Marketing Letters

, Volume 18, Issue 1, pp 85–99

Affective responses to service failure: Anger, regret, and retaliatory versus conciliatory responses

Authors

    • School of Business AdministrationUniversity of Vermont
  • Catherine Cole
    • Henry B. Tippie College of BusinessUniversity of Iowa
Article

DOI: 10.1007/s11002-006-9006-6

Cite this article as:
Bonifield, C. & Cole, C. Market Lett (2007) 18: 85. doi:10.1007/s11002-006-9006-6

Abstract

After a service failure, consumers make appraisals or assessments about the characteristics of this failure. These appraisals, in turn, affect how a consumer responds emotionally and behaviorally. Using an appraisal-tendency framework, we predict that two negatively valenced emotions (anger and regret) underlie or mediate the effects of consumers’ appraisals about service failure on post-purchase behaviors. Consistent with the predictions, in a laboratory study, we find that anger plays a powerful role in explaining retaliatory behaviors, and that both anger and regret account for the effect of appraisals on conciliatory behaviors. We extend the same appraisal-tendency framework to predict how changes in emotions underlie the effects of recovery efforts on post-purchase behaviors. Again consistent with predictions, in the laboratory study and in a web-based study, we find that recovery efforts that reduce anger decrease retaliatory behaviors. However, both studies provide less clear-cut evidence about the emotional mediators between recovery efforts and conciliatory behaviors. Because conciliatory behaviors are important behaviors for businesses to promote, future research should explore what other emotions explain recovery effort effects on conciliatory behaviors.

Keywords

Post-purchaseAngerRegretRecovery

 

Think about the last time your airplane flight was cancelled. Did you try to analyze the cause of this service failure? Did you feel angry? Or did you feel regret over your decision to take the last flight of the day? And, is there anything the airline could have done, at the time that they cancelled the flight, to make you a loyal customer today?

To address these questions, we introduce Lerner and Keltner's (2000) appraisal-tendency framework to the literature on service failure. We use this framework to justify hypotheses about how two important emotions (anger and regret) will have different influences on consumer post-purchase behaviors after a service failure. Previous work has established the dotted lines in Fig. 1. These lines link characteristics of the service failure and the recovery effort to post-purchase behaviors. Our research enhances this prior work by incorporating anger and regret, which are two similarly valenced emotions with different antecedents and consequences. Additionally, our refined measures of two types of post-purchase behaviors (retaliatory and conciliatory) will enhance future research on post-purchase behavior. Finally, we gain managerial insights about how to change post-purchase behavioral responses after a service failure.
https://static-content.springer.com/image/art%3A10.1007%2Fs11002-006-9006-6/MediaObjects/11002_2006_9006_Fig1_HTML.gif
Fig. 1

Conceptual model

We proceed in the following fashion. In the first section, we develop hypotheses about the linkages between appraisals, anger, regret, post-purchase behaviors, and recovery efforts. Then we present the methodology and the results from an experimental study, in which we examine counterfactual thinking as an intervention strategy (imagining how things could have turned out differently). In the subsequent web-based study, we analyze consumer web postings about restaurant failures, this time focusing on the relationships among recovery efforts, emotions, and post-purchase behavior. Finally, we discuss implications for theory and practice.

Literature and hypotheses

Appraisals, anger, regret, and post-purchase behaviors

Numerous studies have examined Path C in our conceptual model which shows the direct relationships between consumers’ appraisals about a service/product failure and post-purchase behaviors, particularly in the areas of complaining, negative word-of-mouth, repurchase intentions, and consumer vengeance (see, for example, Folkes et al., 1987; Folkes, 1984; Bearden and Teel, 1983; Richins, 1987; Singh, 1988; Bechwati and Morrin, 2003). Most of this research has studied the direct consequences of consumers’ causal appraisals on post-purchase behaviors without taking into account the role of emotions (except Folkes et al., 1987; Tsiros and Mittal, 2000).

However, increasingly, the literature recognizes that affect, not just cognition, influences judgment, decision making and even post-purchase behaviors (Lerner and Keltner, 2001). For example, Smith and Bolton (2002) examine how negative emotional responses to a service failure influence how consumers evaluate an organization's recovery efforts. Because they combine negative emotions, but recognize that different types of negative emotions, with different antecedents, may have different effects on recovery effort evaluations, they urge future researchers to study how different types of emotional response affect evaluations after a service failure.

Although recent literature indicates that specific similarly valenced emotions can influence judgment in different ways (see, for example, DeSteno et al., 2000), Lerner and Keltner (2000) present a particularly cogent framework for explaining why specific emotions, with different antecedents, influence judgment and choice in different ways. Drawing on cognitive-appraisal theories of emotion, such as Smith and Ellsworth (1985), Lerner and Keltner's framework assumes that different emotions arise from different appraisal patterns along six cognitive dimensions including pleasantness, responsibility, certainty, attentional activity, situational control, and anticipated effort.

For example, the key dimensions that distinguish anger from other negative emotions are certainty, control, and responsibility. Anger occurs as a result of individuals’ appraisals of high other-responsibility for negative events and high other-control over these negative events (Averill, 1983), but regret, another negative affect, occurs as a result of individuals’ appraisals of high self-responsibility for and high self-control over negative events. In contrast to anger, regret arises when we realize or imagine that our present situation would have been better, had we acted differently. According to regret theory, consumers feel regret when they compare an actual outcome with possible alternative outcomes had they chosen another action—that is, between “what is” and “what might have been.” Recent literature suggests that regret and anticipated regret influence consumer behaviors such as purchase intentions, purchase timing, and choices between brand name and price (Creyer and Ross, 1999; Hetts et al., 2000; Simonson, 1992; Tsiros and Mittal, 2000). Because anger and regret operate in sharp contrast to one another in the area of responsibility, and attribution is a judgment that clearly relates to the appraisal theme of responsibility, we would expect that when consumers perceive others as responsible for a service failure, they will experience anger, but when consumers perceive themselves as responsible, they will experience regret.

Lerner and Keltner (2001) further argue that different emotions trigger different changes in cognition, physiology, and action (different appraisal tendencies), which persist and influence judgments and decision making until the emotion-eliciting problems are resolved. Since anger and regret have different appraisal patterns, we predict that anger and regret will affect behavior in different ways after a service failure. Anger, arising from perceived other-responsibility, is a strong feeling of displeasure or hostility, accompanied by a desire to attack the source of anger. So compared to less angry people, angry consumers are more likely to engage in retaliatory behaviors and less likely to engage in conciliatory negotiations with this blameworthy other. On the other hand, compared to consumers experiencing low levels of regret, people experiencing high levels of regret will be more likely to want to “undo their own choice” and engage in conciliatory negotiations to get what they want. It is unlikely, though, that regret levels will affect intentions to engage in retaliatory behaviors. Thus, our key argument is that anger (but not regret) will mediate between appraisals about responsibility for a service failure and retaliatory behaviors, while both anger and regret will mediate between these appraisals and conciliatory behaviors.
  • H1: Anger will mediate the effect of appraisals about responsibility for a service failure on retaliatory behaviors.

  • H2: Anger and regret will mediate the effect of appraisals on conciliatory behaviors.

These hypotheses require that we refine existing measures of post-purchase behaviors. Several useful typologies and lists, which provide a starting point for identifying behaviors, include Hirschman's exit, voice, and loyalty typology, consumer complaint behavior typologies, and Bettencourt's list of customer voluntary performance behaviors (Bettencourt, 1997; Hirschman, 1970). From this literature, we developed a list of post-purchase behaviors; we then conducted 14 in-depth interviews about failed service experiences and refined our initial listing. We a priori categorize these behaviors into two types. The first type, retaliatory behaviors, occurs when consumers try to hurt the firm, and includes aggressive complaining, negative word-of-mouth, and insisting on a cash discount. The second type, conciliatory behaviors, are important because service providers understand that increasing their frequency will lead to improved customer retention which in turn will lead to improved profits. Conciliatory behaviors include positive word-of-mouth, willingness to accept a discount on a future purchase, willingness to return to a service provider, and feeling sympathy for the service provider. (See Appendix A for a list of items used in Study 1.)

Managerial interventions

In view of retailers’ interest in customer retention, finding effective managerial interventions in the event of failed service outcomes is of great interest. Several researchers have studied factors impacting consumers’ repatronage intentions following a dissatisfactory consumption experience. Blodgett et al. (1993) found perceived justice to be an important determinant of customer retention. Hess et al. (2003) studied how past encounters with a firm impact consumers’ expectations of relationship continuity and their responses to service failures and recoveries, and found that customers with higher expectations of relationship continuity have lower service recovery expectations. In our paper, we explore whether reducing consumers’ negative emotions has an impact on retaliatory and conciliatory behaviors.

In our first study, we examine the effects of one type of intervention (instructions to engage in downward counterfactual thinking), which is aimed at reducing anger, and thus, decreasing retaliatory and increasing conciliatory behaviors. A downward counterfactual (thoughts about how things could have been worse) is a comparative judgment that can improve mood and that can fulfill important psychological functions (Olson et al., 2000). However, these comparative judgments do not reduce all negative emotions. For example, Roese et al. (1999) find that counterfactual thinking has different effects on feelings of dejection and agitation. We predict that downward counterfactual thinking will not alleviate regret, because regretful people spontaneously use counterfactual thinking in a preparative way: to think about how if they behaved differently, they could have a better outcome (upward counterfactuals) (Markman et al., 1993). Downward counterfactual thinking might direct their thinking to what they could have done to make things worse. As a result, we predict that the effects of a downward counterfactual thinking intervention on retaliatory and conciliatory behaviors will be mediated by anger, but not by regret. This leads to hypotheses H3(a) and H3(b):
  • H3: Anger will mediate the effects of downward counterfactual thinking on (a) retaliatory behaviors and (b) conciliatory behaviors.

Studies

To test these hypotheses, we conducted two studies, one a laboratory experiment using undergraduate students as subjects, and the other a web-based study. In both cases, we used restaurants as the service category because not only are they familiar to students, but also there are many web sites where consumers post comments about restaurant service.

Study 1

Method

One hundred and forty-three undergraduate college students participated in a 2 (appraisal: internal vs. external) × 2 (intervention: none vs. self-generated downward counterfactuals) between-subjects experiment outside of class time. Within each session, we randomly assigned participants to one of the four conditions by distributing randomly ordered questionnaire packets. The participants completed a short survey about their experiences at local restaurants, reviewed and selected an item from one of two versions of a professionally designed menu for a new local restaurant, read a written scenario that further manipulated the appraisal, completed a thought-listing task that manipulated the intervention, and finally, completed the main questionnaire with items measuring anger, regret, post-purchase behavioral intentions, and manipulation checks.

The scenario described a student couple who decided to eat at a new local restaurant before going to a musical production. Their meals took a long time to arrive to their table (either because of the restaurant's policies, which was the external appraisal condition, or because of the special meal they selected which the menu indicated took extra time to prepare, which was the internal appraisal condition). As a result, they missed the first act. After reading the scenario, participants completed a thought-listing task related to the intervention. In the no intervention (control) condition, participants were instructed to “think over the situation carefully and list any feelings you have regarding any aspect of the situation.” In the intervention condition, participants were instructed to “please try to think of and list as many ways as possible ways that things could have turned out worse than they did.”

On seven-point Likert-type scales, the participants responded to questions about anger, regret, and conciliatory and retaliatory post-purchase behaviors. We measured anger with a three-item scale (e.g., “I would feel angry about my experience at this restaurant”) (Coefficient alpha=.90) and regret with a six-item scale (e.g., “I would regret choosing the menu item that I chose”) (Coefficient alpha=.96). Participants answered 13 questions designed to assess intention to engage in different post-purchase behaviors (Coefficient alpha for retaliatory=.88; Coefficient alpha for conciliatory=.88). A nine-item causal dimension scale measured participants’ appraisal (Russell, 1982) (Coefficient alpha=.90).

Results

Table 1 shows the means and standard deviations for all variables. Manipulation checks indicated that participants perceived the causal ascriptions as intended in the scenarios. An ANOVA with appraisal, intervention, and their interaction as the independent variables and scores on the causal dimension scale as the dependent variable indicated that, consistent with expectations, consumers in the internal appraisal condition were more likely to attribute the cause of the poor dining experience to internal causes than consumers in the external appraisal condition (2.58 External vs. 3.35 Internal, F(1, 138)=14.70, p<.05). An additional ANOVA on the number of downward counterfactuals in the thought-listing protocols as the dependent variable indicated, consistent with expectations, that consumers in the downward counterfactual intervention condition generated more counterfactuals in their verbal protocols than consumers in the control condition (.15 Control vs. 4.10 Intervention, t (76.4)=−16.18, p<.05). The number of thoughts listed did not vary by condition (4.12 Control vs. 4.28 Intervention, t (118)=−0.5456, p<.60). At the end of the survey, we asked consumers to indicate how realistic the restaurant scenario was. All indicated that they found the scenarios highly realistic and the realism scores (see Table 1) did not vary by condition (F(3, 137)=0.98, p<.42).
Table 1

Means and standard deviations—Study 1

 

Internal no intervention

Internal intervention

External no intervention

External intervention

Dependent variables

    

 Anger

5.54 (1.47)

5.28 (1.43)

6.59 (0.59)

6.00 (0.89)

 Regret

5.10 (1.54)

4.88 (1.12)

1.89 (1.08)

2.15 (1.05)

 Retaliatory

4.93 (1.67)

4.76 (1.40)

5.82 (1.03)

5.03 (1.26)

 Conciliatory

3.23 (1.22)

3.51 (1.01)

2.10 (0.92)

2.84 (1.14)

Manipulation checks

    

 Causal

3.45 (1.19)

3.25 (1.15)

2.43 (1.21)

2.74 (1.18)

 Realistic

6.15 (0.96)

5.97 (1.12)

5.73 (0.90)

5.86 (1.24)

Prior to testing the hypotheses, we conducted a factor analysis to verify that the statements measuring anger and regret were separate emotions. As expected, a two-factor solution emerged in which the two factors are interpreted as the expected anger and regret emotions. Both eigenvalues are greater than one, and the two factors explain 78% of the variance in the data. We conducted a second factor analysis to verify that a two-factor measurement model adequately captured our two types of individuals’ post-purchase behaviors. As expected, a two-factor solution emerged in which the two factors are interpreted as the expected retaliatory and conciliatory behaviors. Both eigenvalues are greater than one, and the two factors explain 55% of the variance in the data.

To test H1 and H2, we initially conducted two Multivariate Analyses of Variance, using two independent variables (appraisal and intervention) and their interaction and using two correlated dependent variables. One MANOVA used retaliatory and conciliatory behaviors as dependent variables, which were negatively correlated (r=−.59, p<.05); another used both anger and regret, which were negatively correlated (r=−.42, p<.05). In each MANOVA, the results indicated that there were significant effects for appraisal and intervention on the dependent variables, but there were not significant interactions between the independent variables. Because the appraisal and intervention effects differed for different dependent variables, we proceeded to test our hypotheses using the results from the univariate ANOVA's.

As expected, consumers who blamed the service provider (external appraisal condition) indicated a higher likelihood of engaging in retaliatory behaviors and lower likelihood of engaging in conciliatory behaviors than consumers who blamed themselves (internal appraisal condition) (Retaliatory: 5.43 External vs. 4.80 Internal, F(1, 138)=6.46, p<.01; Conciliatory: 2.47 External vs. 3.37 Internal, F(1, 138)=24.68, p<.01).

A multiple variable mediation analysis (Baron and Kenny, 1986) shows that the effect of appraisals on retaliatory behaviors is mediated by anger, but not regret, as specified in H1. The results show that the independent variable appraisal (regarding internal vs. external cause) is a significant predictor of retaliatory behaviors (F=6.34. p<.05), anger (F=15.92, p<.05), and regret (F=104.79, p<.05). In a regression of anger and regret on retaliatory behaviors, anger is significant (Beta=4.94, t=10.08, p<.01), but regret is not (Beta=−.41, t=−1.28, p<.20). Further, when anger, regret, and appraisal are all included in the ANCOVA model for retaliatory behaviors, anger is significant (F=48.78, p<.05), regret is not significant (F=3.38, p>.05), and appraisal is no longer significant (F=2.34, p>.15). Thus, we have support for H1: anger (but not regret) mediates the effects of appraisals on retaliatory behaviors.

Consistent with H2, a multiple variable mediation analysis shows that the effects of appraisals on conciliatory behaviors are mediated by both anger and regret. The results show that appraisal is a significant predictor of conciliatory behaviors (F=17.67, p<.05), anger (F=15.92, p<.05), and regret (F=104.79, p<.05). Also, in the regression of anger and regret on conciliatory behaviors, anger and regret are significant (Anger: Beta=−2.53, t=−6.15, p<.01, Regret: Beta=.693, t=2.60, p<.01). When anger, regret, and appraisal are all included in the ANCOVA model for conciliatory behaviors, anger and regret are significant (F=9.38, p<.05; F=5.02, p<.05, respectively), but appraisal is no longer significant (F=1.57, p>.22). Thus, we have support for H2: anger and regret mediate the effects of appraisals on conciliatory behaviors.

To test H3(a) and (b), we conducted a mediation analysis on the intervention. The downward counterfactual intervention reduced retaliatory behaviors and increased conciliatory behaviors (Intervention on retaliatory behaviors: No Intervention 5.37 vs. Intervention 4.89, F(1, 138)=4.44, p<.05; Intervention on conciliatory behaviors: No Intervention 2.66 vs. Intervention 3.17, F(1, 138)=7.90, p<.01). In addition, the intervention had a significant effect on anger, but not regret (Anger: No Intervention 6.08 vs. Intervention 5.63, F(1, 138)=4.81, p<.05; Regret: No Intervention 3.43 vs. Intervention 3.55, F(1, 139)=0.01, p=.98). This result, which is consistent with our expectation that the intervention would decrease anger, but not regret, means that regret cannot function as a mediator between the intervention and post-purchase behaviors. In a regression of just anger on retaliatory behaviors, anger is significant (Beta = 4.18, t=11.68, p<.01); in the regression of just anger on conciliatory behaviors, anger is significant (Anger: Beta=−3.02, t=−7.94, p<.01). When anger was added to an ANCOVA on retaliatory behaviors, anger achieved statistical significance and the intervention was no longer significant (Anger: F(1, 139)=133.38, p<.01; Intervention: F(1, 139)=0.72, p=.39). When anger was added to the ANCOVA on conciliatory behaviors, anger achieved statistical significance (F(1, 139)=75.42, p<.01) and the intervention was marginally significant (F(1, 139)=3.09, p<.10). As discussed by Baron and Kenny (1986), this last result indicates that anger is a potent mediator between the intervention and conciliatory behaviors, but a reduction in anger is not a sufficient condition for an increase in conciliatory behaviors.

To summarize, we find that consistent with H3(a) and (b), anger is mediating the effects of a downward counterfactual intervention on retaliatory and conciliatory behaviors. Like any laboratory study, in designing our study we made tradeoffs between control over the stimuli and ecological validity. By giving student subjects an actual menu and describing the situation as that of two college students at a new local restaurant, we heightened realism, while controlling the experience that consumers reacted to. Several responses on the questionnaire suggest that consumers felt that the somewhat artificial task was interesting and relevant to them: 98% of the respondents had ideas about what the restaurant could do to make things right, 95% of the respondents named a favorite restaurant in the community, and 60% reported that a similar event had happened to them.

Perhaps a more serious criticism of this first study is that although our intervention is theoretically interesting and potentially useful as a question on written consumer complaint forms, it would be hard for front line people to use in interactions with unhappy consumers. As a result, we conducted a second study to investigate several different interventions that are important to consumers, relevant to retailers, and theoretically interesting. In this second study, we turn to randomly selected consumer reports about failed restaurant experiences posted on the Internet.

Study 2

Background

In Study 2, we select and analyze 299 actual restaurant experience reports1 from Internet web sites to further examine a general form of H3: after a service failure, does anger mediate the effects of interventions on retaliatory and conciliatory behaviors? We coded each report for the presence or absence of four interventions, the consumer's expressed level of anger, and the presence or absence of eight different post-purchase behaviors.

We focus on four common and theoretically interesting restaurant interventions: downward counterfactual statements, making an apology, behaving rudely toward the consumer, and trying to shift the blame from the restaurant to the consumer. One (downward counterfactual statements) was studied in Study 1, while the other three have been previously studied (e.g., Bechwati and Morrin, 2003; Wirtz and Matilla, 2004). These interventions differ in their effect on retaliatory and conciliatory behaviors. For example, Bechwati and Morrin (2003) find that behaving rudely and implying that the failure is the customer's mistake (shifting the blame) increases retaliatory behaviors, while Wirtz and Matilla (2004) find that apologizing increases conciliatory behaviors. Based on this prior research, we expect shifting the blame and behaving rudely to increase retaliatory behaviors, while apologizing and making a downward counterfactual statement will increase conciliatory behaviors. We label the former aggravating interventions and the latter calming interventions.

Method

To select reports, we randomly selected and printed off reports listed in the restaurant/food category at puddinghouse.com, epinions.com, planetfeedback.com, complaints.com, and my3cents.com. Within each site, we first sorted the reports alphabetically by restaurant name (if possible), then took a random start between 1 and k, where k was the number of entries divided by 75 and then coded every kth report. Of the reports, 6 came from puddinghouse.com, 73 from epinions.com, 75 from planetfeedback.com, 73 from complaints.com, and 73 from my3cents.com.

Two coders, who were naïve about the study's hypotheses, were trained in coding the protocols. Initially, they each coded 20 protocols (not in the data set) and then met with the researchers to revise the coding questionnaire. After the discussion, we developed a coding manual that each judge used independently. For each report, one judge coded the following: the presence or absence of each of the four interventions (behaving rudely, trying to shift blame, apologizing, and making a downward counterfactual statement), the expressed level of anger (on a one to five scale with one very low and five very high), the presence or absence of four retaliatory behaviors drawn from the previous study (discouraging others from using the restaurant, demanding an immediate discount, aggressively complaining to management, and telling others about the experience), and the presence or absence of four conciliatory behaviors drawn from the previous study (suggesting a solution or service improvement, recommending the restaurant to others, returning the next time the consumer goes out to eat, and returning to the restaurant at some unspecified time in the future) in the entries. The Kuder-Richardson (KR20) reliability coefficient, which assesses the reliability of scales with dichotomous items, was .61 for retaliatory behaviors and .62 for conciliatory behaviors, which falls within acceptable levels for short scales with dichotomous items.

The second coder coded a randomly selected 20% of the reports. To assess inter-rater reliability, we calculated the correlation between the two coders on the anger measures, obtaining a correlation of .92. We also calculated the percentage agreement on the number of different retaliatory (96%) and the number of conciliatory (91%) post-purchase behaviors, and the number of aggravating (92%) and calming (93%) interventions.

Table 2 contains information about the variables in our analysis, example statements from the reports, and the percentage of reports that contained the behavior or the intervention.
Table 2

Frequencies of conciliatory and retaliatory post-purchase behaviors and calming and aggravating interventions in Study 2

   

Percent of consumers

Type

Behavior

Example

reporting

Retaliatory post-purchase behavior

 

Aggressively complaining to management

My wife stood up, walked over to the manager and told him in a very loud voice exactly what was wrong with this place.

20

 

Tell others about the experience

I know there is nothing we can do except quit going to the restaurant and to tell others about what happened.

14

 

Discourage others from using the restaurant

I have told all my friends, college connections, business associates to avoid this restaurant because of the treatment I received not just by the employee, but by the company itself.

13.4

 

Demand discount

I wanted to be compensated for my full meal which came to $14.00.

19.73

Conciliatory post-purchase behavior

 

Suggest a solution/service improvement

What the manager should have done after we waited for him for over 10 minutes was take care of the check.

50.5

 

Recommending to others

Also, I probably still will recommend you to others.

10.37

 

Return the next time

We will eat again at your restaurant this Friday night.

9.03

 

Return at some time in the future

I might be curious enough to try their biscuits and gravy at breakfast.

37.12

Calming interventions by retailer

 

Making a downward counterfactual

She told us that earlier in the day, they had had worse problems with the kitchen and that some customers had waited over an hour for their food.

1.34

 

Apologize

I got his attention and explained the problem. He apologized.

6.35

Aggravating interventions by retailer

 

Trying to shift the blame from the service provider to the consumer

I only wanted jam on my biscuit, but somebody put butter on it. Then when I complained, the waitress lied and said that I had asked for a buttered biscuit.

10.37

 

Behaving rudely

This kind of rudeness ruins your day.

2.68

Analysis and checks

We classify the 299 reports into those that have at least one calming intervention and those that have none (91.6% of the reports). We also classify the reports as having no aggravating interventions (94.65% of the reports) or at least one aggravating intervention. Consumers who reported at least one aggravating intervention also reported increased retaliatory behaviors (.75 No aggravating intervention vs. 1.25 At least one aggravating intervention, F(1, 297)=3.86, p<.05). Consumers who reported at least one calming intervention reported increased conciliatory behaviors (1.02 No calming intervention vs. 1.76 At least one calming intervention, F(1, 296)=11.92, p<.01). However, aggravating interventions did not have a significant effect on conciliatory behaviors (F(1, 297)=1.61, p<.21), nor did calming interventions affect retaliatory behaviors (F(1, 297)=1.16, p<.29). These results are consistent with our previous definitions of aggravating and calming interventions.

Results

To test whether anger mediates the effects of interventions on post-purchase behavior, we first tested the effect of aggravating interventions on retaliatory behaviors, finding a significant relationship as reported above (F(1, 297)=3.86, p<.05). Then we tested the effect of aggravating interventions on anger, finding a significant effect (2.4 No aggravating intervention reported vs. 3.5 Aggravating interventions reported, F(1, 297)=7.62, p<.01). We next regressed anger on retaliatory behaviors, finding that anger was significant (Beta for anger: .286, t=5.69, p<.01). Consistent with H3(a), when anger was added to the ANCOVA with retaliatory behaviors as the dependent variable, aggravating interventions as the independent variables and anger as a covariate, anger was significant (F(1, 296)=26.05, p<.01) and aggravating interventions were no longer significant (F(1, 296)=1.47, p=.22). Thus, we have evidence that anger mediates the effects of aggravating interventions on retaliatory behaviors. However, calming interventions do not have a significant effect on anger (F(1, 296)=.05, p<.83), so anger does not mediate the effects of calming interventions on conciliatory behaviors.

Admittedly, there are limitations with this data. First, because consumers posted the reports on the Web, the consumers may have biased their reports to be more socially desirable. In addition, because the reports are based on memory, the informants may have left out key information. Nonetheless, we again see that interventions that affect anger can change post-purchase behavior.

Theoretical and managerial implications

Summarizing across the two studies, we make two primary theoretical contributions. First, we identify the powerful role anger plays in explaining retaliatory post-purchase behaviors. When consumers blame the service provider for a service failure, they experience anger which causes their intentions to retaliate against the service provider to increase. An intervention, which reduces this anger, decreases intentions to retaliate. The second study confirms this result by showing that retailers who try to shift the blame from the service provider to the consumer or who behave rudely to the consumer increase anger and increase consumers’ intentions to retaliate. Study 1 suggests what to say and do; Study 2 suggests what not to say and do when faced with an angry consumer.

Second, we have learned that anger is one of several emotional mediators which can affect conciliatory behaviors. Study 1 suggests that appraisals about responsibility which decrease anger and increase regret will also increase conciliatory behaviors. However, both studies indicate that we do not understand well the emotions that mediate between interventions and conciliatory behaviors. In Study 1, our mediation analysis of the effects of the downward counterfactual intervention on conciliatory behaviors showed that anger was only a partial mediator. In the content analysis of the web sites (Study 2), we found that interventions (making an apology, making a downward counterfactual statement), which did not affect anger, could increase conciliatory behaviors.

From a managerial perspective, by understanding the emotions intervening between consumer appraisals about a service failure and post-purchase behaviors, managers can reduce the incidence of retaliatory behaviors and increase the frequency of conciliatory behaviors. Based on Study 1, managers can try to cue downward counterfactual thinking after a service failure either by training employees about what to say or by including a question on a complaint form. For example, an employee could say, “Wow—we really messed up here. I wonder if we could have done anything to make it worse?” A complaint form could solicit the basic information, but also ask the consumer to list ways that things could have turned out worse. However, we need to add a caveat that these ideas need to be tested because in practice they could inflame the situation.

Limitations and future research

Our limitations suggest several directions for future research. First, we study just the responsibility dimension of appraisals. Clearly, this dimension is important in understanding how service failures affect anger, regret, and post-purchase behaviors as well as in understanding how to design recovery efforts to manage the ensuing emotions. Nonetheless, appraisal theories of emotions assert that we interpret our experiences along multiple dimensions and experience a variety of emotions. Future research could explore and categorize the different dimensions along which consumers appraise a service failure and examine whether other emotions (disappointment, self-pity, anxiety, pride) mediate the effect of appraisals on conciliatory behaviors. Then, one could develop and test interventions to increase conciliatory behaviors by consumers.

Additionally, future research may want to examine different aspects of anger and regret. Our finding that anger, but not regret, mediates the effects of appraisals on retaliatory behaviors may be due to the fact that we measured anger toward the service provider and regret toward the menu choice. The effects for regret may have been stronger if we had assessed it at the level of the choice to utilize the service provider. This argument is weakened because we found that regret played an important role in conciliatory behaviors, but the topic deserves additional research.

Because we employed two very different methodologies—a laboratory study and a content analysis of web complaints—each of our studies contributes unique insights. However, future research should continue to explore the relationships between appraisals, interventions, emotions, and post-purchase behaviors in more externally valid contexts.

Footnotes
1

In a report, an anonymous consumer posts a comment describing an experience at a specific restaurant. Only one report described an incident which the consumer thought was their fault. As a result, we eliminated this observation, giving us 299 reports, all services failures in which the consumers blamed the service provider.

 

Acknowledgments

This work is based on Carolyn Bonifield's dissertation. We thank the members of that committee, especially Irwin Levin and Baba Shiv, for their extensive comments. In addition, we would like to acknowledge the helpful suggestions from the reviewers and editors at Marketing Letters, especially those of Professor Charles B. Weinberg.

Copyright information

© Springer Science+Business Media, LLC 2006