Sie sind auf Seite 1von 17

Intern. J. of Research in Marketing 23 (2006) 45 61 www.elsevier.

com/locate/ijresmar

Antecedents and purchase consequences of customer participation in small group brand communities
Richard P. Bagozzi a,, Utpal M. Dholakia b
a

University of Michigan, Ross School of Business, 701 Tappan St., Ann Arbor, MI 48109-1234, USA b Rice University, Jesse H. Jones Graduate School of Management, Houston, TX 77005, USA

Abstract This paper investigates behavior and determinants of the behavior of small group brand community participants. A small group brand community is a friendship group of consumers with a shared enthusiasm for the brand and a well-developed social identity, whose members engage jointly in group actions to accomplish collective goals and/or to express mutual sentiments and commitments. Group activities centered on the brand intermingle with other social activities in these brand communities. A comprehensive model is proposed based on a broadening and deepening of the theory of planned behavior to incorporate social intentions, three aspects of social identity (cognitive self-awareness of membership in the brand community, affective commitment, and evaluative significance of membership), anticipated positive and negative emotions toward achieving or failing to achieve group participation goals, and desire as a transformative mechanism translating reasons for acting into social intentions to do so. The proposed theoretical framework is tested on a sample of 154 members of Harley-Davidson Motorcycle small group brand communities and another sample of 255 members of motorcycle riding groups not organized around specific brands. 2006 Elsevier B.V. All rights reserved.
Keywords: Brand communities; Consumer behavior; Social influence; We-intentions

1. Introduction The small group tends to restore, structurally, the symbolic power. Step by step, one can see a mystical network being built, carefully yet solidly connected, leading one to speak of a cultural resurgence in social life. This is the lesson taught by these eras of the masseseras based mainly on the concatenation of groups with splintered but exacting intentionalities (Maffesoli, 1996, p. 83, emphasis added). Amidst the increasing media clutter and growing resistance of consumers to traditional marketing programs, marketers' interest in novel means of engaging consumers continues to grow. One approach that has received attention recently is that of organizing and nurturing brand communities of customers (Algesheimer, Dholakia, & Herrmann, 2005; McAlexander,
Corresponding author. Tel.: +1 713 647 6435; fax: +1 734 936 8715. E-mail addresses: bagozzi@umich.edu (R.P. Bagozzi), dholakia@rice.edu (U.M. Dholakia). 0167-8116/$ - see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.ijresmar.2006.01.005

Schouten, & Koenig, 2002; Muniz & O'Guinn, 2001). Brand communities are defined as specialized, non-geographically bound communit(ies), based on a structured set of social relationships among admirers of a brand. Examples of successful brand communities include the Harley-Davidson Motorcycles' Harley Owners Groups (HOGs), Apple Computer's MacIntosh user groups (e.g., Belk & Tumbat, 2002), Volkswagen drivers (Brown, Kozinets, & Sherry, 2003) and Sun Micosystems' Java center community (Williams & Cothrel, 2000). To many marketers, brand community building appears as an effective marketing program for at least two reasons. First, brand communities are not subject to many of the problems increasingly associated with traditional marketing approaches such as fragmentation of media and the accompanying clutter faced by mass advertising campaigns, and the resistance of consumers to receiving marketing communications faced by direct marketing campaigns. In contrast, brand communities are venues where intense brand loyalty is expressed and fostered, and emotional connections with the brand forged in customers. Second, brand communities coincide with the increasingly

46

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

popular movement of consumer empowerment (Prahalad & Ramaswamy, 2000) which encourages firms to treat their customers as partners, cede control over information gathering and decision making to them to a significant degree, and coopt their competence in ways that are mutually beneficial and profitable. Underlying the prevalent views of the effectiveness of brand communities is the assumption that forming relationships with other like-minded consumers who share one's interest in the brand will be credible and impactful in persuading and bonding customers to the brand, leading them to make more purchase behaviors and be more loyal. While recent marketing studies of brand communities (Algesheimer et al., 2005; Muniz & O'Guinn, 2001; McAlexander et al., 2002) have provided some support for this assumption, what is missing from these extant analyses is a detailed understanding of the social and psychological variables that lead a firm's customers to participate in collective action within the brand communities. Furthermore, these studies have also not distinguished clearly between brand communities that are mainly comprised of small friendship groups and other types of community influences on consumers such as those from subcultures (e.g., Schouten & McAlexander, 1995) and networks (e.g., Muniz & O'Guinn, 2001). Our goals in this research are to address these issues. In particular, we elaborate on the concept of small group brand communities using the example of Harley-Davidson's HOG members to do so. We then investigate the social and psychological antecedents of group- and brand-related behaviors of two consumer communities, small group brand communities organized around the Harley-Davidson brand and a comparison group of motorcycle riders not organized around any particular brand. Our analyses provide an in-depth conceptualization of the variables leading to collective customer behaviors in small group brand communities. 2. The concept of small group brand community In the published research, Harley-Davidson's HOGs are advanced as the prototypical example of a brand community (e.g., Fournier, McAlexander, Schouten, & Sensiper, 2000; Schouten & McAlexander, 1995; Yates, 1999). It is worth noting that even though HOG chapters are organized by geography and often comprise hundreds of registered members, for most active members, a majority of the social interactions occur with a small group of friends. We call these, small group brand communities, which are comprised of clusters of Harley riders, typically fewer than ten or so riders plus a few passengers, that have close friendships with one another and engage in regular and frequent face-to-face interactions. For such communities, the demarcation blurs in that brand-related activities intermingle with the group's social activities. This is in contrast to network-based brand communities, which Muniz and O'Guinn (2001, p. 415) characterize as explicitly commercial, where brand-related activities predominate, and social relationships are tenuous and based on narrow individualistic motives, and interactions occur exclusively through virtual media and in firm-orchestrated venues such as internet

bulletin-boards or chat-rooms (Dholakia, Bagozzi, & Pearo, 2004). For Harley riders belonging to small group brand communities, social interactions occur through group rides for purely recreational purposes or for more formal goals such as fundraising (e.g., a rally for raising funds for the Asian Tsunami victims), competitions, political protests (e.g., anti-helmet law rallies), or community service. Even more frequently, small group members come together to meet at a pub or restaurant or to mutually examine the latest bikes and accessories at a nearby dealership. The latter activities for the groups we studied occur not only weekly but frequently at multiple times during any particular week, making this type of small group brand community one of high personal and social involvement for consumers. In comparison to consumer networks, the interpersonal relationships among community members are stronger and multi-faceted, going beyond brand-related interactions. Small groups are members of the local HOG chapter, which is usually funded by the Harley-Davidson Motorcycle company, and the local Harley dealership(s). Indeed, such organized support through funding, through organizing various events, by providing a place to meet, through providing a variety of ancillary products and services, etc. from the firm and its dealers, is critical in sustaining small group brand communities At the same time, the forces that seem to energize these communities are the on-going sentiments and sensuality that occur in fellowship as members relive past happenings, share current concerns, and plan for future outings together. There is an aesthetic component to these forces as well. Consequently, small group brand communities are strongly sociocentric, with members exuding a strong identity with their group as well as a strong identification with the brand. Another aspect of small group brand communities is that they correspond to Maffesoli's (1996) concept of affectual tribes, in the sense that they lack formal hierarchies. Informal hierarchies within them tend to be limited to the infrequent occurrences when new members join the group (as in initiationlike rites), or in defacto ways, where one or a few members take the lead in suggesting and organizing the next joint activity, or when one or two members exert their (implicit) status to head a group riding formation. Women sometimes are subtly accorded lower status, such as during group rides where men, who typically drive more powerful bikes and exhibit less caution, often speed away from women members. However, it would be misleading to overemphasize the informal hierarchies in the Harley small group brand communities, for most tend to be egalitarian collectivities that are horizontally organized by friendship ties.1
1 A good example of this, witnessed in the qualitative phase of the study, was the reaction of one woman rider who let it be known in no uncertain terms that the group would not be a group anymore if the men were to once again so thoughtlessly and selfishly pull away from the pack. Efforts to create status differences tended to be subtly and not so subtly counteracted by such reactions in the small group brand communities we studied. Most members of these communities were what are known in the marketing literature as Mom-andPop bikers, plus a few rich urban bikers (see Schouten & McAlexander, 1995, p. 49).

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

47

To recap, small group brand communities are highly involving, sustained social groups. The affectual bonds and social communion in such customer communities might be termed tribalism plus linking value (Cova, 1997). Importantly, the magic and mystique of owning and riding a Harley motorcycle and its associated positive connotations provide the glue binding members together and are visible in both the foreground and background of social life in these brand communities. 3. Theoretical framework and research hypotheses In this research, we attempt to develop an in-depth understanding of the social and psychological processes that lead small group brand communities to influence consumers and encourage collective participation. To do so, we begin with a discussion of a well-known approach in social psychology: the theory of planned behavior (TPB, Ajzen, 1991). The TPB is frequently employed in contemporary research and has been used successfully in hundreds of applied studies (e.g., Armitage & Conner, 2001; Eagly & Chaiken, 1993). It maintains that behavior is directly influenced by one's decision to act (i.e., intention) and the control one perceives one has over the behavior; intention to act, in turn, is dependent on attitudes toward the act, subjective normative pressure to act, and perceived behavioral control. While the variables and processes under the TPB seem relevant to certain aspects of small group brand communities, it fails to capture important social processes, emotional behavior, and goal-directed features of participation in them. As a consequence, we endeavor to both

broaden and deepen the TPB, to derive an approach more explanatory of brand community participation than is afforded by the predominantly individualistic TPB. The TPB will be broadened by adding a number of new antecedents; it will be deepened by introducing a mediator between the antecedents and intentions, which better explains how and why the antecedents influence the intentions to participate in collective actions in small group brand communities. Note that while previous studies have suggested various enhancements to the TPB in many different and non-brand contexts such as weight loss, eating at restaurants, blood donation, etc. (e.g., Armitage & Conner, 2001; Bagozzi & Lee, 2002; Perugini & Conner, 2000), our intended contribution is to propose and test a coherent and integrated framework that provides insights and helps explain collective consumer participation in small group brand communities. No study to date has integrated the personal and social aspects of behavior we introduce in modification and expansion of the TPB, and no study has used the TPB or modifications thereof in the study of brand communities. From a practical standpoint, we suggest that in addition to cognitive (e.g., evaluations of product performance), and motivational (e.g., commitment to the brand) variables, brand loyalty can also be influenced by firms through encouraging interactions with small groups of other enthusiastic customers in a setting that is controlled and managed more-orless by the customers themselves. Our presentation begins first with a discussion of social processes added to the TPB. Next, we consider new emotional and goal-directed (psychological) content to be introduced into the theory. Fig. 1 summarizes the proposed model developed below.

Fig. 1. Proposed model for explaining antecedents and consequences of participation in small group brand communities.

48

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

3.1. Social processes Intentions under the TPB, and indeed in virtually every approach taken to date in attitude theory, have been construed in a strictly individual sense. For example, Eagly and Chaiken (1993, p. 168) interpret an intention as a person's motivation in the sense of his or her conscious plan to exert effort to carry out a behavior by him or herself. We might term this an Iintention; an example would be, I intend to ride my motorcycle this afternoon. The TPB and contemporary models of attitude explain I-intentions by individual-level reasons for performing a personal act (i.e., attitude toward the personal act; perceived behavioral control over the personal act) and perceived interpersonal pressure to perform the personal act (i.e., subjective norms). This is also how most consumer behavior models predicting purchase, loyalty, etc. are formulated. Such a formulation raises three questions. First, is it meaningful for consumers to form intentions to perform something other than personal acts? Second, can the referents or objects of reasons for acting constitute group acts? Third, how can social reasons for acting be better formulated than found in subjective norms? Note that these questions have considerable relevance not only for the present case of collective participation in small group brand communities, but also for other important marketing issues such as selling products which are consumed jointly with others, formulating community-based marketing programs, devising effective strategies to influence socially conscious consumers, etc. With regard to the content of the consumer's intentions, it seems possible that intentions can be in some sense social. For example, a consumer in a relationship might speak about our intention to see a Broadway play; a soccer player might mention the team's plan to attend an end of season party at a local restaurant, and an executive editor might announce Consumer Reports' decision to purchase and test hybrid vehicles. These examplesreferring respectively to a two person dyad, a small group, and an organizationillustrate that people often use social notions of intentions in ordinary speech, whether designating informal or formal groups. Social scientists have generally not treated social intentions of the sort illustrated above, but philosophers have considered the logical foundations of such intentions (for an analysis and review, see Bagozzi, 2000). Bratman (1997) calls this kind of intention a shared intention and expresses it in the form, I intend that we act. In a somewhat similar manner, Tuomela (1995, p. 2) defines a we-intention as a commitment of an individual to participate in joint action and involves an implicit or explicit agreement between the participants to engage in that joint action. Using these definitions, a social intention might be termed as a consumer's shared intention to perform a group act. This can take the form, We intend to rally for Tsunami relief next month. Here a person plans to participate in a joint activity, but conceives of the activity, not so much as individuals performing personal acts that atomistically accumulate and contribute to a group performance, but rather as a group action where one is a

member of the group. It can also be of the form, I intend that we (my three friends and I) play bridge this weekend. It is noteworthy that in these formulations of social intentions, the activity (and the intention) only has meaning if the group acts in concert. Individual action by oneself is not enough to attain one's goal in either case. Therefore, when formulating a social intention, the consumer acts as an agent of, or an agent with, the group, in coordination and cooperation with its other members. Social intentions form the basis of participation in small group brand communities, since the actions sustaining the community are by the group as a whole. It is worth pointing out that social intentions conceptualized this way are equally significant for marketers of products that are usually purchased and consumed jointly by friendship and/or family groups such as new homes, package tours, online games, etc. In our theoretical framework, it is these social intentions that constitute group decisions and explain group behavior, where the latter is defined as activities engaged in by the consumer with the group of members self-designated as Harley-Davidson club friends that one regularly meets together with (see Fig. 1). These activities include such joint behaviors as recreational rides, sponsored rides, attending club meetings, mutual fund raising efforts, gathering for meals in a restaurant, socializing in pubs or at picnics, and sharing in shopping trips or other special events. The TPB only allows for one type of social determinant of intentions and behavior: namely, subjective norms. Subjective norms reflect the impact of expectations from other people, which are largely based on the need for approval. Kelman (1974) termed this mode of social influence, compliance. However, this is only one type of social influence that is relevant for consumer participation in small group brand communities. As Terry and Hogg (1996, p. 778) point out, the operation of subjective norms in many situations may be problematic because it describes norms in terms of the context to which people perceive that others want them to perform the behavior, yet for behaviors that do not influence other people or behaviors that are not so directly associated with [impactful] outcomes, such pressures may be latent and may not be perceived. Eagly and Chaiken (1993, p. 683) note further that features embedded in certain types of role relationships, rather than compliance-targeted stimuli or cues, need to be considered when investigating the social context of action. A different and more subtle aspect of social influence needs to be considered than found in subjective norms, if we are to capture the consciousness of kind aspect of brand community that Muniz and O'Guinn (2001) identified. One way to deepen and elaborate this aspect is to assemble together three notions from social identity theory and social categorization theory (e.g., Hogg & Abrams, 1988; Tajfel, 1978). Tajfel (1978) suggested that a person achieves a social identity through selfawareness of one's membership in a group and the emotional and evaluative significance of this membership. Building on these insights, Ellemers, Kortekaas, and Ouwerkerk (1999, p. 372) recently proposed that three components comprise one's social identity: a cognitive component (a cognitive

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

49

awareness of one's membership in a social groupselfcategorization), an evaluative component (a positive and negative value connotation attached to this group membershipgroup self-esteem), and an emotional component (a sense of emotional involvement with the groupaffective commitment). Construct and nomological validity for measures of these components have been demonstrated by Bergami and Bagozzi (2000) and Bagozzi and Lee (2002). Social identities prescribe and instigate behaviors for the benefit of group members. Ellemers et al. (1999) studied experimentally formed groups and found that aspects of social identity influenced in-group favoritism (in terms of evaluative ratings and outcome allocations). Similarly, Bergami and Bagozzi (2000) found in an investigation of food service employees that social identity influenced organization citizenship behaviors (a form of in-group favoritism). In the study by Algesheimer et al. (2005), identification with the car club led to involvement and engagement in group initiatives among members. For small group brand communities, social identity is expected to influence social intentions to participate with group members in collective activities. Similar to the rationale for the effects of group identity on in-group favoritism (e.g., Ellemers et al., 1999), and cooperation and organizational altruism (e.g., Bergami & Bagozzi, 2000), one reason for this is the self-enhancement that the person receives as a consequence of group membership and group participation. Self-enhancement is thought to arise through a boost in personal self-esteem (Tajfel & Turner, 1986) as well as collective self-esteem (Crocker, Luhtanen, Blaine, & Broadnax, 1994). Social identity with one's brand community is also expected to impact the consumer's brand identification. Brand identification is the extent to which the consumer sees his or her own self-image as overlapping with the brand's image. As one's social identity with the brand community increases and participation in group activities is fostered, a greater involvement with the brand occurs, which should promote the assimilation of the brand's images into one's identity. This hypothesized path is also supported by Escalas and Bettman (2003) who show that an important set of brand associations is that regarding reference groups, and such associations influence brand equity to the extent that the consumer belongs to the group. Note also that such a path is different from the study by Algesheimer et al. (2005) where a positive relationship with the brand preceded, and led to, identification with the car club. Our prediction follows from our focus on small group brand communities, whereas they studied European car clubs of all sizes, comprised of consumer networks as well, where members tended to join only after forming a relationship with one's vehicle brand. In contrast, we predict that on-going activities within the small group brand community will impact the consumer's bond with the brand, affirming and strengthening it. From a practical standpoint, we note that this hypothesized path is important in placing small group brand community marketing programs on a comparable footing with other established marketing approaches such as image advertising that also seek to strengthen the consumer's identification with the brand.

3.2. Psychological processes Social intentions, group behavior, and social identity are social processes not addressed by the TPB. Instead, the TPB is primarily a model of individual decision making and focuses on two psychological variables (in addition to subjective norms): attitudes and perceived behavioral control. Attitudes are evaluative (cognitive) reactions to an action and are thought to reflect predispositions to respond in a favorable or unfavorable manner (Eagly & Chaiken, 1993). Attitudes are believed also to arise through learning, whereby a person acquires a reaction to an action over a period of time. Once learned, the attitude can be triggered automatically when one is exposed to the action or thinks about it (Fazio, 1995). Perceived behavioral control refers to a decision maker's judgment that he or she can perform an act (e.g., how easy or difficult the act is to do) and is thought to help account for behavior not fully under volitional control (Ajzen, 1991). Although attitudes and perceived behavioral control have been shown to be important antecedents to I-intentions (e.g., Armitage & Conner, 2001), a number of shortcomings can be pointed out that we try to overcome in the present research. An important limitation of the TPB is its failure to take into account emotional reactions to a contemplated action (note also that the emotional reactions we introduced aboveaffective commitmentare with respect to one's felt social identity, not behavior). Early attempts to incorporate emotion into the TPB were limited to the emotion of regret and at the same time exhibit problems noted below (e.g., Richard, van der Pligt, & de Vries, 1995). A comprehensive approach to the role of emotion in decision making was proposed by Bagozzi, Baumgartner, and Pieters (1998) who argued that people, when considering whether or not to act in goal-directed situations, take into account the emotional consequences of both achieving and not achieving a sought after goal. In both cases, the processes are believed to be predicated upon a type of thought process analogous to counterfactual thinking, but perhaps better termed prefactual appraisals (Gleicher et al., 1995), whereby a decision maker imagines the affective consequences of goal attainment and goal failure before deciding to form an intention and act. In our analysis of small group brand community members, and based on the emotion literature, we include a total of 21 emotions (9 positive, 12 negative; see Methodological considerations below). Anticipated positive and negative emotions differ from attitudes in three ways, namely the referents, underlying theoretical processes, and manner of measurement (e.g., Perugini & Bagozzi 2001; for other distinctions, see Bagozzi, 2006). First, attitudes toward the act under the TPB focus upon what one does or can do, whereas anticipated emotions focus not upon action, but rather upon the attainment and failure to attain a goal. Under the TPB, all antecedents have action as their referents, and indeed, Ajzen and Fishbein have made a point to explicitly exclude objects or goals from their formulations (e.g., Ajzen & Fishbein, 1980, pp. 2930). This perspective fits the theoretical stance of attitude theory (see below), but to the

50

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

extent that decisions and behavior are goal-directed, it likely provides an incomplete explanation of action, in particular many consumer actions which are goal-directed (Bagozzi & Dholakia, 1999). Thus, the incorporation of anticipated emotions in the sense presented above introduces goal criteria into consumer decision making as proximal determinants of volitional processes. A second difference between attitudes and anticipated emotions concerns the theoretical processes underlying the effects of the two classes of variables. Attitudes are presumed to be relatively stable, learned predispositions that are retrieved or activated to influence decisions or behavior (Eagly & Chaiken, 1993; Fazio, 1995). Attitudes are typically constant over reasonable periods of time and are not formulated as responses contingent on the occurrence or possible occurrence of happenings to be appraised. Rather than active deliberations, attitudes are passive reactions. By contrast, the processes behind the functioning of anticipated emotions are dynamic and entail self-regulation in anticipation of achievement and failure of particular consumption goals (Carver & Scheier, 1998). That is, one has a consumption-related goal, appraisals are made of expected consequences of attaining and not attaining the goal in a particular situation to be appraised, and positive emotions arising as a result of anticipated goal attainment (e.g., joy) and negative emotions as a result of anticipated goal failure (e.g., disappointment) motivate one to act so as to promote goal achievement and avoid goal failure. An implicit comparison is made between one's goal as a standard or reference value and achieving and failing to achieve the goal under specific situational contingencies. The third distinction between attitudes and anticipated emotions occurs at the level of measurement and its implications for how the variables form and exist in human memory. When people are asked to express their attitudes, they are forced to make a choice of favorability or unfavorability. This is a consequence of the common practice of using bipolar items to indicate the respondent's attitudes. In contrast, anticipated emotions are measured with unipolar items (e.g., the experience of excitement is indicated along a not at all to very much continuum to capture the extent to which one feels this positive emotion). Attitude items in effect assume that attitudes are positive or negative by definition, whereas emotion items permit respondents to reveal whatever they are feeling, which can expose independence or novel associations between positive and negative emotions (e.g., Bagozzi, Wong, & Yi, 1999). Previous research has found that both positive and negative anticipated emotions influence decision making and are positively correlated, yet clearly distinct emotions (e.g., = .50 in Bagozzi et al., 1998). This is a consequence of the prefactual processes where respondents are asked to consider how negatively they would feel if they failed to achieve their goal. The more negative the expressed feeling, the more motivated respondents were to put forth effort and hence the positive association between positive and negative anticipated emotions (see Bagozzi et al., 1998). A second broad limitation of the TPB in a psychological sense that we wish to overcome in the present study with respect

to psychological processes is the following. It has been argued that the TPB fails to consider how intentions become energized (Bagozzi, 2006; Calder & Ross, 1973, p. 7; Fazio, 1995, pp. 271272). The antecedents in the TPB provide rational reasons for acting, it is claimed, and may be significantly correlated with intentions, but do not incorporate explicit motivational content needed to induce an intention to act. A consumer, for example, may value or be attracted to a luxury automobile (i.e., have a positive attitude toward it), and may have the means to acquire it, but may not want the automobile and therefore not form an intention to buy it. Consumers often have favorable attitudes, feel normative pressure, and have the wherewithal to make a purchase, but these reasons for acting are not necessarily owned by them so to speak and may lack motivational impact. In other words, a consumer may not be committed to or have conviction in his/her rational reasons for acting. Similar to Gollwitzer's designation for wants and wishes with regard to goals in his action phases model (Gollwitzer, Heckhausen, & Stellar, 1990, p. 1119), Bagozzi (1992, pp. 184 186) proposed that desires provide the motivational impetus for intentions to act and suggested that attitudes and other antecedents in decision making work through desires enroute to influencing intentions. This happens in one or both of two ways, depending on the context and person. One, with volitive desires, reasons for acting are taken into account to form a selfcommitment to act. Some philosophers maintain that desires have a particular kind of relationship to intentions in the sense that, once one is aware of and accepts his or her desire to act, this will motivate him or her to form an intention to act. Davis (1984a, p. 53) calls this the connection condition for intentions. Two, with appetitive desires (Davis, 1984b), reasons for acting serve as catalysts to release or free-up a hidden or latent desire related to such biological needs as food, sex, or safety, in contrast to their arousing function for volitive desires. Desires, then, transform reasons for acting into self-regulated motives to act which, in turn, are the proximal causes of intentions. Recent research demonstrates that volitive desires are distinct from intentions and attitudes (Perugini & Bagozzi, 2004), and has shown appetitive desires to play an important role in a variety of consumer behaviors (Belk, Ger, & Askegaard, 2003). To recap, the main dependent variables in our framework of small group brand community participation are the frequency of engagement in shared activities with one's brand community (group behavior) and also brand behavior, which is defined as the frequency of on-going purchase of Harley-Davidson motorcycle accessories, the amount of money spent on these accessories, and the number of shopping trips to dealers selling accessories (see Fig. 1).2 To explain these behaviors of the consumer, we began with the TPB but found the need to refer to
2 It is important to point out that such a conceptualization of brand behavior is different from past behavior. Rather, we conceptualize (and operationalize) brand behavior as ongoing behavior of the customer regarding the brand. Note that such a view is hypothesized to account for the specific social influence of the small group brand community on the individual member's personal actions.

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

51

social intentions and to introduce social identity (cognitive selfawareness of group membership, affective commitment, and evaluative significance) as a key antecedent. Anticipated positive and negative emotions were also imported to better account for the affective implications of achieving and failing to achieve one's group participation goals. Finally, desires were added to the TPB as key mediators between the reasons for acting and social intentions to act. Taken together, our proposed framework provides an in-depth description of the social and psychological processes that lead to participation in the joint activities of small group brand communities. 3.3. Methodological considerations This study is a survey, and to address challenges to validity we took a number of precautions. First, rather than just testing a saturated model where everything is related to everything, we performed formal tests of mediation wherever appropriate. For example, in our framework, attitudes are expected to influence social intentions but only through desires. To test this hypothesis, we compared the model shown in Fig. 1 to the model where a direct path from attitude to social intentions was added. The comparison was done with a 2-difference test and implements a procedure essentially equivalent to the one recommended by Baron and Kenny (1986) for mediation under multiple regression models. In all, given the hypothesized model, there are a total of 11 such rival hypotheses to test in each consumer community, for a total of 22 formal tests of mediation. Second, we tested the generalizability of the measurement models for two consumer community samples, one conforming to our definition of a small group brand community, and a second sample of consumers belonging to riding clubs not organized around any particular brand. This ascertains whether or not the correspondence between latent variables and their hypothesized measurements are invariant across the two samples, which is a test of the generalizability of the psychometric properties of the measures. Third, given a satisfactory test of measure invariance, we test boundary conditions between the models. If the social and psychological processes proposed above are valid, then we would expect stronger relationships among the paths shown in Fig. 1 for the small group brand communities organized around the Harley brand than non-brand-specific small group consumer communities. This is a test of moderation or interaction and thus explores boundary conditions. The effects proposed in Fig. 1 are expected to be contingent on the nature of the consumer community under study, as specified above. This aspect of our study constitutes a form of quasi-experimentation. To summarize, we bolster the survey findings in this research by testing for rival hypotheses, measurement generalizability, and interaction effects. 4. Empirical study 4.1. Participants and procedures (Harley group) Extensive interviews were conducted with three members of a local HOG chapter, including one member active with

leadership and administrative tasks, one highly involved as a participant, and one low to moderately involved. Notes taken from these interviews were used to adapt existing scales and to develop new items for the questionnaire. We re-contacted the first two members on a number of occasions to clarify issues and get feedback. The interviewees also gave feedback on an early draft of the questionnaire. After securing permission from the board of this HOG chapter, a questionnaire with an accompanying cover letter was sent to all its members. Although the chapter had nearly 600 members on its roster at the time of mailing, it is estimated that only 180 or so were active and belonged to small group brand communities.3 Of these, 161 members returned questionnaires, and 154 filled out all items. Early in the questionnaire, respondents were asked, Imagine that you are participating in a club activity, such as riding together. You have a number of club friends that you regularly engage in activities together with. Please picture briefly in your mind each of these club friends. Then write your first name, and their first names in the table below. Please be sure to include only club friends that are part of the group that engages in activities together with you now and then: The table contained 11 rows, corresponding to the respondent's own name and 10 Riding Club Friend X's name, plus next to each name was a space asking for the first name of a Passenger, if any. Thus, respondents could write up to 21 names, which permitted us to capture most small groups, as preliminary interviews with club members indicated that the largest riding groups typically contained 11 drivers or less, plus a few passengers. Respondents were assured that their responses will be kept strictly anonymous and confidential. As an incentive, the cover letter from the researchers indicated that a check for $500 would be donated to their HOG chapter to use in any way it sees fit. Table 1 provides the measures used in the study. Sample characteristics of this group were as follows. One hundred and fourteen were men (74%), and 40 were women (26%). Ages of respondents ranged from 23 to 73 (mean = 47.5, S.D. = 9.5). Riding experience ranged from 1 to 45 years (mean = 18.3, S.D. = 13.4). Respondents had on average been members of the local HOG chapter for 4.7 years (S.D. = 3.8). Fifty-two percent of respondents had a spouse who was also a member. On average, respondents rode with 6.1 other bikers (S.D. = 3.5) plus 3.0 passengers (S.D. = 2.2), resulting in an average small group size of approximately 9 people. 4.2. Participants and procedures (non-Harley group) These participants were members of the Southern Cruisers Riding Club (SCRC), a motorcycle club for riders of any
3 This estimate is based on personal communication with Mrs. Janet Nightingale, editor of the local HOG chapter monthly newsletter. The remaining (inactive) members contain a number of first-time members who receive a one-year complimentary membership as a result of purchasing a Harley Davidson motorcycle, but who fail to attend meetings or join small riding groups during the first year. A large number of other members either are loners, belong to other non-Harley groups where they spend most of their recreation time, or for whatever reason are not engaged with the HOG chapter or a riding group.

52 Table 1 Details of measures in the study Constructs and measures

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561 Table 1 (continued) Constructs and measures Evaluative social identity (two measures) I am a valuable member of the group I indicated above (7-point does not describe me at alldescribes me very well scale) I am an important member of the group I indicated above (7-point scale; same as first measure) Desires (three measures) I desire to participate in activities such as riding or meeting together with the group of my club friends I mentioned above during the next one month (7 point agreedisagree scale) My desire for participating in activities such as riding or meeting with the group of my club friends I mentioned above during the next one month can be described as (7-point very weak desirevery strong desire scale) I want to participate in activities such as riding or meeting together with the group of my club friends I mentioned above during the next month (7point does not describe me at alldescribes me very much scale) Social intentions (two measures) I intend that our group [i.e., the group of club friends you identified above] participate in activities such as riding or meeting together sometime during the next one month (5-point strongly agreestrongly disagree scale) We [i.e., the group of club friends identified above] intend to participate in activities such as riding or meeting together sometime during the next one month (5-point strongly agreestrongly disagree scale) Brand identification (single measure) How would you express the degree of overlap between your own personal identity and the identity of Harley-Davidson as you interpret it? (8-point graphical not at allvery much scale) Group behavior (two measures)1 About how many outings do you think you will participate in with the group of club friends you mentioned above in the next month? (Openended response scale) About how many outings do you think you will participate in next year in a typical month? (Open-ended response scale) Brand-related behavior (three measures)1, 2 How often do you purchase motorcycle accessories for your Harley? (Seven-item once a yearseven or more times a year scale) About how much money do you spend per year on motorcycle accessories for your Harley? (Open-ended response scale) About how often do you visit Harley bike dealer showrooms in a year? (Open-ended response scale)
1

Attitudes (four measures) On the following scales, please express your attitude toward participating in activities such as riding or meeting together with the group of club friends you identified above sometime during the next one month. (7-point scales) Foolishwise Harmfulbeneficial BadGood PunishingRewarding Positive anticipated emotions (nine measures) If I am able to participate in activities such as riding or meeting together with the group of club friends I identified above during the next month, I will feel (7-point not at allvery much scales) Excited Delighted Happy Glad Satisfied Proud Self-assured Relief Contentment Negative anticipated emotions (twelve measures) If I am unable to participate in activities such as riding or meeting together with the group of club friends I identified above during the next month, I will feel (7-point not at allvery much scales) Angry Frustrated Guilty Ashamed Sad Disappointed Depressed Worried Uncomfortable Anxious Agitated Nervous Subjective norms (two measures) Please express how strongly most people who are important to you feel you should or should not participate in activities such as riding or meeting together with the group of club friends you normally ride with (7-point scales) Most people who are important in my life think I (circle appropriate number): should 1: 2: 3: 4: 5: 6: 7: should not ride or meet with my club friends sometime during the next one month. Most people who are important to me would (circle appropriate number): approve 1: 2: 3: 4: 5: 6: 7: disapprove of me participating in activities such as riding or meeting together with my club friends sometime during the next one month Perceived behavioral control (two measures) How much control do you have over participating in activities such as riding or meeting together with the group of club friends you identified above during the next one month? (7-point no controltotal control scale) For me to participate in activities such as riding or meeting together with the group of club friends I mentioned above during the next month is (7point difficulteasy scale) Cognitive social identity (two measures) How would you express the degree of overlap between your personal identity and the identity of the group of club friends you mentioned above when you are actually part of your group of club friends and engaging in group activities such as riding, meeting together, etc.? (8-point graphical not at allvery much scale) Please indicate to what degree your self-image overlaps with the identity of the group of your club friends as you perceive it (7-point not at allvery much scale) Affective social identity (two measures) How attached are you to the group you mentioned above? (7-point not at all attachedattached very much scale) How strong would you say your feelings of belongingness are toward the group you indicated above? (7-point not at all strongvery strong scale)

To improve distributional properties, natural log transformations of these measures were used in the analysis; 2For the non-brand-specific rider sample, the word Harley was replaced with motorcycle.

motorcycle brand. The SCRC currently has over 12,000 active members with chapters all over the United States, and a strong history of using the Internet to organize club activities.4 The questionnaire was similar in content and format to the one sent to the Harley group and sought information regarding respondents' rider groups. It was placed on the SCRC web site (www.southerncruisers.net). Members were requested to participate in the survey to help club officers to better understand members' needs, wishes, and ideas regarding the
4 The SCRC's founder credits its web-site as playing a very important role in increasing its popularity and the rapid multiplication of SCRC chapters throughout the United States (see http://www.southerncruisers.net/history.htm for details).

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561 Table 2 Means, standard deviations, and reliabilities of scales Scale Harley small group community Mean Attitudes Subjective norms Positive anticipated emotion Negative anticipated emotion Desires Cognitive social identity Affective social identity Evaluative social identity Perceived behavioral control Brand identification Social intention Group behavior Brand-related behavior
1

53

Non-Harley rider communities Reliability1 .94 .87 .95 .95 .93 .90 .91 .96 .57 .90 .74 .62 Mean 5.59 5.77 4.97 2.24 6.07 4.21 4.88 4.74 4.86 2 4.32 .95 2.01 Standard deviation 1.05 1.45 1.39 1.19 1.03 1.62 1.57 1.92 1.48 .90 .60 .63 Reliability .88 .80 .91 .90 .85 .88 .87 .94 .62 .82 .51 .55

Standard deviation 1.15 1.50 1.59 1.08 1.43 1.71 1.64 1.89 1.64 2.04 .93 .73 .66

5.31 5.76 4.59 1.82 5.19 4.14 4.61 4.30 4.71 4.79 3.86 1.02 2.15

Note that reliabilities are calculated based on all the measures used in the study. As an example, the reliability of attitudes is based on its four measures (see Table 1). In cases where the construct has two measures, Pearson rs are reported instead of Cronbach's s. 2Brand identification was not measured for the non-Harley rider communities.

club's functioning. In addition, the club's national membership director sent e-mails to all chapter officers requesting them to encourage their members to participate in the survey. The survey, using the measures in Table 1, was available on the SCRC web-site for a period of two weeks. Participation was voluntary and no monetary incentives were given to respondents. As a token of appreciation, the researchers developed a detailed marketing research report for the SCRC leadership and helped them in the creation of a marketing program to attract new members. Characteristics of this sample were as follows. Of the 298 completed surveys received, 249 (83.6%) were men, while 49 (16.4%) were women. Ages of respondents ranged from 20 to 67 (mean = 43.2 years, S.D. = 9.4). Riding experience ranged from 1 to 51 years (mean = 16.6 years, S.D. = 13.4). Respondents had been members of SCRC for an average of 1.4 years (S.D. = 1.1). Forty one percent of respondents had a spouse who was also a member. On average, respondents rode with 6.5 other riders (S.D. = 3.4) resulting in a group size of approximately eight members. Details regarding passengers were not collected from this group. Forty-three (14.4%) respondents owned a Harley-Davidson, 112 (37.6%) owned a Honda, 44 (14.8%) owned a Kawasaki, and 70 (23.5%) owned a Yamaha motorcycle. Other brands mentioned less frequently included BMW, Triumph, Guzzi, and Indian. Harley owners were removed from the SCRC sample prior to analysis.5 4.3. Test of hypotheses Confirmatory Factor Analysis (CFA) and Structural Equation Models (SEM) were used to test our proposed theoretical
5 Note that this was done to prevent the possibility that the SCRC Harley riders could concurrently belong to Harley small group brand communities, and their survey responses could be colored by any such membership. Our results do not differ materially whether or not we include this sub-sample in the analysis.

framework. The LISREL 8.54 program was employed for this purpose (Jreskog & Srbom, 1999). The goodness-of-fit of the estimated models was assessed with 2 tests, the root mean square error of approximation (RMSEA), the non-normed fit index (NNFI), and the comparative fit index (CFI). Discussions of these indices may be found in Bentler (1990), Browne and Cudeck (1993), and Marsh, Balla, and Hau (1996). Satisfactory model fits are indicated by non-significant 2 tests, RMSEA values less than or equal to .08, and NNFI and CFI values greater than or equal to .90. Tests of mediation were performed with 2difference tests. Two indicators were used to operationalize each latent variable, except for brand identification where only one item was available. For latent variables with more than two items, these were combined to produce two indicators, using the so-called partial disaggregation model (Bagozzi & Heatherton, 1994). Item parcels were created by averaging multiple measures (e.g., the first five measures of positive anticipated emotions were averaged to form the first indicator, and the remaining four measures were used to form the second indicator). This approach yielded models with less parameters to estimate and reasonable ratios of cases to parameters, while smoothing out measurement error to a certain extent. All analyses were performed on covariance matrices (Cudeck, 1989). 4.4. Results 4.4.1. Reliability Table 2 summarizes the means, standard deviations, and reliabilities for the 11 scales with multiple measures separately for the Harley small group brand communities and the nonbrand-specific riding groups. In most cases, the reliabilities are much above the minimum of 0.70 recommended in the literature (Nunnally, 1978). Exceptions occur for perceived behavioral control (r = .57 for Harley and r = .62 for non-Harley group members), brand-related behavior ( = .62 for Harley and = .55 for non-Harley group members) and group behavior (r = .51 for non-Harley group members).

54

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

4.4.2. Discriminant validity For the Harley and the non-Harley samples, we separately evaluated discriminant validity using two different approaches. First, CFA models were built for each sample in LISREL. The Harley brand community sample had 13 latent constructs and 25 measures, whereas the non-Harley group had 12 constructs and 24 measures (brand identification was not available for this group). Results, as interpreted by the goodness-of-fit measures, showed that the models fit the data well in both cases. For the Harley brand community sample, the goodness-of-fit measures were as follows: 2(198) = 238.65, p .02, RMSEA = .03, SRMR = .04, NNFI = .99, CFI = .99. For the non-Harley group, the goodness-of-fit measures were: 2(186) = 245.80, p .00, RMSEA = .03, SRMR = .04, NNFI = .98, CFI = .99. The matrices (correlations between constructs, corrected for attenuation) for the two samples are provided in Table 3. As a first test of discriminant validity, we checked whether the correlations among the latent constructs were significantly less than one. Since none of the confidence intervals of the values ( two standard errors) included the value of one, this test provides evidence of discriminant validity. Second, for each pair of factors, we compared the 2-value for a measurement model constraining their correlation to equal one to a baseline measurement model without this constraint. A 2-difference test was performed for each pair of factors (a total
Table 3 -matrix of latent constructs for Harley small group brand community sample GRPB GRPB BRNB SINT DES BRNI CSI ASI ESI ATT PAE NAE SN PBC 1 .31 .52 .40 .12 .23 .44 .50 .13 .28 .21 .22 .34 BRNB 1 .16 .14 .39 .20 .19 .15 .06 .02 .02 .12 .01 SINT DES BRNI CSI

of 76 tests in all for each sample), and in every case resulted in a significant difference, again suggesting that all measures of constructs in the measurement model achieve discriminant validity. 4.4.3. Path model for Harley small group brand community sample Fig. 2 presents the findings for the path model for the Harley Brand Community Groups (HBCGs). This model fits the data well: 2(272) = 507.06, p .00, RMSEA = .069, NNFI = .91, and CFI = .93. It can be seen, as hypothesized, that desire to participate in the small group brand community is significantly influenced by attitude (1 = .13, p b .05), positive anticipated emotions (2 = .23, p b .01), negative anticipated emotions (3 = .10, p b .05), subjective norms (4 = .24, p b .001), and social identity (7 = .43, p b .001). We have presented standardized parameter estimates for ease of interpretation, where it can be seen that social identity has the strongest impact and is about 1.8 times more powerful than the next largest determinants of participation desire (i.e., positive anticipated emotions and subjective norms). Attitudes and negative anticipated emotions have the smallest effects on participation desire, showing roughly half the impact of positive anticipated emotions and subjective norms and about one quarter the impact of social identity on participation desire, respectively. Reflecting the

ASI

ESI

ATT

PAE

NAE

SN

PBC

1 .81 .21 .53 .64 .63 .49 .54 .32 .49 .45

1 .27 .68 .76 .66 .57 .72 .37 .64 .36

1 .48 .30 .16 .30 .27 .08 .17 .01

1 .72 .59 .40 .55 .18 .55 .23

1 .84 .45 .59 .27 .52 .32

1 .37 .57 .21 .48 .46

1 .46 .27 .34 .28

1 .42 .45 .14

1 .18 .08

1 .23

-matrix of latent constructs for non-Harley rider community sample GRPB GRPB BRNB SINT DES CSI ASI ESI ATT PAE NAE SN PBC 1 .42 .17 .28 .30 .25 .23 .15 .10 .05 .08 .31 BRNB 1 .42 .34 .39 .40 .44 .14 .30 .22 .11 .33 SINT DES CSI ASI ESI ATT PAE NAE SN PBC

1 .63 .16 .43 .39 .38 .27 .23 .27 .51

1 .22 .66 .51 .43 .56 .33 .22 .45

1 .47 .40 .01 .32 .08 .10 .05

.80 .29 .48 .30 .20 .46

1 .25 .40 .22 .18 .52

1 .34 .35 .05 .20

1 .44 .07 .22

1 .02 .06 1 .39 1

GRPB = Group behavior, BRNB = Brand behavior, SINT = Social intentions, DES = Desire, BRNI=Brand identification CSI = Cognitive SI, ASI = Affective SI, ESI= Evaluative SI, ATT = Attitude, PAE = Positive anticipated emotions, NAE = Negative anticipated emotions, SN = Subjective norms, PBC = Perceived behavioral control. Significant at = .05 level. Note: All correlations are significantly less than 1.0.

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

55

Fig. 2. Findings for structural equation model: Harley riding group (N = 154). (Parameter estimates presented for structural model only for simplicity.)

effects of five significant antecedents, quite a bit of explained variance results for desire (R2 = .80). Also as hypothesized, desire had a strong influence on social intentions (1 = .78, p b .001), social intentions had a strong effect on group behavior (2 = .47, p b .001), and group behavior had a moderate effect on brand behavior (3 = .32, p b .01). The respective explained variances for these endogenous variables were .64, .22, and .10, respectively. In addition, as hypothesized, social identity had a significant impact on brand identification (8 = .24, p b .01). However, only 6% of the variance in brand identification was accounted for by social identity. This could be because, as Algesheimer et al. (2005) showed in their study, the consumer's relationship with the brand is often a function of a host of individualistic factors concerning the functional and symbolic value derived from its use. Our finding is that in addition to such factors, brand community identification also contributes to increasing the brand's value for the consumer through increasing the overlap between one's own image and the brand's image. Three hypothesized paths failed to reach statistical significance. Perceived behavioral control neither predicted desire (5 = .05, ns) nor social intentions (6 = .10, ns) significantly.6 Likewise, brand identification did not significantly affect brand behavior (4 = .01, ns).
6 Following a reviewer's suggestion, we also considered the possibility that perceived behavioral control functions as a moderator between attitudes or desires on the one hand and social intentions on the other. Regression models revealed the coefficients of interaction terms to be insignificant, suggesting this was not the case. Details of this analysis are available upon request.

All three components of social identity loaded at high levels on the second-order social identity factor. These loadings roughly provide an indication of the relative contributions of the components to overall social identity. It can be seen that the affective component is the strongest component (and by implication contributes the most to desire, 10 = .95, p b .001), while the evaluative component is somewhat less strong (11 = .88, p b .001), and the cognitive component the least strong (9 = .65, p b .001). Nevertheless, all three components of social identity are of consequence. 4.4.4. Tests of mediation for the HBCGs Table 4 shows the results for the formal tests of mediation implied by the hypotheses and the model in Fig. 1. The first row in the table displays the goodness-of-fit findings for the model as pictured in Fig. 2. This model serves as a baseline for 2difference tests of direct paths from antecedents to consequences, one and two steps removed. For example, the second row in Table 4 presents the 2-values for the model of Fig. 2 with a direct path from social identity to social intentions added. The difference in 2-values between the baseline model and this 2 model (d (1) = .66), with one degree of freedom, is a test of the significance of the added path. As this difference is not significant (p N .30), we may conclude that the direct path from social identity to social intentions is insignificant, and therefore desire mediates all the effects of social identity on social intentions, as hypothesized. Row 3 in Table 3 reveals that the direct path from attitude to 2 social intentions is non-significant (d (1) = .13, p N .70), and row

56

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561


2 that neither subjective norms (d (1) = .67, p N .30) nor 2 perceived behavioral control (d (1) = .34, p N .57) significantly affect group behavior. Thus, as anticipated, social intentions fully mediate the effects of subjective norms on group behavior (perceived behavioral control failed to predict either desire or social intentions and therefore there is nothing to mediate here).

Table 4 Summary of findings for Harley small group brand community sample Model 1. Baseline model: hypothesized paths (Fig. 1) Goodness-of-fit 2(272) = 507.06 p .00 RMSEA = .069 NNFI = .91 CFI = .93 2(271) = 506.40 2(271) = 506.93 2(271) = 506.90 2(271) = 506.51 2(271) = 506.58 2(271) = 503.68 2(271) = 506.58 2(271) = 506.83 2(271) = 506.90 2(271) = 506.39 2(271) = 506.72 Tests of hypotheses

2. Social identity intentions 3. Attitude intentions 4. Positive anticipated emotions intentions 5. Negative anticipated emotions intentions 6. Subjective norm intentions 7. Social identity group behavior 8. Attitude group behavior 9. Positive anticipated emotions group behavior 10. Negative anticipated emotions group behavior 11. Subjective norm group behavior 12. Perceived behavioral control group behavior

M1 M2: (2(1) = .66, p N .30) d M3 M1: 2 (d(1) = .13, p N .70) M4 M1: (2(1) = .16, p N .68) d M5 M1: (2(1) = .55, p N .46) d M6 M1: (2(1) = .48, p N .49) d M7 M1: (2(1) = 3.38, p N .07) d M8 M1: (2(1) = .48, p N .49) d M9 M1: (2(1) = .23, p N .66) d M10 M1: (2(1) = .16, p N .68) d M11 M1: (2(1) = .67, p N .30) d M12 M1: (2(1) = .34, p N .57) d

4 shows that the direct path from positive anticipated emotions 2 is likewise non-significant (d (1) = .16, p N .68). Thus, desire mediates all the effects of attitude and positive anticipated emotions on social intentions, as predicted in our proposed framework. Rows 5 and 6, respectively, summarize the tests of direct effects from negative anticipated emotions and subjective norms on social intentions. As anticipated, these paths are 2 2 non-significant (d (1) = .55, p N .46) and (d (1) = .48, p N .49), respectively), and hence desire fully mediates the effects of negative anticipated emotions and subjective norms on social intentions. The last six rows in Table 4 present the tests of mediation for social intentions. Rows 7 and 8 show the results for the tests of direct paths from social identity and attitude, respectively, on group behavior. It can be seen that the respective paths are non2 2 significant (d (1) = 3.38, p N .07), and (d (1) = .48, p N .49), and therefore, as hypothesized, social intentions fully mediate the effects of social identity and attitude on group behavior. Rows 9 and 10 display the findings for the tests of direct effects of positive and negative anticipated emotions on group behavior, respectively. The 2-difference tests reveal that these effects are 2 2 non-significant, as predicted (d (1) = .23, p N .66), and (d (1) = .16, p N .68), respectively). Hence, social intentions mediate all the effects of anticipated emotions on group behavior. Finally, rows 11 and 12 in Table 4 summarize the results for tests of mediation of subjective norms and perceived behavioral control, respectively, by intentions. It can be seen

4.4.5. Path model for non-Harley rider group members Fig. 3 presents the findings for the path model for the nonHarley Rider Groups (NHRGs). This model fits the data well: 2 (225) = 366.60, p .00, RMSEA = .046, NNFI = .94, and CFI = .95. It can be seen, as hypothesized, that desire is significantly influenced by attitude (1 = .24, p b .001), positive anticipated emotions (2 = .28, p b .001), subjective norms (4 = .11, p b .05), and social identity (7 = .39, p b .001). As with the HBCGs, social identity is the strongest determinant of desire, where it is approximately half again as impactful as attitude and positive anticipated emotions and four times as impactful as subjective norms. Contrary to hypotheses, neither negative anticipated emotions (3 = .01, ns) nor perceived behavioral control (5 = .07, ns) significantly influenced desire for the NHRGs. As forecast, social intentions were significantly affected by desire (1 = .78, p b .001) and perceived behavioral control (6 = .14, p b .01), group behavior was significantly influenced by social intentions (2 = .37, p b .001), brand behavior was significantly determined by group behavior (3 = .33, p b .01), and brand behavior was significantly affected by social identity (8 = .17, p b .05). The amounts of explained variance in desire, social intentions, group behavior, and brand behavior were .58, .41, .14, and .15, respectively. These are lower than the values found for the HBCGs, except for brand behavior (compare Figs. 2 and 3; again, brand identification was not measured for the NHRGs because many different brands were purchased and coefficients associated with causes and effects of brand identification could not be interpreted as referring to a singular brand in this case). All three components of social identity for the NHRGs loaded highly on the second order factor. It can be seen in Fig. 3 that the affective component is again the strongest component (10 = .99, p b .001), while the evaluative component is somewhat less strong (11 = .85, p b .001), and the cognitive component is the least strong (9 = .57, p b .001). Nevertheless, all three components of social identity are of consequence. 4.4.6. Tests of mediation for the NHRGs Table 5 presents the findings for the formal tests of mediation implied by the hypotheses in Fig. 1 for the NHRGs. The first row again displays the goodness-of-fit results for the model as pictured in Fig. 3. Row 2 shows the results for the test of a direct path from social identity to social intentions, where it can be 2 seen that the path is non-significant, as hypothesized (d (1) = .37, p N .56). Desire fully mediates the effects of social identity on social intentions.

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

57

Fig. 3. Findings for structural equation model: non-Harley riding group (N = 255). (Parameter estimates presented for structural model only for simplicity.)

The third row in Table 5 reveals, contrary to hypotheses, that attitude has a direct effect on social intentions 2 (d (1) = 5.29, p b .05). Thus, given the additional finding of a significant effect of attitude on desire, we conclude that desire partially mediates the effects of attitude on social intentions. Rows 46 indicate that, as predicted, neither positive 2 anticipated emotions (d (1) = 1.50, p N .21), negative anticipat2 ed emotions (d (1) = .55, p N .48), nor subjective norms 2 (d (1) = 2.68, p N .10) significantly influence social intentions. Thus, given the results for the paths shown in Fig. 3, we conclude that desire fully mediates the effects of positive anticipated emotions and subjective norms on social intentions (negative anticipated emotions failed to affect social intentions, so there is no influence to mediate here). Row 7 in Table 5 reveals, contrary to expectations, that social identity has a direct effect on group behavior for the NHRGs 2 (d (1) = 9.07, p b .001). Therefore, social intentions only partially mediate the effects of social identity on group behavior. We find in row 8 in Table 5 that attitude does not 2 directly affect group behavior (d (1) = .99, p N .35). Hence, as predicted, social intentions fully mediate the effects of attitude on group behavior. As shown in row 9 in Table 5, positive anticipated emotions have a significant direct effect on group 2 behavior (d (1) = 4.54, p b .05). Thus, contrary to forecasts, social intentions only partially mediate the effects of positive anticipated emotions on group behavior. Finally, rows 1012 demonstrate that neither negative anticipated emotions 2 2 (d (1) = 1.38, p N .24), nor subjective norms (d (1) = .81, 2 p N .40), nor perceived behavioral control ( d (1) = 2.00,

p N .17) influences group behavior, consistent with hypotheses. In other words, social intentions fully mediate the effects of subjective norms and perceived behavioral control on group behavior (because negative anticipated emotions did not affect desire, there is nothing to mediate here). 4.4.7. Test of relative differences in key associations across HBCGs and NHRGs We also conducted a simultaneous confirmatory factor analysis for the latent variables common to the two groups. This twelve-factor model fit the data well: 2(374) = 595.06, p .00, RMSEA = .047, NNFI = .95, and CFI = .96. The test of equality of factor loadings showed that we cannot reject the 2 hypothesis of invariance: (d (12) = 19.99, p N .05). Table 6 presents the results for tests of differences in correlations between the correlated factors corresponding to the key paths in Fig. 1. It can be seen that eight of the eleven correlations are statistically different across groups, with the values corresponding to the HBCGs greater than the NHRGs in every case. The correlation between cognitive identity and desire and between subjective norms and desire is especially greater for the HBCGs than NHRGs (.72 vs .14 and .69 vs .22, respectively). This suggests that cognitive identity and subjective norms have much greater effects on desire for the HBCGs than NHRGs. Nevertheless, the six other correlations also point to greater impacts for the respective paths for the HBCGs versus NHRGs. Three correlations did not differ significantly between the two groups: perceived behavioral control with desire (.13 vs .30), negative anticipated emotions

58

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

Table 5 Summary of findings for non-Harley rider community sample Model 1 Baseline model: hypothesized paths (Fig. 2) Goodness-of-fit 2(225) = 366.60 p .00 RMSEA = .046 NNFI = .94 CFI = .95 2(224) = 366.23 2(224) = 361.31 2(224) = 365.10 2(224) = 366.05 2(224) = 363.92 2(224) = 357.53 2(224) = 365.61 2(224) = 362.06 2(224) = 365.22 2(224) = 365.79 2(224) = 364.60 Tests of hypotheses

2. Social identity intentions 3. Attitude intentions 4. Positive anticipated emotions intentions 5. Negative anticipated emotions intentions 6. Subjective norm intentions 7. Social identity group behavior 8. Attitude group behavior 9. Positive anticipated emotions group behavior 10. Negative anticipated emotions group behavior 11. Subjective norm group behavior 12. Perceived behavioral control group behavior

M2 M1: (2(1) = .37, p N .56) d M3 M1: 2 (d(1) = 5.29, p b .05) M4 M1: (2(1) = 1.50, p N .21) d M5 M1: (2(1) = .55, p N .48) d M6 M1: (2(1) = 2.68, p N .10) d M7 M1: (2(1) = 9.07, p b .001) d M8 M1: (2(1) = .99, p N .35) d M9 M1: (2(1) = 4.54, p b .05) d M10 M1: (2(1) = 1.38, p N .24) d M11 M1: (2(1) = .81, p N .40) d M12 M1: (2(1) = 2.00, p N .17) d

with desire (.37 vs .33), and group behavior with brand behavior (.46 vs .37). 5. General discussion The findings indicate that customer participation in collective activities within small group brand communities can be explained by a combination of social and psychological variables. One of these variables captures an aspect of shared volition of group members and was termed social intentions. We argued that social intentions have hitherto been accorded relatively little attention by marketers, but are important for a variety of consumer behaviors, being particularly useful for characterizing participation in small group brand communities. We found social intentions to be strong proximal determinants of group behavior in our study. A second kind of social variable we considered in order to explain brand community activities was the social identity of brand community members with their specific small circle of friends with whom they interact frequently. We note that the construct of social identity as proposed here is a refinement of, and elaboration on, the notion of consciousness of kind, designated by Muniz and O'Guinn (2001) as a core component of brand community. It consists of three interrelated parts: the customer's self-awareness that s/he is a member of a particular brand community (cognitive identification), the attachment to, or feelings of belongingness with, the brand community (affective commitment), and the evaluative significance of

membership (sometimes called, collective or group-based selfesteem). Social identity in relation to one's small group brand community was an important indirect determinant of group behavior for both Harley brand community and the non-Harley rider club members. At the same time, the magnitude of effects was statistically stronger for the small group brand community versus the non-brand-specific rider club members. This result is noteworthy, providing evidence that customer communities organized in small groups around a brand engender greater social identification, when compared to similar communities of customers organized around the product category more generally. The drivers of this greater social identity are worth studying further in future research. Here, there are multiple possibilities in this regard. One is that customers in the Harley small group brand communities are more homogeneous in their psychographic characteristics (although demographically they were equally varied in our samples), leading to greater overlaps between the personal identities of members and the group in the former when compared to the later case. A second distinct possibility is that in the former case, the organizing brand serves to magnify or accentuate the similarities among group members leading them to experience stronger social identity. Equally interesting was the finding that the association between cognitive identification and desire was much greater for Harley brand community members than the non-brand-specific rider club members. This result suggests that awareness may go beyond identification with one's group to strongly influence the desire to, and ultimately actual participation in, group activities. We also studied subjective norms which reflect quite different social processes than social identity. They capture the experienced social pressure, are based on the need for approval and are interpersonal (as opposed to group-oriented) in focus. Both subjective norms and social identity were found to be important antecedents of group behavior. However, scrutinizing the findings within each of the two consumer communities we studied, we found social identity to be between about 2 to 3 times more impactful than subjective norms. Comparing the two samples, the effect of subjective norms was
Table 6 Differences in correlations between key variables Correlation Desire-intention Affective ID-desire Cognitive ID-desire Evaluative ID-desire Subjective norm-desire Perceived behavioral control-desire Attitude-desire Positive AE-desire Negative AE-desire Intention-group behavior Group behavior-brand behavior p b .05, p b .01, p b .001. Harley small group brand community .79 .79 .72 .72 .69 .13 .62 .72 .37 .50 .46 Non-Harley rider group .61 .63 .14 .51 .22 .30 .47 .57 .33 .36 .37

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561

59

much stronger for the Harley small group brand communities than for the non-brand-specific rider groups. These findings support the notion that small group brand communities are egalitarian associations. There is still another sense of social influence that we did not examine in the present study, group norms, or what Kelman (1974) termed, internalization. This could be an important aspect of social influence in small group brand communities (e.g., Bagozzi & Lee, 2002) and would seem to be a worthy topic for future research. In our study, we also found the need to deepen and broaden psychological explanation within the TPB. Desire was a key mediator introduced into the TPB to better account for how rational and other reasons for acting are transformed into a decision or intention to act. Building on recent work in philosophy, we constructed desire as a motivational variable functioning in an integrative and executive fashion to activate the social intention to act. Desires have only recently begun to be studied in this sense. Strong tests of mediation were performed in this study by the use of 2-difference tests. We found for the Harley small group brand communities that desire fully mediated the effects of all antecedents on social intentions shown in Fig. 1 (except for perceived behavioral control which had no significant direct or indirect effects and does not appear to be relevant for the brand communities under investigation). For the non-Harley rider communities, desire fully mediated the effects of attitudes on intentions but only partially mediated the effects of social identity, subjective norms, positive anticipated emotions, and negative anticipated emotions on intentions. It is unclear why desire performed as predicted for the former group and showed direct and indirect effects on intentions for the latter. One possibility is that, given the nature and focus of Harley brand communities, desires are more well-formed and crystallized for Harley brand communities than the non-Harley rider groups. Non-Harley group members may be less cohesive and relatively more individualistic and thus desires to participate in group activities may be less organized and focused. If true, this provides another instance of the brand community exerting greater social influence on its members than a community organized around a product category. The above explanation for the different outcomes for mediation by desires receives indirect support when we examine the non-hypothesized, rival direct effects on group behavior. Again, desire and intentions fully mediated all effects of antecedents on group behavior, as hypothesized and as depicted in Fig. 1, for Harley brand community members. But for non-Harley rider community members, we found that intentions fully mediated the effects of attitudes, negative anticipated emotions, subjective norms, and perceived behavioral control, and only partially mediated the effects of social identity and positive anticipated emotions on group behavior. The two direct effects on group behavior for non-Harley rider community members (by social identity and positive anticipated emotions) imply that these variables have non-volitional, automatic impacts, in addition to indirect volitional effects. As with desires, intentions for non-Harley members may not be

well-formed or fully crystallized or may be unstable. Whether this interpretation is correct, and what the conditions affecting the formation and stability of desires and intentions are, deserve further study. The other psychological content we introduced into the TPB concerns anticipated emotions. We found that positive and negative anticipated emotions influenced desire to participate in group activities significantly for Harley small group brand communities, but only positive anticipated emotions influenced desire for non-brand-specific rider groups. It seems that Harley versus non-brand-specific riders are more goal-directed and were bothered to a greater extent by the prospects of failing to achieve their group goals. The differential aspects of anticipated emotions may be a consequence of greater solidarity and affinity for the former versus the latter. Brand-related behaviors were found to be consequences of group behavior for both Harley small group brand communities and non-Harley rider communities, but the amounts of explained variance in both cases were low. This is likely because brand-related behaviors may be driven by individualistic and situational considerations to a significant extent, which are not considered in our analysis. Showing that brand-related behaviors of small group brand community members are influenced by their participation in the community is an important step toward the goal of legitimizing brand community building programs and placing them within the portfolio of marketing programs available to brand and marketing managers in consumer settings. It could also be that more refined and multidimensional measures of brand-related behavior could help to bolster the strength of the relationship between group and brand behavior. Purchase diaries or panel studies where brand-related behaviors are measured more frequently could also help in this regard. Brand identification was found to be dependent on social identity with one's small group brand community, but brand identification did not significantly predict brand-related behavior for Harley riders. This is probably due to measurement error, because brand identification, unlike the other constructs in the model, was indicated by a single item, making predictions based on it particularly susceptible to unreliability. Our findings regarding brand identification, along with those of the recent study by Algesheimer et al. (2005) raise interesting questions regarding the relationship between brand identification and brand community belonging and participation. On the one hand, as we found, social identity with the brand community can contribute to the participant's identification with the brand. Practically, this link suggests that brand community may be effective means of recruiting novice customers and making them attached and loyal to the brand. On the other hand, as Algesheimer et al. (2005) showed, brand identification can be an antecedent to the consumer's participation and affiliation with the community. Such a relationship implies that the brand community may be more suitable and effective for experienced and already-engaged customers of the firm. Are these two perspectives inconsistent? We believe that there is interplay between the consumer's relationship with the band and the brand community. It is possible that among others,

60

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561 Algesheimer, R., Dholakia, U. M., & Herrmann, A. (2005, July). The social influence of brand community: Evidence from German car clubs. Journal of Marketing, 64, 1934. Armitage, C. J., & Conner, M. (2001). Efficacy of the theory of planned behavior: A meta-analytic review. British Journal of Social Psychology, 40 (4), 471499. Bagozzi, R. P. (1992). The self-regulation of attitudes, intentions, and behavior [Special issue]. Social Psychology Quarterly, 55(2), 178204. Bagozzi, R. P. (2000). On the concept of intentional social action in consumer behavior. Journal of Consumer Research, 27(3), 388396. Bagozzi, R. P. (2006). Consumer action: Automaticity, purposiveness, and self-regulation. In N. Malhotra (Ed.), Review of marketing research, vol. 2. (pp. 342). Armonk, NY: Sharpe. Bagozzi, R. P., Baumgartner, H., & Pieters, R. (1998). Goal-directed emotions. Cognition and Emotion, 12(1), 126. Bagozzi, R. P., & Dholakia, U. M. (1999). Goal setting and goal striving in consumer behavior [Special issue]. Journal of Marketing, 63(4), 1932. Bagozzi, R. P., & Heatherton, T. F. (1994). A general approach to representing multifaceted personality constructs: Application to state self-esteem. Structural Equation Modeling, 1(1), 3567. Bagozzi, R. P., & Lee, K. -H. (2002). Multiple routes for social influence: The role of compliance, internalization, and social identity. Social Psychology Quarterly, 65(3), 226247. Bagozzi, R. P., Wong, N., & Yi, Y. (1999). The role of culture and gender in the relationship between positive and negative affect. Cognition and Emotion, 13(6), 641672. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51 (6), 11731182. Belk, R. W., Ger, G., & Askegaard, S. (2003). The fire of desire: A multisited inquiry into consumer passion. Journal of Consumer Research, 30(3), 326351. Belk, R. W., & Tumbat, G. (2002). The cult of MacIntosh. Salt Lake City, UT: Odyssey Films. Bendapudi, N., & Berry, L. L. (1997). Customer receptivity to relationship marketing. Journal of Retailing, 73(1), 1537. Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107(2), 238246. Bergami, M., & Bagozzi, R. P. (2000). Self-categorization, affective commitment and group self-esteem as distinct aspects of social identity in the organization. British Journal of Social Psychology, 39(4), 555577. Bratman, M. E. (1997). I intend that we J. In G. Homstrm-Hintikka, & R. Tuomela (Eds.), Contemporary action theory (pp. 4963). Dordrecht, the Netherlands: Kluwer. Brown, S., Kozinets, R. V., & Sherry Jr., J. F. (2003). Teaching old brands new tricks: Retro branding and the revival of brand meaning. Journal of Marketing, 67(3), 1933. Browne, M., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen, & J. S. Long (Eds.), Testing structural equation models (pp. 136162). Newbury Park, CA: Sage. Calder, B. J., & Ross, M. (1973). Attitudes and behavior. Morristown, NJ: General Learning Press. Carver, C. S., & Scheier, M. F. (1998). On the self-regulation of behavior. Cambridge: Cambridge University Press. Cova, B. (1997). Community and consumption: Towards a definition of the linking value of product or services. European Journal of Marketing, 31 (34), 297316. Crocker, J., Luhtanen, R. K., Blaine, B., & Broadnax, S. (1994). Collective self-esteem and psychological well-being among White, Black, and Asian college students. Personality and Social Psychology Bulletin, 20, 503513. Cudeck, R. (1989). Analysis of correlation matrices using covariance structure models. Psychological Bulletin, 105(2), 317327. Davis, W. A. (1984a). A causal theory of intending. American Philosophical Quarterly, 21, 4354. Davis, W. A. (1984b). The two senses of desire. Philosophical Studies, 45, 181195.

current strength of the brand, the importance of the product class, and the disposition of the consumer could all be potential moderators in determining the direction(s) and the strength(s) of the relationship between brand identification and brand community social identity. Further research is needed to examine this important issue. To summarize the results, small group brand communities were found to express their decisions in social intentions, which in turn influenced group behavior. In addition to the effects of the classic antecedents specified under the TPB (i.e., attitudes and subjective norms), we found that dynamic appraisals in the form of anticipated emotions and group identification processes in the form of social identity were key explanatory variables. All antecedents were further found to work through desires enroute to their effects on decisions and behavior. In concluding, it is important to position the present research with respect to the large and vibrant research body on customer satisfaction, brand loyalty, and customer equity (e.g., Dick & Basu, 1994; Oliver, 1999). Also based on attitudetheoretic models, this research has shown customer loyalty to be explained by consumers' product-related cognitions (e.g., expectations and evaluations of performance), commitment toward the brand, and identification with the brand's image (e.g., Bendapudi & Berry, 1997; Fullerton, 2003; Rust, Lemon, & Zeithaml, 2004). Rather than supplanting it, our research complements and adds to these analyses of customer loyalty. In particular, we proposed and found that in addition to cognitive and motivational variables related to the product and the brand that these studies have examined, commitment to a brand can also be influenced (positively) through encouraging interactions with small groups of like-minded customers and identification with the group in a social context offered (and sponsored) by the firm and the brand, but controlled and managed primarily by the consumers themselves (see also Escalas & Bettman, 2003). As such, encouraging participation in small group brand community may be yet another means of increasing loyalty and customer equity, in addition to the variables uncovered by the loyalty modeling research. Acknowledgements The authors would like to thank the Ann Arbor HOG Chapter, John Lonskey, Janet Nightingale, and Bill Jarhead Strebel and Rick Rickster Perry of the Southern Cruisers Riding Club, and the many anonymous motorcycle enthusiasts who participated in this study. When this paper was written Richard P. Bagozzi was both part-time, Cardiff Business School, Cardiff, Wales, and full-time at the Jesse H. Jones School of Management, Rice University. References
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior & Human Decisions Processes, 50(2), 179211. Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall.

R.P. Bagozzi, U.M. Dholakia / Intern. J. of Research in Marketing 23 (2006) 4561 Dholakia, U. M., Bagozzi, R. P., & Pearo, L. R. K. (2004). A social influence model of consumer participation in network- and small-group-based virtual communities. International Journal of Research in Marketing, 21(3), 241263. Dick, A., & Basu, K. (1994). Customer loyalty: Toward an integrated conceptual framework. Journal of the Academy of Marketing Science, 22(2), 99113. Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX: Harcourt Brace Jovanovich. Ellemers, N., Kortekaas, P., & Ouwerkerk, J. W. (1999). Self-categorization, commitment to the group and group self-esteem as related but distinct aspects of social identity. European Journal of Social Psychology, 29(23), 371389. Escalas, J. E., & Bettman, J. R. (2003). You are what you eat: The influence of reference groups on consumers' connections to brands. Journal of Consumer Psychology, 13(3), 339348. Fazio, R. H. (1995). Attitudes as object-evaluation associations: Determinants, consequences, and correlates of attitude accessibility. In R. E. Petty, & J. A. Krosnik (Eds.), Attitude strength: Antecedents and consequences (pp. 247282). Mahwah, NJ: Erlbaum. Fournier, S., McAlexander, J. H., Schouten, J. W., & Sensiper, S. (2000). Building brand community on the Harley-Davidson posse ride. Boston, MA: Harvard Business School Publishing. Fullerton, G. (2003). When does commitment lead to loyalty? Journal of Service Research, 5(4), 333344. Gleicher, F., Boninger, D. S., Strathman, A., Armor, D., Hetts, J., & Ahn, M. (1995). With an eye toward the future: The impact of counterfactual thinking on affect, attitudes, and behavior. In N. J. Roese, & J. M. Olson (Eds.), What might have been: The social psychology of counterfactual thinking (pp. 283304). Mahwah, NJ: Erlbaum. Gollwitzer, P. M., Heckhausen, H., & Stellar, B. (1990). Deliberative and implemental mind-sets: Cognitive tuning toward congruous thoughts and information. Journal of Personality and Social Psychology, 59(6), 11191127. Hogg, M. A., & Abrams, D. (1988). Social identification: A social psychology of intergroup relations and group processes. London: Routledge. Jreskog, K. G., & Srbom, D. (1999). LISREL8 user's reference guide, 2nd ed. Chicago: Scientific Software International. Kelman, H. C. (1974). Further thoughts on the processes of compliance, identification, and internalization. In J. T. Tedeschi (Ed.), Perspectives on social power (pp. 126171). Chicago: Aldine. Maffesoli, M. (1996). The time of the tribes: The decline of individualism in mass society. Thousand Oaks, CA: Sage. Marsh, H. W., Balla, J. R., & Hau, K. -T. (1996). An evaluation of incremental fit indices: A clarification of mathematical and empirical properties. In G. A. Marcoulides, & R. E. Schumacker (Eds.), Advanced structural equation modeling: Issues and techniques (pp. 315353). Mahwah, NJ: Erlbaum.

61

McAlexander, J. H., Schouten, J. W., & Koenig, H. F. (2002). Building brand community. Journal of Marketing, 66(1), 3854. Muniz Jr., A. M., & O'Guinn, T. C. (2001). Brand community. Journal of Consumer Research, 27(4), 412432. Nunnally, J. C. (1978). Psychometric theory, 2nd ed. New York, NY: McGrawHill. Oliver, R. L. (1999). Whence customer loyalty? (Special issue). Journal of Marketing, 63(4), 3344. Perugini, M., & Bagozzi, R. P. (2001). The role of desires and anticipated emotions in goal-directed behavior: Broadening and deepening the theory of planned behavior. British Journal of Social Psychology, 40(1), 7998. Perugini, M., & Bagozzi, R. P. (2004). The distinction between desires and intentions. European Journal of Social Psychology, 34(1), 6984. Perugini, M., & Conner, M. (2000). Predicting and understanding behavioral volitions: The interplay between goals and behavior. European Journal of Social Psychology, 30(5), 705731. Prahalad, C. K., & Ramaswamy, V. (2000). Co-opting customer competence. Harvard Business Review, 78(1), 7987. Richard, R., van der Pligt, J., & de Vries, N. (1995). Anticipated affective reactions and prevention of AIDS. British Journal of Social Psychology, 34 (1), 921. Rust, R. T., Lemon, K. N., & Zeithaml, V. A. (2004). Return on marketing using customer equity to focus marketing strategy. Journal of Marketing, 68(1), 109127. Schouten, J. W., & McAlexander, J. H. (1995). Subcultures of consumption: An ethnography of the new bikers. Journal of Consumer Research, 22(1), 4361. Tajfel, H. (1978). Social categorization, social identity, and social comparison. In H. Tajfel (Ed.), Differentiation between social groups: Studies in the social psychology of intergroup relations (pp. 6167). London: Academic Press. Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S. Worchel, & W. G. Austin (Eds.), Psychology of intergroup relations (pp. 724). Chicago: Nelson-Hall. Terry, D. J., & Hogg, M. A. (1996). Group norms and the attitude-behavior relationship: A role for group identification. Personality and Social Psychology Bulletin, 22(8), 776793. Tuomela, R. (1995). The importance of us: A philosophical study of basic social notions. Stanford, CA: Stanford University Press. Williams, R. L., & Cothrel, J. (2000, Summer). Four smart ways to run online communities. Sloan Management Review, 8191. Yates, B. (1999). Outlaw machine: Harley-Davidson and the search for the American soul. New York, NY: Broadway Books.

Das könnte Ihnen auch gefallen