ENGAGEMENT IN ONLINE BRAND COMMUNITIES AND MARKETING RESEARCH ONLINE COMMUNITIES (MROCs) By Brian J. Baldus A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Business Administration – Doctor of Philosophy 2013 i ABSTRACT ENGAGEMENT IN ONLINE BRAND COMMUNITIES AND MARKETING RESEARCH ONLINE COMMUNITIES (MROCs) By Brian J. Baldus The focus of this dissertation is to develop insights regarding the use of online brand communities (i.e., discussion forums, blogs, or other social media sites centered around a single brand or branded product) and marketing research online communities (MROCs) as strategic marketing assets to influence brand and community outcomes. This is one of the first academic research studies of MROCs. Technological advances in the last decade have fundamentally changed the tools available to marketers and enabled them to socially interact with a broader base of consumers. I developed and validated scales to measure the diverse motivations online brand community members have to participate in online brand communities using 5 rounds of quantitative and qualitative data collection from a total of 660 online brand community members. I found that there are 11 distinct motivations online brand community members have to interact with online brand communities. Additionally, I developed and tested a model to test the effects of fit between marketing activities and community member motivations to leverage social dynamics and influence brand and community outcomes using longitudinal surveys and secondary data from 256 members of 8 MROCs. There is strong support that online brand communities and MROCs can be used as strategic marketing assets to enhance brand assessments (i.e., brand commitment and oppositional brand loyalty), supportive brand behaviors (i.e., word-of-mouth, defending the brand, and willingness to pay a price premium) and ii supportive community behaviors (i.e., participation intentions and community participation). In addition, I also found counterintuitive results for the leveraging effects of marketing activities and engagement. Overall, this research contributes to the strategic marketing literature, marketing practice, marketing research firms, and marketing consultants by assessing the degree to which online brand communities and MROCs can be used as strategic marketing assets to influence a loose hierarchy of effects for brand and community outcomes. Furthermore, this research contributes to the relationship marketing literature by classifying the motivations online brand community members have to form relationships with firms, brands, and other community members. iii Copyright by Brian J. Baldus 2013 iv DEDICATION To my wife, daughter, family, friends, and colleagues, I am very grateful for your love, support, friendship, and mentoring throughout my time at Michigan State University. Miss Elizabeth, your faith, courage, laughter, and contagious smiles inspire us all. To the many doctors, nurses, and staff at Sparrow Hospital, Michigan State University’s Pediatric Hematology Oncology Group, and affiliated childhood leukemia programs of 2012-2013. Thank you for working tirelessly to save my daughter’s life in her battle with leukemia. v ACKNOWLEDGEMENTS Thank you to my committee (Dr. Roger Calantone, chair; Dr. Clay Voorhees, co-chair; Dr. Cornelia Droge; and Dr. Tomas Hult) for all their generous support, expertise, and mentoring throughout my time at Michigan State University. Thank you to the executives and managers of the firm who worked with me to collect the data for essay two. Thank you to Jacci Weber, Sam Herzing, and Cody Potter for your help as research assistants. Thank you to Jared Pratt and Scott DuHadway. vi TABLE OF CONTENTS LIST OF TABLES ....................................................................................................................... .vii LIST OF FIGURES ....................................................................................................................... ix General Overview ........................................................................................................................... 1 ESSAYS ...................................................................................................................................... 3 LITERATURE GAPS ................................................................................................................. 5 Gaps Addressed ......................................................................................................................11 Essay One...................................................................................................................................... 13 INTRODUCTION .................................................................................................................... 13 RESEARCH BACKGROUND ................................................................................................ 16 Behavioral, Psychological, or Sociological Segmentation of Online Brand Community Members .......................................................................................................................... 21 PROPOSITION DEVELOPMENT .......................................................................................... 29 Online Brand Community Engagement Typology................................................................ 29 METHODS AND RESULTS .................................................................................................... 37 Scale Development Procedure .............................................................................................. 37 Online Brand Community Engagement Typology Development ......................................... 58 Differences Across Online Brand Community Engagement Segments ................................ 62 Comparison of Predictive Ability of Online Brand Community Engagement Typology and Rival Role-Based Typology...................................................................................... 68 DISCUSSION ........................................................................................................................... 74 Managerial Implications ....................................................................................................... 77 Limitations and Future Research .......................................................................................... 79 Essay Two ..................................................................................................................................... 81 INTRODUCTION .................................................................................................................... 81 LITERATURE BACKGROUND ............................................................................................. 85 HYPOTHESIS DEVELOPMENT............................................................................................ 92 Marketing Activities.............................................................................................................. 93 Online Brand Community Engagement ................................................................................ 96 Leveraging Marketing Activities and Online Brand Community Engagement .................... 98 Loose Hierarchy of Effects from Psychological Sense of Community .............................. 102 METHODS ............................................................................................................................. 106 Perceptual and Secondary Measures ................................................................................... 106 Analysis and Results ............................................................................................................110 DISCUSSION ......................................................................................................................... 121 Managerial Implications ..................................................................................................... 125 Limitations and Future Research ........................................................................................ 127 Executive Summary and Learning Implications ......................................................................... 129 REFERENCES ........................................................................................................................... 133 vii LIST OF TABLES Table 1 General Overview: Brand Community Literature Research Gaps ................................... 7 Table 2 Essay One: Bases for Segmentation ............................................................................... 23 Table 3 Essay One: Summary of Steps For Brand Community Engagement Scale Development Process ..................................................................................................... 38 Table 4 Essay One: Online Brand Community Engagement Construct Definitions .................. 41 Table 5 Essay One: Scale Items, Descriptive Statistics, and Factor Loadings ........................... 48 Table 6 Essay One: Results of Measurement Model Assessment and Scale Statistics Final Validation Study Short Form Scale ................................................................................ 57 Table 7 Essay One: Paticipation Intentions Latent Class Regression Model—Results From Short Form of Engagement in Study 5 Final Validation Study ..................................... 59 Table 8 Essay One: Brand Purchase Intentions Measures .......................................................... 63 Table 9 Essay One: Differences in Brand Purchase Intentions Across Classes—Multivariate Results for MANOVA on Study 5 Final Validation Study Data .................................... 65 Table 10 Essay One: Differences in Brand Purchase Intentions Across Classes—Univariate Results for MANOVA on Study 5 Final Validation Study Data ................................... 65 Table 11 Essay One: Rival Role-Based Typology Measures (Fournier and Lee 2009) ................ 70 Table 12 Essay One: Frequency Table for Roles in Study 5 Final Validation Study Data ........... 71 Table 13 Essay One: Comparison of Engagement and Role-Based Typlogy at Predicting Participation Intentions Using Study 5 Final Validation Study Data ............................. 72 Table 14 Essay One: Classification Tables Using Study 5 Final Validation Study Data .............. 73 Table 15 Essay Two: TURF Analysis of Online Brand Community Engagement ....................... 97 Table 16 Essay Two: Summary of Surveys Completed ............................................................... 111 Table 17 Essay Two: Measurement Model Summary .................................................................112 Table 18 Essay Two: Correlation Matrix .....................................................................................115 Table 19 Essay Two: Structural Equation Models .......................................................................116 viii LIST OF FIGURES Figure 1 Essay One: Differences in Brand Purchase Intentions Medians Across Classes Study 5 Final Validation Study ...................................................................................... 66 Figure 2 Essay Two: Model of Leveraging Online Brand Community Engagement ................... 95 Figure 3 Essay Two: Regions of Significance Testing.................................................................119 ix General Overview $3.6 billion dollars were spent on social media ads alone in 2012, up 40 percent from 2011 representing roughly 10 percent of online advertising spending (Elkin et al. 2012). $3.6 billion underestimates the tremendous amount of money spent creating and maintaining online brand communities and marketing research online communities (MROCs). For example, in early 2012 General Motors cut its $10 million spending on Facebook ads, while maintaining their $30 million spending on generating content for their online community of followers. As digital media usage increases among consumers and expenditures by firms reach record levels, consumers and brands are increasingly forming social relationships online. A surprising variety of firms have utilized online brand-centered communities to form lasting relationships with their customers (e.g., WD-40, Quickbooks, Jones Soda, Nike Plus, Maker’s Mark Embassy, Spread Firefox, Lululemon, Dell Ideastorm, Harley Davidson; Moffitt 2008) and to generate insights into their customers. A brand community is a “specialized, non-geographically bound community, based on a structured set of social relationships among admirers of a brand” (Muniz and O'Guinn 2001, p. 412), and it has become increasingly important to consumers, marketers, and academic researchers. Marketers have developed and proliferated online brand communities as there has been a tremendous evolution of the technology to interact with consumers socially online and a rapid expansion of fast and inexpensive communication technology. Online brand communities are discussion forums, blogs, or other social media sites centered around a single brand or branded product. Marketing research online communities (MROCs) are online communities created for the purpose of observing customers and generating insights about customers. The 1 MROCs studied in this dissertation are branded private communities run by a marketing research firm. It has been estimated that 84 percent of Internet users have contacted or participated in online brand communities (Madupu and Cooley 2010). A worldwide study of Internet users found that 75 percent of regular Internet users visited brand web sites, 32 percent had joined an online brand community in the past six months and 18 percent had created an online brand community in the past six months (Hutton and Fosdick 2011, p. 569). Consumers utilizing these technological advances to discuss brands develop a sense of community, even if they have never met face-to-face (Muniz and O'Guinn 2001, p. 413). MROCs are “dedicated online communities for qualitative market research purposes” (Bortner et al. 2008) and can be either branded or unbranded (Yazbeck 2011). MROCs enable marketers to enter the “consumer’s world and allows for comfortable, convenient participation. Creating an intimate, natural space for consumers to relate to each other and generate their own discussions yields authentic and detailed insights that are likely to be missing from traditional, artificial approaches” (Austin and Schlack 2012, p. 82). One leading industry report has estimated that 45 percent of respondents (marketing research suppliers and clients) currently use and an additional 38 percent of respondents plan to use MROCs in the future (GreenBook Winter 2013, p. 22). The rise and prevalence of online brand communities and MROCs is an interesting and relevant phenomenon to consumers, marketers, and academic researchers because of the numerous benefits these communities can provide members and firms. For example, these communities can benefit consumers who are members by providing them with a sense of affiliation with others, social interaction, and valuable information regarding the brand and its 2 use (Martin 2009). Online brand communities can also provide numerous benefits to firms “at a fraction of the cost of traditional marketing programs” (Dholakia and Vianello 2009). For example, some of the benefits brand communities can provide firms are sustainable brand loyalty (Madupu 2006), oppositional brand loyalty (Madupu 2006; Thompson and Sinha 2008), regulation of community member behavior (Hogg et al. 1995), ideas for new products (Füller et al. 2008; Schau et al. 2009; Schouten and McAlexander 1995), test areas for new products (Dholakia and Vianello 2009), enhanced brand equity (Muniz and O'Guinn 2001), prompt and high-quality customer service (Dholakia and Vianello 2009), a stable customer recruiting platform, publicity, and word-of-mouth (Dholakia and Vianello 2009). Some firms have found online brand communities so valuable to the firm and consumers that they have fundamentally changed how they deliver post-sales service and support. For example, “unsourcing” is the practice of turning over technical support to online communities of consumers helping each other instead of employees or contractors in overseas call centers. It is estimated that unsourcing saves firms 50 percent on their customer support costs (Economist 2012). Best Buy estimates that it saves $5 million annually from its community (Economist 2012). Brand communities are also interesting and relevant to academic researchers because of the social dynamics and potential influence on consumer decision making. For example, academic researchers should explore mechanisms through which social dynamics in online brand communities could be leveraged to increase the efficiency of marketing activities. ESSAYS There have been numerous calls for additional research into how online brand communities and marketing research online communities (MROCs) affect firm performance (Andersen 2005; 3 Bagozzi et al. 2011; Cova and Pace 2006; Dholakia and Vianello 2011; Gabisch and Gwebu 2011; Hickman 2005; MSI 2012; Thompson 2009; Yu-Chen 2006). This dissertation is one of the first academic research studies of MROCs and helps explain the mechanisms through which online brand communities and MROCs can affect firm performance. Therefore, the focus of this dissertation is to look deeper into online brand communities and MROCs from a marketing strategy perspective by developing and testing a typology of online brand community members, quantifying the effect on brand purchase intentions (i.e., new product trial, new product adoption, and purchase intentions). In addition, this dissertation explores marketing activities for leveraging online brand communities to enhance brand assessments, supportive brand behaviors, and supportive community behaviors. The focus of essay one is on what motivates consumers to interact with online brand communities and what are the consequences of these motivations. Specifically, I developed a measure for online brand community engagement. Building on prior research, I define online brand community engagement as the compelling intrinsic motivations to continue interacting with an online brand community (Algesheimer et al. 2005, p. 21). As a stable emotional commitment, it propels individuals to continue interacting with the community because it is an aroused state. In other words, engagement is the feeling that drives people to keep interacting in the online brand community. I then used this measure of brand community engagement to develop a typology of online brand community members. After developing the typology, I examined differences in brand purchase intentions across engagement segments. Furthermore, I compared the predictive ability of this typology to a rival typology to assess its relative performance. Therefore, essay one contributes to the literature by developing a measure of 4 online brand community member engagement and creating an improved way to classify brand community members. The focus of essay two is on understanding the mechanisms through which online brand community engagement affects brand assessments, supportive brand behaviors, and supportive community behaviors. My premise is that a firm that understands the mechanisms through which online brand community engagement affects brand assessments, supportive brand behaviors, and supportive community behaviors can leverage the brand community to achieve strategic marketing objectives. Essay two addresses important gaps in the brand community literature by assessing how communities can be used by marketers to achieve strategic objectives like improved brand assessments, increased sales, and increased community participation. Furthermore, essay two contributes to the brand community literature by assessing the importance of marketing activities in MROCs and the motivations members have for interacting with online brand communities. “For the brand to serve as a legitimate relationship partner, it must… actually behave as an active, contributing member of the [community]” (Fournier 1998, p. 346). LITERATURE GAPS Brand communities have been the subject of several ethnographic studies (e.g., Muniz 1998; Muniz and O'Guinn 2001; Muniz and Schau 2007; Schau et al. 2009; Schouten and McAlexander 1995) and an increasing number of quantitative studies (e.g., Algesheimer et al. 2005; Bagozzi and Dholakia 2002; Carlson 2005; Dholakia et al. 2004; Hickman 2005; Madupu 2006; Martin 2009; McAlexander et al. 2002; Thompson and Sinha 2008). Several of these studies have examined motivations to participate in brand communities (e.g., Algesheimer et al. 5 2005; Dholakia et al. 2004; Madupu 2006). There has however been a dearth of research on the motivations online brand community members have to interact with online brand communities since the development of Web 2.0 (Gabisch and Gwebu 2011). There has been even less academic research on marketing activities marketers can use within brand communities to achieve financial and other important strategic objectives (see Table 1). More importantly, this is the first academic research that I am aware of that studies MROCs. Over a decade ago, Williams and Cothrel (2000) suggested that successfully managing online brand communities “will become a distinguishing feature of nearly every successful business” (p. 81). MROCs have had tremendous success in generating insights in a timely and efficient manner. For example, Colgate found that an conducting the same research in an MROC instead of conducting the same research traditional means community “would pay for itself within the first three or four months” (Austin and Schlack 2012, p. 82). Similarly, Kraft found that using MROCs to gain consumers insights earlier and throughout the new product development process was highly effective. Through using an MROC, Kraft created the “100 Calorie pack and a whole new product category along with it” (Austin and Schlack 2012, p. 82). Firms engaging in behavior contrary to what their online brand communities want frequently incur significant negative publicity. For example, Porsche’s introduction of the Cayenne SUV faced stiff opposition from community members who did not consider the Cayenne to be a real Porsche. Community members did not like the Cayenne’s noticeable design changes from traditional Porsche models (Fournier and Lee 2009). Brand communities have also outright rejected and actively opposed brands from making strategic changes. For example, GAP attempted to revitalize the GAP logo in 2010, but was forced to retract the new logo quickly due to vehement opposition from GAP brand community 6 members. Pfizer also got itself into a lot of negative publicity when they posted a ChapStick ad and then censored negative reactions from community members following the ad campaign (Bhasin 2011; Nudd 2011). As interest in and research on brand communities and MROCs continues to increase, more theoretical development needs to take place to better explicate how marketers can leverage brand communities to enhance brand performance. Understanding marketing activities marketers can use in brand communities is essential to successfully implementing brand communities as part of a firm’s relationship marketing strategy. Table 1 General Overview: Brand Community Literature Research Gaps Author(s) Fox (1987) Research Focus Studied the social structure of the punk subculture. Future Research/Gaps Ethnographic study of punk subculture lacks clear empirical measure of how to segment community members Schouten and McAlexander (1995) Studied Harley Davidson Owners Group as a subculture of consumers to understand how they use a brand as the basis of social interaction. How symbols of consumption are used, altered, or reinterpreted when embedded in a nonnative host culture with differing cultural categories and principles Muniz (1998) Studied the existence of brand communities and their key social processes and characteristics (consciousness of kind, moral responsibility, homogenous culture, rituals and traditions, symbols, and structure). Description of several online brand communities and created suggestions for how to manage them. Suggests that we need to update mass communications model with social influence in decision making. New media allows rapid dissemination of information and active involvement of audience. Williams and Cothrel (2000) 7 Explore community asset management (e.g., content, relationships, and commitment of members) Table 1 (cont’d) Author(s) Bagozzi and Dholakia (2002) Research Focus Studied text based community participation as intentional social action, compliance, internalization, and social identity processes of influence. Future Research/Gaps Perhaps consider using internalization (adoption of a decision based on the congruence of one’s values with the values of another) as a less frequently used mechanism relative to compliance and identification. McAlexander et al. (2002) Changed from triadic to consumer centric study of brand communities. Ethnographic and some quantitative aspects, suggests that future research needs to look more to product types/characteristics, brand characteristics, and links to other communities. Wang et al. (2002) Proposed framework of functional, psychological, and social needs related to community participation. Conduct an empirical test to framework proposed. Dholakia et al. (2004) Drivers of virtual community participation and typology of community structures. See how communities evolve over time. This study is one of the first into online brand communities. Wang and Fesenmaier (2004) Proposed and tested framework of functional, social, psychological and hedonic needs driving online community participation. Emphasized the way that the Internet was reshaping communication and redefining markets. Thus, as the Internet evolves, need to update this research. Algesheimer et al. (2005) Brand relationship is important driver of relationship with community and brand related outcomes in face-to-face settings. Private v. public community engagement behaviors may lead to different outcomes (reactance and other negative outcomes may be primarily due to public engagement behavior) Carlson (2005) Created and validated measures for perceived connection among users of Internet communities. Examine the effect community information has on non-community members. Is it more or less persuasive than advertising? 8 Table 1 (cont’d) Author(s) Madupu (2006) Research Focus Studied the effects of functional, psychological, social, and hedonic needs on individual level consequences of online brand community participation. Future Research/Gaps How are people participating in communities? Created a single dimensional measure of participation, but discuss multiple different aspects of participation. Muniz and Schau (2007) Netnographic study of consumer advertising in light of firm canceling product. Research consumer response to consumer generated content. Thompson and Sinha (2008) Studied online brand community effects on the adoption of new products. Relative roles of information exposure and identification in the influence of adoption behavior. Fournier and Lee (2009) Described brand community management myths, common community member roles, and the structure of communities. Need to capture customer centric view, assess structure of the community, the roles members play, offline as well as online important, and capture firm involvement. Martin (2009) Online brand collectives v. brand communities Suggests that there are important constructs missing from the model that would help explain antecedents and consequences of brand community membership. Schau et al. (2009) Review processes in communities that create value. Suggested that future research should look for generalizable drivers of brand community participation and value creation. Thompson (2009) Online community roles in new product adoption. Need to examine the generalizability of findings beyond products such as cars, computers, and personal digital assistants (e.g., consumer packaged goods, and less visibly consumed products). Bagozzi and Yi (2011) Studied the link between identities to actions and action tendencies Future research should employ longitudinal methods to study online brand communities 9 Table 1 (cont’d) Author(s) Gabisch and Gwebu (2011) Research Focus Impact of virtual brand experience on purchase intentions. Hutton and Fosdick (2011) Longitudinal study of online communities across countries. Future Research/Gaps Future research needs to explore how behavior is influenced by emerging Web. 2.0 technologies that enable enhanced functional and social interactivity. . Online brand communities will continue to be important areas to study because consumers are interested in interacting with brands in social settings online. Much of what is documented in the academic literature about online brand communities originated from face-to-face brand community studies. However, face-to-face brand communities are significantly different from online brand communities. Harley Owners Group, one of the most prominent examples of a face-to-face brand community illustrates the powerful social dynamics that take place during face-to-face brand community gatherings. At these faceto-face meetings, members spend several days interacting or traveling with each other, participating in activities, and exchanging personal information (Fournier 2000; McAlexander et al. 2002). The Harley Owners Group has developed such strong social dynamics that it has been described as a subculture (Schouten and McAlexander 1995). Unlike members of face-to-face brand communities, online brand community members may actually never meet or know the true identities of other community members (e.g., members create their own profiles and frequently use pseudonyms). Therefore, a community member’s experience in an online brand community is qualitatively different from the experience of a community member in a face-to-face brand community (McAlexander et al. 2002, p. 43). Given the variety of brands that online brand communities have been founded around 10 (e.g., motorcycles, cars, computers, apparel, diapers, services, sports teams, and beverages), and the selection of prominent communities for ethnographic study, the communities studied previously are generally not representative of face-to-face communities and are even less representative of online brand communities in general. Additionally, there are significant differences between online and face-to-face brand communities, which suggests that there could be entirely different motivations to participate in the two different types of brand communities. For example, online brand communities facilitate one to many interaction, enable minority groups to gain strength in numbers, and online settings typically have more relaxed social norms for conversation (Prykop and Heitmann 2006; Schlosser 2003, p. 192). Therefore, it is not clear how well a generic measure of engagement developed in face-to-face European car clubs will accurately assess the motivations of community members in the plethora of online brand communities. Gaps Addressed Cova and Pace (2006) state that “above and beyond check-lists drawn up by consultants claiming to be familiar with the keys to success in this area [brand communities], which they call ‘tribal branding’ or ‘cult branding’ …, it is clear that efforts have to be made to strengthen good marketing practices in this respect” (p. 1089). There are many anecdotal suggestions for how marketers can and should design and manage online brand communities based on a few visible examples of brand communities. This explosion of information has left marketers with numerous and sometimes contradictory explanations of how to design and manage a community (Fournier et al. 2005). Astute marketers will want to assess the effectiveness of marketing activities. 11 This research helps marketers measure online brand community engagement, a set of key metrics that has been poorly defined and measured in the past. It helps marketers evaluate how brand communities affect brand assessments (i.e., brand attachment, brand commitment, and brand identification), brand purchase intentions (i.e., new product trial, new product adoption, and purchase intentions), supportive brand behaviors (i.e., word-of-mouth, defending the brand, and willingness to pay a price premium), and supportive community behaviors (i.e., participation intentions and community participation). Segmentation based on online brand community engagement provides better prediction of brand assessments and brand purchase intentions than other classification techniques available. The online brand community engagement typology is very easy to deploy, requiring a single short survey to segment community members. Furthermore, this research helps marketers identify and efficiently engage in marketing activities to enhance effectiveness of online brand communities. Specifically, this research demonstrates that online brand communities and MROCs can influence a loose hierarchy of effects for brand and community outcomes. Additionally, this research proposes that marketers can leverage online brand communities by using marketing activities that fit with community member engagement. 12 Essay One INTRODUCTION Marketers must understand their brand community members’ engagement, or his/her intrinsic motivation to interact with the brand community (Algesheimer et al. 2005, p. 21), to achieve strategic marketing objectives (e.g., Fournier et al. 1998; Kumar et al. 2010). Technology continues to rapidly evolve, enabling consumers to have unique experiences with the brand and other brand users. These unique experiences facilitate the formation of relationships with the brand and other brand users (McAlexander et al. 2002). Online brand communities are discussion forums, blogs, or other social media sites centered around a single brand or branded product (McWilliam 2000; Muniz and O'Guinn 2001). This definition of an online brand community is consistent with the definition of an online community set forth by Dröge et al. (2010) with the added condition that the community must be centered around a single brand or branded product (p. 68). Prominent examples of online brand communities include Audi’s virtual lab, Dell’s Ideastorm, IBM’s numerous online communities, Intuit’s business community, Nike’s Facebook Fan Page, Proctor and Gamble’s Pampers Village, Salesforce.com’s support community, RedSox Nation, and Starbucks’ MyStarbucksIdea. Online brand communities represent a network of relationships between consumers and the brand, product, fellow consumers, and the marketer (McAlexander et al. 2002, p. 39). The networks of relationships in online brand communities provide marketers a valuable strategic relationship-marketing platform. Marketers trying to reap a variety of benefits are investing heavily and interacting frequently in online brand communities to increase brand and product awareness, customer loyalty, supplement consumption experience, and gain insights into consumers. 13 While each brand community has a unique purpose, the primary universal is that they represent an explicit marketing investment on behalf of the firm to develop long-term connections with their current and potential consumers. In order to increase returns on these substantial investments, marketers require better consumer insights into the motivations to participate in brand communities and the resulting attitudinal and financial benefits to the brand. Despite this practical need, academic research on the consumer motivations to participate in online brand communities has struggled to keep pace with the changing landscape of the industry. While early investigations in brand communities provide us with operational definitions of these investments: “Online brand communities represent a network of relationships between consumers and the brand, product, fellow consumers, and the marketer” (McAlexander et al. 2002, p. 39) and insight into early motivations for community engagement (Dholakia et al. 2004), they fail to capture the complexity of motivations driving consumer engagement in communities due to recent and rapid technological innovations and substantial investments in these communities by their brands. In the past 15 years, online brand communities have become accessible to a broader range of consumers as the Internet has become increasingly available worldwide. Online brand community members are no longer just innovators and heavy users, instead online brand communities are comprised of a broader range of consumers. Furthermore, during this same time, there has been a tremendous shift in the focus of the Internet from information storage and retrieval to a transactional system where consumers can not only learn about a brand and its products, but can also purchase them online. The web has also sustained tremendous growth in social networks and usage (e.g., LinkedIn, Facebook, YouTube, Twitter, Vimeo, MySpace, 14 Pinterest, and Instagram). Therefore, existing measures for motivations to participate in online brand communities developed almost a decade ago lack sufficient contextual specificity for the current state of online brand communities and are likely to have poor predictive ability (Warshaw 1980). To effectively allocate marketing resources, marketers need to understand the motivations consumers have to interact with online brand communities (MSI 2012). Following a grounded theory approach, I develop a measure of online brand community engagement. This research contributes to the brand community literature by providing a building block for causal analysis in online brand community research and way to classify online brand community members. Four research questions guide this essay. First, what motivations do people have for interacting with an online brand community? Following the Gerbing and Anderson (1988) update of the Churchill (1979) procedure for scale development, I develop scales to measure online brand community engagement. Second, how many segments of online brand community members are there? Third, how do brand purchase intentions differ across segments? And fourth, does a motivational typology of online brand community members predict community participation intentions better than a rival role-based typology? My findings provide important insights to academics, marketers, marketing research firms, and consultants in regards to the motivations of online brand community members. I show that online brand community engagement is not unidimensional, but is in fact multidimensional. Therefore, existing conceptualizations of engagement are far too narrow to capture the diverse motivations online brand community members have for interacting with a brand community. I identify 11 distinct motivations to interact with online brand communities. 15 Segmentation analysis reveals the emergence of two segments of online brand community members (“brand passionate helpers” and “individualistic information seekers”). Further analysis reveals important differences between segments in terms of their brand purchase intentions. For example, rewards (utilitarian) (e.g., cash, coupons, or discounts) motivate “brand passionate helpers” to participate but actually reduce participation for “individualistic information seekers.” This suggests that providing deals and discounts to community members may not actually increase participation as intended. Furthermore, the online brand community engagement typology demonstrates comparable ability to its rival role-based typology to predict participation intentions. The online brand community engagement typology, however, can be implemented at a fraction of the cost and facilitate rapid generation of insights for marketers. These insights can then be used to tailor the community experience and better utilize scarce marketing resources in the community. RESEARCH BACKGROUND Brand communities have been called the “Holy Grail for brand loyalty” (McAlexander et al. 2002, p. 38) because they are a place where marketers can find active and loyal customers. Some of the features that make brand communities especially attractive to marketers are that communities are explicitly commercial, relatively stable, and that members are generally strongly committed and feel a sense of moral responsibility to the community as well as fellow members (Muniz and O'Guinn 2001). These desirable characteristics make brand communities especially attractive to marketers trying to obtain loyal customers and a sustainable competitive advantage in the market. However, very little research has explored the motivations members have to interact with online brand communities. 16 Consumers’ motivations to interact with a brand community are essential in the formation of how they evaluate their experiences in online brand communities, as well as the consequences for those interactions. While ethnographic studies have explored and described the characteristics of brand communities, as a whole there has been very little work done on what motivates members to interact with online brand communities. Thus, a substantial gap in the brand community literature exists and needs to be addressed. Describing Online Brand Community Member Motivations to Interact Algesheimer et al. (2005) introduced the concept of brand community engagement to capture the member’s intrinsic motivation to interact with other community members (p. 21). Few other studies have looked at engagement (e.g., Hatch and Schultz 2010; Lee et al. 2011; Schau et al. 2009; Shih et al. 2010). However, several other studies have explored the antecedents to interaction with online brand communities (Dholakia et al. 2004; Madupu 2006; Madupu and Cooley 2010; Wang and Fesenmaier 2004; Wang et al. 2002), which are motivations to start interacting with an online brand community (Madupu and Cooley 2010). Dholakia et al. (2004) asserts that people start to interact in online brand communities for five reasons: purposive value, self-discovery, maintaining interpersonal connectivity, social enhancement, and entertainment. People who start to interact with online brand communities for purposive value are looking to accomplish “some pre-determined instrumental purpose (including giving or receiving information) through virtual community participation” (Dholakia et al. 2004, p. 244). People start to interact with online brand communities for self-discovery are interested in “understanding and deepening salient aspects of one’s self through social interactions” (Dholakia et al. 2004, p. 244). People who start to interact with online brand 17 communities for maintaining interpersonal connectivity are looking for “social benefits derived from establishing and maintaining contact with other people such as social support, friendship, and intimacy” (Dholakia et al. 2004, p. 244). Online brand communities also provide opportunities for social enhancement. People who start to interact with online brand communities for social enhancement want the “value that the participant derives from gaining acceptance and approval of other members and the enhancement of one’s social status within the community on account of one’s contributions to it” (Dholakia et al. 2004, p. 244). Lastly, Dholakia et al. (2004) states that people who start to interact with online brand communities for entertainment value are seeking “fun and relaxation through playing or otherwise interacting with others” (Dholakia et al. 2004, p. 244). Similarly, Wang et al. (2002), Wang and Fesenmaier (2004), and Madupu (2006) describe motivations to start to interact with an online brand communities using slightly different terminology and operationalizations from Dholakia et al. (2004). They argue that people are motivated to start to interact with an online brand community to fulfill functional, social, psychological, and hedonic needs. Functional needs are very similar to purposive value, and are defined as needs that “are met when community members go online to fulfill specific activities” (Wang and Fesenmaier 2004, p. 262). Examples of functional needs include information, efficiency, and convenience (Wang and Fesenmaier 2004, p. 265). Social needs are described as being based on the purpose of the community. Examples of social needs include communication, relationship, involvement, and trust (Wang and Fesenmaier 2004, p. 265). Psychological needs are described as the basic needs individuals have. Examples of psychological needs include affiliation, belonging, and identification (Wang and Fesenmaier 2004, p. 265). Hedonic needs are described as those centered around enjoyment and entertainment. Examples of hedonic 18 needs include entertainment, enjoyment, amusement, and fun (Wang and Fesenmaier 2004, p. 265). Collectively, these studies indicate that people start interacting with online brand communities to satisfy a variety of needs and specific objectives (Dholakia et al. 2004; Madupu 2006; Madupu and Cooley 2010; Wang and Fesenmaier 2004; Wang et al. 2002). The broad range of motivations to start interacting with online brand communities also suggests that community interaction could lead to a wide range of member behaviors and firm outcomes and that motivations to continue interacting with online brand communities could be numerous and diverse. Describing Brand Community Member Behavior Brand community member behavior has received relatively more attention in the literature than engagement. Several different conceptualizations and operationalizations of interaction in a brand community have emerged. For example, brand community member behavior has been viewed as simply interacting “with other members of the online community” (Madupu 2006, p. 31), word-of-mouth recommendations, leadership behavior in face-to-face brand related events (Algesheimer et al. 2005, p. 22, 33), functional roles in the community (Fournier and Lee 2009, p. 109), and in terms of the frequency of participation (Dholakia et al. 2004, p. 252). The diverse nature of interaction behaviors suggests a diverse set of motivations for interacting in a brand community. Reviewing the behaviors that have emerged in the literature will shed light on potential motivations for those behaviors. The first type of behavior one sees in an online brand community is that members actively interact with each other (Madupu 2006). In an online setting, brand community 19 members can interact in chat rooms, on message boards, posting to forums, or commenting on other user submitted materials. Another type of behavior is the spreading of information about the brand (e.g., word-ofmouth). Individuals who are very active at spreading positive things about the brand have been described as brand evangelists (Muniz and O'Guinn 2001; Yeh and Choi 2011, p. 146). Evangelists promote the brand not only within, but also outside of the brand community, and zealously recruit new members to the community. Leadership in brand communities is not limited to a firm’s employees, but community members also exhibit leadership behaviors. Community members even assume leadership roles entirely on their own accord (Algesheimer et al. 2005, p. 22; Israel 2012; McAlexander et al. 2002, p. 42). Community members taking on leadership roles sacrifice their own time and resources to help others, recruit new members, and defend the brand community (Algesheimer et al. 2005, p. 22). For example, when the Dell IdeaStorm community was faltering during 2011, community members began to become very vocal critics of Dell and how the community was being managed. When Dell realized that the IdeaStorm community had devolved into a “reverse-monologue,” they recruited one of the most vocal critics of the community to take part in the redesign of the brand community (Israel 2012). The feedback and advocacy of this critic has been instrumental in reviving IdeaStorm into a cutting-edge brand community once again. Fournier and Lee (2009) identified 18 different roles (i.e., patterns of behavior) spanning many leadership and non-leadership behaviors in brand communities. They argue that these roles are “critical to a community’s function, preservation, and evolution” (Fournier and Lee 2009, p. 109). Behaviors they identify include giving and receiving help, serving as a role model, bringing in new information, recruiting new members, and promoting the community to 20 outsiders. Another interesting aspect of community member behavior is the frequency with which community members do something. The definition of interaction used by Dholakia et al. (2004) is based on the frequency with which a community member does something. An interesting extension of this is the idea that behaviors can and should change over time. Fournier and Lee (2009) state that “successful communities give members opportunities to take on new roles, alternate between roles, and negotiate tensions across roles in conflict–without ever leaving the fold” (p. 109). Therefore, the frequency of specific behaviors and/or roles may change over time. Behavioral, Psychological, or Sociological Segmentation of Online Brand Community Members Brand community members are not as homogenous as one might think based on the focused nature of the brand community (i.e., a community centered on a brand). Online brand communities are especially diverse (even relative to face-to-face brand communities) because of geographical dispersion of members and very low barriers to entry (e.g., the community is available anytime, no travel is required to participate, and it is typically free to join the community). Prior ethnographic and descriptive research confirms that brand community members can be segmented within communities (e.g., Fournier and Lee 2009; Fox 1987; Moreland and Levine 2002). Customer lifetime value research and the market segmentation literature suggest that not all segments of consumers are equally profitable for marketers to serve (e.g., Kumar et al. 2010). Segmenting community members based on distinct characteristics (see Table 2) can help marketers more efficiently and effectively allocate marketing resources that 21 better match the needs of members in the segment (Dickson and Ginter 1987). Some have argued that numerous one-to-one marketing attempts made by marketers lead to consumers viewing “many marketing initiatives [as] trivial and useless instead of unique and valuable” (Fournier et al. 1998, p. 44). Shifting from viewing marketing initiatives as unique and valuable to trivial and useless suggests that there is a point of diminishing returns for marketing communications (Fournier et al. 1998; Kumar et al. 2010), and if so, marketers should carefully craft the right message for the right segment to minimize alienating community members. 22 Table 2 Essay One: Bases for Segmentation Construct Behavior: Roles (Madupu 2006) (Dholakia et al. 2004) (Fournier and Lee 2009) Psychological: Involvement (Petty et al. 1983) See also (Celsi and Olson 1988) (Zaichkowsky 1985) (Zaichkowsky 1994) (Mittal 1995) Definition “Consumer’s active participation in brand-related events and his/her interactions with other members of the online brand community” (Madupu 2006, p. 31) Roles occupied by members of brand communities (Fournier and Lee 2009, p. 109) Mentor: Teaches others and shares expertise Learner: Enjoys learning and seeks self-improvement Back-Up: Acts as a safety net for others when they try new things Partner: Encourages, shares, and motivates Storyteller: Spreads the community’s story throughout the group Historian: Preserves community memory; codifies rituals and rites Hero: Acts as a role model within the community Celebrity: Serves as a figurehead or icon of what the community represents Decision Maker: Makes choices affecting the community’s structure and function Provider: Hosts and takes care of other members Greeter: Welcomes new members into the community Guide: Helps new members navigate the culture Catalyst: Introduces members to new people and ideas Performer: Takes the spotlight Supporter: Participates passively as an audience for others Ambassador: Promotes the community to outsiders Accountant: Keeps track of people’s participation Talent Scout: Recruits new members Perceived personal relevance, or the expectation that it can have a “significant consequences for their own lives” (Petty et al. 1983, p. 81) 23 Table 2 (cont’d) Construct Sociological: Engagement (Algesheimer et al. 2005) (This Study) Definition “Consumer's intrinsic motivation to interact and cooperate with community members” (Algesheimer et al. 2005, p. 21) Online brand community engagement is the compelling intrinsic motivations to continue interacting with an online brand community. It is a stable emotional commitment, and it propels individuals to continue interacting with the community because it is an aroused state. In other words, engagement is the feeling that drives people to keep interacting in the online brand community. Behavioral: Roles Online brand communities provide an information rich medium for gathering data about brand community member behavior. A strength of online brand communities is that they enable marketers to gather and analyze member behavior in a very unobtrusive way. Specifically, member interactions with other members can be organized and recorded without the members ever even knowing their behavior is being carefully monitored and analyzed. However, categorizing community member behavior can be very difficult and prohibitively costly due to the tremendous volume of data, the subtle nature of social interactions (e.g., capturing the subtle differences between mentoring, guiding, and being back-up; Fournier and Lee 2009, p. 109), and requiring trained ethnographers or highly sophisticated software and powerful computers to code the data. Despite the widespread use of role-based classifications of community members in the popular press (e.g., Bernoff 2010), these types of classifications of community members are problematic for several reasons. First, describing a member based on a particular role tends to oversimplify the description of their behavior in the community. For example, suppose a firm labels a community member an “evangelist” based on their pattern of evangelizing the brand in 24 the past. What if members perform multiple roles or change roles? Describing that member as an “evangelist” could easily lead the firm and researchers to overlook numerous other behaviors that the member performs within the community. Second, role-based classification of community members inhibits causal analysis of the effects brand communities have on member behavior. Specifically, there are three ways in which role-based classifications inhibit causal analysis. Relying on a role (i.e., past pattern of behavior) to serve as the basis for predicting the same behaviors in the future fails to account for what started the behavior(s) and what causes changes in the frequency with which members engage in the behavior(s). Next, it takes time for new patterns of behavior to be recognized and classified. Role-based classifications of members will have poor predictive ability following role changes due to the need for a sufficient number of observations to recognize and classify changes in patterns of behavior. Lastly, role-based classifications could have a large amount of error variation that could inhibit statistical analysis. Role-based classifications by definition require sufficient observations to establish a pattern of behavior. The researcher classifying community members into roles has to specify the frequency at which a repeated behavior should be classified as a role. Lower observed frequencies of behavior (e.g. a few observations of the behavior) could be used to identify “patterns” of behavior quickly, but the variety of member behaviors could lead to improper assignment of the member to a role. Reliable classification of members into roles requires larger frequencies of behavior and consistent behaviors. 25 Psychological: Involvement Enduring involvement, or the perceived personal relevance of the community (i.e., the community is instrumental in achieving the individual’s goals) (e.g., Celsi and Olson 1988; Petty et al. 1983; Zaichkowsky 1985; Zaichkowsky 1994), could also be used to segment online brand community members. Involvement can be used to capture the intrinsic aspects which drive personal relevance (Celsi and Olson 1988), suggesting linkages to situation specific as well as enduring personal characteristics which may be predictive of behavior. However, a limitation of involvement as defined in the literature is that personal relevance does not necessarily imply motivation to act. The community could therefore be very relevant to an individual, but fail to motivate his/her behavior. Sociological: Engagement Brand community engagement is the “consumer’s intrinsic motivation to interact and cooperate with community members” (Algesheimer et al. 2005, p. 21). Engagement captures the intrinsic drive members have to interact with the community rather than trying to capture the specific behaviors members perform. Engagement, unlike involvement, does imply “members are interested in helping other members, participating in joint activities, and otherwise acting volitionally in ways that the community endorses and that enhance its value for themselves and others” (Algesheimer et al. 2005, p. 21). Brand community engagement in face-to-face settings has been shown to lead to community participation, community recommendation, and desires to continue membership in the community (Algesheimer et al. 2005). An engaged community member is therefore motivated to be an actively involved citizen of the brand community and can interact with the community in many ways. 26 Building on prior research, I define online brand community engagement as the compelling intrinsic motivations to continue interacting with an online brand community. As a stable emotional commitment, it propels individuals to continue interacting with the community because it is an aroused state. In other words, engagement is the feeling that drives people to keep interacting in the online brand community. There are several reasons why engagement instead of roles or involvement should be used as a basis to segment community members. First, engagement is a motivational state, and should function as a “leading indicator,” such that higher levels of engagement in one time period should correspond with higher levels of interaction in following time periods, whereas lower levels of engagement in one time period should correspond with lower levels of interaction in following time periods. Second, roles in a brand community encompass a wide range of behaviors, making it difficult to identify and classify specific behaviors (e.g., posting comments online, attendance at offline events, conspicuously leading face-to-face events) as a role. Third, perceived personal relevance fails to explicitly account for the social aspects of interacting with other community members. Consequences of Interaction with Online Brand Community Consequences of interaction with a brand community has also been the subject of several studies. In one of the seminal works on brand communities, Muniz and O'Guinn (2001) indicate that brand communities exhibit several important “markers” of community: consciousness of kind, shared rituals or traditions, and a sense of moral responsibility (p. 413). Member perceptions of consciousness of kind, shared rituals and traditions, and a sense of moral responsibility influence their behavior and the relationships they have with other members (e.g., a consciousness of kind 27 creates a connection between community members and a sense of moral responsibility compels one member to help another member in need). These perceptions of community are especially important when researching comparative groups, as brand community membership is strong enough to induce in-group/out-group biases affecting decision-making regarding consumption of the focal and competing brands. Madupu (2006) found that interaction with a brand community leads individuals to feel a consciousness of kind, develop a sense of shared rituals, and have a sense of moral responsibility toward other community members. Furthermore, Madupu (2006) found that feelings of consciousness of kind leads to two kinds of brand loyalty (oppositional and sustainable) (see also, Thompson and Sinha 2008), and that moral responsibility leads to brand recommendations. Interaction with a brand community can also influence a consumer’s decision-making process. Fellow brand community members could function as “surrogate consumers.” A surrogate consumer is someone who enters the vertical market structure to assist the consumer in making a purchase decision (Solomon 1986, p. 208). Interaction with a brand community provides access to information and other resources which can be used to simplify the market, evaluate alternatives, and even manipulate the market. For example, the firm could provide discounts, exclusive information, and special promotions to the community members. In addition, other community members could also suggest where to purchase the product at the lowest price, how to use or modify the product in new ways to obtain additional value from it, and detailed product reviews. Access to these community resources should lead consumers to make different purchase decisions than had they not had access to and interacted with the brand community (Baligh and Richartz 1967; Solomon 1986). 28 PROPOSITION DEVELOPMENT Online Brand Community Engagement Typology Regarding typologies, Bailey (1994) states: “A well constructed typology can be very effective in bringing order out of chaos. It can transform the complexity of apparently eclectic congeries of diverse cases into well-ordered sets of a few rather homogenous types, clearly situated in a property space of a few important dimensions. A sound typology forms a solid foundation for both theorizing and empirical research. Perhaps no other tool has such power to simplify life for the social scientist.” (p. 33) Likewise, Hunt (2002) states that classification helps build theory. An important question for academic researchers and managers is therefore whether brand community members can be segmented into unique and meaningful groups. Wedel and Kamakura (2000) argue that the segmentation method should be determined by the strategic objectives for segmenting. “The strategic purposes of segmentation determine the bases and methods used in market research; different segments may be identified in the same population of customers in different segmentation studies with different purposes” (Wedel and Kamakura 2000, p. 336). Since many of the benefits of online brand communities are contingent upon the social interaction of members, it makes sense that one of the primary objectives of a typology would be to describe what drives participation in the community. From this basis, latent class regression is the most appropriate tool to use for segmenting community members because it segments community members on the relationships (i.e., betas) between predictors (i.e., engagement dimensions) and the response variable (i.e., participation intentions) rather than means. From a managerial perspective, Kotler and Keller (2009) state that there are five criteria for evaluating the usefulness of a segmentation scheme: measurable, substantial, accessible, differentiable, and actionable (p. 228). Using engagement to segment community members is measurable because community member engagement can readily be assessed with the items 29 developed in this paper. Online brand communities are sufficiently large to estimate substantial groups based on engagement (even very small communities generally have several hundred members and large communities can have millions of members). Furthermore, Dröge et al. (2010) point out that “sometimes the community is not ‘large’ in any absolute sense but does tap nearly all important decision makers, opinion leaders, trendsetters, or lead users, making that… community’s impact disproportionately large relative to its actual size” (p. 69). Members of online brand communities are readily accessible by the firm in as much as the firm tracks visitors to its online brand communities (e.g., contact information required for membership). Each of the segments should be differentiable in terms of their brand purchase intentions. Lastly, the engagement typology provides a strong foundation for development of actionable marketing strategies to leverage communities to enhance brand purchase intentions. Segmenting community members is not a novel idea in and of itself, (e.g., Fournier and Lee 2009; Fox 1987; Moreland and Levine 2002), but using online brand community engagement as the basis for segmenting is novel and yields important insights. Fournier and Lee (2009) segment member roles in a brand community (e.g., mentor, learner, hero, celebrity, supporter, p. 109), and emphasize how these roles contribute to the functioning of the community. Specifically, Fournier and Lee (2009) state that communities need members performing a variety of tasks to ensure the “function, preservation, and evolution [of the community]” (Fournier and Lee 2009, p. 109). Unlike Fox (1987), Fournier and Lee (2009) do not organize brand community roles into a social hierarchy. Instead, Fournier and Lee (2009) emphasize a greater diversity of roles within the community. One could however infer that some of the roles Fournier and Lee (2009) describe have relatively higher social status than other roles, based on community members valuing particular roles more than other roles. Thus, while the 30 dimensions of segmentation for Fournier and Lee (2009) role-based typology are not clear, but the emergence of segments of members within the community is observable. Fox (1987) studied the “punk” subculture. While Schouten and McAlexander (1995) points out that subcultures and brand communities are different, they do have many important things in common. Subcultures of consumption and brand communities are similar in that they both exhibit shared values and beliefs, rituals and traditions, social structure, and membership transcends many boundaries (e.g., national, cultural, demographic, class, race, ethnic, etc.). Subcultures of consumption also differ from brand communities in several respects. Social structure in a subculture tends to be viewed more from a role perspective with a clear emphasis on a well-defined social hierarchy. Brand communities tend to have a less formal social hierarchy, but do exhibit social hierarchies nonetheless (Fournier and Lee 2009; Muniz and O'Guinn 2001, p. 414). Brand communities place greater emphasis on the non-geographically bound nature of the communities (Muniz and O'Guinn 2001) and tend to place the brand at a greater focus of the community. Furthermore, brand communities tend to have socially negotiated meaning, rather than a fixed meaning across contexts which are characteristic of subcultures of consumption (Muniz and O'Guinn 2001, p. 414), suggesting that brand communities are perhaps a little more dynamic across contexts than subcultures of consumption. Therefore, given the substantial amount of congruence in the social structures of brand communities and subcultures of consumption, it is appropriate to discuss the work of Fox (1987) in a brand community context. Fox (1987) segments community member roles from a hierarchical perspective, suggesting that there are concentric rings of member roles which are determined by the member’s commitment to the community (p. 50). Moving from the center out, rings are 31 described as hardcore, soft-core, preppie, and spectator (Fox 1987, p. 350). A concentric rings conceptualization of community member roles highlights the notion that the degree of engagement with the community is not constant across a community, but instead varies across the community. The relative size of each ring suggests relatively few members are at the center (core) and the relative amount of members increases with each succeeding ring of the community. Individual and situational factors determine motivational states. The variation in needs and situations consumers have will naturally lead to different levels of engagement with the online brand community. Furthermore, the differences in the community in terms of the format, content, and social dynamics will be more or less appealing to different people. Thus, differences among their motivations to participate in the community should exist among members of the online brand community. Furthermore, as discussed previously, ethnographic and descriptive studies have observed different categories of behavior. Assuming motivations drive behaviors then suggests that different categories of motivations that drive intentions to participate in the community should emerge. P1: The motivations that drive community members’ participation intentions for online brand communities will differ across community members. Brand communities are typically regarded as being good for brands. Marketers accepting this belief invest substantial resources in creating, monitoring, and sustaining brand communities. In 2012, it was estimated that spending for social media advertisements alone will exceed $3.63 billion dollars (excluding the costs for establishing and maintaining online brand communities, Elkin et al. 2012). While it is difficult to estimate the cumulative spending on 32 creating and maintaining an online brand community and MROCs, it was recently reported that a majority of global firms make these investments. For example, a study conducted by IBM found that seventy-nine percent of global firms maintain a social media presence on social networking sites and fifty-two percent of global firms engage in micro blogging (IBM 2013). This demonstrates the tremendous resources being deployed to create and maintain community involvement in the evolving Internet landscape. One of the primary drivers for investing in online brand communities and MROCs is simply the volume of users looking to interact with brands online. It has been estimated that 84 percent of Internet users have contacted or participated in online communities (Madupu and Cooley 2010). A worldwide study of Internet users found that 75 percent of regular Internet users visited brand web sites and 32 percent had joined an online brand community in the past six months (Hutton and Fosdick 2011). As consumers continue to proactively reach out to socially engage brands online, marketers will continue to invest in these relationships. Just to create content for its Facebook fan page General Motors invests $30 million annually (this does not include other nontrivial costs associated with creating and maintaining a social presence online, and GM even continued to spend $30 million a year on content while they cut their $10 million a year spending on advertising in Facebook, Barkholz and Rechtin 2012). However, greater scrutiny and vetting of strategies for creating and maintaining a social presence online is needed before marketers adopt strategies utilizing social presences as a core component of their marketing strategy to enhance brand performance. Brand communities provide a forum for the active interpretation and creation of meaning of the brand (Muniz and O'Guinn 2001; Muniz and Schau 2007). Engaged community members are motivated to participate in this creation by sharing “context-rich” and “meaningful 33 consumption experience[s] [which] strengthens interpersonal ties and enhances mutual appreciation for the product, the brand, and the facilitating marketers. Virtual ties become real ties. Weak ties become stronger. Strong ties develop additional points of attachment” (McAlexander et al. 2002, p. 44). These unique experiences shape community member evaluations of the brand and can form the basis of a competitive advantage for the firm. It is engagement, their intrinsic motivations that propels them to interact with the community. Interaction with the community exposes the member to community-generated content. Some community-generated content can rival the quality of professional ad agencies and is it is increasingly common to see marketers using consumer generated content as the basis for national advertising campaigns (e.g., Muniz and Schau 2007; Doritos' Super Bowl ads). The context rich interaction of community members (e.g., affiliation with others, social interaction, and valuable information regarding the brand and its use; Martin 2009) should affect their brand purchase intentions. Additionally, being part of a brand community has been described as becoming “part of the family” (McAlexander et al. 2002, p. 46). The brand community member interviews conducted by McAlexander et al. (2002) revealed an interesting observation that brand communities can influence willingness to try a product for the first time (McAlexander et al. 2002, p. 46). Specifically, they found that following the trial and adoption of a brand by a community member, the community member’s family tended to also try and adopt the brand. More engaged community members tend to have closer relationships with other community members, which suggests that brand community members could influence each other in similar ways as that observed by McAlexander et al. (2002). 34 Engagement is likely to be positively related to new product trial and adoption intentions for three reasons. First, because of their interaction with other community members, they are more likely to hear more about new products. The explicit focus of the community on the brand makes it a great source for cutting-edge information about the brand and its products. Awareness of the brand’s existing and new products is essential for purchase intentions to be formed (Mahajan et al. 1995; Mahajan et al. 1990). Second, their relationships with other community members (especially ones experienced with the brand) can reduce the risks associated with trial and adoption of new products. Perceived risk is a key barrier to trial and adoption of new products. Community members have the additional support and information from the community which will decrease their perceived risk of new product trial and adoption. A reduction in perceived risk should therefore increase the community member’s willingness to purchase the brand’s new products. Third, close relationships also enhance the meaningfulness of consuming the products, leading to increased utility for trial and adoption of the brand’s new products. Brand community engagement should also affect purchase intentions. Muniz and O'Guinn (2001) discuss community members as feeling a moral responsibility or in other words, feeling a sense of duty or obligation to the community as a whole. They state that this felt moral responsibility to the community helps ensure the survival of the community. A key way to help ensure the survival of a brand, and consequently its community, is to patronize the brand instead of competitor brands. More engaged members should be more interested in the community’s survival and therefore be more willing to spend more on its products than less engaged community members. In addition, engaged members are exposed to group level influences on their attitudes and behavior (e.g., compliance, internalization, and identification; Dholakia et al. 35 2004), which tend to be pro-brand (Dröge et al. 2010), and therefore positively contribute to spending more on the brand. P2a: Mean levels of new product trial will differ across online brand community engagement segments. P2b: Mean levels of new product adoption will differ across online brand community engagement segments. P2c: Mean levels of purchase intentions will differ across online brand community engagement segments. The engagement typology should provide enhanced predictive accuracy of brand community participation intentions over role-based classifications of online brand community members (Fournier and Lee 2009). Role-based classifications of community members tend to oversimplify the description of community member behavior(s), fail to account for what started and what causes changes in the frequency of the behavior(s), and are unstable depending on classification timeframe selected. Therefore, a motivationally-based typology of community members should be superior at predicating brand community participation intentions as compared to role-based typologies for three reasons. First, a motivationally-based typology can be used to develop a better picture of what behaviors should emerge in an online brand community based on specific motivations. Second, motivation is an aroused state which leads to action; thus it can serve as a leading indicator of behavior. Third, online brand community engagement can be measured at a single point in time, overcoming the need to arbitrarily define frequencies of behavior to classify it as a role. These three desirable features of the engagement 36 typology suggest that it should have better predictive ability of brand community participation intentions than role-based classification of community members. P3: A typology created using online brand community engagement will explain more variance in brand community participation intentions than the role-based typology created by Fournier and Lee (2009). METHODS AND RESULTS Scale Development Procedure Online brand community engagement is an emerging area for research, so I follow a grounded theory approach (Spiggle 1994) to develop the scales for online brand community engagement. Please see Table 3 for a summary of the steps for developing the online brand community engagement scale. 37 Table 3 Essay One: Summary of Steps For Brand Community Engagement Scale Development Process Steps in the Process 1. Study 1 - Focus Groups • Focus Group Interviews 2. Study 2 – Open-Ended Surveys • Extension of focus group questions into open-ended questionnaire format 3. Item Generation 4. Item Reduction and Expert Review 5. Study 3 - Exploratory Study • Exploratory Factor Analysis 6. Study 4 - Initial Validation Study • Confirmatory Factor Analysis o Dimensionality o Factor Loadings • Validity o Convergent Validity o Discriminant Validity • Reliability 7. Study 5 - Final Validation Study • Confirmatory Factor Analysis o Convergent Validity o Discriminant Validity • Structural Equation Modeling o Nomological Validity Details • Two focus groups • Qualitative analysis of focus group transcripts for motivational themes • Open-ended surveys of active brand community members • Qualitative analysis of responses to open ended survey questions for motivational themes • Generation of 494 items by research team based on 11 constructs • 138 items selected for expert review • 2 Marketing Faculty Members and 6 Marketing Doctoral Students Judged Items • 94 Items remained • Online survey of online brand community members (student sample) • Online survey of online brand community members (adult sample) • Online survey of online brand community members (adult sample) 38 1. Study 1: Focus Groups Initial literature reviews were conducted to assess the domain of extant research on motivations brand community members have to interact with the community. This research suggested that engagement and reactance are two powerful motivations in brand community members (Algesheimer et al. 2005). Since the research on online brand community member engagement and reactance is sparse, focus groups were conducted to explore the domain of the constructs. Focus group participants were recruited from a large Midwestern university’s undergraduate marketing courses. In total, 30 students applied for the focus groups and 11 were selected for being active members of online brand communities (6 male, 5 female). Focus group sessions were conducted by a moderator and assistant moderator following a questioning route developed specifically for this study (Krueger and Casey 2009). Focus groups started by asking participants about brand communities they participate in, what they like and dislike about brand communities. A key questioning route was then used to explore how and why brand community members interact with the brand community. The questioning route concluded by talking with participants about what they thought an ideal brand community would look like. Participants were then debriefed and thanked for their participation. All focus group sessions were recorded and transcribed. Following the focus group sessions, the researcher and assistant moderators met to discuss the transcripts. The researcher and assistant moderators (two research assistants not aware of the theoretical background) reviewed the transcripts and identified themes related to brand community engagement and brand community reactance. From this initial analysis, the differences between face-to-face and online brand communities became apparent (McAlexander et al. 2002). Reactance, a construct introduced to the face-to-face brand community literature by Algesheimer et al. (2005) study of European Car Clubs, did not emerge as a theme for online 39 brand community members we interviewed. Therefore, reactance was dropped from further consideration as a key motivation for the interaction with the brand community. 2. Study 2: Open-Ended Surveys An online panel company was then used to collect a broader sample of very active (participate 23 times per week or more) online brand community members. Very active online brand community members were selected for this study because they would likely exhibit stronger motivations and clearer themes for scale development purposes. In subsequent studies, this screening criteria was not used. Open-ended survey questions based on the focus group questioning route were used to interview respondents. Of the completed 70 surveys returned, 44 surveys were screened because they were not very active members of online brand communities (i.e., participate in an online brand community more than 2-3 times a week). In addition, 2 surveys were dropped for data quality concerns (e.g., lack of elaboration on qualitative questions and speeding). The remaining 24 responses were analyzed for themes surrounding the motivations brand community members have to interact with the brand community. See Table 4 for the 11 main themes that emerged from the responses. Following Rossiter (2002), tentative construct definitions were created based on the themes. The researcher team reviewed the themes again and then refined the construct definitions according to the iterative approach advocated in a grounded theory development (Spiggle 1994). 40 Table 4 Essay One: Online Brand Community Engagement Construct Definitions Engagement Dimensions Construct Definition Brand Influence The degree to which a community member wants to influence the brand. Brand Passion The ardent affection a community member has for the brand. Connecting The extent to which a community member feels that being a member of the brand community connects them to some good thing bigger than themselves. Helping The degree to which a community member wants to help fellow community members by sharing knowledge, experience, or time with other community members. Like-minded Discussion The extent to which a community member is interested in talking with people similar to themselves about the brand. Rewards (Hedonic) The degree to which the community member wants to gain hedonic rewards (e.g., fun, enjoyment, entertainment, friendly environment, and social status) through their participation in the community. Rewards (Utilitarian) The degree to which the community member wants to gain functional rewards (e.g., monetary rewards, time savings, deals or incentives, merchandise, and prizes) through their participation in the community. Seeking Assistance The degree to which a community member wants to receive help from fellow community members who share their knowledge, experience, or time with them. Self-Expression The degree to which a community member feels they can express their true interests and opinions. Up-to-date Information The degree to which a community member feels that the brand community helps them to stay informed or keep up-todate with brand and product related information Validation A community member’s feeling of the extent to which other community members affirm the importance of their opinions, ideas, and interests. 41 3. Item Generation Following the identification of themes from the focus groups and depth surveys, the research team generated items using the construct definitions. In total, 2 researchers and 2 research assistants independently generated a total of 494 items (approximately 10 items per theme per person) to measure the 11 dimensions of brand community member motivations. 4. Item Reduction and Expert Review After creating a large pool of potential items, the research team met to refine construct definitions, eliminate redundant items, and select items that had good face validity for expert review. In total, 138 items were selected for expert review. Given the large number of constructs, experts were randomly assigned half of the total items. A review of the item pool was conducted by two marketing PhD’s and six marketing doctoral students not familiar with or associated with the research project in any way. Each expert was presented 69 items one at a time in random order. With each item presented, the experts were given a multiple choice list of constructs definitions from which to select. If a majority of the experts (three or more of the four that reviewed each item) correctly assigned the item to its intended definition, then the item was retained for further testing. In total, 94 items were retained based on the expert review. In tests following the expert review, each item was measured by an 11 point Likert ranging from “0 Strongly Disagree” to “10 Strongly Agree” with a scale midpoint of “5 Neither Agree nor Disagree” and numerals for each of the remaining scale points. 5. Study 3: Exploratory Study The properties of the online brand community engagement items were explored using a sample 42 of 262 undergraduate business students. 126 completed surveys were returned. Two independent raters coded whether or not the remaining responses were in fact brand community members (91 percent agreement on classifying members and nonmembers of brand communities). 43 responses were screened because the respondents did not meet the study criteria of being a member of an online brand community. Therefore, 83 usable responses were collected in this round of data collection (49 percent of respondents were male; average age = 21 years old). Principal components exploratory factor analysis in SPSS using an oblique rotation (i.e., Direct Oblimin) was then used to assess the dimensionality of the scales. Using an oblique rotation accounts for the expected covariance among the dimensions when extracting the factors. Items that did not load on the same factor as the majority of the other items for that construct were flagged for the next round of data collection. In addition, items that had factor scores of less than .50 on the factor for their construct were also flagged for the next round of data collection. A comparison was also conducted using an orthogonal Varimax rotation; this led to a similar pattern of items flagged for the next round. No items were dropped from analysis based on this pilot test. 6. Study 4: Initial Validation Study An online panel company was contracted for the next round of data collection to get a sample of 2,839 US Internet users who are 18 and older. 1,190 respondents indicated they were not members of brand communities and were screened immediately from the study. 737 of the remaining responses were screened by the panel company for speeding through initial survey questions. In total, 911 completed surveys were returned. Of the completed surveys returned, 282 responses were dropped for data quality (i.e., speeding, straight-lining, lack of elaboration 43 and gibberish responses to open-ended questions). Two independent raters coded whether or not the remaining respondents were actually members of a brand community (86 percent agreement on classifying members and nonmembers of brand communities). 285 responses were screened because the respondents were not active members of brand communities. Therefore, 344 valid responses were used for further analysis (38 percent of respondents were male; average age = 44 years old; median education 2 year college degree). There are several popular and somewhat different approaches to scale development. The classical method is based on the Churchill (1979) paradigm. One of the primary limitations and subsequent criticisms of scale development using the Churchill (1979) paradigm is the use of item-total correlations with reflective scales. Jarvis et al. (2003) establish four criteria for determining if a scale is reflective. First, the direction of causality from construct to measure is implied by the conceptual definition of each scale. Each of the items selected from the initial pool of items in our study was chosen because it represents a manifestation of the scale definition. In scale development, item reduction is one of the primary goals, therefore it is important that the elimination of items does not substantially alter the conceptual definition of the construct. Thus, changes in the latent construct should affect changes in the items. Second, interchangeability of the indicators/items is important for reflective scales. Each of the items selected from the initial pool share a common theme based on the definition of the construct. Items were written and selected with the knowledge that through the scale development process items would be dropped. Therefore the research team carefully selected items so that if one were dropped, it would not change the conceptual domain of the construct it was intended to measure. Third, covariation among the indicators is expected for reflective scales. Fourth, the nomological net of the construct indicators for each of the items in reflective constructs should 44 have the same antecedents and consequences. Based on the above four criteria established by Jarvis et al. (2003), the scales developed in this study are reflective. The primary shortcoming of using item-total correlations to eliminate items is that itemtotal correlations are created using a unitary weighted sum of the each item score in the scale (i.e., a correlation is calculated between each item and the sum of the other items). Using a unitary weighted sum models the items as formative, reversing the causal direction from the latent construct causing the indicators to the indicators causing the latent construct. This represents a fundamental violation of the characteristics of a reflective construct. In addition, item-total analysis is not characterized by statistical significance testing and relies heavily on rules of thumb to make decisions regarding the retention of items into the final scale. However, due to the relative ease of item-total analysis it remains a popular technique to this day. In addition, this approach uses exploratory factor analysis which typically only accounts for approximately 80 percent of variance. Alternative approaches have been developed that properly model reflective constructs and model all of the variance in the data. Gerbing and Anderson (1988) updated the Churchill (1979) paradigm with the use of confirmatory factor analysis (CFA) to overcome some of the specific limitations of the item-total and exploratory factor analysis procedure. Their methods more directly assess the dimensionality of a scale, which is a key criteria for any scale development procedure (Gerbing and Anderson 1988). Therefore, analysis of the scales developed in this study was conducted following Gerbing and Anderson (1988) approach. Specifically, I iteratively estimated a series of measurement models where the 11 engagement dimensions were estimated as first order reflective scales. In these models, “bad” items were removed and then the entire model was reestimated and re-assessed. During this process, items with large standardized residuals were 45 removed as they negatively affected the unidimensionality of each dimension (Gerbing and Anderson 1988). Items with large standardized residuals indicate a lack of external consistency, meaning they correlate highly with other factors, thus the items composing the scale lacks unidimensionality (Gerbing and Anderson 1988). In order to enhance the rigor of this initial investigation, I included measures for two related variables: expectations for community member behavior and intentions to share information from the community with others. These variables were included to provide a more rigorous assessment of discriminant validity and ensure that items for the engagement dimensions were strictly measuring engagement and not expectations or outcomes of brand community participation. Initial overall model fit for the CFA was modest (χ2 = 12,380, df = 5,916; CFI = .84; SRMR = .08; RMSEA = .06; AIC = 548). Through a series of inspections and iterative model estimations, we removed a total of 22 items due to (1) large standardized residuals (> .25) (Gerbing and Anderson 1988, p. 189), (2) lambdas below .707 (lambda’s below .707 indicate that random error determines more variation in the item than what is determined by the latent construct), and (3) significant cross-loadings as detected through an examination of Lagrange Multiplier indices. After deleting these items, a final measurement model was estimated that offered improved fit (χ2 = 7,168, df = 3,662; CFI = .90; SRMR = .05; RMSEA = .05; AIC = -156). Each construct demonstrates adequate convergent validity as each average variance extracted is greater than .50. Lambda loadings and descriptive statistics for each item retained in the final model are presented in Table 5 in an effort to establish scale norms. Discriminant validity of the scales was assessed following Fornell and Larcker (1981). The average variance extracted for each construct are all greater than .60. The average variance extracted per construct were then 46 compared to each squared correlation between the construct and all other potential pairs of constructs. No squared correlations between constructs were greater than the average variance extracted for each construct. Therefore, discriminant validity between the scale dimensions is supported. Following assessment of convergent and discriminant validity, reliability was analyzed. Latent construct reliability ranged from .78 to .92 which supports reliability for the constructs. 47 Table 5 Essay One: Scale Items, Descriptive Statistics, and Factor Loadings Study 4 – Initial Validation Range Mean SD λ Factor Item Brand Influence 1. I am motivated to participate in this brand community because I can help improve the brand and its products 2. This brand community provides the company valuable insights to help improve the brand and its products 3. I like to know that my comments and suggestions can influence the brand and its products 4. Increasing the influence I have on the brand and its products makes me want to participate more in this brand community 5. I hope to improve the brand or product through my participation and expression in this brand community 6. I participate in this brand community to offer my insight to the company 7. I want the company to listen and respond to my opinions in this brand community Brand Passion 1. I am motivated to participate in this brand community because I am passionate about the brand 2. If it weren’t for the positive feelings I have about the brand, I wouldn’t participate in this brand community 3. I participate in this brand community because I care about the brand 4. I would only belong to a brand community for a brand I care deeply about 5. I would not belong to a brand community if I did not have passion for the brand 48 Study 5 – Final Validation Range Mean SD λ 1-11 8.05 2.65 .94 1-11 7.02 2.84 .92 1-11 8.48 2.41 .88 1-11 7.84 2.50 .80 1-11 8.54 2.48 .92 1-11 7.35 2.78 .85 1-11 8.07 2.61 .90 1-11 6.98 2.75 .89 1-11 8.25 2.63 .92 1-11 7.18 2.79 .90 1-11 1-11 8.00 8.68 2.63 .88 2.44 .85 1-11 6.85 1-11 7.96 2.91 .83 2.56 .79 1-11 8.56 2.39 .89 1-11 8.75 2.20 .91 1-11 8.33 2.59 -- -- -- 1-11 1-11 1-11 8.89 8.14 8.23 2.02 .85 2.61 .80 2.59 .78 2-11 8.90 1-11 8.30 1-11 8.79 -- -- 2.12 .89 2.63 .81 2.31 .84 Table 5 (cont’d) Factor Item 6. If I was not passionate about the brand, this brand community would not interest me 7. The other members of this community fuel my passion for the brand 8. My passion for this brand's products makes me want to participate in this brand community 9. Brand communities for products that I am passionate about interest me Connecting 1. Increasing the strength of the connection I have with this brand community makes me want to participate more in the community 2. I think that the brand community extends beyond just me 3. Being part of this brand community makes me feel more connected to the brand 4. Being part of this brand community makes me feel more connected to other consumers of the brand 5. I participate in this brand community to feel more connected to the brand 6. Interacting with community members makes me feel like I am part of something that can really make a difference 7. I am part of something bigger than myself when I interact with this community Helping 1. I like participating in the brand community because I can use my experience to help other people 49 Study 4 – Initial Validation Range Mean SD λ 1-11 8.22 2.59 .76 Study 5 – Final Validation Range Mean SD λ 1-11 8.31 2.68 .75 1-11 1-11 7.28 8.55 2.68 -2.42 .90 --1-11 8.76 --2.25 .88 1-11 8.83 1.97 -- -- -- 1-11 7.68 2.42 .83 1-11 7.89 2.17 .89 1-11 1-11 9.19 8.60 1.88 -2.07 .82 --1-11 8.41 --1.91 .74 1-11 8.35 2.20 .83 1-11 8.53 1.92 .82 1-11 1-11 7.90 7.71 2.45 .81 2.52 -- 1-11 7.54 --- 2.38 ---- 1-11 7.93 2.51 -- -- -- 1-11 7.57 2.38 .87 1-11 7.58 -- -- -- -- 2.46 .85 Table 5 (cont’d) Factor Item 2. 3. 4. 5. 6. 7. 8. 9. If it weren’t for being able to help other community members, I wouldn’t participate in this brand community The more help I am able to give in this brand community, the more I feel motivated to participate in this community I like to share my experience and knowledge with others in this brand community to help them be more educated about the brand Being part of this brand community makes me feel needed by others I feel like I offer a unique perspective that can help other members of this brand community I really like helping other community members with their questions I feel good when I can help answer other community member’s questions The thing I like doing the most in this community is helping others Study 4 – Initial Validation Range Mean SD λ 1-11 4.68 2.90 -- Study 5 – Final Validation Range Mean SD λ ----- 1-11 7.38 2.58 .83 1-11 7.40 2.48 .83 1-11 7.84 2.51 .87 1-11 7.74 2.39 .84 1-11 1-11 6.21 7.56 2.87 .77 2.47 .83 1-11 5.88 1-11 7.40 2.79 -2.38 .78 1-11 1-11 1-11 7.66 8.24 6.81 2.51 .89 2.29 .85 2.67 .86 1-11 7.85 1-11 8.34 1-11 6.72 2.40 .89 2.29 .85 2.67 .75 7.81 2.66 -- -- -- 7.76 2.65 .89 1-11 8.12 2.35 .88 7.83 7.69 2.59 .94 2.73 .92 1-11 8.37 1-11 8.14 2.16 .85 2.24 .83 7.96 2.70 .90 1-11 8.00 2.48 .85 7.43 2.85 .90 1-11 7.99 2.64 .86 Like-Minded Discussion 1. I am motivated to participate in this brand community because I can talk 1-11 with other people like myself about the brand 2. I look forward to discussing my opinions about the brand with others who 1-11 share the same interest as me 3. I enjoy conversing with people similar to myself in this brand community 1-11 4. I like to talk to other brand community members with similar interests to 1-11 myself 5. Participating in this brand community is a good way to discuss my interests 1-11 with people who share those same beliefs 6. I look to this brand community when I want to discuss a topic with people 1-11 who have similar interests 50 -- -- Table 5 (cont’d) Study 4 – Initial Validation Range Mean SD λ 1-11 7.04 2.87 .84 Factor Item 7. 8. Discussing my views with other members is my favorite part of this community Having conversations with people in this brand community who share the 1-11 same views about this brand is important to me Rewards (Hedonic) 1. I like participating in this brand community because it is entertaining 2. Having fun is my main reason for participating in this brand community 3. I participate in this brand community because I think it is fun 4. I find participating in this brand community to be very entertaining 5. I want to participate in this brand community to be a part of its friendly environment 6. Participating in this brand community feels like an escape 7. I enjoy being immersed in this brand community Rewards (Utilitarian) 1. My main purpose for belonging to this community is access to deals 2. Without the special deals provided by this community, I probably would stop being a member 3. I am motivated to participate in this brand community because I can earn money 4. If it weren’t for the money, I wouldn’t participate in this brand community 5. Receiving more money makes me want to participate more in this brand community 51 Study 5 – Final Validation Range Mean SD λ 1-11 7.57 2.75 .83 7.24 2.76 .89 1-11 7.51 2.45 .87 1-11 1-11 1-11 1-11 1-11 7.84 7.01 8.25 7.87 8.00 2.63 2.87 2.44 2.55 2.39 1-11 1-11 1-11 1-11 1-11 2.23 2.71 2.22 2.18 2.37 1-11 1-11 6.47 7.62 3.04 .71 2.54 .81 1-11 6.48 1-11 8.00 2.91 -2.24 -- 1-11 1-11 5.77 4.69 3.48 -3.31 -- --- --- 1-11 4.37 3.57 .94 1-11 2.83 2.84 .91 1-11 1-11 3.89 5.13 3.21 .83 3.64 .89 1-11 2.47 1-11 3.62 2.42 .89 3.28 .78 .93 .83 .87 .91 .81 8.29 7.58 8.44 8.17 7.94 --- .91 .83 .89 .88 -- --- Table 5 (cont’d) Factor Item 6. 7. 8. Receiving more sales and coupons makes me want to participate more in this brand community I like participating in this brand community because I can earn free merchandise from the brand This brand community motivates me to participate by offering me the chance to win prizes Study 4 – Initial Validation Range Mean SD λ 1-11 6.69 3.60 -- Study 5 – Final Validation Range Mean SD λ ----- 1-11 5.51 3.65 -- -- -- -- -- 1-11 5.72 3.49 -- -- -- -- -- 7.59 2.80 .90 1-11 7.86 2.49 .90 7.70 2.67 .92 1-11 8.09 2.50 .91 7.53 2.80 .88 1-11 7.95 2.47 .92 7.60 2.70 .82 1-11 7.75 2.54 .83 8.57 2.25 .79 1-11 8.89 2.10 .78 7.89 2.59 .84 1-11 8.38 2.14 .80 8.51 2.44 .77 1-11 8.68 2.30 .80 7.35 2.93 .73 1-11 8.22 2.50 .79 8.07 2.54 .82 1-11 8.23 2.42 .84 Seeking Assistance 1. I am motivated to participate in this brand community because I can 1-11 receive help from other community members 2. I am motivated to participate in this brand community because community 1-11 members can use their knowledge to help me 3. I like participating in this brand community because it gives me an 1-11 opportunity to receive help from other community members 4. Increasing the help I receive from this brand community makes me want to 1-11 participate more in the community 5. I appreciate when members of this brand community share their knowledge 1-11 and experience with me 6. I participate in this brand community so that other consumers can share 1-11 their knowledge with me 7. This community is a great way to get assistance with questions about the 1-11 brand 8. I usually interact with this community when I have questions about the 1-11 brand 9. It is important to me to be able to use this community to find answers to my 1-11 questions about the brand 52 Table 5 (cont’d) Study 4 – Initial Validation Range Mean SD λ Factor Item Self-Expression 1. I feel that I can “be myself” in this brand community more than I am able 1-11 to be in other settings 2. If it weren’t for being able to express my true interests and opinions in this 1-11 brand community, I wouldn’t participate in this brand community 3. Increasing the ability to express my true interests and opinions in this brand 1-11 community makes me want to participate more in this brand community 4. I feel that I can freely share my interests in the brand community 1-11 5. This brand community allows me to express my true feelings about a 1-11 product or brand 6. I am not afraid to express my opinion in this brand community 1-11 7. I would express any opinion or idea I had about this brand in this brand community 8. I can always be myself when interacting with others in this community 9. This community makes it easy for me to express my true beliefs about the brand 10. I am able to share my true feelings with this community without fear of ridicule 7.03 2.99 .87 -- -- -- -- 5.95 3.02 .90 -- -- -- -- 5.33 3.10 .84 -- -- -- -- 5.43 6.29 3.15 .88 3.13 .89 2-11 8.78 1-11 8.80 6.90 -- 1.96 .87 2.00 .82 1-11 6.20 3.01 .89 - --3.27 .89 1-11 8.64 1-11 1-11 7.01 6.65 3.02 .76 3.11 .93 1-11 8.55 1-11 8.74 2.32 .86 2.08 .90 1-11 5.91 3.13 .89 -- -- 8.78 2.11 .85 1-11 8.70 2.13 .83 9.05 1.97 .87 1-11 9.02 1.89 .83 Up-To-Date Information 1. I am motivated to participate in this brand community because it helps me 1-11 keep up-to-date with the brand 2. I like participating in this brand community because it helps me stay 1-11 informed about the brand 53 Study 5 – Final Validation Range Mean SD λ -- -- 2.28 .85 -- Table 5 (cont’d) Study 4 – Initial Validation Range Mean SD λ 1-11 8.28 2.51 .85 Factor Item 3. This brand community is my critical connection for new and important information about the brand and its products 4. Belonging to this brand community helps me to stay informed about the 1-11 brand 5. I feel more up-to-date with this brand's products because I belong to the 1-11 community 6. When I want up-to-date information about this brand, I look to this brand 1-11 community 7. This community keeps me on the leading edge of information about the 1-11 brand 8. This community is the best way to stay informed about new developments 1-11 with this brand 9. This brand community is my essential connection for exclusive information 1-11 about the brand and its products 10. I feel like I have access to more information about a brand when I am part 1-11 of its brand community Validation 1. I am motivated to participate in this brand community because other 1-11 members value my ideas 2. If it weren’t for other community members affirming the value of my 1-11 interests, I wouldn’t participate in the brand community 3. Receiving more affirmation of the value of my comments, makes me want to 1-11 participate more in the brand community 4. What other community members think of my ideas is important to me 1-11 5. I feel good about myself when other community members share my ideas 1-11 54 Study 5 – Final Validation Range Mean SD λ 1-11 8.60 2.26 .85 9.25 1.82 .88 3-11 9.38 1.56 .77 8.98 2.08 .88 1-11 9.04 1.85 .83 8.83 2.29 .87 1-11 8.95 2.15 .89 8.78 2.15 .89 1-11 9.00 1.98 .84 8.91 2.13 .89 2-11 8.86 2.11 .86 8.31 2.48 .85 1-11 8.21 2.52 .76 9.06 1.92 .81 -- -- -- -- 6.96 2.54 -- -- -- -- -- 4.77 2.88 -- -- -- -- -- 7.35 2.60 .82 1-11 8.00 2.31 .79 6.42 7.87 2.85 .76 2.24 .83 1-11 6.93 1-11 8.00 2.70 .75 2.10 .89 Table 5 (cont’d) Factor Item 6. I appreciate when others agree with the ideas I express in this brand community 7. When others support my ideas and opinions in this brand community, I feel better about myself 8. I like the support I get from other members when I express my ideas or opinions in this brand community 9. If no one agrees with my idea in this brand community, I feel bad 10. Members of this brand community validate my opinions about the brand Study 4 – Initial Validation Range Mean SD λ 1-11 8.07 2.18 .80 Study 5 – Final Validation Range Mean SD λ 1-11 8.26 2.12 .84 1-11 7.08 2.56 .84 1-11 7.54 2.30 .88 1-11 7.81 2.24 -- -- -- -- -- 1-11 1-11 4.61 7.70 2.88 -2.29 -- --- --- --- --- Note: All Scales measured on a 0 - 10 Likert-Type Scale with Anchors 0 = Strongly Disagree and 10 = Strongly Agree. Prior to analysis all values were recoded to a 1 – 11 range, which is presented in all results tables. Items in italics were used for validating a short-form of the scale. 55 7. Study 5: Final Validation Study A second validation dataset was collected to assess the reliability and validity in a second setting, as well as the nomological properties of the scales developed. This dataset was collected through Amazon’s Mechanical Turk service. A brief description of the study with link to the survey was posted for 18 and older US residents to complete. Respondents who completed the survey online were paid a nominal amount for participating in the study. 198 valid responses were returned. Once again, I assessed the scales by iteratively estimating models to identify any items that may be negatively affecting the scale or each dimension by assessing (1) standardized residuals, (2) lambda loadings, and (3) cross-loadings. The initial measurement model provided good fit to the data (χ2 = 16,636, df = 2,556; CFI = .85; SRMR = .07; RMSEA = .07; AIC = 371). As a result of this final screening 5 items were removed. Following the removal of these items, the measurement model offered good fit (χ2 = 15,372, df = 2,211; CFI = .87; SRMR = .06; RMSEA = .06; AIC = -414). Moreover, I found evidence for the validity and reliability of each scale based on Fornell and Larcker (1981) criteria. Table 5 includes means, standard deviations, and loadings for each scale. Development of a Short-Form Scale Ultimately, the goal of this research was to develop a short-form scale of online brand community engagement. In an effort to do this, I re-estimated one final set of measurement models using only the 4 items for each dimension that had the highest lambda loadings in the Study 5 Final Validation data. This measurement model provided the best fit (χ2 = 8,354, df = 861; CFI = .94; SRMR = .05; RMSEA = .06; AIC = -280). AVEs, construct reliabilities, and correlations for all constructs are reported in Table 6. 56 Table 6 Essay One: Results of Measurement Model Assessment and Scale Statistics Final Validation Study Short Form Scale 1 2 3 4 5 6 7 8 1. Brand Influence .25 2. Brand Passion .35 .54 3. Connecting .37 .36 .59 4. Helping .26 .59 .67 .66 5. Like-minded Discussion .27 .58 .58 .51 .62 6. Rewards (Hedonic) .08 -.53 -.37 -.30 -.41 -.44 7. Rewards (Utilitarian) .23 .31 .46 .56 .56 .38 -.26 8. Seeking Assistance .29 .37 .42 .47 .50 .51 -.27 .38 9. Self-Expression .31 .35 .40 .25 .31 .43 -.26 .38 10. Up-to-date Information .26 .36 .62 .63 .56 .38 -.21 .32 11. Validation .79 .77 .67 .74 .74 .77 .74 .80 Average Variance Extracted .87 .87 .78 .86 .86 .87 .82 .88 Construct Reliability 7.13 8.80 8.27 7.88 8.00 8.12 2.97 8.03 Mean 2.56 2.02 1.77 2.14 2.16 2.13 2.58 2.27 Standard Deviation 1 1 2 1 1 1 1 1 Minimum 11 11 11 11 11 11 11 11 Maximum Note: All correlations greater than .11 and less than -.11 are significant at the α = .05 level. 57 9 10 11 .31 .37 .76 .86 8.68 1.95 2 11 .10 .75 .86 8.85 1.92 2 11 .72 .85 7.95 1.96 1 11 Online Brand Community Engagement Typology Development Consistent with Wedel and Kamakura (2000), once the measurement model was confirmed, I began to assess the heterogeneity of online brand community members. I used latent class regression in Latent Gold version 4.5 to model the segments of online brand community members. This method of estimation does not constrain the parameter estimates between the dimensions of online brand community engagement and the outcome variable (i.e., participation intentions) to be equal across potential sub-populations among the brand community population. Latent class regression is preferable to two-stage means-based segmentation approaches (e.g., Homburg et al. 2008; Wong et al. 2010) to assess the heterogeneity of online brand community members for two reasons. First, latent class regression forms homogenous segments based on the relationship between the predictors and dependent variable (Wedel and Kamakura 2000). This is advantageous in this research over a means based approach because it groups of community members whose motivations have similar effects on participation intentions which facilitates the use of similar marketing activities. Additionally, motivationally based groups are likely to respond similarly to marketing activities. Means-based clustering works to create groups of community members with similar levels of motivations. This approach is primarily descriptive and not prescriptive. Therefore, latent class regression creates a normative segmentation of community members for use by academics and marketers that is most effective for predicting community member behavior (Wedel and Kamakura 2000). Second, latent class regression classifies segment membership on a probabilistic basis. Probabilistic assignment to segments provides a more accurate view of the degree to which the individual is similar to the clusters into which they are placed. The probabilistic assignment to segments can then be used to create nominal cluster assignments. 58 Specifically, I assessed the ability of the 11 dimensions of online brand community engagement to predict intentions to participate in the brand community (Algesheimer et al. 2005, p. 33). Since there are 11 dimensions of engagement, BIC which is sometimes used as a heuristic for determining the number of classes in latent class regression (e.g., Homburg et al. 2005), is not appropriate to use in this situation because it overly penalizes the model for complexity. Significance of chi-square difference tests between number of classes were used as the primary cutoff rule for determining the number of classes (-2LL diff = 59.72, p < .01 for going from 1 class to 2; -2LL diff = 49.78, p > .10 for going from 2 classes to 3 classes) (Wedel and Kamakura 2000). Overall, the two class model fit the data well (Log-Likelihood = -336.08, BIC = 815, CAIC = 842, R2 = 69). Please see Table 7 for descriptives of each cluster. Table 7 Essay One: Paticipation Intentions Latent Class Regression Model—Results From Short Form of Engagement in Study 5 Final Validation Study “Brand Passionate Helpers” Independent Variables Intercept Brand Influence Brand Passion Connecting a “Individualistic Information Seekers” Ward Statistic a Class 1 n = 130 Class 2 n = 68 .64** (1.16)** .55** -.04** (.06)** -.54** .20** (.09) * 2.29** .21†* (.11) * 1.88** 8.46** (.70)** 12.14** .09** (.04)** 2.09** -.20** (.08)** -2.50** -.15** (.07)** -2.19** 59 p value 32.23 <.01 2.47 .12 10.49 <.01 7.28 .01 Table 7 (cont’d) “Brand Passionate Helpers” Independent Variables Like-minded Discussion Rewards (Hedonic) Rewards (Utilitarian) Seeking Assistance SelfExpression Up-to-date Information Validation R a Ward Statistic a Class 1 n = 130 Class 2 n = 68 p value .38** (.09) * 4.13** .23** (.10) * 2.43** .03** (.09) * .33** .15** (.07) * 2.27** -.18** (.08) * -2.22** .22** (.09)** 2.50** -.04** (.09)** -.49** -.16†* (.09)** -1.74** -.06** (.08)** -.71** -.02** (.08)** -.31** -.11†* (.06)** -1.67** -.23** (.06)** -4.17** -.17** (.06)** -2.94** .14** (.06)** 2.32** .44** (.05)** 8.24 -.30** (.09)** 3.20** 10.74 .51** Helping 2 a “Individualistic Information Seekers” .83** .69 <.01 4.28 .04 1.50 .22 20.53 <.01 .01 .92 .57 .45 21.60 <.01 10.87 <.01 Estimate, (Standard Error), Z Statistic; * significant at α = .05; † significant at α = .10 Results of the analyses revealed that all 11 dimensions had a significant impact on the dependent variable across all classes. However, there was substantial heterogeneity in the sign and significance of these effects from class to class. Intentions to actively participate in a community for the first class (“Brand Passionate Helpers”) were primarily driven by a desire for helping other brand users, discussing the brand, expressing themselves, and their passion for the 60 brand. This suggests that this class participates in communities to be of assistance to other brand users. Importantly, this group is motivated by rewards (utilitarian), so it is important to provide compensation to this group. The second class (“Individualistic Information Seekers”) is primarily motivated by up-todate information and a desire for validation. Additionally, they are ambivalent with respect to helping, and like-minded discussion. This suggests that this class participates in communities to be on the leading edge of information about the brand and to receive validation from the group. While this class also has higher mean levels of participation, they also have more negative significant coefficients than the first class does (class 1: seeking assistance and validation versus class 2: brand passion, connecting, rewards (hedonic), rewards (utilitarian), and seeking assistance). The significant negative coefficients for dimensions of online brand community engagement were an unexpected finding. Online brand community engagement is defined as the compelling intrinsic motivations to continue interacting with an online brand community. Participation intentions was operationalized by a scale developed by Algesheimer et al. (2005), “I intend to actively participate in the brand community’s activities” (Algesheimer et al. 2005). The negative coefficients indicate that while the community members are more motivated to interact with the community, they are less likely to actively participate in the community. These findings suggest that these motivations tend to cause members to passively participate (e.g., "lurk," Yu-Chen 2006) instead of more active or visible contributions to the community. For example, seeking assistance and validation for the first class (“Brand Passionate Helpers”) had significant negative effects on participation intentions. This suggests that when seeking assistance and validation these members interact with the community through viewing rather 61 than posting. Specifically, when seeking assistance instead of asking the community for help, they search the community for answers to their questions. For validation, instead of soliciting feedback on the value and importance of their ideas from the community, these members scan the community for validation of their ideas. Ultimately, the results suggest that the online brand community engagement dimensions do an excellent job predicting consumer motivations for participating in online brand communities. Moreover, the results reveal substantial heterogeneity in the effects of motivations on community participation intentions. The sub-populations of community members differed so much in the effects of their motivations that some classes experience opposite signs for the effects on a number of key dimensions (e.g., brand passion, connecting, rewards (utilitarian), and validation). In addition, the significance of several dimensions differed across several dimensions (e.g., brand influence, helping, like-minded discussion, rewards (hedonic), and up-todate information). Failing to account for heterogeneity in online brand community engagement could lead to biased interpretations of the effects of these motivations on community member behavior. For one class the motivation may drive active participation in the community whereas for the other class it may drive passive participation (e.g., “lurking”). Differences Across Online Brand Community Engagement Segments Based on the proposed and observed heterogeneity of online brand community member motivations to participate in online brand communities, I explore the implications for brand purchase intentions. The following analysis explores the differences in brand purchase intentions across the latent classes. 62 Table 8 Essay One: Brand Purchase Intentions Measures Construct a New Product Trial (Kempf and Smith 1998, p. 34) a Item Assuming that this brand introduced a new product, and you have a need for a product like it: How likely would you be to buy one of the brand’s new product in the next couple of weeks? New Product Adoption (Kempf and Smith 1998, p. 34) Assuming that this brand introduced a new product, and you have a need for a product like it: How likely do you think you would be to buy that product again? Purchase Intentions (Algesheimer et al. 2005, p. 33) I intend to buy this brand in the near future Unless indicated otherwise, all items measured on an 11 point Likert ranging from “0 Strongly a Disagree” to “10 Strongly Agree”; 11 point Likert ranging from “0 Zero Likelihood” to “5 Even Chance” to “10 Completely Certain.” For analysis, all values were recoded 1 to 11. Brand Purchase Intentions Measures Brand purchase intentions are the member’s intentions to purchase the brand. Specifically, new product trial is the member’s intentions to try new products from the brand (Cardozo et al. 1988). New product adoption is the member’s intentions to repeat buying the new product (Cardozo et al. 1988). Purchase intentions is the intentions to purchase the brand in the near future and was adapted from existing scales (Algesheimer et al. 2005). The 11 dimensions of engagement and three dimensions of brand purchase intentions were assessed using confirmatory factor analysis. The measurement model indicates that there is a good fit with the data (independence model χ2 = 8,766, df = 993, p < .01; CFI = .93; SRMR = .05; RMSEA = .06). The three brand purchase intentions constructs were all measured with single items, so their discriminant validity was assessed using the fix and free method of 63 evaluating the effect on model fit of constraining correlations between the three constructs at unity. Discriminant and convergent validity of all other constructs was assessed using Fornell and Larcker (1981). The single item constructs demonstrate adequate discriminant validity (new product trial and new product adoption χ2 = 3.51, p = .06; new product trial and purchase intentions χ2 = 42.87, p < .01; new product adoption and purchase intentions χ2 = 29.35, p < .01). All constructs with more than 1 item demonstrated discriminant validity (Fornell and Larcker 1981), convergent validity (average variance extracted ranged from .67 to .80), and latent construct reliability (ranged from .78 to .88) (Hair et al. 2006). The general linear model multivariate analysis of variance (MANOVA) was used to evaluate the differences in the levels of brand purchase intentions (i.e., new product trial, new product adoption, and purchase intentions) across the two classes identified with the mixture model. MANOVA is an efficient way to assess the differences across classes using one multivariate procedure, rather multiple univariate procedures which increase the probability of a Type I error. MANOVA assumes that the distributions of dependent variables are each normal and visual analysis of histograms of the dependent variables showed that brand purchase intentions are skewed left. However, MANOVA is robust to violations of univariate normality with large enough sample sizes (Hair et al. 2006, p. 410). Outlier analysis was also conducted and there were several outliers for each of the dependent variables, but there were no consistent cases that appeared as outliers across all of the dependent variables, so no cases were dropped from the analysis. Next, Box’s M-test revealed that there were not significant differences in the variance/covariance matrices of the dependent variables (Box’s M = 11.48, p = .08). Levene’s test for the equality of error variances indicated that the error variance of new product trial (F = 64 0.24, p = .62) and new product adoption (F = 2.38, p = .13) did not differ across classes, but that the error variance for purchase intentions differed across classes (F = 5.32, p = .02). Because MANOVA is robust to moderate violations of this assumption (Phillips et al. 1999, p. 946), I proceeded with the analysis. Additionally, I calculated and report the eta-square statistic as a measure of effect size. Eta-square represents the percentage of variance in the dependent variable explained by the independent variable. Multivariate results are reported in Table 9 and univariate results are reported in Table 10. A graphical depiction of the data across dependent variables based on classes is presented Figure 1 with median plots. Table 9 Essay One: Differences in Brand Purchase Intentions Across Classes—Multivariate Results for MANOVA on Study 5 Final Validation Study Data Multivariate results Wilk’s λ F Online Brand Community Engagement Typology .96* 2.72* 2 η .04 * Significant at α = .05 Table 10 Essay One: Differences in Brand Purchase Intentions Across Classes—Univariate Results for MANOVA on Study 5 Final Validation Study Data New Product Trial 2 F η 4.38** .02 New Product Adoption 2 F η 6.81* .03 Online Brand Community Engagement Typology * Significant at α = .05; † significant at α = .10 65 Purchase Intentions 2 F η 3.02†* .02 Figure 1 Essay One: Differences in Brand Purchase Intentions Medians Across Classes 0 2 4 6 8 10 12 Study 5 Final Validation Study New Product Trial New Product Adoption Class 1: Brand Passionate Helpers Purchase Intentions Class 2: Individualistic Information Seekers Purchase intentions was measured on an 11 point Likert ranging from “0 Strongly Disagree” to “10 Strongly Agree”; new product trial and new product adoption were measured on an 11 point Likert ranging from “0 Zero Likelihood” to “5 Even Chance” to “10 Completely Certain.” Prior to analysis all values were recoded 1 to 11. Whiskers are 95 percent confidence intervals for median. The MANOVA reveals support for proposition 2. There were significant differences across the classes for mean levels of new product trial (P2a supported), new product adoption (P2b supported) and purchase intentions (P2c supported). Analysis of the mean levels shows that Class 1, “Brand Passionate Helpers,” had significantly higher new product trial, new product adoption, and purchase intentions. For new product trial and new product adoption, “Brand 66 Passionate Helpers” were on average an entire scale point higher (x= .93) and (x= .95) � � respectively than Class 2, “Individualistic Information Seekers.” Purchase intentions differed scale point (x= .62). � between “Brand Passionate Helpers” and “Individualistic Information Seekers” by over half of a Based on the MANOVA analysis, the data suggests that there are several significant differences across classes. First, from a managerial perspective, not all community members have the same brand purchase intentions. Significant differences in brand purchase intentions suggests that promotions aimed at stimulating brand purchase behavior should be targeted rather than applied in mass. If the desired outcome of promoting brand purchase behavior is increased share of wallet, then “Individualistic Information Seekers” should be targeted to increase their purchase levels. If the desired outcome is increased brand purchases, then “Brand Passionate Helpers” should be targeted because they have higher intentions to purchase the brand. Second, the two classes of community members have distinct motivational profiles for active participation with the community. This analysis links community engagement with community member brand purchase intentions. While no causal analysis is addressed in this part of the analysis to explore the mechanisms through which differences in brand purchase intentions should occur, it is nonetheless interesting to note that differences do exist. “Brand Passionate Helpers” have the fewest passive participation (“lurking”) motivations (i.e., seeking assistance and validation) and have the highest brand purchase intentions. Conversely, “Individualistic Information Seekers” have the greatest number of passive participation motivations (i.e., brand passion, connecting, rewards (hedonic), rewards (utilitarian), and seeking assistance) and how lower brand purchase intentions. Therefore, active participation in the community is related to higher levels of brand purchase intentions. 67 Lastly, the mean differences across classes of new product trial and new product adoption intentions were marginally larger with than purchase intentions. Prior research has suggested that participation in online brand communities enhances new product trial and new product adoption (e.g., McAlexander et al. 2002; Thompson and Sinha 2008). Perhaps the slightly larger mean differences for new product trial and new product adoption intentions relative to purchase intentions is the reduction in risk for trying new products that could be greater for community members actively participating in the community. Overall, these differences highlight how engagement is related to brand purchase intentions. Surprisingly, “Brand Passionate Helpers” tend to have the highest levels of brand purchase intentions instead of “Individualistic Information Seekers.” Further analysis is needed to explore the mechanisms through which engagement influences brand purchase intentions. In Essay Two, I build on the findings here to address mechanisms through which engagement can influence both brand and community outcomes. Comparison of Predictive Ability of Online Brand Community Engagement Typology and Rival Role-Based Typology A rival paradigm to looking at motivations for participation is to look at behaviors of community members. The rival typology used for comparison with the online brand community engagement typology is the role-based typology created by Fournier and Lee (2009). This typology identifies 18 roles “critical to a community’s function, preservation, and evolution” (Fournier and Lee 2009, p. 109). These roles are supposed to keep members involved in and add value to the community. While role-based typologies require observation of behavior to code members into the various roles, due to the breadth of communities covered in the data collection monitoring 68 community member behavior across the public and private communities was simply not possible. Additionally, the various roles are highly idiosyncratic to the research team and there is not enough published information to replicate their coding scheme. Respondents in the final validation study were asked to select the role they most often perform in the community based on the role descriptions published by Fournier and Lee (2009). The roles and descriptions were presented to respondents in a randomized order to reduce any potential order effects (please see Table 11). Having community members self-select into the roles they most often perform is a very conservative test of the rival typology because it reduces researcher error in classification of behaviors. It is however, still limited by the ability of the respondents to reflect on their behavior in the community and select the most suitable matching description of their behavior. Given the breadth of the roles, there are ample options for community members to select. 69 Table 11 Essay One: Rival Role-Based Typology Measures (Fournier and Lee 2009) Role-based Typology (Fournier and Lee 2009, p. 109) Which of the following roles do you most often perform in b this brand community? Accountant: Keeps track of people’s participation Ambassador: Promotes the community to outsiders Back-Up: Acts as a safety net for others when they try new things Catalyst: Introduces members to new people and ideas Celebrity: Serves as a figurehead or icon of what the community represents Decision Maker: Makes choices affecting the community’s structure and function Greeter: Welcomes new members into the community Guide: Helps new members navigate the culture Hero: Acts as a role model within the community Historian: Preserves community memory; codifies rituals and rites Learner: Enjoys learning and seeks self-improvement Mentor: Teaches others and shares expertise Partner: Encourages, shares, and motivates Performer: Takes the spotlight Provider: Hosts and takes care of other members Storyteller: Spreads the community’s story throughout the group Supporter: Participates passively as an audience for others Talent Scout: Recruits new members b Single option response Latent class regression in Latent Gold 4.5 was used to compare the predictive ability of the Fournier and Lee (2009) typology with the online brand community engagement typology. Analysis of the data shows that some of the roles are much less common than others (e.g., only 6 of the 18 roles had more than 5 respondents classify themselves into that role), therefore there are sparsely populated cells in the matrix analyzed for the rival role-based typology. A sparsely populated matrix can create problems with convergence as the number of clusters identified increases (see Table 12). A two cluster solution for the rival typology was selected for two 70 reasons. First, there is a significant improvement in model fit by going from one cluster to two clusters (χ2 = 58.70, p < .01). Second, the model fails to converge with three or more clusters. Several different classification statistics were used to compare the predictive ability of the two typologies (see Table 13). Table 12 Essay One: Frequency Table for Roles in Study 5 Final Validation Study Data Role Accountant Ambassador Back-up Catalyst Celebrity Decision Maker Greeter Guide Hero Historian Learner Mentor Partner Performer Provider Storyteller Supporter Talent Scout Total Frequency 0 5 2 9 0 Percent 0.0 2.5 1.0 4.6 0.0 0 0.0 2 6 1 3 75 13 23 1 0 3 53 2 1.0 3.0 0.5 1.5 37.9 6.6 11.6 0.5 0.0 1.5 26.8 1.0 198 100.0 71 Table 13 Essay One: Comparison of Engagement and Role-Based Typlogy at Predicting Participation Intentions Using Study 5 Final Validation Study Data Prediction Statistic Average Weight of Evidence Online Brand Community Engagement Typology 1,165.63 Rival Role-Based Typology (Fournier and Lee 2009) 1,262.33 BIC 814.95 910.47 CAIC 841.95 941.47 Entropy R-squared .48 .62 Classification errors .17 .12 Comparison of the prediction statistics between the two typologies indicates that the online brand community engagement typology has at least as good predictive ability to the rival role-based typology. The Average Weight of Evidence criteria (AWE) is smaller for the online brand community engagement typology, indicating that the online brand community engagement typology performs better, has better fit, and is more parsimonious than its rival role-based typology. In addition the BIC and CAIC are substantially smaller for the online brand community engagement typology suggesting it also fits the data better than the rival role-based typology. The entropy R-squared is slightly higher for the role-based typology, suggesting that it provides slightly better separation between the clusters (Wedel and Kamakura 2000, p. 92). Because both values are moderately high, it suggests that future research and larger sample sizes are needed to assess the heterogeneity of communities at a more refined level. A key challenge pointed out by Dröge et al. (2010) is that size of communities, and by extension size of segments, can be relatively small in terms of absolute numbers, but can be very important. Therefore, the challenge for researchers is to balance identifying a sufficiently large enough number of latent segments of community members to capture the important small segments without overfitting the 72 data. It is especially challenging in segmentation analysis to work with classes that have substantial differences in size. The primary purpose of this evaluation however was achieved by assessing at a broad level how well a motivationally based typology predicts participation intentions compared to a role-based typology being used to predict participation intentions. Analysis of the classifications tables for both models suggests that class 1 in general tends to be easier to define than class 2 (see Table 14). A potential explanation for the relatively large percent of misclassification for class 2 for both typologies is that there are more latent classes in the actual data than are captured with the use of a two class solution. Larger sample sizes would be needed to evaluate models with 3 or more latent classes. Conceptually it makes sense that the remaining class may be more heterogeneous than the first class because there could still be small distinct classes of community members (e.g., hardcore members of the community) that will be masked in the analysis as part of a similar, but larger class of individuals. Table 14 Essay One: Classification Tables Using Study 5 Final Validation Study Data Online Brand Community Engagement Typology Rival Role-Based Typology (Fournier and Lee 2009) Modal Probabilistic Class 1 Class 2 Class1 116.81 20.93 Class2 13.19 Total 130.00 Total Class 1 Class 2 137.74 114.61 20.03 134.64 47.07 60.26 3.39 59.97 63.36 68.00 198.00 118.00 80.00 198.00 73 Total DISCUSSION This paper has contributed to marketing theory and practice by developing a scale for online brand community engagement. This scale represents a significant improvement to the brand community literature and an important contribution for marketing strategy, social media marketing and relationship marketing research for several reasons. First, the breadth of motivations online brand community members have is substantially more than the extant conceptualizations of engagement capture. Failing to recognize the diverse motivations dramatically understates the effects of engagement on consumer behavior and makes it difficult to determine the sign and significance of the effects of motivation. Furthermore, the breadth of motivations identified in this study open the door to analysis of differential effects for the dimensions of online brand community engagement on constructs of critical importance (e.g., brand purchase intentions). Second, this paper contributes to the brand community literature by showing that online brand community members can be meaningfully segmented based on their levels of engagement. Understanding that brand communities are more heterogeneous than they are currently being treated, can help turn an investment in a social presence into a strategic marketing asset. Understanding the different motivations the segments of community members have provides a basis for better tailoring the experience online brand community members have, reducing the potential of wasted marketing expenditures and alienating members with too frequent or irrelevant marketing activities. Third, the engagement typology can be used to explain passive participation, or “lurking” (Madupu and Cooley 2010). Community members may be highly motivated to interact with the community, but seldom contribute or visibly participate. Community managers trying to create 74 and maintain social interactions with community members struggle to get community members talking and actively participating which is often used as a visible metric of how vibrant the community is. “Lurking” is an especially pressing problem in an online environment, where “lurking” rates can be as high as 90 percent or more (Madupu and Cooley 2010, p. 130). The engagement typology shows that community members who “lurk” also have other significant motivations to actively participate in the community. Seeking assistance was the only motivation that I found to have negative effects on participation intentions for all classes of community members. This was an unexpected finding, but makes sense when considering the nature of the communities. Community members tend to perceive other members as similar to themselves and therefore likely assume that other community members have faced similar problems to the ones they are facing. Thus, seeking assistance, while it can be a strong motivator to interact with the community, the interaction can take the form of searching rather than posting. Fourth, the conservative test of differences between classes provides an illustration of the subtle but significant differences that exist in online brand community members. While the effects are relatively small, this is not entirely unexpected. The latent classes were developed using all 11 engagement dimensions to identify segments of community members with similar drivers to participate in online brand communities. Larger effect sizes would be expected had separate latent class regression been performed using the 11 dimensions of engagement for each brand purchase intention. However, the primary point of the MANOVA analysis was to explore differences in classes on important dependent variables. This analysis does support high likelihood of community members to try and adopt new products as mentioned by McAlexander et al. (2002, p. 46). Additionally, community members across classes show very high levels of purchase intentions for the brand which supports the moral responsibility identified by Muniz 75 and O'Guinn (2001) which in a brand community context can be conceptualized as purchasing the brand to ensure its survival. Lastly, the online brand community engagement typology developed in this study substantially improves both theory and practice in online brand community research. Instead of adapting generic engagement measures developed in different contexts, or complex role-based segmentation techniques, the engagement typology can be used to segment brand community members very efficiently. The rival analysis of the engagement typology shows mixed results regarding the predictive ability of the online brand community engagement relative to the rolebased typology. While fit statistics are comparable between the two typologies, the benefits of the engagement typology are primarily visible in the use of the typology. Ease of use and switching costs are a key determinant of adoption of new technology (Gentry and Calantone 2002). Likewise, the online brand community engagement typology is substantially easier to use than a role-based typology because it can be administered as a survey and does not require a large staff of trained ethnographers to observe and classify community member behaviors for extended periods of time. The contextual specific behaviors community members engage in highlight the subtle nature of social interactions. Identical behaviors (e.g., comments) can have a range of meaning based on the context. Identifying when a community member is being a “Performer” versus when a community member is being a “Guide” or “Hero” is a subtle and difficult task. Furthermore, it would require a large number of observations of each community members behavior to accurately classify them into a role (i.e., a pattern of behavior over time). Large and active communities could prove to prohibitive to analyze using a role-based perspective from a practical standpoint due to the tremendous volume of data online communities generate. The 76 potential savings and increased speed of classifying community members using an engagement typology would give marketers a competitive advantage above marketers using role-based typologies or no typologies at all. Managerial Implications Generic discussions of community members being motivated to interact with a community and not scientifically validated tips from individuals “claiming to be familiar with the keys to success in this area” (Cova and Pace 2006, p. 1089) are very common in the online brand community popular press. This research provides significant additions to this dialogue by introducing a much more refined and rigorously developed and validated study of online brand community member motivations. Specifically, this research has especially important implications for marketing research firms and marketing consultants to better understand online brand community members. Online brand community engagement is much more complex than previously thought. While there are diverse motivations for members to interact with a brand community, there are two general types of brand community members: brand passionate helpers and individualistic information seekers. It is important to understand the different motivations these types of members have because applying marketing activities broadly to a brand community could in fact reduce member participation. I have developed short-form scales to help community managers assess the segments of engagement that drive participation in their community. With the high costs of establishing and maintaining a community, understanding member motivations could provide substantial additional savings by reducing turnover, increasing response to marketing activities, and increased satisfaction. Targeted marketing activities should 77 be employed within a brand community based on the engagement groupings members belong to. Based on this research, there are three implications for managers that I will focus on. First, which group of community members should community managers focus on, brand passionate helpers or individualistic information seekers? Community managers can gain different benefits by focusing on the two different classes of community members. If the strategic objective of the community is to have a support focus, then it would be best for managers to focus on the brand passionate helper segment. However, if the goal of the community is to have a highly active community to achieve higher exposure rates or participation frequency, then targeting the individualistic information seekers would be a more appropriate group for them to focus on. Second, online brand communities are frequently used by marketers as a promotions platform in an integrated marketing communication strategy. Based on the results in this study, sending coupons, deals, and discounts to individualistic information seekers is the best strategy to drive traffic in a community. Individualistic information seekers have the highest mean level of participation and have the strongest relationship with rewards (utilitarian). Promoting to the individualistic information seekers will not however increase their active participation in the community. Promotions targeted to brand passionate helpers will stimulate an increase in active participation in the community. Third, this study is one of the first studies to identify the specific motivations of passive participators (“lurkers”) in an online brand community. Passive participation (“lurking”) is a substantial problem in online brand communities, especially in marketing research online communities (MROCs) where respondents are typically carefully recruited and paid to participate. My findings indicate that seeking assistance is the only motivation of the 11 78 identified in this study to consistently lead to passive participation across classes. Interestingly, the other passive participation motivations differ across the classes. “Individualistic Information Seekers” have many more passive participation motivations (i.e., brand passion, connecting, rewards (hedonic), rewards (utilitarian), and seeking assistance) than “Brand Passionate Helpers” (i.e., seeking assistance, and validation). One way that community managers can help convert passive participation into active participation (i.e., reduce “lurking”) is to tailor marketing activities based on their passive participation motivations. For example, a firm could use community analytics (e.g., page views) to capture which pages and areas passively participating community members visit. Community managers could then identify which motivations these areas of the site fit and then tailor activities and discussions centered around these motivations to convert members from passive to active participation in the community. Further analysis, like total unduplicated reach and frequency (TURF) analysis can be used to identify the primary dimensions of engagement that will highly motivate members of the community. Using this insight, community managers can then tailor community structure and activities to appeal to the broadest base of community members. In addition, response surface analysis can be used to explore potential changes in participation based on engagement. This analysis would provide insights into what types of community members should be recruited to maximize participation intentions, as well as when refreshing community membership who should be retained. Limitations and Future Research This research makes a substantial departure from the dominant paradigm of small sample case 79 analysis from a few communities pervasive in the brand community research. This research program has made every effort to assure the generalizability of the results by sampling from as large and broad a pool of online brand community members as possible. This research shows differences in face-to-face and online brand community member motivations to interact with the brand community. The tremendous diversity in format and content of online brand communities is staggering. Just as caution is warranted in generalizing the findings of small sample community ethnographies, the tremendous diversity of online brand communities necessitates calibrating generalized findings to specific communities. The heterogeneity of members in class two, the “Individualistic Information Seekers,” indicates that perhaps with larger sample sizes additional classes of community members could be identified. Future research should work to categorize and explore the engagement norms of the diverse sub-types of brand communities (e.g., research oriented online brand communities, public and private online brand communities, etc.). In addition, future research should explore the mechanisms through which online brand community engagement affects brand and community performance, as well as marketing activities marketers can use within brand communities to enhance brand and community performance. 80 Essay Two INTRODUCTION Marketing research online communities (MROCs) are the number one emerging technique in marketing research (GreenBook Winter 2013) and represent the next step in the evolution of online brand communities from consumer created, to firm sponsored, to marketing research firm created and managed. The brand community literature has developed considerable knowledge regarding the effect brand communities have on brand loyalty. From the first seminal article on brand communities, Muniz and O'Guinn (2001) discuss brand loyalty as one of the core outcomes of a brand community. McAlexander et al. (2002) further focused brand community research on brand loyalty by calling brand communities the “Holy Grail of brand loyalty” (McAlexander et al. 2002, p. 38). Since these two seminal articles, the brand community literature has heavily focused on brand loyalty as the outcome for brand communities. While this focus on brand loyalty has yielded important insights into online brand communities, it has also overlooked an important aspect of online brand communities and their evolving role in modern marketing research. The literature has become overly focused on what people say in brand communities, overlooking why they are there and why they visibly participate. This fundamental paradigm shift permits greater insights into the strategic implications of online brand communities as strategic marketing assets. “Membership and participation in the brand community should… have an impact on the consumer’s brand-related behaviors... because a key marker of community membership is ongoing purchase and use of the brand” (Algesheimer et al. 2005, p. 23). MROCs provide an excellent setting to execute marketing activities and marketing research. Leveraging online brand communities and more specifically MROCs to influence brand 81 and community outcomes is of critical importance to academic and professional audiences as marketers and consumers are increasingly using them and the academic literature provides little guidance. Increasingly, marketers for firms of many different sizes in a vast array of industries are creating and maintaining social presences online. One of the primary effects of creating this social presence is the formation of communities around brands. Community managers launch marketing activities and communications in their online communities to create and maintain this social presence, but also to help curate other benefits for the brand (including a loose hierarchy of effects for brand assessments, brand related behaviors, and community related behaviors). However, marketing activities have a diminishing return and can alienate consumers if the content is not what members want or if the members receive too many from the firm (Fournier et al. 1998; Kumar et al. 2010). To obtain increased benefit from these activities, that is leverage them, a more refined view of the motivations online brand community members have and how community members perceive different types of marketing activities is needed to increase the benefits of MROCs and online brand communities. This research essentially builds on the notion that human beings tend to be social creatures and their behavior tends to be influenced by their reference groups (Brewer 2003; Brewer 1991; Fournier and Avery 2011; Leonardelli et al. 2010). As such, community members who are engaged with MROCs and online brand communities tend to have their behavior influenced by the community reference group. However, the literature provides mixed reviews on the effectiveness of online brand communities being able to achieve strategic objectives using online brand communities. Much of what is documented in the academic literature focuses on qualitative objectives and outcomes and the substantially smaller amount of empirical research typically focuses on loyalty and word-of-mouth as the primary outcomes from online brand 82 communities. There have been several calls in the academic literature to develop metrics to evaluate online brand communities and many sponsors of online brand communities place continual pressure on marketing research firms providing community management services to demonstrate value added for their services (e.g., increased purchases of the brand). Overall, the effectiveness of MROCs and online brand communities as a strategic marketing asset is a complex and emerging area of research and importance. The academic literature has failed to provide sufficient theoretical or practical guidance to academic researchers and marketers. First, there have been several ethnographic studies and numerous case studies focused on describing brand communities. Very few of these have actually been conducted in an online environment and few if any have been conducted in recent years where the interface consumers have to interact online has changed dramatically with advances in multimedia sharing on the Internet. By selecting unique and successful online brand communities (e.g., Bernoff and Li 2008; Dholakia et al. 2009; Fournier and Avery 2011; Israel 2012; Muniz and Schau 2005) they have provided rich descriptions and important insights, but what about the communities beyond these highly successful online communities and beyond highly successful face-to-face communities like Harley Owners Group (e.g., Fournier 2000; Muniz and Schau 2005; Schouten and McAlexander 1995), Jeep Jamborees (e.g., McAlexander et al. 2002), and European car clubs (e.g., Algesheimer and Dholakia 2006)? What about communities with everyday consumers, not just brand zealots and lead users? Additionally, many studies and marketers tend to treat MROCs and online brand communities as homogenous marketing platforms, with the exception of a few studies that describe the heterogeneity of online brand communities. Lastly, the academic literature has not addressed leveraging MROCs and online brand communities to achieve specific objectives. While the relationship between online 83 brand communities and outcomes of interest have been documented, it falls short of assessing the degree to which marketing activities (e.g., creating discussion forums, running surveys, etc.) can actually affect a loose hierarchy of brand and community effects. This study contributes to the marketing literature in two primary ways. First, leveraging marketing activities and community member motivations is introduced as a key interaction for brand and community outcomes. Leverage is the amplifying of the effect of marketing activities on psychological sense of community with matching marketing activities that fit with the community member’s engagement dimensions. A marketing activity is an activity the firm creates (e.g., contests, sweepstakes, discussion forums, media galleries, brainstorming sessions, quick polls, and surveys) for community members to participate in. Using the refined view of community member motivations for participating in online brand communities, I test the leveraging of marketing activities and online brand community engagement dimensions to achieve greater psychological sense of community. Matching marketing activities that resonate with community member motivations should provide added benefit to marketers. Second, this study assesses the degree to which an MROC can achieve a loose hierarchy of effects for the brand and community. The social dynamics of online communities are very different from the social dynamics of face-to-face brand communities. Consequently, the degree to which online communities can influence more resource intensive behaviors needs to be tested. I found that MROCs are important strategic marketing assets for firms that can achieve a loose hierarchy of effects for brand assessments (i.e., brand commitment and oppositional brand loyalty), supportive brand behaviors (i.e., word-of-mouth, defending the brand, and willingness to pay a price premium) and supportive community behaviors (i.e., participation intentions, and community participation). Marketing activities and online brand community engagement help 84 create a psychological sense of community in community members. This psychological sense of community positively affects brand assessments, supportive brand behaviors, and supportive community behaviors. I found limited support in the data for leveraging online brand communities by matching marketing activities with related online brand community engagement dimensions. This essay is organized as follows. I review the background literature regarding brand communities, relationships consumers form with communities, and leveraging online brand communities. Next, I develop hypotheses for the mechanisms through which these marketing activities can leverage online brand community engagement to affect brand and community outcomes. Lastly, I describe the methods used to test hypotheses, the research setting, analysis and results. This essay concludes with a discussion of the contributions, limitations, and directions for future research. LITERATURE BACKGROUND “Organizations both large and small have jumped on the social media bandwagon, feeling their way around to make sense of its usefulness. They have tweeted on Twitter, created fan pages on Facebook, and posted videos on YouTube. Perhaps akin to the development of websites in the latter part of the 20th century, organizations today sense that social media is–—and will remain–—an important fabric of commerce, and that they must get on board. However, given the frequent demand by management for ‘proof’ of return on investment (ROI), it appears that there is a fair degree of uncertainty with respect to allocating marketing effort and budget to social media, and limited understanding of important distinctions among various types of social media.” (Weinberg and Pehlivan 2011, p. 275) Humans are inherently social beings, and as social beings social context influences their behavior (Brewer 2003; Brewer 1991; Fournier and Avery 2011; Leonardelli et al. 2010). Brockner (1983) described the degree to which an individual is susceptible to social influence as 85 behavioral plasticity. Lasting social influence on individual behavior has been described as conversion (e.g., Kanter 1972), where individuals take on the values of the organization. Social influence tactics to effect conversion have also been studied for a long time (e.g., “insisting on ideological conversion for membership, requiring vows to change behavior on the part of recruits, requiring a probationary period, rejecting potential members as unacceptable, and requiring some sort of ‘test of faith’”; Kanter 1972, p. 122). In addition, sociological studies of communities and conversion led to the study of communities from a hierarchical perspective (e.g., Fox 1987). We are witnessing the evolution of the study of relationships consumers have with brands from dyadic to sociological foundations. Early research on consumer relationships with brands focused primarily on the dyadic relationship of a consumer with the brand (e.g., brand loyalty, brand identification, brand commitment). Aaker’s brand personality and anthropomorphization study of consumer relationships with brands showed that brands can and are humanized by consumers, enabling brands to form relationships with consumers like consumers do with other individuals (Aaker 1997). Social psychology ushered in a new era of research looking at the network of relationships consumers form around brands. With the seminal works of Muniz and O'Guinn (2001) and McAlexander et al. (2002) the fundamental emphasis of studying relationships consumers formed with brands evolved to studying the network of relationships consumers form around the brand. This next step in the evolution of the study of consumerbrand relationships spawned a lot of interest in the brand related consequences of these relationships. Many of these studies were conducted in face-to-face settings (for example, Algesheimer et al. (2005) studied European car clubs, Muniz and O'Guinn (2001) studied cars and computers, and Schouten and McAlexander (1995) and Fournier (2000) studied Harley 86 Owners Groups). However, with the corresponding evolution of the Internet, the tremendous advances in technology that enabled widespread availability of high-speed Internet and enhanced website experiences with more multimedia and interactive capabilities, online brand communities emerged as the next step in the evolution of brand communities. Marketers were no longer constrained by having to organize rallies and relying on their community members to travel large distances to meet at specific places at specific times, instead their communities could be available 24 hours a day 7 days a week to anyone anywhere with Internet access. With these advances, several descriptive studies and few empirical studies have been conducted in online environments. However, with the descriptive approach, highly visible online communities (e.g., Audi’s virtual lab, Dell’s Ideastorm, IBM’s numerous online communities, Intuit’s business community, Nike’s Facebook Fan Page, Proctor and Gamble’s Pampers Village, Salesforce.com’s support community, RedSox Nation, and Starbucks’ MyStarbucksIdea) have been selected and studied, omitting a more comprehensive assessment of online brand communities in general. Effectively utilizing social relationships around a brand to achieve strategic objectives has been debated in the literature. Some have argued that since activities in communities are broadcast they will not affect individuals in the same way social studies have indicated (e.g., Schlosser 2003). Fournier et al. (2005) provides an excellent illustration of the opposing views of trying to strategically utilize online brand communities. For example, on the one hand Fournier et al. (2005) state: “Successful brand communities will always attract the attentions of management seeking to leverage them more fully so as to capture more value for the firm. And, it is in this, the ultimate paradox, that success breeds failure, for the community that is tampered with is quickly destroyed.” (p. 19). 87 Marketers following this philosophy will leave the community management to the community members, letting the members direct the course of the community. Such a visible and long-term social medium without management involvement left to its own devices could quickly degenerate to the reverse monologue that threatened the continued existence of Dell’s IdeaStorm in early 2011, one of the most prominent and successful online brand communities soon after it launched. On the other hand, Fournier et al. (2005) state: “[building a] brand community demands an unusual amount of patience, hard work, and devotion on the part of the company. Unfortunately, many companies desirous of community benefits have been unwilling to invest the necessary and involved labor to gain them.” (p. 19). Therefore, Fournier et al. (2005) state that firms “must invest the necessary and involved labor” to create a community (p. 19). Discussions with marketing research firms responsible for creating and maintaining online brand communities indicate that one of their primary frustrations is that clients vary in terms of the degree to which they are interested in making online brand communities a key component of their overall marketing strategy. Fournier et al. (2005) conclude that “managers must walk the fine line between commercial marketing activity and credible authenticity when solving this equation for the brand” (p. 20). Interestingly, some very strong brand communities have emerged without any support from the firm (e.g., Nikonians and Nikonites for Nikon, Lugnut for Lego, and almost countless Yahoo Groups centered on specific brands). Sometimes online brand communities have even formed around a brand or product under direct opposition from the firm (e.g., Apple's Newton; Muniz and Schau 2007). With diversity of online brand communities and the rapid evolution of technology enabling firms to form social relationships with consumers in an online environment, there has been a lack of theory developed to help guide research and practice in this area. Several 88 publications have recognized the limitations of the marketing literature with respect to poor theory development and testing in this area. For example, the current Marketing Science Institute’s (MSI) research priorities reports states “Customers are moving and connected more than ever. We need to get a better sense of what is on their minds and what they are doing at the same time.” The MSI research priorities report also called for new insights into people in their roles as consumers and into the way that marketers are using new technologies and methods (e.g., online brand communities) to generate insights and enhanced customer experiences (MSI 2012). The MSI report also explicitly calls for answers to the questions “Which experiences make a difference with consumers and which are not worth the investment?” and “What new tools for generating consumer insights are valid and which ones are just a fad?” Similarly, Cova and Pace (2006) called for “above and beyond check-lists drawn up by consultants claiming to be familiar with the keys to success in this area…, it is clear that efforts have to be made to strengthen good marketing practices in this respect...” (p. 1089). Online brand communities are likely to remain an important issue for academic researchers and marketers for many years to come because as online brand communities fundamentally tap into “one of the most basic human motivations: the desire to feel accepted, to fit in, and to belong” (Brewer 2003; Brewer 1991; Fournier and Avery 2011, p. 195; Leonardelli et al. 2010). Therefore, both academic and practitioner audiences are calling for better theory to guide research and practice in the area of online brand communities and MROCs. Highly engaged brand community members are a significant source of competitive advantage for firms because they have strong relationships with the brand and community of brand users. These relationships, called covenantal relationships, are characterized by affective ties, a moral obligation to watch out for each other’s interests, mutual respect, support, and 89 accountability (Graham 1991, p. 252; Muniz and O'Guinn 2001). Covenantal relationships form an enduring bond between the member and the brand, reducing customer acquisition and retention costs, increasing word-of-mouth, and enhancing the stability of the relationship over time (even in the presence of failures that would terminate normal relationships between customers and firms) (Graham 1991). Highly engaged brand community members are also desirable for firms to have in a brand community for several reasons. First, they possess two key characteristics—motivation and ability—for a firm to influence their attitudes in a lasting and resistant manner. Highly engaged members are the most involved members in a brand community, and as such are highly motivated to process information regarding the brand. The Elaboration Likelihood Model (ELM) operationalizes motivation, the prerequisite to effortful (central) processing of information, as personal relevance (e.g., Celsi and Olson 1988). Highly engaged members also tend to have been involved in the community for a long period of time, accumulating extensive knowledge of and experience with the brand and its use. ELM research treats knowledge and experience as the ability to process information (Celsi and Olson 1988; Gotlieb and Swan 1990). Therefore, highly engaged members possess the motivation and ability to form veridical attitudes that persist over time, are resistant to change, and tend to be favorable toward the brand. Dröge (1989) found that the type of processing highly engaged brand community members engage results in a stronger relationship between attitudes and behavior than less engaged community members (p. 202). Second, highly engaged member interaction in the community not only reduce marketing costs for the firm (e.g., high levels of loyalty, evangelist marketing, and providing customer support), but can also function as an indicator of the health of the brand community. For example, reductions in the proportion of highly engaged members in a 90 community following a strategy change can serve as an early warning sign to the firm. Specifically, highly engaged members will likely be among the first affected by strategic changes. If the change is bad enough to alienate highly engaged members, then it will likely have an even more dramatic effect on other members. The relationships between highly engaged brand community members and the brand take actions and commitments by both the member and the firm to develop such enduring relationships. There are two perspectives on the process of forming these relationships (i.e., the process through which enduring relationships are formed and psychological sense of community is developed): the individualist perspective which emphasizes the individual’s role in building the relationship (Schouten 1991; Schouten and McAlexander 1995) and the structuralist perspective (Kanter 1972) which emphasizes what actions marketers should take to build these relationships. While not all community members will have enduring relationships with the brand, the relationships they do have can be strengthened in the process of both individual and firm actions aimed at strengthening these relationships. Highly engaged online brand community members are “prototypical” community members (Hogg 2003, p. 468). As Hogg (2003) explains, prototypical group members are ones who demonstrate norm consistent behavior (p. 470), are generally liked more than less prototypical group members (p. 471), and can be very influential in the group (p. 472). Prototypicality is a “highly salient yardstick of group life” (Hogg 2003, p. 471) which means that as members become more prototypical, they move up the social hierarchy (e.g., from a peripheral member to softcore or softcore to core member; Fox 1987). Each stage of successively strengthening the relationship carries specific behavioral expectations and connotations, at the top of which are prototypical community members. 91 HYPOTHESIS DEVELOPMENT In addition to the brand community literature, this essay builds on the social psychology group processes and information processing literature. Specifically, the literature on social influences on individual attitudes (e.g., Baron and Kerr 2003), processing of persuasive cues (Petty and Briñol 2011; Petty and Cacioppo 1981; Petty and Cacioppo 1986; Petty and Wegener 1999), and the literature on social exchanges (Cook and Emerson 1978; Homans 1958; Homans 1974; Lovaglia 2007). One of the many interesting things about online brand communities and MROCs is that they are both a social group and a marketing asset. As such, social dynamics as well as deliberate marketing activities sponsored by the firm are at play. When studying online brand communities and MROCs it is important to account for not only the individual motivations, but also the firm marketing activities. Online brand communities that deliver what people want leads to stronger relationship with the community and ultimately a source of competitive advantage. Brand communities have traditionally been described as having three markers of community: consciousness of kind, shared rituals and traditions, and a sense of moral responsibility (Muniz and O'Guinn 2001), which is very similar to how Graham (1991) described covenantal relationships. Recent research into online brand communities has suggested that these three specific markers of community do not exist to the same degree across all types of brand communities (Carlson 2005; Carlson et al. 2008). Carlson (2005) and Carlson et al. (2008) introduced the concept of psychological sense of community, which has been defined as “the degree to which an individual perceives relational bonds with other brand users” (Carlson et al. 2008, p. 286). Psychological sense of community is a more general measure of the sense of 92 community than that provided by Muniz and O'Guinn (2001) and is more appropriate for an online setting. Brand communities in general are often more “imagined,” meaning they are constructed in the consumers mind, than distinct objectively observable communities per se, and in online communities the members rarely know the true identities of the other community members. What really matters is not necessarily the degree to which Muniz and O'Guinn (2001) markers of community are observable by others per se, but the degree to which members think they are part of a community. Prior research suggests that identification is the primary mechanism which drives the psychological sense of community (Carlson 2005, p. 50). Interaction with the online brand community or MROC ultimately facilitates identification with the community and brand. McAlexander et al. (2002) state that sharing “context-rich” and “meaningful consumption experience[s] [which] strengthens interpersonal ties and enhances mutual appreciation for the product, the brand, and the facilitating marketers. Virtual ties become real ties. Weak ties become stronger. Strong ties develop additional points of attachment” (McAlexander et al. 2002, p. 44). Marketing Activities Marketers tend to communicate frequently to community members by sending them marketing communications (e.g., survey invitations, discussion topics and summaries, important news about the brand, and special events or opportunities). These marketing communications to community members remind members that they are part of the community. Thus the marketing communications that are congruent with what community members had hoped for trigger a selfcategorization effect. As marketers send marketing communications to community members, 93 self-categorization enhances the salience of membership in the community, leading members to develop a psychological sense of community (Carlson 2005; Carlson et al. 2008). In addition, interactive marketing activities could help community members develop even further the psychological sense of community because community members are directly interacting with other community members. The four marketing activities I studied in this essay are firm sponsored brainstorming sessions, discussion forums, media galleries, and surveys. Brainstorming sessions are live discussions led by a community manager around a specific topic or idea. These sessions are open-ended activities and provide rich opportunities to influence the brand and its products. Discussion forums are conversational threads where members can contribute to an ongoing discussion around a certain topic. Media galleries are multimedia activities (typically photographs and movies) where members post and discuss pictures of themselves using the brand and or its products. Surveys are questionnaires around a specific topic where community members are asked to answer questions. These surveys typically are incentivized with some sort of reward to entice members to participate in them. While the quantity or depth of interaction between individuals across any specific single marketing activity may not be large or deep, the cumulative effect on building psychological sense of community can be quite large (e.g., like small-talk can be effective at building relationships; Stelzner and Dragon 2012). Please see Figure 2 for a graphical depiction of hypotheses. H1a-d: Marketing activities (a. brainstorming sessions, b. discussion forums, c. media galleries, and d. surveys) have a positive effect on psychological sense of community. 94 Figure 2 Essay Two: Model of Leveraging Online Brand Community Engagement Measured at Time 1 Measured at Time 3 Measured at Time 2 • Marketing Activities Brainstorming Sessions • Brand Assessments Oppositional Brand Loyalty • Discussion Forums • Brand Commitment • Media Galleries • Surveys Interaction Marketing Activities x Online H1 H4 • H3 H5 Psychological Sense of Community Supportive Brand Behaviors Willingness to Pay a Price Premium Online Brand Community Engagement • Brand Influence H2 • Defending the Brand • Brand Community Engagement Word-of-Mouth H6 • Rewards (Hedonic) • Rewards (Utilitarian) • Supportive Community Behaviors Community Participation • Self-Expression • Participation Intentions 95 Online Brand Community Engagement Online brand community engagement is the intrinsic motivation to interact with the community (Essay one; Algesheimer et al. 2005, p. 21). In essay one, I identified 11 motivations that people have to interact with online brand communities. Based on conversations with 18 community managers and executives at a major marketing research firm that specializes in MROCs, I focused on several specific engagement dimensions to test. In addition, I conducted total unduplicated reach and frequency (TURF) analysis on the 11 engagement dimensions from 620 MROC community members. TURF analysis can be used in marketing research to select the combination of motivations that give the broadest coverage for highly motivating community members. The TURF algorithm identifies the optimal bundle of motivations to maximize the number of highly motivated (those who selected the top two scale points for the dimension of online brand community engagement were considered highly motivated by that dimension) people in the community by sequentially identifying the motivation that gives the largest incremental reach. In each successive iteration of the TURF analysis, the members reached are removed from further analysis (as it not necessary to reach members twice). The point of diminishing returns can be seen when additional motivations provide minor incremental gains in reach (see Table 15). 96 Table 15 Essay Two: TURF Analysis of Online Brand Community Engagement Order Online Brand Community Engagement Dimensions Community Members Reached (Percent of Total) 1 Self-Expression 61.1 2 Brand Influence 69.8 3 Rewards (Utilitarian) 74.8 4 Helping 76.3 5 Rewards (Hedonic) 77.6 6 Validation 78.4 7 Like-minded Discussion 78.9 8 Seeking Assistance 79.4 9 Connecting 79.5 10 Brand Passion 79.7 11 Up-to-Date Information 79.7 Total Cumulative Reach 79.7 I selected 4 of the 11 motivations to test in this essay: brand influence, rewards (hedonic), rewards (utilitarian), and self-expression. These four motivations are highly relevant in the context of MROCs which are created and managed to help generate insights from customers for firms. Often these community members are recruited from client lists and compensated with financial or reward program incentives. Brand influence is the degree to which a community member wants to influence the brand (see Table 15). Rewards (hedonic) is the degree to which the community member wants to gain hedonic rewards (e.g., fun, enjoyment, entertainment, friendly environment, and social status) through their participation in the community. Rewards (utilitarian) is the degree to which the community member wants to gain functional rewards (e.g., 97 monetary rewards, time savings, deals or incentives, merchandise, and prizes) through their participation in the community. Self-expression is the degree to which a community member feels they can express their true interests and opinions in the community. As Muniz and O'Guinn (2001) and McAlexander et al. (2002) observed, interaction in online brand communities can be a context rich and meaningful experience that helps form relational bonds with other community members. Thus, the motivations will propel them to interact and that will help them develop a sense of community. H2a-d: Online brand community engagement (a. brand influence, b. rewards (hedonic), c. rewards (utilitarian), and d. self-expression) have a positive, direct effect on psychological sense of community. Leveraging Marketing Activities and Online Brand Community Engagement Matching marketing activities to the member’s online brand community engagement is important in achieving leveraging of the marketing activities to facilitate greater development of psychological sense of community (e.g., Fournier et al. 1998; Kumar et al. 2010). For a marketing activity to leverage online brand community engagement, it needs to address the reasons for which the member is motivated to interact with the online brand community. Inasmuch as the marketing activity matches or fits the motivations the member has for interacting with the community, then there is fit and leverage occurs. As defined previously, leverage is the amplifying of the effect of marketing activities on psychological sense of community with matching marketing activities that fit with the community member’s engagement dimensions. The extent to which a marketing activity does not fit the member’s 98 motivations for interacting with the community, then there is a lack of fit and leverage does not occur. Thus, fit should enhance the effectiveness of marketing efforts and enable marketers to enhance the development of psychological sense of community. In a brand community setting it is the combination of the individual motivations and marketing activities with the firm that enhances the psychological sense of community. Community member motivations lead them to reconceptualize themselves as part of the community through a self-transformation process (e.g., changing attire, self-concept or physically altering his/her appearance; Schouten 1991; Schouten and McAlexander 1995). Marketers have the opportunity to mold the “malleable perceptions of the new [members]... socialization brings about a transformation of the individual that entails an evolution of motives for involvement and a deepening of commitment to the [community] and its ethos” (Schouten and McAlexander 1995, p. 56). “By understanding the process of self-transformation undergone by individuals within a [community], a marketer can take an active role in socializing new members and cultivating the commitment of current ones. Harley-Davidson cultivates consumer commitment through means such as supplying a steady stream of information geared to the needs of newcomers and providing a full range of clothing, accessories, and services that function as involvement-enhancing side bets and exit barriers” (Schouten and McAlexander 1995, p.57). Additionally, marketing activities that resonate with the consumer help them to develop an even greater sense of community. For example, the “demands of the organization [are seen] as being morally necessary because of their relation to ultimate values” (Kanter 1972, p. 122). This level of psychological sense of community should create an enduring and exhibited lifestyle change. The marketing literature on the elaboration likelihood model supports the notion of leverage leading to superior and lasting change. A fundamental proposition of the elaboration likelihood model is that as personal relevance increases, the community member will be more motivated to 99 process the message (Petty et al. 1983, p. 137), thereby increasing the likelihood that the message will have an effect on attitudes. Thus, it is the combination of member motivations and marketing activities that can result in leveraging psychological sense of community. Brainstorming activities should fit with brand influence motivations to participate in the community creating a leveraging effect on psychological sense of community. Brainstorming activities are less structured and typically used to explore new ideas or concepts with members. Comments from brainstorming activities are used to generate actionable insights for brands and their products. Thus, brainstorming activities should fit with brand influence motivations. H3a: The interaction of brainstorming sessions and brand influence will have a positive effect on psychological sense of community. Discussion forums should fit with rewards (hedonic) motivations to participate in the community creating a leveraging effect on psychological sense of community. Discussion forums are more focused and typically focused on a single thread or comment and tend to be visible for long periods of time. Participating in discussion forums provide members opportunities to demonstrate their expertise and gain status within the community. Schouten and McAlexander (1995) found that “status is conferred on members according to their seniority, participation and leadership in group activities… expertise and experience, [and brand]-specific knowledge” (p. 49). In addition, community members are likely to enjoy discussing things with other like-minded community members. Such discussions should provide enjoyment and entertainment, which are powerful hedonic needs. 100 H3b: The interaction of discussions and rewards (hedonic) will have a positive effect on psychological sense of community. Surveys should fit with reward (utilitarian) motives to participate in online brand communities creating a leveraging effect on psychological sense of community. While not emphasized in brand community literature (e.g., Dholakia et al. 2004), qualitative interviews conducted in essay one found that promotions for brand community members is a motivator for individuals to participate in a brand community. These promotions can include special members only sales, coupons, and rewards points redeemable for merchandise. The omission of utilitarian type rewards in prior academic literature is indicative of the transformation the Internet has undergone in the past 10 to 15 years and the need to update the dated studies of member motivations. Initially the Internet was primarily an information storage and retrieval system, but the Internet has become a thriving commercial channel with advances in technology and shipping. Online commerce has grown tremendously during this time period and with the increase in online commerce, it has been a natural extension for firms to link their social presence to their commercial sites as part of an integrated marketing strategy. Surveys, which are less interactive than other community activities, tend to be incentivized with monetary rewards. Therefore, surveys should fit with reward (utilitarian) motivations to participate in the community. H3c: The interaction of surveys and rewards (utilitarian) will have a positive effect on psychological sense of community. 101 Media galleries should fit with self-expression motivations because media galleries allow members the opportunity to show themselves using the brand and its products. Community members are sharing much richer content with a picture or movie in a media gallery than perhaps they could express in words, particularly if the picture is of themselves with the brand or branded products. H3d: The interaction of media galleries and self-expression will have a positive effect on psychological sense of community. Loose Hierarchy of Effects from Psychological Sense of Community In a content analysis of online communities, Dröge et al. (2010) found that positive comments were twice as frequent as negative comments (p. 77, 80). Muniz and O'Guinn (2001) defined brand communities as a “structured set of social relationships among admirers of a brand” (p. 412, emphasis added). Therefore, the prevailing assessments of a brand among members of a brand community should be positively inclined toward the brand. This shared attitude toward the brand is the requisite condition described by Baron and Kerr (2003) for the polarization of attitudes in a group setting (p. 108). Furthermore, psychological sense of community activates and leads to the self- and community regulation of attitudes and behaviors (Carlson 2005; Hogg et al. 1995). Hogg et al. (1995) explains this phenomena in greater detail: “People have a repertoire of such discrete category memberships that vary in relative overall importance in the self-concept… Each of these memberships is represented in the individual member’s mind as a social identity that both describes and prescribes one’s attributes as a member of that group-that is, what one should think and feel, and how one should behave. Thus, when a specific social identity becomes the salient basis for selfregulation in a particular context, self-perception and conduct become in-group stereotypical and normative, perceptions of relevant out-group members become outgroup stereotypical, and intergroup behavior acquires competitive and discriminatory 102 properties to varying degrees depending on the nature of relations between the groups.” (p. 259-260) Therefore, through normative and informational means the greater the psychological sense of community an individual feels, the greater the positive polarization should occur for their brand assessments and brand related behaviors. The positive polarization of brand assessments, supportive brand behaviors, and supportive community behaviors through psychological sense of community is also consistent with social exchange theory. Social exchange theory suggests that important and valued relationship should lead to the maintenance of those relationships (Cook and Emerson 1978; Homans 1958; Homans 1974; Lovaglia 2007). While much of the earlier brand community research has been conducted in face-to-face brand communities (e.g., Harley Owners Group, Jeep Jamborees, and European car clubs), it remains an open question how much of an effect the psychological sense of community with an online brand community can have on brand and brand related outcomes. Therefore, I hypothesize and test a loose hierarchy of effects to see if psychological sense of community can influence not only brand assessments, but also supportive brand behaviors and supportive community behaviors. Brand assessments (i.e., brand commitment and oppositional brand loyalty) are important indicators for marketers to examine community member attitudes toward their brand. Brand commitment is the extent to which a community member has an affective commitment to remain in a relationship with the brand (Bansal et al. 2005). Formally, brand commitment “reflects an emotional attachment to, identification with, and involvement in an organization” (Meyer and Smith 2000, p. 320). Oppositional brand loyalty has been observed in brand communities as a different type of loyalty. Specifically, oppositional brand loyalty is the rejection and criticism of competing brands (Muniz and O'Guinn 2001, p. 420). Thompson and Sinha (2008) have 103 described oppositional as leading “members of the community to take an adversarial view of competing brands” (p. 65). It has been suggested that “oppositional loyalty may benefit companies by reducing the likelihood that members will purchase products from competing brands” (Thompson and Sinha 2008, p. 65), as well as can even lead people to form brand communities structured entirely around opposition to the competing brand (e.g., Team MacSuck), (Hollenbeck and Zinkhan 2010; Muniz and O'Guinn 2001, p. 421). H4a-b: Psychological sense of community will have a positive effect on brand assessments (a. brand commitment, and b. oppositional brand loyalty). Supportive brand behaviors are a higher level of effect for psychological sense of community to influence. One of the important aspects about online brand communities is that the members provide social context around the brand. Word-of-mouth is the “positive information communicated in social situations (e.g., in conversations with friends and acquaintances)” (Arnett et al. 2003, p. 96). In addition, supportive brand behaviors can also take the form of individual actions. The premise of difficult brand behaviors is that brand related behaviors can be stratified based on their “economic, social, psychological, temporal, or other physical resources” requirements (Park et al. 2010, p. 4). Difficult brand behaviors are “those [behaviors] that [require consumers to] use more of their own resources” (Park et al. 2010, p. 5). These behaviors are indicative of the consumer’s willingness to maintain the relationship with the brand (e.g., willingness to pay a price premium for the brand and willingness to defend the brand from criticism). 104 H5a-c: Psychological sense of community will have a positive effect on supportive brand behaviors (a. word-of-mouth, b. defending the brand, and c. willingness to pay a price premium). Supportive community behaviors are also a higher level of effect for psychological sense of community. One thing that online brand communities are notorious for is that many people simply observe and do not participate. Members that watch and participate little to none at all are called “lurkers” (Madupu and Cooley 2010). “Lurking” rates have been estimated as high as 90 percent and even 100 to 1 (Madupu and Cooley 2010, p. 130). Given the large percent of community members who “lurk”, can psychological sense of community affect participation intentions? Brand community participation intention is the member’s intended level of participation in the brand community (Algesheimer et al. 2005, p. 22). Additionally, can psychological sense of community lead to greater participation in the community? Muniz and O'Guinn (2001) discuss community members as feeling a moral responsibility to the community as a whole. This felt moral responsibility to the community helps ensure the survival of the community by not only driving members to patronize the brand instead of competitor brands, but also contribute to the community. The brand community literature is consistent with social exchange theory which suggests that important and valued relationship should lead to the maintenance of those relationships (Cook and Emerson 1978; Homans 1958; Homans 1974; Lovaglia 2007). H6a-b: Psychological sense of community will have an effect on supportive community behavior (a. participation intentions, and b. community participation). 105 METHODS To address the primary research question of being able to leverage online brand communities and MROCs (i.e., obtain fit between marketing activities and online brand community engagement), I created a unique dataset that combines longitudinal surveys from MROC members in eight different private online branded market research communities for brands in three different industries. In addition, I matched survey respondents to marketing activities sponsored by the firm and individual level participation data and demographics tracked by a marketing research firm who hosts the MROCs. The perceptual measures were collected in three surveys in each community across one month. Due to survey length restrictions required by the community managers, short-form versions of the scales were used in this study to fulfill survey length requirements. The secondary data was collected for 6 weeks on both ends of the survey timeframe creating a 16 week window of data collection. Merging the surveys and numerous secondary data files was accomplished using SAS version 9.1. The MROCs studied were for global brands from the following industries: business to business: firms like CDW, Dell, Novartis, and UCB; services and retail: firms like BestBuy, Caesars, FedEx, Home Depot, and Macy’s; and consumer packaged goods: firms like Colgate, ConAgra, Coca-Cola, Frito-Lay, Kraft, Unilever, and Welch’s. Perceptual and Secondary Measures Marketing Activities Operationalization of marketing activities can be done several different ways. First, marketing activities could be operationalized as the count of marketing activities created by community managers during a specific time period. Using the count of the activities at the community level 106 would not have enough variation to use in analysis. While it would capture the activities run by the community managers, it would not account for individuals noticing or participating in the activities. Second, marketing activities could be operationalized as the count of the individual contributions for each type of marketing activity during a given time period. While this operationalization would improve upon the former, it would fail to capture the overall experience or subjective evaluation of the marketing activities. Third, marketing activities could be operationalized as the perceived congruence of the marketing activities. Operationalizing marketing activities this way captures the members’ overall evaluations of the marketing activities they participated in. This provides the best operationalization of members being exposed to and participating in marketing activities in the community. Therefore, marketing activities is operationalized as congruence of the marketing activities in this study. There are two popular ways to measure the congruence of marketing activities in the literature: subjective and empirical. Subjective assessments of congruence are recorded from the respondents. Spreng et al. (1996) used an additive difference specification which is derived by asking respondents to compare what they wanted and what they got, then weighting the difference based on their evaluation of the difference. Congruence is then the sum of all the weighted differences. Alternatively, congruence can be empirically assessed by identifying respondents with the highest 5 to 10 percent of responses for the dependent variable, computing the mean levels of each predictor they received and then calculating difference scores for all respondents (Vorhies and Morgan 2003). Thus, congruence is then the sum of all the difference scores. In this study, I adapted the Spreng et al. (1996) measure of congruence to assess the congruence of marketing activities by asking respondents “How close was the to what 107 you hoped to do in this community?” This approach was chosen over the empirical approach because asking respondents directly to assess congruence captures their evaluation of the gap and valence of the difference. Due to survey length requirements with the partner firm, only single items were used for assessing the congruence of each marketing activity. Online Brand Community Engagement Online brand community engagement is the intrinsic motivation to interact with the brand community. As defined previously, brand influence is the degree to which a community member wants to influence the brand. Rewards (hedonic) is the degree to which the community member wants to gain hedonic rewards (e.g., fun, enjoyment, entertainment, friendly environment, and social status) through their participation in the community. Rewards (utilitarian) is the degree to which the community member wants to gain functional rewards (e.g., monetary rewards, time savings, deals or incentives, merchandise, and prizes) through their participation in the community. Self-expression is the degree to which a community member feels they can express their true interests and opinions. Psychological Sense of Community Psychological sense of community is “the degree to which an individual perceives relational bonds with other brand users” (Carlson 2005; Carlson et al. 2008, p. 286). Psychological sense of community is a more general measure of the sense of community than that provided by Muniz and O'Guinn (2001). In addition, Muniz and O'Guinn (2001) do not develop empirical measures for the markers of community they describe, whereas (Carlson 2005; Carlson et al. 2008) developed valid and reliable items to measure psychological sense of community. 108 Brand Assessments Brand commitment is the extent to which a community member has an affective commitment to remain in a relationship with the brand (Bansal et al. 2005). Formally, brand commitment “reflects an emotional attachment to, identification with, and involvement in an organization” (Meyer and Smith 2000, p. 320). I used items developed by Morgan and Hunt (1994) and also validated in an online brand community research setting by Carlson et al. (2008). Oppositional brand loyalty has been observed in brand communities as a different type of loyalty. Specifically, oppositional brand loyalty is the rejection and criticism of competing brands (Muniz and O'Guinn 2001, p. 420). Thompson and Sinha (2008) have described oppositional as leading “members of the community to take an adversarial view of competing brands” (p. 65). I adapted existing items to measure oppositional brand loyalty (Madupu 2006, p. 88). Supportive Brand Behaviors I adapted existing measures to capture word-of-mouth (Arnett et al. 2003, p. 103). Two dimensions of difficult brand behaviors were collected (Park et al. 2010). I collected the willingness to pay a price premium dimension and the defending the brand dimension of difficult brand behaviors. These behaviors are conceptualized as requiring a greater “economic, social, psychological, temporal, or physical resources” for consumers to engage in (Park et al. 2010, p. 4). Supportive Community Behaviors In order to assess the effects of psychological sense of community on a community member’s supportive community behaviors, I measured both intentions to participate in the community, and 109 actual community participation for two months following the second survey. Community participation intentions were measured using a scale developed by Algesheimer et al. (2005). Community participation was calculated as the sum of count of contributions the community member made for two months following the administration of survey 2 based on secondary data. Other Secondary Data Additional data was collected on respondents from secondary sources. These included their age, gender, education, household income, and tenure with the community. This data helps provide a better picture of the MROC members. Analysis and Results In total, the sampling frame for primary and secondary data collection was 7,258 community members during the study timeframe. The number of community members in each community at the time the first survey launched ranged from 295 to 1,318, with an average size of 495 members per community. In total, 1,127 individuals completed at least one survey during the study time period giving an overall participation rate of approximately 16 percent (see Table 16). 54 percent of respondents were female, median age of respondents was 30-44 years old with a median household income of $50,000 to $79,000 per year. The median education for respondents is college graduate, and the average tenure in the community is 136 weeks. The average number of days between the time respondents took survey 1 and 2 was 7.2 days, and the average number of days between the time they took survey 2 and 3 was 12.4 days. Respondents’ experience with the brand (Mishra et al. 1993) was also assessed using a 4 item 11-point semantic differential scale (Mishra et al. 1993, p. 334). Responses ranged from 1 (e.g., novice 110 buyer of the brand) to 11 (e.g., expert buyer of this brand), with a mean of 9 and a standard deviation of 1.8. Comparison of mean levels of items used in the measurement model across respondents who completed different sets of the survey waves (e.g., 1 v. 1, 2, and 3) revealed trivial differences in mean levels. Table 16 Essay Two: Summary of Surveys Completed Survey(s) Total Respondents 193 206 129 127 80 125 267 1,127 1 only 2 only 3 only 1 and 2 only 1 and 3 only 2 and 3 only 1, 2, and 3 Total Measurement Model Analysis for this study was conducted in EQS version 6.1 using maximum likelihood (ML) estimation. Of the 267 respondents who completed all three rounds, missing contribution data led to dropping of 11 cases, leaving a usable sample of 256 community members. Missing values were mean imputed. Common method variance for self-reported data was tested for and found not to be a problem using Harmon’s single factor test (χ2 = 5,575, df = 464, p < .01, CFI = .40, IFI = .40, RMSEA = .21). I began the analysis by conducting a CFA to evaluate the convergent and discriminant validity of the constructs in my model. Following Bagozzi and Yi (1988) I looked for negative error variance or “Heywood cases”, correlations greater than 1.00, and extremely large parameter estimates. None of the aforementioned were found in the CFA. The next step was to assess the data for multivariate normality using Maridia’s coefficient. 111 Analysis indicated that there are some problems with normality (Mardia’s coefficient normalized estimate = 54.43, but the model had acceptable fit statistics (χ2 = 656, df = 382, p < .01, CFI = .97, IFI = .97, RMSEA = .05 and AIC = -108) and are therefore used in the analysis of this data (see Table 17). Table 17 Essay Two: Measurement Model Summary Range Mean SD λ 1-11 8.32 1.76 1.00 1-11 9.13 1.93 .99 1-11 9.14 1.94 .94 1-11 9.14 1.92 .99 1-11 9.79 1.44 .94 1-11 9.55 1.61 .82 1-11 9.91 1.41 .92 0-145 23.84 18.22 1.00 1-11 7.8 2.14 .95 1-11 7.37 2.45 .92 a Brainstorming Sessions 1. How close were brainstorming sessions to what you hoped to do in this community? Brand Commitment 1. I am committed to maintaining my relationship with this brand 2. I intend to maintain my relationship with this brand indefinitely 3. I am committed to maintaining my relationship with this brand Brand Influence 1. I am motivated to participate in this brand community because I can help improve the brand and its products 2. Increasing the influence I have on the brand and its products makes me want to participate more in this brand community 3. I hope to improve the brand or product through my participation and expression in this brand community Community Participation The sum of the count of contributions from the end of survey 2 until 6 weeks following the end of survey 3 (covering a period 2 months). Defending the Brand 1. If someone spoke negatively about this brand, I would confront them 2. I would argue with anyone who spoke badly about this brand 112 Table 17 (cont’d) Range Mean SD λ 4-11 8.32 1.77 1.00 1-11 7.91 1.64 1.00 1-11 1-11 5.26 4.99 2.68 2.87 .86 .88 1-11 5.02 2.73 .83 6-11 10.02 1.33 1.00 1-11 7.31 2.48 .93 1-11 7.54 2.45 .92 1-11 8.07 2.26 .94 1-11 7.62 2.35 .94 1-11 8.2 2.12 .89 2-11 9.14 1.69 .95 1-11 8.98 1.78 .97 1-11 8.54 2.64 1.00 5-11 10.05 1.32 .88 5-11 9.88 1.45 .88 1-11 9.95 1.44 .85 a Discussion Forums 1. How close were discussions to what you hoped to do in this community? a Media Galleries 1. How close were media galleries to what you hoped to do in this community? Oppositional Brand Loyalty 1. I have a negative attitude towards competing brand(s) 2. I would never buy competing brand(s) 3. On discussion forums, when the topic about competing brand(s) comes up, I don't say anything good about them Participation intentions 1. I intend to actively participate in this brand community’s activities Psychological Sense of Community 1. I feel strong ties to other members of this brand community 2. I find it very easy to form a bond with other members of this brand community 3. I feel a sense of being connected to other members of this brand community 4. A strong feeling of camaraderie exists between me and other people who visit this brand community 5. I feel a sense of community with other people who visit this brand community Rewards (Hedonic) 1. I like participating in this brand community because it is entertaining 2. I find participating in this brand community to be very entertaining Rewards (Utilitarian) 1. Receiving more money makes me want to participate more in this brand community Self-Expression 1. I feel that I can freely share my interests in the brand community 2. I can always be myself when interacting with others in this community 3. This community makes it easy for me to express my true beliefs about the brand 113 Table 17 (cont’d) Range Mean SD λ a Surveys 1. How close were surveys to what you hoped to do in 3-11 8.17 1.84 1.00 this community? Willingness to Pay a Price Premium 1. I will continue to buy this brand even if its prices 1-11 6.35 2.83 .91 increase somewhat 2. I will pay a higher price than competitors charge for 1-11 6.39 2.85 .93 the benefits that I currently receive with this brand Word-of-Mouth 1. I “talk up” this brand to people I know 1-11 8.16 2.31 .94 2. I bring up this brand in a positive way in conversations 1-11 8.29 2.29 .94 I have with friends and acquaintances 3. In social situations, I often speak favorably about this 1-11 8.3 2.22 .94 brand Note: All scales measured on a 0 - 10 Likert-type Scale with Anchors 0 = Strongly Disagree and a 10 = Strongly Agree unless otherwise indicated. 0 = Much Worse Than I Hoped To Do and 10 = Much Better Than I Hoped To Do. Prior to analysis all values were recoded to a 1 – 11 range, which is presented in all results tables. Convergent and Discriminant Validity Following adequate overall fit, the individual items were then analyzed to assess convergent validity of the measurement model. Each of the items loaded on their respective construct at greater than .70. Each of the scales shows desirable composite reliability with average variance extracted (AVE) all greater than .50 (Bagozzi and Yi 1988). In addition, each scale shows adequate latent construct reliability with each greater than .70. Discriminant validity of the scales with more than one item was assessed and confirmed using the procedure described by Fornell and Larcker (1981) where the AVE of each construct was compared to the squared multiple correlation between the latent constructs (see Table 18). 114 Table 18 Essay Two: Correlation Matrix 1 1. Psychological Sense of Community 2. Brand Commitment 3. Oppositional Brand Loyalty 4. Participation Intentions 5. Word-of-Mouth 6. Community Participation 7. Willingness to Pay a Price Premium 8. Defending the Brand 9. Brand Influence 10. Rewards (Hedonic) 11. Rewards (Utilitarian) 12. Self-Expression 13. Surveys 14. Discussion Forums 15. Brainstorming Sessions 16. Media Galleries Average Variance Extracted Construct Reliability 2 3 4 5 .44 .43 .28 .35 .53 .25 6 7 8 9 .49 .70 .25 .08 .46 .09 .35 .27 .25 .55 .58 .52 .33 .74 .26 .52 .42 .56 .41 .52 .49 .71 .11 .25 .15 .39 .47 .64 .52 .45 -.07 -.02 .43 .54 .34 .30 .38 .24 .41 .20 .07 .15 .18 .18 .22 .07 .52 .26 .27 .14 .39 .24 .17 .85 .94 .92 .93 10 11 12 13 14 .07 .19 .28 .59 .47 .45 .24 .24 .61 -.05 .49 .28 .32 .24 .08 .28 .12 .29 .22 -.13 .43 .27 .33 .30 -.06 .24 .15 .23 .28 .10 .23 .22 .24 .73 1.00 .88 1.00 .81 1.00 .88 1.00 .01 .68 .27 .25 .17 -.01 .63 .36 .33 .24 .04 -.01 -.05 .03 .29 .29 .18 .68 .54 .72 .20 .14 .25 -.06 .21 .41 .45 .84 .87 .80 .92 1.00 .76 1.00 1.00 1.00 1.00 .81 .83 .84 .87 1.00 .82 1.00 1.00 1.00 1.00 Boldface correlations are non-significant at α = .05 115 15 16 .57 Structural Equation Models After establishing the validity of the measurement model, I proceeded to estimate the structural equation models. In addition to the hypothesized effects, several additional paths were estimated in both models among the endogenous factors to account for the numerous relationships among the similar but distinct endogenous factors comprising the loose hierarchy of effects tested in this model. The paths added are consistent with consumer behavior literature on cognitions leading to behavior (Eagly and Chaiken 1993). Since these paths are not of primary concern in this essay, they have been omitted from formal hypothesis testing. Following the Ping (1995) approach to estimating latent variable interactions, I estimated two models, Model 1 was estimated without the interaction effects and Model 2 was estimated with the interaction effects (see Table 19). Following Bagozzi and Yi (1988) and Hu and Bentler (1999) criteria for model fit of a CFI close to .95, IFI greater than .90, and RMSEA close to .06, Model 1 demonstrates acceptable fit (χ2 = 926, df = 462, CFI = .95, IFI = .95, and RMSEA = .06). Furthermore, the chi-square for Model 1 demonstrates a significant improvement over the independence model (Δχ2 = 9,022, df = 535, p < .01). Model 2 also demonstrates acceptable fit (χ2 = 1,330, df = 596, CFI = .95, IFI = .95, and RMSEA = .07) and the chi-square for Model 2 demonstrates a significant improvement over the independence model (Δχ2 = 13,985, df = 677, p < .01). Table 19 Essay Two: Structural Equation Models Model Parameter Psychological Sense of Community Marketing Activities Brainstorming Sessions Model 1 H1a 116 .23* (.10) Model 2 Interactions .24* (.10) Table 19 (cont’d) Model Parameter Discussion Forums H1b Media Galleries H1c Surveys H1d Online Brand Community Engagement Brand Influence H2a Rewards (Hedonic) H2b Rewards (Utilitarian) H2c Self-Expression H2d Interaction Effects Brainstorming Sessions * Brand H3a Influence Discussion Forums * Rewards H3b (Hedonic) Surveys * Rewards (Utilitarian) H3c Media Galleries * Self-Expression H3d Brand Assessments DV: Brand Commitment Psychological Sense of Community H4a DV: Oppositional Brand Loyalty Psychological Sense of Community H4b Word-of-Mouth Supportive Brand Behaviors DV: Word-of-Mouth Psychological Sense of Community H5a Brand Commitment Participation Intentions Willingness to Pay a Price Premium DV: Defending the Brand Psychological Sense of Community H5b Brand Commitment Word-of-Mouth DV: Willingness to Pay a Price Premium Psychological Sense of Community H5c Supportive Community Behaviors DV: Participation Intentions Psychological Sense of Community H6a DV: Community Participation Psychological Sense of Community H6b Word-of-Mouth Standard errors in parentheses; * significant at α = .05 117 Model 1 -.01* (.10) .16* (.08) -.01* (.08) .09* .39* -.07* .10* Model 2 Interactions -.02* (.10) .14* (.08) -.01* (.08) (.13) (.10) (.04) (.16) .09* .38* -.06* .10* (.13) (.10) (.04) (.16) --* -- -.03* (.06) --* -- .10* (.04) --* --* --- .02* -.12* (.02) (.02) .43* (.05) .44* (.05) .24* .34* (.08) (.08) .24* .34* (.07) (.08) .11* .40* -.05* .46* (.05) (.06) (.08) (.07) .12* .40* -.05* .46* (.05) (.06) (.08) (.07) .26* -.06* .54* (.07) (.09) (.09) .26* -.06* .54* (.07) (.09) (.09) .54** (.05) .54* (.05) .34* (.04) .34* (.04) .17* .16* (.60) (.61) .17* .16* (.60) (.61) Hypothesis Testing Results Both models provide partial support for H1, H2, and full support for H4, H5, and H6. H1 proposed that marketing activities positively affect psychological sense of community. This analysis revealed that brainstorming sessions and surveys significantly affected psychological sense of community. H2 proposed that online brand community engagement positively affects psychological sense of community. This analysis revealed that in the context of all the exogenous factors used to predict psychological sense of community, only rewards (hedonic) had a significant effect on psychological sense of community. H4, H5, and H6, proposed that psychological sense of community could lead to a loose hierarchy of effects for brand assessments, supportive brand behaviors, and supportive community behaviors. Indeed, there is strong support for each of the proposed relationships. Psychological sense of community is an important driver for a range of brand assessments, supportive brand behaviors, and supportive community behaviors. The chi-square difference tests between the Model 2 and Model 1 reveals that Model 2 does not overall provide a statistically significantly better fit to the data than Model 1 (Δχ2 = 404, df = 134, p < .01) because only 1 in 4 of the interactions is significant. Thus there is limited support for H3. However, regions of significance analysis and a significant interaction effect (media galleries and self-expression) provides important insights into leveraging the effects of online brand community engagement and marketing activities (Aiken et al. 1991; Curran et al. 2004). Post hoc analysis of interaction effects using regions of significance testing (Aiken et al. 1991; Curran et al. 2004) provided some additional insight into the leveraging effect of fitting marketing activities and online brand community engagement to influence psychological sense 118 of community. Regions of significance show at what level of the moderator the main effect becomes non-significant. Regions of significance testing was conducted for each of the interaction effects (see Figure 3 Panels A through D). Figure 3 Essay Two: Regions of Significance Testing 1 6.95 Brand Influence B: Discussion Forums x Rewards (Hedonic) Simple Slopes of Discussion Forums -.6 0 1 1 p < .05 0 p > .05 -.8 Simple Slopes of Brainstorming A: Brainstorming x Brand Influence 1 11 C: Surveys x Rewards (Utilitarian) 1 p < .05 7.24 Rewards (Hedonic) 11 D: Media Galleries x Self-Expression p < .05 Simple Slopes of Media Galleries 0 .5 9 p < .05 8 Simple Slopes of Surveys 0 .4 p > .05 p > .05 2.72 Rewards (Utilitarian) 11 1 11 Self-Expression 119 Hypothesis 3a proposes a positive interaction between brainstorming activities and brand influence. Figure 3 Panel A shows that when brand influence is below 6.95 on a 1-11 scale, the effect of brainstorming on psychological sense of community becomes non-significant. This means that while there is not a significant interaction of brainstorming sessions and brand influence, at low levels of brand influence the effect of brainstorming becomes non-significant. So, community members who have neutral to low levels of motivation to interact with the community to influence the brand will should not develop a sense of community through brainstorming activities. Hypothesis 3b proposes a positive interaction between discussion forums and rewards (hedonic). Figure 3 Panel B shows that when rewards (hedonic) is below 7.24 on a 1-11 scale, the effect of discussion forums on psychological sense of community becomes non-significant. This means that while there is not a significant interaction effect of discussion forums and rewards (hedonic) and, at moderate to low levels of rewards (hedonic) the effect of discussion forums becomes non-significant. Community members who have moderate to high levels of rewards (hedonic) motivations will develop a psychological sense of community through discussion forums. Hypothesis 3c proposes a positive interaction between surveys and rewards (utilitarian). Figure 3 Panel C shows that when rewards (utilitarian) is very low, below 2.72 on a 1-11 scale, the effect of surveys on psychological sense of community becomes non-significant. In comparing Figure 3 Panel B and Figure 3 Panel C, one can see that there is a wider zone of tolerance for surveys than discussion forums, suggesting that perhaps community members have more defined expectations of discussion forums than surveys. 120 Hypothesis 3d proposes a positive interaction between media galleries and selfexpression. Surprisingly, Figure 3 Panel D shows that there is a significant negative interaction between media galleries and self-expression throughout the range of self-expression. As selfexpression increases, the effect of media galleries on psychological sense of community is diminished. Therefore, community members highly motivated to express themselves will have develop less of a psychological sense of community through media galleries than community members with very low levels of self-expression. Instead of the hypothesized leveraging effect of media galleries and self-expression, I found a dampening effect. This negative interaction suggests that media galleries and self-expression could be alternative drivers of psychological sense of community. Several interesting findings include the number of non-significant paths from marketing activities and engagement to psychological sense of community (H1b, H1d, H2a, H2c, and H1d), as well as the interactions of marketing activities with engagement (H3a, H3b, and H3c). Further examination of these paths shows small effect sizes, but large standard errors which could be the result of significant heterogeneity within the sample. Thus, these paths may be significant if a larger sample size was collected which would reduce sample variation. As discussed and observed in essay one, heterogeneity within online brand communities is expected. However, due to sample size limitations, a tradeoff had to be made between testing a more complex model with a loose hierarchy of effects and a simpler model without the loose hierarchy of effects. DISCUSSION Strategic marketing assets are “customer focused measures of the value of the firm (and its offerings) that may enhance the firm’s long-term value” (Rust et al. 2004, p. 78). Strong 121 relationships with customers can provide a source of competitive advantage and revenue generation. The focus of this essay has been to evaluate whether social relationships in an online environment (i.e., marketing research online community) should be considered a strategic marketing asset. By exploring the importance of leveraging marketing activities by the firm and an individual’s online brand community engagement, this research found that marketing research online communities (MROCs) should be considered strategic marketing assets (i.e., they are valuable, rare, and difficult to imitate). This essay contributes to the marketing strategy literature and brand community literature by building and testing a model of how marketers can leverage their online brand communities. As brand communities have evolved, MROCs have been developed to foster closer relationships with customers. Marketers are sponsoring marketing activities within these online communities and actively working to develop a sense of community for the members. In addition, community members have their own intrinsic drives to participate in the communities. I found that the social dynamics in these communities do have significant effects on brand and community outcomes. This research highlights that MROCs are not currently being utilized to their full potential. MROCs are valuable to firms because they can help marketers develop insights (like focus groups) and because they can influence brand and community outcomes. Similar to the findings from the brand community literature, I found that MROCs can be used to enhance brand assessments (i.e., oppositional brand loyalty and brand commitment). Interaction with MROCs strengthens the relationships community members have with the brand and helps create a competitive advantage for the brand. In addition, MROCs also increase supportive brand behaviors (i.e., word-of-moth, defending the brand, and willingness to pay a price premium). It is important for marketers to recognize the largely untapped potential of utilizing MROCs to 122 increase financial returns above and beyond insights into their customers wants and needs. Lastly, similar to the brand community literature, the more that managers can get MROC members to feel part of a community, the more supportive community behaviors the members will engage in. Marketing activities play an important role in developing this sense of community in conjunction with the intrinsic motivations of community members. Importantly, this sense of community can be developed in an online environment without face-to-face interaction. MROCs are a rare and difficult to imitate strategic marketing asset because they are branded and possess unique social dynamics in the community. The community structure, composition, and activities create the unique of the community. Communities using badges and other differentiators between members are structured hierarchically whereas communities without badges and other differentiators between members tend to be more egalitarian in structure. Thus, using rewards (hedonic) such as badges and other differentiators can create an entirely different community than one that uses rewards (hedonic) that are focused more on pleasure and enjoyment. Additionally, the composition of members also contributes to the uniqueness of the community. The heterogeneity observed in the results (i.e., large standard errors for engagement) suggest that different motivational types of community members could be present within an MROC. Lastly, MROCs are used by marketers to field marketing activities. The type and content of the activities contributes to the uniqueness of the community. As the results demonstrate, marketing activities are important contributors to the development of a sense of community. Together, these three factors illustrate how MROCs are an important link in creating relationships with customers in a unique way. Leverage is the amplifying of the effect of marketing activities on psychological sense of 123 community with matching marketing activities that fit with the community member’s engagement dimensions. While there is strong conceptual support for leveraging to occur under the elaboration likelihood model and the social psychology literature, I did not find empirical support for leveraging in MROCs. Post hoc analysis revealed that the fit between marketing activities and engagement tends to only be important at moderate to low levels of engagement. At moderate levels of brand influence and rewards (hedonic) and at very low levels of rewards (utilitarian) the effect of brainstorming, discussion forums, and surveys (respectively) on psychological sense of community is suppressed. Surprisingly, the effect of media galleries on psychological sense of community is dampened as self-expression increases. One of the most surprising findings in this essay is the significant negative interaction of media galleries and self-expression on psychological sense of community. As hypothesized, matching marketing activities with online brand community engagement should amplify the effect of marketing activities on psychological sense of community. However, the data has shown the opposite is true for media galleries and self-expression. This dampening of the effect of media galleries on psychological sense of community as self-expression increases suggests that media galleries are especially effective at developing a psychological sense of community for community members who have moderate to low levels of self-expression. Conversely, media galleries are less effective at developing psychological sense of community for community members who have moderate to high levels of self-expression. While media galleries and selfexpression may be redundant drivers of psychological sense of community, there is perhaps important implications for the use of the elaboration likelihood model in social settings such as online brand communities and MROCs. The negative interaction, or dampening effect, of media galleries and self-expression is a 124 surprising finding because it is counter to the leveraging effect expected by the elaboration likelihood model. The elaboration likelihood model was developed in the advertising literature at a time when advertising was primarily unidirectional (i.e., advertiser to consumer) which is no longer the case in online communities where social interaction and dialogue are the norm (e.g., firm to member and member to member interactions). Therefore the elaboration likelihood model may be underspecified to capture the complexity of how community members process social interactions. One potential explanation for this specific counterintuitive finding is that a picture is not worth a thousand words when it comes to feeling part of a community. In other words, sharing and viewing pictures are more effective at developing a sense of community for members with low levels of self-expression than high levels of self-expression. For members with high levels of self-expression they need to say more than what they can express in a picture to feel that they are interacting enough to develop a sense of community. High self-expression community members participating in a media gallery should tend to include more text captions and additional details with their photographs than low self-expression community members. Managerial Implications Marketing research online communities (MROCs) are becoming an increasingly prevalent tool for marketers to try to create insights into their customers. This research suggests that MROCs are currently not being utilized to their full potential. Specifically, I found that MROCs can be used to enhance what community members think about the brand, behaviors they engage in to support the brand, and their participation in the community. Therefore, MROCs are not simply the next generation of focus groups, but they are in fact an important strategic marketing asset 125 with social dynamics that can be beneficial for the brand. Of the four marketing activities that I studied, I found that brainstorming and media galleries are effective activities for marketers to sponsor in their MROC to develop a sense of community. There are numerous more marketing activities that marketers can perform in online communities and undoubtedly, as technology continues to evolve, many more types of marketing activities will become available to community managers. As additional marketing activities become available, this research shows that it is important to consider how the marketing activities contribute or diminish the sense of community for participants. Marketing activities that diminish the sense of community for participants should be avoided because it will adversely affect the brand and the participants participation in the MROC. When utilizing MROCs as a strategic marketing asset, it is important to consider the interplay between marketing activities and online brand community engagement. Increasing community member motivations is frequently the subject of checklists and blog posts. However popular talking about increasing motivations may be, it may be more difficult to change a community member’s intrinsic motivations than to work more efficiently with the motivations the community member already has. Taking a more refined view of online brand community engagement than is currently recognized in the popular press and leveraging marketing activities that fit community member online brand community engagement should provide increased gains (in terms of brand assessments, supportive brand behaviors, and supportive community behaviors) from MROCs. While it may be a bit idealistic to expect single activities in MROCs to lead to lift in brand and community outcomes, the cumulative effect of these marketing activities does have an important effect on the member’s sense of community. My analysis also suggests that a member’s level of engagement is important to consider 126 when making recruiting and turnover decisions. Figure 3 illustrates that when community members have below certain levels of engagement (e.g., 6.95 on a 1-11 scale of brand influence), marketing activities (e.g., brainstorming) should not have an effect on the development of psychological sense of community. Therefore, when making recruitment and turnover decisions in MROCs, managers can utilize these baseline levels of engagement to select members to join or retain as part of the community. Members that fall below these baseline levels should probably not be retained in the community unless there are specific reasons for retaining them even if they are not likely to be active participants in the community or supporters of the brand. Limitations and Future Research Model complexity and parsimony are two competing factors in any research program. The more refined view of engagement and the tremendous variety of activity types make it prohibitive to test a model with all possible interaction effects of marketing activities and engagement. This essay has taken a first step at exploring the potential leveraging effects of a reduced set of marketing activities and online brand community engagement. Model complexity and having sufficient sample size to estimate effects was an important constraint of this study. I ran an additional model with partial mediation of all marketing activities and online brand community engagement dimensions. While this model had overall better fit to the data, it still supported the importance of psychological sense of community as a mediator. However, the number of significant paths from the exogenous variable to brand assessments, supportive brand behavior and supportive community behavior indicates that there may be additional mechanisms through which marketing activities and engagement affect brand and community outcomes. 127 A limitation of the data used in this study is that media galleries and brainstorming had some missing data. Because respondents who completed all three rounds were needed to estimate the overall model, dropping the respondents with missing data on brainstorming and media galleries would have made the data set too small to estimate the model proposed. Therefore, mean imputation was used for brainstorming and media galleries to replace missing values. While all missing value imputation techniques have limitations, mean imputation’s primary limitation is that it reduces the variation in the variable, making it more difficult to find significant effects. Future research should search for additional marketing activities not covered in the academic literature or popular press. Additionally, finding marketing activities directly based on online brand community engagement dimensions could aid in testing leveraging effects of fit between marketing activities and engagement on psychological sense of community. The MSI 2012-2014 research priorities report calls for additional study of emerging tools for generating consumer insights (MSI 2012). As MROCs continue to become more mainstream marketing research tools and components of integrated marketing plans, additional research should be conducted on them as strategic marketing assets. Lastly, future research could also explore the public policy and ethical issues of studying consumers in online communities. What kind of policies are marketers likely to encounter when utilizing MROCs? What ways can and should marketers collect information about community members? 128 Executive Summary and Learning Implications Andersen (2005) states, “the measurement of brand community effects on market performance is a critical issue” (p. 48). Similarly, Yu-Chen (2006) states that “empirical studies of how virtual communities influence business are relatively rare” (p. 400). In total, 916 online brand community members from numerous online brand communities and MROCs were studied across the 6 studies in essay one and essay two. Based on this large-scale research project, there are four key takeaways for marketers. First, marketers need to take a much more refined view of motivation when trying to study and manage online brand communities and MROCs. Traditionally, engagement, or the intrinsic motivation to participate in a brand community has been measured as a single generic motivation (Algesheimer et al. 2005). However, interviews and surveys of 660 online brand community members indicates that consumers have a much broader set of motivations to interact with an online brand community. Specifically, consumers are motivated to interact with online brand communities for 11 reasons: brand influence, brand passion, connecting, helping, likeminded discussion, rewards (hedonic), rewards (utilitarian), seeking assistance, self-expression, up-to-date information, and validation (see Table 4 for definitions of each motivation). Three separate studies I conducted confirm the distinct nature of each of these dimensions and led to the development of three to four key questions to reliably and validly measure each dimension. I was able to identify two types of online brand community members, “brand passionate helpers” and “individualistic information seekers,” using a special type of segmentation analysis (latent class regression) on the 11 dimensions of engagement (see Table 7 for specifics on each class). Important differences in motivations exist between the two different types of community members. For example, coupons, deals, and discounts motivates (i.e., rewards utilitarian) 129 motivate “brand passionate helpers” to actively participate in the community, but instead motivate “individualistic information seekers” to watch or “lurk” rather than contribute to the community. Furthermore, subsequent analysis revealed that “brand passionate helpers” were more likely to purchase the brand and try its new products. Therefore, the more refined view of engagement provides important suggestions for marketers in terms of recruiting and managing community members. If the objective of the community is to have active participation, then “brand passionate helpers” should be recruited. However, if the community is simply intended to be a promotional platform, then “individualistic information seekers” should be recruited. Additionally, when conducting marketing activities within the community, managers should be aware of the type of community members in order to get the right activities to the right members to achieve the desired strategic objective. Lastly, the more refined view of engagement provides a basis for management of community member turnover. Community members who have a motivational profile that does not match the strategic objectives of the community (e.g., active participation v. promotional platform) can be dropped from the community at scheduled “refreshes” or mass turnover. Second, the social dynamics of online brand communities and MROCs do influence brand and community outcomes. While the experiences of face-to-face and online brand communities are very different (McAlexander et al. 2002), technological advances are helping to create rich interactive experiences for community members in an online setting closing the gap between face-to-face and online communities. Thus, similar to their face-to-face counterparts, these online communities do have social dynamics that influence member behaviors. Therefore, these online communities should be considered much more than just an online focus group where the community members talk while marketers just listen. Specifically, I found that the sense of 130 community that community members develop through the marketing activities and their intrinsic motivations leads to significant lift in brand assessments (i.e., brand commitment and oppositional brand loyalty), supportive brand behaviors (i.e., word-of-mouth, defending the brand, and willingness to pay a price premium) and supportive community behaviors (i.e., participation intentions and community participation). Third, I found that community member response to marketing activities depends on their motivations. In essay one, I found that active participation and passive participation (“lurking”) depends on the type of community member (“brand passionate helper” or “individualistic information seeker”). In essay two, I build on these findings to explore how marketing activities and engagement interact to drive brand and community outcomes. Specifically, I found that fit between marketing activities and engagement is important to consider when trying to achieve specific objectives (e.g., lift in brand assessments, supportive brand behaviors, and supportive community behaviors). Very low levels of engagement negate the effect of marketing activities on psychological sense of community (see Figure 3 for graphical depiction) for three of the four pairs of marketing activities and engagement dimensions. However, I found that media galleries and self-expression have a significant negative interaction. Media galleries and self-expression could be redundant drivers to developing a sense of community. Therefore, media galleries will be less effective at helping highly self-expressive community members develop a sense of community than community members with low levels of self-expression. Lastly, while online brand communities and MROCs are very similar, there are a few pertinent differences to consider when selecting one for implementing as a strategic marketing asset. First, online brand communities are typically open to the public whereas MROCs are 131 typically private communities requiring registration to interact with the community. Depending on the type of product and the desired audience, public versus private community may be an important factor in recruiting community members. Second, online brand communities are generally open to anyone, and therefore tend to have much higher “lurker” rates than MROCs do. For example, “lurking” rates for online brand communities have been estimated as high as 90 percent and 100 to 1 (Madupu and Cooley 2010, p. 130). MROC community managers monitor participation closely and can weed out members who don’t participate, so MROCs tend to have much lower “lurker” rates. Third, MROCs tend to be more focused than online brand communities. Specifically, MROC community managers typically field a specific number of marketing activities per week to generate insights about specific ideas or concepts. Furthermore, MROC community members are typically compensated with various financial incentives for completing marketing activities. Thus, financial constraints also dictate the level of marketing activities fielded in an MROC. Lastly, MROCs are either branded or unbranded (e.g., focused on a product type but not a specific brand). Branded MROCs were studied in this dissertation to aid in comparing to online brand communities. However, unbranded or product category focused communities can help marketers gain a broader understanding of consumers. In conclusion, online brand communities and MROCs offer marketers powerful tools to generate consumer insights and to directly influence firm performance through brand and community outcomes. Online brand communities and MROCs can be more fully utilized as strategic marketing assets as marketers use a more refined view of the motivations of community members and the mechanisms through which brand and community outcomes can be enhanced. Future research into and management of online brand communities and MROCs should build upon the refined conceptualization of engagement put forth in this dissertation. 132 REFERENCES 133 REFERENCES Aaker, Jennifer L. (1997), “Dimensions of Brand Personality,” Journal of Marketing Research, 34 (3), 347-56. Aiken, Leona S., Stephen G. West, and Raymond R. Reno (1991), Multiple Regression: Testing and Interpreting Interactions. Newbury Park, CA: Sage Publications. Algesheimer, René and Paul M. Dholakia (2006), “Do Customer Communities Pay Off?,” Harvard Business Review, 84 (11), 26-30. ----, Utpal M. Dholakia, and Andreas Herrmann (2005), “The Social Influence of Brand Community: Evidence from European Car Clubs,” Journal of Marketing, 69 (3), 19-34. Andersen, Poul Houman (2005), “Relationship Marketing and Brand Involvement of Professionals Through Web-enhanced Brand Communities: The Case of Coloplast,” Industrial Marketing Management, 34 (1), 39-51. Arnett, Dennis B., Steve D. German, and D. Hunt Shelby (2003), “The Identity Salience Model of Relationship Marketing Success: The Case of Nonprofit Marketing,” Journal of Marketing, 67 (2), 89-105. Austin, Manila and Julie Wittes Schlack (2012), “From Research to Relationship,” Marketing News, 46 (8), 82-82. Bagozzi, Richard P., Massimo Bergami, Gian Luca Marzocchi, and Gabriele Morandin (2011), “Customer–organization Relationships: Development and Test of a Theory of Extended Identities,” Journal of Applied Psychology (0021-9010). ---- and Utpal M. Dholakia (2002), “Intentional Social Action in Virtual Communities,” Journal of Interactive Marketing, 16 (2), 2-21. ---- and Youjae Yi (1988), “On the evaluation of structural equation models,” Journal of the Academy of Marketing Science, 16 (1), 74-94. ---- (2011), “Specification, evaluation, and interpretation of structural equation models,” Journal of the Academy of Marketing Science, 1-27. Bailey, Kenneth D. (1994), Typologies and Taxonomies: A Itroduction to Cassification Techniques. Thousand Oaks, CA: Sage Publications. Baligh, Helmy H. and Leon E. Richartz (1967), Vertical Market Structures. Boston: Allyn and Bacon. Bansal, Harvir S., Shirley F. Taylor, and Yannik St James (2005), “‘Migrating’ to New Service 134 Providers: Toward a Unifying Framework of Consumers’ Switching Behaviors,” Journal of the Academy of Marketing Science, 33 (1), 95-115. Barkholz, David and Mark Rechtin (2012), “Net Worth: GM’s Paid-ad Pullback Highlights a Key Question for Auto Makers: Can Facebook Sell Cars?,” in Automotive News: Crain. Baron, Robert S. and Norbert L. Kerr (2003), Group Process, Group Decision, Group Action (2 ed.). Buckingham, England: Open University Press. Bernoff, Josh (2010), “Social Technographics: Conversationalists get onto the Ladder,” in Empowered Vol. 2012. Bernoff, Josh and Charlene Li (2008), “Harnessing the Power of the Oh-So-Social Web,” MIT Sloan Management Review, 49 (3), 36. Bhasin, Kim (2011), “Chapstick is in Double Trouble After Censoring Women Who Called This Ad Sexist,” (accessed April 5, 2012), [available at http://www.businessinsider.com/sexistchapstick-ad-facebook-comments-2011-10]. Bortner, Brad, Ellen Daley, Heidi Shey, and Madiha Ashour (2008), “Will Web 2.0 Transform Market Research?,” Forrester. Brewer, Marilynn B. (2003), “Optimal Distinctiveness, Social Identity, and the Self,” in Handbook of Self and Identity, Mark R. Leary and June Price Tangney, eds. New York: Guilford Press. ---- (1991), “The Social Self: On Being the Same and Different at the Same Time,” Personality and Social Psychology Bulletin, 17 (5), 475-82. Brockner, James (1983), “Low Self-Esteem and Behavioral Plasticity,” in Review of Personality and Social Psychology. United States of America: Sage. Cardozo, Richard N., David K. Smith Jr, and Madhubalan Viswanathan (1988), “Identifying Key Customers for Novel Industrial Products,” Journal of Product Innovation Management, 5 (2), 102-13. Carlson, Brad D. (2005), “Brand-based Community: The Role of Identification in Developing a Sense of Community Among Brand Users,” Ph.D., Oklahoma State University. ----, Tracy A. Suter, and Tom J. Brown (2008), “Social Versus Psychological Brand Community: The Role of Psychological Sense of Brand Community,” Journal of Business Research, 61 (4), 284-91. Celsi, Richard L. and Jerry C. Olson (1988), “The Role of Involvement in Attention and Comprehension Processes,” Journal of Consumer Research, 15 (2), 210-24. 135 Churchill, Gilbert A., Jr. (1979), “A Paradigm for Developing Better Measures of Marketing Constructs,” Journal of Marketing Research, 16 (1), 64-73. Cook, Karen S. and Richard M. Emerson (1978), “Power, Equity and Commitment in Exchange Networks,” American Sociological Review, 43 (5), 721-39. Cova, Bernard and Stefano Pace (2006), “Brand Community of Convenience Products: New Forms of Customer Empowerment - The Case “My Nutella the Community.”,” European Journal of Marketing, 40 (9-10), 1087-105. Curran, Patrick J., Daniel J. Bauer, and Michael T. Willoughby (2004), “Testing Main Effects and Interactions in Latent Curve Analysis,” Psychological Methods, 9 (2), 220-37. Dholakia, Utpal M., Richard P. Bagozzi, and Lisa Klein Pearo (2004), “A Social Influence Model of Consumer Participation in Network- and Small-group-based Virtual Communities,” International Journal of Research in Marketing, 21 (3), 241-63. ----, Vera Blazevic, Caroline Wiertz, and René Algesheimer (2009), “Communal Service Delivery: How Customers Benefit From Participation in Firm-hosted Virtual P3 Communities,” Journal of Service Research, 12 (2), 208-26. ---- and Silvia Vianello (2009), “Business Insight (A Special Report): E-Commerce ---The Fans Know Best: When it Comes to Building Online Brand Communities, Do Unto Yourself as Others Already Do Unto You,” in Wall Street Journal. United States, New York, N.Y. ---- (2011), “Effective Brand Community Management: Lessons from Customer Enthusiasts,” IUP Journal of Brand Management, 8 (1), 7-21. Dickson, Peter R. and James L. Ginter (1987), “Market Segmentation, Product Differentiation, and Marketing Strategy,” Journal of Marketing, 51 (2), 1-10. Dröge, Cornelia (1989), “Shaping the Route to Attitude Change: Central versus Peripheral Processing through Comparative versus Noncomparative Advertising,” Journal of Marketing Research, 26 (2), 193-204. ----, Michael A. Stanko, and Wesley A. Pollitte (2010), “Lead Users and Early Adopters on the Web: The Role of New Technology Product Blogs,” Journal of Product Innovation Management, 27 (1), 66-82. Eagly, Alice Hendrickson and Shelly Chaiken (1993), The Psychology of Attitudes. Fort Worth, TX: Harcourt Brace Jovanovich College Publishers. Economist (2012), “Outsourcing is so Last Year,” in The Economist (Online). London, United Kingdom: The Economist Newspaper NA, Inc. Elkin, Noah, David Hallerman, Monica Peart, Tracy Tang, Martin Utreras, Haixia Wang, and 136 Debra Aho Williamson (2012), “US Digital Ad Spending: Key Trends for 2012,” eMarketer. Fornell, Claes and David F. Larcker (1981), “Evaluating Structural Equation Models with Unobservable Variables and Measurement Error,” Journal of Marketing Research, 18 (1), 39-50. Fournier, Susan (2000), “Building Brand Community on the Harley-Davidson Posse Ride,” in Harvard Business School Case. Harvard, MA: Harvard Business School Publication Corp. ---- (1998), “Consumers and Their Brands: Developing Relationship Theory in Consumer Research,” Journal of Consumer Research, 24 (4), 343-53. ---- and Jill Avery (2011), “The Uninvited Brand,” Business Horizons, 54 (3), 193-207. ----, Susan Dobscha, and David Glen Mick (1998), “Preventing The Premature Death of Relationship Marketing,” Harvard Business Review, 76 (1), 42-51. ---- and Lara Lee (2009), “Getting Brand Communities Right,” Harvard Business Review, 87 (4), 105-11. ----, Kathrin Sele, and Marcus Schogel (2005), “The Paradoxes of Brand Community Management,” Thexis (Special Issue On Community Marketing), 16-20. Fox, Kathryn Joan (1987), “Real Punks and Pretenders,” Journal of Contemporary Ethnography, 16 (3), 344. Füller, Johann, Kurt Matzler, and Melanie Hoppe (2008), “Brand Community Members as a Source of Innovation,” Journal of Product Innovation Management, 25 (6), 608-19. Gabisch, Jason A. and Kholekile L. Gwebu (2011), “Impact of Virtual Brand Experience on Purchase Intentions: The Role of Multichannel Congruence,” Journal of Electronic Commerce Research, 12 (4). Gentry, Lance and Roger Calantone (2002), “A Comparison of Three Models to Explain ShopBot Use on the Web,” Psychology & Marketing, 19 (11), 945-56. Gerbing, David W. and James C. Anderson (1988), “An Updated Paradigm for Scale Development Incorporating Unidimensionality and Its Assessment,” Journal of Marketing Research, 25 (2), 186-92. Gotlieb, Jerry and John Swan (1990), “An Application of the Elaboration Likelihood Model,” Journal of the Academy of Marketing Science, 18 (3), 221-28. Graham, Jill W. (1991), “An Essay on Organizational Citizenship Behavior,” Employee Responsibilities and Rights Journal, 4 (4), 249-70. GreenBook (Winter 2013), “Research Industry Trends Report,” Leonard Murphy (Ed.). 137 http://www.greenbook.org/PDFs/GRIT-W13.pdf. Hair, Joseph F., William C. Black, Barry J. Babin, Rolph E. Anderson, and Ronald L. Tatham (2006), Multivariate Data Analysis (6th ed.). Upper Saddle River, NJ: Pearson Prentice Hall. Hatch, Mary Jo and Majken Schultz (2010), “Toward a Theory of Brand Co-creation With Implications For Brand Governance,” Journal of Brand Management, 17 (8), 590-604. Hickman, Thomas (2005), “Intergroup Rivalry in Brand Communities: A Social Identity Theory Perspective,” Ph.D., Arizona State University. Hogg, Michael A. (2003), “Social Identity,” in Handbook of Self and Identity, Mark R. Leary and June Price Tangney, eds. New York: Guilford Press. ----, Deborah J. Terry, and Katherine M. White (1995), “A Tale of Two Theories: A Critical Comparison of Identity Theory with Social Identity Theory,” Social Psychology Quarterly, 58 (4), 255-69. Hollenbeck, Candice R. and George M. Zinkhan (2010), “Anti-brand Communities, Negotiation of Brand Meaning, and the Learning Process: The Case of Wal-Mart,” Consumption, Markets & Culture, 13 (3), 325-45. Homans, George C. (1958), “Social Behavior as Exchange,” American Journal of Sociology, 63 (6), 597-606. ---- (1974), Social Behavior; Its Elementary Forms (Rev. ed.). New York, NY: Harcourt Brace. Homburg, Christian, Ove Jensen, and Harley Krohmer (2008), “Configurations of Marketing and Sales: A Taxonomy,” Journal of Marketing, 72 (2), 133-54. ----, Nicole Koschate, and Wayne D. Hoyer (2005), “Do Satisfied Customers Really Pay More? A Study of the Relationship between Customer Satisfaction and Willingness to Pay,” Journal of Marketing, 69 (2), 84-96. Hu, Li-tze and Peter M. Bentler (1999), “Cutoff Criteria for Fit Indexes in Covariance Structure Analysis: Conventional Criteria Versus New Alternatives,” Structural Equation Modeling: A Multidisciplinary Journal, 6 (1), 1-55. Hunt, Shelby D. (2002), Foundations of Marketing Theory: Toward a General Theory of Marketing. Armonk, NY: M.E. Sharpe. Hutton, Graeme and Maggie Fosdick (2011), “The Globalization of Social Media: Consumer Relationships With Brands Evolve in the Digital Space,” Journal of Advertising Research, 51 (4), 564-70. IBM (2013), “Share of Global Companies With a Profile on Social Sites in 2010.” Statista. 138 Israel, Shel (2012), “Dell Modernizes IdeaStorm,” in Technology Vol. 2012: Forbes. Jarvis, Cheryl Burke, Scott B. MacKenzie, and Philip M. Podsakoff (2003), “A Critical Review of Construct Indicators and Measurement Model Misspecification in Marketing and Consumer Research,” Journal of Consumer Research, 30 (2), 199-218. Kanter, Rosabeth Moss (1972), Commitment and Community; Communes and Utopias in Sociological Perspective. Cambridge, MA: Harvard University Press. Kempf, Deanna S. and Robert E. Smith (1998), “Consumer Processing of Product Trial and the Influence of Prior Advertising: A Structural Modeling Approach,” Journal of Marketing Research, 35 (3), 325-38. Kotler, Philip and Kevin Lane Keller (2009), Marketing Management (13 ed.). Upper Saddle River, New Jersey: Pearson Prentice Hall. Krueger, Richard A. and Mary Anne Casey (2009), Focus Groups : A Practical Guide for Applied Research (4th ed.). Los Angeles, CA: SAGE. Kumar, V., J. Andrew Petersen, and Robert P. Leone (2010), “Driving Profitability by Encouraging Customer Referrals: Who, When, and How,” Journal of Marketing, 74 (5), 1-17. Lee, Doohwang, Hyuk Soo Kim, and Jung Kyu Kim (2011), “The Impact of Online Brand Community Type on Consumer’s Community Engagement Behaviors: Consumer-created vs. Marketer-created Online Brand Community in Online Social-networking Web Sites,” Cyberpsychology, Behavior, and Social Networking, 14 (1-2), 59-63. Leonardelli, Geoffrey J., Cynthia L. Pickett, and Marilynn B. Brewer (2010), “Chapter 2 Optimal Distinctiveness Theory: A Framework for Social Identity, Social Cognition, and Intergroup Relations,” in Advances in Experimental Social Psychology, P. Zanna Mark and M. Olson James, eds. Vol. Volume 43: Academic Press. Lovaglia, Michael J. (2007), “Social Exchange Theory,” in Blackwell Encyclopedia of Sociology Online, George Ritzer (Ed.) Vol. 2011. Madupu, Vivekananda (2006), “Online Brand Community Participation: Antecedents and Consequences,” Ph.D., The University of Memphis. ---- and Delonia O. Cooley (2010), “Antecedents and Consequences of Online Brand Community Participation: A Conceptual Framework,” Journal of Internet Commerce, 9 (2), 12747. Mahajan, Vijay, Eitan Muller, and Frank M. Bass (1995), “Diffusion of New Products: Empirical Generalizations and Managerial Uses,” Marketing Science, 14 (3), G79-G88. 139 ---- (1990), “New Product Diffusion Models in Marketing: A Review and Directions for Research,” Journal of Marketing, 54 (1), 1-26. Martin, William C. (2009), “Investigating the Antecedents and Consequences of Perceived Connectedness to Brand Users: Brand Communities Versus Brand Collectivities,” Ph.D., Mississippi State University. McAlexander, James H., John W. Schouten, and Harold F. Koenig (2002), “Building Brand Community,” Journal of Marketing, 66 (1), 38-54. McWilliam, Gil (2000), “Building Stronger Brands through Online Communities,” Sloan Management Review, 41 (3), 43-54. Meyer, John P. and Catherine A. Smith (2000), “HRM Practices and Organizational Commitment: Test of a Mediation Model,” Canadian Journal of Administrative Sciences, 17 (4), 319-31. Mishra, Sanjay, U. N. Umesh, and Donald E. Stem, Jr. (1993), “Antecedents of the Attraction Effect: An Information-Processing Approach,” Journal of Marketing Research, 30 (3), 331-49. Mittal, Banwari (1995), “A Comparative Analysis of Four Scales of Consumer Involvement,” Psychology and Marketing, 12 (7), 663-82. Moffitt, Sean (2008), “9 Types of Brand Community - A New Model,” (accessed December 21, 2010), [available at http://buzzcanuck.typepad.com/agentwildfire/2008/03/9-types-of-bran.html]. Moreland, Richard L. and John M. Levine (2002), “Socialization and Trust in Work Groups,” Group Processes and Intergroup Relations, 5 (3), 185-201. Morgan, Robert M. and Shelby D. Hunt (1994), “The Commitment-Trust Theory of Relationship Marketing,” Journal of Marketing, 58 (3), 20-38. MSI (2012), “2012-2014 Research Priorities.” http://www.msi.org/MSI_RP12-14.pdf: Marketing Science Institute. Muniz, Albert M. (1998), “Brand Community,” Ph.D., University of Illinois at UrbanaChampaign. ---- and Thomas C. O’Guinn (2001), “Brand Community,” Journal of Consumer Research, 27 (4), 412-32. ---- and Hope J. Schau (2005), “Religiosity in the Abandoned Apple Newton Brand Community,” Journal of Consumer Research, 31 (4), 737-47. ---- and Hope Jensen Schau (2007), “Vigilante Marketing and Consumer-Created Communications,” Journal of Advertising, 36 (3), 35-50. 140 Nudd, Tim (2011), “ChapStick Gets Itself in a Social Media Death Spiral: A Brand’s Slient War Against Its Facebook Fans,” (accessed April 19, 2013), [available at http://www.adweek.com/adfreak/chapstick-gets-itself-social-media-death-spiral-136097]. Park, C. Whan, Deborah J. MacInnis, Joseph Priester, Andreas B. Eisingerich, and Dawn Iacobucci (2010), “Brand Attachment and Brand Attitude Strength: Conceptual and Empirical Differentiation of Two Critical Brand Equity Drivers,” Journal of Marketing, 74 (6), 1-17. Petty, Richard E. and Pablo Briñol (2011), “The Elaboration Likelihood Model,” Handbook of Theories of Social Psychology: Volume One, 224. ---- and John T. Cacioppo (1981), Attitudes and Persuasion--Classic and Contemporary Approaches. Dubuque, Iowa: W.C. Brown Co. Publishers. ---- (1986), Communication and Persuasion: Central and Peripheral Routes to Attitude Change. New York: Springer-Verlag. ----, John T. Cacioppo, and David Schumann (1983), “Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement,” Journal of Consumer Research, 10 (2), 135-46. Petty, Richard E. and Duane T. Wegener (1999), “The Elaboration Likelihood Model: Current Status and Controversies,” in Dual Process Theories in Social Psychology, Shelly Chaiken and Yaacov Trope, eds. New York: Guilford Press. Phillips, Mark R., Bradley D. McAuliff, Margaret Bull Kovera, and Brian L. Cutler (1999), “Double-Blind Photoarray Administration as a Safeguard Against Investigator Bias,” Journal of Applied Psychology, 84 (6), 940-51. Ping, Robert A., Jr. (1995), “A Parsimonious Estimating Technique for Interaction and Quadratic Latent Variables,” Journal of Marketing Research, 32 (3), 336-47. Prykop, Catja and Mark Heitmann (2006), “Designing Mobile Brand Communities: Concept and Empirical Illustration,” Journal of Organizational Computing & Electronic Commerce, 16 (3/4), 301-23. Rossiter, John R. (2002), “The C-OAR-SE Procedure for Scale Development in Marketing,” International Journal of Research in Marketing, 19 (4), 305-35. Rust, Roland T., Tim Ambler, Gregory S. Carpenter, V. Kumar, and Rajendra K. Srivastava (2004), “Measuring Marketing Productivity: Current Knowledge and Future Directions,” Journal of Marketing, 68 (4), 76-89. Schau, Hope J., Albert M. Muñiz, and Eric J. Arnould (2009), “How Brand Community Practices Create Value,” Journal of Marketing, 73 (5), 30-51. 141 Schlosser, Ann E. (2003), “Come Together, Right Now, Virtually: An Examination into Online Communities,” Advances in Consumer Research, 30 (1), 192-95. Schouten, John W. (1991), “Selves in Transition: Symbolic Consumption in Personal Rites of Passage and Identity Reconstruction,” Journal of Consumer Research, 17 (4), 412-25. Schouten, John W. and James H. McAlexander (1995), “Subcultures of Consumption: An Ethnography of the New Bikers,” Journal of Consumer Research, 22 (1), 43-61. Shih, Pai Cheng, Hsin-Yun Hu, and Cheng-Kiang Farn (2010), “Lead User Participation in Brand Community: The Case of Microsoft MVPS,” International Journal of Electronic Business Management, 8 (4), 323-31. Solomon, Michael R. (1986), “The Missing Link: Surrogate Consumers in the Marketing Chain,” Journal of Marketing, 50 (4), 208-18. Spiggle, Susan (1994), “Analysis and Interpretation of Qualitative Data in Consumer Research,” Journal of Consumer Research, 21 (3), 491-503. Spreng, Richard A., Scott B. MacKenzie, and Richard W. Olshavsky (1996), “A Reexamination of the Determinants of Consumer Satisfaction,” Journal of Marketing, 60 (3), 15-32. Stelzner, Michael and Ric Dragon (2012), “Social Media Science: How Behavior Impacts Social Media Marketing,” in Social Media Examiner Vol. 2012. http://www.socialmediaexaminer.com/social-media-science/. Thompson, Scott A. (2009), “From Rumor to Release: Leveraging Social Identification in Marketing Strategy,” Ph.D., Arizona State University. ---- and Rajiv K. Sinha (2008), “Brand Communities and New Product Adoption:The Influence and Limits of Oppositional Loyalty,” Journal of Marketing, 72 (6), 65-80. Vorhies, Douglas W. and Neil A. Morgan (2003), “A Configuration Theory Assessment of Marketing Organization Fit with Business Strategy and Its Relationship with Marketing Performance,” Journal of Marketing, 67 (1), 100-15. Wang, Youcheng and Daniel R. Fesenmaier (2004), “Modeling Participation in an Online Travel Community,” Journal of Travel Research, 42 (3), 261-70. ----, Quaehee Yu, and Daniel R. Fesenmaier (2002), “Defining the Virtual Tourist Community: Implications for Tourism Marketing,” Tourism Management, 23 (4), 407-17. Warshaw, Paul R. (1980), “Predicting Purchase and Other Behaviors from General and Contextually Specific Intentions,” Journal of Marketing Research, 17 (1), 26-33. 142 Wedel, Michel and Wagner A. Kamakura (2000), Market Segmentation: Conceptual and Methodological Foundations (2nd ed.). Boston: Kluwer Academic. Weinberg, Bruce D. and Ekin Pehlivan (2011), “Social Spending: Managing the Social Media Mix,” Business Horizons, 54 (3), 275-82. Williams, Ruth L. and Joseph Cothrel (2000), “Four Smart Ways to Run Online Communities,” MIT Sloan Management Review, 41 (4), 81-81-91. Wong, Charles, Ian Wilkinson, and Louise Young (2010), “Towards an Empirically Based Taxonomy of Buyer-seller Relations in Business Markets,” Journal of the Academy of Marketing Science, 38 (6), 720-37. Yazbeck, Bob (2011), “Where do we go from here? Thoughts on how to address important questions about MROCs,” in Quirk’s Marketing Research Review. 4 ed. Vol. XXV. Yeh, Yi-Hsin and Sejung Marina Choi (2011), “MINI-lovers, Maxi-mouths: An Investigation of Antecedents to eWOM Intention Among Brand Community Members,” Journal of Marketing Communications, 17 (3), 145-62. Yu-Chen, Chen (2006), “The Value of Participation in Virtual Consumer Communities on Brand Loyalty,” Internet Research, 16 (4), 398-98. Zaichkowsky, Judith L. (1985), “Measuring the Involvement Construct,” Journal of Consumer Research, 12 (3), 341-52. ---- (1994), “The Personal Involvement Inventory: Reduction, Revision, and Application to Advertising,” Journal of Advertising, 23 (4), 59-70. 143