On the nature of influence: identifying and characterizing superdiffusers in seven countries

Abstract

An important component of theoretical and applied work on social influence is identifying influential people. Boster et al.’s theoretical framework on superdiffusers provides one method of doing so, but important questions on the nature of influence remain. In particular, because existing studies have primarily sampled U.S. college students, it remains unclear whether (a) the framework adequately characterizes superdiffusers in different populations and (b) our current understanding of superdiffusers applies outside of the United States. To address these questions, we used an online survey to examine factorial validity, metric invariance, and correlates of superdiffuser characteristics in the United States, the United Kingdom, Singapore, South Africa, India, Pakistan, and Australia (total N = 3,476). Results suggest the superdiffuser framework can fruitfully be used to describe and identify influential individuals in diverse contexts. Influence also appears to be a relatively trait-like individual difference rather than a matter of unique fit to a particular country or culture.

In the communication field, both theoretical and applied work on social influence has emphasized the importance of identifying influential people. Work on the diffusion of innovations, for instance, has pointed to the important role of opinion leaders in promoting adoption of new behaviors and technologies (Rogers, 2003). Opinion leaders have been a key component of several health campaigns, such as reducing high-risk behaviors for HIV (Kelly et al., 1992; Theall et al., 2015) and encouraging mammography screening (Earp et al., 2002).

Boster et al. (2011) proposed a theoretical framework that could be used to describe and efficiently identify these highly influential individuals, referred to as superdiffusers. In particular, they argued that influence has three dimensions, capturing the extent to which someone knows others in their community (connector), argues effectively (persuader), and is highly trusted and motivated to share their expertise on a topic (maven). They also developed scales that could be used to measure these dimensions, which have been extensively validated in existing research (Boster et al., 2012, 2015; Carpenter et al., 2022a; Carpenter et al., 2015).

Although the superdiffuser framework represents an important theoretical contribution to our understanding of the nature of interpersonal influence, existing work has all been conducted in the United States, mainly with student populations, which are not necessarily representative of other groups (e.g., see Afifi & Cornejo, 2020). The validity of the framework cross-nationally and cross-culturally remains unclear, and there is also no evidence as to whether our current understanding of superdiffusers is U.S.-specific. For instance, without collecting comparison data from other countries, it is impossible to tell whether findings about the personality characteristics of influential individuals (Carpenter et al., 2022b) describe a trait-like profile shared by superdiffusers across countries or a context-dependent profile specific to U.S. superdiffusers. Accordingly, this study sought to identify and characterize superdiffusers in seven countries: the United States, the United Kingdom, Singapore, South Africa, India, Pakistan, and Australia. The objectives were to examine whether Boster et al.’s (2011) theoretical framework was adequate for identifying influential individuals in diverse contexts and provide a preliminary examination of how characteristics of superdiffusers might differ across countries.

Defining and identifying superdiffusers

As noted above, the Boster et al. (2011) framework proposes that superdiffusers have three characteristics: they are connectors, persuaders, and mavens. Connectors are people who have connections to many other people, particularly through weak ties (Granovetter, 1973) that allow them to bridge different groups. Persuaders are people who are eager to engage in influence attempts and are able to express and debate arguments effectively. And mavens are experts who are willing to give and sought after for advice on a given topic. Someone who has these characteristics in combination (i.e., a superdiffuser) thus possesses the access, ability, and motivation to persuade others in their communities, which grants them considerable influence. Boster et al. (2011) focused on health mavens in their original study, but subsequent research has examined mavens in areas as diverse as politics, music, environmental issues, clothing, and sports (Boster et al., 2015; Carpenter & Averbeck, 2020; Carpenter et al., 2022b).

Existing research suggests this framework is effective in capturing different aspects of influence and identifying highly influential individuals. Measurement work shows the three dimensions to be distinct (i.e., Boster et al., 2011, 2015; Carpenter et al., 2009) and to correlate with related constructs in expected ways (i.e., their nomological network; Cronbach & Meehl, 1955). Persuaders, for instance, are more argumentative (Boster et al., 2011) and perceived to be more persuasive by acquaintances (Carpenter et al., 2022a). Connectors know and are known by more of their peers (Carpenter et al., 2015), are perceived by acquaintances to be well connected (Carpenter et al., 2022a), make friends somewhat more quickly (Carpenter et al., 2019), and show savviness in their awareness of social networks (Smith & Carpenter, 2018; Smith & Fink, 2015). Finally, mavens tend to be high in value-relevant involvement for topics they care about (Boster et al., 2011), active in and knowledgeable about those domains (Boster et al., 2015; Jensen et al., 2020), likely to communicate about those topics (Carpenter & Averbeck, 2020; Francis et al., 2022), rated by others as good sources of advice (Boster et al., 2015; Carpenter et al., 2022a), and motivated to share that advice widely (Smith & Carpenter, 2018). True superdiffusers—meaning those who score in the top 75th percentile on all three dimensions of influence—also appear to be more influential than non-superdiffusers. For instance, a field experiment demonstrated that they can successfully promote health behavior change (Boster et al., 2012).

There has also been some exploration of features that affect the likelihood that someone will fill a superdiffuser role in their community; or, to put it differently, features that distinguish between those who become superdiffusers and those who do not. Previous research has explored personality characteristics (Carpenter et al., 2022b) and communicative characteristics such as social anxiety, argumentativeness (Boster et al., 2011), civility (Carpenter & Averbeck, 2020), argumentation style (Carpenter et al., 2009), and willingness to communicate (Francis et al., 2022). However, this evidence base is still fairly limited.

Lingering questions

Despite the encouraging evidence for the utility of the superdiffuser framework, it remains largely unexamined in non-student populations and entirely unexamined outside of the United States. Ultimately, this raises broad questions about the generalizability of the existing knowledge base on superdiffusers. For example, a number of authors have criticized the use of student samples on the grounds that the attitudes and beliefs of younger individuals are less fully formed than those of older adults (e.g., see Visser et al., 2000; though cf Miller & Manata, 2023). Likewise, Afifi and Cornejo (2020) recently pointed out that communication research has focused in large part on North American (particularly United States) samples, mainly of White and Asian participants, which may not adequately reflect the experiences of people in other groups and cultures. Thus, making confident claims about our knowledge of communication phenomena requires more research with diverse populations.

With this in mind, examining Boster et al.’s (2011) superdiffuser framework in diverse racial and national groups has the potential to bolster our understanding of the theoretical nature of influence. In particular, it is unclear to what extent existing findings may be specific to the United States. If superdiffusers share a similar profile across groups and contexts (i.e., influence is relatively trait-like), findings from U.S. samples could fruitfully be used to make inferences about superdiffusers more generally. However, if this profile varies based on the characteristics valued most within a particular group or culture (i.e., influence is contingent upon context), this would suggest the need for more country- and culture-specific research and a more refined theoretical characterization of what makes someone influential.

Influence as trait-like

One possibility is that people have a particular profile or set of traits that makes them superdiffusers, meaning the same kinds of people end up in that role across time and in different contexts.¹ This perspective appears to be an implicit assumption in much of the work on opinion leaders and superdiffusers. For instance, although Carpenter et al. (2009) acknowledge the possibility that “influential people can be made as well as born” (p. 151), their focus is mainly on investigating individual differences in how people argue. They suggest—and find in their study—that some people naturally craft better arguments than others. In other words, these people appear to have a particular set of traits that predispose them to be more influential.

To be sure, this does not imply that the influence of superdiffusers is entirely independent of context. Given that mavenness is topic specific (Boster et al., 2011, 2015; Carpenter et al., 2022b), there will logically be some differences across contexts in who is influential, depending on what topical expertise is relevant. Still, this work largely takes a trait-oriented view of influence. Furthermore, there is no reason to expect that this overall profile will differ across groups and cultures. If extraversion is an innate characteristic of influential people, for instance (Carpenter et al., 2022b), then superdiffusers across countries should consistently score higher in extraversion than non-superdiffusers.

Influence as contingent on context

An alternative possibility, one that has not been explored much, if at all, in the existing literature, is that one’s superdiffuser status is mainly context dependent. In other words, there is not a universal trait profile that bolsters one’s influence; instead, it is more a matter of one’s fit to a particular situation, time, or group of potential persuasive targets. After all, influence is only possible if community members are actually willing to connect with, listen to, and seek advice from a would-be superdiffuser. In this sense, it may be useful to think of superdiffusers as the people who others choose to be influenced by rather than people who are inherently influential.

This perspective also emphasizes the important role that group preferences and cultural factors might play in determining who is influential. Hofstede (2001), for instance, argues that relatively stable characteristics of national culture affect the preferences and behaviors of people in different countries. Although it would be difficult to make any firm predictions about how national cultures might give rise to country-level differences in who becomes influential, it is plausible that at least some of these dimensions play a role. For instance, masculinity (versus femininity) reflects the extent to which people value stereotypically masculine (e.g., competition, achievement) or feminine (e.g., cooperation, consensus) traits. Accordingly, people may be more likely to become influential if they exhibit traits that match these preferences. For example, perhaps superdiffusers tend to be more verbally aggressive and willing to engage in arguments (i.e., they have a more stereotypically masculine communication style) in countries that value masculinity versus femininity. In other cases, perhaps people are more likely to become influential if they belong to a segment of the population that is highly regarded in a particular culture. For instance, power distance reflects the extent to which people prefer and accept uneven divisions of power and long-term (versus short-term) orientation reflects the extent to which people value tradition and norms versus pragmatism and preparation for the future. People in countries that value high power distance and a strong long-term orientation may thus be more likely to prefer and trust advice from people with seniority and experience, such as respected elders. Accordingly, superdiffusers in these countries may tend to be older, on average, than superdiffusers in other countries.

Hofstede’s other cultural dimensions include individualism (versus collectivism), which reflects the extent to which people, on average, define themselves and expect others to define themselves based on social connections and in-groups (i.e., as “I” versus “we”); uncertainty avoidance, which reflects the extent to which people are comfortable with ambiguity and deviance; and indulgence (versus restraint), which reflects the extent to which people are expected to express versus suppress hedonistic needs and desires. The possible connection between these dimensions and influence is less clear, though they may likewise lead to unique preferences that impact who emerges as a superdiffuser in a given country. Regardless, for any of these dimensions, such cross-national differences would indicate that influence is not associated with a consistent set of characteristics, but a matter of fit to the profile valued by the culture to which one belongs.

The present study

To investigate these alternative possibilities, this study focused on identifying and characterizing superdiffusers in a racially and ethnically diverse U.S. sample and samples from the United Kingdom, Singapore, South Africa, India, Pakistan, and Australia. These countries were intended to capture a geographically diverse sample, including participants from five of six inhabited continents. The first goal was to establish preliminary evidence as to the adequacy of the superdiffuser framework for capturing key dimensions of influence across countries. To do so, we focused on the factorial validity (i.e., fit of the predicted factor structure to the data) and measurement invariance (i.e., similarity in fit across groups) of the superdiffuser scales:

RQ1: Do the superdiffuser scales exhibit (a) factorial validity and (b) measurement invariance across countries?

Provided that evidence for the tenability of the framework could be established, the second goal was to examine predictors of the superdiffuser scales. Drawing on past research, we focused on traits related to personality and communicative tendencies, as well as demographic characteristics. As Carpenter et al. (2022b) note, there are a number of plausible links between the Big Five personality attributes and superdiffuser characteristics. Their results revealed that extraversion, which reflects the extent to which someone is outgoing and sociable, was positively associated with connectivity, persuasiveness, and mavenness for some topics. Openness, which reflects intellectual curiosity and imagination, was also positively associated with persuasiveness and mavenness on many topics. On the other hand, associations with conscientiousness, neuroticism, and agreeableness were found to be weak or inconsistent. However, even these traits may still bear examining, especially in an international sample. For instance, agreeableness has been found to be a positive predictor of friendship network size (Selfhout et al., 2010), which suggests it may be linked to connectivity in at least some contexts.

To characterize communication tendencies, we focused on two variables that would contribute new evidence on the nomological network for the scales (Cronbach & Meehl, 1955): verbal aggressiveness and communication apprehension. Verbal aggressiveness is a tendency to “attack the self-concepts of other people instead of, or in addition to, their positions on topics of communication” (Infante & Wigley, 1986, p. 61). Previous research suggests that persuasiveness is positively associated with argumentativeness (Boster et al., 2011), but has not examined associations with more hostile communication characteristics like verbal aggressiveness. Verbal aggressiveness might plausibly hamper one’s superdiffuser potential by driving other people away or encouraging them to avoid engaging in arguments in the first place. However, verbal aggressiveness could feasibly be persuasive if not directed at the message recipient. For instance, one could conceivably convince a friend to visit one business by attacking the moral character of the owner of a rival business. This might also help explain why Carpenter and Averbeck (2020) found that people who were higher in connectivity and political mavenness were also more willing to use uncivil, disrespectful language when arguing with others on Facebook.

Communication apprehension is “fear or anxiety about communication” in different contexts, including small groups, meetings, dyads, and public speeches (McCroskey, 1977, p. 78). Previous research suggests that social anxiety is negatively associated with connectivity (Boster et al., 2011), but has not examined communication apprehension directly. Some forms of communication apprehension, particularly anxiety about communicating in one-on-one situations, would presumably hamper one’s ability to persuade, make connections with others, or give effective advice. However, this may not be true for other forms, such as public speaking apprehension. Given the difficulty of anticipating the effects for these variables and the U.S.-specific nature of the existing evidence, we simply asked:

RQ2: How are superdiffusers scores associated with (a) communicative tendencies (verbal aggressiveness, communication apprehension), (b) personality characteristics (Big Five), and (c) demographics?

More importantly, we also sought to examine whether these relationships were consistent across countries. To reiterate, we were interested in whether the above variables would be equally predictive of superdiffuser scores across countries, which would suggest that influence is relatively trait-like; or differ on the basis of cultural variables, which would suggest that influence depends more on one’s match to a particular cultural context. Given the lack of evidence on which we could base specific predictions about similarities or differences in superdiffusers across countries, we asked:

RQ3: Do predictors of superdiffuser characteristics vary based on country-level dimensions of culture?

Method

Sample

Participants for the study were recruited through Qualtrics Services from August to September 2022 with the incentive that they could receive a small amount of money for their participation. A sample of at least 250 subjects was requested from each country of interest and from each of five major racial groups in the United States (White, Black, American Indian/Alaska Native, Asian, and Native Hawaiian/Pacific Islander), though Qualtrics was unable to recruit enough participants to meet this quota for the Native Hawaiian/Pacific Islander group. This resulted in a full sample size of N = 3,476 (see Table 1 for demographics). Based on Koran’s (2020) simulation work, an n of approximately 250 per group was adequate to estimate the fit of our desired models.

Procedure

The survey was conducted online using the Qualtrics survey platform. In all cases the survey was conducted in English, so facility with the English language was a screening requirement for participation in the study.² This decision was made because the superdiffuser scales were originally written in English. Had we translated the scales into a variety of languages and then found cross-national differences in their factorial validity, it would have been impossible to determine whether this outcome was due to true differences in scale validity or to differences in translation quality. Conducting all surveys in the same language eliminated this potential source of variance. Subjects gave their informed consent to participate prior to filling out the survey.³

Measures

Superdiffuser scales

The superdiffuser scales were measured using the items developed by Boster et al. (2011). The scales include five items each for connectivity, persuasiveness, and mavenness, all measured on 5-point Likert-type scales (1 = strongly disagree, 5 = strongly agree). In this case, the mavenness items were worded to focus on physical fitness, because we reasoned that this topic would be applicable regardless of country (see Supplementary  Table S1 for full item wording).

Individual-level predictors

Verbal aggressiveness was measured using the 10 aggressively worded items from the scale developed by Infante and Wigley (1986). Existing measurement work has shown that the original 20-item scale captures two separate factors (Levine et al., 2004), so we excluded the reverse-coded, benevolently worded items for the purposes of this study. The items were measured on 5-point Likert-type scales (1 = almost never true, 5 = almost always true).

Communication apprehension was measured using the PRCA-24 (McCroskey et al., 1985). The scale is intended to measure anxiety about communicating in four contexts: small-group discussions, meetings, dyadic conversations, and public speeches. Each was measured with six items on five-point Likert-type scales (1 = strongly disagree, 5 = strongly agree).

Personality characteristics were measured using the self-descriptive adjective items presented in Zimprich et al. (2012). This version of the inventory includes four items each to measure neuroticism and conscientiousness, five items each to measure extraversion and agreeableness, and seven items to measure openness. All items were measured on seven-point Likert-type scales (1 = strongly disagree, 7 = strongly agree).

Demographic measures included age, sex, education, political ideology, and religiosity. Education was measured by asking participants to indicate their age of last school attendance (their current age, if still attending school). This measure was used because terms used for different levels of education vary by country. Political ideology was measured with a single item (“In political matters, people sometimes talk about ‘the left’ and ‘the right.’ How would you place your views on this scale, generally speaking?”; Evans et al., 1996) on a 10-point semantic differential scale. A simple left-right measure was used because names of particular political parties and positions on specific political issues differ by country. Finally, religiosity was measured using the CRS-5 (Huber & Huber, 2012). This scale includes five items (e.g., “To what extent do you believe that God, deities, or something divine exists?”) measured on five-point Likert-type scales (1 = not at all, 5 = very much so). It is intended to measure the importance of religion in someone’s life, regardless of their specific faith.

Cultural dimensions

Country-level scores on Hofstede’s dimensions of national culture (power distance, individualism, masculinity, uncertainty avoidance, long-term orientation, and indulgence) were taken from the data reported on his website (hofstede-insights.com; see Supplementary Table S2). Although Hofstede’s work has been criticized in various ways (e.g., McSweeney, 2002; Spector et al., 2001), there are few agreed-upon alternatives and none for which data are available for all seven countries studied here. Given that our goal was a preliminary exploration of possible country-level differences, we considered Hofstede’s dimensions to be a reasonable starting point.

Analysis procedure

Factorial validity

The first stage of analysis focused on the factorial validity of the superdiffuser scales (RQ1a). For this purpose, a confirmatory factor analysis (CFA; see Hunter & Gerbing, 1982) was performed using the lessR package in R version 4.2.3 (Gerbing, 2023; Hunter & Gerbing, 1982; R Core Team, 2023), which uses iterated centroid estimation (see Hunter & Gerbing, 1982). Fit was assessed primarily by inspecting residuals to identify items consistently associated with large errors (i.e., invalid items; see Manata & Boster, 2024) for removal from the model. The comparative fit index (CFI) and standardized root mean residual (SRMR) were also used as indicators of global model fit, using recommendations from Hair et al. (2019). These indices were obtained using the lavaan package (Rosseel et al., 2023), which uses maximum likelihood estimation.

Measurement invariance

The second stage of the analysis focused on the measurement invariance of the superdiffuser scales across countries (RQ1b). Although there are several means by which to assess measurement invariance (for a discussion, see Kim et al., 2017), a multi-group CFA approach was adopted here. The multi-group approach was preferred over other methods (e.g., multi-level approaches) because it works well when group size is small, as was the case here (i.e., group k = 7).

The analysis began by testing for configural invariance, which provides an assessment of whether the structure of the measurement model remains invariant between groups (i.e., whether the same items load on the same factors in each group). Next, different equality constraints were added in sequential order to assess whether more specific parameters remained invariant (i.e., equal factor loadings and item residuals). If model fit was not significantly attenuated despite the added equality constraints, then the measurement model was considered invariant (Chen, 2007; Cheung & Rensvold, 2002).

Based on preliminary simulation studies, current recommendations are that invariance in item factor loadings and residuals can be inferred when changes in CFI values are ≤.01 when adding each constraint (Chen, 2007; Cheung & Rensvold, 2002). A simulation by Chen (2007) also showed that with large sample sizes (N > 300), invariance in residuals and factor loadings can be inferred when changes in SRMR values are ≤.01 and ≤.03, respectively. These decision criteria were adopted when assessing the extent to which the measurement model remained invariant between countries, though we emphasize that these decision criteria were treated as general guidelines, not strict rules (Chen, 2007; Marsh et al., 2004). This is because simulation work in this corpus is limited by the scope of the designs implemented, which have so far assessed invariance in a limited number of conditions (e.g., two groups only). All analyses were performed using the lavaan package in R version 4.2.3 (Rosseel et al., 2023).⁴

Main analyses

Provided evidence for the validity of the superdiffuser scales could be established, we then planned to examine predictors of scores on each subscale (connectivity, persuasiveness, and mavenness). Superdiffusers have been defined in past research as individuals scoring above the 75th percentile on each subscale (see Boster et al., 2012, 2015), so it would hypothetically have been possible to run an analysis of predictors of this dichotomous variable (0 = non-superdiffuser, 1 = superdiffuser). Doing so in a cross-national sample, however, is not exactly straightforward. In particular, one could classify people as superdiffusers based on their scores relative to the entire sample—which would identify the most influential people across nations—or relative to others within the same nation—which would identify the most influential people within nations. The choice of method also has a non-trivial impact on estimates of superdiffuser prevalence. For example, 13.3% of the South African sample (n = 34) would be classified as superdiffusers using the first method, as compared to 5.5% (n = 14) using the second method. Focusing simply on participants’ continuous scores on each subscale rather than choosing arbitrarily between these classification methods thus provided a simpler and more straightforward method of examining predictors of influence (see Supplementary Table S4 and Supplementary Figures S1–S3 for additional descriptive information for each subscale at the country level).

To examine RQ2–3, we ran a series of multilevel regression models (MLMs) with connectivity, mavenness, and persuasiveness as the dependent variables. MLMs were used to account for the fact that participant responses (N = 3,476) were nested within country (k = 7). To build each model, we began with the unconditional mean model to examine the extent to which scores clustered by country. Next, we added the level-1 predictors (communication variables, personality variables, and demographics) using forward-stepping procedures (see Nezlek, 2008), such that predictors were added one by one to check for significance, and only significant predictors were retained in the final model. Then, we examined the extent to which there was variance in slopes between countries for this final set of predictors, as it is only useful to examine interaction effects with level-2 variables if there is variance in slopes to explain. To do so, we used likelihood-ratio (LR) tests to compare fit of an augmented intermediate model—estimating random slope variance for all significant predictors—to the fit of a constrained intermediate model—using a fixed slope for a given predictor—for each predictor in turn (see Sommet & Morselli, 2017). Although absence of evidence is not necessarily evidence of absence (e.g., Nezlek, 2008), significant LR tests can indicate whether there is meaningful variance in slopes to be explained. Finally, for predictors for which there was evidence of variance in slopes, we examined the extent to which two-way interactions with each level-2 variable (cultural dimension), in turn, could account for that variance. Each dimension was examined separately because there was substantial multicollinearity between the cultural dimensions within the sample of seven countries selected for the study (e.g., individualism and masculinity were correlated r = .93), so there was no meaningful way to examine their effects together in a single regression analysis. For all analyses, continuous level-1 predictors were group-mean centered, and level-2 predictors were grand-mean centered (see Enders & Tofighi, 2007).

Results

Factorial validity and measurement invariance

The CFA results indicated that Boster et al.’s (2011) proposed first-order three-factor model for the superdiffuser scales provided good fit to the data, χ² (87) = 702.25, CFI = .98, SRMR = .03. An initial test of the configural model also suggested a good fit to the data, χ² (168) = 289.68, CFI = .99, SRMR = .02, indicating that the same items provided an adequate fit to the data in each country. Subsequent nested model comparisons also indicated that both factor loadings (ΔCFI < .01; ΔSRMR = .01) and item residuals (ΔCFI = .01; ΔSRMR < .01) remained largely invariant across nations.⁵ This indicates that participants in each nation were interpreting and responding to the superdiffuser items in a similar manner. As such, both RQ1a and RQ1b were answered in the affirmative: The superdiffuser scales exhibit factorial validity and measurement invariance (see Supplementary Table S3 for descriptives, reliabilities, and correlations). In other words, the framework appears to capture dimensions of influence that are both recognizable and distinct in a variety of countries.

Main analyses

Proceeding with the main analyses, the unconditional mean models suggested some clustering of scores on the superdiffuser scales by country: for connectivity, $σ02$ = 0.11, ICC(1) = .10; for persuasiveness, $σ02$ = 0.04, ICC(1) = .06; for mavenness, $σ02$ = 0.39, ICC(1) = .25. Multilevel regression analyses were thus used to examine the extent to which the measured communication variables, personality characteristics, and demographics (level 1) and cultural characteristics (level 2) could explain the observed within- and between-country variance. To assess RQ2, we focused on the models including only level 1 predictors (see Table 2) before moving on to models including level 2 predictors for RQ3.⁶ To reiterate, connectivity, persuasiveness, and mavenness were examined with separate analyses because it was simpler and more straightforward than arbitrarily defining a cutoff to separate superdiffusers from non-superdiffusers.

Connectivity

Consistent with previous research (e.g., Boster et al., 2011; Carpenter et al., 2022b), connectivity scores were negatively associated with interpersonal apprehension (though not with public speaking apprehension) and positively associated with extraversion. However, whereas previous research found no significant associations between connectivity and other personality variables (Carpenter et al., 2022b), both openness and pro-sociality were found to be positive predictors here. People who were more religious, younger, and more verbally aggressive also tended to score higher in connectivity.

Furthermore, LR tests suggested that the variance in slopes for most of these predictors was small enough that estimating random slope variance did not improve model fit. This suggests that associations between connectivity and these predictor variables were largely the same across countries rather than varying between countries. The only exception where estimating random slope variance improved model fit was for pro-sociality, χ² (1) = 9.32, p = .002. However, there was no evidence to suggest the cultural variables moderated this association. Instead, it appeared that the variance in slopes was due primarily to a uniquely strong relationship between pro-sociality and connectivity in the sample from India (β = .47) relative to more modest relationships evident in other countries (combined, β = .10). Still, the results suggest that for connectivity, RQ3 can be answered mainly in the negative; there was limited evidence to suggest that predictors of connectivity varied by country and no evidence to indicate that these relationships were moderated by cultural variables. Thus, it appears that connectivity may be more trait-like than context-dependent.

Persuasiveness

Persuasiveness was largely predicted by the same variables as connectivity, with the exception that no demographic variables were significant predictors in this model. These findings are also largely in line with previous findings, with the exception of pro-sociality; whereas Carpenter et al. (2022b) report a negative association between persuasiveness and agreeableness, the agreeableness-conscientiousness composite was a positive predictor of persuasiveness here.

LR tests also suggested that there were no cases in which estimating random slope variance improved model fit for the persuasiveness model. Thus, the results provide no evidence to suggest that the associations between these predictors and persuasiveness scores vary across countries. For persuasiveness, RQ3 can again be answered in the negative; there was no evidence to suggest that predictors of persuasiveness varied by country. Again, it appears that persuasiveness may be more trait-like than context-dependent.

Mavenness

Mavenness was also associated with largely the same predictors as both connectivity and persuasiveness. Once again, the analyses revealed positive associations with extraversion, openness, and pro-sociality (similar to findings in Carpenter et al., 2022b). Mavenness was positively associated with verbal aggressiveness and negatively associated with interpersonal apprehension (but not with public speaking apprehension). As with connectivity, people who were younger and more religious also tended to score more highly on mavenness. The main distinction from the other two models was that mavenness scores also tended to be higher among men, political conservatives, and those with lower neuroticism scores.

For this model, results also indicated that model fit improved when estimating random slopes for age, χ² (1) = 7.36, p = .007; interpersonal apprehension, χ² (1) = 3.85, p = .049; and verbal aggressiveness, χ² (1) = 4.81, p = .03. Thus, there were some cases where relationships between mavenness and predictor variables differed across countries, though it is also worth noting that the variance in slopes tended to be very small (⁠$σ12$ < 0.01).

In the case of interpersonal apprehension, there was no evidence to suggest that the variance in slopes could be accounted for by any of the cultural variables. Instead, it appeared that this variance was due to the comparatively weak relationships between apprehension and mavenness in the samples from the United Kingdom (β = .08) and Australia (β = −.06) relative to the other countries (combined, β = −.25). On the other hand, the evidence for moderation was somewhat more compelling in the case of verbal aggressiveness and age. In particular, the relationship between verbal aggressiveness and mavenness was moderated by country-level indulgence scores (unstandardized B = 0.004, p = .04; see Figure 1), such that aggressiveness had a stronger positive association with predicted mavenness scores in high-indulgence countries (e.g., Australia) than low-indulgence countries (e.g., Pakistan).⁷ Furthermore, the relationship between age and mavenness was moderated by country-level power distance (unstandardized B = 0.005, p < .001; see Figure 2), such that age had a positive association with predicted mavenness scores in countries with higher power distance (e.g., South Korea), but a negative association in countries with lower power distance (e.g., the United Kingdom).⁸ Thus, although this evidence is limited, it suggests that the answer to RQ3 is more ambiguous in the case of mavenness. Although mavens share a similar profile across the countries examined here, there also appear to be some country-specific preferences in terms of what makes someone a valued source of advice.

Discussion

The purpose of this study was to identify and characterize superdiffusers (Boster et al., 2011) in a diverse, cross-national sample. In general, the results reflect very favorably on Boster et al.’s theoretical framework. The three dimensions of influence they propose clearly represented distinct constructs in all seven sampled countries, as indicated by the evidence for the measurement model’s factorial validity, reliability, and the relative invariance of the factor structure, factor loadings, and residuals. Furthermore, the results suggest that superdiffuser scores tend to be associated with a particular demographic, psychographic, and communicative profile that was largely the same across the countries sampled here. When differences between countries did emerge, they were also as likely to be idiosyncratic as attributable to any of Hofstede’s (2001) proposed cultural characteristics.

The findings have important theoretical implications for how influence is conceptualized. An implicit assumption of much of the literature on opinion leaders and superdiffusers appears to be that influence is relatively trait-like (e.g., see Carpenter et al., 2009), in that some people exhibit a particular profile of characteristics that predisposes them to be more influential than others. Examining the predictors of superdiffuser characteristics across countries suggests that there is some truth to this statement. The demographics, personality characteristics, and communicative tendencies associated with influence were remarkably similar across the seven countries examined here, despite the cultural and geographic differences among them. The main exception to this pattern was in the case of mavenness. Mavens tended to be older in countries that were higher in power distance. Furthermore, mavens’ communicative tendencies tended to differ across countries. Interpersonal apprehension had clear negative associations with mavenness in most of the samples, but not in the United Kingdom or Australia. Mavens also tended to be more verbally aggressive in countries that were high in indulgence, but less so in countries high in restraint. In other words, it appears that whereas one’s connectivity and persuasiveness may be trait-like, one’s mavenness is at least partly a matter of fit to country-specific preferences and values. Of course, these results cannot speak to why these patterns emerged, though it is plausible to expect they may have to do with communication preferences in different cultures. For example, perhaps verbally aggressive individuals are seen as particularly inappropriate in high-restraint countries, which makes them less appealing sources of advice.

It is also worth noting that this finding is specific to mavenness for the topic of physical fitness, so the findings might differ for different topics. For example, we might expect mavens’ characteristics to be particularly dependent on cultural context for topics that are uniquely sensitive or controversial in particular countries. The issue of climate change, for instance, is more politically divisive in the United States than in most other countries (Poushter et al., 2021). As a result, who is seen as a trusted source of climate change-related advice in the United States may be influenced much more strongly by factors such as partisanship, political ideology, or even demographics that differ in clear ways between political parties (e.g., religiosity; Hartig et al., 2023) than it is in other countries. Indeed, the topic-dependent nature of mavenness may be one reason why the results revealed this dimension to be more context-dependent than the other dimensions of influence. Future work on the mechanisms that facilitate different types of influence over time, in varied cultural contexts and for varied topics, would help clarify why these different patterns emerged and continue to advance our understanding of the nature of interpersonal influence.

These findings have also practical implications for future use of the superdiffuser scales. Corroborating existing evidence as to their validity (e.g., Boster et al., 2011, 2015; Carpenter et al., 2009) the results here provide strong evidence that not only are they valid and reliable, but they also measure key superdiffuser characteristics with equal factorial validity and precision in different countries. This suggests that the scales can be used with some confidence when conducting cross-national or cross-cultural research.

From an applied perspective, the results also indicate that the scales could fruitfully be used as a tool to identify influential people in diverse contexts and point to some tactics that may be useful when incorporating the scales into campaign design. In particular, the finding that religiosity was a significant predictor of connectivity and mavenness across countries suggests that religious communities may be good places to start when attempting to identify community members to serve as partners in promoting prosocial behaviors. This finding is not entirely surprising in that formal organizations such as religious groups are often a source of weak ties (Granovetter, 1973), but still points to a strategy that may not have been otherwise intuitive. At the same time, however, this finding also suggests that there may be challenges to using opinion leader approaches (e.g., Kelly et al., 1992) for campaigns that religious groups find objectionable, such as a fundraising drive for Planned Parenthood. To wit, if the most influential people have strong religious beliefs, campaigns that appear to conflict with those beliefs will be more difficult to promote.

Strengths, limitations, and future directions

Although the findings contribute important evidence as to the utility of the superdiffuser scales and the nature of influence more generally, they should also be interpreted in light of the study’s strengths and limitations. A clear strength of the study is that the sample was much more diverse than those employed in most research in communication (Afifi & Cornejo, 2020). We deliberately oversampled for diverse racial groups in the United States to ensure that the results were not specific to any particular group. We also captured data from five continents, including North America, Asia, Africa, Europe, and Australia. Still, for practical reasons, the study was only able to examine samples from a limited selection of English-speaking countries. We cannot assume the findings as to metric invariance or the consistency of the superdiffuser profiles would hold for other countries or when the scales are translated into other languages. Accordingly, this study should only be viewed as a starting point for understanding influence cross-nationally.

Likewise, although we examined a range of demographic variables, personality characteristics, and communication tendencies as predictors, there are many other variables that might be important for predicting influence in different countries. For example, it is possible that superdiffusers tend to be people who fill important and valued roles in their communities by working in a respected occupation, having a high socioeconomic status, or participating in local voluntary organizations. Given that the occupations and memberships that communities value are likely to vary by culture (as is the accessibility of those occupations and organizations to different segments of the population), it is possible that these variables also lead to important differences in who becomes influential. Thus, although the finding that superdiffusers had a similar profile of traits across seven countries is encouraging, different results might emerge were we to examine other variables not investigated here.

Furthermore, although the results of the study provide evidence that the superdiffuser framework works well beyond the college student samples used in most existing research, they do not directly address questions about substantive ways that superdiffusers may differ in younger versus older populations. For instance, if younger people indeed have weaker attitudes and beliefs (e.g., Visser et al., 2000), they may be less capable of crafting strong persuasive arguments for their positions or of serving as appealing sources of advice on various topics. Thus, the young people who become influential among their peers may differ from older people who fill similar roles. Additional research that more explicitly compares superdiffusers in younger and older populations would help address these questions.

More sophisticated methods of examining cultural predictors of between-country differences would also be beneficial in future research. Hofstede’s (2001) dimensions are not the only way to capture culture. His approach to measuring these dimensions has been criticized (e.g., Spector et al., 2001) and the scores, even if valid, only capture a country’s culture at a particular point in time. Additional aspects of culture and other measures of cultural characteristics might bear examining in the future (e.g., see Triandis, 2004). For example, future research might explore differences between high- and low-context cultures (Hall, 1959). In high-context cultures, people tend to assume that others will share extensive knowledge and information, such that meaning can be communicated effectively in implicit, subtle ways. In low-context cultures, in contrast, people communicate meaning in more explicit, direct ways that do not rely on a shared understanding. There is some evidence that this factor affects how people approach persuasion, in that advertising strategies differ in some ways between high- and low-context cultures (Liang et al., 2011; Zhou et al., 2005). By extension, it may be that superdiffusers also need different tactics to be persuasive depending on the communication styles that are normative in the culture to which they belong. Another useful cultural variable to explore may be societal levels of trust. Social trust describes a broad tendency for people in a particular society to think of others as trustworthy, which tends to be associated with, among other things, the strength of informal social networks (Delhey & Newton, 2003). Accordingly, superdiffusers might need different strategies to establish social connections in cultures with high social trust than low social trust, which would lead to different predictors of connectivity. Investigating cultural characteristics such as these may help account for the apparently idiosyncratic associations with pro-sociality and interpersonal apprehension in India and in the United Kingdom and Australia, respectively, as well as identify other variables that may be important in particular cultures.

Employing other measures of the cultural dimensions examined here might also help avoid some of the limitations of these measures that arose in this study. In particular, the limited number of countries resulted in range restriction for some of the cultural dimensions, which may have limited their ability to explain variance in observed effects. There was also substantial multicollinearity among the scores on the different cultural dimensions, which made distinguishing their effects challenging. Still, it is important to note that there was very little variance in slopes to be explained in the first place, so the impact of these limitations on the final results and conclusions is limited.

As evidence on cross-cultural similarities and differences in the characteristics that predict influence continues to build, future research might also begin to consider the processes that give rise to these associations. For example, the fact that religiosity consistently predicted influence across countries in our sample does not necessarily imply that these superdiffusers became influential in the same ways. In some countries, perhaps religion plays a role mainly because it serves a community-building function, which helps would-be superdiffusers become influential by expanding their social networks and building trust among their contacts. In other countries, in contrast, perhaps religion plays a role mainly because there are strong social expectations around faith and religious attendance, such that less religious individuals automatically face distrust or rejection. Investigating these possibilities will continue to advance our understanding of who becomes influential and why.

Conclusion

Overall, this study suggests that Boster et al.’s (2011) superdiffuser framework effectively captures key dimensions of interpersonal influence both in diverse U.S. populations and cross-nationally. Bolstering existing evidence, our results revealed the superdiffuser scales to have strong factorial validity, reliability, and metric invariance across seven countries, which emphasizes their utility for continued use in both theoretical and applied research in the future. Examining predictors of the subscales cross-nationally also suggests that influence—perhaps with the exception of mavenness—may be relatively trait-like in that its associations with demographics, personality characteristics, and communicative tendencies are largely the same in diverse cultural and national contexts. Future cross-national and cross-cultural research would continue to bolster our understanding of what makes people influential.

Supplementary material

Supplementary material is available online at Human Communication Research online.

Funding

This project was supported by a faculty research award from the Department of Communication Arts & Sciences at The Pennsylvania State University.

Data availability

Data associated with this project are available upon request to the corresponding author.

Conflicts of interest

None declared.

Footnotes

References