opinion leaders - do they know more than others about their area of interest

Upload: liebe2772

Post on 02-Jun-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    1/22

    Opinion leaders Do they know more than othersabout their area of interest?

    SABINE TREPTE and HELMUT SCHERERE-Mail: [email protected]; [email protected]

    AbstractThe knowledge of opinion leaders (in their area of interest) to a certainextent has always been taken for granted by communication scholars. Thisarticle investigates what opinion leaders really know. Two studies will be

    presented to answer the above question. Participants (N 119) of the firststudy were assessed according to ratings on three scales of opinion leader-ship (Katz and Lazarsfeld, 1955; Troldahl and van Dam, 1965; Childers,1986), personality strength (Noelle-Neumann, 1983) and political knowl-edge. In the second study, respondents (N 727) were assessed according

    to ratings of opinion leadership (Childers, 1986) and political knowledge.In both studies, it was found that opinion leaders can be divided into in-

    formed opinion leaders, who know a lot in their area of interest, and unin-formed opinion leaders, who are ill-informed about the field they claim astheirs. In both studies informed opinion leaders read newspapers approxi-mately one hour longer per week than uninformed opinion leaders.

    Keywords: opinion leaders, followers, political knowledge, information,Multi-Step-Flow

    Introduction

    The discovery of the concept of opinion-leadership is one of the meritsof The Peoples Choice a study that was conducted by Lazarsfeld,Berelson and Gaudet in 1944. Opinion leaders are termed as people whoare more influential within their social networks than others. They con-sider themselves experts in a specific area of interest (e. g. home policyor fashion) and are asked for advice in this area (Katz and Lazarsfeld,1955). Opinion leaders select information in these areas and then pass it

    on to others. In the process of reporting to others they more or lessconsciously modify the information they transmit. Katz and Lazarsfeld

    Communications 35 (2010), 119140 03412059/2010/0350119DOI 10.1515/COMM.2010.007 Walter de Gruyter

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    2/22

    120 Sabine Trepte and Helmut Scherer

    (1955) summarize that ideas often flow from radio and print to theopinion leaders and from them to the less active sections of the popula-tion (p. 309). Research on opinion leadership and the flow of informa-tion has focused on the questions of how information is transmitted insocial networks (Lenart, 1994; Schenk, 1985), who is seeking advice andwho is giving advice (Katz and Lazarsfeld, 1955), and what role themedia plays in these processes (Levy, 1978).

    Initial ideas on opinion leadership seem to suggest that opinion leadersare particularly skilled, and that they have more knowledge and infor-mation than their followers. However, the opinion leaders ability toselect, to process and to retain information has been widely neglected.There are first studies indicating that not all opinion leaders are knowl-edgeable (Trepte and Bcking, 2009), but today, after 65 years of re-

    search on opinion leadership, important questions are still unsettled.There might be different types of opinion leaders: those, who know alot, influence others and are asked for advice; and opinion leaders withcomparatively low levels of information, but good communicative skillsto compensate for their lack of knowledge. In the following, a very muchneglected criterion of opinion leadership will be addressed: knowledge.We will explore, whether there are different types of opinion leaders informed and uninformed leaders and how they might differ in terms

    of media use.

    Opinion leadership and the flow of information

    The term opinion leader was developed by Lazarsfeld, Berelson andGaudet (1944). In their study, the authors first suggested that opinions

    are influenced by the media. Referring to the stimulus-response para-digm, they hypothesized a causal influence from the media to the general

    public. Contrary to previous research, Lazarsfeld, Berelson and Gaudet

    (1944) discovered that interpersonal communication was much more in-

    fluential than direct media effects were. Respondents of the study namedfriends and affiliates as sources of information more often than newspa-pers or the radio. In the course of their studies, the authors identified

    people who had a particular influence on others. These so-called opinionleaders were characterized as selecting and transmitting information on

    the election as well as modifying and facilitating it. Rather than being

    neutral reporters, opinion leaders were passing on useful items of infor-mation. Also, they used the media more often and had a greater interest

    in the election and the campaigns than others. On the basis of their

    findings, Lazarsfeld, Berelson and Gaudet (1944) developed the hypothe-sis of the two-step-flow of communication, which implies that informa-

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    3/22

    Opinion leaders 121

    tion flows from the media to opinion leaders and from them to lessactive sections of the population.

    However, the results of the first studies were solely based on a self-report measure questionnaire. For that reason Merton (1949) chose an-other method to identify opinion leaders. In his study, he asked subjectsabout the person that gave them advice. Katz and Lazarsfeld (1955)linked both, the self-report measure and the third-party ratings, in theirDecatur-Study. They conducted follow-up interviews with the peoplewho had been given advice. The authors showed that on average 70percent of the people who had taken advice remembered the person andeven the information that was given. Complete refutation of the desig-nating respondents claim was found in only ten percent of the personsasked. Katz and Lazarsfeld (1955) suggested that some information was

    transmitted directly by the mass media to all people and that other infor-mation was passed on by opinion leaders. Based on these findings themodel of multi-step flow was established. Personal influence was said toreinforce information given by the mass media on the one hand and tocomplement it by providing people with in-depth and special-interestinformation on the other. Further, Katz and Lazarsfeld (1955) founddifferent types of opinion leaders some who select and transmit infor-mation early after its occurrence, and others who receive informationlater and refer to opinion leaders themselves.

    Subsequent research has criticized the separation of opinion leadersand followers and has suggested distinguishing between those opinionleaders who give advice (opinion givers) and those who both give andreceive advice (opinion sharers) (Robinson, 1976). Consequently, the fol-lowers were distinguished into opinion seekers and non-discussants ac-cording to their habit of receiving information from others or not.

    The main concerns in the research on opinion leadership were relatedto the opinion leaders media use and their communicative habits. How-ever, only few studies apply methods such as network analysis, observa-tional studies or experiments (OKeefe, 1982). For instance, Schenk

    (1994) did research on the relationship between the media, opinion lead-ers and followers. Using the snow-ball technique, he identified the peopleopinion leaders usually talked to and interviewed them as well. Lenart(1994) applied different experiments and focused on person-to-personinfluence, group level influences and opinion climate influences. The re-sults of both studies show that the coherence between personal relation-ships and mass media are moderated by political socialization, personal-ity, the information that is processed, and properties of the social grouppeople belong to.

    Research on opinion leadership has often been criticized, because in-formation and influence are not differentiated (Bostian, 1970; OKeefe,

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    4/22

    122 Sabine Trepte and Helmut Scherer

    1982; Robinson, 1976). Many studies concentrate on the flow of infor-mation but neglect the question ofwhat exactly is flowing. Particularly,the question of knowledge has not been addressed. Only Troldahl (1966)suggests a two-cycles model. He assumes that an opinion leader needsthe head start of having knowledge to pass on and a certain personalityto exert influence on others.

    Opinion leadership has been applied to all kinds of different areas ofinterest, e. g. public affairs (Katz and Lazarsfeld, 1955), farming pro-ducts (Rogers and Cartano, 1962), consumer products such as packagedfood, household cleansers, detergents, fashion, cosmetics, personal groom-ing aids, drugs, pharmaceutical products (Katz and Lazarsfeld, 1955;King and Summers, 1970), jeans, professional clothing (Flynn, Gold-smith and Eastman, 1994), rock music, and records (Goldsmith and

    Hofacker, 1991). Many areas have been investigated, because scholarswanted to find out, whether opinion leadership refers to one area ofinterest only (monomorphic) or more (polymorphic) (Marcus and Bauer,1964). Katz and Lazarsfeld (1955) had already stated that opinion lead-ers for public affairs, fashion, and marketing issues have to be distin-guished, but that there is a certain probability of finding an opinionleader for two areas of interest. Most scholars proclaim that opinionleaders are monomorphic in terms of broader areas of interest (Gold-smith and Hofacker, 1991; Rogers and Cartano, 1962), but that there

    are intersections in their knowledge with regard to similar fields of inter-est (King and Summers, 1970).Opinion leadership has been discussed within the context of a number

    of correlates to prove the scales validity and to analyze its realm (Katzand Lazarsfeld, 1955; Marcus and Bauer, 1964). Variables that have fre-quently been associated with opinion leadership are the opinion leadersareas of interest, their group behavior, and demographics.

    Opinion leaders group behavior and participation is an importantvariable, as it reflects the definition of opinion leaders as communicatingabout certain issues earlier and more frequently than others. Concordant

    with the multi-step flow model, correlations show that opinion leadersare more likely to be gregarious (Katz and Lazarsfeld, 1955), to partici-pate in political organizations, and to have peers and affiliates belongingto heterogeneous groups (Rogers and Cartano, 1962; Weimann, 1994).This has similarly been found for personality strength (see section Re-view of scales to measure opinion leadership), and it has been shown,that people with high personality strength are affiliated to larger socialnetworks and talk to others more often (Schenk and Rssler, 1997).

    According to demographics such as social background, gender, and

    even age, no major differences between opinion leaders and followerscould be found (Hamilton, 1971; Myers and Robertson, 1972). Only in

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    5/22

    Opinion leaders 123

    the early studies, status was shown to be different between opinion lead-ers and followers (Katz and Lazarsfeld, 1955; Troldahl and van Dam,1965).

    Review of scales to measure opinion leadership

    In this article and in other work on the subject matter (Trepte andBcking, 2009) it is suggested that knowledge might be a neglected vari-able in research on opinion leadership. To elaborate on this, it is crucialto understand how opinion leadership is operationalized, if and howthese measures refer to knowledge. The opinion leaders knowledge ispresent in theoretical definitions, but is it present in operationalizations?Four different ways of measuring opinion leadership have been used in

    previous research: sociometric techniques, interviews with key infor-mants, observation, and self-designating techniques (Jacoby, 1974; Rog-ers and Cartano, 1962; Weimann, 1994). The most prominent measure-ments have always been self-designating techniques, because they caneasily be administered in surveys.

    The first opinion leadership scale was introduced in 1944. Based ontwo items (see Appendix), Lazarsfeld, Berelson and Gaudet (1944) andKatz and Lazarsfeld (1955) computed an index. They distinguished be-tween opinion leaders and followers. This first idea of operationalizingopinion leaders as convincing others and being asked for advice has hadtremendous impact on follow-up studies and scale development. How-ever, the first attempt to measure opinion leadership has been criticizedthroughout the field, because of statistical problems and low instrumen-tal validity (Hamilton, 1971). Since Katz and Lazarsfeld (1955), sevenscales have been developed that are built on and refer to preceding ver-sions (chronological: Rogers and Cartano, 1962; Troldahl and van Dam,1965; King and Summers, 1970; Levy, 1978; Childers, 1986; Flynn,Goldsmith and Eastman, 1994). The most important scales, in terms ofhow often they have been applied and how well they were able to en-

    hance the preceding scales, will be outlined in brief.Rogers and Cartano (1962) were inspired by the idea of Katz and

    Lazarsfeld (1955), but they criticized the simplicity of defining the con-cept of opinion leadership using just two items. Also, they assumed thatopinion leadership is a continuous variable rather than a dichotomy ofleaders and followers. Consequently, the authors altered and expandedthe two-item scale and introduced a scale that consisted of six items.Similar to Katz and Lazarsfelds (1955) scale, the six items deal withtwo components of opinion leadership: the respondents self-image as an

    opinion leader and the respondents perception of past behavior wheninteracting with others. The scale yielded a split-half reliability ofr .76

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    6/22

    124 Sabine Trepte and Helmut Scherer

    in contrast to the former two-item scale, which yielded a split-half reli-ability of r .48 (Rogers and Cartano, 1962). Also, validity measuresproved that the six-item scale significantly correlated with composite keyinformants ratings (r .64) and sociometric choices (r .30).

    The next step in scale development was taken by Troldahl and vanDam (1965). They criticized that all former scales were measuring per-ceived opinion leadership. To score high on the scales, a respondentwould have to see himself as influential, whether or not she or he actu-ally is. For that reason, authors introduced measures that were closer tothe idea of perceived opinion leadership, such as: Would you like tobe thought of as a person who others depend upon in making up theirminds about major issues in the news. Additionally, they suggested ask-ing respondents, if they could recall any topics that had been getting a

    lot of attention in the news lately (see Appendix for items). This way,the authors were able to prove that opinion leaders were up to date oncertain issues. If respondents said yes, they were asked another eightquestions to check the validity of their answers. Afterwards, respondentswere asked the traditional opinion leadership items. These were similarto the scale of Rogers and Cartano (1962).

    King and Summers (1970) revised Rogers and Cartanos (1962) scaleas well and adapted the scale for marketing issues. King and Summers(1970) scale was used for a long time to measure opinion leadership and

    yielded satisfactory internal consistency reliability (Riecken and Yavas,1983). However, it was considered difficult to proceed in the field and indata-analysis, because it allowed different responses such as dichoto-mous and nominal scales.

    Childers (1986) revised the scale six years after it had been publishedwith the aim of simplifying the response format and checking validityand internal consistency. His modified scale has six items (see Appendix)and works with a Likert scale response format. Childers (1986) scaleyielded an internal consistency reliability of Cronbachs alpha .79 upto Cronbachs alpha .84 (Goldsmith and Desborde, 1991). It is the

    most recent scale in the tradition of Rogers and Cartano (1962).Similar to the concept of opinion leadership, Noelle-Neumann (1983)

    developed the scale Personality Strength to assess the trait of influenc-ing other people (see Appendix). The scale was established after a facto-rial reduction of a 34-item questionnaire. Similar to the opinion leaderscales, Noelle-Neumann (1983) included items asking for self-appraisalof giving advice and being a source of information. Unlike previousscales, her questionnaire contained items that touched upon the psycho-logical trait extraversion. Additionally, behavioral references are part

    of the scale, e. g. whether respondents participate in a political party orhold offices in a club or an organization. The different scales presented

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    7/22

    Opinion leaders 125

    above, address knowledge more or less explicitly. The scales will be inves-tigated in terms of how they relate to results in knowledge testing.

    RQ 1: Do opinion leadership scales differ in their predictive power toforecast knowledge in a certain area of interest?

    Knowledge and opinion leadership

    Katz and Lazarsfeld (1955) showed that opinion leaders tend to knowmore about public affairs: In their Decatur study, 46 percent of the peo-ple with a high information level were considered opinion leaders, 35percent with a medium information level, and only 3 percent with alow information level (p. 275). Also, Troldahl and van Dam (1965)defined knowledge as a presupposition for opinion leadership. Beforeletting respondents answer questions on their advice giving to other peo-ple, respondents were asked to recall topics that had been getting a lotof media coverage lately. Also, the authors showed that the nationalnews information level and opinion leadership are correlated (r .18).Similarly, Schenk (1985) observed that opinion leaders are able to referto political issues that are not currently being reported in the mass me-dia. Schenk and Rssler (1997) showed a positive relationship betweenpersonality strength and the outcome of a four-item scale on political

    knowledge (r .28). Additionally, Myers and Robertson (1972) ad-dressed the issue of opinion leaders knowledge using self-assessments(How much do you feel you know about each topic or area, in compari-son to your friends and relatives?, p. 42) and consequently yielded cor-relations between r .52 and r .81.

    The empirical association between opinion leadership and knowledgehas been found to be amazingly weak, if objective knowledge testing isapplied (Schenk and Rssler, 1997). The variable of knowledge has neverbeen said to be the only presupposition of opinion leadership and yet,

    communication scholars assumed opinion leaders to be knowledgeablein their area of interest: It is evident, therefore, that subjective interest,or advanced knowledge, in the arena of public affairs as in the otherarenas cannot operate independently to generate opinion leadership.Along with information and interest, there must be some objective en-abling factor which makes it possible to translate subjective predisposi-

    tions into the actual give-and-take of participation and leadership (Katzand Lazarsfeld, 1955, p. 275).

    Considering the low level of relationships between opinion leadership

    and knowledge in subsequent studies, we assume that there might bedifferent types of opinion leaders and non-leaders. Opinion leadership

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    8/22

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    9/22

    Opinion leaders 127

    results of previous research are manifold. According to the Decaturstudy, opinion leaders use more media and are more influenced by themthan others (Katz and Lazarsfeld, 1955). Preceding studies stated thatopinion leaders are more likely to read the newspaper, but not to watchtelevision or listen to the radio (Troldahl and van Dam, 1965; Schenkand Rssler, 1997). In contrast, other scholars found that opinion leadersdo not use media more often than other people do (Can and Shekar,1990; Lin, 1973; Levy, 1978; Robinson, 1976), but that they use mediamore for cognitive orientation (Levy, 1978), and that they are morelikely to talk about media content with friends and relatives (Hamilton,1971).

    It can be assumed that knowledge is acquired by media use. And,consequently, knowledge might be an intervening variable, moderating

    the causal effect of media use and opinion leadership. However, otheralternatives have to be thought of. There might be people who are en-abled to influence others by high media use and acquired knowledge butlack the leadership abilities to do so. Also, there might be opinion leaderswho do not have knowledge in their supposed area of interest and arenot able or willing to acquire it by media use, but nevertheless are ableto influence others. The latter group might compensate their lack ofknowledge with opinion leadership abilities. To explore these open ques-tions, first, the relationship between media use and opinion leadership

    will be investigated.

    RQ 3: Do informed opinion leaders use mass media more extensivelythan uninformed opinion leaders?

    Two empirical studies will be used to answer the research questions. Thefirst study was carried out in a class setting, and tested knowledge, butwith the disadvantage of a small sample size. The second study is basedon an online survey and a large-scale sample.

    Study 1

    Procedure

    A survey was conducted in universities of two large cities in NorthernGermany. Participants filled out a paper-pencil questionnaire during in-troduction to communication and introduction to psychology classes.Respondents received course credit for participation. It took about 20minutes to fill out the questionnaire. The questionnaire was divided into

    three sections. First, the scales for opinion leadership were assessed. Ifitems were very similar, they were not used twice. The second section

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    10/22

    128 Sabine Trepte and Helmut Scherer

    consisted of 20 items on German home policy and public affairs. In thethird section, respondents were asked for their sex, age, major subject ofstudy and about their media use habits.

    Sample

    A total of 119 students took part in the survey. The average age ofrespondents was M 24.22 (s 4.47; range 1941 years old). Twothirds of respondents were female (69 %), one third was male (29 %).Students were majoring in humanities (32 %), media management andeconomics (24 %), linguistics (22 %), public relations, and journalism(19 %).

    Measures

    Knowledge was assessed by a scale of 20 items. 13 questions were askedin a multiple choice format (e. g. In what city is the German FederalConstitutional Court located? Please mark the right answer: Berlin,Bonn, Frankfurt, Karlsruhe; When did Switzerland join the UNO?Please mark the right answer: 1945, 1977, 1999, 2002). Seven questionswere asked as open questions and participants filled in their answers(e. g. What is another name for the UN peacekeeping forces?; Whois the president of UNMOVIC?). The respondents answers range from119.

    Media use was assessed by the question How long do you use thefollowing media per day: television, newspaper, radio, internet and maga-zines?. Participants were asked to fill in blank spaces and to indicatetheir media use by hours (television, radio) or minutes (newspaper, in-ternet, magazines). Hours were rounded to full (e. g. 3) and half (e. g.3.5) hours and thus a metric scale is used for further analyses.

    Opinion leadership was measured by using four different scales andapproaches: opinion leadership scales of Katz and Lazarsfeld (1955),

    Troldahl and van Dam (1965), and Childers (1986). In the first study,items of the Childers scale that were represented in the other scales werenot requested to keep the questionnaire short. For that reason not all ofthe items followed a Likert scale response format. The respondents an-swers range from 723. Additionally, Noelle-Neumanns (1983) person-ality strength scale was applied.

    To assess the scales reliability and congruency, a three-step analysiswas conducted. The first step was to determine the scales internal consis-tency with Cronbachs alpha. The second step was to determine by factor

    analyses whether the scales were one-dimensional. In the third step theopinion leadership scales were correlated in order to establish whether

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    11/22

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    12/22

    130 Sabine Trepte and Helmut Scherer

    strength scale with the other scales are rather poor with values betweenr .30 and r .40.

    To summarize, Katz and Lazarsfelds (1955) scale measures with theleast accuracy. The personality strength scale supposedly measures aconcept different to opinion leadership. The most reliable scale appearsto be Childers (1986) scale as it has the highest internal consistencyreliability and shows high congruency with the other scales. Accordingto these results, the following analyses are based on Childers (1986)scale of opinion leadership (pls. see Appendix for scale)

    Results

    RQ1: The first research question asks, whether opinion leadership scalesdiffer in their predictive power to forecast knowledge in a certain areaof interest. According to previous research, knowledge and opinion lead-ership yielded rather weak correlations (Schenk and Rssler, 1997;Schenk; 1985). To check the analogy of our findings with these studies,we computed bivariate correlations (c. f. table 3). Between three of thefour scales and the variable knowledge a significant correlation was ap-parent. With a range ofr .342 tor .431, correlations are higher thanin preceding research. The correlations show a positive relationship witha medium strength. People, attaining high values on the opinion leader-

    ship scales, are presumably in possession of more knowledge than fellow-persons. This is not the case for personality strength. On average (forthree significant opinion leadership scales) 16 percent of variance in

    opinion leadership is explained by the knowledge variable.RQ 2: With regard to the second research question, we investigated

    the assumption that the following four types result from the combina-tion of knowledge and opinion leadership: informed opinion leaders,

    uninformed opinion leaders, silent experts and others. A hierarchical

    cluster analysis was conducted (ward treatment, squared euclidian dis-

    tance, z-transformed variables). We used two variables in the analysis:

    Table 3. Correlation between opinion leadership scales and knowledge index (Pearsoncorrelation coefficient).

    Index of Knowledge

    Childers Scale index ,431**Troldahl and van Dams Scale index ,418**Katz and Lazarsfelds Scale index ,342**Noelle-Neumanns Personality Strength index ,113

    **p< .01

  • 8/10/2019 Opinion Leaders - Do They Know More Than Others About Their Area of Interest

    13/22

    Opinion leaders 131

    Table 4. Study 1: Description of clusters by clustering variables and media use.

    informed uninformed silent others Allopinion opinion experts Respond-leaders leaders ents

    Respondents (n) 9 55 7 48 119

    Childers Scale Total IndexMean1) 19.1a 17.4b 11.9c 11.1d 14.7Standard Deviation 3.3 2.0 1.1 1.9 3.8

    Total Knowledge IndexMean1) 15.6a 7.0b 11.3c 4.8d 7.0Standard Deviation 2.1 2.5 2.0 1.9 3.7

    Media UseNewspaper1) (min./day) 39a 26b 16c 14c 22

    Magazine use (min./day) 20 14 9 17 16Television use (h/day) 2.8 1.8 2.1 1.8 1.9Radio use (h/day) 1.4 1.5 1.1 1.4 1.4Internet use (min./day) 42 68 36 67 64

    1)F-value is statistically significant (p < .01); means in the same row that do not sharesubscripts differ at p < .05 in the post-hoc test (Scheffe).

    opinion leadership (Childers, 1986) and knowledge. The distance coeffi-cient indicated that a solution of four clusters would be best.

    Considerable differences are seen in the number of respondents as-signed to the resulting clusters. The following statements can only betentative, based on the low sample size. Table 4 shows the resulting meanof the variables for each individual cluster. Differences between clusterswere computed with one-way analyses and post-hoc tests (Scheffe). Sta-tistical significance was yielded regarding knowledge (F(3,115) 68.02,

    p< .001) and regarding opinion leadership (F(3,115) 97.838, p