The Engagement Model of Opinion Leadership: Testing Validity Within a European Context

Opinion leadership has been examined with a range of conceptualizations and measurements ever since Lazarfeld and his colleagues (1944) introduced the concept sixty years ago. Explicating and measuring opinion leadership continues to be an important theoretical topic, as personal influence and interpersonal discussion are widely recognized as shaping public opinion, political behavior, and idea diffusion (Katz, 1957; Keller & Berry, 2003; Lazarsfeld, Berelson, & Gaudet, 1944; Rogers, 2003; Scheufele & Shah, 2000; Sperber, 1990; Weimann, 1994; Yin, 1999). Though somewhat overlooked in the original conceptualization, information-giving and information-seeking behaviors are key underlying mechanisms of opinion leadership and have been the focus of considerable research on opinion leadership and the consequences for public opinion or political participation (Burt, 1999; Childers, 1986; Katz, 1957; Lazarsfeld et al., 1944; Rogers, 2003; Scheufele & Shah, 2000; Shah & Scheufele, forthcoming; Weimann, 1994).

However, the concept of opinion leadership arose out of empirical research conducted primarily within the United States. Whether it may be applied to other political or cultural contexts remains to be determined. Some scholars have begun to explore how the concept and measurement of opinion leadership may be applied to other national or cultural contexts outside the United States (Cosmas & Sheth, 1980; Dawar, Parker, & Price, 1996; Marshall & Gitsoduramo, 1995; Weimann, 1994). Yet, the issue of international generalizability and validity of American models of opinion leadership to other national or cultural contexts has not been widely considered and the available research is scant (Dawar et al., 1996).

This paper will directly address the ecological validity of an established concept and measurement of political opinion leadership formulated within the United States by Keller and Berry (2003) by examining how it may be applied to a European context. Keller and Berry, based upon research conducted by Roper ASW since the 1970s, created their measure of opinion leadership using measures of participation in political and civic activities, which I shall term the engagement model of opinion leadership. This model identifies opinion leaders by the number of political activities in which the respondent has engaged within the last 12 months. I will begin by exploring the assumptions underlying their engagement model and how it relates to other existing models of opinion leadership. Furthermore, opinion leadership has been previously defined as exhibiting three primary dimensions: social embeddedness, information-giving, and information-seeking behaviors (Katz, 1957; Flynn, Goldsmith, & Eastman, 1996; Weimann, 1994). However, the scant cross-cultural research on opinion leadership has noted that the nature and type of information-seeking behaviors are the most culturally contingent of the three dimensions (Dawar et al., 1996). Thus, an additional focus will be to explore how different types of communication behaviors that have been traditionally associated with opinion leadership within the United States may vary across national or cultural contexts.

Employing the 2002–3 European Social Survey (ESS), I construct a similar engagement measure of opinion leadership to that employed by Roper ASW. I then examine the ecological validity, or applicability, of the engagement model to a European context by testing its construct validity within fifteen different European countries. This test will examine the significant relative influences and variance explained by characteristics (i.e. education, social gregariousness) and behaviors (i.e. media use, interpersonal political discussion) of opinion leaders previously identified within the United States and by other cross-cultural research.

Europe should be the optimal region outside the United States for this model as it has the closest structural (i.e. economic development, media penetration, political system, political communication patterns, social structure) and cultural (i.e. religion, ideology, culture, history) similarities with the United States when compared to any other global region, while sharing the same broad patterns of political participation and civic engagement (Almond & Verba, 1963; Dalton, 2000; Inglehart, 1997; Norris, 2000, 2002; Verba, Nie, & Kim, 1978).

CONCEPTUALIZING OPINION LEADERSHIP

Defining and Measuring Opinion Leadership

From its inception, there has been a duality within the concept of opinion leadership, as opinion leaders were viewed as both persuaders, or information givers, and information seekers, though the latter was implicit. The initial research by Lazarsfeld and his colleagues focused on asking individuals about attempts to persuade others and providing advice; theoretical associations between opinion leaders and information-seeking were only developed later (Katz, 1957).

Research on opinion leadership continues to primarily focus on the information-giving behaviors or persuasive influence of opinion leaders as an individual characteristic (Flynn et al., 1996; Weimann, 1994). Research and measurement developed by Rogers (2003), Childers (1986), and Weimann (1991, 1994) have all focused on information-giving behavior and ability, or the individual personality strength of opinion leaders, rather than on information-seeking. A few others have looked at both aspects of opinion leadership, information-giving and information-seeking, as complementary behaviors that must be both explicated and measured together (Flynn et al., 1996; Schiffman, Dash, & Dillion, 1975; Shah & Scheufele, forthcoming). In these cases, opinion leaders were measured and defined not only by their propensity to influence others or provide information, but also by how much they seek information through interpersonal discussion or the mass media.

A third dimension of opinion leadership has been developed in recent years, one that moves the focus away from individual traits and characteristic to the social ‘embeddedness’ of opinion leaders (Weimann, 1994). In this conceptualization, opinion leaders are defined by their number of ties to different social networks or groups, or by the density of their social relations (Rao & Bhaskaran, 1978; Weimann, 1983, 1994). Some of this shift was driven by the research emphasizing the influence of social networks and bridging social ties in general, and in particular on political engagement (Granovetter, 1973, 1985; Huckfeldt & Sprague, 1988, 1995; Knoke, 1990a, 1990b).

Burt (1999) synthesized these three dimensions of opinion leadership—information-giving, information-seeking, and social relations—when he conceived of opinion leaders as information brokers. In this definition, opinion leaders are viewed as individuals who have a high density of social relations and act as exchangers of information through their information-giving and seeking behaviors across social contexts (rather than within), furthering the diffusion of information and ideas.

The Engagement Model of Opinion Leadership

The engagement model of opinion leadership developed by Roper ASW, as I have labeled the framework described by Keller and Berry (2003), incorporates several of the conceptions of opinion leadership and measurement discussed above. Primarily, the model focuses on identifying opinion leaders, or ‘influentials’ as Keller and Berry term them, through the number of political or civic activities in which they are actively engaged.

The assumption behind the development of the model was based upon research conducted by Roper in the 1950s and 1960s that demonstrated that ‘Americans who were more politically and socially active . . . appear to be the thought leaders on public affairs’ (Keller & Berry, 2003, p. 19).

Beginning in the early 1970s, Roper ASW developed a battery of questions in their annual surveys that asked respondents whether they had engaged in any of ten activities within the last 12 months. A respondent was identified as an influential if they answered affirmatively to engaging in at least three of the activities. To guard against social desirability and response bias, an eleventh dummy question asking respondents if they had signed a petition within the last twelve months was added to the battery (Keller & Berry, 2003). This addition reduced the probability of respondents answering false affirmatives to the other participation measures. Figure 1 provides the list of questions and activities developed by Roper ASW.

Questions on the scale primarily reflect political activities or behaviors, especially those that include some type of opinion expression. Employing this method, Roper ASW has identified an overall trend of 10 percent of the American population being characterized as opinion leaders since the 1970s, ranging between 8 and 14 percent on any one survey (Keller & Berry, 2003).

In regard to measurement, the engagement model defines opinion leadership as a combination of social embeddedness (or density) and persuasion (or information-giving). It incorporates the social relations of individuals across political activities by asking about membership or participation in differing types of groups or organizations. The model also addresses opinion-giving behavior by asking about persuasive activities such as contacting public officials, making speeches, or writing an article. From a measurement perspective, it also combines a sociometric technique of examining social networks with a self-report method that utilizes an objective criterion (three or more activities) for determining influential status.

However, the Roper ASW engagement measure has two major drawbacks. The first is that assessments of information-seeking behaviors or activities are missing from the overall measure. Secondly, and more importantly, the Roper ASW engagement measure relies on self-reported behavior and only indirectly measures opinion leadership or personal influence. The engagement measure does not incorporate any direct measures of actual personal influence or opinion leadership within social networks. However, Roper ASW did conduct a series of studies between 1978 and 2002 to evaluate the constructive validity of the model with other external criteria commonly associated with opinion leadership (Keller & Berry, 2003).

Cross-National Concepts of Opinion Leadership?

Do models of opinion leadership hold across national and cultural contexts? This question has increasingly become important, especially for global marketing and business (Cosmas & Sheth, 1980; Dawar et al., 1996; Marshall & Gitosudarmo, 1995). However, the scant literature on the question is mixed. Early on, Bostian (1970) argued that the ‘two-step flow’ theory of opinion leadership was limited in its international generalizability and could not be applied to developing countries, but rather only those with highly similar levels of media, economic, social, and political development. For example, lower media penetration would mean less reliance on the media as an instrumental source of information for opinion leaders and interpersonal discussion would play a greater role as both a persuasive tool and information-seeking behavior for opinion leaders.

More recently, Marshall and Gitosudarmo (1995) tested a scale of opinion leadership across eight different countries to examine commonality of individual characteristics. They found that opinion leaders did share traits across cultures, such as higher media exposure, interest, and sociability (i.e. being talkative, outgoing, and a good mixer), though other characteristics such as education or ambition were not shared. Marshall and Gitsodarmo also observed that some characteristics of opinion leaders are stable across contexts, though with some variability within each characteristic. Therefore, they concluded that a general conception of opinion leadership is valid and may be employed across cultures.

Dawar et al. (1996) also explored opinion leadership across cultural contexts. Following previous researchers, they conceived opinion leaders as both opinion-seekers and opinion-givers. Creating a multi-level model, they looked for the effects of differing types of cultural and individual characteristic on both these types of behaviors by opinion leaders. Their results demonstrated that behaviors associated with information seeking vary by cultural context, specifically the mix between personal (discussion) and impersonal (media) sources, but information providing does not. Furthermore, individual traits and behaviors were much more strongly associated with opinion leadership than any cultural influences.

A third study, conducted by Weimann (1991, 1994), tested a scale of personality strength to identify opinion leaders in both Germany and Israel. Conducting factor analysis and reliability tests, his findings demonstrated the ‘stability [of the construct] across social and cultural settings’ (Weimann, 1991, p. 271). Opinion leaders in both countries appeared to have the same general patterns of media use: higher levels of newspaper and magazine readership and lower levels of television use (Weimann, 1994).

Roper ASW has also attempted to apply their engagement model of opinion leadership outside the United States by testing it in a series of surveys across thirty different countries (Crispell, 2003). Roper ASW used a similar list of activities as their American measurement, though modifying it slightly to include opinion expression through economic behavior (i.e. making a complaint to a store or company), widening the types of organizations in which influentials may participate (i.e. religious, cultural, etc.), and reducing the list to nine activities instead of ten. Roper ASW’s initial results suggest that the model was applicable in an international context, though there was some variation in the number of activities which indicated opinion leadership, with the number of activities ranging from two to four across countries. Roper ASW found that opinion leaders identified by their measure shared broad characteristics across national contexts: They were less likely to hold materialist/hedonistic values, watched less entertainment television, read more, and were more likely to engage in social behaviors (Crispell, 2003).

Although all these findings appear to suggest that models of opinion leadership may be successfully employed across cultural or national contexts, other comparative scholarship warns against this type conclusion. Pierce and Pride (1972) assert that one of the common mistakes in cross-national micro-analyses of political and civic engagement is not taking into account differences in political, social, and economic structures. They note, ‘people participate in an environment composed of structural and cultural limitations on their behavior . . . differences [between states] in the formal and informal political structures . . . constrain and/or stimulate the participation of individuals’ (p. 20). Furthermore, they warn against applying ‘ideal-type’, or universal, conceptual models and measurement as some concepts or measures may be ‘culture-bound’ and not applicable to all societies.

In this study, even though I have selected a region that is similar to the United States in political, economic, and social development, as well as levels of social capital and engagement (Norris, 2002), one must be cognizant of variations in national and social contexts. Slomp (2000) notes that there are fundamental differences between American and European civil society due to more extensive European state services and higher levels of centralized organization in European institutions. Within Europe as well, there are variations in the nature and strength of civil society across the various sub-regions (Norris, 2002; Slomp, 2000). Furthermore, though Europe has moved toward higher levels of political integration than any other region, there is still significant variation in political cultures between individual countries (Eatwell, 1997). Therefore, though previous research may lead to an expectation that a model of opinion leadership may transcend borders, other research warns against taking this fact for granted.

PAST RESEARCH ON CHARACTERISTICS AND BEHAVIORS OF OPINION LEADERS

What are the major characteristics and behaviors of opinion leaders upon which to test the constructive validity of the engagement model? Previous research has identified five major categories of indicators associated with opinion leadership: demographics, information-seeking behaviors, social gregariousness, efficacy and trust, and values and satisfaction.

Demographics

In their study of influentials, Keller and Berry (2003) have found higher levels of education to be the most definitive demographic characteristic of opinion leaders. According to their collected data, influentials are twice as likely to have graduated from college and three times as likely to have done post-graduate work as the general population. Berelson and Steiner (1964) and Shah and Scheufele (forthcoming) also found similar relationship between opinion leadership and education. Otherwise, no stable demographic factors have been found to be significantly linked to opinion leadership, especially since the social, cultural, and issue spheres of opinion leadership can vary so widely (Weimann, 1994). However, Keller and Berry (2003) did note that influentials identified by the engagement model tended to be somewhat older than the general population.

Information-Seeking

As discussed above, opinion leaders are not only defined as information providers, they are also information seekers. Early work focused on interpersonal discussion as the primary source of information for opinion leaders, though this emphasis was most likely the result of the limited media density (prior to television, Internet, etc.) of the time (Lazarsfeld et al., 1944; Weimann, 1994). Within the channel of interpersonal discussion, opinion leaders are more likely to both seek opinions from others, as well as provide opinions and recommendations to other people (Keller & Berry, 2003; Weimann, 1994). According to Keller and Berry (2003), influentials in the Roper studies are 40 percent more likely to turn to interpersonal discussion for information than non-leaders.

Beyond interpersonal discussion, media use and opinion leadership has been the focus of extensive research ever since Lazarsfeld et al.’s (1944) ‘two-step flow’ hypothesis. In follow-up research to these findings, Katz (1957) not only confirmed opinion leaders’ generally higher levels of media use and exposure, but added an additional dimension as well. He found that across spheres of influence, opinion leaders paid a higher level of attention to media and news specific to that sphere. In other words, political opinion leaders paid higher attention to political news than the general population, fashion leaders paid higher attention to fashion news, and so on.

More recent research on the relationship between opinion leadership has refined the understanding of this relationship. Surveying a breadth of research, Weimann (1994) notes that many studies have found that opinion leaders tend to use print media at much higher levels than the general population. Keller and Berry (2003), Shah and Scheufele (forthcoming), and Chan and Shekhar (1990) all found a similar relationship. According to Keller and Berry (2003), ‘influentials are, as a rule, big readers . . . reading is ingrained on their daily lives’ (p. 156). However, the converse is true for television. Weimann (1994), Keller and Berry (2003), and Levy (1978) found that opinion leaders tend to watch either the same amount or less television than the general population. Shah and Scheufele (forthcoming) also found less of a relationship between opinion leadership and attention to television news than print news.

The Internet has emerged as another form of media use that has a strong relationship with opinion leadership. Roper ASW has reported that while opinion leaders in their surveys make up 10 percent of the general population, they compose 15 percent of those who regularly use the Internet within the United States (Keller, 2003). Furthermore, according to Roper ASW, influentials in their study rate the Internet as the media on which they spend the most time during the week (Keller, 2003). Keller and Berry (2003) also note that influentials in their study turn to the Internet as a source of information 75 percent more than the general population. Shah and Scheufele (forthcoming) also determined a strong relationship between Internet use and opinion leadership, stronger than television or print media in fact. Conceptualizing opinion leaders also as information seekers, one can understand why newspapers and the Internet are their media of choice. The form, content, and attributes of each align with information seekers’ needs and motivations more so than other media (Chaffee & Kanihan, 1997; Eveland, 2003).

Social Gregariousness

As discussed above, one key aspect of opinion leaders is their level of social embeddedness (e.g. density of social relations) measured by either the number of social groups to which an opinion leader belongs or by the density of personal relationships. This, in part, has been attributed to opinion leaders possessing a very high level of ‘social gregariousness’ resulting in opinion leaders having ‘more social ties, more friends, and more social contacts’ than non-leaders (Booth & Babcuck, 1972; Weimann, 1994, p. 80).

Efficacy and Trust

Differing levels of efficacy, or self-confidence, and trust from those of the general population have both been noted as significant individual attributes of opinion leaders. Reexamining data collected by Bockman and Gayk (1977) on levels of political efficacy and trust among opinion leaders and non-leaders, Weimann (1994) noted that opinion leaders tended to have higher levels of efficacy but lower levels of political trust than non-leaders. However, in terms of social or personal trust, Shah and Scheufele (forthcoming) demonstrated a positive relationship with opinion leadership. Keller and Berry (2003) also note that, in their studies, influentials have much higher levels of self-confidence and self-reliance in a variety of spheres, including politics, than the general public. However, as compared to other social institutions, influentials generally have lower confidence in the political system. In general, higher levels of efficacy and personal trust have been associated with higher levels of political and civic engagement, as well as higher levels of public affairs media use (Berelson, Lazarfeld, & McPhee, 1954; McLeod et al., 1996; McLeod, Scheufele, & Moy, 1999; Moy & Scheufele, 2000; Putnam, 1995, 2000; Verba, Schlozman, & Brady, 1995).

Values and Satisfaction

Opinion leaders have been found to share certain value orientations. The influentials identified by Keller and Berry (2003) tended to be less concerned with goals of material gain and financial success, or less materialist or hedonistic, than the general public. Opinion leaders also tend to exhibit higher levels of social responsibility, political tolerance, civic-mindedness, environmental concern, and are less traditionalist, all values associated with post-materialist or ‘Aquarian’ value constructs (Adams, 2003; Inglehart, 1997; Keller & Berry, 2003; Weimann, 1994). In turn, post-materialism has also been shown to encourage higher levels of engagement while materialism has been shown to lessen participation (Sotirovic & McLeod, 2001).

Scheufele and Shah (2000) note that a range of research has linked feelings of satisfaction or personal contentment to higher levels of civic and political engagement (Inglehart, 1997; Ototo & Synder, 1995; Sullivan & Transue, 1999). Furthermore, they go on to demonstrate a significant relationship between personal satisfaction and opinion leadership (Scheufele & Shah, 2000).

Application of the Engagement Model: Hypotheses and Research Questions

As outlined at the outset of the paper, my goals are to assess the validity of the engagement model as applied to a European context and examine how communication behaviors associated with opinion leadership may vary across national context. Therefore, my hypotheses focus on testing the applicability of the model by examining the construct validity of key measures associated with opinion leadership reviewed above. Specific characteristics, communication behaviors, and socio-psychological traits should be significant predictors of opinion leaders in each country if the general model is valid across national contexts. Furthermore, my analysis will assess the patterns of significant influences and amount of explained variance the model demonstrates across each country. As my measure of opinion leadership, I shall employ a similar political engagement scale as Keller and Berry (2003) to create a dichotomous indicator of opinion leadership for each country included in the study.

Based upon the above review of previous research, I have selected eight key measures for stating hypotheses of construct validity: five communication behaviors (interpersonal discussion, TV exposure, exposure to TV political news, exposure to newspaper political news, and Internet use); a measure of social gregariousness, or how much the individual socializes, makes personal contact with others, and values the importance of personal relations; and lastly two socio-psychological traits, internal political efficacy and post-materialist values. My formal hypotheses for each country included in the study are:

H1 Higher general (non-news) television exposure will be negatively associated with the indicator of opinion leadership.

H2 Higher television exposure to political news will be positively associated with the indicator of opinion leadership.

H3 Higher newspaper exposure to political news will be positively associated with the indicator of opinion leadership.

H4 Higher Internet use will be positively associated with the indicator of opinion leadership.

H5 Higher levels of interpersonal political discussion will be positively associated with the indicator of opinion leadership.

H6 Higher levels of social gregariousness will be positively associated with the indicator of opinion leadership.

H7 Higher levels of internal efficacy will be positively associated with the indicator of opinion leadership.

H8 Higher levels of post-material/Aquarian values will be positively associated with the indicator of opinion leadership.

The literature review also demonstrated possible positive and negative relationships between other individual characteristics or behaviors and opinion leadership that were not included in the hypotheses above. Thus, in order to explore how other behaviors or traits of opinion leadership may vary across national or cultural contexts, I pose the following research question:

RQ1 What additional traits or behaviors are significant predictors associated with the indicator of opinion leadership within each national context?

METHOD

For my analysis I use data from the 2003 European Social Survey (the ESS), a biennial multi-country survey covering over 22 nations within Europe. The fieldwork for the survey was conducted in late 2002 and early 2003. The survey was conducted through face-to-face interviews of approximately one hour in length. Within each country, strict random probability sampling and rigorous translation protocols were observed, though achieved response rates varied by country.¹ For this paper, fifteen countries that participated in the first round of the ESS and had complete datasets available at the time of the study were selected: United Kingdom, Ireland, Belgium, the Netherlands, Germany, Denmark, Portugal, Spain, Hungary, Greece, Slovenia, Poland, Sweden, Norway, and Finland. The design weight created by the administrators of the ESS was used for all country analyses. For complete information on the ESS questionnaire, survey methodology, response rates, and results see European Social Survey (n.d.).

Since the dependent variable (whether an individual is an influential or not using the engagement model measure of political engagement) is a dichotomous categorical variable, the analyses are based on a hierarchical logistic likelihood regression model for each country. In these models, blocks of variables are entered according to their assumed causal order. In other words, exogenous variables, such as demographics, are entered first, followed by other antecedent endogenous variables (Cohen & Cohen, 1986).²

Demographics

The demographic variables within my model include a set of demographic characteristics most associated with opinion leadership: gender, age, education, and employment. Gender is dummy coded with women coded high (51.7 percent).³Age was measured as a continuous variable with a respondent range of 15–103 years of age (M = 46.5, SD = 18.1). Education was measured by a 6-point scale, with no completed primary education the lowest value and a completed second stage of tertiary education as the highest value (M = 2.8, SD = 1.5). Whether the respondent was employed was also coded as a dummy variable, with employment coded high (52.8 percent).

Communication Behaviors

I included five measures of communication behaviors in the model. There were three measures of media exposure. General (non-news) television exposure was calculated by subtracting how much time respondents spent watching news or current affairs programming from their total time spent watching television on an average weekday. The 7-point scale ranged from 0 to more than 3 hours in one-half hour increments (M = 2.29, SD = 1.19). Exposure to TV news measured how much time respondents spent watching news, politics, and current affairs programming on an average weekday using a 7-point scale ranging from 0 to more than 3 hours (M = 2.0, SD = 1.3). Exposure to newspaper news was measured similarly to television news, asking respondents how often they read political or current affairs news on an average weekday on the same time scale (M = .91, SD = .94). Internet use asked respondents how often they personally used the Internet, email or world wide web, using a 6-point scale from ‘never/no access’ to ‘every day’ (M = 2.3, SD = 2.9). Finally, Political discussion was a single measure asking respondents how often they discussed political topics or current affairs on a 7-point scale of ‘never’ to ‘every day’ (M = 4.1, SD = 2).

Social Gregariousness

The measure for Social gregariousness was a composite scale of three different measures. The first asked the respondent ‘how often they socially meet with friends, relatives, or colleagues’ on a 7-point scale of ‘never’ to ‘every day’ (M = 5.0, SD = 1.6). The second measure asked respondents their perception of how often they participated in social activities compared to others of their same age on a 5-point scale ranging from ‘much less’ to ‘much more’ (M = 2.7, SD = 0.93). The last measure asked the respondent how important friends were in their life on a 10-point scale from ‘extremely unimportant’ to ‘extremely important’ (M = 8.48, SD = 1.7). Each score was standardized and added together to form the complete scale (α = .52).

Efficacy and Trust

Four measures were included in this category, two measures of internal and external efficacy and two measures of political and personal trust. Internal efficacy was an additive index of three 5-point scales measuring how complicated respondents’ found politics, how difficult it was for them to make political decisions, and if they believed they were capable of taking an active role in a political group (M = 8.1, SD = 2.7, α = .63). External efficacy was an additive index of two 5-point scales measuring respondents’ perceptions toward political and governmental figures, specifically if they cared about what people like the respondent think and if they care more for votes than for people (M = 4.6, SD = 1.9, r = .62).

Personal trust was measured by an additive index of two 10-point scales, one asking if most people can be trusted and the other if most people try to be fair (M = 10.7, SD = 4.4, r = .60). Political trust was an additive index of four 10-point measures of trust in various national political institutions, including the parliament, the legal system, the police, and the politicians (M = 20.1, SD = 8.0, α = .83).

Individual Values

Two measures of individual values were included within the model, materialist values and post-material values. These measures were constructed by utilizing value statements that respondents were asked to rate on a 6-point scale. Materialist values were measured through an additive index of six statements associated with material gain, success, ambition, and narcissism. They were ‘important to be rich, have money and expensive things,’ ‘important to have a good time,’ ‘important to be successful and that people recognize achievements,’ ‘important to seek adventures and have an exiting life,’ ‘important to seek fun and things that give pleasure,’ ‘important to get respect from others,’ (M = 21.4, SD = 5.5, α = .75).

Post-materialist values were measured similarly through an index of seven measures associated with aspects of post-materialism such as personal liberties, political tolerance, and the environment, such as: ‘important that people are treated equally and have equal opportunities,’ ‘important to make own decisions and be free,’ ‘important to be loyal to friends and devote yourself to people close,’ ‘important to care for nature and environment,’ ‘important to understand different people,’ ‘important to help people and care for others well-being,’ and ‘important to think new ideas and being creative’ (M = 32.6, SD = 5.0, α = .76).

Personal Satisfaction

A measure of personal happiness and satisfaction was included in the model by combining two closely related measures. The first question asked respondents ‘how happy are you’ on a 10-point scale where 1 was ‘extremely unhappy’ and 10 was ‘extremely happy’ (M = 7.4, SD = 1.9). The second question asked to respondents was ‘how satisfied with life as a whole are you’ and also used a 10-point scale with 1 as ‘extremely unsatisfied’ and 10 as ‘extremely satisfied’. Both variables were standardized and added together (r = .68).

Criterion Variable

As previously mentioned, the criterion measure was constructed by utilizing an engagement model of opinion leadership. This model uses the number of political or civic activities in which the respondent has actively participated within the last 12 months as an indicator of opinion leadership. Using the Roper ASW list of activities as a model, nine similar activities were selected from the European Social Survey to create a measure of engagement. Respondents were asked to indicate whether they had actively participated in any of the following nine activities in the last 12 months:⁴ active member of political party, displayed campaign badge/sticker, donated money to a candidate or political organization, contacted a politician or government official, worked in a political party or action group, worked in another political organization or action group, had either boycotted a product or bought a product for political or social reasons, took part in a lawful demonstration, and volunteered time for a civic or community group⁵ (M = 1.1, SD = 1.4, α = .68). Employing this scale, a dichotomous criterion measure of opinion leadership was created for each country included in the study, depending upon the number of activities in which each respondent engaged.

The number of activities that indicated opinion leadership varied by country depending on the overall level of civic or political activity within that country’s population. In considering the number of indicator activities for each country, I referred to Keller and Berry’s (2003) report that influentials ranged from 8 to 14 percent for each of their surveys since the 1970s, with an overall trendline of 10 percent within the United States, as well as Roper ASW’s latest research on global influentials referenced above (Crispell, 2003). Thus, the number of selected activities for each country was based on this principle and attempted to replicate a percentage of identified opinion leaders within each country closest to this range. Table 1 provides an overview of the number of activities needed to qualify as an opinion leader, the percentage of each country’s sample that qualified, the total size of each country’s sample, the mean number of activities for each country and corresponding standard deviation.

RESULTS

As previously mentioned, to test the hypotheses, hierarchical logistic regression was used with the standardized coefficients for each dependent variable and incremental and total Nagelkerke variance reported for each country in the study. Table 2 provides the results for each country included in the analysis.

Results for Primary Hypotheses

The first hypothesis regarding the negative association of general television exposure was supported in eight different countries, specifically the United Kingdom, Ireland, the Netherlands, Denmark, Germany, Spain, Greece, and Finland. The relative strength and significance of television exposure as compared to other predictors was greatest in the United Kingdom, Germany, and the Netherlands.

The second hypothesis that theorized a positive relationship between television news exposure and the indicator of opinion leadership was only supported in one country, Belgium. Counterintuitively, higher levels of television news exposure were negatively associated in five countries, specifically the United Kingdom, the Netherlands, Portugal, Greece, and Slovenia. This finding may be a result of the time-spent approach to measuring television news exposure employed by the ESS and is addressed in further detail below in the discussion of limitations.

In contrast to television news exposure, the third hypothesis regarding newspaper news exposure was supported in seven countries, with newspaper news exposure positively related to opinion leadership in Belgium, the Netherlands, Slovenia, Hungary, and strongest in Germany, Portugal, and Poland. Higher Internet use, the fourth hypothesis, was the most widely significant media behavior associated with the indicator of opinion leadership. Internet use was a significant predictor in eight countries, specifically the United Kingdom, the Netherlands, Denmark, Germany, Spain, Portugal, Finland, and Sweden. Internet use had the largest coefficients and greatest significance in Denmark, the Netherlands, and Germany.

Hypotheses 5, 6, and 7 were the most widely accepted of any of the hypotheses, with interpersonal political discussion and internal efficacy positively associated with the indicator of opinion leadership in every country included in the study. Social gregariousness was also positively associated in all but three countries, Poland, Ireland, and Portugal. The eighth hypothesis, that post-material values would be positively associated with the indicator of opinion leadership, was supported in 10 countries. It was absent in Portugal, Slovenia, Hungary, Finland, and Norway.

Research Questions

Beyond the primary attributes and behaviors of opinion leaders cited in my hypotheses, I asked a research question regarding what other attributes may be associated with the indicator of opinion leadership across national contexts. With regards to demographic characteristics, education was significantly associated with identified opinion leaders in seven countries, specifically the United Kingdom, the Netherlands, Spain, Slovenia, Hungary, Poland, and Norway. Education as a predictor appears to play the strongest role in Poland and Hungary, both in terms of the size of its standardized coefficient and amount of incremental variance which it explains. Besides education, age was significantly associated with opinion leadership across eight different countries, though its direction of association was mixed. In Spain and Greece, age was negatively associated with the indicator of opinion leadership. Conversely, in the United Kingdom, the Netherlands, Denmark, Finland, Norway, and Sweden, age was positively associated. This difference may have been the result of a curvilinear relationship between age and participation, and therefore test regressions for all countries in the study were conducted using age-squared with no change in results.

External political efficacy was also significantly associated with the indicator of opinion leadership in eight countries, most notably Sweden, Germany, Denmark, and Portugal. Confirming previous research, political trust was negatively associated with the indicator of opinion leadership in four countries, Greece, Hungary, Poland, and Sweden. Personal trust was only a significant positive predictor in one country, Poland. Materialist values were negatively associated with the indicator in three countries, the United Kingdom, Ireland, and especially the Netherlands. Lastly, personal satisfaction was found to be negatively associated with opinion leadership in Ireland, Slovenia, and especially Spain.

Model Fit and Constructive Validity Results

How well does the model fit across countries, and what are the overall results of the constructive validity test? Table 3 provides a summary of supported hypotheses and total explained variance for each country in the study. From these results, a pattern emerges where the model fits best (in terms of supported hypotheses and explained variance) in the United Kingdom and western European nations of Belgium, the Netherlands, Denmark, and Germany. Outside of this region, only Spain had a noteworthy number of confirmed hypotheses and amount of explained variance. Aside from predictors common to all 15 countries —or at least the majority—(political discussion, social gregariousness, and internal efficacy), general television exposure, Internet use, and post-material values were the significant predictors most commonly associated with the indicator of opinion leadership within these six countries. Although, even in these six countries, there were notable variations between countries and with traits and behaviors of opinion leaders found in the United States, especially in education, television news exposure, external political efficacy, and personal satisfaction. The implications of these findings for the validity of the engagement model within Europe will be addressed further in the discussion.

Limitations

There were some limitations to this study in the area of measurement. First, the nature of the media exposure measures included in the survey may have contributed to some of the contradictory findings and surprising lack of significant relationships associated with television news and newspaper news exposure. Specifically, the newspaper and television measures of exposure were constructed to measure time spent, rather than frequency of use or relative amount of attention. A time-spent approach may not sufficiently measure either the frequency of use or level of attention to news content, which as noted above, previous research has found to be much higher for opinion leaders. In other words, actively engaged influentials may spend less time on the average weekday reading the newspaper or watching television news, but pay closer cognitive attention (not necessarily measured as time) to news and current affairs on the television or within the paper. Previous research has suggested that opinion leaders process media content and information differently than non-leaders, with simple measures of time-spent not capturing qualitative differences in media use between leaders and non-leaders (Levy, 1978). Other communication research has confirmed the superiority of attention measures for television or newspaper news rather than simple time-spent measuring for assessing media use impact on political knowledge and political behavior (Chaffee & Schleuder, 1986). In this light, the results of this study regarding television news exposure, and to a lesser degree newspaper exposure, are consistent with these previous assertions that highly politically engaged individuals use the media in a qualitatively different manner, with greater gains, than can be explained by merely the degree of media exposure.

The measure of Internet use also had some limitations, with the maximum value on the use scale equal to ‘every day’, rather than extending the scale to ‘several times a day’. Considering that opinion leaders within the United States have been characterized as ‘web junkies’ and rate the Internet as the medium on which they spend the most time during the week (Roper ASW, 2003), this measure may underestimate the relationship of Internet use. Though structural factors, such as limited Internet penetration and level of economic development, may also explain the lack of Internet use as a significant predictor in several countries.

A third limitation of the study is that the measure of political engagement is only an approximation of the scale used by Roper ASW, though many of the items were identical and represented the major dimensions encompassed in the Roper ASW list of political activities, including the dimension of economic behavior employed in Roper’s international surveys. Nevertheless, there may be some variation in the measurement or percentage of the population that would be identified as opinion leaders using the ESS scale as opposed to the Roper ASW scale.

CONCLUSION

At the outset of this paper, I stated my two goals were to (a) evaluate the validity of the engagement model within an European context and (b) explore how information-seeking behaviors of opinion leaders identified by the engagement measure varied across national contexts. With regards to the latter, previous cross-national or cross-cultural research has shown that media information-seeking behaviors of opinion leaders are the most culturally contingent. The results of this study appear to confirm these findings, with wide variation in media behaviors across countries, the aforementioned measurement issues notwithstanding. Less non-news television exposure and higher Internet use were the most widely associated behaviors, yet they were only significant in slightly more than half the countries in the study. In addition, exposure to newspaper news was also only significant in seven countries.

Though there was significant variation, some patterns did emerge that may be related to differing levels of information infrastructure between each country. For example, one pattern that emerged was that Internet use was significant in each country of western Europe, the United Kingdom, Finland, and Sweden. According to indicators collected by the World Bank Institute (2004), these nations within Europe have the highest levels of Internet infrastructure measured in terms of online users, PCs per capita, and per capita spending on information infrastructure. Conversely, the majority of countries in which newspaper use was significant have overall low levels of information infrastructure (i.e. Slovenia, Hungary, Poland, and Portugal). Internet use, as may be expected, was not widely significant in these countries as well.

Thus, these findings build upon Bostian’s (1970) assertion that opinion leaders’ use of media in developing countries would be dependent on the level of media infrastructure. In this case, we also see how the nature of the national information infrastructure may channel individual information-seeking behavior in developed nations. Opinion leaders, as active information seekers, will primarily turn to the interactive form and content of the Internet and newspapers, over television, which allow the greatest utility for active information seekers with high motivation (Chaffee & Kanihan, 1997; Eveland, 2003). However, which specific information source of the two (Internet or newspaper) opinion leaders turn to depends on the availability and level of penetration of each media, with the Internet primarily favored over newspaper if available. However, other structural or cultural factors may also play a role. For example, Slovenia has a higher percentage of its population online and more PCs per capita than either Spain or Portugal, yet Internet use is not significant in Slovenia but is significant within the other two nations.

In contrast to media behaviors, interpersonal political discussion, the key communicative mechanism of opinion leadership that embodies both information-seeking and information-giving behaviors, was universally predictive across all countries in the study. However, the relative strength, significance, and amount of explained variance for political discussion did vary greatly within each national context. Thus, though interpersonal discussion may be seen as the least contingent communicative activity of highly engaged individuals, its strength and significance may also be moderated by structural or cultural factors as noted by Bostian (1970).

Collectively, what do the results of this study tell us about the ecological validity of the engagement model of opinion leadership outside the United States and its use as a measurement tool? The results discussed above suggest that the model is most valid within the United Kingdom and nations in the western portion of Europe. This region shares the most structural and cultural similarities with the United States in terms of economic development, media infrastructure, shared cultural and historical traits, and political participation. Table 1 shows a clear consistency in the average level of participation within these nations as opposed to other regions which either had lower (eastern and southern Europe) or higher (northern Europe) mean rates of participation. This suggests that the engagement model is ecologically valid within a narrow band of nations that share the same patterns of political participation and engagement as the United States. Even though the critical number of activities used to indicate influential status was adjusted, the validity of the model in national contexts where either the level of political engagement was low, or was qualitatively different from the types of engagement found in the United States or western Europe (and thus not captured in the measures used), is questionable. The same is true for regions such as northern Europe where the general level of participation is high. The results suggest that opinion leaders may not be identified or distinguished as well in such societies through measuring overall political engagement.

Do these results then suggest that a general model of opinion leadership or personal influence cannot be employed across national contexts? Key demographic characteristics often associated with opinion leaders (education and age) and socio-psychological traits (personal trust, political trust, and external political efficacy) did vary considerably across nations. However, other important characteristics and behaviors commonly associated with opinion leadership and personal influence did not, namely interpersonal discussion, social gregariousness, and internal political efficacy. Also consistent across nations were higher levels of information-seeking behavior using some form of media, though the type of media used varied across countries depending on structural factors as mentioned above.

Thus, turning back to the earlier discussion of previous conceptualizations and measurements of opinion leadership and considering these commonalities, some lessons may be developed for future research and measurement of political opinion leadership and personal political influence across national contexts. First, political activity measures appear to be the most structurally or culturally contingent as exhibited by the variability in the levels of national engagement shown in Table 1. Measuring the sociometric status through degree of group membership may also be problematic depending on the level of civic engagement in each society and the qualitatively different types of organizations and groups across nations. Alternatively, combining measures of personality strength and personal efficacy, for example the one originally developed by Noelle-Neumann (1985) and further developed by Weimann (1994), with measures of interpersonal discussion and social gregariousness, may result in a measurable construct of opinion leadership that may be valid across different countries. Media information-seeking behaviors may also be used, but they may be contingent upon levels of media infrastructure and economic development, even in highly developed regions such as Europe. This may be addressed by first examining the types of media available within each national context and identifying which media type has content and form that most aligns with the motivations and needs of high information seekers. In other words, in nations where the Internet is readily available, opinion leaders will most likely turn to the Internet as a primary information source. In other societies, it may be newspapers, or even possibly radio in less developed nations. Other structural factors, such as education and literacy may also influence information-seeking behaviors.

Continuing to explicate the concept of opinion leadership and personal influence within the United States and other nations is important in comprehending comparative mechanisms of opinion formation and political behavior. Increasing globalization and awareness that opinion formation across nations, such as the case of anti-Americanism in the Muslim world, have wide-ranging implications and reinforce the need for better understanding of how opinion leadership may be measured across national or political contexts. In this regard, this paper has provided some important qualifications on applying models of opinion leadership derived in the United States to other global regions. However, it has also provided some direction on how future research may develop models of opinion leadership that are more generally applicable across borders. Furthermore, this study has primarily cited how structural differences, such as level of media infrastructure and media availability, may interact with individual information-seeking needs to determine mass media use. Future scholarship should also examine cultural or social factors, for example political culture (Swidler, 1986), that may influence patterns of information-giving and information-seeking behavior within different national and political contexts.

The author would like to thank Prof. James Shanahan of Cornell University, Prof. Matthew Nisbet of Ohio State University, and reviewers for their comments on earlier drafts of this paper.

REFERENCES