Introduction

Increased coverage of negative business events in mainstream and social media has caused firms to become more sensitive regarding any negative information directed at their companies (Einwiller et al. 2006). When such negative information is spread, companies rely on the relationship they have with their customers for visible support in the market place (Einwiller et al. 2006; Ahluwalia et al. 2000). Firms that have built strong brand identification with their customers can count on active support of their brands (Tuškej et al. 2013; Bhattacharya and Sen 2003). Brand identification, the extent to which consumers see their own identity or self-image as matching with the brand’s image (Bagozzi and Dholakia 2006), leads consumers to ignore the negative information, and consumers’ (positive) attitude toward the brand is unaffected by the event (Lisjak et al. 2012; Liu et al. 2010). However, these findings prove valid only when the unethical situation is at a low to moderate level. In an extremely negative unethical situation, positive impact of brand identification tends to disappear because consumers can no longer ignore such negative information (Bhattacharya and Sen 2003; Einwiller et al. 2006; Liu et al. 2010). In this research, we investigate the impact of strong brand identification that can lead to support even in extreme unethical situations.

For example, consider this hypothetical scenario; A consumer learns that her favorite brand she strongly identifies with has engaged in extremely unethical practices such as taking advantage of child labor, bribing government officials, or even both. Extant research as discussed above indicates that her relationship with the brand will not be enough to prompt support. However, prior literature has not considered the impact of the passionate love she might have for the brand. This consideration is theoretically and managerially important as brand love is the strongest relationship consumers could form with a brand (Huber et al. 2015). It is a good predictor of vital managerial variables (Bagozzi et al. 2017), and emotions have been found to play a key role in situations involving morality (Ditto et al. 2009).

In this study, we examine the role of identification and love in situations when a brand acted extremely unethically and we extend the literature in two specific ways. First, we introduce the concept of brand love, an emerging marketing construct (Langner et al. 2016), to the business ethics literature. Brand love is defined as the passionate affection consumers have toward a brand (Carroll and Ahuvia 2006), extending beyond brand attachment (Thomson et al. 2005). Our results indicate that identified consumers report the situation highly immoral (direct effect). However, increased levels of identification also lead to greater brand love. These consumers who are in love with the brand find the situation less immoral. Therefore, this indirect effect leads to a positive impact of high brand identification on ethical judgment (total effect). In other words, contrary to existing research, brand identification has a beneficial impact even during extremely unethical situations but only if the brand has created a strong, passionate, love type relationship with consumers. These results are both theoretically and managerially intriguing for companies as they invest in their brands seeking to enjoy relational benefits with their customers (Lam et al. 2013). However, unless these relationships develop into a passionate emotional love, their customers will not only reject the company but in fact they will punish it even more than other customers when the company is involved in an extremely unethical situation.

Second, we extend the word of mouth (WOM) literature by introducing two new constructs especially suitable for consumers’ supportive behaviors, namely sin of omission and brand defense. Our results indicate that consumers who love their brands support them by either purposefully not talking about the negative situations (i.e., sin of omission) or by actively defending the brands to other consumers (i.e., brand defense). We will describe these constructs in more detail.

In the next section, we discuss theoretical foundations and hypotheses and illustrate their relationship in the conceptual model. Later, we outline the methodology and present our results from the covariance-based structural equation modeling. Finally, the paper concludes with a discussion of findings and their implications both for increasing knowledge in the marketing literature and assisting practitioners actively engaged in improving consumer–brand relationships.

Theory and Hypotheses

Brand Identification

Brand identification is defined as the extent to which consumers see their own identity or self-image as matching the brand’s image (Bagozzi and Dholakia 2006). Over time, this self-brand matching process stimulates consumers’ epistemological evolution, both consciously and unconsciously. The result is a personalized brand relationship that consumers selectively choose to maintain (Currás-Pérez et al. 2009). Consumers benefit from this relationship by expressing their self-image to others through their possessed brands. Likewise, companies benefit from increased consumer loyalty, ability to charge premium prices, and consumers’ higher positive (lower negative) WOM intentions (Wolter et al. 2016; Wolter and Cronin 2016).

In the extant literature on ethics, it has been found that if such identification is sufficiently strong, consumers tend to “look the other way” when the brand is caught in unethical situations in order to defend their own identities (Einwiller et al. 2006). However, this effect is found only under moderately unethical situations and not during extremely unethical situations (Bhattacharya and Sen 2003; Einwiller et al. 2006; Liu et al. 2010; Schmalz and Orth 2012). This is apparently because consumers reach a point where they can no longer justify the brand’s actions (Einwiller et al. 2006; Liu et al. 2010). For example, in a retailing context, Liu et al. (2010) show that consumers seem to easily rationalize when the retailer brand they identify with does not apply the same promotion to all items in the store even though the company had communicated that promise (a moderately unethical situation). However, they find it very hard to excuse the same company when it sells expired products that potentially could cause health problems (an extremely unethical situation).

Ethical (mis)conduct, such as the examples provided, involves personal judgment and the weighing of relevant facts which lead to an assessment of the company’s conduct. This, in turn, will affect behavioral intentions (Hunt and Vitell 1986, 2006). Ethical judgment is defined as “an individual’s personal evaluation of the degree to which some behavior or course of action is ethical or unethical” (Sparks and Pan 2010, p. 409). This implies that ethical judgment is specific to a given situation (Barnett and Karson 1987). In marketing, it is shown that consumers evaluate companies’ actions not in isolation, but also by taking into account the overall actions along with prior perceptions of that brand (Brunk 2012). Therefore, it is important to understand how consumers’ judgments are shaped given their personal relationship with a brand.

Strongly identified consumers consider themselves to be the “same as the brand” (Chaplin and Roedder John 2005; Dutton et al. 1994) and utilize the brand to communicate their own identity to others (Underwood et al. 2001). Therefore, they feel personally threatened when the brand is caught in an unethical situation (Lisjak et al. 2012). This is consistent with the literature on moral judgments, arguing that people inherently would like to see themselves as ethical beings (Mazar et al. 2008), holding to high moral standards (Aquino and Reed 2002). In moderately unethical situations, consumers use motivated reasoning to justify brand’s actions and protect their own identity (Einwiller et al. 2006). When the situation falls into the “extreme” domain, however, consumers are unable to bring themselves to excuse the company’s behavior as the situation is too egregious to ignore (Einwiller et al. 2006). Similarly, Bhattacharya and Sen (2003) also note that identification should shield the company from the detrimental effects of negative information as long as the situation is “within the zone of tolerance” (p. 84). Beyond this zone of tolerance, identified consumers are expected to react more strongly. In a moral domain, when the situation is too extreme from consumers’ expectations, the abhorrent behaviors prompt adverse reactions (Bandura 1991). In these cases, consumers are likely to distance themselves from the transgression as much as possible in order to show their disapproval (Bandura 1991, 1986).

Based on this reasoning, we argue that in extremely unethical situations where strongly identified consumers can no longer find reasons to support the companies’ actions, they will judge the companies’ actions more negatively (i.e., less ethical) than consumers who do not possess similar identification to the company. Therefore, we hypothesize;

H 1

Brand identification has a direct negative effect on ethical judgments under extremely unethical situations, such that the brand is evaluated as more unethical.

Brand Love

As described in the introduction section, brand love is an emerging concept in consumer brand relationships (Bagozzi et al. 2017; Hegner et al. 2017; Langner et al. 2016; Sarkar 2014) and has been shown to be a distinct construct (e.g., Barker et al. 2015). Consumers are able to form “love” relationships with brands in different product categories (Fetscherin et al. 2014) as well as in service sectors (Long-Tolbert and Gammoh 2012). Consequently, this love leads to positive reactions from consumers such as active engagement (Bergkvist and Bech-Larsen 2010; Sarkar 2014), commitment (Albert et al. 2013), loyalty (Batra et al. 2012; Carroll and Ahuvia 2006), positive word of mouth (Albert et al. 2013; Batra et al. 2012;), and willingness to pay a price premium (Bauer et al. 2009).

The central idea of the positive impact of brand identification on brand love is the intensity of the emotional bond. According to a recent study by Huber et al. (2015), when a brand reflects one’s inner self, the consumer feels some sense of comfort, which then triggers the emotional response toward the brand. In addition, Bergkvist and Bech-Larsen (2010) have also shown that brand identification has a positive effect on brand love. Taken together with prior studies of consumer brand relationships, the brand literature supports the argument that brand love is the outcome of the integration of a consumer’s identity and the brand (Carroll and Ahuvia 2006; Batra et al. 2012). Therefore, we hypothesize the following:

H 2

Brand identification has a direct positive effect on brand love under extremely unethical situations, such that the stronger the brand identification, the stronger is the brand love.

Emotions have been shown to be the key drivers of moral behavior (Greene and Haidt 2002). Specifically, ethical judgments are influenced by emotions (Ditto et al. 2009), and brand love is a very powerful emotion (Carroll and Ahuvia 2006). In fact, it is the strongest emotion a consumer can have for a brand (Huber et al. 2015). In several aspects, it is analogous to interpersonal love (Carroll and Ahuvia 2006; Sarkar et al. 2012). When we are in a close, loving relationship with someone (e.g., our child, or spouse), we are partial toward these people (Velleman 1999). This partiality is also extended into the domain of morality (Cottingham 1986). More specifically, one’s moral judgment of the situation about a person is often affected by the love one has for that person. For example, assume that your child gets caught cheating in school. While you are fully aware of the unethical nature of the situation, you are likely to be less harsh in your ethical judgment by suggesting extraordinary circumstances that your child might have experienced that contributed to the poor decision. However, others with no relationship to the child focus on the unethical nature of the situation more than the special factors. This illustration parallels decision-making explanations of deontologists who employ different standards of normative principles (Kleiser et al. 2003; Robin and Reidenbach 1987). (We discuss the inherently unethical nature of sin of omission and brand defense below.)

While some marketing scholars criticize this type of analogy of interpersonal love (e.g., Bengtsson 2003), it is still argued to be acceptable to use interpersonal love theory for theoretical arguments (Batra et al. 2012; Sarkar 2014). Therefore, brand love that generates strong emotion toward the brand is likely to positively impact ethical judgment even under extremely unethical situations. Based on the discussion above, we hypothesize:

H 3

Brand love has a direct positive effect on ethical judgments under extremely unethical situations, such that the brand is evaluated as more ethical.

Brand Supportive Behaviors: Brand Defense and Sin of Omission

When information about a company’s unethical activities becomes public, it is important for the company to have loyal customers who continue to talk positively and support the brand (Einwiller et al. 2006). This interpersonal communication, or word of mouth (WOM) (Westbrook 1987), has a measurable effect on the behavior of the recipients of this communication (Berger 2014) and plays an important role in promoting a brand’s image (Sweeney et al. 2008). From this perspective, it is also strongly linked to a company’s success (East et al. 2007) since consumers find this spontaneous and voluntary communication to be more reliable than company-produced materials (Trusov et al. 2009).

The valence of WOM is varied and generally categorized as positive, negative, and neutral (Anderson 1998). In every situation, nevertheless, emotions have been one of the main drivers of generating WOM (de Matos and Rossi 2008). In an extremely negative context, if consumers still choose to support the brand, they could do so either actively by defending the brand to others (brand defense) or passively by purposefully not bringing up the issue to others (sin of omission).

One of the most important motivators for positive WOM is a strong emotional relationship with the brand (Dick and Basu 1994). As discussed above, brand love is defined as the passionate emotion a consumer has toward a brand, and it is one of the strongest emotional relationships. More importantly, passion is recognized as intense, positive feelings (Thomson et al. 2005), and this has been shown to impact WOM positively (Albert et al. 2013). Recent research in consumer brand relationship literature identified brand defense as a distinct construct which is conceptualized as the extreme positive WOM due to the love consumers have for their brands (Javed et al. 2015). The question remains, however, as to whether the passion consumers have for their brand will lead to brand defense even under extremely unethical conditions. Literatures on interpersonal love as well as brand love offer convincing explanations that it will.

According to the triangular theory of (interpersonal) love (Sternberg 1986), the passionate component of love, in particular, causes an idealized view of the loved one. This idealization is related to preserving the other person’s well-being (Rempel and Burris 2005). While we are not arguing that the love for a brand is completely comparable or unconditional (Batra et al. 2012), we believe that these intense emotions will elicit responses similar to what one might have for close loved ones. In fact, consumers could sometimes be more interested in brands than their human loved ones (McEwen 2005). Also, brand love is important in accepting the brand and advocating it to others (Wallace et al. 2014). Based on the discussion above, we hypothesize;

H 4

Brand love has a direct positive effect on brand defense under extremely unethical situations, such that consumers engage more in brand defense.

Another strategy we propose that consumers who love their brands can employ is sin of omission. We define the sin of omission as motivated inaction in which consumers actively choose not to bring up or completely ignore (when it is mentioned) the unethical issues linked with the brand for the purpose of protecting the brand. Unlike brand defense, however, the sin of omission can be implemented with minimal effort because consumers can be quickly freed from the uneasiness of supporting the brand by simply ignoring or not bringing up the details of the ethical failure to others. For example, if they are in the presence of others who are unaware of the negative information, consumers who love their brand could purposefully choose not to bring it up in conversation. Or, they could pretend they did not know the information if it is exposed. No matter what the reason, we argue that brand love could make people engage in this type of behavior. In other words, they could support the brand they love with intentional inaction. We are arguing that the sin of omission does not correspond to low levels of negative WOM, but that it is a distinct construct. In the literature on negotiation, this type of behavior is defined as “concealing the information on purpose” (Jensen et al. 2011). Therefore, we hypothesize:

H 5

Brand love has a direct positive effect on sin of omission under extremely unethical situations, such that consumers engage more in sin of omission.

In business ethics literature, ethical judgment is shown to be an important factor explaining consumers’ behavioral intentions (Chiu 2003; Hunt and Vitell 1986; Rest 1986). In marketing contexts, too, the effect of ethical judgment on behavioral intention, specifically WOM, has been well studied in the areas of consumers’ commitment (Ingram et al. 2005) and their online expertise (Román and Cuestas 2008). Not surprisingly, this area of research indicates the more positive the judgment, the more likely consumers are to support the brand (i.e., positive WOM). In our research context, supporting the brand would mean increased levels of brand defense and sin of omission. Therefore, we hypothesize;

H 6

Ethical judgment has a direct positive effect on brand defense under extremely unethical situations, such that consumers engage more in brand defense.

H 7

Ethical judgment has a direct positive effect on sin of omission under extremely unethical situations, such that consumers engage more in sin of omission.

Research Design

Data Collection

A pretest and a main study were conducted using Mechanical Turk (mTurk) samples collected from 2015 to 2016. The pretest (n = 59) had two objectives. First, the purpose was to establish a scenario to communicate an extremely unethical condition and second, to confirm the research context that consumers perceived brand defense and the sin of omission unethical. The main data collection (n = 403) was used for the data analysis involving the measurement model test and structural model evaluation.

The mTurk was used to recruit participants for this study. The participants from mTurk were deemed to offer substantial variation of focal constructs that was better than that of a student sample. Although interest and popularity of mTurk continue to grow, researchers using mTurk are beginning to give careful consideration for the quality of data. For this study, we ensured quality of the data by recruiting only participants who obtained good reputation scores mTurk regularly assesses and announces to mTurk buyers, and by monitoring the duration of survey completion time offered by the online survey program. Results showed that average participants spent between three and four minutes to complete all 16 individual questions tied with constructs as well as other questions presented (e.g., survey instruction, control variables, scenarios, and basic demographics). After careful examination of the data collected, a total of 399 responses were used to test the measurement and theoretical models.

Scenario and Administration Procedure

In the pretest, we varied ethical misconduct scenarios between subjects in two ways. In one version, a brand bribed local officials in an Asian country where its factory was located. In the other version, a brand was using child labor and bribing the local officials to cover it up in an Asian country where its factory was located. We did not use a specific brand in the scenario in order to avoid contamination of the results with brand-related emotions. In both versions, before showing them the scenario, we first asked participants their ethical judgments of people who (a) commit a sin of omission and (b) defend the brand after knowing the brand engaged in ethical misconduct (4-item scales). By not mentioning the specificity of the misconduct, we tried to measure a general understanding of people’s attitude toward our main dependent variables. After responding to those questions, participants saw one of the scenarios mentioned above and indicated their judgment of the situation using the same 4-item scale.

In the main study, participants first completed the scale items measuring their identification with, and love for, the Apple brand. This allowed us to measure these two constructs before we introduced the scenario describing ethical violations. After responding to the scale items, participants were presented with a scenario explaining that they read in a trusted local newspaper an article that described Apple employing child labor in one of its Asian factories and bribing the local officers to cover up this story. We chose Apple brand as we thought it would create variation on the brand love construct because Apple could generate both very positive and negative emotions for consumers (Hutcheon 2014). Brand love is conceptualized as ranging from nonexistent to very intense (e.g., Zarantonello et al. 2016). While every brand could theoretically generate love (Fetscherin et al. 2014), in reality, not all brands achieve the level of very intense (Bagozzi et al. 2017; Zarantonello et al. 2016). Fetscherin et al. (2014), therefore, suggest researchers studying the brand love construct should employ brands in their studies that could, realistically, generate these emotions (Fetscherin et al. 2014). Moreover, both Bagozzi et al. (2017) and Bergkvist and Bech-Larsen (2010) show that the impact of brand love (on managerial variables) is captured best when there is a high variance on the brand love construct. Therefore, it is important in our research context to use a brand that could generate both very weak and very intense emotions.

After reading the scenario, participants responded to scale items on negative WOM, brand defense, and sin of omission, followed by ethical judgment questions. We asked negative WOM questions to establish discriminant validity between negative WOM and sin of omission (discussed below). Participants then responded to manipulation check items on newspaper credibility and their affection for the newspaper along with a few additional questions used to assess common method bias and to perform endogeneity checks. Lastly, the survey included some questions regarding demographic information.

Measures

A careful review of prior studies and industry literature was conducted to select appropriate survey instruments with proven psychometric properties. We made only minor modifications to the existing scale items. Only when necessary did we develop new scale items to measure past behavior and decisions. Brand identification refers to the degree that a consumer identifies herself with the brand; this was evaluated using five scale items modified from Einwiller et al. (2006). Brand love refers to the passionate affection a consumer has for the brand; seven scale items were modified from Fetscherin et al. (2014). Ethical judgment refers to the moral evaluation of the situation by the consumer; four scale items were modified from Reidenbach and Robin (1990). The WOM-related dependent variables refer to actively choosing not to talk about a situation and defending the brand to other consumers, measuring sin of omission, and brand defense constructs, respectively. We developed three items for each of these WOM-related constructs based on our conceptualization.

Based on the scale development guidelines by Anderson and Gerbing (1988), and common procedural recommendations by Baumgartner and Homburg (1996) and Iacobucci (2009), a two-step procedure was applied to assess the reliability and validity of the constructs in this research. Prior to the data analysis, we inspected all responses to check for outliers and missing data. Out of the 403 surveys we obtained, only 4 surveys were completely removed from further data analysis because more than 50% of the responses were blank. For the surveys with a small number of missing data, we used the means substitutions method offered by IBM SPSS 20.0. In deciding which scale items should be included in each construct, we first conducted exploratory factor analysis (Anderson and Gerbing 1988; Shook et al. 2004). Results showed that while the total number of original scale items for both sin of omission and brand defense remained the same, constructs involving brand identification (reduced from 5 to 3 items), brand love (reduced from 7 to 4 items), and ethical judgment (reduced from 4 to 3 items) were intentionally modified from their original instruments because of low factor loadings (below .7) and/or failure to exceed a cutoff value of Cronbach’s alpha (less than .6). Final scale items for each construct are shown in Table 1. In addition, the correlation matrix of all the items, along with item means and standard deviations, is summarized in Table 2.

Table 1 Scale items and sources
Full size table
Table 2 Correlation matrix of variables
Full size table

Pretest Results

In the pretest, we first analyzed whether participants actually considered sin of omission and brand defense as unethical behaviors. We ran two one-sample t tests (N = 59) on the composite ethical judgment scales for sin of omission and brand defense (Cronbach’s alphas: .94 and .93, respectively) against the scale midpoint (4 = neither unethical/ethical). It is important to note that people are fundamentally motivated to punish perpetrators of moral violations (Pizarro and Helzer 2010; Pizarro and Tannenbaum 2011). Therefore, supporting the brand under those conditions is not considered morally acceptable behavior. Results revealed that participants judged both sin of omission (M: 2.84, t = − 6.5, p < .01) and brand defense (M: 2.72, t = − 7.6, p < .01) to be unethical, as predicted.

We then analyzed the judgments of the two different scenarios. Our dependent variable was the same 4-item ethical judgment scale, this time for the scenarios (Cronbach’s alpha: .96) and the independent variable was the specific version of the scenario (Moderate vs. Extreme Scenario, Between Subjects). ANOVA indicated a significant main effect of version [F(1,58) = 7.9, p < .01; MExtreme = 1.68 vs. MModerate = 2.72]. We observed that even the “moderate” scenario is well below the scale midpoint (t = − 4.52, p < .01). Our main objective was to create an even more unethical scenario in order to have a stronger test of our model. Moreover, our extreme scenario involves two acts of unethical behavior, harm done to others (i.e., child labor violations) and the intentional cover up by the perpetrator (i.e., bribing officials). Therefore, based on these results, we have chosen to use the scenario that included both child labor and bribing officials to represent an extremely unethical situation.

Results of Construct Validation

The confirmatory factor analysis with a total of 399 data records was used to estimate the parameters of the measurement model. Table 3 presents the results of overall goodness-of-fit estimates and results of convergent and discriminant validity. Although the Chi-square statistic for the measurement model was disappointing and significant (χ2 = 225.09, df = 94), other comparative fit indices met and exceed the cutoff value (Bentler 1993). The results suggested that the factor structures that we hypothesized based on theory successfully reproduced the observed correlations of data that we had collected (CFI = .98; GFI = .94; NFI = .97; TLI = .98; RMSEA = .06).

Table 3 Parameters for measurement model
Full size table

To assess the convergent validity, we followed the guidelines developed by Anderson and Gerbing (1988). As shown in Table 3, all scale items loaded on their constructs as we hypothesized. All factor loadings ranged from a low of .836 to a high of .975, and the estimates were all positive and significant. Thus, these results provided evidence of convergent validity (Anderson and Gerbing 1988; Iacobucci 2009).

Discriminant validity was assessed by comparing the difference between the average variance extracted (AVE) from scale items and the shared variance among other constructs (Anderson and Gerbing 1988; Bagozzi and Yi 1988). For example, the average variance extracted (AVE) for brand identification was 85.29%, which was higher than the highest shared variance among other constructs (i.e., a low of 6% and a high of 63%). Results showed that the AVE values from other constructs (brand love = 85.31%; ethical judgment = 88.08%; sin of omission = 81.94%; brand defense = 90.74) were all higher than the shared variance (highest shared variance = 34, 48, 44, and 44%, respectively). Thus, we concluded that the measurement model had passed the discriminant validity test. Taken together, results confirmed that the measurement model was acceptable to test our proposed structural model and substantive hypotheses described early.

While prior research had conceptualized and empirically shown that brand defense is a distinct construct indicating extremely positive WOM (Javed et al. 2015), the new construct sin of omission could be argued that it is, in fact, similar to low levels of negative WOM. To test this idea, we conducted a separate discriminant validity test between sin of omission and negative WOM. As shown in “Appendix 1”, results indicate that sin of omission is not the same as the negative WOM construct. In addition, we compared our proposed model (without negative WOM) and the rival model (including negative WOM) to see which one produces better goodness of model fit indices. The results in “Appendix 2” show that the Chi-square difference between our proposed model and the rival model is statistically significant (χ2 = 264; df = 46; p < 0.001). All fit indices of our proposed model are superior to that of the rival model (CFI = .973 vs. .952; GFI = .914 vs. .867; NFI = .962 vs. .938; TLI = .967 vs. .943 l).

Hypotheses Tests and Results

Before testing the empirical model, we analyzed our manipulation check about the source (i.e., newspaper) credibility. We tested the 7-point (“not credible”/“credible;” “not believable”/“believable”), two item composite scale (Cronbach’s alpha: .94) against the scale midpoint. A one-sample t test indicated that participants found the newspaper article credible (M = 5.17, t = 16.2, p < .01); therefore, our manipulation of the newspaper article was effective. To test the theoretical model depicted in Fig. 1, structural equation model using IBM SPSS AMOS 2.0 was used.

Fig. 1
figure 1

Empirical model and hypotheses

Full size image

Table 4 presents overall goodness-of-fit estimates, t value, and summary of hypothesis results. First, the Chi-square statistic for the structural model is significant (χ2 = 312.097, df = 97). In large samples the Chi-square statistic is not reliable; thus, we investigated other model fit indices (Bollen 1989). All comparative fit indices including CFI (.97), GFI (.91), NFI (.96), and TLI (.97) were higher than Bentler’s (1993) cutoff value (i.e., higher than .90). In addition, RMSEA was .08, which suggested that the structural model provides a good fit to the data. Thus, we concluded that results from the structural equation model supported turning our attention to the statistical estimates of the hypothesized structural paths shown in Fig. 1.

Table 4 Parameter estimates for structural model
Full size table

Tests of Hypotheses

As we predicted in H1, results showed that the effect of brand identification on ethical judgment was negative and statistically significant (\(\gamma_{31} = - \,.262,t = - \,2.699)\). Therefore, H1 was supported. Next, we tested H2 that brand identification had a positive effect on brand love. As we hypothesized, brand identification had a statistically significant impact on brand love \(\left( {\gamma_{21} = .842,t = 22.141} \right)\). Therefore, H2 was also supported. As we predicted, brand love had a significant positive impact on ethical judgment \(\left( {\beta_{32} = .618,t = 6.29} \right)\) supporting H3. As these results show, brand identification has both negative (direct) and positive (indirect through brand love) effects on consumers’ evaluation of the event. To understand the overall impact of identification on judgment, we also looked into the total effect. As shown in Table 5, the total effect of the brand identification on the ethical judgment is positive (.259), where the indirect effect of the path brand identification to brand love to ethical judgment (.521) is almost two times higher than the direct effect of the path brand identification to ethical judgment (− .262).

Table 5 Direct, indirect, and total effects of parameter estimates
Full size table

To analyze the impact of brand love on brand supportive behaviors, we first tested the relationship between brand love and brand defense (H4). The coefficient of the path was positive and statistically significant \(\left( {\beta_{52} = .399,t = 10.297} \right)\) as we predicted in H4. In H5 we also estimated the relationship between brand love and sin of omission. The result was positive and statistically significant \(\left( {\beta_{42} = .226,t = 4.779} \right)\) and thus supported our argumentation shown in H5. Finally, we tested the effects of ethical judgment on both of the brand supportive behaviors involving brand defense (H5) and sin of omission (H6). Results supported both H5\(\left( {\beta_{53} = .516, t = 13.0687} \right)\) and H6\(\left( {\beta_{43} = .480,t = 9.603} \right)\) as hypothesized.

Common Method Variance Test

Common method variance may have inflated the strength of the path coefficients that we have estimated. We assessed the potential impact of the common method bias by employing the marker variable technique (Lindell and Whitney 2001). One of the additional questions we had asked, “Participants’ interest in the scenario,” was used as the marker variable (rwith brand identification = .204; rwith brand love = .186; rwith ethical judgment = .098; rwith sin of omission = .053; rwith brand defense = .042), which showed the lowest correlation score with all constructs used in this study. Results showed that when the portion of the variances in the model was controlled by the marker variable (rs = .042), the overall pattern of the relationships did not change, and all adjusted path coefficients were also significant (lowest t value among constructs in model = 4.746; p < 0.01). Thus, we concluded that the common method variance bias appeared to have no significant effect on the results displayed in the test of hypotheses section above.

Endogeneity Test

Although our results reflected our predictions well, we additionally tested our model more rigorously by employing an endogeneity test. We followed the widely used procedure for potential endogeneity problem using instrumental variable (IV) approach. First, based on prior research, we selected a list of (IVs) using the Wooldridge (2012) guideline specifying two important criteria in selecting instrumental variables: (1) IVs should not be correlated with the error term (Corr (X, e) equal to zero), and (2) IVs should be highly correlated with independent variables in the model. [Corr (X, IV) is not equal to zero.] In this study, we selected the brand reputation as an instrumental variable for the brand love (Kuenzel and Halliday 2010) and the acceptability for the ethical judgment (Cojuharenco and Squera 2015) (Table 6).

Table 6 Selection of instrumental variables (IV)
Full size table

Second, we tested to see if brand love and ethical judgment were endogenous variables when they were regressed on sin of omission and brand defense separately. Results from the Durbin test (χ2 score = 24.448 (df = 2); p < 0.001) and Wu–Hausman test [F(2, 393) = 12.826; p < .001] revealed that the null hypothesis stating that variables are exogenous was rejected. We also found that brand defense is endogenous variable [Durbin test χ2 = 36.963 (df = 2), p < 001]; Wu–Hausman test [F(2,393) = 20.062, p < 0.001]. Third, as follow-up according to the endogeneity literature (Bollen 1996, 2012; Burgess et al. 2014; Heller et al. 2009; Jedidi et al. 1997), we further tested the strength of the brand reputation. Results from Adjusted R-Square and the 2SLS relative bias showed that both instrumental variables have F values greater than 10 and higher than the 95% confidence level of 2SLS relative bias test. Fourth, we estimated the over-identification problem due to the addition of instrumental variables. Our expectation from this test is that parameter estimates would be sufficiently stable even after adding instrumental variables. Results indicated that the hypothesis that over-identification restriction is valid was not rejected, when the model was regressed on the sin of omission (Sargan χ2 = 1.534 (p = .216); Basmann χ2 test = 1.521 (p = .218)). Likewise, when the model was regressed on the brand defense, (H0: over-identification restriction is valid) the model was also not rejected (Sargan χ2 = 1.819 (p = .178); Basmann χ2 test = 1.804 (p = .179)).

Finally, we compared the parameter estimates between the model without instrumental variables against the model with the two instrumental variables discussed above. Results showed that similar to the findings from SEM, the impact of brand love and ethical judgment on the sin of omission and the brand defense was positive while that of brand identification was negative. This result is not consistent with the finding from the SEM where the brand identification has positive impact on brand love. We believe that the IV (instrumental variable) model based on the OLS estimate might have failed to capture the mediation effect of brand love on the ethical judgment because of the limitation of the linear model against the SEM model (Min and Mishra 2010; Wooldridge 2012) (Table 7).

Table 7 Results of unstandardized coefficient, testing endogenous variables, strengths of IVs, and over-identification
Full size table

Discussion

This study focuses on consumer behaviors resulting from extremely unethical situations in which their relationship with the brand impacts their supportive behavior for the brand in the marketplace. Drawing on various theories and perspectives from the ethics literature, identity theory, epistemological perspective, triangular theory of love, and brand management literature, we first explain how brand identification affects the moral evaluation of the event (i.e., ethical judgment) both negatively (direct effect) and positively (indirect effect through brand love). Our proposed model then shows how consumers support the brand even in extreme unethical situations by either actively defending the brand to other consumers (brand defense) or by deliberately not talking about the situation (sin of omission), even though these support mechanisms themselves are considered unethical.

From this perspective, we suggest that academics and practitioners should carefully consider the useful impact of these results. For example, brand supportive behaviors in the marketplace are indeed attractive from the brand management perspective. In contrast, from the deontologist perspective, brand defense and sin of omission are both unethical behaviors. This paradox provides an interesting environment for consumer brand relationships. Within that context, we provide theoretical and managerial implications in the next section.

Theoretical Implications

Extant research in business ethics indicates that consumer brand relationships prove beneficial for companies only to a certain extent when the brand has acted unethically. For instance, when the unethical situation falls within the extreme domain, consumers can no longer justify their support (Einwiller et al. 2006; Bhattacharya and Sen 2003; Liu et al. 2010). Building upon traditional brand identification, a fundamental relationship where consumers identify their individual self with the brand, we contribute to this research stream by showing a more nuanced role of identification in situations containing moral issues. We do so by introducing brand love into business ethics literature. The direct impact of identification on ethical judgment is negative, but when brand love is included, brand identification’s total effect is positive. The ethical judgment, in other words, evaluation of the relative level of immorality in the situation is an important aspect in forming consumers’ behavioral intentions in marketing (Hunt and Vitell 1986, 2006). Therefore, our results show that brand identification helps the brand even in extremely unethical situations but only if a (passionate) love type of relationship exists between the brand and the consumer.

Second, our research also contributes to the recent and growing literature on consumer–brand relationships. Prior literature indicates the role of consumer brand relationship when a brand engages in ethical misconduct. For example, Huber et al. (2010) investigated the impact of brand relationship quality in the case of an ethical misconduct and found that repurchase intentions are vulnerable. Similarly, Trump (2014) showed that connected consumers are not forgiving when the misconduct is in the ethical domain. On the other hand, our research context deals specifically with brand love, the strongest emotional relationship consumers could form with a brand that includes intense emotions and passion (Langner et al. 2016). This inclusion is theoretically valid in this research context as emotions are important drivers of moral behavior (Greene and Haidt 2002; Ditto et al. 2009).

Finally, our results bolster the effect of WOM as consumers’ supportive behavior in the marketplace even for extremely unethical situations. We do so first by introducing a distinct construct to the WOM literature, sin of omission. This construct is especially suitable for studying ethical issues as it relates to protecting brands during interactions with other consumers, even when the brand is perceived to be unethical. Additionally, we introduce the newly established construct, brand defense, to the business ethics literature for the same reason as above. These results are important from an ethical perspective, especially since brand defense and sin of omission are perceived as unethical behaviors as indicated by the results of the pretest.

Managerial Implications

The authors encourage firms to continue supporting the importance of brand management, creating and strengthening their relationship with consumers. From this perspective, managers could tailor their communications to increase consumers’ brand identity. For example, Bergkvist and Bech-Larsen (2010) suggest using image advertising to strengthen the fit between brands’ and consumers’ identities. However, as the results show, it is important that this strong identification should turn into a love type relationship. As Barker et al. (2015) discuss, the BERA platform, which includes different consumer variables related to brand love offers actionable items to brand managers in 200 categories.

However, this study does not focus on how long a suggested mechanism would be effective. We suspect that if a company fails to resolve the unethical activities in a timely manner, keeps repeating the unethical behavior, or simply fails to manage the relationship with consumers, the support due to brand love would gradually weaken. The relationship might even become toxic over time. This logic is consistent with the recent work of Huber et al. (2015) defining brand love as a dynamic relationship where consumers might fall in and out of love with the brands. Along with our empirical results, we argue that managers should be proactive in resolving the ethical issues in a timely manner and also keeping the “passion alive.”

Limitations and Future Research

Our results indicate that strong brand identification leads to increased negative judgment following the event. An interesting question for future researchers is how this negative judgment will have an impact on brand identification in the long term. Research on brand disidentification (e.g., Wolter et al. 2016), for example, suggests that consumers may choose to actively distance themselves from brands. Future research, therefore, could investigate when (i.e., if repeated transgression is unavoidable) and how (i.e., the exact process) the negative judgment will lead to a decrease in identification. Similarly, we expect to see a boundary condition of brand love as discussed above. More specifically, we expect, analogous to interpersonal love, that people might withstand a certain number of unethical acts, but will eventually fall out of love. How consumers fall out of love with brands remains an unexplored area in consumer brand relationships literature (Langner et al. 2016).Footnote 1

Moreover, our theorizing relies on the moral partiality the strong love type of relationship creates. As discussed in the methodology section, it is important for researchers to choose brands that could generate both very low and strong feelings to study the impact of the brand love construct (Bagozzi et al. 2017; Bergkvist and Bech-Larsen 2010). However, not all brands in the marketplace are able to generate these feelings (Fetscherin et al. 2014). Therefore, our results need to be evaluated with caution. Further research should also look into whether or not a brand with a less variance on love construct could still prompt supportive behaviors in the marketplace. We would speculate that it could not, and our prediction is paralleled with the previous research that has investigated the role of identification in extremely unethical situations. Consistent with their research questions, the role of brand identification was not specifically tested under the existence of brand love. In fact, some studies used hypothetical brands for their theoretical manipulations, generating no emotional reaction for consumers.

In the conceptual background, we give plausible explanations as to why we see these brand supportive behaviors. However, we do not consider which explanation is dominant (e.g., the passion vs. moral intuition), and even more interesting, how these constructs might interact with each other. This area of study offers an important direction for future research as well.

Another intriguing question for future research is with whom customers communicate during WOM interactions. For example, both moral judgment (e.g., Haidt 2001) and WOM (e.g., Brown and Reingen 1987) literatures show that it is important to understand the relationship between the sender and receiver. Future research could investigate this issue further by manipulating the love affect that the sender has for different recipients of WOM. Similarly, online versus offline WOM could be an interesting area to investigate. For example, one could expect sin of omission to be stronger in online environments due to the inherent anonymity in the medium. Social networking communities offer a rich context for studying these communications.

Finally, our objective was to manipulate an extreme scenario to investigate the relationship between our constructs. Specifically, the unethical issue we chose is a certain domain of child labor and bribery. What would happen if the unethical behavior occurred in another domain? Future research, therefore, could hold the relative degree of extremely unethical behavior constant and investigate how different contexts (e.g., environmental, product-related, personal, or political) affect the outcomes or impact judgment of morality.