Abstract
This study employs content and network analysis techniques to explore the predictors of opinion leadership in a political activism network on Twitter. The results demonstrate the feasibility of using user-generated content to measure user characteristics. The characteristics were analyzed to predict users’ performance in the network. According to the results, Twitter users with higher connectivity and issue involvement are better at influencing information flow on Twitter. User connectivity was measured by betweenness centrality, and issue involvement was measured by a user’s geographic proximity to a given event and the contribution of engaging tweets. In addition, the results show that tweets by organizations had greater influence than those by individual users.
In the summer of 2012, citizens of Wisconsin followed recall efforts to remove Wisconsin Governor Scott Walker. The recall was triggered by Walker’s proposed budget repair bill which restricted unions’ collective bargaining rights. In the days leading up to the recall election, Wisconsin constituents used Twitter to diffuse political information through the #wirecall hashtag. Recent Twitter studies have illustrated the usefulness of Twitter. It is evident from the literature that Twitter helps to disseminate information and mobilize the public (Larsson & Moe, 2012; Otterbacher, Shapiro, & Hemphill, 2013). Nevertheless, the extent to which individual users influence information diffusion on Twitter is less explored.
We explore the predictors of opinion leadership in the #wirecall Twitter activism network. Based on the activities of #wirecall Twitter users, we examine the characteristics of opinion leaders through a network analysis in conjunction with a content analysis. We refer to the diffusion literature (Vishwanath & Barnett, 2011) to identify the key antecedents of opinion leadership and measure them based on user-generated content. More specifically, we test the relationships of Twitter users’ connectivity, involvement, and identity to their ability to influence information flow on Twitter.
The rest of this article is organized as follows: We first discuss the current political use of social media. We then review previous research on opinion leadership. We later present arguments about online opinion leadership in a political context. The foundation for the hypotheses and research questions are provided based on the literature review.
Social Media and Politics
Despite Putnam (1995) and Morozov’s (2013) cautionary tales about technological effects on politics, social media continue to evolve as platforms for political engagement (Larsson & Moe, 2012). According to a 2012 Pew Research Center survey (Pew Research Center, 2012), 67% of all American adults used social networking sites (SNSs) such as Facebook and Twitter, among which 66% used SNSs for political purposes. Aside from citizens’ use, news media, government agencies, and advocacy groups have leveraged social media to influence public discourse (Bennett & Segerberg, 2012; Nielsen, 2012).
Governments and other institutions increasingly focus on mining social media content to gauge public opinion (Kavanaugh et al., 2012). For example, Twitter discussions have been used to predict election outcomes (Tumasjan, Sprenger, Sandner, & Welpe, 2010) and identify the stages in which the public can be targeted for persuasive communication (Russell, Flora, Strohmaier, Poschko, & Rubens, 2011). These approaches identify popular, and thus influential, ideas in online discourse. We argue that it is equally important to identify influential users, who not only possess or disseminate a large amount of information but also influence others’ viewpoints and behaviors.
Traditional Opinion Leadership in New Media Environments
Opinion leadership is an individual’s unequal influence on others’ attitude and behaviors (Rogers, 2003). The concept originates from two-step flow theory (Lazarsfeld, Berelson, & Gaudet, 1948). This theory posits that the influence of mass media first reaches opinion leaders who pass on to others what they read and hear about (Katz, 1957). Therefore, opinion leaders have better access to information and can turn the information into asymmetric influence on others (Rogers, 2003). Diffusion studies have identified opinion leaders to be innovative individuals who are highly involved and who have relative high social status and vast social connections (Rogers, 2003; Vishwanath & Barnett, 2011).
Traditionally, opinion leaders have both greater access to information and more platforms for disseminating information. However, digital technologies have changed the dynamics. Ordinary Internet users can produce and broadcast to mass audiences. In some instances, their content becomes a desirable alternative to mass media content and even influences mass media agendas (Himelboim, 2008; Himelboim, Gleave, & Smith, 2009). In addition, digital technologies level the field for people who have limited influence in the offline world. In the traditional framework of opinion leadership, social, cultural, and demographic characteristics of individuals drive their influence. In online communication, which tends to limit participants’ identity and social cues, the question of whether offline authority continues to influence online audiences becomes a debatable issue (Suler, 2004).
Online opinion leadership rests on the ability to influence information flow (Cha, Haddadi, Benevenuto, & Gummadi, 2010; Sun, Youn, Wu, & Kuntaraporn, 2006). This ability has two dimensions: the ability to contribute information (Weimann, Tustin, Van Vuuren, & Joubert, 2007) and to lead others in disseminating information (Cha et al., 2010; Phelps, Lewis, Mobilio, Perry, & Raman, 2004). The low cost of digital communication helps people of varying socioeconomic status to produce massive amounts of information. Nevertheless, competition for attention remains fierce. In the era of abundant or even redundant online content, attention represents the most valuable asset. Therefore, users who are successful not only in getting attention but also in persuading others to maximize attention (e.g., leading others to disseminate information) are the most influential ones.
The traditional opinion leadership framework has been applied to online word-of-mouth (WOM) intentions in marketing contexts (see Sun et al., 2006). Previous studies have shown that topic involvement and social connectivity predict online WOM intentions (Sun et al., 2006). In addition, identity matters in online WOM communication. Himelboim et al. (2009) found that the mass media has considerable influence in online political discourse. These findings suggest that that the two-step process of information flow and the traditional opinion leadership framework still hold significance in online communication. Therefore, we in this study extend the traditional opinion leadership framework to political activism on social media.
Political activism on social media provides an interesting context for understanding opinion leadership because social media are used by a range of users whose influence may vary across offline audiences. For example, government agencies, nonprofit organizations, mass media outlets, politicians, and celebrities—along with citizens—engage in information diffusion through social media. Some are offline opinion leaders because of their offline popularity and leadership positions (Valente & Pumpuang, 2007). This offline influence may be translated into online dominance when social media users reveal information reflecting their offline status—such as by saying who they are or with whom they are affiliated? Therefore, despite the equalizing aspects of online communication, users’ offline identity may still provide an advantage in influencing online discourse.
Structural Dimensions of Opinion Leadership in New Media Environments
Early opinion leadership studies used subjective information from self-report surveys or interviews (Katz & Lazarsfeld, 1955). To address the limitations caused by subjectivity, scholars have used the network analysis technique to explore opinion leaders within the structure of social relationships. Coleman, Menzel, and Katz (1959) related users’ positions in social networks to the extent of their influence. The structural aspect of opinion leadership posits that opinion leaders are central players in the social network and have wide social connections (Valente & Pumpuang, 2007). The nature of connections varies according to social interactions, the exchange of resources, and information flow (Borgatti, Mehra, Brass, & Labianca, 2009).
Twitter Network
Twitter is a social media platform used for disseminating information through social networks. Twitter users post 140-character messages called tweets, which can be forwarded by followers to reach broader audiences. A forwarded tweet is referred to as a retweet (RT). In contrast to the traditional sender–message–receiver communication mode, Twitter communication includes information providers who send original messages and information transmitters who forward those messages to wider audiences (Kim & Park, 2012). Both information providers and transmitters influence the information flow on Twitter.
Retweeting is consistent with the notion of opinion leadership. When a user is retweeted, the user’s message is likely to have caught the attention of others (Gruzd, Wellman, & Takhteyev, 2011). Users retweet when they agree or disagree with the viewpoints expressed in the original tweets or when they acknowledge the informational value of the tweets (Bruns & Burgess, 2011; Cha et al., 2010). Either way, an RT reflects influence. That is, the original sender influences others by drawing their attention and stimulating responses to the message (Bruns & Burgess, 2011; Gruzd et al., 2011).
Opinion Leadership in Political Communication on Twitter
Although Twitter levels the playing field in political discourse, elite groups of journalists and institutions continue to dominate online discourse. Journalists are often the hub in a Twitter discussion network (Bruns, 2011), indicating that the influence of traditional media extends to social media. However, citizen users such as bloggers and activists can emerge as opinion leaders when they share valuable information. More specifically, citizen users can dominate during the initial stages of discussions (Papacharissi & de Fatima Oliveira, 2012). Traditional media and elite groups are likely to participate only in later stages, competing with citizen users for public attention.
Predicting Opinion Leaders in the Twitterverse
As discussed earlier, traditional opinion leadership is associated with high involvement and wide social connectivity. Conventionally, involvement and social connectivity are measured through self-report surveys. As detailed later, alternatively, these constructs can be measured by user-generated social media content.
Social Connectivity
Connectivity is indicated by users’ positions in online social networks (Park & Thelwall, 2008). Social networks are based on social interactions and the exchange of resources, and as in the case of this study, they can be based on information flow (Borgatti et al., 2009). On Twitter, relationships take the form of followings or followers. Users can follow others to receive information provided by those others (followings). By following, users can be exposed to more information. Comparatively speaking, by having more followers, users have a broader audience base (Bruns & Burgess, 2011; Gruzd et al., 2011; Takhteyev, Gruzd, & Wellman, 2012).
In social network analysis (SNA), a central network position indicates high centrality. A central and strategic network position implies control of information flow and thus others’ attitudes and behaviors (Burt, 1999). Given the relationship between connectivity and traditional opinion leadership, the following hypothesis is proposed:
Hypothesis 1: Users’ centrality in Twitter networks is related to influence on the diffusion of political information such that the higher the centrality, the more likely users’ messages are retweeted by other users.
User Involvement
Marketing studies define involvement as interest, enthusiasm, excitement, and personal relevancy with respect to a given product (Zaichkowsky, 1985). Likewise, this study refers to involvement as one’s interest and personal relevancy with respect to a given political issue. Enduring involvement and situational involvement are two forms of involvement—the former reflects a long-term attachment, whereas the latter, transitory situational feelings or states, related to specific behaviors and contexts (Bergadaa, Faure, & Perrien, 1995). In the context of Twitter-based political activism, enduring involvement is one’s general interest in politics, whereas situational involvement is one’s interest with regard to a specific political issue (in the case of this study, the Wisconsin recall election). Higher involvement is associated with knowledge and expertise (Lazarsfeld et al., 1948; Sun et al., 2006; Tsang & Zhou, 2005), and this association tends to hold in offline settings as well as in online WOM communication (Sun et al., 2006; Tsang & Zhou, 2005).
Users’ social media behaviors indicate varying levels of involvement. For example, users can reveal or not reveal political information on themselves through social media profiles. Such information can range from their political affiliation to their views on specific issues. Therefore, users’ political involvement can be assessed by their disclosure or nondisclosure of political information on profiles.
In addition to political involvement, issue involvement can be inferred from users’ geographic proximity to a given political event. Researchers have found that social media users’ central role in news diffusion networks is related to their geographic proximity to news events (Yardi & Boyd, 2010). Because events occur in and affect specific communities, being at the news scene means personal relevance and local knowledge of the event (Gruzd et al., 2011; Yardi & Boyd, 2010). Therefore, based on the assumption that local users are more likely to be involved in local events than nonlocal ones, we can rely on users’ location information to infer users’ issue involvement.
Tweets may also indicate the variation in issue involvement. Lovejoy and Saxton (2012) introduced a typology for classifying tweets into three categories: information, communities, and actions. This typology represents a hierarchy of involvement—when users send tweets in the information category, they simply pass on information, and the tweets may or may not spark a discussion. Real engagement occurs when users send tweets in the community category. Users not only pass on information but also provide feedback, thereby starting discussions. When sending tweets in the action category, users not only provide feedback but also explicitly call for action. Typical action tweets are messages that call for attending events, signing petitions, making donations, and voting.
In sum, Twitter profiles and tweets can indicate the level of user involvement. More specifically, involved users are likely to be those who self-disclose personal political information, those who send tweets about local events, and those who send engaging tweets—that is, tweets that are likely to spark discussions and action. In this regard, the following hypotheses are proposed:
Hypothesis 2: The more politically involved the users are, based on the level of self-disclosure of personal political information, the more likely users’ messages are retweeted by other users.
Hypothesis 3: The more involved the users are in a given political issue, based on their geographic proximity to the political event, the more likely their messages are retweeted by other users.
Hypothesis 4: The more involved the users are in a political issue, based on their contribution of engaging tweets, the more likely their messages are retweeted by other users.
User identity
Identity information regarding whether a Twitter user is an ordinary citizen, journalist, or institute can be determined from Twitter profiles. Classifying a user’s identity is important because the user’s offline identity may drive online influence. For instance, personal sources are influential in WOM communication because they provide personalized information and are independent of organizations’ efforts to persuade the public (Silverman, 2001; Sun et al., 2006). Twitter is a personalized media platform through which users can follow like-minded individuals based on shared interests and viewpoints (Hsu & Park, 2012; Hsu, Park, & Park, 2013). Like-minded Twitter friends filter and deliver customized information (Yoon & Park, 2013). Hence, average Twitter users can appeal to other users based on similar viewpoints and interests. However, as discussed earlier, established organizations continue to enjoy considerable online influence (Weber & Monge, 2011). More specifically, traditional news media still play an important in the Twitterverse (Papacharissi & de Fatima Oliveira, 2012). In this regard, the following research questions are proposed:
Research Question 1: What types of Twitter users (individuals or organizations) are more likely to be retweeted by other users?
Research Question 2: Are members of traditional news media more likely to be retweeted by other users?
Measures
Social connectivity
Users’ social connectivity was measured by betweenness centrality. Betweenness centrality captures how many connections a user has, like in-degree centrality, and also the user’s strategic location in reaching all other users in the network. More precisely, betweenness centrality measures the frequency with which a user lies in the shortest path connecting everyone else in the network (Freeman, 1979). Higher betweenness centrality means higher level of connectivity (Freeman, 1979).
We measured betweenness centrality based on follower/following relationships of users using the #wirecall hashtag. Twitter hashtags are words or phrases prefixed with the symbol #. Twitter hashtags are used to categorize tweets pertaining to the same topic. In addition to #wirecall, there were other hashtags related to the recall election, such as #wiright, #wiunion, and #wipolitics. However, the hashtags #wiright and #wiunion clearly indicated particular political preferences and thus contained skewed user bases. Therefore, we only considered #wirecall in this study because it was actively used by users across all political spectrum and showed higher activity compared to other related hashtags.
Tweets were gathered through the Twitter API every day from May 29, 2012, to June 5, 2012, the election day. As a result, a total of 8,957 users sending messages with #wirecall were identified. These users were then randomly sampled for a list of 1,000 users. For this, NodeXL, a network analysis tool, was used to extricate following/follower relationships between the sampled users and calculate each user’s betweenness centrality. In the process, isolates were deleted because they did not have explicit relationships with others in the network. In addition, several isolates were inactive/suspended accounts or accounts with restricted privacy settings. Because of the skewness of betweenness centrality, we used the log transformation to correct the distribution.
Involvement and identity
We retrieved Twitter profile information of the sampled users, including their biographical data, profile images, and locations. The users were grouped based on whether they were individuals or organizations. The individual group included ordinary citizens, journalists, citizen reporters, activists, and politicians, whereas the organizational group included government agencies, political parties, nonprofit organizations, and news outlets. In addition, media users were classified as Twitter accounts of traditional media or of individuals affiliated with traditional media.
The users were then coded based on whether they revealed political information on Twitter profiles. Frequently used political keywords on profiles included conservative, Republican, right, GOP, Tea Party, liberal, Democrat, progressive, libertarian, Obama supporter, and left. Some users used Twitter profile images to signal political affiliation. For example, some used comic images that demonize President Obama or Governor Walker. As discussed earlier, self-disclosure of political information was employed as a proxy for users’ political involvement. The users who disclosed political information were coded as showing a high level of political involvement, whereas those who did not were coded as showing a low level of political involvement.
Based on location information on profiles, Wisconsin users were coded as having a high level of issue involvement, whereas those outside Wisconsin were coded as having a low level of issue involvement. We acknowledged the arbitrariness of this coding method for the following two reasons: First, some Wisconsin users might not have disclosed their true location, and second, the coding scheme might have overlooked those who were former Wisconsin residents but were just as concerned about the recall election as current residents. However, this coding approach was a feasible one given resource constraints.
Among the sampled users, 89.3% were individual Twitter users, and 10.7% were organizational users. In addition, 3.8% were traditional media outlets and personnel working for those outlets. Further, 82.2% indicated that they lived in Wisconsin, and 17.8% revealed no location or indicated that they lived outside Wisconsin. Finally, 60.5% chose to explicitly disclose their political affiliation.
To measure users’ issue involvement reflected by the contribution of engaging tweets, we retrieved tweets sent by the sampled users during the election week, which resulted in a total of 3,546 tweets. These tweets were classified into three categories: information (I), commentary/community (C), and action (A) categories. Each category represented a different level of involvement. In this study, the latter two were referred to as engaging tweets. This coding scheme, which reflected a hierarchy of involvement, was based on Lovejoy and Saxton (2012). The three categories were mutually exclusive. Information tweets were used to refer to RTs with no original comments. Here users serve as messengers by sending information tweets. Commentary and community tweets were used to refer to two types of tweets: Commentary tweets were personal comments about shared links or RTs, as well as descriptions of personal activities and thoughts. Community tweets were directed at particular users and aimed at showing appreciation and support and requesting feedback. Action tweets called for taking concrete action such as voting, volunteering, and forwarding information (asking for RTs). As explained, commentary and community tweets, along with action tweets, indicated a higher level of involvement than information tweets. In the final analysis, commentary and community tweets were combined with action tweets as engaging tweets. The ratio of engaging tweets to all tweets sent by a user was used as a behavioral indicator of the user’s issue involvement.
Three coders were employed to code a total of 3,546 tweets. Here Krippendorff’s α (Krippendorff, 2004) was used to assess intercoder reliability based on a sample of 355 tweets, representing 10% of all tweets. The coefficient was within the acceptable range (.76). Among the 3,546 tweets, information tweets (62.2%) were most frequent, followed by commentary and community tweets (25.7%) and action tweets (12.1%).
Opinion leadership
The number of times user’s tweets are retweeted indicates the user’s influence in the information diffusion process. A keyword search was conducted to sort all RTs, and the number of times each user was retweeted was determined. In the final data set, the average user was retweeted seven times (SD = 50.16). Because the data were skewed, we applied the log transformation to correct the distribution.
Results
This study explored the predictors of opinion leadership in social media environments through a network analysis and a content analysis. The predictors—users’ social connectivity, involvement, and identity—were measured using user-generated content. More accurately, social connectivity was measured based on positions in the networks of following/follower relationships. Political involvement was measured by the presence of statements or symbols on Twitter profiles indicating political viewpoints. Issue involvement was measured by the proxy of engaging tweets and geographic proximity to a political event. We hypothesized that more connected and involved users were more successful in influencing information flow within Twitter networks. The results provided general support for the hypotheses.
Figure 1 visualizes the network describing follow relationships between the sampled Twitter users. Table 1 shows the descriptive statistics for the variables and their correlations. We conducted regression analysis using the number of RTs as the dependent variable (see Table 2). The independent variables in the model were betweenness centrality (for social connectivity), the dichotomous categories of whether the user disclosed political viewpoints on profile (for political involvement), of whether the user was a local Wisconsin resident (for issue involvement), and the proportion of engaging tweets (for issue involvement). To address the potential issue of heteroscedasticity, we adopted a robust regression approach called the Huber-White sandwich estimator (Huber, 1967; White, 1982).
Network of directional following/follower relationships between sampled Twitter users. Nodes represent Twitter users who are connected by following/follower relationships. It appears that there is a considerable amount of mutual following between users. Such mutual following is not surprising given users’ shared topic interest. In addition, the network is fragmented into two subgroups and ties are dense within the subgroups. The fragmentation might reflect divisions in political ideologies, which points to potential future research questions.
Descriptive Statistics for Variables and Their Zero-Order Correlations (Means and Standard Deviations Are Presented Along the Diagonal).
RT = the number of log-transformed RTs + 100; Betw. = log-transformed betweenness centrality; Location = issue involvement based on geographic proximity to the event, a categorical variable coded as 2 for Wisconsin users and 1 otherwise; Poli. = political involvement, a categorical variable coded as 1 for a user disclosing his or her political affiliation and 0 otherwise; Identity = the user’s identity coded as 1 for an individual user and 2 for an organization; Media = the user’s affiliation with news media coded as 1 for a user from traditional media and 0 otherwise; Tweets = issue involvement based on engaging tweets—the log-transformed ratio of user’s engaging tweets to all tweets by the user.
p <. 05. **p < .01.
Regression Model for the Number of RTs for a User.
RT = the number of log-transformed RTs + 100; Betw. = log-transformed betweenness centrality; Location = issue involvement based on geographic proximity to the event, a categorical variable coded as 2 for Wisconsin users and 1 otherwise; Poli. = political involvement, a categorical variable coded as 1 for a user disclosing his or her political affiliation and 0 otherwise; Identity = the user’s identity coded as 1 for an individual user and 2 for an organization; Media = the user’s affiliation with news media coded as 1 for a user from traditional media and 0 otherwise; tweets = issue involvement based on engaging tweets—the log-transformed ratio of user’s engaging tweets to all tweets by the user.
p < .05. **p < .01.
The model explained 26% of the variance, F(6,593) = 8.22, p < .001. The first hypothesis addressed users’ social connectivity. The finding showed a pattern consistent with prediction; betweenness centrality was positively related to the number of RTs (β = .26). This relationship may be explained by the following factors: Higher betweenness centrality indicates a broad audience base, which facilitates the dissemination of messages. Likewise, higher betweenness centrality means that the user has a diverse range of information and viewpoints from Twitter connections (Freeman, 1979). The diversity may contribute to the quality of shared content to grab attention from other users.
The second hypothesis posited that users’ political involvement positively predicted the ability to get messages retweeted. However, this hypothesis was not supported. Political involvement may not necessarily translate into political knowledge and the contribution of valuable information. Twitter users may prefer objective and informative messages over partisan and opinion-laden comments.
The third hypothesis examined users’ issue involvement indicated by geographic proximity to a political event. The hypothesis was supported; local users were more likely to be retweeted (β = .20), indicating that issue involvement based on geographic proximity to the event predicted online influence. This relationship may be explained by the following reasons: Local users were more informed about the event and hence were better at contributing attention-grabbing tweets. Local users were witnesses to the event. Being on the ground meant having firsthand personal accounts and insightful observation. Alternatively, local users might have shared offline relationships, leading to increased retweeting of one another’s tweets. The fourth hypothesis discussed issue involvement indicated by sending engaging tweets. The finding supported the hypothesis that issue involvement based on engaging tweets (β = .21) positively predicted the number of RTs.
We also proposed research questions to explore the role of user identity. The first research question examined whether individuals and organizations differed in their ability to get RTs; we found that organizational users were retweeted more than individual ones (β = .11). The results indicated that organizations were more likely to influence Twitter-based political activism than individual users, suggesting that organizations’ offline authority may be consistent with their online influence. However, the number of organizational users was highly disproportional to individual users. Hence, readers are advised to interpret the results with caution. The second research question discussed whether news organizations were better at getting retweeted. The finding showed no difference in the number of RTs between news organizations and other users.
Discussion
This study provides an exploratory analysis of various predictors of opinion leadership within a Twitter activism network. Although many studies have examined the flow of information on Twitter, specifically through the network analysis technique, they have generally been descriptive in nature and focused on network-level factors. Few studies have related individual users’ characteristics and attributes to their ability to get attention and responses from other online users. The present study provides an exploratory analysis of node-level factors and performance in a social network and demonstrates the flexibility of using user-generated content to measure user characteristics. In this regard, the results suggest that user-generated content can be useful for understanding human behavior. Because of the popularity of social media, behavioral data can help scientists reach conclusions based on direct observations of user behaviors instead of self-reported survey responses.
However, the current analytical framework requires further improvements. Social connectivity was measured by snapshot data on follow relationships gathered after the election. However, follow relationships between users undoubtedly changed throughout the election cycle, and therefore, the measure might not have fully captured the level of connectivity between users at the time of their tweets. In this regard, future research should develop a composite measure of connectivity that accounts for the level of the user’s connectivity throughout the election cycle. In addition, the validity of using user-generated content to measure the user’s involvement should be tested. This study’s measure assumes relationships between the profile disclosure of political information and users’ political involvement and between users’ contribution of engaging tweets and their issue involvement. However, these relationships were not empirically tested, and therefore, future research should combine surveys with content analyses to establish the empirical link. It should be noted that using the proportion of engaging tweets to all tweets as a proxy for issue involvement is a novel and convenient approach. Here a more precise approach may be to assign a numerical value to each category of tweets such that the higher the value, the higher the level of involvement.
Future research should extend this study’s framework by considering user connectivity in various types of networks. This study narrowly defines the network by considering only those users who send the #wirecall hashtag. In addition, this study’s SNA measure is based on connectivity within this bounded network. However, Twitter users might have had few connections with other users within the #wirecall network but many outside the network. This indicates a need for considering the user’s overall connectivity in the Twitterverse.
In addition, future research should employ this study’s framework to examine information diffusion in settings not limited to specific issues or social media platforms. This study focuses on a specific context, and therefore, although the Wisconsin recall election was a salient political issue that attracted considerable local and national attention, the event might have had certain characteristics that clearly distinguished it from typical political events. Furthermore, it may be useful to consider political activism on other social media platforms. Although Twitter is a popular platform for political activism, there may be some systematic differences between its users and those of other platforms such as Facebook and Youtube (Hsu & Park, 2011).
Future research can employ this study’s framework to provide a longitudinal analysis of the emergence of opinion leaders. As mentioned earlier, the data were collected during the last week of the election. However, the influence of Twitter users may change over time. That is, a few individual users tend to dominate the dialogue in the initial stages, followed by traditional news media, which implies that the effects of various predictors of opinion leadership may vary according to the stage of a given political event.
Footnotes
Acknowledgements
We would like to thank Ji-young Kim and Seong-cheol Choi from YeungNam University, South Korea, for their assistance in data gathering.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.