Consumer Complaining and Word of Mouth Activities: Field Evidence


Steven P. Brown and Richard F. Beltramini (1989) ,"Consumer Complaining and Word of Mouth Activities: Field Evidence", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 9-16.

Advances in Consumer Research Volume 16, 1989      Pages 9-16


Steven P. Brown, University of Texas at Austin

Richard F. Beltramini, Arizona State University

Complaining and the spreading of negative word of mouth as alternative responses to consumer dissatisfaction have attracted considerable scholarly attention (e.g., Westbrook 1987; Richins 1983; Folkes, Koletsky and Graham 1987; Folkes 1984). Research has investigated whether the same determinants predict both complaining and word of mouth (Richins 1983), what cognitive mechanisms lead to the selection of one response over the other (Richins 1983), what types of causal attributions lead to complaining and negative word of mouth (Folkes 1984; Folkes, Koletsky and Graham 1987; Richins 1983), and the role affect plays in complaining and/or negative word of mouth (Westbrook 1987, Folkes, Koletsky and Graham, 1987).

One difficulty with complaining and negative word of mouth research has been its reliance on retrospective self-reports for measurement of the constructs of interest (Richins 1983; Folkes 1984; Westbrook 1987). This technique asks subjects to remember a dissatisfying experience and, relying on their memory, to complete measures assessing the cognitive and affective dimensions of the experience which caused them to complain or to spread negative word of mouth. While this approach is ecologically valid in the sense that it taps into a real experience, it also creates potential sources of bias in subjects' responses.

First, asking respondents to report on past experiences may not give an accurate portrayal of the factors which caused them to respond as they did when the dissatisfying was temporally salient. Forgetting and abatement of affect caused by the experience may bias retrospective measures.

Second, asking survey respondents to reflect on a dissatisfying experience with a broadly defined product category (e.g., clothing items or appliances) is likely to elicit a great diversity of contexts within which those experiences took place. That is, each respondent reports measures based on a different experience which occurred in a context unique to that experience.

The present study investigates complaining and negative word of mouth in a naturalistic setting under conditions of problem salience. It measures constructs related to complaining and word of mouth at the time of problem occurrence, and measures the reactions of all study participants to the same problem. The study reexamines the question raised by Richins (1983) of whether the same variables affect both word of mouth and complaining and extends her work by providing a measure of the extent of word of mouth activity. This research introduces perceived inconvenience as a variable related to responses to dissatisfaction and tests it as a possible mediator of the relationship between problem severity and response. It also measures hypothesized determinants of complaining and negative word of mouth diachronically to gain insights into the behavior of these variables over time and into the question of whether dissatisfaction and responses to it vary differentially across levels of severity over time.

While level of involvement has not been systematically varied to date in studies of consumer response to dissatisfaction, the present study captures the-effects of problem severity, inconvenience, attributions or controllability of the cause of product failure, and perceived responsiveness of management s a highly involved consumption situation.


Richins (1983) has investigated whether the same variables which affect complaining also affect word of mouth. Using retrospective self-reports she found evidence among consumers who had problems with clothing items or appliances that the severity of the problem and the attribution of blame for the problem influence the amount of effort spent seeking redress. The consumer's perception of the marketer s responsiveness, in turn, influences the form of response chosen.

Richins found a positive correlation between management responsiveness and complaining and a negative one between management responsiveness and word of mouth. This suggests that the more responsive the dissatisfied consumer perceives management to be, the more likely complaining is to be chosen as a response. Similarly, the less responsive the consumer perceives management to be, the more likely word of mouth is to be chosen.

Richins attributes this positive correlation between perceived management responsiveness and complaining to a cognitive process in which the consumer compares the effort expended in complaining with the expected return on that effort. The more responsive management is perceived to be, the higher the ratio of expected return to effort expanded, and the more likely the consumer is to consider the effort worthwhile.

Folkes and her colleagues (Folkes 1984; Folkes, Koletsky and Graham 1987) investigated complaining behavior from the perspective of attribution theory. They found that attributions of causal locus for a product failure, controllability of the cause of failure, and the likelihood of recurrence influence the incidence of complaining. Folkes, Koletsky and Graham (1987) developed a model of complaining which includes the affective variable anger as a mediator between attributions of causal locus, controllability, and stability (i.e., likelihood of recurrence) and the intention to complain or to repurchase from the same vendor Supporting evidence for the model came from a field investigation of airline passengers awaiting delayed flights. Folkes (1984) provided other empirical evidence of the importance of attributions and of the affective dimension in complaining from a study using retrospective self-reports of unsatisfactory restaurant experiences. Westbrook (1987) has also provided strong evidence of the importance of negative affect on complaining and word of mouth by showing that the effect of negative affect is not completely mediated by disconfirmation and satisfaction/dissatisfaction constructs.



The first research objective is to examine the effects of problem severity, perceived controllability of Fe cause of the problem, and perceived management responsiveness on complaining and word of mouth under conditions of problem salience. If there are no differences in study participants' response styles between retrospective self-reports and problem-salience self-reports, results similar to those reported in studies using retrospective methodologies (Richins 1983; Folkes 1984) would be expected. Hypothesis one (H1) states these expectations.

H1) Problem severity, attribution of controllability of the cause of the problem, and perceived management responsiveness should correlate with complaining in the following manner:

a) problem severity should correlate positively with both complaining and word of mouth.

b) attribution of controllability of the cause of the problem should correlate positively with both complaining and word of mouth (i.e., the more controllable one perceives the problem to be, the more likely one is to engage in complaining and/or word of mouth).

c) perceived management responsiveness should be positively related to complaining and negatively related to word of mouth.

In addition to the variables which have been previously examined as predictors of complaining and word of mouth, inconvenience is introduced as a potential mediator between problem severity and the response variables. The inconvenience construct is posited to be closely related to but discriminably distinct from problem severity.

H2) Perceived inconvenience resulting from - product failure increases as the severity of the problem increases and serves as a mediator between problem severity and complaining and word of mouth, (Figure 1).

Finally, an exploratory test is made into the behavior of variables related to response to dissatisfaction over time. This test examines whether perceptions of the inconvenience caused by product failure will change differentially across levels of problem severity. No a priori hypothesis is offered regarding this exploratory test.


The present study investigates complaining and negative word of mouth among residents of a 417-unit apartment complex. Dissatisfaction arose among the residents when failure of the main gas line into the complex disrupted gas service partially or totally to 113 of the units. The failure affected units within the complex differently, with 26 losing all gas services (i.e., heating, stove gas, hot water), 87 losing heating and stove gas but not hot water, and 304 losing nothing in the way of gas services. [The broken gas line belonged to the apartment complex. The actual cause of the break in the line (children playing had punctured it with a large rock) was never communicated to the residents by the management and was generally unknown to residents.]

The different ranges of gas services lost by different units provides a natural experimental manipulation of problem severity. Those deprived of all gas services constitute a "severe" group, those who lost heat and cooking but not hot water form a "moderate" group, and those who remained unaffected form a pool from which a control group was randomly chosen.

Of a total of 25 buildings in the complex, six buildings contained the 113 affected units. In the six buildings affected, all units suffered the same level of service disruption. Because all units within each particular building of the complex were affected similarly, randomization with respect to problem severity was not possible. Demographic variables, including length of residence, number of adults, and number of children occupying the unit, were subjected to analysis of variance across levels of severity. No significant differences resulted, indicating considerable homogeneity across the severity groupings. All units within the complex were constructed at the same time, have similar maintenance histories, and according to the apartment management, have experienced similar turnovers rates during the thirteen years of the complex's existence.

This field investigation counterbalances the difficulty of randomization with regard to problem severity with the corresponding benefit of high ecological validity (Brunswik 1956; Cook and Campbell 1975). The time measurement of determinants of and responses to dissatisfaction under highly salient and highly involving product failure conditions sacrifices a small degree of internal control in order to provide a test of these phenomena in a naturalistic setting.

Shortly after the onset of-the loss of gas services, residents of each affected units were approached and asked to complete a questionnaire. A total of 163 questionnaires was distributed among the three levels of problem severity: 26 severe, 87 moderate, and 50 control. [Because the number of unaffected units (304) greatly exceed the number of affected units (113), a random sample of 50 of the unaffected units was deemed sufficient.]

One hundred three responses were obtained by repeated call backs, resulting in a 63% response rate. The severe group yielded 17 responses (65% response), and the control group 25 (50%).

The questionnaire included items to measure (as dependent variables) whether the study participants had complained to management about the loss of gas services (yes/no), whether they had engaged in word of mouth activities regarding the loss of gas services, and, if they had engaged in word of mouth activities, how many people had they spoken to about it.

Independent variables included problem severity (a three-level variable determined by number of gas services lost), perceived inconvenience, perceived controllability of the cause of the problem, and perceived management responsiveness. The inconvenience measure consisted of a single item on a seven-point bipolar scale anchored by "not inconvenienced at all" and "extremely inconvenienced." Controllability of the cause consisted of two items on seven-point bipolar scales which asked study participants how likely they thought the cause of the gas service disruption was poor maintenance (very likely--not at all likely) and how likely they thought the cause was management error (very likely--not at all likely). A test of the internal consistency of these two measures yielded a coefficient alpha of .75.

Perceived management responsiveness was measured by three seven-point bipolar items relating to management's communication effectiveness, concern with residents' well-being and effectiveness in responding to residents' needs during the loss of services. Coefficient alpha for these three items was .80.

Questionnaires were distributed before gas services were restored, and all responses were obtained within a week of restoration of services, after the affected units had been without gas services for eight days. [The time of the onset of the failure and of the restoration of service was the same for all units affected.] Thus the product failure event was highly salient to the affected study participants at the time of the study.

Two months after the first study, a second questionnaire was distributed to remeasure the same constructs well after the problem had been resolved to assess changes in perceptions. The questionnaire was delivered to the 103 units which had responded to the first questionnaire and collected over a one-week period with repeated personal call-backs. In nine of the units, respondents to the first survey had moved during the two-month interval between surveys. Of the remaining 94 units, 85 (90%) provided responses to the second survey.


The correlation matrix in Table 1 reveals generally high correlations between the independent and dependent variables and generally low ones between pairs of independent variables. The one notable case of correlated independent variables occurs for severity and inconvenience (r = .69). A strong positive correlation between these two variables was predicted by H2. However, the two constructs, though closely related, appear to be distinguishable, as an F test for significance of gain in H2 by adding inconvenience to the regressions proved highly significant for all three models (Kerlinger and Pedhazur 1973).

To test the hypotheses, multiple regressions were conducted as reported in Table 2. Problem severity, attribution of controllability of the cause, and perceived management responsiveness were entered into the regression models as independent variables. Then perceived inconvenience was added as a covariate to test its role as a mediator. If its addition caused the beta coefficient for problem severity to drop from significance to nonsignificance, evidence of inconvenience's mediating role as predicted in H2 would be provided (Turnbull, Insko, and Yandell 1974; Westbrook 1987).

In the extent of word of mouth model, parameter estimates for problem severity, attribution of controllability of the cause and perceived management responsiveness all proved significant as predicted in H1. In the existence of word of mouth model, however, only the parameter estimate for problem severity was significant.

In the complaining model, the parameter estimate for problem severity was significant. The parameter estimate for management responsiveness reached statistical significance, but in the opposite (negative) direction from what was predicted in H1c, suggesting that the less responsive management is perceived to be, the more likely complaining is to occur. This result leads to rejection of H1c for the complaining model and is discussed below. The significant negative beta coefficient for management responsiveness in the extent of word of mouth model accords with H1c.







Addition of inconvenience as a covariate in the complaining model caused the beta coefficient of problem severity to drop to non-significance, suggesting that in convenience does mediate the problem severity-complaining relationship. The word of mouth results furnish no support for the mediating role of inconvenience. Parameter estimates for both problem severity and inconvenience showed significance, suggesting that both variables influence word of mouth directly. Figure A presents path coefficients for the complaining and word of mouth models.

The attribution of controllability of the cause, according to H1b, was expected to vary positively with both complaining and word of mouth. Only in the extent of word of mouth model does the parameter estimate for controllability achieve significance. Thus, the extent of word of mouth model provides limited support for H1b, while the existence of word of mouth and complaining models fail to provide support.

Table 3 shows the differences between measures of inconvenience, attributions of controllability of the cause of the problem, and perceived management responsiveness taken at the time of the problem occurrence and then repeated two months later. ANOVA was conducted using perceived inconvenience, attribution of controllability, and perceived management responsiveness successively as dependent variables and problem severity as the independent variable. Analyses compared means resulting from the first measurement, the second measurements and the mean difference between the first and second measurements. Significant differences resulted between levels of problem severity for both the first and second measures of perceived inconvenience, but the change in perceived inconvenience between the two measures did not differ significantly. No other analyses yielded significant differences.



Thus the results of the exploratory test into the behavior of variables related to response to dissatisfaction over time show very little change in the perception of inconvenience as the problem situation fades into the past. The change in perception of inconvenience did not differ across levels of problem severity. One possible explanation for these results is that the loss of gas services was sufficiently highly involving that the affective response to it was long-lasting. Study participants may also have tried to respond consistently across the two studies.


The results of this study include one striking departure from reports of previous studies into complaining and word of mouth behavior -- i.e, the negative correlation between complaining and perceived management responsiveness. While several possible explanations might be offered for the strong difference found in this relationship between the present study and previous studies (Richins 1983) (e.g., differences in operationalization of constructs, differences in product context studied, differences in study participants' level of involvement with the product category investigated) perhaps the most obvious explanation is that people tend to respond differently depending on whether the problem is temporally salient or long since past.

Consumer's perceptions at a time when they have just undergone considerable stress and inconvenience are likely to differ from those of consumers who calmly reflect on past events. Particularly, it seems reasonable that the former would be more likely to respond in an affective (i.e., angry) manner, rather than engaging in a calculation of the expected return on effort expended in complaining. Thus, salience of the product failure is likely to bc an important determinant of consumer response to dissatisfaction. A hypothesis meriting investigation is whether an initially affective response to dissatisfaction later gives way to a cognitive response.

The product failure studied here involved individuals' living arrangements and essential domestic facilities for which they exchange a substantial proportion of their monthly income. Disruption of use patterns of such basic facilities would likely arouse a stronger affective response than failure of a less essential product or service. Thus, level of involvement may in part account for the obtained negative relationship between perceived management responsiveness and complaining.

The correlational analyses presented here do not provide a basis to conclude that problem severity, perceived inconvenience, attributions of controllability of the cause of the problem, or perceived management responsiveness cause complaining and word of mouth. Thus, the negative relationship found between complaining and perceived management responsiveness could be attributable to -the possibility that study participants who complained might not have received prompt action in response and hence perceived management to be unresponsive. his possible explanation is rendered somewhat less likely in consideration of the fact that 37% (32) of the study participants affected by the gas problem reported complaining, while only 15% (12) of them rated management responsiveness below four on a seven-point scale (which was the average of three items ).

Previous findings have posited both cognitive (Richins 1983) and affective (Westbrook 1987; Folkes, Koletsky and Graham 1987) interpretations of complaining. The cognitive interpretation predicts consumers will be more likely to complain when they perceive the marketer to be responsive to such complaints. The results of this study do not confirm this prediction.

Supporting evidence for the cognitive interpretation has come from retrospective self reports, in which consumers reflect on past dissatisfactions. It seems likely that reliance on that type of methodology introduces a cognitive bias into responses, as the passage of time allows for the abatement of negative affect aroused by a product failure. Self reports are also used in the present study, but here they reflect study participants' reactions at the time of problem occurrence.

The present research also removes another potential source of error variance by measuring all study participants' responses to the same product failure. In retrospective self report surveys, most or all of the study participants respond to dissatisfying events which are unique to their own experience.

This study has also revealed a considerable degree of heterogeneity between predictive models of complaining and word of mouth, suggesting that the same set of correlates does not necessarily predict both responses to dissatisfaction. Of the variables tested, problem severity and inconvenience were the best predictors in the word of mouth models, with attribution of controllability of the cause of the problem also proving significant in the extent of word of mouth model. These results suggest that problem severity alone is sufficient to generate negative word of mouth activities, but that the extent of those activities is a function of the extent to which the consumes perceives the cause of dissatisfaction to be under management's control and the degree of responsiveness which the consumer attributes to management.

In the complaining model, inconvenience exerted the strongest influence on the criterion and acted as a mediator of the problem severity-complaining relationship. Perceived responsiveness of management also accounted for a significant amount of variation, but its negative sign ran counter to expectations as discussed above.

The limitations of this study include the difficulty of maintaining strict controls in a naturalistic setting. It was not possible, for example, to randomize with respect to the problem severity variable. Problem severity, however, figured as only one of a number of independent variables, and a check of demographic variables indicated considerable homogeneity among the groupings. In a dynamic situation, it is also difficult to manipulate systematically the variables under examination. Measures of study participants perceptions of management responsiveness are taken at a point in time during a continuous process of interaction between management and the participants. Perhaps, as Sherry (1984) suggested, a mix of qualitative and quantitative techniques may be needed in future research efforts to capture the richness of this oral tradition.

In spite of these limitations, the present study employs a methodology (problem-salience self-reports) which contrasts with the retrospective self-report method of data collection which has characterized previous response to dissatisfaction research, and a different result was obtained. It is argued that using retrospective self-reports may tend to bias responses toward a cognitive interpretation of complaining. The study shows that the same correlates do not necessarily predict both complaining and word of mouth. It also shows that perceived inconvenience is highly related to both complaining and word of mouth and suggests that problem salience may have important effects on the form of response to dissatisfaction.


Brunswik, Egon (1956), Perception and the Representative Design of Psychological Experiments, Berkeley and Los Angeles: University of California Press.

Cook, Thomas and Donald Campbell (1975), "The Design and Conduct of Experiments and Quasi-Experiments in Field Settings," in Handbook of Industrial and Organizational Research, ed. Martin Dunnette, Chicago: Rand McNally & Co.

Folkes, Valerie S. (1984), "Consumer Reactions to Product Failure: An Attributional Approach," Journal of Consumer Research, 10 (March), 398-409.

Folkes, Valerie S., Susan Koletsky, and John L. Graham (1987), "A Field Study of Causal Inferences and Consumer Reaction: The View from the Airport," Journal of Consumer Research, 13 (March), 534-539.

Insko, Chester A., W. Turnbull, and B. Yandell (1974), "Facilitative and Inhibiting Effects of Distraction on Attitude Change," Sociometry, 34, 508-528.

Kerlinger, Fred N. and Elazar J. Pedhazur (1973), Multiple Regression in Behavioral Research, New York: Holt, Rinehart and Winston.

Richins, Marsha L. (1983), "Negative Word-of-Mouth by Dissatisfied Consumers: A Pilot Study," Journal of Marketing, 47 (Winter), 68-78.

Sherry, John F., Jr. (1984), "Some Implications of Consumer Oral Tradition for Reactive

Westbrook, Robert A. (1987), "Product/Consumption-Based Affective Responses and- Postpurchase Processes," Journal of Marketing Research, 24 (August), 258-270.



Steven P. Brown, University of Texas at Austin
Richard F. Beltramini, Arizona State University


NA - Advances in Consumer Research Volume 16 | 1989

Share Proceeding

Featured papers

See More


D3. Social Exclusion and WOM about Past versus Future Experiences

Melis Ceylan, Koc University, Turkey
Ezgi Akpinar, Koc University, Turkey
Selin Atalay, Frankfurt School of Finance and Management, Germany

Read More


Consuming Products with Experiences: Why and When Consumers Want Mementos

Charlene Chu, Chapman University
Suzanne Shu, University of California Los Angeles, USA

Read More


Who Gets Credit? Who Gets Blame? The Role of Agency in Ethical Production

Neeru Paharia, Georgetown University, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.