A multinational examination of the symbolic–instrumental framework of consumer–brand identification

Abstract

The authors propose a symbolic–instrumental interactive framework of consumer–brand identification (CBI) and explore its predictiveness across 15 countries. Using multinational data, they show that the negative impact of the misalignment between self–brand congruity and perceived quality on CBI is universal. The interaction among CBI, perceived quality, and uncertainty avoidance orientation in motivating consumers’ identity-sustaining behavior is weak. However, the synergy between CBI and perceived quality in motivating consumers’ identity-promoting behavior is stronger among collectivist consumers. The authors derive a typology of symbolic–instrumental misalignments to help international marketing managers motivate consumers to identify with and promote brands.

Appendices

APPENDIX A

Theoretical Foundations of the Symbolic–instrumental Framework

Economic roots

Economists generally view consumer choices as means to achieve maximization of functional utility (McFadden, 1986). Also, the common practice among marketing researchers is to model consumer brand choices based on product attributes and marketing mix (e.g., Guadagni & Little, 1983, and many subsequent extensions). However, according to the original multi-attribute utility theory (Lancaster, 1966), consumer utility includes not only a brand's functional attributes but also its sociopsychological attributes. Similarly, McFadden (1986: 284) posits that it is necessary to incorporate psychometric data in choice models, because these factors also shape the utility function. Surprisingly, only recently has research on choice models revived the need to incorporate softer, non-product-related attributes, such as customers’ attitudes and perceptions, into models of brand choice and brand switching (e.g., Ashok, Dillon, & Yuan, 2002; Erdem et al., 2006). This research, however, treats functional utility (corresponding to the instrumental driver in Katz's, 1960, conceptualization) and other utilities as additive rather than interactive. Note that Katz (1960) uses the word “functional” to refer to all the functions an attitude serves.

Consumer Psychology Roots

In psychology, Katz (1960) posits that an attitude can serve various functions. Two important functions that have received the most attention in marketing research are the symbolic and instrumental functions. However, our survey of the literature reveals a multitude of perspectives of how these drivers are linked together.

The linear, sequential perspective

Walker and Olson (1991) use the means–end chain goal to theorize that product knowledge affects self-knowledge linearly. In other words, they treat the self as the end in the means–end relationship. This perspective is also consistent with Young and Feigin’s (1975) “grey benefit chain,” which links product attributes to benefits and, ultimately, to emotional payoff. Sirgy and colleagues (1991) treat self–brand congruity (a symbolic construct) as an antecedent to functional congruity (an instrumental construct). From this perspective, self–brand congruity as a cognitive scheme at higher levels in the cognitive hierarchy is more accessible, and is likely to be processed before concrete schemes, such as functional congruity. Thus it follows that symbolic variables may bias perceptions of instrumental variables. Note, however, that Sirgy and colleagues find that the bias produced by top-down thinking is only moderate. Nevertheless, their perspective, which goes from abstract cognition to more specific cognition, is in contrast with Walker and Olson’s (1991) means–end chain model.

The independent, dual-drivers perspective

The independent, dual-drivers perspective originates from the influential work of Gardner and Levy (1955). Through qualitative research, they demonstrate (p. 39) that “products and brands have interwoven sets of characteristics and are complexly evaluated by consumers.” In Keller's (2008) customer-based brand equity pyramid, brand performance and brand imagery are two key components of brand meanings. Similarly, Mittal (2006) contends that possessions become the extended self in two ways: (1) they become instrumental because of their functional benefits; and (2) they become identity implementing, because they enhance self-expression through self–brand congruity. In line with this dual-driver perspective, Brown and Dacin (1997) show that the association between corporate ability (i.e., capability for producing products) and corporate social responsibility can have different effects on consumer responses to products. Homburg, Wieseke, and Hoyer (2009) propose the social identity perspective as an alternative to the traditional satisfaction-based service–profit chain, but do not examine the interaction between customer satisfaction and customer–company identification. Several empirical studies based on social identity theory have also hinted at the differential effects of the instrumental component (e.g., customer satisfaction with functional attributes) and the symbolic component (e.g., company prestige) on customer identification with companies and nonprofit organizations (Ahearne et al., 2005; Arnett et al., 2003; Bhattacharya, Rao, & Glynn, 1995; Donavan et al., 2006; Kuenzel & Halliday, 2008). Other research has also treated the symbolic and instrumental components of people's attitude and behavior as drivers that function independently (e.g., Lievens & Highhouse, 2003).

Conceptually, the dual-drivers perspective is consistent with hygiene theory (Herzberg, 1966), which contends that individual needs can be classified into two independent categories: (1) basic or hygiene needs, and (2) growth or higher-order needs. In other words, factors that drive dissatisfaction are not the same as satisfiers. From the means–end model we have just discussed, we can infer that instrumental drivers are lower-order hygienes, and symbolic drivers are higher-order motivators. In fact, Agustin and Singh (2005) view transactional satisfaction (which is highly correlated with perceived quality) as a lower-order need. If the instrumental and symbolic drivers are processed independently, the prediction is that the interaction between them in driving consumer brand attitude and brand-related behavior will not be statistically significant. Herzberg's theory is not without controversy (Bockman, 1971). In Lindsay, Marks, and Gorlow’s (1967) survey of the literature, they found that empirical research that relies on the theory produced mixed results. They revised Herzberg's theory, and found a positive interaction between hygiene factors and growth motivators in predicting job satisfaction.

The symbolic–instrumental interactive perspective that we propose reconciles the existing perspectives. First, it takes into consideration both symbolic and instrumental drivers of consumer–brand relationship correlates. Second, it allows for the linear sequential flowing from functional means (e.g., perceived quality) to symbolic ends (e.g., CBI). Third, it complements Sirgy and colleagues’ (1991) perspective by hypothesizing that the bottom-up process that Walker and Olson (1991) propose is as important as the top-down bias that self–brand congruity exerts on functional congruity. In other words, our interactive perspective conceptualizes the bias that Sirgy and colleagues (1991) mentioned as a mutually compensatory process between instrumental and symbolic drivers.

APPENDIX B

Table B1

Table B1 Sample description

APPENDIX C

Table C1

Table tC1 Construct measures

APPENDIX D

Analytical Notes

Exploratory factor analysis and measurement invariance

Consistent with Van de Vijver and Leung’s (1997) work, the preliminary analyses began with an exploratory factor analysis to explore across countries, which included the item characteristics, item-to-total correlations across the measuring items of each construct across countries, and factor loading pattern of all items across all constructs. After we dropped five cultural orientation items (see Appendix C), the final set of scale items yielded the hypothesized factor structure across countries, and none of the items cross-loaded heavily onto other unintended factors. Therefore we proceeded with multigroup confirmatory factor analyses to test measurement invariance before testing the structural model (Steenkamp & Baumgartner, 1998). Because PLS estimation does not provide a fit index for the models that allows for testing measurement invariance, and the structural equation modeling estimation does not allow for a measurement invariance test of formative constructs, we specified CBI as a second-order reflective construct (instead of a second-order formative construct) and used the fit index of the full structural equation modeling as a proxy. The configural invariance model in which all loadings were freely estimated yielded good fit (χ²(4590)=15,411.25, comparative fit index [CFI]=0.918, Tucker–Lewis index [TLI]=0.906, root mean square error of approximation [RMSEA]=0.02, Akaike information criterion [AIC]=17,571.25, and Browne–Cudec criterion [BCC]=17,759.08). In multiple-group analyses, the BCC imposes a slightly greater penalty than AIC for model complexity. Because the Bayes information criterion is normally reported only for a single-group analysis, we do not report it here. All factor loadings were highly significant across all 15 countries, and most of the standardized factor loadings exceeded 0.60. The full metric invariance model in which all the factor loadings were constrained to be invariant across countries also yielded good fit (χ²(4856)=16,175.91, CFI=0.915, TLI=0.907, RMSEA=0.02, AIC=17803.91, and BCC=17,945.48). Although the increase in chi-square is significant (Δχ²(266)=764.66, p<0.00), the alternative fit indexes showed only minimal changes. Steenkamp and Baumgartner (1998) recommend that, in comparing model fit, researchers should not rely exclusively on the chi-square difference test, because of its sensitivity to sample size, and that other fit indexes such as CFI, TLI, and AIC should also be used. Using this heuristic, we concluded that cross-national invariance was supported. Then, we specified a main-effects-only structural model in PLS to extract latent scores for each country. We create interactions of these latent scores to test the model with interactions (Chin, 1998).

Discriminant validity

Table 1 shows that the square of the zero-order correlation between any two constructs is smaller than the average variance extracted by the measurement items of the corresponding constructs, demonstrating discriminant validity (Fornell & Larcker, 1981). We also tested the discriminant validity of the focal constructs. Because there is no procedure for testing discriminant validity between a reflective construct and a formative construct, we model CBI as a reflective construct, and use the result of that model as a proxy for discriminant validity. The comparison of the AVE and the square of pairwise correlation suggests that discriminant validity was established. The models in which the correlations between the latent constructs were constrained to unity also yielded significantly worse fits, providing evidence of discriminant validity (for the focal pair of constructs, CBI-perceived quality, the change in model fit is Δχ²(1)=48.5, p<0.01).

Common method biases

The conceptual framework of this study has two formative constructs (CBI, self–brand incongruity), and is focused primarily on interaction effects. Prior research has shown that, statistically, interaction effects cannot be artificially created by common method biases (Siemsen, Roth, & Oliveira, 2010). Conceptually, consumers who fill out the survey are much less likely to be able to guess the interaction effects that we test. Podsakoff et al. (2003: 900) recommend that “when formative-indicator constructs are an integral part of a study, researchers must be even more careful than normal in designing their research because procedural controls are likely to be the most effective way to control common measurement biases.” Consistent with this recommendation, we designed our survey such that it included a variety of question types and response formats, including Venn diagram and Likert scales of different lengths; the self–brand incongruity score is a Euclidean score calculated from a number of questions. The items of the focal construct of CBI appeared in three different sections in the survey, some of which appeared after we asked consumers about the dependent variables. These procedural remedies help with reducing common method biases (Podsakoff et al., 2003).

APPENDIX E

Table E1

Table E1 Factor loadings and path weights for focal constructs