local governments did not prioritize this strategy or lack the resources to implement it consistently.
Limitations
While emergency management provides a compelling case study, as demonstrated during Hurricane Florence, the insights gained are not necessarily applicable across all service delivery areas. However, by focusing on core objectives such as transparency, reputation management, coproduction, customer service, and citizen participation, this research can be replicated across different contexts. The context of each case study will likely influence the specific needs and public preferences. Therefore, a customized approach is needed.
Another limitation of this article is the omission of the structural aspects of social media communication, as discussed by Meijer and Thaens (2013), Mergel (2013), and Hu et al. (2024). While this article focuses on the content of government goals such as transparency, customer service, and coproduction, it does not address communication structure, such as one-to-many, one-to-one, or many-to-many interactions. This omission is noteworthy because such interactions with the public may reflect efforts to achieve transparency (Villodre, 2024) or citizen participation (Wukich, 2022, 2023). Future studies should examine both the content and structural dimensions of social media engagement to provide a more comprehensive understanding of how government messages are received and interpreted by the public.
Another limitation concerns the small sample sizes for several variables tested. Specifically, there were relatively few posts representing the variables tested in Hypothesis 4 (intelligence gathering with 41 posts), Hypothesis 6 (customer involvement with 8 posts), Hypothesis 7 (rumor management with 24 posts), and Hypothesis 8 (conversation starters with 5 posts). These small sample sizes raise concerns about the robustness and reliability of the results for these hypotheses, since they constitute half of the hypotheses tested. This limitation should be acknowledged, for it may affect the confidence in the conclusions drawn from these specific variables. Future studies should collect a larger number of posts for these categories to ensure more robust and reliable results.
Other limitations include the uniform modeling of the reaction variable, which does not account for differences among reaction types (e.g., like, love, haha, wow, sad, angry). Future studies should consider a more detailed analysis of reactions, because different content may evoke varied emotional responses. Additionally, the identity of the public who engage with city government content remains unclear. It is uncertain whether the Facebook users who interacted with posts were primarily local residents or outside observers. Furthermore, this article's approach to embedded multimedia could be more nuanced by incorporating the substantive meanings conveyed by photos, memes, infographics, and videos (Gherheș et al., 2023; Imran et al., 2020; Renshaw et al., 2021).
Recommendations for Practitioners
Engagement metrics should be carefully considered. While event disruptions, intelligence gathering, and rumor management messages presumably provide public value, they had little effect on engagement rates. To increase engagement, cities might modify their tactics, such as asking for direct feedback on Facebook instead of promoting other channels. Practitioners can also employ private sector marketing techniques to boost engagement scores by prioritizing positive content, curating high-quality visuals, being relatable, interacting with audiences, and asking questions (McLachlan & Newberry, 2021).
However, authorities should not deviate from evidence-based, risk communication best practices simply to achieve higher engagement scores. Strategies that involve negative, needlessly emotional, or unnecessarily provocative content should be avoided. Instead, public goals and values should be prioritized, as illustrated in this article.
Engagement metrics alone do not generate substantive community engagement. Nabatchi and Leighninger (2015) described reactions, comments, and shares as the thinnest options for engagement. Without government follow-up, these interactions remain one-way ties. Public officials who want to generate substantive dialog should diversify their engagement efforts beyond Facebook by using participatory budgeting, citizen assemblies, and other applications designed for deliberative dialogue (Hofstra et al., 2023; Nabatchi & Leighninger, 2015; Zingale, 2013).
Conclusion
The research approach outlined in this article acknowledges the context-specific nature of government social media use and is replicable. Future researchers could adopt a two-step process of first identifying case-specific message types and then connecting them conceptually with government communication goals applicable across different service delivery areas and policy domains. This approach provides granular specificity useful for practitioners while accounting for fundamental concepts in public administration theory.
Footnotes
Acknowledgements
The author thanks Kirby Suntala, Gabrielle Parsson, Jose Mendez, LaNiqua Jones, and Alex Mills for their data collection and analysis efforts. The project was supported by Cleveland State University's Office of Research and the Maxine Goodman Levin School of Urban Affairs. The author used ChatGPT to support the revision of this article, particularly for editing text to improve clarity and grammar. Special thanks to Ronda Harris for her meticulous copyediting as well.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.