In some instances, loyal customers can take extreme actions to hurt the firm, and thereby become its worst enemies. Consider the story of Jeremy Dorosin, a former loyal customer of Starbucks Coffee. Mr. Dorosin purchased an espresso machine from Starbucks as a wedding gift for a friend, but found it to be defective. After a series of phone calls to Starbucks, he became infuriated when his complaint was not resolved to his satisfaction. As a result, Mr. Dorosin purchased a series of ads in the Wall Street Journal that criticized Starbucks, and he publicized his story with a book, a website, and media interviews on national television.

Customers like Mr. Dorosin can pursue a variety of strategies to punish service firms, including complaining to the media to generate negative publicity, insulting frontline employees, engaging in negative word-of-mouth, and creating weblogs (Huefner et al. 2002; Ward and Ostrom 2006). Stories such as Mr. Dorosin’s may not be that unusual. In fact, practitioner surveys on customer rage (Customer Care M&C 2005), anecdotes in the business press (The Economist 2006; Lyons 2005) and the abundance of customer advocacy websites (Yahoo! 2007) suggest that customer retaliation is becoming prevalent in today’s society. Yet, this issue has received limited attention in academic literature.

In this article, retaliation is defined as a customer’s actions that are designed to punish and cause inconvenience to a firm for the damages the customer felt it caused (Bechwati and Morrin 2003; Grégoire and Fisher 2006). The natural question that arises in this context is the following: Why do loyal customers, such as Mr. Dorosin, invest time, energy, and even money to retaliate when limited material gain exists for doing so? Research on why customers retaliate is needed because the potential for customers to harm firms has grown exponentially with the proliferation of online protection agencies (e.g., consumeraffairs.com), complaint websites (Rip-offReport.com) and anti-corporation websites (e.g. starbucked.com; Ward and Ostrom 2006).

The degree to which Mr. Dorosin’s relationship with Starbucks played a role in how he retaliated is of special interest in this research. It is important to understand how and to what extent relationship strength or quality affects a customer’s response to a service failure and recovery, because strong-relationship customers are one of the key assets of a firm (Heskett et al. 1997). Overall, firms have to do everything in their power to ensure that their best customers do not become their worst enemies, as was the case with Mr. Dorosin.

Although previous research has argued that strong relationships can amplify customers’ unfavorable responses to negative service encounters (Bhattacharya and Sen 2003; Grégoire and Fisher 2006; Hess et al. 2003; Mattila 2001, 2004; Tax et al. 1998), findings have offered limited support for the “love becomes hate” thesis (see Mattila 2004 for an exception). In most cases, strong relationships were found to have a positive effect on satisfaction with a recovery (Hess et al. 2003), to mitigate the effects of a poor recovery on a reduction in trust, commitment, and loyalty (Mattila 2001; Tax et al. 1998), and to reduce the desire for retaliation when low controllability is inferred (Grégoire and Fisher 2006). Overall, the literature suggests that strong relationships protect firms against the possible damage that service failures and poor recoveries can cause (Berry 1995). Hence, because of its managerial importance, we propose a mechanism that explains the counterintuitive “love becomes hate” effect.

This research develops and tests a justice-based theory that incorporates perceived betrayal as the means to understand why customers retaliate and examine the controversial effect of a strong relationship on responses to service recoveries. As a first contribution, this research asserts that betrayal is a key motivational force that leads customers to take actions to restore fairness, and as such, it provides new insights into why customers retaliate. Compared to established causing variables (i.e., anger and dissatisfaction), perceived betrayal is argued to be particularly influential in predicting retaliation. As a second contribution, the concept of betrayal is used to explain the psychological mechanism that underlies the “love becomes hate” effect. This effect is embedded in justice theory and takes place when strong relationship customers perceive a violation of the fairness norm, a situation conceptualized as a low level of fairness related to both the outcomes and the process (Brockner and Wiesenfeld 1996; Tax et al. 1998). As a relationship gains in strength, our model proposes that a fairness violation leads to an increased sense of betrayal, which in turn drives strong relationship customers to retaliate with greater intensity. Our theory and hypotheses are tested on a sample of travelers who had experienced a service failure with an airline company, and subsequently complained to a consumer agency after an unsuccessful recovery.

Theoretical development

Our justice-based model (see Fig. 1) asserts that a violation of the fairness norm creates a sense of betrayal among customers (Elangovan and Shapiro 1998; Koehler and Gershoff 2003), which in turn pressures customers to restore fairness through two basic mechanisms: demanding reparation and retaliating (Walster et al. 1973). The “love becomes hate” effect is represented by the moderation effect of relationship quality. As illustrated in Fig. 1, the robustness of our theoretical model is tested by controlling for a variety of factors, including failure severity (Smith et al. 1999), firm’s blame (Hess et al. 2003), as well as anger and dissatisfaction (Bougie et al. 2003).

Figure 1
figure 1

Theoretical model.

Full size image

This model is based on customers’ responses after both a service failure and a recovery process, a unit of analysis that is consistent with prior research (Smith et al. 1999; Ward and Ostrom 2006). For simplicity of exposition, we only refer to the notion of recovery in the rest of the manuscript. This unit of analysis is also consistent with justice theory that suggests that customers consider reparation first (through the recovery process) before engaging in retaliatory behaviors (Walster et al. 1973).

Demands for reparation versus retaliatory behaviors

Demands for reparation

In a service setting, reparation is a positive mechanism for restoring fairness and refers to anything a service firm provides to customers in order to compensate them for the failure and to redress their grievances (cf., Walster et al. 1973). Reparation includes initiatives whereby firms exchange or repair a defective product, offer a discount or reimbursement, apologize, or respond in a timely fashion (Bowen et al. 1999; Smith et al. 1999).

In this research, the higher-order construct demands for reparation incorporates two specific behaviors. The most direct way for customers to demand reparation is through problem-solving complaining to a firm, which constitutes a customer’s effort to contact a firm in order to find a solution to their problem (Hibbard et al. 2001). In this case, customers can voice their concerns to a firm and ask for reparation. In difficult situations when discussions with a firm appear to fail, customers can seek assistance from a third-party organization. Specifically, third-party complaining for dispute resolution is defined as a customer’s effort to consult a consumer agency or legal counsel to gain assistance in reaching a settlement with a firm. Most consumer agencies, such as the Better Business Bureau, provide such help at no cost.

Retaliatory behaviors

As noted earlier, customer retaliation represents the efforts made by customers to punish and cause inconvenience to a firm for the damages it caused them. In contrast to reparation, whereby customers seek to improve their own situations by receiving something, retaliation is motivated by a desire to “bring down” a firm in some fashion (Walster et al. 1973). Where reparation is fundamentally corrective, retaliation is punitive in essence. In this research, the higher-order construct retaliatory behaviors refers to the three following behaviors.

Retaliation can be direct and takes the form of vindictive complaining, when customers contact a firm to inconvenience and abuse its employees (Hibbard et al. 2001). Other types of direct retaliation exist such as physical aggression, stealing, and vandalism. However, we do not examine these forms here because they are illegal and only infrequently used by customers (Huefner et al. 2002).

Spreading negative word-of-mouth (i.e., a customer’s efforts to denigrate a firm to their family and acquaintances) can be viewed as an indirect form of retaliation (Wangenheim 2005). By sharing their bad experiences with others, customers hope to tarnish a firm’s reputation and to encourage others to avoid patronizing it. Customers can also indirectly retaliate by contacting a third-party organization (i.e., an agency, the media, or a complaint website) to publicize their failure to a vast audience, a behavior named third-party complaining for publicity (cf., Ward and Ostrom 2006). Part of the mission of consumer agencies such as the Better Business Bureau and ConsumerAffairs.com is to inform the public about the number and types of complaints against firms. The underlying goal is to protect public interest, but this process can also harm a firm’s business by making its actions public. Also, customers can contact the media directly and publicly report their misadventures through advertising, opinion letters, or complaint websites.

It should be noted that previous research has not considered relationship exit as retaliatory per se (Bechwati and Morrin 2003; Huefner et al. 2002). Relationship exit is principally motivated by a desire to avoid a future negative experience, and it does not explicitly aim to cause harm and inconvenience to a firm. Although this behavior is not formally part of our conceptualization, it is included in our analysis because of its managerial relevance.

Why do customers retaliate: the effects of betrayal

It is important to understand what motivates customers to retaliate even when there is no material gain for doing so. This research asserts that the concept of betrayal explains why loyal customers like Mr. Dorosin both retaliate and persist in their demands for reparation.

In this research, perceived betrayal is defined as a customer’s belief that a firm has intentionally violated what is normative in the context of their relationship (Elangovan and Shapiro 1998; Koehler and Gershoff 2003; Ward and Ostrom 2006). Research on betrayal has focused on the context of close relationships (Finkel et al. 2002) and employee–employer relationships (Elangovan and Shapiro 1998), and a recent study refers to betrayal as the mechanism explaining the motivation underlying online consumer protest (Ward and Ostrom 2006). Findings from these studies reveal that acts of betrayal are extremely difficult to forgive and forget (Finkel et al. 2002), and they are associated with greater punishments in terms of jail time and punitive damages in a criminal setting (Koehler and Gershoff 2003). In service contexts, acts of betrayal include situations in which customers believe that firms have lied to them, taken advantage of them, tried to exploit them, violated their trust, cheated, broke promises, or disclosed confidential information (cf., Elangovan and Shapiro 1998).

As stated in its definition, betrayal is experienced within the context of a relationship, and this characteristic makes this construct especially relevant to explain a “love becomes hate” effect. Indeed, the relational foundation of betrayal represents an important distinction with other established mediating variables, such as dissatisfaction and anger (Bougie et al. 2003; Smith et al. 1999). Betrayal involves a reference to the norms regulating a relationship, whereas anger and dissatisfaction can be experienced without reference to any relational context. Betrayal also differs from dissatisfaction and anger on other attributes.

Compared to dissatisfaction, betrayal relies on a violation or infringement of normative standards (Elangovan and Shapiro 1998; Ward and Ostrom 2006), and it involves more extreme cognitions than expectation disconfirmation (Oliver 1996), and “thoughts of what one had missed out” (Bougie et al. 2003, p. 380). Compared to customers who feel betrayed, dissatisfied customers are not as inclined to invest energy to restore fairness through retaliation and persist at demanding reparation.

Compared to anger, an intense and action-driven emotion (Bougie et al. 2003), betrayal involves the formation of beliefs about a violation. Although betrayal is expected to lead to anger, these two concepts are not synonymous because of their respective cognitive and emotional natures (Koehler and Gershoff 2003). Overall, our model rests on the assumption that actions directed to restore fairness are deliberate and conscious (Bechwati and Morrin 2003; Walster et al. 1973), and as such, retaliation and reparation are expected to be cognitively driven to a great extent.

These conceptual distinctions being made, our model asserts that betrayal is a key motivational force that energizes customers to restore fairness by all mechanisms or means available to them. Specifically, our model posits that a perceived violation of normative expectations makes retaliation viewed as acceptable, and this perception also creates a greater propensity to demand reparation. Betrayed customers see retaliation as a justified option for restoring fairness and re-establishing a form of social order (Ward and Ostrom 2006). Along with a heightened desire for retaliation, betrayal is also associated with a customer’s difficulty in forgiving, and as a consequence they persist in demanding reparations in the recovery process (cf., Finkel et al. 2002). Formally:

  1. H1a:

    Perceived betrayal is a key antecedent of demands for reparation such as problem-solving complaining and third-party complaining for dispute resolution.

  2. H1b:

    Perceived betrayal is a key antecedent of retaliatory behaviors such as vindictive complaining, negative word-of-mouth, and third-party complaining for publicity.

The “love becomes hate” effect

This research posits that a violation of the fairness norm creates a greater sense of betrayal for customers who have a stronger relationship with a firm. Here, we first explain the “fairness violation → perceived betrayal” effects, and then elaborate on the moderation effect of relationship quality on these former paths (see Fig. 1).

The basic “Fairness Violation → Betrayal” effects

The importance of the fairness norm at the recovery stage is well established (Blodgett et al. 1997; Smith et al. 1999; Tax et al. 1998), and our model asserts that the violation of this norm leads to betrayal. Consistent with the recovery literature, fairness judgments are conceptualized in three dimensions. Distributive fairness refers to the outcomes or the compensation received by a customer at the recovery stage. In turn, customers evaluate the recovery process by making judgments about procedural fairness (i.e., the procedures, policies, and methods used by a firm to address a customer’s complaint) and interactional fairness (i.e., the manner in which employees treat customers when the procedures are enacted).

A norm violation involves more than the individual effects of the fairness components. Based on extensive research in service recovery (Blodgett et al. 1997; Tax et al. 1998), workplace retaliation (Barclay et al. 2005), and justice theory (Brockner and Wiesenfeld 1996), the fairness components are expected to interact in predicting perceived betrayal. Indeed, customers should experience a heightened sense of betrayal when a recovery episode is characterized by a low level of fairness related to both the outcomes and the recovery process. In other words, we expect to find a substantially higher level of betrayal when customers perceive low levels of (1) both distributive fairness and procedural fairness, or (2) both distributive fairness and interactional fairness (Tax et al. 1998). The basic logic underlying these effects is that a fair process has the potential to compensate for a lack of outcome fairness, and vice versa (see Brockner and Wiesenfeld (1996) for detailed explanations). These interactions effects are well established in the literature, and they do not make the object of formal hypotheses.

The moderation effect of relationship quality

Consistent with previous research (Crosby et al. 1990; De Wulf et al. 2001; Hennig-Thurau et al. 2002), the concept of relationship quality is employed to conceptualize the strength of a relationship. Previous research defines relationship quality as a second-order construct consisting of several dimensions. Initial work identified trust (i.e., a customer’s confidence that a firm is dependable and can be relied on) and relationship satisfaction (i.e., a customer’s affective state resulting from the evaluation of all aspects of a relationship over time) as the key constructs that capture the quality of a relationship (Crosby et al. 1990). Subsequent research added commitment to the conceptualization (De Wulf et al. 2001), a construct defined as a customer’s desire to continue a relationship and a willingness to maintain a relationship with a firm.

The “love becomes hate” effect implies that, as relationship quality increases, customers experience a greater sense of betrayal when they perceive low levels of fairness related to both the outcomes and the process (i.e., a fairness violation). The amplifying effect of relationship quality on the “distributive fairness × procedural fairness → betrayal” and “distributive fairness × interactional fairness → betrayal” paths can be explained by the “higher you are, the harder you fall” effect described in organizational psychology (Brockner et al. 1992). Customers who perceive a high level of relationship quality are more likely than others to take offense if they feel they are the victims of an unambiguously unfair recovery. Being treated poorly by a firm with which customers feel a strong connection can be especially disconcerting and hurtful. This situation could also have negative effects on the self-perceptions of high relationship quality customers. Findings in social psychology support the existence of this effect whereby individuals become more upset when they receive criticism from a group with which they feel a strong sense of cohesion (Moreland and McMinn 1999), or if they had a close relationship with the transgressor (McCullough et al. 1998). Formally, we suggest:

  1. H2a:

    In a recovery context, relationship quality moderates the “distributive fairness X procedural fairness → betrayal” path. As relationship quality increases, customers experience a greater sense of betrayal when they perceive a low level of both distributive fairness and procedural fairness (i.e., a fairness violation).

  2. H2b:

    In a recovery context, relationship quality moderates the “distributive fairness X interactional fairness → betrayal” path. As relationship quality increases, customers experience a greater sense of betrayal when they perceive a low level of both distributive fairness and interactional fairness (i.e., a fairness violation).

Research method

Context

This research involved a survey of travelers who experienced poor recoveries with an airline and subsequently complained to the Canadian Transportation Agency (CTA). The CTA serves as a third-party in the settlement of disputes between consumers and airline companies. It also aims to protect the interests of travelers by publicizing online how complaints were handled by the airlines. CTA’s policy is to intervene only when an airline has failed to resolve a situation after a reasonable delay (60 days). Accordingly, the sample contains travelers who experienced a service failure and a failed recovery with an airline. This context is viewed as appropriate because it offers a unique opportunity for understanding customer betrayal and retaliation in a natural setting (cf., Ward and Ostrom 2006).

Method and procedure

Consistent with prior research on service recovery (e.g., Bougie et al. 2003; Tax et al. 1998) and retaliation (Aquino et al. 2001; Barclay et al. 2005; McCullough et al. 1998), we conducted a field study based on the retrospective experiences of customers. Respondents were asked to recall the experiences that led them to complain to the CTA. First, respondents were asked to answer a series of questions related to their relationship with the airline before their service failure. Following this, they were asked to recall the thoughts and feelings they experienced throughout their failure and recovery episode. This method for studying the effects of relationship has been used frequently in the channel literature (Hibbard et al. 2001).

Because of strict privacy guidelines, the CTA was responsible for contacting prospective respondents and informing them about the study. The researchers were not allowed to have any direct communication with the agency’s clients. The complete anonymity of the respondents offers a methodological remedy that reduces common method bias (Podsakoff et al. 2003) and social desirability bias (Fisher 1993). The CTA sent an introductory email to prospective respondents that briefly described the project, and provided a link to the online questionnaire. The introductory email was followed up 3 weeks later with a reminder.

Sample

The CTA sent emails to the 2,057 complainers for whom they had accurate email addresses. Two hundred and fifty questionnaires were returned for a response rate of 12.2%. Twenty-four respondents were eliminated for missing responses. The final sample included 226 usable questionnaires. Sixty-three percent of respondents were male, and the modal age ranges were 30–39 and 40–49 comprising 25 and 30% of respondents, respectively. Before the service failures, respondents had been customers of the airline for an average of 10.5 years (SD = 10.97) and flew it an average of 4.10 times per year (SD = 8.93).

The characteristics of the respondents were very similar to the characteristics of the travelers reported by the CTA. Respondents complained for similar reasons and in similar proportions to those registered in the CTA database: quality of service (29 versus 33% in the CTA database), schedules and denial of boarding (21 versus 23%), loss of baggage (14 versus 13%), tickets and fares (14 versus 18%), frequent flyer program (5 versus 5%), reservations (10 versus 5%), and safety (7 versus 3%). In addition, 92% of the sample spoke English as their first language, compared to 93% for the population.

Potential non-response bias was assessed through an extrapolation method comparing early and late respondents. No significant differences in either the mean score or variance were found for any constructs between the early (introductory email) and late (reminder email) respondents (all p’s > 0.29).

Questionnaire and measurement

Most measures are influenced or adapted from previous work. Unless otherwise indicated, the measures are based on seven point Likert scales (scale end points: 1 = strongly disagree to 7 = strongly agree). The scale items (after purification) are provided in Appendix.

We performed a pilot study with 216 undergraduate students from a major public university in the USA to assess the psychometric properties of perceived betrayal (the results are summarized in the next section). A complete version of the questionnaire was also pre-tested with 31 employees of a major North American university. The final questionnaire was electronically sent to a random sample of 50 consumers. As no problems were identified, the agency proceeded with the electronic mass mailing.

Fairness judgments

Distributive fairness is reflected in three items that include “Overall, the outcomes I received from this experience were fair.” In turn, interactional fairness is reflected in four items such as “The employee(s) who interacted with me treated me in a polite manner,” whereas procedural fairness is reflected in five items including “The airline company was flexible in the way it responded to my concerns.” These scales are based on the work of Smith et al. (1999) and Tax et al. (1998).

Relationship Quality is a reflective second-order construct that is based on trust, commitment, and relationship satisfaction. Trust was measured in four semantic differential items, such as “I felt the airline was very undependable (versus dependable)” (Sirdeshmukh et al. 2002). Commitment was reflected in three items, such as “I was very committed to my relationship with the airline.” Last, relationship satisfaction was measured in three items that included “I was satisfied with my relationship with the airline” (De Wulf et al. 2001).

Perceived Betrayal

is measured with a five-item scale adapted from Bardhi et al. (2005) work. Perceived betrayal was reflected in three items that measure the extent to which customers felt (1) betrayed, (2) lied to, and (3) cheated by the airline. This was followed by two items that measured the extent to which customers perceived that the airline (4) intended to take advantage of them, and (5) tried to abuse them.

Demands for Reparation

is a second-order construct that is reflected in (1) problem-solving complaining, and (2) third-party complaining for dispute resolution. Problem-solving complaining is reflected in three items adapted from Hibbard et al. (2001) that include “I complained to the airline in order to constructively discuss the problem.” In turn, third-party complaining for dispute resolution is measured with four newly developed items such as “I complained to the CTA so it could advise me on the best way to reach a settlement.”

Retaliatory Behaviors

is a second-order construct reflected in (1) negative word-of-mouth, (2) vindictive complaining, and (3) third-party complaining for publicity. Negative word-of-mouth is measured by adapting a three-item scale developed by Wangenheim (2005) that includes items like “I spread negative word-of-mouth about the airline company.” Vindictive complaining is reflected in three items adapted from Hibbard et al. (2001). These items include “I complained to the airline in order to give the representatives a hard time.” In turn, third-party complaining for negative publicity is measured by four newly developed items including “I complained to the CTA to have it report my experience to other travelers.”

Control variables

We controlled for a variety of causes that could have explained the variance of perceived betrayal, demands for reparation, and retaliatory behaviors (see measures in Appendix). First, we controlled for the effects of age and gender (Aquino et al. 2001). Second, we controlled for the effects of the severity of the failure and firm’s blame, two variables that were found to affect customer responses to service recovery in previous research (Smith et al. 1999). Third, we also examined the effects of two behavioral aspects of a relationship, its length in months and frequency of interaction (Hess et al. 2003). Finally, we controlled for the effect of dissatisfaction (Smith et al. 1999) and anger (Bougie et al. 2003; Shaver et al. 1987) on demands for reparation and retaliatory behaviors.

Results

Measurement properties

Pilot study on perceived betrayal

Because of the importance and novelty of perceived betrayal, we performed a pilot study to assess its psychometric properties. A total of 216 undergraduate students were asked to recall a recent service recovery and to complete a series of questions related their sense of betrayal (five items) and behavioral responses (i.e., negative word-of-mouth and vindictive complaining with three items each). A confirmatory factor analysis (CFA) model using the 11 items produced a satisfactory fit with a comparative fit index (CFI) of 0.97, a Tucker–Lewis index (TLI) of 0.96, a root mean square error of approximation (RMSEA) of 0.062, and a χ 2 of 75.03 (df = 41, p = 0.001). In terms of convergent validity, the loadings (λ’s) of perceived betrayal were large (between 0.65 and 0.82) and significant, and its average variance extracted was 0.56. In addition, the internal consistency of this scale was adequate with a Cronbach’s alpha of 0.86. As a sign of discriminant validity, the covariances (Φ’s) of betrayal with the two other constructs were significantly less than one. The covariances were also positive and significant (p < 0.01), a result that provides evidence of nomological validity. Overall, this study provides initial evidence of the validity of this scale.

CFA models

We ran three CFA models for the main study in order to maintain a five to one ratio of observations to parameters (Bentler and Cho 1988). The first CFA model included distributive fairness (three items), procedural fairness (five items), interactional fairness (four items), firm’s blame (three items), dissatisfaction (three items), and failure severity (three items). After deleting two items, this model fit the data acceptably with a χ 2 of 212.11 (df = 137, p = 0.000), a CFI of 0.97, a TLI of 0.96, and a RMSEA of 0.05. The second CFA model included perceived betrayal (with five items), anger (three items), and relationship quality, a second-order construct that was reflected in commitment (three items), trust (four items), and relationship satisfaction (three items). This model fit the data acceptably with a χ 2 of 233.36 (df = 128, p = 0.000), a CFI of 0.96, a TLI of 0.96, and a RMSEA of 0.06. Finally, the third CFA modelFootnote 1 contained the second-order constructs demands for reparation, which was reflected in problem-solving complaining (three items) and third-party complaining for dispute resolution (four items), and retaliatory behaviors, reflected in negative word-of-mouth (three items), vindictive complaining (three items), and third-party complaining for publicity (four items). After deletion of one item, this model fit the data acceptably with a χ 2 of 159.43 (df = 98, p = 0.000), a CFI of 0.97, a TLI of 0.96, and a RMSEA of 0.05.

In all the models, the loadings (λ’s) were large and significant (p’s < 0.001), and the average variance extracted exceeded or approached 0.50 for all constructsFootnote 2 (see Appendix for details). In turn, the standardized loadings (γ’s) of the first-order constructs being the reflection of higher-order constructs were substantive and significant (p’s < 0.05). Cronbach’s alphas were also greater than or approaching the 0.7 guideline (see Appendix). In addition, the covariances (Φ’s) were significantly less than one. Overall, the CFA models indicated that our constructs possessed satisfactory psychometric properties. As a result, construct scores were calculated and used in the regression analyses. Table 1 displays the descriptive statistics and correlations.

Table 1 Descriptive statistics and correlation matrix
Full size table

Tests of hypotheses

The effects of betrayal

To test H1a and H1b, we examined the effects of perceived betrayal on demands for reparation and retaliatory behaviors.Footnote 3 As recommended by Aiken and West (1991), we conducted regression analyses in which the control variables were first entered and then followed by the hypothesized main effect. Because many control variables—age (all p’s > 0.09), blame (all p’s > 0.37), relationship length (all p’s > 0.42), and interaction frequency (all p’s > 0.70)—did not have significant effects, they were excluded from the analysis for the sake of simplicity. Based on collinearity diagnostics, multicollinearity was minimal in both regressions: variance inflation factors varied between 1.12 and 1.74 and were substantively below the 10 guideline. Table 2 displays the standardized coefficients of the final regressions.

Table 2 The effects of perceived betrayal on demands for reparation and retaliatory behaviors
Full size table

The regression analyses provide results that are supportive of H1a and H1b: betrayal had a positive and large effect on both demands for reparation (β = 0.240; p < 0.001) and retaliatory behaviors (β = 0.344; p < 0.001). As expected, the effects of betrayal on these two mechanisms for restoring fairness differed from those of dissatisfaction and anger. Dissatisfaction affected demands for reparation (β = 0.194; p < 0.01), but not retaliatory behaviors (β = −0.098; p > 0.19). In turn, anger had a positive influence on retaliatory behaviors (β = 0.198; p < 0.05), but a negative effect on demands for reparation (β = −0.193; p < 0.05).

Consistent with prior research, the propensity to retaliate was influenced by gender, whereby male respondents report using more retaliatory behaviors than females (β = −0.402; p < 0.01; male is the reference category). In addition, failure severity (β = 0.329; p < 0.001) was found to significantly explain demands for reparation.

Love becomes hate

The moderation effect of relationship quality on the two “fairness violation → perceived betrayal” paths was tested with moderated regression analyses. Our hypotheses (H2a and H2b) were formally tested by examining the significance of the following three-way interactions: (1) a distributive fairness (DF) by procedural fairness (PF) by relationship quality (RQ) interaction, and (2) a DF by interactional fairness (IF) by RQ interaction. In each regression, the control variables were followed by the main effects, the two-way interaction terms, and the three-way interaction terms (Aiken and West 1991). All the interacting predictors were centered, and the interaction terms were created by multiplying the centered predictors (Cohen et al. 2003). This procedure is recommended because it eliminates nonessential multicollinearity between the predictors. Indeed, multicollinearity was minimal with variance inflation factors that varied between 1.11 and 2.04. The final results of the regression analyses are displayed in Table 3. As before, only the significant control variables are included in the analyses.Footnote 4

Table 3 Three-way interactions between relationship quality, distributive fairness, and process fairness (procedural or interactional) predicting perceived betrayal
Full size table

In terms of significant control variables,Footnote 5 failure severity and a firm’s blame had significant main effects on betrayal (all p’s < 0.05) in both regression models. In addition, the main effects of DF and PF were also significant in both models (all p’s < 0.05). In terms of two-way interactions, DF was found to interact with both process components (i.e., PF and IF respectively; all p’s < 0.05), a set of results that was consistent with the assumption of our model and the justice literature. DF was also found to interact with RQ in both models (all p’s < 0.05). Importantly, main effects and two-way interaction terms were qualified by two significant three-way interaction terms. Consistent with H2a, the DF by PF by RQ interaction had a significant effect on perceived betrayal (β = 0.148; p < 0.01). Our results are also supportive of H2b: the DF by IF by RQ interaction term achieved significance (β = 0.138; p < 0.05).Footnote 6

The procedure recommended by Cohen et al. (2003) was used to understand the significant three-way interactions obtained in the regression analysis. Standardized values of “−1” and “1” were input in the regression models for all interacting predictors, and the predicted values of perceived betrayal (for each combination of predictors) were plotted. Figure 2 represents a summary of this procedure. Compared to low RQ customers, high RQ customers experience the strongest sense of betrayal when they perceive a low level of both DF and PF (Fig. 2a), or both DF and IF (Fig. 2b). The level of betrayal experienced by high RQ customers was relatively low for all other combinations of fairness. Overall, these results support the logic underlying H2a and H2b and the “love becomes hate” effect. The results obtained for low RQ customers are also worth describing. Overall, their sense of betrayal seems principally conditioned by their judgment about the fairness of the process (procedural and interactional) and minimally explained by their perceptions of DF.

Figure 2
figure 2

Plotted three-way interactions. a The distributive fairness by procedural fairness by relationship quality interaction (H2a). b The distributive fairness by interactional fairness by relationship quality interaction (H2b).

Full size image

Discussion

Theoretical contribution

Overall, the two research hypotheses were supported by data collected from a national sample of airline customers who had complained to a third-party organization. Our justice-based theory highlights the key role perceived betrayal plays in explaining customer retaliation, and the existence of a “love becomes hate” effect.

Retaliation and reparation

This research contributes to our understanding of how customers respond to poor recoveries by differentiating between two mechanisms for restoring fairness: retaliation and reparation. Building on research that examines general complaining behaviors and third-party responses (Singh 1988, 1990), this research makes the extra step by conceptualizing these behaviors based on their ability to restore fairness in different fashions. Because of evident managerial implications, this research insists on the importance to distinguish between complaining motivated by vindictive versus problem-solving reasons, and between third-party complaining for negative publicity versus dispute resolution.

This research expands previous work on retaliation by introducing an extended list of behaviors that customers can use to hurt a firm. Although negative word-of-mouth has been the object of previous work (Wangenheim 2005), vindictive complaining to a firm and third-party complaining for publicity are relatively new and have received limited attention. We believe that the potential for customers to create negative publicity requires special attention because of the advent of online organizations and communities that make public complaining more convenient and accessible (cf., Ward and Ostrom 2006). Online consumer agencies, complaint websites, consumer weblogs and also the recent phenomena of social networking (i.e., Facebook) and user-generated video sites (i.e., YouTube) have given tremendous power to consumers, and these new behaviors need to be better understood in the context of service failure and recovery.

This research also highlights the positive ways that customers can use to restore fairness, such as problem-solving approaches and dispute resolution via a third-party intermediary. It is encouraging to find that customers, even those who feel betrayed after a poor recovery, remain favorably predisposed to restore fairness through constructive discussions with the firm.

Why do customers retaliate?

A central objective of this research was to understand why customers, like Mr. Dorosin, invest time and resources to retaliate against firms. Overall, our results indicate that betrayal is a key motivational force that explains why customers not only retaliate but also persist in demanding reparation after a poor recovery. In this research, betrayal constitutes the only antecedent that is associated with a greater emphasis upon engaging in both fairness restoration mechanisms. When customers believe that they have been betrayed, they urgently try to restore fairness by all possible means.

The effects of betrayal differ from those observed for dissatisfaction and anger, and these results support our contention that betrayal conceptually differs from these established variables. In contrast to betrayal, dissatisfaction does not create a sufficient drive to lead to retaliation, and this variable is only associated with a greater inclination to demand reparation. Interestingly, anger was found to have differentiated effects on the two restoration mechanisms. We found that anger led customers to try to hurt the offending firm but reduced their demands for reparations. Although the effect of anger on retaliation is consistent with previous research (Bougie et al. 2003), its negative effect on reparation is surprising. We speculate that anger might lead customers to focus on punishing the offending firm rather than restoring fairness.

Perceived betrayal explains more variance in retaliation than any other variables included in this study. These results suggest that customers do not simply retaliate as a response to an emotion or an impulse (i.e., anger). They also do so because they “coldly” judge that a firm has violated fundamental norms related to their relationships. We suggest that the cognitive and normative basis of retaliation may explain why customers like Mr. Dorosin display such a high tenacity and persistence in their efforts.

Antecedents of betrayal

This research also proposed and tested an initial model of the antecedents of betrayal. Consistent with justice theory (Brockner and Wiesenfeld 1996), a high level of betrayal was observed when customers perceived low levels of fairness related to both the outcomes and the recovery process. As predicted, we found that distributive fairness interacted with both procedural and interactional fairness in predicting betrayal. Our analyses also suggest that failure severity and blame attribution are important predicting variables that should be considered in future research.

When love becomes hate

We find a pattern of results that supports the “love becomes hate” effect. Compared to low relationship quality customers, high relationship quality ones feel more betrayed when they perceived a low level of both distributive fairness and process fairness (a situation conceptualized as a fairness violation). These results are counterintuitive because they challenge previous findings (Grégoire and Fisher 2006; Hess et al. 2003; Mattila 2001; Tax et al. 1998) and managerial intuitions (Berry 1995) that suggest that strong relationships function as a safety cushion of tolerance and forgiveness in most service recovery episodes. Although a favorable effect of a strong relationship is likely to characterize “regular” recovery efforts, we propose that this effect becomes unfavorable when customers perceive a violation of the fairness norm. In this case, high relationship quality customers feel more betrayed, and this greater sense of betrayal leads them to retaliate to a greater extent. In other words, betrayal is at the core of the psychological mechanism that explains why loyal customers like Mr. Dorosin may turn against firms and become their worst enemies.

Overall, this research provides further insights into the betrayal effects reported in the literature on service recovery (Mattila 2004) and social psychology (Moreland and McMinn 1999). Previous findings suggest that individuals may develop, after a transgression, a stronger negative attitude toward a firm or institution with which they have a relationship. The current article augments our understanding of the betrayal effect by formally measuring betrayal, demonstrating its strong link with retaliation, and explaining the justice context in which this effect effectively takes place.

In addition, our findings indicate that the betrayal of low relationship quality customers is primarily conditioned by the interactional and procedural aspects of a recovery. Indeed, one of the two-way interactions (i.e., the significant distributive fairness by relationship quality interaction term) suggests that the fairness related to the outcomes have minimal effect on the sense of betrayal experienced by this segment of customers. Here, the process heuristic theory developed by Van den Bos et al. (1997) provides a rationale for these unexpected findings. Because low relationship quality customers tend to lack the experience and knowledge to adequately assess the appropriateness of their outcomes, they may rely more on their immediate evaluation of the process to identify an act of betrayal. According to Van den Bos et al. (1997), process fairness is easier to assess than distributive fairness because the former relies on a comparison with universal norms.

Implications for managers

The potential costs of customer retaliation are significant, and they are above and beyond the loss of a customer’s lifetime patronage. For instance, direct acts of retaliation, such as making a public scene or insulting personnel, place a great deal of pressure on frontline employees. Retaliation is also insidious because many of its facets (i.e., word-of-mouth and negative publicity) take place outside the borders of a firm, leaving managers unaware of these actions. To identify retaliators, Heskett et al. (1997) recommend devoting special resources to tracking online and public acts of retaliation. Although initiatives such as these are important, they are insufficient because they only detect public actions and therefore only the visible part of the retaliation “iceberg.” The present study suggests that measuring perceived betrayal after a recovery could provide an early warning system for retaliation.

Although identification procedures should be implemented, prevention is arguably the best strategy against retaliation. To prevent betrayal and retaliation, service firms must have a clear understanding of the normative expectations of their customers, and identify the point at which poor service recoveries are viewed as normative violations. Our findings suggest that the interactional and procedural aspects of a recovery process are especially important in the creation of this belief for both low and high relationship quality customers. If customers perceive that recovery procedures and interpersonal interactions are fair, they will not feel betrayed regardless of the outcomes received and their levels of relationship.

To prevent “love becomes hate” situations, firms should also develop recovery procedures that are specifically designed to satisfy the needs of high relationship quality customers. Firms should make special efforts to identify these customers, and to insure that their grievances are effectively “repaired” early in the recovery process. Yet, the identification of these customers could be more challenging than expected. Customers who perceive a high level of relationship quality are not necessarily those who had the longest relationship with a firm or who had the highest frequency of interactions. Accordingly, firms should collect data about their customers’ perceptions of relationship quality, and not exclusively rely on the behavioral measures available in their information system.

Limitations and future directions

We faced challenges and limitations that stemmed from the exigencies of studying customers who complained to a third-party organization. First, given CTA’s privacy guidelines, we had no control over the procedures for contacting the complainers, and as a result, our ability to increase our response rate was restrained. Second, our context could involve an amplified sense of betrayal because all of the respondents externally reported their failure. However, we believe that these limitations are, on balance, minor compared with the opportunity to study real examples of customer betrayal and retaliation that took place in the context of a third-party organization with an online presence. In addition, we argue that the nature of our data offers a conservative test of the “love becomes hate” effect, as respondents may have reported a diminished level of relationship quality after experiencing a poor recovery.

Given the specificity of our context, replication and extension through alternative methodologies and with different samples are desirable. It would be useful to manipulate service recovery characteristics and relationship quality within an experimental design to validate the proposed model. In addition, a longitudinal design would help increase our understanding of the effects of time on our model. Importantly, a longitudinal design should examine whether a sense of betrayal is experienced directly after a service failure, or whether it necessarily involves a poor recovery. Finally, the robustness of our findings should be tested with samples of customers who experienced service failures and recoveries in different industries.

Although our decision to refer to relationship quality is consistent with previous work (Crosby et al. 1990), we recognize that other conceptualizations—such as firm identification (Bhattacharya and Sen 2003) and emotional bonding (Mattila 2004)—could shed new light on the “love becomes hate” effect. It would be interesting to examine whether these different conceptualizations would amplify or attenuate our findings. Social exchange theory (Blau 1962) also proposes a mechanism that could be useful for studying customer retaliation. As relationships gain in strength, customers may feel that firms “owe them” more than they owe others. In a recovery context, the inability of a firm to reciprocate the investments made by high relationship customers could cause them to feel a more intense desire for retaliation.

Finally, recent research in psychology has studied the effects of personality traits such as agreeableness, negative affect, and neuroticism on retaliation (McCullough et al. 1998; Skarlicki et al. 1999). Although the effects of personality traits were found to be relatively weak (McCullough et al. 1998), the potential role they play in incidents of customer retaliations warrants further examination.