Connecting Mayors: The Content and Formation of Twitter Information Networks

Abstract

Politicians interact with each other in polarized and partisan ways when they correspond on social media. Meanwhile, local governments share more substantive information about their service delivery efforts. Mayors operate in both spaces simultaneously (i.e., politics and administration), yet little has been reported about how they balance these positions online. This article examines (1) the extent to which mayors who govern the 100 largest U.S. cities form ties with each other, (2) the messages they create, and (3) the factors that influence their associations. Content analysis identifies common message types, including favorable presentation, symbolic acts, and political positioning statements. Results from quadratic assignment procedure indicate mayors form ties with high-profile counterparts from the same political party who represent larger cities and possess more followers (i.e., more popular accounts). The findings suggest tie formation has less to do with the exchange of best practices—as in policy networks—and more to do with self-promotion and political marketing to constituents at home.

Keywords

social media information networks mayors

Introduction

Advocates of social media applications such as Twitter initially hailed their potential to facilitate collaboration between diverse actors (Mergel 2012; Roberts 2011). Although politicians primarily use Twitter to engage constituents, they also share information with each other (Gainous and Wagner 2014), and polarization and partisanship largely characterize the subsequent political discourse (Boulianne 2016; Jungherr 2016). For example, evidence from Twitter reveals that members of the U.S. Congress use it to launch political attacks (Russell 2018b) and advance their political brands (Golbeck, Grimes, and Rogers 2010). However, mayors purport to operate in a different political space. They describe themselves as more pragmatic and willing to cross-ideological lines (Emanuel 2020; Nutter 2018). Some are innovators who coordinate with each other to confront difficult challenges (Barber 2017; Hughes 2017). One might surmise that these behaviors and tendencies manifest on social media. But do they? To what extent do mayors interact with each other on Twitter, what types of information do they share, and what factors influence their interactions?

Scholars distinguish between how politicians and government agencies use social media, and as a result, two separate research streams have emerged. The former focuses on electioneering and self-promotion by officeholders (mainly legislators), candidates, and party activists (Evans, Cordova, and Sipole 2014; Gainous and Wagner 2014; Golbeck, Grimes, and Rogers 2010; Jungherr 2016), while the latter focuses on policy implementation by agency staff (De Widt and Panagiotopoulos 2018; Mergel 2012, 2016; Wukich and Mergel 2016; Zavattaro 2016). Mayors would seem to operate in both spaces simultaneously (i.e., politics and administration), yet little has been reported about how they balance these positions online.

This article examines (1) the extent to which U.S. city mayors use Twitter to connect with other mayors, (2) the messages they produce, and (3) the determinants of tie formations that are enabled by Twitter’s interactive functions, including mentions, replies, follows, and retweets. Network analysis of 32 months of Twitter data reveals that while most large city mayors—the top 100 in the United States by population—form ties with other mayors, they do so intermittently as opportunities arise. Content analysis identifies common message types including favorable presentation, symbolic acts, and political positioning statements. Results from quadratic assignment procedure indicate mayors tend to form ties with high-profile counterparts from the same political party who represent larger cities and possess more followers (i.e., more popular accounts).

These findings suggest that tie formation has less to do with the exchange of best practices (something that might be expected in a more substantive policy network) and more to do with what Mayhew (1974) called electorally oriented activities and relatedly projecting likability, competence, and worthiness to constituents (DePaula, Dincelli, and Harrison 2018). These findings show mayors prioritize political marketing over more substantive dialogue, as do members of Congress (see Golbeck, Grimes, and Rogers 2010; Russell 2018a). Nevertheless, the content of the mayors’ social media messages appears to be more positive than that of members of Congress (see Russell 2018b), and when mayors do post negative rhetoric, they direct it at federal officials from the opposing party rather than toward other mayors.

Twitter Functions That Facilitate Tie Formation

Twitter allows politicians to communicate directly with a number of different actors, and many prefer this method to Facebook and other social media applications (Evans, Brown, and Wimberly 2018; Jungherr 2016; Russell 2018a). Twitter offers “an alternative environment by which opinion leaders, politicians, and citizens can engage with each other . . . without limits from borders or geography” (Gainous and Wagner 2014, p. 3), and they can do so by posting and reading other people’s messages or more directly by using a set of Twitter functions, which include mentions, replies, following, and retweets.¹

A mention occurs when a user includes another account’s username in a Tweet. The mentioned account then receives a notification. A reply represents a response to another person’s tweet. Again, that account receives a notification. By following another account, a user subscribes to that content to see new tweets as soon as they are posted. The followed account receives notification. Finally, a retweet occurs when a user forwards another account’s Tweet to their followers. Often used to pass along news or other valuable discoveries, retweets retain original attribution, and the original account receives a notification when retweeted (Twitter 2018). Each of the functions—mentions, replies, follows, and retweets—create a public record of asymmetric ties between accounts.

Because technical capability alone does not determine how people employ technology (Fountain 2001; Rethemeyer 2007), interactions between politicians are not inevitable. Politicians and staff must decide to employ the options; otherwise, social media use simply becomes an extension of preexisting one-way communication practices in which accounts are used to disseminate messages (i.e., broadcasting) with no intent to interact or engage in dialogue with others (Golbeck, Grimes, and Rogers 2010; Jungherr 2016). Mergel (2012) referred to this approach as a press release style for social media use. Nonetheless, many politicians do engage in direct communication with other Twitter users, whether within ideologically bounded echo chambers with party activists (Barberá 2017) or as part of negative attacks against opposing parties (Russell 2018b). To better understand mayors’ behavior, this article bridges two different streams of literature on Twitter use—Twitter use by politicians and Twitter use by local governments. These literature streams are presented in the following sections.

Twitter Use by Politicians

On social media, politicians develop and market their political brands both during campaigns (Boulianne 2016; Evans, Cordova, and Sipole 2014) and while in office (Hong 2013; Russell 2018a). According to Golbeck, Grimes, and Rogers (2010), their overarching goal is self-promotion. Mayhew (1974) concluded that the top priority of members of Congress is reelection and that they engage in three basic types of electorally oriented activities while in office: advertising, credit claiming, and position taking. Applying Mayhew’s (1974) typology, Russell (2018a) reported that U.S. Senators use Twitter most frequently to comment on proposed legislation and roll call votes (i.e., position taking). However, they also advertise their public appearances (i.e., advertising) and claim credit for government actions (i.e., credit claiming), although not as regularly. These message types (e.g., position taking, advertising, and credit claiming) align with what DePaula, Dincelli, and Harrison (2018, p. 102) referred to as favorable presentation, or messages designed to “seek attribution of likability, competency or worthiness” by reporting “positive activity . . . with positive imagery or self-referential language.” Therefore, favorable presentation is a popular approach for U.S. Senators.

While campaigning, politicians communicate their values, describe their policy positions, and critique their opponents (Evans, Brown, and Wimberly 2018; Evans, Cordova, and Sipole 2014). Messages that offer a critique of a politician’s opponents warrant attention. During campaigns, challengers are more likely to use negative rhetoric (words such as angry, sad, and anxious) than do incumbents, and incumbents are more likely to react negatively when faced with more competitive opponents (Gervais, Evans, and Russell 2020). However, Russell (2018b) demonstrated that negative rhetoric continues while in office. She argued that “social media introduce a fundamentally different relationship between elite actors and the spread of partisan rhetoric by creating an easily accessible and transparent record of party-polarizing priorities” (Russell 2018b, p. 698). Relatedly, partisan rhetoric, both negative blame shifting messages and positive loyalty statements, is common, particularly among party members in opposition to the president. Still, not all rhetoric posted on social media contains negative or partisan comments.

While many scholars focus their attention on Congress, literature on how mayors use social media is far less prevalent. Criado, Martínez-Fuentes, and Silván (2012) reported on Twitter use by Spain’s 2011 mayoral candidates to amplify their message and increase their community of contacts. Sobaci and Karkin (2013) analyzed Twitter content from 16 municipal mayors in Turkey during a six-month period in 2012. They concluded the mayors also used Twitter “for the purposes of self-promotion and political marketing” (p. 417). Occasionally, some encouraged residents to access public services or attend community events. This suggests mayors at least acknowledged their roles as executives who were implementing local government service delivery in addition to their self-promotion.

Twitter Use by Local Governments

While many mayors seek re-election and some desire higher office (Einstein et al. 2020), they also engage in day-to-day governing and, therefore, blur the lines of the so-called politics-administration dichotomy (Svara 1998). Consequently, understanding how local governments share information and interact with each other is pertinent to this study because mayors may employ similar tactics.

Local governments disseminate a range of content on Twitter, some of which is more conducive to engaging other local governments than others. Message types include (1) public service announcements, (2) the promotion of local government programs, and (3) symbolic acts used to express congratulations, celebrate holidays, and reference cultural symbols like sporting events (DePaula, Dincelli, and Harrison 2018, p. 102). Some content is employed to market community businesses and quality-of-life (Kagarise and Zavattaro 2017; Zavattaro 2016) and respond to public concerns (Brainard and Edlins 2014). While community members represent the target audience, elected officials and staff from other local governments may be interested in that content—perhaps to learn about best practices or replicate it in their communities.

Fewer municipalities share best practices by engaging other local governments more directly via conversations. In one example from the United Kingdom, De Widt and Panagiotopoulos (2018) described Twitter conversations between local government personnel that dealt with austerity measures and were organized by the hashtag #localgov. Participants (1) shared and commented on related news, (2) criticized the budget cuts, and (3) put forth ideas to cope specific to their policy domains and service areas. Mayors may not employ all message types or engage in similar dialogic conversations. Mergel (2016, p. 147) warned against conflating tactics across all Twitter accounts in this manner: “Mayors have different intentions in their online interactions than for example police departments. Political actors, such as elected officials, are constantly on the campaign trail and are less likely to deliver online services through social media channels.” To some extent, Sobaci and Karkin (2013) validated Mergel’s (2016) argument in the Turkish context, although mayors did promote public services to some extent.

Determinants of Tie Formation

Mergel (2017) illustrated the process through which government personnel scan social media conversations for opportunities to provide feedback or engage in dialogue that advances their social media goals. Based on these findings, mayors and their communications teams may monitor for opportunities to cultivate their online networks with other mayors who might help them. Election cycles provide opportunities for this engagement (Boulianne 2016; Jungherr 2016) as do major policy decisions that affect multiple jurisdictions (De Widt and Panagiotopoulos 2018). Other focusing events such as disasters lead authorities to express condolences, offer support, and try to make meaning from what transpired during an extreme event (Stern 2017). It is within this context that mayors form ties.

While events such as elections, major policy decisions, and disasters create opportunities for interaction between mayors on Twitter, additional factors shape and constrain the decision to form ties (i.e., mention, reply to, follow, or retweet another mayor). Theories of electorally oriented behavior while in office (Mayhew 1974) and prestige prominence, or the popularity a mayor exhibits among their peers (Knoke 1994), are relevant with regard to the extent that mayors self-promote by associating with other high-profile mayors. Theories of policy network formation and institutional collective action between elected officials may be more applicable when mayors seek out best practices and substantive dialogue (Feiock et al. 2010). By leveraging those theories and taking into consideration the growing body of scholarship on social media, several possible factors are germane, including (1) political party homophily, (2) cities’ geographic proximity, (3) colocation within larger jurisdiction, (4) city population and popularity, (5) technical engagement capacity and message salience, (6) a mayor’s tenure in office, and (7) the norm of reciprocity as an endogenous network process that influences tie formation.

Political Party Homophily

Both polarization and partisanship characterize political dialogue on Twitter (Boulianne 2016; Jungherr 2016). By and large, social media users seek out content that aligns with their viewpoints (Gainous and Wagner 2014); therefore, social media users create echo chambers in which few users are exposed to cross-ideological political views by the accounts they follow or the conversations in which they engage (Himelboim, McCreery, and Smith 2013). According to Ott (2017, p. 58), “Twitter privileges discourse that is simple, impulsive, and uncivil.” Members of Congress are apparently rewarded for this kind of partisan behavior. For example, members of the U.S. House of Representatives from the 111th Congress who publicized more extreme political rhetoric received more followers (Hong and Kim 2016) and more campaign contributions (Hong 2013). Jungherr (2016, p. 76) concluded “interactions between politicians seem to occur predominantly between candidates of the same party.”

While partisan divides may be less apparent at the local level than the national, policy differences continue to exist (Gerber and Hopkins 2011), and within the larger, hyper-partisan political climate, mayors may be more inclined to associate with members of the same party to avoid drawing the ire of their political bases of support. After all, ties on Twitter are public and, therefore, are open to scrutiny. Moreover, when an individual’s belief systems are shared with others, those similarities provide the basis for ties between like-minded potential allies (Sabatier 1988). The advocacy coalition framework literature, for example, advances the “belief homophily” hypothesis that “actors systematically seek out ideologically similar collaborators” (Henry, Lubell, and McCoy 2011, p. 426). Along those lines, a mayor’s political party affiliation provides an indication of values, ideology, and policy preferences (Gerber, Henry, and Lubell 2013). Like other types of professional networks (Carr, LeRoux, and Shrestha 2009), party affiliation may engender trust by simply conveying shared values and experiences.

Hypothesis 1 (H1): Mayors who share the same political party affiliation are more likely to form ties with each other rather than with members of another party.

Geographic Proximity

While some cities coordinate policy positions on climate change, for example, with their faraway counterparts (Barber 2017; Hughes 2017), mayors govern locally. Cities in close geographic proximity are more likely to cooperate in the provision of public goods and services (Feiock 2007) and within more informal information networks as well (Lee, Feiock, and Lee 2012). Moreover, they are more likely to have met in-person. Therefore, mayors may view Twitter as an extension of their traditional networking responsibilities and may seek to form ties with others who are in close geographic proximity.

Hypothesis 2 (H2): Mayors are more likely to form ties with mayors who are geographically closer rather than with those who are farther away.

Colocation within Larger Jurisdiction

Lee, Feiock, and Lee (2012) suggested that tie formation is due less to geographic proximity than to colocation within the same, larger political jurisdiction. Mayors within the same state, for example, share information and coordinate policy preferences to then communicate them to state policy makers (Lindstrom 1998; Thurmaier and Wood 2016). Lindstrom (1998) described a forum of mayors operating in the Chicago metropolitan region that works to build consensus and present regionally friendly proposals to the Illinois state legislature; thus, those state political institutions shape subsequent collective action. In addition to sharing similar policy problems, mayors from the same state (1) adhere to the same state laws and regulations, (2) operate within the same political context, and (3) work with the same state officials. Therefore, it is logical to assume that mayors operating in the same state will more likely form ties with each other rather than with those from different states.

Hypothesis 3 (H3): Mayors are more likely to form ties with other mayors who are colocated in the same state.

Population and Popularity

As opposed to their neighbors, mayors may want to associate with well-connected and high-profile colleagues from farther away. New York City, Los Angeles, and Chicago represent America’s largest cities, and their mayors receive a level of national prominence. The officials possess a certain amount of gravitas for other mayors to leverage through association, perhaps to develop relationships for future political support (Jungherr 2016) or convey competence to constituents at home (DePaula, Dincelli, and Harrison 2018; Mayhew 1974). Knoke’s (1994, p. 10) notion of prestige prominence is especially useful to describe the latter dynamic in which ties are asymmetrical (i.e., one mayor mentions another without a reply), and in this situation, “prominence increases with the extent to which a position receives many relations but does not reciprocate.” Therefore, mayors may value the symbolism of association rather than valuing the resources or information that might be accrued.

At the same time, the act of following another mayor’s content provides a tangible resource, that of information. Assuming the mayors are at the political forefront, following their content or engaging them more directly increases an individual’s access to valuable information. This strategy is intuitive. “In many collaborative situations, connecting to a key popular actor, rather than making efforts to create direct ties with many other actors, can be efficient” (Lee, Lee, and Feiock 2012, p. 556). Larger cities, for example, provide more in-depth and advanced website content (Bearfield and Bowman 2017; Feeney and Brown 2017). They possess more capacity to implement both information sharing policies (Gil-Garcia, Pardo, and De Tuya 2019) and a more interactive social media presence (Mossberger, Wu, and Crawford 2013). The more populous cities are more likely to have larger staffs that include digital communications teams who regularly develop relevant content. Furthermore, large cities may be more likely to provide innovative policy ideas or acknowledge emerging political trends on Twitter and therefore, they may provide the kind of information their colleagues find to be valuable (see Calanni et al. 2015).

Hypothesis 4 (H4): Mayors from larger cities are more likely to receive ties from others than mayors from smaller cities.

Technical Engagement Capacity and Message Salience

Follower counts, along with high tweet counts, provide initial measures for what Welch, Feeney, and Park (2016) referred to as technical engagement capacity or “the ability of the organization to digitally interface with external groups and organizations” (p. 394). High follower counts potentially indicate the presence of salient content. High tweet counts indicate the degree to which mayors regularly use Twitter and, thus, these mayors “are more likely to be highly visible” (Barberá 2017, p. 78).

Bakshy et al. (2011) reported that influential Twitter users (i.e., those who generate conversations) often have large numbers of followers; this may be indicative of high-quality information from the influential user or (at least) information from the user that might be of interest to many people. What garners interest on social media (i.e., what is salient) is subjective across audiences (Alonso, Marshall, and Najork 2013), and, in general, public officials often use their discretion to determine from whom to draw knowledge and advice (Quick and Feldman 2014). Regardless of specific content characteristics, those with high follower counts provide something of interest to their audiences, and, therefore, their tweets may be of interest to other mayors as well. Furthermore, those who tweet more often create more content. This may increase the likelihood their content will be seen by other mayors.

Hypothesis 5 (H5): Mayors with higher follower counts are more likely to receive ties than mayors with lower follower counts.

Hypothesis 6 (H6): Mayors who tweet more are more likely to receive ties than mayors with lower tweet counts.

Tenure in Office

In offline situations, mayors with longer tenure in office have had more time to (1) adjust to the demands of their jobs, (2) network formally within organizations such as the U.S. Conference of Mayors, and (3) reach out to others informally via social media. Siciliano (2015) reported that government personnel with more experience are more likely to seek advice from their peers. However, evidence suggests that younger and less experienced politicians use social media more frequently, for Jungherr (2016) reported, “Young politicians appear to be more likely to use Twitter than old” (p. 74), and Ahn and Berardino (2014) demonstrated that newly elected governors in the United States are more likely to use Twitter than incumbents. Therefore, more recently elected mayors may be more eager to establish contact with other mayors and their staff, and social media provides a low-cost communication channel to do so.

Hypothesis 7 (H7): Mayors with fewer years in office are more likely to initiate (or send) ties than mayors with more time in office.

Reciprocity

Reciprocity is the social norm of responding to an action in kind, and Axelrod (1984) suggested that reciprocity holds together cooperative partnerships. Participants feel obligated to sustain relationships because they fear that deviation or defection will be punished; evidence suggests that this applies to policy networks as well (Berardo and Scholz 2010; Feiock et al. 2010). However, Twitter’s asymmetric relationships and the context in which information is shared challenge the applicability of this concept.

Twitter permits one person to create a tie regardless of the other person’s level of interest; therefore, a relationship is only symmetrical when reciprocated (Kane et al. 2014). Moreover, prominent Twitter accounts seldom reciprocate ties. For example, they tend to have many followers yet follow few in return. While mayors and their staff do not have to expend excessive amounts of time monitoring social media to identify opportunities to reply to mentions or follow those who follow them, reciprocity is not likely an established norm among high-profile Twitter users.

Hypothesis 8 (H8): Mayors will not reciprocate ties from other mayors on Twitter.

Data and Methods

Case Selection

In this article, tie formation between U.S. mayors on Twitter is examined by considering the 100 largest cities by U.S. Census population estimates.² An underlying assumption was that the mayors possess more resources than smaller cities, which enable them to dedicate more personnel and attention to their strategic online presence (Feeney and Brown 2017; Mossberger, Wu, and Crawford 2013). While more people use Facebook, Twitter facilitates more open conversations between diverse actors, including politicians (Jungherr 2016) and public sector organizations (Hu 2019; Mergel 2012, 2016). Therefore, Twitter represents an appropriate choice as far as a social media application. An extended period of observation (January 2016 to August 2018) captured patterns of tie formation from a day-to-day perspective while also accounting for prominent focusing events including (1) multiple election cycles, (2) major policy decisions such as the Paris Agreement on climate change withdrawal and the forced separation of immigrant families at the Mexican border, and (3) several large-scale disasters (such as the 2017 hurricanes in Florida, Texas, and Puerto Rico).

The 100 cities encompassed 35 states and all four U.S. Census regions, which provided for geographic diversity. Forms of municipal governing bodies varied (47 mayor-council cities in which a mayor serves as the chief executive and a city council serves as the legislative body, 46 council-manager cities in which a mayor and council serve as the legislative body and appoint a city manager to administer day-to-day operations, six hybrid cities in which a mayor serves the chief executive, but also hires a city manager to oversee day-to-day operations, and one city commission [Portland, Oregon] in which the elected mayor and additional board of commissioners together serve as the city’s legislative and administrative body).³ Political party affiliation also varied, albeit to a lesser extent (63 Democrats, 28 Republicans, and 9 Independents).

Data Collection

The dependent variable was whether mayors use their Twitter accounts to form ties with other mayors. The unit of analysis was a directional dyad (i.e., the activity of one mayor’s account directed toward another via mentions, replies, follows, and retweets). To collect data, city websites were first reviewed to identify accounts, an approach intended to ensure the collection of only official accounts (Sutton et al. 2013). Next, Twitter searches yielded additional accounts, and steps were taken to ensure account authenticity (e.g., the identification of Twitter’s blue verified badges and/or links to city or campaign webpages). In all, 90 of the 100 mayors maintained at least one account.⁴ Twenty-two mayors had both an official account promoted on the city website and an additional account (i.e., personal and campaign). In those cases, both accounts were included to ensure the capture of all relevant ties.

The resulting usernames provided a seed list. NodeXL™ was employed to collect all tweets from Twitter’s application programming interface that captured ties between those accounts as demonstrated via mentions, replies, follows, and retweets. In all, during the period of observation, mayors formed 838 mention ties, 181 reply ties, 847 follower ties, and 239 retweet ties. NodeXL™ captured each tweet’s text, the date issued, who initiated the tie, and the account to which it was directed. Those interactions provided the basis for four separate information networks, including (1) a mention network, (2) a retweet network, (3) follower network, and (4) a reply network. The decision to analyze each network separately stemmed from the different nature of each interaction type.

Additional data (i.e., independent variables) offered insights on individual mayors, their cities’ institutions, and the communities they represent. A review of official mayoral websites and newspaper reports provided data on each mayor’s age, gender, and race.⁵ Twitter account metrics (i.e., the number of followers and number of tweets posted) were collected.⁶ In line with past research (Lee, Feiock, and Lee 2012), community attributes were obtained from U.S. Census data in the form of median household income (e.g., median household income) and race (e.g., percentage of White not Hispanic); these attributes were used as a base measure of community homogeneity.⁷ In addition, Ballotpedia.org provided useful data including the mayors’ political party affiliation, tenure in office, and the type of government (strong vs. weak mayor) for each city.⁸ Geographic proximity between each city was calculated by using an inverse-distance construct of proximity. This approach accounted for wider spatial distributions between cities in the Western United States and the Great Plains compared with the regions such as the Eastern United States.⁹ Table 1 summarizes the independent variables and related hypotheses.

Table 1.

Hypotheses and Independent Variables.

Hypothesis	Independent Variable	Expected Effect
Homophily effects
H1: Same political party	Same political party	+
H2: Geographic proximity	Inverse distance weighting	+
H3: Colocation in the same state	Same state	+
—	Median income dissimilarity	Control
—	Same gender	Control
—	Same race	Control
—	Race dissimilarity (community)	Control
—	Same form of government (mayor-council)	Control
Main effects
H4: Population	City population	+
H5: Twitter followers	Followers per 100K population	+
H6: Tweets	Tweets per year	+
H7: Tenure in office	Years in office	–
—	Age	Control

Note. A dyad reciprocity measure (i.e., the percentage of reciprocated ties in each network) was used to evaluate the hypothesis (H8) that mayors are unlikely to respond to a mention, reply, follow, or retweet in kind.

Data Analysis

This research had three basic objectives. The first objective was to identify the extent to which mayors form ties with other mayors on Twitter and was pursued by an initial set of network diagrams structured by geographic location. These maps, created with the software program ORA-GIS™, visually depict the spatial distribution of cities and their interactions. In addition, out-degree and in-degree network statistics identify particularly active (and inactive) mayors. For example, in the mention network, a mayor with an out-degree centrality score of five indicates that they mentioned five other mayors in their tweets. Conversely, an in-degree score of five indicates that a mayor was mentioned by five of their colleagues. Relatedly, measures of central tendency and dispersion (i.e., mean, median, and standard deviation) offer insight on the extent to which mayors (on the whole) were connected within the networks.

The second objective was to identify the types of message content and the frequency at which mayors posted them. Previous research informed the creation of message typology (see Table 2). Categories were not mutually exclusive; therefore, one tweet could contain elements of more than one. A coder manually assigned the attributes to each mention (838) and retweet (239).¹⁰ A second rater then analyzed 35% of those tweets, using the same coding scheme. The resulting lowest Cohen’s kappa score (0.908) for key variables indicated an extremely high agreement rate between the two coders.

Table 2.

Message Types and Definitions.

Message Type	Definitions
Actions	Describing mayor’s activities or schedule (DePaula, Dincelli, and Harrison 2018; Sobaci and Karkin 2013)
Candidate endorsement	Endorsing a political candidate (Evans, Brown, and Wimberly 2018)
Critique	Critiquing another political actor or their positions (Evans, Brown, and Wimberly 2018)
Favorable presentation	Describing activities with positive imagery or self-referential language; seeks attribution of likability, competency, or worthiness (DePaula, Dincelli, and Harrison 2018, p. 102)
Marketing local businesses	Mentioning or showcasing local businesses or their products (DePaula, Dincelli, and Harrison 2018)
Marketing services	Encouraging residents to take advantage of city services (DePaula, Dincelli, and Harrison 2018)
Marketing the city	Promoting aspects of the city favorably, perhaps to attract people to visit, relocate, or do business there (DePaula, Dincelli, and Harrison 2018; Kagarise and Zavattaro 2017)
Policy statements & political positioning	Stating a position on a political or policy matter (DePaula, Dincelli, and Harrison 2018; Evans, Brown and Wimberly 2018; Sobaci and Karkin 2013)
Public service announcements	Providing recommendations for safety, public health, and well-being (DePaula, Dincelli, and Harrison 2018, p. 102)
Social sharing	Informative content that may be of interest to others, but not directly about the mayor or city government (DePaula, Dincelli, and Harrison 2018, p. 102)
Symbolic acts	Expressing congratulations, gratitude, or condolences; celebrating holidays; or referencing other cultural symbols such as sporting events (DePaula, Dincelli, and Harrison 2018, p. 102)

The third objective was to identify the factors that influence whether mayors form ties with other mayors. The use of traditional statistical models was inappropriate because of the interdependent nature of these networks. The researcher could not assume that each tie was independent of one another, a detail that violated a key assumption of traditional statistical models (Snijders 2011). Instead of traditional models, quadratic assignment procedure (QAP) offered an alternative for network regression to ascertain estimates of statistical significance (Dekker, Krackhardt, and Snijders 2007; Krackhardt 1988). Lubell et al. (2012, p. 357) offer a description:

[QAP] tests whether or not two matrices are correlated, either with bivariate or multiple regression measures of association. QAP uses a bootstrapping approach to randomly “relabel” the networks and examine the distribution of network statistics from the resulting population of networks. If the observed correlation or measure of association is outside the 95% confidence interval obtained from the set of bootstrapped networks, the statistic is considered significantly different from zero.

In this article, the dependent variable was binary because either a mayor forms a tie with another mayor’s account via a mention, reply, follow, or retweet or does not; therefore, logistic regression QAP (LRQAP) was used.¹¹ In this model, the dependent variable took the form of a square matrix that reported which Twitter accounts formed ties with others and which did not. The data for the independent variables were also organized in square matrices. Wukich et al. (2017) applied this method with regard to Twitter to examine the factors that influenced whether disaster response organizations followed one another’s accounts.

Two types of independent variables were applied within this model: (1) homophily effects and (2) main effects. Homophily effects “capture differences in the probability of tie formation that result due to actor similarity” (Wukich et al. 2017, p. 11). For example, a related hypothesis is that mayors from the same political party are more likely to interact than with mayors from different parties. In the resulting square matrix, the i, j cell indicates whether two accounts share the same party—a binary variable.

The second type of independent variable is that of main effects. Main effects “measure the differences in the odds of [one account directing activity towards another account] based on the characteristics of the actors themselves” (Wukich et al. 2017, p. 11). Both receiver and sender effects are, therefore, relevant. For example, some accounts may receive more attention due to a specific attribute (e.g., a city with a large population or a mayor’s short tenure in office might make tweets from related accounts more popular). A receiver effect would estimate that influence. Relatedly, a specific attribute of the sender may influence tie formation. An example would be a mayor who has less tenure in office and might therefore be more active in the network. This represents an example of a sender effect based on tenure (i.e., years in office).¹²

Findings

Measuring Interaction

Follows between mayors were relatively commonplace (847 in all), mentions were notable (838), retweets were less common (239), and replies were infrequent (181).¹³ Figure 1 depicts the geographic relationships created by those ties.

Figure 1.

Twitter mention, reply, follower, and retweet networks among U.S. mayors.

Of the 90 mayors with Twitter accounts, 70 (77.8%) followed 84 other mayors’ content (93.3%). These were asymmetric ties, one mayor following another. Mayors from cities such as Los Angeles (39 followers), Chicago (33), and Charlotte (29) were particularly popular and received the most follows. Therefore, these were the highest in-degree centrality scores. In total, mayors followed, on average, 9.41 other mayors (median = 4, SD = 12.31). Mayors from Memphis (52 follows), Pittsburgh (48), and Henderson, Nevada (47) were the most active following other mayors; however, 20 mayors failed to follow a single account. Yet the fact that 84 mayors (93.3%) were followed by at least one other mayor indicates a baseline level of interest within the larger network.

In total, 65 mayors (72.2%) mentioned at least one other mayor over the 32-month period of observation. Reno, for example, mentioned 23 other mayors, and Los Angeles mentioned 14. Conversely, 68 mayors (75.7%) were mentioned by at least one other mayor. Commonly mentioned mayors included Los Angeles (mentioned by 25 other mayors) and Houston (16). However, not all mayors were as active or as popular in this network. On average, mayors mentioned 3.66 counterparts (see out-degree centrality) over the period of observation (out-degree median = 2, SD = 4.24). The in-degree mean and median scores were the same with a slightly higher standard deviation of 4.32, which indicated that more popular mayors received a slightly larger share of the total number of mentions. In all, 65 mayors mentioned another mayor a total of 838 times over 32 months. While mentioning another mayor is not an everyday practice for individual mayors, mayors in total averaged 26 mentions per month across the whole network.

Fewer mayors retweeted content from their colleagues. Fifty mayors (55.6%) retweeted 239 tweets from a total of 49 other mayors (54.4%). While this represents a small number of mayors and total messages, the retweets demonstrate that some mayors amplified the content of others to their followers. Oakland, Austin, and St. Petersburg retweeted the most mayors (8). Conversely, Houston (11), Los Angeles (10), and Washington, D.C. (8) were retweeted by the most mayors. Fewer mayors directly replied to the tweets of other mayors; a total of 35 mayors (38.9%) replied to 32 others (35.6%) for a total of only 181 tweets.

Message Content

Table 3 reports the frequency of mentions and retweets by specific message types. A plurality of mentions (44.1%) aligned with DePaula, Dincelli, and Harrison’s (2018, p. 102) definition of favorable presentation, which is “describing activities with positive imagery or self-referential language.” For example, Los Angeles Mayor Eric Garcetti tweeted the following:

Welcome to L.A., @MayorHancock! It’s a pleasure to host you and @DowntownDenver to talk about what our cities can learn from each other and our shared work to build vibrant downtowns that serve as the backbone of economic growth. https://t.co/AWhH77SGM1

Other mentions (40.8%) demonstrated examples of symbolic acts (expressing congratulations, gratitude, condolences) or referencing cultural symbols (holidays and sporting events) (DePaula, Dincelli, and Harrison 2018, p. 102). Following the October 2017 mass shooting on the Las Vegas Strip, for example, Dallas Mayor Mike Rawlings tweeted, “Thinking of my friend @mayoroflasvegas Carolyn Goodman and all those impacted by this horrific attack. Dallas weeps for Las Vegas.” Similar disaster-related messages included calls for action. During Hurricane Harvey in 2017, Washington, D.C., Mayor Muriel Bowser tweeted, “We can help our fellow Americans affected by #HurricaneHarvey: @RedCross. Our hearts are w/you Mayor @SylvesterTurner. https://t.co/FESSXRbxUZ.” The link directed readers to an American Red Cross tweet with information on how to donate.

Table 3.

Message Type Frequency.

	N	%
Mentions
Favorable presentation	370	44.1
Symbolic acts	342	40.8
Policy/political positioning	309	36.8
Actions	299	35.6
Social sharing	44	5.2
Marketing the city	42	5
Candidate endorsement	39	4.6
Marketing businesses	33	3.9
Critiques	25	3
Public service announcements	7	0.8
Marketing services	1	0.1
Retweets
Policy/political positioning	119	49.8
Actions	100	41.8
Favorable presentation	66	27.6
Symbolic acts	56	23.4
Critiques	13	5.4
Marketing events	11	4.6
Social sharing	10	4.2
Marketing city	9	3.8
Candidate endorsement	4	1.7
Marketing businesses	2	0.8
Public service announcements	0	—

Note. There were 838 mentions and 239 retweets. Categories were not mutually exclusive; therefore, one tweet could contain elements of more than one message type.

Some tweets were more cheerful, and many of those pertained to sporting events (17.6% overall). Mayor Sam Liccardo of San Jose, California, for example, made a bet with Anaheim Mayor Tom Tait in the lead-up to a National Hockey League playoff series between the two cities’ teams and tweeted:

Anaheim @MayorTomTait & I have placed a friendly wager as our @SanJoseSharks & the @AnaheimDucks face off: I’m betting @eggo waffles, he’s betting churros, & win or lose we’ll both donate to a non-profit working to rehouse homeless residents in the other’s city. #PlayoffModeOn

By promoting Eggo waffles, a San Jose brand, the mayor showcased that product for a wider audience. Along those lines, several mayors used the messages as opportunities to promote their cities and local businesses (see rows Marketing the City and Marketing Local Businesses in Table 3). In another example, mayors from Pittsburgh and Philadelphia focused on nonprofits and service work. Pittsburgh’s Mayor Bill Peduto tweeted: “Hey @PhillyMayor—it’s on. Pens-Flyers Round One. What do you wanna bet . . . Aid to Homeless Shelter? Assistance to Veterans Employment Center? Snow Plow? You call it @ACE_Fitzgerald & I will match it.” Philadelphia Mayor Jim Kenney responded in a separate tweet: “@billpeduto! How about a service bet? If the @NHLFlyers win you can come to Philly and volunteer at the Hub of Hope with @SEPTAPHILLY, @ProjectHOME and @PHLCityHomeless! https://t.co/2fCQat3W8F”

Many messages focused on policy statements and political positions. Of all mentions, 36.8% contained some type of policy statement or political position. Houston Mayor Sylvester Turner, for example, publicized a letter to the U.S. Departments of Justice and Homeland Security that he signed along with the mayors of Los Angeles, Tucson, and Albuquerque.

Today I joined @MayorofLA @JRothschildAZ @MayorKeller in asking White House cabinet members @JeffSession and @SecNielsen to stop the inhumane, un-American separation of children from their migrant families. https://t.co/EPaR1Idl7V https://t.co/DWxmtUWvgS

Within that tweet, Mayor Turner criticized federal immigration policy. While the critiques were relatively rare (3.0%), almost all of them pertained to federal actors and their policy preferences.

Other tweets addressed joint advocacy efforts within the same state. The San Jose mayor, for example, described a meeting in which several California mayors advocated policies to address homelessness. “Great to join @ericgarcetti @mayoredlee @LibbySchaaf @Steinberg4Sac & mayors working for solutions to homelessness. https://t.co/anHyZOkF0D.” The mayor of Long Beach mentioned the Los Angeles mayor and their effort to reduce air pollution in their region in this way: “Proud to join @MayorOfLA today to sign principles of new Clean Air Action Plan at @portoflongbeach & @PortofLA https://t.co/0zOSXoNBtS.” In all, 35.6% described some type of joint action.

Additional events such as the U.S. Conference of Mayors’ meetings in Boston, Indianapolis, Miami, and Reno provided opportunities for tie formation. During a 2018 meeting in Boston, Mayor Marty Walsh invited his counterparts to participate in the city’s pride parade. Kansas City Mayor Sly James posted a photo of the group and tweeted, “Thanks #MayorMartyWalsh for inviting us to participate in the #WickedProud pride parade today. #USCM18 #KansasCity.” Sacramento, California Mayor Darrell Steinberg tweeted a photo of several mayors sitting at a dinner table.

Great time at mayors’ get together! U.S. conference of Mayors chair and Mayor of Columbia, S.C. @SteveBenjaminSC, Stockton Mayor @MichaelDTubbs, Oakland Mayor @LibbySchaaf, Da Mayor Willie Brown, me, and San Jose Mayor @SamLiccardo. Mayors and cities are leading the way. https://t.co/9EaBlWzJ1u

In addition to those meetings, the U.S. Conference of Mayors organized policy advocacy efforts. Reno Mayor Hillary Schieve promoted an event when she tweeted, “Join the #ACA discussion! Thank you @BilldeBlasio @MayorGiles @MayorGinther @ Mayor’s Tele Town Hall #MayorsStandForAll CALL 202/224-3121.” In this tweet, the mayor recognized her counterparts while encouraging residents to learn more about the Affordable Care Act.

Some mentions (5.2%) highlighted the accomplishments of others and represented attempts at social sharing—or informative content that was not specifically about the mayor who was tweeting or their city, but instead their counterpart (see DePaula, Dincelli, and Harrison 2018, p. 102). Denver Mayor Michael Hancock tweeted, “Colorado Springs (way to go, @COSpringsMayor!) with the strong showing in this year’s @usnews Best Places to Live! #ColoradoProud https://t.co/K7DXlTMNGD.” Other mentions related to political campaigning. For example, 4.6% showcased a candidate endorsement. In one tweet, Jersey City, New Jersey Mayor Steven Fulop praised Newark Mayor Ras Baraka.

I’m proud to have supported @rasjbaraka 4 yrs ago when most of the political establishment in NJ was against us. He’s done a phenomenal job as mayor + has been a good partner for #JerseyCity + dealing w/challenges facing cities. looking forward to another term working together. https://t.co/RDloyAxdOk

This tweet included a link to a newspaper article about a fundraiser Mayor Fulop had hosted for Mayor Baraka earlier that week. Also common were congratulations following election victories. In January 2018, Charlotte Mayor Vi Lyles commented, “Congratulations to @KeishaBottoms who today became Atlanta’s 60th mayor. Atlanta is lucky to have you!”

Regarding retweets (see Table 3), mayors frequently amplified policy statements and political positioning (49.8%) as well as descriptions of their counterparts’ actions (41.8%). Baton Rouge Mayor Sharon Weston Broome retweeted Houston Mayor Sylvester Turner’s tweet and stated,

It does not matter if you’re a Democrat or a Republican, Mayors are on the front lines working to address the day to day issues our cities face daily. Proud to join my fellow Mayors as we kickoff the annual conference along with our @houstonpolice Chief @ArtAcevedo. #USCM2018.

While most messages conveyed positive statements like the one posted by Mayor Turner, some critiqued federal policies and/or offered political contrasts. Following the decision to withdraw from the Paris Agreement on climate change, for example, mayors from Austin and Los Angeles both retweeted Pittsburgh Mayor Bill Peduto’s tweet that stated, “As the Mayor of Pittsburgh, I can assure you that we will follow the guidelines of the Paris Agreement for our people, our economy & future.” The President had referenced the people of Pittsburgh and their legacy as a steel producing region as justification for withdrawing from the agreement, which required a reduction in industrial carbon emission.

Factors That Influence Tie Formation

Table 4 reports the LRQAP results first by homophily effects and then by main effects, ordered by hypotheses. Across the four networks, several common determinants were evident. For example, starting with homophily effects, shared political party affiliation helped to account for tie formation. Coefficients were positive across each network and statistically significant at the .05 level for mentions (and significant at the .10 level for followers and retweets). The findings offer some support for H1.

Table 4.

QAP Logistic Regression Results.

	Mention		Reply		Follower		Retweet
	Coefficient	Odds Ratio	Coefficient	Odds Ratio	Coefficient	Odds Ratio	Coefficient	Odds Ratio
Intercept	−24.2347***		−24.0125***		−17.6553***		−24.6882***
Homophily effects
Same party	0.4545*	1.5755	0.3529	1.4231	0.3351^†	1.3981	0.5775^†	1.7815
Geographic proximity	6.4844***	654.8196	3.3022^†	27.1718	8.1913***	3609.3049	4.5798*	97.4907
Same state	2.6636***	14.3481	3.0922***	22.0263	2.3623***	10.6157	2.9803***	19.6933
Median income dissimilarity	0.0066	1.0066	0.0021	1.0021	0.0019	1.0019	−0.0036	0.9964
Same mayor-council form	−0.1685	0.8449	0.2663	1.3052	−0.2049*	0.8147	−0.0808	0.9224
Same gender	−0.3191	0.7268	−0.0421	0.9588	0.0597	1.0616	−0.0072	0.9929
Same race (mayor)	0.026	1.0263	−0.114	0.8922	0.247	1.2802	0.0566	1.0583
Race dissimilarity (community)	0.0012	1.0012	0.0033	1.0033	0.007	1.007	0.0006	1.0006
Main effects
Receiver: Population	1.4772**	4.3808	0.6306	1.8788	1.4283***	4.1715	0.9112*	2.4872
Sender: Population	0.5544	1.7409	0.0707	1.0732	−0.7425	0.4759	0.1601	1.1736
Receiver: Followers	0.6636***	1.9419	0.6305*	1.8785	0.6128***	1.8455	0.4815*	1.6185
Sender: Followers	0.3006^†	1.3507	0.2015	1.2232	0.4622*	1.5876	0.5746**	1.7764
Receiver: Tweets	−0.0315	0.969	0.6563*	1.9276	0.0323	1.0328	0.4571*	1.5794
Sender: Tweets	0.3360*	1.3994	0.5215*	1.6845	0.4115*	1.509	0.089	1.093
Receiver: Tenure in office	−0.0564	0.9452	0.0739*	1.0767	−0.0075	0.9926	−0.0936	0.9106
Sender: Tenure in office	−0.0031	0.9969	0.1182	1.1255	−0.1406*	0.8688	−0.1078*	0.8978
Receiver: Age	0.0084	1.0085	−0.0014	0.9986	0.001	1.001	0.0212	1.0214
Sender: Age	−0.0189	0.9813	−0.0255	0.9748	0.0006	1.0006	0.0191	1.0193
Dyad reciprocity	25.10%		23.64%		22.58%		20.87%

Note. Geographic proximity between each city was calculated by using an inverse-distance construct of proximity. For the homophily and main effects of median income, tweets, and population, the values were log transformed. Followers were calculated as a standardized measure of the mayors’ Twitter followers per 100,000 population and log transformed. QAP = quadratic assignment procedure; OR = odds ratio.

†

p < .10. *p < .05. **p < .01. ***p < .001.

Geographic proximity was statistically significant in each network. This finding supports H2 that proposed mayors were more likely to form ties with counterparts located in closer proximity than those farther away. Relatedly, colocation within the same state positively influenced tie formation, which supports H3. However, it is important to note that these networks were not totally place bound. From time to time, mayors also formed ties with their cross-country counterparts. Each graphic in Figure 1 depicts a variety of long-distance ties (e.g., Seattle to Boston or Miami to Los Angeles) and refers to the distance spanning nature of the networks.

Table 4 reports several main effects as being statistically significant across the four networks. For example, mayors of cities with larger populations were more likely to receive a mention, follow, and retweet from other mayors. This finding supports H4 regarding population. Furthermore, mayors with more followers (per 100,000 population) were more likely to receive ties across networks, which lends credence to H5. Relatedly, mayors with more tweets per year were more likely to mention, reply to, and follow other mayors. They were also more likely to have their messages replied to and retweeted. The findings offer support for H6 about follower and tweet counts.

Regarding tenure, fewer years in office correlated with tie formation. Mayors with less tenure were more likely to follow and retweet the accounts of others. This finding supports H7, which puts forth that less experienced mayors would be more likely to initiate ties.

Finally, measures of dyad reciprocity were used to evaluate the extent to which mayors reciprocate ties (see bottom of Table 4). A certain percentage of ties were reciprocated, but not at a particularly high rate. Of all the pairs (or dyads) in which one mayor mentioned another, 25.1% were reciprocated (both mayors mentioned each other in their tweets). Fewer replies (23.6%), follows (22.6%), and retweets (20.9%) were reciprocated. However, in terms of online communication, mentions, replies, and retweets did not take place in a vacuum. In analyzing all messages in aggregate, reciprocity took place at a rate of 33.2%. This offers some evidence against H8 that mayors do not reciprocate ties, but it does not indicate these practices represent a pervasive norm.

Discussion and Conclusion

The purpose of this article was to examine the extent to which mayors form ties with other mayors on Twitter, the messages they create, and the determinants of tie formation. Related social media scholarship distinguishes between use by politicians (Golbeck, Grimes, and Rogers 2010; Russell 2018a) and service delivery networks (Mergel 2012, 2016; Wukich and Mergel 2016). Mayors occupy both spaces simultaneously—politics and administration (Svara 1998). This article contributes to this scholarship by exploring the extent to which mayors balance these positions and the extent to which they behave like politicians (e.g., members of Congress) versus administrators.

While mayors did not form ties with each other on a daily basis, the vast majority initiated one or more ties with other mayors over the period of observation. Dialogue was rare, supplanted by asymmetric ties. For example, while one-way mentions were commonplace, they were seldom replied to or reciprocated. Common message types included favorable presentation, symbolic acts, and political positioning. Regarding tie formation, mayors tended to choose other high-profile mayors from the same political party who represented larger cities and possessed more followers (i.e., more popular accounts). The results suggest that tie formation had less to do with the two-way technical exchange of best practices (something that might be expected in a more substantive policy network) and more to do with what Mayhew (1974) referred to as electorally oriented activities among elected officials that project likability, competence, and worthiness to constituents (DePaula, Dincelli, and Harrison 2018). Therefore, the findings showed that mayors prioritized political marketing over more functional, policy-based dialogue, as do members of Congress (Evans, Cordova, and Sipole 2014; Golbeck, Grimes, and Rogers 2010; Russell 2018b).

Each interactive Twitter function (e.g., mentions, replies, follows, and retweets) established a public association and perhaps facilitated information exchange, but among these functions, differences existed. For example, when one mayor mentioned another, a direct communication was initiated. Few mayors who were mentioned actually reciprocated those ties by mentioning the other mayor or replying to the tweet, and on social media, a user does not necessarily expect a reply from a prominent account. That is not a norm on Twitter, particularly between high-profile accounts. In this case, mayors appeared to associate political brands rather than engage in dialogue, and their constituents were likely their primary audiences based on the high frequency of messages such as favorable presentation, symbolic acts, and political positioning statements.

Of all Twitter functions in this case, mayors used replies the least. The reply capability provides a specific dialogic function that enables more interactive discussions (Brainard and Edlins 2014; Mergel 2016, 2017). The lack of replies further underlies the one-way, asymmetric nature of the networks; coupled with the low dyad reciprocity scores across each network, the few replies underscore the lack of reciprocity in this case (see H8). One limitation of this article should be noted, however. It fails to account for Twitter likes, or reactions used to signal appreciation for content. When mentioned or replied to, a mayor might like that content, which would be a way to acknowledge and reciprocate the tie. This article did not capture those measures.

Following another account was common and may be more akin to the act of listening than other functions, especially to the degree that mayors and their staff consume Twitter content. By following an account, the follower automatically receives posted content. The follow, itself, may be the result of a strategic communication decision, and there may be an additional networking reason for following another mayor, although in this case, the information flowed one-way, and most follows were not reciprocated. Therefore, mayors with high in-degree centrality represent those whose content was either most sought after or with whom other mayors most wanted to associate.

The last function was retweeting. When a mayor retweets, they forward that content to their followers. Mayors and their staff demonstrate the importance of a message by amplifying it publicly, and to the extent that the mayor who was retweeted monitors this activity, they may elicit symbolic meaning in terms of recognition or validation. Mayors with less tenure were more likely to retweet and follow other mayors (see H7), yet were not likely to mention or reply to content. Perhaps they were less familiar with other mayors and were reluctant to engage them directly through those functions. It could be that mayors consider mentions and replies to be more direct forms of engagement, and therefore perhaps more personal, than following and retweeting.

Results from the quadratic assignment procedure models across all Twitter functions identify specific tie formation determinants. Political party homophily shaped the networks, whereas mayors were more likely to mention other mayors from the same political party and were more likely to follow and retweet that content as well (see H1). This aligns with findings related to political polarization and partisanship on Twitter (Boulianne 2016; Gainous and Wagner 2014; Jungherr 2016). However, some cross-party ties were evident (e.g., Dallas-Fort Worth and Long Beach-Los Angeles). Compared with members of Congress, content appears to be more positive and less partisan (Evans, Cordova, and Sipole 2014; Gervais, Evans, and Russell 2020; Russell 2018b). When mayors posted negative rhetoric, which was rare, they directed their critiques toward federal officials from the opposing party rather than other mayors. Therefore, in total, the levels of negative rhetoric, polarization, and partisanship demonstrated by mayors were not as pronounced as that of members of Congress (see Russell 2018b).

A challenge for those who study social media networks is assessing the applicability of theories that were not developed to account for the unique functions and context related to social media communication (Kane et al. 2014). On one hand, existing theories of policy network formation (Berardo and Scholz 2010) and institutional collective action among elected officials (Feiock et al. 2010) seemingly inform our understanding of how mayors communicate, particularly when mayors seek to leverage other mayors’ positional power for substantive information or resources. However, capabilities novel to Twitter such as asymmetric relationships (e.g., publicly mentioning an account without receiving a reply) challenge the applicability of these theories because they were “developed primarily through studies of offline social networks” and based on situations in which information was reciprocally exchanged (Kane et al. 2014, p. 274). The evidence presented in this article suggests that mayors do not publicly seek out substantive information or other resources on Twitter, an assumption made in the policy network and institutional collective action literature. Rather, they appear to leverage what Knoke (1994) referred to as a mayor’s prestige prominence for their own self-promotion purposes. Instead of reciprocal interactions, mayors created asymmetric ties to publicly associate with other high-profile mayors.

Although this Twitter network differs from service delivery networks and appears to leverage prestige prominence, two well-known variables from the policy network and institutional collective action literature proved particularly germane to tie formation. Mayors were more likely to interact with other mayors from geographically closer cities (see H2) and cities colocated in the same state (H3). Those mayors were more likely to have interacted face-to-face at some point in time (see Lee, Feiock, and Lee 2012), and that interaction may have led to Twitter ties. To that point, tweets were noted that referred to state and regional policy initiatives that included proposed policies about homelessness and air pollution advocated in California. Related messages relayed joint statements to the public rather than enabling substantive, more spontaneous dialogue between officials. These joint statements require coordination between mayors and their staff. Although this article does not account for the role of existing offline relationships, future studies should consider accounting for existing relationships.

As reported, mayors tended to form ties with high-prestige counterparts. Specifically, mayors who represented larger cities (H4) and who had higher follower counts (H5) were more likely to receive attention. Furthermore, mayors who tweeted more frequently were also more likely to receive replies and retweets (H6). The population measure attests to the role of prestige prominence in this network. The follower and tweet count measures capture a baseline technical engagement capacity that relates to mayors’ abilities to engage external actors (see Welch, Feeney, and Park 2016). However, this operationalization is limited. Scholarship demonstrates that officials who enact social media confront several challenges of capacity (e.g., staff, resources, and training), and they possess varying levels of commitment to engage external actors (Bryer and Zavattaro 2011; Mergel 2012; Welch, Feeney, and Park 2016; Zhang and Feeney 2018). In addition, certain message strategies promote engagement between users based on mutual interests and salient content (Kagarise and Zavattaro 2017; Mergel 2016; Zavattaro and Brainard 2019). These are all factors that might be in play and deserve attention in future research.

This article has additional limitations; it does not include data on (1) smaller municipalities that are in closer geographic proximity to larger cities or (2) city and department-level accounts that concentrate more on service delivery and other priorities including the promotion of public safety or economic development (see Brainard and Edlins 2014; Kagarise and Zavattaro 2017; Zavattaro 2016). Because many accounts prioritize different goals, and personnel have varied demands placed on their time (Mergel 2016), this article’s findings are not necessarily generalizable. Future research should investigate patterns of communication between mayors who operate in closer geographic proximity perhaps delineated by county boundaries, metropolitan statistical areas, or individual states. Existing relationships as well as the more realistic specter of interlocal agreements may alter these interactions. As a result, more substantive policy networks may be created, and researchers may observe higher levels of reciprocity.

Service delivery networks may prove to be substantively different from these more political networks as well. Social media managers at the department level likely concentrate more time and effort on the technical operations of their service areas such as police, fire, emergency medical services, emergency management, and public works (see Brainard and Edlins 2014; Mergel 2016; Sutton et al. 2013). To what extent do departments from one city interact with departments from other cities? If they actually share information that informs cross-jurisdictional decisions on Twitter, the resulting networks would then more resemble the types of information and communications technology-enabled knowledge networks explored in the digital government literature in which participants share specific types of information to solve specific types of problems (see Comfort 2005; Dawes, Gharawi, and Burke 2012). Noveck (2009) and Roberts (2011) reported this to be the case when agencies use wikis, another form of social media, for interagency cooperation. Unlike Twitter, however, these wikis facilitate data storage, allow for secure communication, and do not have additional audiences with whom managers might seek to engage or from whom they might want to conceal certain information. Relatedly, Twitter limits character count (i.e., 280 characters max) and these limitations constrain what is shared between counts, whereas short messages supersede longer statements, memos, and other documents.¹⁴

Future research should also differentiate between how elected officials and appointed officials engage their counterparts, such as whether city managers put as much effort on favorable presentation as elected officials. Another topic would be to determine whether the authorities engage in more dialogic conversations about public problems, specific policy options, and available best practices. De Widt and Panagiotopoulos (2018) suggested this in the case of #localgov in the United Kingdom in which local government personnel discussed how to manage austerity measures.

This article also fails to consider outcomes or the effects of the observed ties concerning whether the information influences policy making or administrators’ decision making. Mayors and their staffs could very well be monitoring content, following counterparts, and accruing ideas that then influence subsequent policy making. But researchers would not necessarily observe the actions on Twitter unless officials publicly posted acknowledgments. Twitter enables these networks among practitioners (see De Widt and Panagiotopoulos 2018). However, findings from this case did not support those conclusions in the context of mayors; rather, they focused on offering favorable presentations and making joint statements rather than engaging in more substantive conversations. Scholars might continue this line of inquiry because it explores the extent to which they engage in meaningfully dialogue (see Brainard and Edlins 2014; Mergel 2017; Zavattaro and Brainard 2019).

Footnotes

Acknowledgements

This article benefited from several contributions. Gabrielle Parsson and Sam Motes identified Twitter accounts, and Kirby Suntala and Carlton Merriweather coded message attributes. Dr. Joanna Ganning provided geospatial data support, and Drs. Heather Evans, Tom Hilde, and Meg Rubado contributed valuable suggestions. The author wishes to thank all of them as well as the three anonymous reviewers and the journal’s editors for their astute guidance. The author also thanks Dr. Michael Siciliano for his insights about quadratic assignment procedure from previous projects. The correct application of the model is due to him; any mistakes belong to the author.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project was supported by Cleveland State University’s Undergraduate Research Award Program and the Maxine Goodman Levin College of Urban Affairs.

ORCID iD

Clayton Wukich

Notes

Author Biography

Clayton Wukich is associate professor and MPA director at the Maxine Goodman Levin College of Urban Affairs at Cleveland State University. He studies organizational theory, innovation, and behavior, and his research explores the role of information and communications technology in enabling various forms of participation and collaboration.

References

Ahn

Michael J.

Michael

Berardino

2014. “The Adoption of Web 2.0 by the State Government: The Role of Political Environment and Governors.” International Journal of Public Administration in the Digital Age 1 (1): 56–73.

Alonso

Omar

Marshall

Catherine C.

Najork

Marc

. 2013. “Are Some Tweets More Interesting Than Others? #HardQuestion.” Proceedings of the Symposium on Human-computer Interaction and Information Retrieval, Vancouver, BC.

Andrew

Simon A.

2009. “Regional Integration through Contracting Networks: An Empirical Analysis of Institutional Collection Action Framework.” Urban Affairs Review 44 (3): 378–402. doi:10.1177/1078087408323941.

Axelrod

Robert M.

1984. The Evolution of Cooperation. New York: Basic Books.

Bakshy

Eytan

Hofman

Jake M.

Mason

Winter A.

Watts

Duncan J.

2011. “Everyone’s an Influencer: Quantifying Influence on Twitter.” Proceedings of the Fourth ACM International Conference on Web Search and Data Mining, Hong Kong.

Barber

Benjamin R.

2017. Cool Cities: Urban Sovereignty and the Fix for Global Warming. New Haven: Yale Univ. Press.

Barberá

Pablo

. 2017. “Birds of the Same Feather Tweet Together: Bayesian Ideal Point Estimation Using Twitter Data.” Political Analysis 23 (1): 76–91. doi:10.1093/pan/mpu011.

Bearfield

Domonic A.

Bowman

Ann

. 2017. “Can You Find It on the Web? An Assessment of Municipal E-government Transparency.” The American Review of Public Administration 47 (2): 172–88. doi:10.1177/0275074015627694.

Berardo

Ramiro

Scholz

John T.

2010. “Self-organizing Policy Networks: Risk, Partner Selection, and Cooperation in Estuaries.” American Journal of Political Science 54 (3): 632–49. doi:10.1111/j.1540-5907.2010.00451.x.

10.

Borgatti

Stephen P.

Everett

Martin G.

Freeman

Linton C.

2002. UCINET 6 for Windows: Software for Social Network Analysis. Harvard: Analytic Technologies.

11.

Boulianne

Shelley

. 2016. “Campaigns and Conflict on Social Media: A Literature Snapshot.” Online Information Review 40 (5): 566–79. doi:10.1108/OIR-03-2016-0086.

12.

Brainard

Lori

Edlins

Mariglynn

2014. “Top 10 U.S. Municipal Police Departments and Their Social Media Usage.” The American Review of Public Administration 45 (6): 728–45. doi:10.1177/0275074014524478.

13.

Bryer

Thomas A.

Zavattaro

Staci M.

2011. “Social Media and Public Administration.” Administrative Theory & Praxis 33 (3): 325–40. doi:10.2753/ATP1084-1806330301.

14.

Calanni

John C.

Siddiki

Saba N.

Weible

Christopher M.

Leach

William D.

2015. “Explaining Coordination in Collaborative Partnerships and Clarifying the Scope of the Belief Homophily Hypothesis.” Journal of Public Administration Research and Theory 25 (3): 901–27. doi:10.1093/jopart/mut080.

15.

Carr

Jered B.

Hawkins

Christopher V.

Westberg

Drew E.

2017. “An Exploration of Collaboration Risk in Joint Ventures: Perceptions of Risk by Local Economic Development Officials.” Economic Development Quarterly 31 (3): 210–27. doi:10.1177/0891242417710325.

16.

Carr

Jered B.

LeRoux

Kelly

Shrestha

Manoj

. 2009. “Institutional Ties, Transaction Costs, and External Service Production.” Urban Affairs Review 44:403–27. doi:10.1177/1078087408323939.

17.

Comfort

Louise K.

2005. “Risk, Security, and Disaster Management.” Annual Review of Political Science 8:335–56. doi:10.1146/annurev.polisci.8.081404.075608.

18.

Criado

J. Ignacio

Martínez -Fuentes

Guadalupe

Silván

Aitor

. 2012. “Social Media for Political Campaigning. The Use of Twitter by Spanish Mayors in 2011 Local Elections.” In Web 2.0 Technologies and Democratic Governance: Political, Policy and Management Implications, edited by Reddick

Christopher G.

Aikins

Stephen K.

, 219–32. New York: Springer.

19.

Dawes

Sharon S.

Gharawi

Mohammed A.

Burke

G. Brian

. 2012. “Transnational Public Sector Knowledge Networks: Knowledge and Information Sharing in a Multi-dimensional context.” Government Information Quarterly 29 (Suppl. 1): S112–S120. doi:10.1016/j.giq.2011.08.002.

20.

Dekker

David

Krackhardt

David

Snijders

Tom A. B.

2007. “Sensitivity of MRQAP Tests to Collinearity and Autocorrelation Conditions.” Psychometrika 72 (4): 563–81. doi:10.1007/s11336-007-9016-1.

21.

DePaula

Nic

Dincelli

Ersin

Harrison

Teresa M.

2018. “Toward a Typology of Government Social Media Communication: Democratic Goals, Symbolic Acts and Self-presentation.” Government Information Quarterly 35 (1): 98–108. doi:10.1016/j.giq.2017.10.003.

22.

De Widt

Dennis

Panagiotopoulos

Panos

. 2018. “Informal Networking in the Public Sector: Mapping Local Government Debates in a Period of Austerity.” Government Information Quarterly 35 (3): 375–88. doi:10.1016/j.giq.2018.05.004.

23.

Einstein

Katherine Levine

Glick

David M.

Palmer

Maxwell

Pressel

Robert J.

2020. “Do Mayors Run for Higher Office? New Evidence on Progressive Ambition.” American Politics Research 48 (1): 197–221. doi:10.1177/1532673X17752322.

24.

Emanuel

Rahm

. 2020. The Nation City: Why Mayors Are Now Running the World. New York: Alfred A. Knopf.

25.

Evans

Heather K.

Brown

Kayla J.

Wimberly

Tiffany

. 2018. “‘Delete Your Account’: The 2016 Presidential Race on Twitter.” Social Science Computer Review 36 (4): 500–508. doi:10.1177/0894439317728722.

26.

Evans

Heather K.

Cordova

Victoria

Sipole

Savannah

. 2014. “Twitter Style: An Analysis of How House Candidates Used Twitter in Their 2012 Campaigns.” PS: Political Science & Politics 47 (2): 454–62. doi:10.1017/S1049096514000389.

27.

Feeney

Mary K.

Brown

Adrian

. 2017. “Are Small Cities Online? Content, Ranking, and Variation of U.S. Municipal Websites.” Government Information Quarterly 34 (1): 62–74. doi:10.1016/j.giq.2016.10.005.

28.

Feiock

Richard C.

2007. “Rational Choice and Regional Governance.” Journal of Urban Affairs 29 (1): 47–63. doi:10.1111/j.1467-9906.2007.00322.x.

29.

Feiock

Richard C.

Lee

In-Won

Park

Hyung Jun

Lee

Keon-Hyung

. 2010. “Collaboration Networks among Local Elected Officials: Information, Commitment, and Risk Aversion.” Urban Affairs Review 46 (2): 241–62. doi:10.1177/1078087409360509.

30.

Fountain

Jane E.

2001. Building the Virtual State: Information Technology and Institutional Change. Washington, DC: Brookings Institution Press.

31.

Gainous

Jason

Wagner

Kevin M.

2014. Tweeting to Power: The Social Media Revolution in American Politics. New York: Oxford Univ. Press.

32.

Gerber

Elisabeth R.

Henry

Adam Douglas

Lubell

Mark

. 2013. “Political Homophily and Collaboration in Regional Planning Networks.” American Journal of Political Science 57 (3): 598–610. doi:10.1111/ajps.12011

33.

Gerber

Elisabeth R.

Hopkins

Daniel J.

2011. “When Mayors Matter: Estimating the Impact of Mayoral Partisanship on City Policy.” American Journal of Political Science 55 (2): 326–39. doi:10.1111/j.1540-5907.2010.00499.x

34.

Gervais

Bryan T.

Evans

Heather K.

Russell

Annelise

. 2020. “Fear and Loathing on Twitter: Exploring Negative Rhetoric in Tweets during the 2018 Midterm Election.” In The Roads to Congress 2018: American Elections in the Trump Era, edited by Foreman

Sean D.

Godwin

Marcia L.

Wilson

Walter Clark

, 31–52. Cham: Springer.

35.

Gil-Garcia

J. Ramon

Pardo

Theresa A.

De Tuya

Manuel

. 2019. “Information Sharing as a Dimension of Smartness: Understanding Benefits and Challenges in Two Megacities.” Urban Affairs Review. Published electronically May 3. doi:10.1177/1078087419843190.

36.

Golbeck

Jennifer

Grimes

Justin M.

Rogers

Anthony

. 2010. “Twitter Use by the U.S. Congress.” Journal of the American Society for Information Science and Technology 61 (8): 1612–21. doi:10.1002/asi.21344.

37.

Henry

Adam Douglas

Lubell

Mark

McCoy

Michael

. 2011. “Belief Systems and Social Capital as Drivers of Policy Network Structure: The Case of California Regional Planning.” Journal of Public Administration Research and Theory 21 (3): 419–44. doi:10.1093/jopart/muq042

38.

Himelboim

Itai

McCreery

Stephen

Smith

Marc

. 2013. “Birds of a Feather Tweet Together: Integrating Network and Content Analyses to Examine Cross-ideology Exposure on Twitter.” Journal of Computer-Mediated Communication 18 (2): 154–74. doi:10.1111/jcc4.12001.

39.

Hong

Sounman

. 2013. “Who Benefits from Twitter? Social Media and Political Competition in the U.S. House of Representatives.” Government Information Quarterly 30 (4): 464–72. doi:10.1016/j.giq.2013.05.009

40.

Hong

Sounman

Kim

Sun Hyoung

. 2016. “Political Polarization on Twitter: Implications for the Use of Social Media in Digital Governments.” Government Information Quarterly 33 (4): 777–82. doi:10.1016/j.giq.2016.04.007.

41.

Qian

. 2019. “Twitter Data in Public Administration: A Review of Recent Scholarship.” International Journal of Organization Theory & Behavior 22 (2): 209–22. doi:10.1108/IJOTB-07-2018-0085.

42.

Hughes

Sara

. 2017. “The Politics of Urban Climate Change Policy: Toward a Research Agenda.” Urban Affairs Review 53 (2): 362–80. doi:10.1177/1078087416649756.

43.

Jungherr

Andreas

. 2016. “Twitter Use in Election Campaigns: A Systematic Literature Review.” Journal of Information Technology & Politics 13 (1): 72–91. doi:10.1080/19331681.2015.1132401

44.

Kagarise

Warren

Zavattaro

Staci M.

2017. “Social Media: How One City Opens the Evidence Black Box.” Public Administration Review 77 (4): 486–88. doi:10.1111/puar.12696.

45.

Kane

Gerald

Alavi

Maryam

Labianca

Giuseppe

Borgatti

Steve

. 2014. “What’s Different about Social Media Networks? A Framework and Research Agenda.” MIS Quarterly 38 (1): 274–304. doi:10.25300/MISQ/2014/38.1.13

46.

Knoke

David

. 1994. Political Networks: The Structural Perspective. New York: Cambridge Univ. Press.

47.

Krackhardt

David

. 1988. “Predicting with Networks: Nonparametric Multiple Regression Analysis of Dyadic Data.” Social Networks 10 (4): 359–81. doi.org/10.1016/0378-8733(88)90004-4

48.

Lee

In-Won

Feiock

Richard C.

Lee

Youngmi

. 2012. “Competitors and Cooperators: A Micro-level Analysis of Regional Economic Development Collaboration Networks.” Public Administration Review 72 (2): 253–62. doi:10.1111/j.1540-6210.2011.02501.x.

49.

Lee

Youngmi

Lee

In-Won

Feiock

Richard C.

2012. “Interorganizational Collaboration Networks in Economic Development Policy: An Exponential Random Graph Model Analysis.” Policy Studies Journal 40 (3): 547–73. doi:10.1111/j.1541-0072.2012.00464.x.

50.

Lindstrom

Bonnie

. 1998. “Regional Cooperation and Sustainable Growth: Nine Councils of Government in Northeastern Illinois.” Journal of Urban Affairs 20 (3): 327–42. doi:10.1111/j.1467-9906.1998.tb00425.x.

51.

Lubell

Mark

Scholz

John

Berardo

Ramiro

Robins

Garry

. 2012. “Testing Policy Theory with Statistical Models of Networks.” Policy Studies Journal 40 (3): 351–74. doi:10.1111/j.1541-0072.2012.00457.x.

52.

Mayhew

David

. 1974. Congress: The Electoral Connection. New Haven: Yale Univ. Press.

53.

Mergel

Ines

. 2012. Social Media in the Public Sector: A Guide to Participation, Collaboration, and Transparency in the Networked World. San Francisco: Jossey-Bass.

54.

Mergel

Ines

. 2016. “Social Media Institutionalization in the U.S. Federal Government.” Government Information Quarterly. 33 (1): 142–48. doi:10.1016/j.giq.2015.09.002.

55.

Mergel

Ines

. 2017. “Social Media Communication Modes in Government.” In Routledge Handbook on Information Technology in Government, edited by Chen

Yu-Che

Ahn

Michael J

, 168–79. New York: Routledge.

56.

Mossberger

Karen

Yonghong

Crawford

Jared

. 2013. “Connecting Citizens and Local Governments? Social Media and Interactivity in Major U.S. Cities.” Government Information Quarterly 30 (4): 351–58. doi:10.1016/j.giq.2013.05.016.

57.

Noveck

Beth Simone

. 2009. Wiki Government: How Technology Can Make Government Better, Democracy Stronger, and Citizens More Powerful. Washington, DC: Brookings Institution Press.

58.

Nutter

Michael A.

2018. Mayor: The Best Job in Politics. Philadelphia: Univ. of Pennsylvania Press.

59.

Ott

Brian L.

2017. “The Age of Twitter: Donald J. Trump and the Politics of Debasement.” Critical Studies in Media Communication 34 (1): 59–68. doi:10.1080/15295036.2016.1266686.

60.

Quick

Kathryn S.

Feldman

Martha S.

2014. “Boundaries as Junctures: Collaborative Boundary Work for Building Efficient Resilience.” Journal of Public Administration Research and Theory 24 (3): 673–95. doi:10.1093/jopart/mut085.

61.

Rethemeyer

R. Karl

. 2007. “Policymaking in the Age of Internet: Is the Internet Tending to Make Policy Networks More or Less Inclusive?” Journal of Public Administration Research and Theory 17 (2): 259–84. doi:10.1093/jopart/mul001.

62.

Roberts

Nancy C.

2011. “Beyond Smokestacks and Silos: Open-source, Web-enabled Coordination in Organizations and Networks.” Public Administration Review 71 (5): 677–93. doi:10.1111/j.1540-6210.2011.02406.x.

63.

Russell

Annelise

. 2018a. “The Politics of Prioritization: Senators’ Attention in 140 Characters.” The Forum 16 (2): 331–56. doi:10.1515/for-2018-0020.

64.

Russell

Annelise

. 2018b. “U.S. Senators on Twitter: Asymmetric Party Rhetoric in 140 Characters.” American Politics Research 46 (4): 695–723. doi:10.1177/1532673X17715619.

65.

Sabatier

Paul A.

1988. “An Advocacy Coalition Framework of Policy Change and the Role of Policy-oriented Learning Therein.” Policy Sciences 21:129–68. doi:10.1093/oxfordhb/9780199646135.013.24

66.

Siciliano

Michael D.

2015. “Advice Networks in Public Organizations: The Role of Structure, Internal Competition, and Individual Attributes.” Public Administration Review 75 (4): 548–59. doi:10.1111/puar.12362.

67.

Snijders

Tom A. B.

2011. “Statistical Models for Social Networks.” Annual Review of Sociology 37:131–53. doi:10.1146/annurevsoc.012809.102709.

68.

Sobaci

Mehmet Zahid

Karkin

Naci

. 2013. “The Use of Twitter by Mayors in Turkey: Tweets for Better Public Services?” Government Information Quarterly 30 (4): 417–25. doi:10.1016/j.giq.2013.05.014.

69.

Stern

Eric K.

2017. “Unpacking and Exploring the Relationship between Crisis Management and Social Media in the Era of ‘Smart Devices.’” Homeland Security Affairs 13 (4): 1–17.

70.

Sutton

Jeannette

Spiro

Emma

Butts

Carter

Fitzhugh

Sean

Johnson

Britta

Greczek

Matt

. 2013. “Tweeting the Spill: Online Informal Communications, Social Networks, and Conversational Microstructures during the Deepwater Horizon Oilspill.” International Journal of Information Systems for Crisis Response and Management 5 (1): 58–76. doi:10.4018/jiscrm.2013010104.

71.

Svara

James H.

1998. “The Politics-administration Dichotomy Model as Aberration.” Public Administration Review. 58 (1): 51–58. doi:10.2307/976889.

72.

Thurmaier

Kurt

Wood

Curtis

. 2016. “Interlocal Agreements as an Alternative to Consolidation.” In City-County Consolidation and Its Alternatives: Reshaping the Local Government Landscape, edited by Carr

Jered B.

Feiock

Richard C.

, 113–30. New York: Routledge.

73.

Twitter. 2018. “Glossary.” Twitter. Accessed November 14, 2018. https://help.twitter.com/en/glossary.

74.

Welch

Eric W.

Feeney

Mary K.

Park

Chul Hyun

. 2016. “Determinants of Data Sharing in U.S. City Governments.” Government Information Quarterly 33 (3): 393–403. doi:10.1016/j.giq.2016.07.002.

75.

Wukich

Clayton

Mergel

Ines

. 2016. “Reusing Social Media Information in Government.” Government Information Quarterly 33 (2): 305–12. doi:10.1016/j.giq.2016.01.011.

76.

Wukich

Clayton

Siciliano

Michael D.

Enia

Jason

Boylan

Brandon

. 2017. “The Formation of Transnational Knowledge Networks on Social Media.” International Public Management Journal 20 (3): 381–408. doi:10.1080/10967494.2016.1238428.

77.

Zavattaro

Staci M.

2016. “Some Ideas for Branding via Social Media.” In Social Media for Government: Theory and Practice, edited by Zavattaro

Staci M.

Bryer

Thomas

, 60–73. New York: Routledge.

78.

Zavattaro

Staci M.

Brainard

Lori A.

2019. “Social Media as Micro-encounters: Millennial Preferences as Moderators of Digital Public Value Creation.” International Journal of Public Sector Management 32 (5): 562–80.

79.

Zhang

Fengxiu

Feeney

Mary K.

2018. “Managerial Ambivalence and Electronic Civic Engagement: The Role of Public Manager Beliefs and Perceived Needs.” Public Administration Review 78 (1): 58–70. doi:10.1111/puar.12853.