Unique influences of cognitive and affective customer-company identification

Journal of Business Research 78 (2017) 172–179

Contents lists available at ScienceDirect

Journal of Business Research journal homepage: www.elsevier.com/locate/jbusres

Unique influences of cognitive and affective customer-company identification

MARK

Jeremy S. Woltera,⁎, J. Joseph Cronin Jrb a

Auburn University, Raymond J. Harbert College of Business, Department of Marketing, 405 W. Magnolia Ave., Auburn, AL 36849, United States Florida State University, The College of Business, Marketing Department, Rovetta Business Annex, Room 420, P.O. Box 3061110, Tallahassee, FL 32306-1110, United States b

A R T I C L E I N F O

A B S T R A C T

Keywords: Customer-company identification Organizational identification Cognitive identification Affective identification Social cohesion

Recent research suggests both the cognitive and affective dimensions of customer-company identification (CCI) influence outcomes of interest such as customer loyalty. Yet no research has empirically examined whether there are separate firm influenced drivers of the cognitive (CCICog) and affective (CCIAff) dimensions of CCI. The current research examines how two sets of drivers, symbolic and social, uniquely affect CCICog and CCIAff in comparison to 21 control variables. The results suggest CCICog is primarily influenced by antecedents that assist in self-definition (e.g., identity similarity and in-group ties) whereas CCIAff is primarily influenced by antecedents that assist in self-evaluation (e.g., organizational prestige and in-group bond). In addition, social drivers enhance the effect of symbolic drivers on CCICog whereas social drivers attenuate the effect of symbolic drivers on CCIAff.

1. Introduction There is growing interest in customer-company identification (CCI), or the extent an organization represents one or more of a customer's social identities, as a potential tool in fostering long-term customer loyalty (Bagozzi, Bergami, Marzocchi, & Morandin, 2012; Lam, 2012). Importantly, CCI is comprised of at least cognitive (CCICog) and affective (CCIAff) dimensions, both of which are instrumental in driving customer loyalty (Bagozzi et al., 2012; Homburg, Wieseke, & Hoyer, 2009). Yet, research that examines antecedents of CCI either ignores the affective dimension or combines it with the cognitive dimension. In other words, methods of uniquely creating CCIAff are unexplored and unknown. Furthermore, CCIAff influences customer outcomes uniquely from CCICog (Wolter & Cronin, 2016). Thus, organizations can better develop and manage meaningful long-term customer relationships if the unique drivers of both CCIAff and CCICog are established. The objective of the current research is to alleviate this lack of attention by exploring whether antecedents uniquely affect CCICog and CCIAff. As shown in Fig. 1, this objective is accomplished by examining the direct effects and interactions of social and symbolic antecedents on CCICog and CCIAff over the effect of 21 control variables. The results suggest CCICog is primarily influenced by antecedents that assist in selfdefinition (e.g., identity similarity and in-group ties) whereas CCIAff is primarily influenced by antecedents that assist in self-evaluation (e.g.,

⁎

organizational prestige and in-group bond). Furthermore, social drivers enhance the effect of symbolic drivers on CCICog whereas social drivers attenuate the effect of symbolic drivers on CCIAff. Because the lack of attention on the unique influences of the CCI dimensions spans across literatures, the current research's findings contributes to the broader theory of social identification while also expanding theory on CCI specifically. 2. Conceptualizing CCICog and CCIAff Emerging theory considers CCICog and CCIAff as part of two systems (see Lam, 2012, Fig. 1). These systems serve different purposes (selfdefinition and self-evaluation, Wolter & Cronin, 2016) based on different self-motives (uncertainty reduction and self-enhancement, Johnson, Morgeson, & Hekman, 2012) that relate to different aspects of a customer's sense of self (self-concept and self-esteem, Lane & Scott, 2007). Self-categorization theory and social identity theory separately specify the two systems that lead to CCI (Johnson et al., 2012). From this viewpoint, social identity theory's rationale that identification develops from customers' desire to feel better about themselves (i.e., the self-esteem hypothesis) is an explanation of CCIAff formation. In accord with this view, CCIAff is the connection between the identity of an organization and the evaluation a customer applies to him or herself and the subsequent emotions this connection engenders

Corresponding author. E-mail addresses: [email protected] (J.S. Wolter), [email protected] (J. Joseph Cronin).

http://dx.doi.org/10.1016/j.jbusres.2017.05.010 Received 7 January 2016; Received in revised form 8 May 2017; Accepted 11 May 2017 0148-2963/ Published by Elsevier Inc.

Journal of Business Research 78 (2017) 172–179

J.S. Wolter, J. Joseph Cronin

Fig. 1. Conceptual models of antecedents to cognitive and affective CCI. Notes: a = symbolic drivers. b = social drivers. Dotted grey lines represent hypotheses not supported by the data. Solid grey lines represent empirically determined non-hypothesized effects. Lines with “*” represent non-hypothesized effects that are controlled.

symbolic member of an organization to acquire the organization's identity attributes and self-definitional properties (Tuškej, Golob, & Podnar, 2013). Because one's identity can be viewed negatively or positively, high identity similarity does not ensure a company reflects well on a customer. For example, a sports team may comprise residents' social identities despite not being the best or even that good. Customers may perceive congruity with Apple's identity even if the company is viewed unfavorably by many of their peers (Arsel & Stewart, 2015). Thus, we propose the following:

(Harris & Cameron, 2005). In contrast, self-categorization theory's rationale that identification develops from customers' desire to reduce painful social uncertainty (i.e., the uncertainty reduction hypothesis) is an explanation of CCICog formation (Johnson et al., 2012). Given this link, CCICog is a connection between the definition of an organization and the definition a customer applies to himself or herself (Dutton, Dukerich, & Harquail, 1994). 3. Antecedents to CCICog and CCIAff

H1. Identity similarity influences CCI primarily through CCICog rather than CCIAff.

3.1. Symbolic drivers of CCI: identity similarity and organizational prestige There are three main groups of antecedents that influence CCI: symbolic, social, and instrumental (Lam, 2012). However, emerging research suggests the effects of instrumental drivers decay quickly leaving symbolic and social drivers as the two primary influences of CCI (Lam, Ahearne, Mullins, Hayati, & Schillewaert, 2013). Symbolic drivers represent the central, distinctive, and enduring aspects of an organization's identity that must be signaled through marketing communications and product design to customers (Bhattacharya & Sen, 2003). Though companies do not fully control these aspects nor how they are communicated, symbolic characteristics are integral to crafting an appealing social identity for customers (Press & Arnould, 2011). Two symbolic characteristics, identity similarity and organizational prestige, are often linked to organizational identification across the marketing and management literatures (e.g., Ahearne, Bhattacharya, & Gruen, 2005; Dukerich, Golden, & Shortell, 2002). However, research has not considered these drivers in relation to both CCICog and CCIAff as the current research does below.

3.1.2. Organizational prestige Organizational prestige is the extent others whose opinion matters consider an organization well-regarded (Bhattacharya & Sen, 2003). Stated differently, prestige represents how a company reflects on a customer in the eyes of people who matter. The value, then, of high prestige is one of self-enhancement (Bhattacharya, Rao, & Glynn, 1995). If a customer's friends think highly of an organization that engages in CSR, then that customer will share in the high regard. As a result, an organization high in prestige is one that is likely to elicit CCIAff (denoted by the path marked H2 in Fig. 1). However, many companies are well regarded. So having high prestige does not specifically help customers separate their self and others into meaningful categories. More simply, being well regarded does not enhance one's self-knowledge nor define the content of the self-concept. Instead, prestige functions as a barometer of the merit of a company and by extension, how the company reflects on a customer's self-worth (Dutton & Dukerich, 1991). As a result of the above reasoning:

3.1.1. Identity similarity Identity similarity is the recognition that a company's identity matches a customer's identity, either on an attribute-by-attribute basis or as a gestalt match (Bhattacharya & Sen, 2003). Engaging with a company that matches one's personality allows a customer to signal relevant identity attributes to others (Ahearne et al., 2005; Lee, Park, Rapert, & Newman, 2012). Thereby, social uncertainty is reduced through the construction of “viable, cognitively consistent social identities” (Bhattacharya & Sen, 2003, p 80). As a result (and as denoted in Fig. 1 by the path marked H1), a customer will self-categorize as a

H2. Organizational prestige influences CCI primarily through CCIAff rather than CCICog. 3.2. Social drivers: in-group ties and in-group bond In comparison to symbolic influences of identification, the social influences are not nearly as well studied. To facilitate exploration of the social forces, we turn to social cohesion which is recognized as a theoretically viable explanation of identification formation (Friedkin, 173

Journal of Business Research 78 (2017) 172–179

J.S. Wolter, J. Joseph Cronin

2004). Social cohesion is not a construct but “the total field of [social] forces which act on members to remain in the group" (Festinger, 1950, p 274). Though there are many “forces” of social cohesion, some play a dominant role because they influence outcomes across many types of groups (Cota, Evans, Dion, Kilik, & Longman, 1995; Dion, 2000). Two such forces are the number of “mutual dyadic ties” between group members and the strength of such ties (Friedkin, 2004, p 417). Other researchers note these forces, in some form, as the “amount of interaction and feeling of friendship” within a group (Moos, 1976, p 327), as interpersonal relations and social attraction (Hogg & Hains, 1996), and as sociometric cohesion and social satisfaction (Hagstrom & Selvin, 1965). Extrapolating from this viewpoint, the current research considers social cohesion within a customer-company framework as consisting of two primary components: 1) in-group ties, the number of interpersonal relationships a customer has to an organization's members (i.e., employees and customers), and 2) ingroup bond, the extent a customer feels a connection and friendship with an organization's members. Importantly, these two states, though likely related, are distinct from each other (Hagstrom & Selvin, 1965) so each can affect CCICog and CCIAff differently as detailed below.

H4. In-group bond influences CCI primarily through CCIAff rather than CCICog. 3.3. Moderating effects of social cohesion The very idea of a social identity is that part of an individual's identity is sourced and developed from groups that impact social relations (Hogg, Sherman, Dierselhuis, Maitner, & Moffitt, 2007). From this standpoint, then, the ability of a company to become part of a social identity will be contingent on the extent said company is connected to a customer's social network (Bhattacharya & Sen, 2003). Stated differently, a customer's social interactions determine whether a social identity is viable. As a company gains a greater share of a customer's social network, identity similarity becomes increasingly important in facilitating a social identity for the customer. H5. As in-group ties increases, the effect of identity similarity on CCICog increases. While in-group ties acts as an enhancing moderator, in-group bond should act as an attenuating moderator of organizational prestige. This difference stems from the idea that customers utilize “contingencies of self-worth” to maintain the positivity of an identity (Crocker & Wolfe, 2001). From this perspective, customers willingly shift focus to positive aspects of an identity, even in the face of negative aspects. As such, a customer who has social inclusion with a company's customers (through in-group bond) is less likely to rely on the regard of others to determine CCIAff. Thus, a final hypothesis:

3.2.1. In-group ties As denoted in Fig. 1 by the path marked H3, in-group ties should influence CCICog because the benefit of a multiplex of interpersonal relationships is one of self-definition. Customers construct the meaning of an organization through “verbal and non-verbal interactions with other individuals” (Ashforth & Mael, 1989, p 27) wherein they “develop a common language or ideology that helps them construct a collective understanding of reality” (Postmes, Baray, Haslam, Morton, & Swaab, 2006, p 216). Also known as sensemaking (i.e., deriving meaning from experience, Ashforth, Harrison, & Corley, 2008), these interactions are crucial because they allow a customer to better understand an organization and connect their self-understanding to aspects of the organization (Press & Arnould, 2011). Thus, the more interpersonal relationships that connect a customer to an organization, the more that organization anchors the customer to society and better satiates selfdefinitional needs (Bhattacharya & Sen, 2003). If a customer is connected to others through an organization, it does not necessarily mean the organization reflects positively on the customer. In fact, organizations that are perceived negatively can still represent a customer's social network. As an extreme example, a customer may know several people through a company that offers vice products (e.g., gambling). Other more mundane examples include customers who have grown-up using the same grocery store as their family or a group of neighbors who use the same lawn service. In each of these cases, a customer may recognize a company as self-definitional because s/he is embedded in a social network tied to the company. However, the company does not necessarily reflect positively on the customer nor engender positive self-evaluation. As a result:

H6. As in-group bond increases, the effect of organizational prestige on CCIAff decreases. 4. Study design To test the proposed drivers of CCICog and CCIAff., we employed a cross-sectional study that controlled many explanatory variables to ensure the posited relationships are robust to confounding influences. Five surveys were necessary (with each one focusing on a different set of controls) to avoid having an overly long survey instrument that could fatigue survey respondents and thereby introduce common method bias (Podsakoff, Mackenzie, & Podsakoff, 2012). Each survey had all six of the primary constructs and four to eight controls which means the surveys ranged from 36 to 51 items. The chosen controls represent a wide range of potential drivers (e.g., symbolic and instrumental) and are shown along with the measures in the online Supplementary material (see Table ST1). We recruited from three groups of potential respondents to diversify the resulting sample. The first group consisted of undergraduate and graduate students from a large Southeastern university who were offered class credit for participation (30% of the resulting sample). The second group consisted of customers from the local area who were recruited by students (separate from the first group) and offered a chance to win prizes through a random drawing (26% of the resulting sample). The third group consisted of paid workers from Amazon's Mechanical Turk (44% of the resulting sample). In each method, participants filled out an electronic version of the survey so that all experienced the same type of survey. To check the validity and uniqueness of the participants, the research team assessed URLs and ensured each respondent completed the survey from a different location. Sample size for each survey fluctuated depending on the number of students in a class that were recruited or used as recruiters for a particular survey. Once they started the survey, respondents provided the name of a company with which they were a customer and liked. The survey software then inserted the company provided name automatically into the scale items. Though this method confounds respondents and

H3. In-group ties influences CCI primarily through CCICog rather than CCIAff. 3.2.2. In-group bond As denoted in Fig. 1 by the path marked H4, in-group bond should lead to CCIAff because obtaining social acceptance validates one's social self as worthy, and as a result, positive self-conscious emotions are experienced (Baumeister & Leary, 1995; Ellemers, Doosje, & Spears, 2004). In other words, a positive bond with others acts as a signal of the solidification of group and interpersonal relationships. Such social support is beneficial for a customer's psychological well-being and should bolster a customer's sense of self-worth (Rosenbaum & Massiah, 2007). Thus, in-group ties should lead to CCI through the cognitive dimension because the ties facilitate self-definition whereas in-group bonds should lead to CCI through the affective dimension because social acceptance facilitates positive self-evaluation. Thus: 174

Journal of Business Research 78 (2017) 172–179

J.S. Wolter, J. Joseph Cronin

companies, it ensures a range of responses are observed for constructs relating to customers' deep connections with companies (Batra, Ahuvia, & Bagozzi, 2012). The software showed the antecedent and control scales in a randomized order, followed by the scales for CCICog and CCIAff (also in a randomized order), followed by demographic questions. Each scale was shown on screen one-at-a-time.

Table 1 Measurement scales for primary constructs. λ CCICog – adapted from Johnson et al. (2012) 1. My identity includes my relationship with ______. 2. ______ is part of my sense of who I am. 3. Being associated with ______ helps me express my identity.

4.1. Measurement of research variables

CCIAff – adapted from Johnson et al. (2012) 1. Generally, being associated with ______ gives me a sense of pride. 2. Overall, I feel happy when I think about myself as a customer of ______. 3. The things that ______ stands for makes me feel good to be connected with it.

For CCICog and CCIAff, we adapted measures from Johnson et al.'s (2012) organizational identification scale because their constructed theory provided much of the theoretical foundation for the current research. Furthermore, the reported relationship between the cognitive and affective dimensions (r = 0.40) suggests the measured dimensions are very distinct and likely to exhibit different relationships with antecedents. Identity similarity was measured using a two-item scale that measures similarity from a holistic perspective (Malar, Krohmer, Hoyer, & Nyffenegger, 2011). This type of scale is generally preferable to one based on an attribute-by-attribute matching sequence (Sirgy et al., 1997). Organizational prestige was measured by using three items from Mael and Ashforth's (1992) widely utilized scale. For the measures of in-group ties and in-group bond, we culled items from several different scales as existing scales seem to mix items from the two constructs together (Cameron, 2004; Henry, Arrow, & Carini, 1999). A pre-test of these scales, based on data from undergraduate students, employed factor analysis to determine if any items had small loadings on an intended construct (< 0.5) or large cross loadings on other constructs (> 0.5). The resulting scales are each comprised of three items. Of the 21 control variables, 16 had established scales and five required scale development. All scales and item loadings for the primary research variables are shown in Table 1 whereas those for the control variables are available in the online Supplementary material (see Table ST1).

0.90 0.92 0.88 0.90 0.89 0.88

Identity similarity – adapted from Malar et al. (2011) 1. The personality of ______ is consistent with how I see myself. 2. The personality of ______ is a mirror image of me.

0.81 0.84

Organizational prestige – from Mael and Ashforth (1992) 1. People who I care about think that ______ is a well-respected company. 2. People who I care about think highly of ______. 3. People who I care about think that ______ has a good reputation.

0.94 0.91 0.95

In-group ties – developed based on Cameron (2004) and Henry et al. (1999) 1. Being a customer of ______ has led me to interact with a lot of people. 2. I am connected to many other people through ______. 3. ______ represents a network of people I know.

0.85 0.91 0.85

In-group bond – developed based on Cameron (2004) and Henry et al. (1999) 1. I feel a bond with customers of ______. 2. I would get along with many customers of ______. 3. When I am around other customers of ______, I feel like part of the group.

0.85 0.74 0.92

Marker variable Respond to the following statements in regards to the last time you went grocery shopping. 1. I knew exactly what I needed before I went grocery shopping. 2. I decided upon what I needed while I was grocery shopping. All measures had a nine-point scale and were anchored by “strongly disagree” and “strongly agree”, and had “neither agree nor disagree” as a neutral point. The respondent's self-elicited company was dynamically placed by the survey software in the blank on each scale item.

4.2. Statistical tests for assessing hypotheses We now establish the statistical means for testing the hypotheses for sake of clarity. To assess the hypotheses, the current research relies on three statistical tests that establish an antecedent's influence is primarily through one CCI dimension. The relevant hypothesis is supported if it passes all three tests. The first test is the direct effect of the antecedent on a given CCI dimension as determined by a coefficient that is statistically different from zero. The second test is the relative effect of an antecedent on a given CCI dimension as determined by comparing whether an antecedent's coefficient on one CCI dimension is significantly stronger than the antecedent's coefficient on the other dimension. A third test takes into account the control variables through a credibility interval based on the observed effects of the CCI drivers within each survey (Field & Gillett, 2010). If a credibility interval contains zero, based on antecedents' coefficients from each survey in which the control variables are included in the analysis, the third statistical test is not passed. The moderation hypotheses (H5 and H6) are tested by assessing whether the relevant interaction coefficient is significantly different from zero. The Latent Moderated Structural Equations (LMS) approach is used to estimate the coefficients of the latent variable interactions through Mplus (Utz, 2004). However, fit indices are not available through LMS, so the direct effect hypotheses are tested without the interactions using maximum likelihood estimation.

respondents (51%) and males (50%) and the average age is 35 years. More demographic information and sample characteristics are available in the Supplementary material (see Table ST2). Some respondents (n = 158, 7%) did not pass the several engagement checks in the survey instrument. Without these respondents, the results exhibit a strengthening of the coefficients that support hypothesis testing. As such, these respondents were kept in the sample to provide a more conservative test of the hypotheses. We next tested for differences in the constructs across the sample sub-groups. Though some small differences exist and are statistically significant, the differences are small and significance is mostly the result of sample size. These differences in constructs between the sub-groups are available in the online Supplementary material (see Table ST3). A confirmatory factor analysis (CFA) assessed the quality of the measures for the primary research variables. We set up the measurement model such that each item loaded on its intended construct, all constructs' covariances (ϕ) freely varied, and the variance of each construct was set to one so the measurement model would be identified. The results of the CFA (as shown in Table 2) suggest the measures are adequate (χ2(104) = 541.97; CFI = 0.99; TLI = 0.98; RMSEA = 0.044). Though the chi-square value and the ratio of chi-square to degrees of freedom is high (5.2), this is normal for studies with large sample sizes (Marsh, Balla, & Mcdonald, 1988). The measures exhibit strong reliability as all composite reliabilities are above 0.7. Furthermore, the constructs exhibit convergent and discriminant validity as all average variances extracted (AVE) are above 0.5 and all shared variances are lower than the corresponding AVE (Fornell & Larcker, 1981). Because some of the constructs (e.g., in-group ties and in-group

5. Results 5.1. Preliminary analyses Of the 2257 surveys collected, 38 had missing data (2%) leaving a final sample of 2219. The sample is comprised mostly of married 175

Journal of Business Research 78 (2017) 172–179

J.S. Wolter, J. Joseph Cronin

Table 2 Measurement properties of primary research constructs across all surveys.

1 2 3 4 5 6 7 8

CCICog CCIAff Identity similarity Organizational prestige In-group ties In-group bond Marker variable 1 Marker variable 2 Mean Standard deviation

CR

1

2

3

4

5

6

7

8

0.92 0.92 0.81 0.95 0.90 0.87 – –

0.81 0.66 0.67 0.32 0.64 0.61 −0.10 0.17 4.3 2.0

0.44 0.80 0.69 0.59 0.54 0.63 − 0.01 0.17 5.8 1.8

0.45 0.48 0.68 0.47 0.51 0.60 −0.04 0.16 5.1 1.8

0.10 0.35 0.22 0.87 0.38 0.41 0.08 0.15 6.6 1.7

0.41 0.29 0.26 0.14 0.76 0.65 − 0.07 0.17 4.9 2.1

0.37 0.40 0.36 0.17 0.43 0.70 −0.04 0.14 5.3 1.7

0.01 0.00 0.00 0.01 0.01 0.00 – − 0.23 6.8 1.9

0.03 0.03 0.03 0.02 0.03 0.02 0.05 – 4.8 2.3

Notes: Model fit ignoring different surveys: χ2(104) = 541.97, p < 0.01; CFI = 0.99; TLI = 0.98; RMSEA = 0.044. Model fit controlling for different surveys: χ2(632) = 2381.12, p < 0.01; CFI = 0.95; TLI = 0.94; RMSEA = 0.080. All items were measured using nine-point scales. Model fit was estimated in a CFA that did not include the two marker variables. Marker variable 1 = the summation of the two marker variable items with the second item reversed. Marker variable 2 = the summation of the two marker variable items without the second item reversed. Correlations are shown in the lower triangle. Average variance extracted is shown on the diagonal. Shared variance is shown in the upper triangle. All correlations with an absolute value > 0.04 are significant at the 0.05 alpha level. Scales were measured with a nine-point scale.

(see Table ST4 and ST5). The structural model fit the data well (χ2(126) = 616.24; CFI = 0.98; TLI = 0.98; RMSEA = 0.042, SRMR = 0.021) and the results generally support the proposed relationships. More precisely, the coefficients indicating the relationships of identity similarity (γ = 0.42, p < 0.01) and in-group ties (γ = 0.34, p < 0.01) with CCICog are strong and significant. Unexpectedly, identity similarity also exhibits a strong relationship with CCIAff (γ = 0.37, p < 0.01) whereas in-group ties exhibits a weak relationship (γ = 0.09, p < 0.01). Despite its relationship with CCIAff, identity similarity (Δ χ = 8.27, Δ df = 1, p < 0.05) exhibits a significantly stronger relationship with CCICog as compared to CCIAff as does in-group ties (Δ χ = 89.00, Δ df = 1, p < 0.01). Thus, H1 and H3 pass the first two statistical tests. The coefficients representing the relationships of organizational prestige (γ = 0.29, p < 0.01) and in-group bond (γ = 0.22, p < 0.01) with CCIAff are also strong and significant. However, in-group bond exhibits a moderate relationship with CCICog (γ = 0.17, p < 0.01) whereas organizational prestige surprisingly exhibits a negative relationship (γ = −0.08, p < 0.01). As such, only organizational prestige (Δ χ = 224.61, Δ df = 1, p < 0.01), and not in-group bond (Δ χ = 1.10, Δ df = 1, p = 0.7), exhibits a significantly stronger relationship with CCIAff as compared to CCICog. Thus, H2 passes the first two statistical tests whereas H4 only passes one.

bond) were theoretically similar and had a high correlation, a more stringent discriminant validity test is needed to ensure the quality of the measures. The heterotrait-monotrait ratio of correlations (HTMT) provides such a test as it is one of the more conservative discriminant validity tests available (Henseler, Ringle, & Sarstedt, 2015). The results of the HTMT analysis strongly supports discriminant validity because the HTMT score for any combination of two constructs is lower than a conservative threshold of 0.75 (Voorhees, Brady, Calantone, & Ramirez, 2015). Common method bias is a concern because the data for the entire model is collected within a single survey instrument (Podsakoff et al., 2012). Given such a problem, the marker variable test is an appropriate tool to assess the degree of common method bias (Lindell & Whitney, 2001). Two items were included in each survey instrument to measure whether a respondent planned their last shopping trip as shown in Table 1. Because these items are worded so that they are opposites, two different marker variables can be created. The first variable reverses one of the items to represent general common method bias. The second variable adds the variables together without reversing one of the items thereby capturing if a respondent is using an acquiescence response style (Podsakoff et al., 2012). Whereas the published literature suggests using the second smallest correlation between all research variables and a marker variable as an assessment of common method bias (Lindell & Whitney, 2001), the current research used the highest correlation (organizational prestige r = 0.17, as shown in Table 2) for a more conservative test. When this correlation is partialled out, no other correlations lose statistical significance. Furthermore, even using the upper bound of the 0.99 confidence interval of the marker variable correlation results in no other correlations losing significance. Thus, this analysis suggests common method bias is not a concern.

5.3. Do the antecedent direct effects hold when examined with controls? We next tested the effects of the antecedents in relation to the measured controls within each survey using SEM. The coefficients across each survey were used to develop credibility intervals (as shown in the far right column of Table 3). For the effects of identity similarity and in-group ties on CCICog as well as that of organizational prestige and identity similarity on CCIAff, the credibility intervals do not contain zero. In contrast, the credibility intervals for all other antecedent relationships, including that of in-group bond and organizational prestige on CCICog as well as that of in-group bond and in-group ties on CCIAff contain zero. Thus, the hypothesis regarding in-group bond (H4) did not pass the last two statistical tests whereas the other antecedent hypotheses passed all three statistical tests.

5.2. Are the direct effect hypotheses supported by the combined data? For hypothesis testing, we utilized SEM (Mplus v7.11) and specified the structural model such that all antecedents affected each CCI dimension (to assess the second statistical test that compares an antecedent's coefficients across the dimensions). The structural model included the marker variables to control for common method bias and the covariance between CCICog and CCIAff to provide information on the dimensions' interrelationship (no differences in the antecedent coefficients were found if the covariance was freed or constrained to zero). We tested the antecedent relationships using other forms of statistical analysis (both recursive and non-recursive) and tested the coefficients across the sample sub-groups. Though there are some differences in coefficients, there are more similarities than differences. The results for these two analyses are available in the online Supplementary material

5.4. Are the moderation effects supported by the combined data? The next analysis focused on the moderation hypotheses. To test that the variables are related to only a single dimension of CCI, the structural model specified both social cohesion antecedents as affecting a direct effect of interest. In other words, both in-group ties and ingroup bond moderated the effect of identity similarity on CCICog and 176

177

0.58

0.42⁎⁎ − 0.08⁎⁎ 0.34⁎⁎ 0.17⁎⁎ 0.30 − 0.04⁎⁎ 0.03+,⁎

CCICog

All

0.61

− 0.01⁎⁎ 0.02⁎⁎

0.37⁎⁎ 0.29⁎⁎ 0.09⁎⁎ 0.22⁎⁎

CCIAff

0.63

0.38⁎⁎ − 0.01⁎⁎ 0.29⁎⁎ 0.27⁎⁎ 0.16 − 0.01⁎⁎ − 0.01⁎⁎ 0.10+,⁎ − 0.03⁎⁎ − 0.06⁎⁎ 0.08⁎⁎

CCICog

0.69

−0.01⁎⁎ 0.06+ ⁎ 0.01⁎⁎ −0.15⁎⁎ 0.09⁎⁎ 0.10+,⁎

0.25⁎⁎ 0.25⁎⁎ 0.02⁎⁎ 0.29⁎⁎

CCIAff

1 (429)

0.57

− 0.00⁎⁎ 0.10⁎⁎ − 0.01⁎⁎ − 0.04⁎⁎ 0.02⁎⁎

− 0.12⁎⁎ − 0.11⁎⁎ 0.10⁎⁎ − 0.06⁎⁎ 0.06⁎⁎

0.52

0.01⁎⁎ 0.03⁎⁎

0.32⁎⁎ 0.16⁎⁎ 0.05⁎⁎ 0.32⁎⁎

CCIAff

0.29⁎⁎ − 0.02⁎⁎ 0.26⁎⁎ 0.27⁎⁎ 0.37 − 0.01⁎⁎ 0.04⁎⁎

CCICog

2 (526)

0.71

0.16⁎⁎ 0.17⁎⁎ − 0.05+,⁎ − 0.01⁎⁎ − 0.03⁎⁎ 0.01⁎⁎ − 0.03⁎⁎

0.36⁎⁎ − 0.09⁎⁎ 0.28⁎⁎ 0.11⁎⁎ 0.21 − 0.02⁎⁎ 0.02⁎⁎

CCICog

0.63

0.11⁎⁎ 0.24⁎⁎ 0.04⁎⁎ 0.08⁎⁎ − 0.12⁎⁎ 0.04⁎⁎ − 0.03⁎⁎

− 0.02⁎⁎ − 0.05+,⁎

0.23⁎⁎ 0.20⁎⁎ 0.12⁎⁎ 0.11⁎⁎

CCIAff

3 (645)

0.64

0.07⁎⁎ 0.09⁎⁎ − 0.09⁎⁎ 0.07⁎⁎ 0.02⁎⁎

0.40⁎⁎ − 0.12+,⁎ 0.29⁎⁎ 0.22⁎⁎ 0.32 − 0.14⁎⁎ 0.02⁎⁎

CCICog

0.76

0.29⁎⁎ 0.12⁎⁎ − 0.03⁎⁎ − 0.02⁎⁎ − 0.04⁎⁎

− 0.08⁎⁎ − 0.05⁎⁎

0.31⁎⁎ 0.14⁎⁎ 0.14⁎⁎ 0.11+,⁎

CCIAff

4 (326)

0.13+,⁎ 0.05⁎⁎ − 0.02⁎⁎ 0.13⁎⁎ 0.49

0.29⁎⁎ 0.01⁎⁎ 0.22⁎⁎ 0.11⁎⁎ 0.46 − 0.07⁎⁎ 0.08+,⁎

CCICog

0.17⁎⁎ 0.04⁎⁎ 0.07⁎⁎ 0.04⁎⁎ 0.65

− 0.05⁎⁎ 0.10⁎⁎

0.36⁎⁎ 0.25⁎⁎ 0.02⁎⁎ 0.20⁎⁎

CCIAff

5 (293)

[0.17/0.51] [− 0.23/0.13] [0.22/0.32] [− 0.12/0.51]

CCICog

[0.10/0.47] [0.04/0.36] [− 0.12/0.27] [− 0.19/0.60]

CCIAff

Credibility intervals

Notes: Model fit for the combined data (as shown in the column titled “All”): χ2(126) = 616.24, p < 0.01; CFI = 0.98; TLI = 0.98; RMSEA = 0.042, SRMR = 0.021. Coefficients shown are standardized. Credibility intervals were computed using non-standardized coefficients. Numbers in parentheses represent the sample size for that survey. Marker variable 1 = the summation of the two marker variable items with the second item reversed. Marker variable 2 = the summation of the two marker variable items without the second item reversed. ⁎⁎ p < 0.01. ⁎ p < 0.05. + p < 0.1.

Identity similarity → Organizational prestige → In-group ties → In-group bond → CCICog CCIAff covariance (Φ) Marker variable 1 → Marker variable 2 → Identity over-inclusiveness → Identity conflict → Identity clarity → Identity stability → Customer familiarity → Customer satisfaction → Customer attitude → Utilitarian value → Hedonic value → Customer entitativity → Identity centrality → Org. personality: Boy scout → Org. personality: Innovative → Org. personality: dominance → Org. personality: stylish → Org. personality: thrift → Customer trust → Memorable experiences → Organizational warmth → Social desirability bias: ERT → Social desirability bias: MRT → Organizational distinctiveness → Brand engagement → Disposition towards orgs → Belief in CCI → R2

Survey (n)

Table 3 Antecedent coefficients from within surveys.

J.S. Wolter, J. Joseph Cronin

Journal of Business Research 78 (2017) 172–179

Journal of Business Research 78 (2017) 172–179

J.S. Wolter, J. Joseph Cronin

both self-definition and self-evaluation. Interestingly, this relationship held even when controlling for the perceived positivity of the company both externally (i.e., prestige) and internally (i.e., attitude). Interestingly, this relationship held even when controlling for the perceived positivity of the company whether the positivity was internal (i.e., attitude) or external (i.e., prestige) to the customer. By examining a broad range of control variables, the current research finds that though the CCI dimensions are driven by key variables (in the form of identity similarity, organizational prestige, and in-group ties), the CCI dimensions are influenced by a multitude of sources. An examination of the controls listed in Table 3 suggests the number that exhibit a significant relationship with CCIAff is approximately double the number that exhibit a significant relationship with CCICog. Thus, it would seem that the determinants of CCIAff are wide ranging whereas the determinants of CCICog are more narrow. Linking this to the underlying self-motives of each CCI dimension, the ways of satisfying self-enhancement seem plentiful whereas the methods for satisfying self-uncertainty rests primarily on a customer perceiving a firm as being similar and having many interpersonal relationships in the context of the firm. The current research further expands theory on CCI by exploring social influences. Social cohesion, to the authors' best knowledge, has not been explored at all as a potential explanation of CCI. Thus, the current research answers calls to reconcile how social and symbolic antecedents converge to form CCI (Lam, 2012). Furthermore, the results provide empirical support for Bhattacharya and Sen's (2003, p 82) conceptualization that “in contrast to arm's-length relationships, consumers' embedded relationships with companies are likely to be strong, intricate, and trusting, resulting in consumers feeling more like insiders than outsiders”. While qualitative evidence also supports the role of social forces in the formation of CCI (Press & Arnould, 2011), the current research evidences the utility of the social cohesion framework and that social forces can affect CCI uniquely through different dimensions.

moderated the effect of organizational prestige on CCIAff. The moderation of the two symbolic antecedents was tested separately. The results reveal that the interaction of in-group ties and identity similarity on CCICog is positive and significant (γ = 0.06, p < 0.01) whereas the interaction between in-group bond and identity similarity on CCICog is not (γ = 0.01, p = 0.43). Additionally, the interaction between ingroup ties and organizational prestige on CCIAff is insignificant (γ = 0.00, p = 0.91) whereas the interaction between in-group bond and organizational prestige on CCIAff is marginally significant (γ = − 0.02, p < 0.1). Notably, the interaction between in-group bond and organizational prestige on CCIAff becomes fully significant (γ = − 0.02, p < 0.05) when the interaction between in-group ties and organizational prestige on CCIAff is not in the model. Thus, the hypotheses (H5 and H6) that the effects of the symbolic drivers on the CCI dimensions are moderated differently by social drivers are supported by the data. 6. Discussion The current research examines the unique antecedents of CCICog and CCIAff. The proposed model is strongly supported as five of the six hypotheses are confirmed across the different analyses. Generally, the current research finds antecedents that determine the overall positivity of an organization (e.g., organizational prestige and in-group bond) have a relationship with CCI primarily through CCIAff whereas antecedents that tie an organization to a customer's self-concept (e.g., identity similarity and in-group ties) have a relationship with CCI primarily through CCICog. Unexpectedly, identity similarly was found to have a strong secondary effect on CCIAff. These findings provide several theoretical implications as discussed below. 6.1. Theoretical implications The unique drivers of the cognitive and affective dimensions of identification are a gap in the literature that extends beyond marketing and management to social psychology and the more broadly defined social identification. Given that the current research evidenced unique drivers of each dimension, empirical evidence is provided supporting Ashforth, Harrison, and Corley's conceptual idea (2008, p 329) that employees or customers can “think or feel” their way into identification. In other words, CCI creation is not simply facilitated by creating CCICog and allowing CCIAff to result. Instead, both dimensions have to be managed separately. This finding is in line with research in management that suggests stakeholders are continually exploring and assessing how companies impact their identities from both a definitional and evaluative standpoint (Ashforth et al., 2008). As such, companies must facilitate this exploration to the desired identity by managing the drivers of each CCI dimension. Regarding unique drivers of the CCI dimensions, the current research finds that organizational prestige and in-group ties are strongly associated with only one dimension of CCI, organizational prestige with CCIAff and in-group ties with CCICog. As such, the current results provide an explanation as to why previous research has sometimes failed to link prestige to CCI (Ahearne et al., 2005), consumerbrand identification (Stokburger-Sauer, Ratneshwar, & Sen, 2012), or employee-organizational identification (Ciftcioglu, 2011). Namely, organizational prestige acts as a barometer of an organization's ability to reflect positively on a customer and this attracts customers through CCIAff. Previous research that focused only on the cognitive dimension of identification missed this fundamental relationship unless affective items were used in the measurement of identification (e.g., Dutton et al., 1994). The finding that identity similarity influenced both CCI dimensions is in line with recent research that links self-company congruity to emotional attachment (Malar et al., 2011). In other words, the extent a company matches a customer's personality and values seemingly assists

6.2. Implications for marketing practice Given the effects observed in the current research, practical implications can be drawn. The first implication is that companies must manage the two CCI dimensions separately. Previous research urges companies to espouse values that are similar to a desired customer segment while also winning acclaim, awards, and recognition simultaneously (Bhattacharya & Sen, 2003). However, an efficiency argument can be made that a company should focus primarily on one dimension based on its unique outcomes or because one dimension is sufficient for customer loyalty attitudes (Wolter & Cronin, 2016). If a company chooses to focus on CCICog, it would be well suited in focusing on a values-led strategy and incorporating ways to create networked relationships among customers. Broad-based acclaim and a company's general reputation are not a necessary consideration when CCICog is the goal. In other words, focusing on representing one's chosen target market may be a more worthwhile endeavor than improving reputational rankings or fighting to keep a net-positive public image. Not that these are worthless pursuits, only that they are unnecessary for developing cognitive identity-based connections with customers. A second implication is that engendering and maintaining CCI is well served by increasing a customer's social cohesion as early as possible no matter which dimension of CCI is desired. This implication is because in-group ties enhance the effect of identity similarity on CCICog and in-group bonds protect CCIAff from downturns in corporate reputation. One potential method is to convert a customer's contacts into new customers by offering free trials for customers' friends and referral rewards. Thereby, a customer becomes tied to the organization through existing social circles. Another method is to facilitate customers meeting each other so that a customer's social circles grow within the reach of the organization. For example, a company could host a 178

Journal of Business Research 78 (2017) 172–179

J.S. Wolter, J. Joseph Cronin

cooperative behaviors of physicians. Administrative Science Quarterly, 47(3), 507–533. Dutton, J. E., & Dukerich, J. M. (1991). Keeping an eye on the mirror: Image and identity in organizational adaptation. Academy of Management Journal, 34(3), 517–554. Dutton, J. E., Dukerich, J. M., & Harquail, C. V. (1994). Organizational images and member identification. Administrative Science Quarterly, 39(2), 239–263. Ellemers, N., Doosje, B., & Spears, R. (2004). Sources of respect: The effects of being liked by ingroups and outgroups. European Journal of Social Psychology, 34(2), 155–172. Festinger, L. (1950). Informal social communication. Psychological Review, 57(5), 271–282. Field, A. P., & Gillett, R. (2010). How to do a meta-analysis. British Journal of Mathematical and Statistical Psychology, 63, 665–694. Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. Friedkin, N. E. (2004). Social cohesion. Annual Review of Sociology, 30, 409–425. Hagstrom, W. O., & Selvin, H. C. (1965). Two dimensions of cohesiveness in small groups. Sociometry, 28(1), 30–43. Harris, G. E., & Cameron, J. E. (2005). Multiple dimensions of organizational identification and commitment as predictors of turnover intentions and psychological well-being. Canadian Journal of Behavioural Science, 37(3), 159–169. Henry, K. B., Arrow, H., & Carini, B. (1999). A tripartite model of group identification: Theory and measurement. Small Group Research, 30(5), 558–581. Henseler, J., Ringle, C., & Sarstedt, M. (2015). A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43(1), 115–135. Hogg, M. A., & Hains, S. C. (1996). Intergroup relations and group solidarity: Effects of group identification and social beliefs on depersonalized attraction. Journal of Personality and Social Psychology, 70(2), 295–309. Hogg, M. A., Sherman, D. K., Dierselhuis, J., Maitner, A. T., & Moffitt, G. (2007). Uncertainty, entitativity, and group identification. Journal of Experimental Social Psychology, 43(1), 135–142. Homburg, C., Wieseke, J., & Hoyer, W. D. (2009). Social identity and the service profit chain. Journal of Marketing, 73(2), 38–54. Johnson, M. D., Morgeson, F. P., & Hekman, D. R. (2012). Cognitive and affective identification: Exploring the links between different forms of social identification and personality with work attitudes and behavior. Journal of Organizational Behavior, 33(8), 1142–1167. Kietzmann, J. H., Hermkens, K., Mccarthy, I. P., & Silvestre, B. S. (2011). Social media? Get serious! Understanding the functional building blocks of social media. Business Horizons, 54(3), 241–251. Lam, S. (2012). Identity-motivated marketing relationships: Research synthesis, controversies, and research agenda. AMS Rev, 2(2–4), 72–87. Lam, S., Ahearne, M., Mullins, R., Hayati, B., & Schillewaert, N. (2013). Exploring the dynamics of antecedents to consumer–brand identification with a new brand. Journal of the Academy of Marketing Science, 41(2), 234–252. Lane, V. R., & Scott, S. G. (2007). The neural network model of organizational identification. Organizational Behavior and Human Decision Processes, 104(2), 175–192. Lee, E. M., Park, S.-Y., Rapert, M. I., & Newman, C. L. (2012). Does perceived consumer fit matter in corporate social responsibility issues? Journal of Business Research, 65(11), 1558–1564. Lindell, M. K., & Whitney, D. J. (2001). Accounting for common method variance in crosssectional research designs. Journal of Applied Psychology, 86(1), 114–121. Mael, F., & Ashforth, B. E. (1992). Alumni and their alma mater: A partial test of the reformulated model of organizational identification. Journal of Organizational Behavior, 13(2), 103–123. Malar, L., Krohmer, H., Hoyer, W. D., & Nyffenegger, B. (2011). Emotional brand attachment and brand personality: The relative importance of the actual and ideal self. Journal of Marketing, 75(4), 35–52. Marsh, H. W., Balla, J. R., & Mcdonald, R. P. (1988). Goodness-of-fit indexes in confirmatory factor analysis: The effect of sample size. Psychological Bulletin, 103(3), 391–410. Moos, R. H. (1976). The human context: Environmental determinants of behavior. Wiley. Podsakoff, P. M., Mackenzie, S. B., & Podsakoff, N. P. (2012). Sources of method bias in social science research and recommendations on how to control it. Annual Review of Psychology, 63(1), 539–569. Postmes, T., Baray, G., Haslam, S. A., Morton, T. A., & Swaab, R. I. (2006). The dynamics of personal and social identity formation. In T. Postmes, & J. Jetten (Eds.), Individuality and the group: Advances in social identity (pp. 215–236). London: Sage. Press, M., & Arnould, E. J. (2011). How does organizational identification form? A consumer behavior perspective. Journal of Consumer Research, 38(4), 650–666. Rosenbaum, M. S., & Massiah, C. A. (2007). When customers receive support from other customers: Exploring the influence of intercustomer social support on customer voluntary performance. Journal of Service Research, 9(3), 257–270. Sirgy, M., Grewal, D., Mangleburg, T., Park, J.-O., Chon, K.-S., Claiborne, C., ... Berkman, H. (1997). Assessing the predictive validity of two methods of measuring self-image congruence. Journal of the Academy of Marketing Science, 25(3), 229–241. Stokburger-Sauer, N., Ratneshwar, S., & Sen, S. (2012). Drivers of consumer–brand identification. International Journal of Research in Marketing, 29(4), 406–418. Tuškej, U., Golob, U., & Podnar, K. (2013). The role of consumer–brand identification in building brand relationships. Journal of Business Research, 66(1), 53–59. Utz, S. (2004). Self-construal and cooperation: Is the interdependent self more cooperative than the independent self? Self and Identity, 3(3), 177–190. Voorhees, C. M., Brady, M. K., Calantone, R., & Ramirez, E. (2015). Discriminant validity testing in marketing: An analysis, causes for concern, and proposed remedies. Journal of the Academy of Marketing Science, 1–16. Wolter, J. S., & Cronin, J. J. (2016). Re-conceptualizing cognitive and affective customer–company identification: The role of self-motives and different customerbased outcomes. Journal of the Academy of Marketing Science, 44(3), 397–413.

social event offered through a local chamber of commerce. The two methods outlined above could be considered more traditional methods for establishing social connections among customers. However, in today's age with mobile computing and social media, the easiest way of engendering in-group ties is probably through online communities. In fact, a strong social media presence for a brand inherently enables the building blocks (e.g., sharing, conversations, and presence) that foster relationships and group formation among customers (Kietzmann, Hermkens, Mccarthy, & Silvestre, 2011). Though not explicitly studied in the current research, it is logical that online connections should foster the same perception of in-group ties as offline connections. 6.3. Limitations A limitation of the current research is the focus on only two dimensions of social cohesion. Though there is a theoretical rationale that in-group bonds and in-group ties are primary because they can apply to all groups (e.g., Dion, 2000), their influence should be compared to other constructs within the social/group cohesion domain. For example, vertical cohesion (the relationship between leaders and followers, Dion, 2000) could be influential to the extent that an organization has recognizable leaders. An additional limitation is the data are cross-sectional. As such, the examined relationships cannot be considered causal and the moderation relationships could be interpreted in the opposite direction (e.g., identity similarity enhances the effect of in-group ties) because they were not measured sequentially. Finally, the current research did not ensure CCICog and CCIAff mediate the effects of antecedents on outcomes such as customer loyalty. Though research supports such mediation (Bhattacharya & Sen, 2003), future research should evaluate the mediating effect of each CCI dimension. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jbusres.2017.05.010. References Ahearne, M., Bhattacharya, C. B., & Gruen, T. (2005). Antecedents and consequences of customer-company identification: Expanding the role of relationship marketing. Journal of Applied Psychology, 90(3), 574–585. Arsel, Z., & Stewart, S. (2015). Identity degrading brands. In S. Fournier, M. Breazeale, & J. Avery (Eds.), Strong brands, strong relationships. New York, NY: Routledge. Ashforth, B. E., Harrison, S. H., & Corley, K. G. (2008). Identification in organizations: An examination of four fundamental questions. Journal of Management, 34(3), 325–374. Ashforth, B. E., & Mael, F. (1989). Social identity theory and the organization. Academy of Management Review, 14(1), 20–39. Bagozzi, R. P., Bergami, M., Marzocchi, G. L., & Morandin, G. (2012). Customer– organization relationships: Development and test of a theory of extended identities. Journal of Applied Psychology, 97(1), 63–76. Batra, R., Ahuvia, A., & Bagozzi, R. P. (2012). Brand love. Journal of Marketing, 76(2), 1–16. Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117(3), 497–529. Bhattacharya, C. B., Rao, H., & Glynn, M. A. (1995). Understanding the bond of identification: An investigation of its correlates among art museum members. Journal of Marketing, 59(4), 46–57. Bhattacharya, C. B., & Sen, S. (2003). Consumer–company identification: A framework for understanding consumers' relationships with companies. Journal of Marketing, 67(2), 76–88. Cameron, J. E. (2004). A three-factor model of social identity. Self and Identity, 3(3), 239–262. Ciftcioglu, A. (2011). The relationship between perceived external prestige and turnover intention: An empirical investigation. Corporate Reputation Review, 13(4), 248–263. Cota, A. A., Evans, C. R., Dion, K. L., Kilik, L., & Longman, R. S. (1995). The structure of group cohesion. Personality and Social Psychology Bulletin, 21(6), 572–580. Crocker, J., & Wolfe, C. T. (2001). Contingencies of self-worth. Psychological Review, 108(3), 593–623. Dion, K. L. (2000). Group cohesion: From “field of forces” to multidimensional construct. Group Dynamics: Theory, Research, and Practice, 4(1), 7–26. Dukerich, J. M., Golden, B. R., & Shortell, S. M. (2002). Beauty is in the eye of the beholder: The impact of organizational identification, identity, and image on the

179