Opinion leaders in Tweeter: Those with stronger motivations to distribute relevant information tended to overestimate their influence in the network

#Opinionleaders: a comparison of self-reported and observable influence of Twitter users. Stephan Winter et al. Information, Communication & Society, Jan 1 2020. https://doi.org/10.1080/1369118X.2019.1705374

ABSTRACT: Social media have become forums of discussions on political and societal debates in which individual users may forward information or influence others. While prior studies either employed network analyses or surveys to identify opinion leaders and their characteristics, the present investigation combines these two approaches to address the relationship between observable and self-perceived influence. For this purpose, a retweet network of Twitter communication on the Brexit debate (N = 15,018) was analyzed in relation to a survey on motives and personality traits that was filled out by a subsample of active users (N = 98). Results showed that users’ eigenvector centrality (as a measure of influence in the network) was significantly related to their political interest and their number of followers, but not to self-perceived opinion leadership. According to a comparison of self-assessment and network position, those with stronger motivations to distribute relevant information tended to overestimate their influence in the network. Implications for the identification of opinion leaders are discussed.

KEYWORDS: Two-step flow, opinion leadership, opinion expression, motivations, network analysis, Twitter

Discussion
The goal of this study was to investigate the characteristics of opinion leaders whose influence can be detected on the basis of observations in contemporary social networks. Furthermore, we aimed to combine prior approaches of survey research and network analyses
(see Kim et al., 2017) in order to compare self-perceptions with more objective data. For
this purpose, we conducted a mixed-method study (a network analysis and survey with a
subsample of the network) with the exemplary case of the Twitter communication regarding the Brexit 2016.
As a measure of observable influence, the present investigation focused on retweets
because this act of passing along information (which is then visible in further users’
feeds) increases the reach of the original message. The measure of eigenvector centrality
(Bonacich, 2007; Dubois & Gaffney, 2014) calculated on the retweet graph captures not
only the number of people who have retweeted one’s original tweets but also how influential the people who further distributed one’s messages are. Results showed that eigenvector centrality was significantly related to political interest (as measured in the
survey). This is in line with prior findings that rendered political interest as a precondition
to actively engage in opinion expression online (e.g., Vraga et al., 2015) and offline (e.g.,
Katz & Lazarsfeld, 1955). The fact that political interest is not only related to self-perceived
influence, as shown in these prior studies, but also to observed influence suggests that
being interested in the domain also makes it more likely that users are able to write tweets
that raise the attention of others. Perhaps these users’ tweets contain higher-quality or
more recent information, which would be a positive outcome for the democratic discourse,
but it is also conceivable that they are more controversial and therefore arouse further
attention. These two opposing assumptions can be tested in future studies that involve
a content analysis of the tweets.
Extraversion and personality strength, which were shown to be predictors of offline
opinion leadership (Gnambs & Batinic, 2012; Weimann, 1991), were not significantly
related to users’ centrality. While extraversion may not be important due to the online
context that may give more opportunities for introverts to carefully prepare their messages, the lack of a relationship with personality strength is in contrast to prior findings
on self-reported opinion leadership in social media (Winter & Neubaum, 2016). Possibly,
this discrepancy can be explained with the more objective measurement and suggests that
personality strength might not serve as a reliable indicator of influence in terms of visibility and reach, at least in this specific debate. However, it has to be noted that persuasive
effects on followers’ attitudes (which are also an integral part of influence) could not be
captured by the present analysis. Thus, the measurement on the basis of retweets may
have overlooked the potential impact that people with high personality strength exert
on others’ attitudes.
With regard to the network structure, eigenvector centrality was positively related to
the number of followers and the number of original tweets a user has generated. In line
with prior research (Xu et al., 2014), these network characteristics can be seen as prerequisites of becoming an opinion leader. While this is unsurprising, the additional influence of
political interest suggests that the potential of a large followership and active posting is
exploited when users are sufficiently knowledgeable in the domain to write messages
that get the attention of others.
In the comparison of self-reports and observed influence, results showed a divergence
between self-perceived opinion leadership and the actual visibility in one’s Twitter network. Although the correlation coefficient was positive, it was rather small in magnitude
and non-significant, which raises the question whether people are able to accurately estimate which impact they have. In most cases, they seem to use the parameters that are
available to them (the number of followers and the number of tweets they generated),
but the further distribution via retweets may be difficult to estimate (experienced Twitter
users may utilize further options of Twitter or monitoring tools for more detailed analyses,
but this is unlikely to be employed by average users). In this stage, it is conceivable that
users with certain characteristics tend to over- or underestimate their influence.
In order to explore these potential biases, we compared the ranks of users according to
their self-perceptions and their more objective influence in terms of eigenvector centrality
(and alternatively the mere number of retweets) and computed a difference score that represents a subjective overestimation of one’s influence. This variable was positively related
with personality strength, suggesting that those who perceive themselves as charismatic
may not always be accurate in their estimation of influence. Although personality strength
has been regarded as a direct measurement of opinion leadership in prior research (e.g.,
Schenk & Rössler, 1997; Shah & Scheufele, 2006), this may raise questions on how suitable
the personality strength scale is when it comes to assess a person’s reach in the online context; however, it has to be noted that results only showed a marginally significant relation
for the index based on eigenvector centrality (and a significant relation in the alternative
analysis based on the number of retweets). Additionally, the extent to which users are
motivated to distribute relevant information was significantly related to an overestimation:
Although this motive may spur users’ activity (Winter & Neubaum, 2016), it appears that
users might be overly optimistic and that their goals of informing others about public
affairs are not always reached. One explanation could be that users who are driven by
the motive of informing others might be inclined to post more factual information,
while those who wish to persuade make use of emotional appeals. Emotional messages
have been shown to be more likely to be shared or reacted to on social media in various
settings (Stieglitz & Dang-Xuan, 2013). However, this explanation has to be tested in
future studies.
In the interpretation of results, the following limitations have to be kept in mind. First,
the present investigation focused on eigenvector centrality as an indicator of visibility and
reach in the network, which is an important aspect of opinion leadership but, as mentioned above, does not cover the actual persuasive influence (Robinson, 1976). Thereby,
the self-perception measurement included a more extensive understanding of influence,
while the analysis was only able to focus on observable quantitative aspects. Future studies
could therefore also employ ratings by other users (Schenk & Rössler, 1997) to assess how
a specific user is regarded by followers and how her/his content leads to actual attitude
change. Second, the analysis did not consider the content of the messages. A next step
for future research could be to investigate whether messages with a greater reach and visibility differentiate from those that do not get as much attention (for instance, are opinion
leaders’ tweets more accurate since these people are more politically interested?). Third,
the survey data is limited by the small sample size. Due to the restrictions of sending messages to Twitter users, the recruitment procedure turned out to be more challenging than
expected and resulted in a low response rate and thus in a relatively small sample.
Furthermore, it is likely that those who decided to take part in the survey were more interested in the topic (results showed a high level of political interest in the sample). Comparing the methods of inviting users publicly versus privately, direct messages led to a higher
turnout, which could be considered in the recruitment for similar studies. Finally, the
community of Twitter users may have particularities that do not apply to the general
population (for instance, Hölig (2018) showed that German Twitter users are more extraverted and report higher personality strength than average Internet users). Thus, the
findings warrant replication with different groups, topics, and platforms.
Despite these limitations, we argue that the present study demonstrates the need for
combined approaches of computer science and social science methods to get further
insights into the characteristics of opinion leadership in social media. The unprecedented
combination of behavioral and survey data indicates a gap between self-perceived opinion
leadership and people’s observable influence in the network. Since Twitter users may not
have knowledge about whether they have successfully passed along information or persuaded others, Twitter users take the mere quantity of followers and posting activities
as a proxy to assess their potential influence. The divergence between self-perceived
opinion leadership and observable influence in terms of reach appears to be fueled by
people’s personality strength and their motivation to distribute relevant information.
These factors seem to lead to an overstatement, indicating that those factors that were proposed to drive opinion expression may also be the variables which dilute the accuracy of
self-reports on online influence.