The Role of Involvement and Opinion Leadership in Consumer Word-of-Mouth: An Imfdidt Model Made Explicit. Marsha L. Richins, University of Massachusetts.
The Role of Involvement and Opinion Leadership in Consumer Word-of-Mouth: An Imfdidt Model Made Explicit Marsha L. Richins, University of Massachusetts Teri Root-Shaffer, Louisiana State University
Abstract Models or frameworks of opinion leadership normally start with opinion leadership and postulate about its impact on recipients and on the success of new products. Researchers have been less interested in modeling opinion leadership itself. This paper examines the opinion leadership literature to determine how consumer behavior researchers have viewed opinion leadership at the sender level and idenitifies the model implicit in their work. This implicit model is expanded and tested empirically.
global measures of WOM such as asking "how much" respondents talk with others about a product category (e.g., Myers and Robertson 1972). Little if any research has examined the content of opinion leaders' comments or attempted to distinguish the conversational content of opinion leaders vis a vis consumers who are not opinion leaders. Thus, while empirical evidence is not especially voluminous, the model of opinion leadership implied by consumer behavior writers can be represented as follows:
Introduction The importance of interpersonal communication in consumer decision processes has been documented again and again in consumer behavior research, with numerous studies describing the frequency of consumer word-ofmouth and its influence on recipients (Arndt 1967; Katz and Lazarsfeld 1955; Leonard-Barton 1985; Technical Assistance Research Programs 1981). Even in this era of mass communications and mass advertising, it has been estimated that as much as 80% of all buying decisions are influenced by someone's direct recommendation (Voss 1984). Opinion leaders are credited with a large amount of this interpersonal communication, and considerable research resources have been directed at identifying the demographic and social characteristics of opinion leaders (e.g. Myers and Robertson 1972; Summers 1970). Much less effort has been directed at identifying the motives underlying opinion leadership or understanding why opinion leadership occurs at all. In one of the few studies on this topic. Dichter (1966) suggested that involvement with the product class is an important determinant of word-of-mouth. While Dichter's analysis has stimulated little further research, a few early studies of opinion leadership that examined a potpourri of consumer variables found significant correlations between opinion leadership and product involvement, among other variables (Reynolds and Darden 1971; Summers 1970). Despite the rather wobbly theoretical and empirical underpinnings for the conclusion, however, involvement seems to be accepted as the motive for opinion leadership. Feick and Price (1987) summarize current thinking as follows:
There are, however, some inadequacies with the implicit model of opinion leadership, primarily with respect to the involvement variable. Dichter (1966, p. 148), in his explanation of how product involvement results in wordof-mouth, suggested that "experience with the product (or service) produces a tension which is not eased by the use of the product alone, but must be channeled by way of talk, recommendation, and enthusiasm...." While any form of product excitement may result in WOM, it is questionable that all types of product "excitement" or involvement result in opinion ieadership. Recent distinctions in the product involvement literature (Bloch and Richins 1983; Houston and Rothschild 1978) suggest that product involvement may be transitory (situational) or it may be long term and enduring. Because of the transitory nature of situational involvement, it is unlikely to result in a relatively permanent state such as opinion leadership. For this reason, the implicit model of opinion leadership is altered for purposes of this research. As shown in Figure 1, enduring involvement is expected to result in opinion leadership, which in turn results in word-of-mouth. Situational involvement is expected to result in word-ofmouth but to have no linkage with opinion leadership. Figure 1 A Model of Opinion Leadersiiip
the implicit assumption in examining the personal influence of opinion leaders is that they are motivated to talk about the product because of their involvement with it.... Product involvement remains the predominant explanation for opinion leaders' conversations about products (p. 84). The essential behavior said to define opinion leaders is that they talk about products, yet the linkage between opinion leadership, as commonly measured, and actual word-of-mouth is not well understood. Only a few studies have measured word-of-mouth separately from opinion leadership, and even these studies have used
The research reported here tests this revised model of opinion leadership. It also explicitly tests the link between opinion leadership and word-of-mouth, and does 32
Advances in Consumer Research Volume 15, ©1988
Advances in Consumer Research (Volume IS) 133 so with specific rather than global WOM measures used in earlier research. Methodology Product Class Automobiles were chosen as a suitable product class for this study for several reasons. First, it was desirable to use a product owned by a large percentage of the general population. Second, to assess situational involvement effects it was necessary to choose a product capable of eliciting high levels of situational involvement, at least at the point of purchase; prior research has shown this to be the case for automobiles (Hupfer and Gardner 1971; Richins and Bloch 1986). Access to a list of new car registrants permitted the identification of respondents who can be said to possess high situational involvement at the time of the survey. Finally, a product capable of eliciting high levels of enduring involvement among some but not all consumers was needed to provide an adequate range on the enduring involvement construct (Bloch 1981). Data Collection Data were collected using a mail survey with follow-up ieminder. Questionnaires were mailed to a randomly drawn sample of 650 adult consumers living in a medium sized Sunbelt city; 217 of these retumed usable forms. To provide responses along the complete range of the situational involvement construct, this general population sample was supplemented with a mailing to 125 new car owners identified from state motor vehicle registration office records. 53 usable forms were obtained from this sample, yielding a total sample size of 270. While the inclusion of the supplemental sample limits the extemal validity of the study, it should be noted that it is not the intention of this study to make specific generalizations from a sample to a population of interest but rather to understand underlying theoretical processes. Measures Opinion leadership was measured using four items taken from the King and Summers (1970) scale. A five point response format was used (Childers 1986) instead of the usual dichotomous response format. The items were summed to form a composite measure of opinion leadership. Cronbach's alpha equalled .82, indicating that the reliability of the scale was not compromised by using a shortened version. Enduring involvement was measured using the 9item version of the automobile involvement scale developed by Bloch (1981; see Richins and Bloch 1986). The summed scale had an alpha of .90. The recent purchase of a new car was used as a dummy variable indicator of situational involvement. The short term temporal nature of situational involvement has been discussed in the literature (Antil 1984; Bloch and Richins 1983; Houston and Rothschild 1978). Following Richins and Bloch (1986), a twomonth period was used to divide respondents into high and low situational involvement groups; those respondents who had purchased a new automobile within the two months prior to the study were placed in the high situational involvement category. Types of word-of-mouth were measured using 13 items generated from depth interviews of adult consumers who described product-related conversations in which
they had recently participated. Questionnaire items were designed to measure frequency of different types of comments about automobiles. Respondents indicated the number of times they had discussed each topic within the last two weeks. Pretesting had revealed that respondents have some difficulty in recalling the precise number of comments made over a two-week period. However, respondents were generally confident that the ordinal relationship among the types of comments was correct. In other words, they believed their estimates accurately reflected which types of comments they made more frequently and less frequently than others. Principal components analysis with oblique rotation resulted in four factors for word-of-mouth comments, which were labeled positive personal experience, advice-giving, product news, and negative word-of-mouth. Because of concem for the level of measurement, this analysis was performed twice, once using a matrix of Pearson r's and the second time a matrix of rank correlations. Results did not differ. Positive personal experience word-of-mouth consists of comments in which respondents made favorable statements about their cars or described how or why they bought their cars. All these topics are related to respondents' personal experiences. Advice-giving word-of-mouth included comments in which respondents gave information about cars or advice about which car to buy. This factor seems to capture the dimension of wordof-mouth communication which has traditionally been studied in the opinion leadership literature (Jacoby 1974; Rogers and Cartano 1962). Product news includes comments about advances in car technology, car model differences, and similar topics. In contrast to the first word-of-mouth dimension, comments loading highly on this factor seem to be based less on personal experience with one's own car and more on general knowledge about automobiles as a product class. Negative word-of-mouth consists of statements in which respondents said something unfavorable about their cars or described something they didn't like about them. Preliminary analysis showed that this type of word-of-mouth had no sigificant relationship with either involvement or opinion leadership, so only the first three word-of-mouth factors were used for the remainder of the study. Items in each of the three remaining factors with loadings greater than .6 on the pattem matrix were summed to create the three word-of-mouth subscales, as follows:
subscale personal experience advice-giving product news
# of items 4 2 5
alpha 0.89 0.68 0.83
Because the distributions of the word-of-mouth variables were highly skewed, square root transformations for these variables were used in subsequent analyses. The three word-of-mouth factors require a slightly more complex model, shown in Figure 2. Analysis and Resuits The path coefficients for the recursive model were estimated by ordinary least squares regression using path analysis techniques arising from the work of Simon and
34 / Involvement, Opinion Leadership, and Consumer Word-of-Mouth Figure 2 Opinion Leadership Modei wtth Tliree WOM Factors
Figure 3 Revised Opinion Leadership Model
Endunng ^ Involvement y
.27
\34
.44/^
f Opinion ^ V Leadership . ^ ^ . ^S ^ 3 /SituationaiN \lnvolvement/
Blalock (see Asher 1983; Blalock 1964; Pedhazer 1982). In order to test whether the path coefficients assumed to be zero in the hypothesized model were indeed zero, these linkages were included in the initial test of the model. Regression results are shown in Table 1. In interpreting regression results, statistical significance was not the only criterion used to determine whether a path should be included in the model. Because of the large sample size, even coefficients of small magnitude may achieve statistical significance. Therefore, an arbitrary cutoff was used in addition to the significance criterion. Path coefficients less than .20 (accounting for less than 4% of variance) were considered marginal and corresponding linkages omitted from the final model. Regression results indicate that the hypothesized model needs to be modified. Two hypothesized linkages were not strong enough to be maintained. The path coefficients for hypothesized links between situational involvement and two forms of word-of-mouth (product
.47
^ Product .News WOM ^''^ ^
Advice Givinj
lir^'Persoriar' Experience V WOM ^
news and advice-giving) were less than .20; these links were eliminated. Regression analysis also revealed that an additional path needs to be added to the model. The coefficient for the path between enduring involvement and product news word-of-mouth was .24 (p opinion leadership—> word-of-mouth model implied in the literature. Involvement does appear to be an important antecedent Path CoeffiPath to opinion leadership, but it is necessary to specify that Explan- cients Coefficients only enduring involvement results in opinion leadership. Dependent atory Full R- Trimmed RSituational involvement bears no relationship at all with Variable Variables Model Square Model Square opinion leadership. Likewise, the implicit relationship between opinion leadership and word-of-mouth is Opinion .05 .12 .11 SI confirmed. It is not surprising that the relationship .34* Leadership(X3) .33* El between opinion leadership and word-of-mouth is strongest for advice-giving, the form of word-of-mouth .26 Product News SI .18* .29 traditionally linked with opinion leadership. In these El .24* .27* WOM (X4) respects, the implicit model is correct. .35* OL .34* Findings also indicate that the implicit model inadequately represents word-of-mouth communication. .18 .23 Advice—Giving Si .19* Word-of-mouth results from situational involvement not El .11 WOM (X5) just from opinion leadership. Situational involvement is OL .37* .42* especially associated with personal experience word-ofmouth. For at least a brief period, consumers seem to engage in product-related conversations after the purchase .30 .45* SI .43* .33 Personal of new automobile because of the excitement generated El .16* Experience by the new item (Dichter 1966) or possible cognitive .28* OL .23* WOM(X6) dissonance (Menasco and Hawkins 1978). This arousal [a] X1=Enduring Involvement (El), X2»Situational dissipates over the course of time, however, and productInvolvement (SI), X3 = Opinion Leadership (OL) related word-of-mouth declines as well. *p