An Investigation of the Interrelationship Between Consumer (Dis)Satisfaction and Complaint Reports

ABSTRACT - The relationships between satisfaction, its antecedents, behavioral intention, and self-reports of complaint activities were examined to estimate the degree to which complaining correlates with states of dissatisfaction in an hypothesized causal structure. Results of a maximum likelihood analysis showed that the basic expectancy disconfirmation (dis)satisfaction motel was supported and that complaining is related to lower levels of satisfaction, but at a relatively modest order of magnitude. The data also suggest that the relation between complaining and subsequent satisfaction and conation cognitions requires further specification.

Citation:

Richard L. Oliver (1987) ,"An Investigation of the Interrelationship Between Consumer (Dis)Satisfaction and Complaint Reports", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 218-222.

Advances in Consumer Research Volume 14, 1987      Pages 218-222

AN INVESTIGATION OF THE INTERRELATIONSHIP BETWEEN CONSUMER (DIS)SATISFACTION AND COMPLAINT REPORTS

Richard L. Oliver, University of Pennsylvania

ABSTRACT -

The relationships between satisfaction, its antecedents, behavioral intention, and self-reports of complaint activities were examined to estimate the degree to which complaining correlates with states of dissatisfaction in an hypothesized causal structure. Results of a maximum likelihood analysis showed that the basic expectancy disconfirmation (dis)satisfaction motel was supported and that complaining is related to lower levels of satisfaction, but at a relatively modest order of magnitude. The data also suggest that the relation between complaining and subsequent satisfaction and conation cognitions requires further specification.

INTRODUCTION

Consumer satisfaction as an outcome of a purchase/usage experience would appear to be an important variable in the chain of purchase experiences linking product selection with other post-purchase phenomena including favorable word-of-mouth and consumer loyalty. Conversely, the study of dissatisfaction is equally important because of its close ties to negative word-of-mouth, complaining, and other redress-seeking (e.g., Richins 1980). The present study focuses on a subset of these relationships, specifically with regard to the satisfaction-complaining linkage. This issue has been subject to some scrutiny because of its relevance to managerial strategies based on the incidence of complaining (Bearden and Teel 1983; Fornell and Westbrook 1984).

Generally, it has been found that fewer consumers complain than would be expected from expressed levels of dissatisfaction (Ash 1980; Day and Bodur 1978; Day and Ash 1979; Leigh and Day 1979). Bearden and Teel (1983) have reviewed this and other literature and find that the percent of dissatisfied individuals who complain is discouraging" because it: (a) prevents the consumer from rectifying an unpleasant purchase experience, (b) limits firm exposure to marketplace problems, and (c) precludes effective policy-making on the basis of complaint data.

The present study was designed to further understanding of the satisfaction-complaining relationship in three areas. First, additional ewidence on the magnitude of the association between satisfaction and complaining is provided. Second, a partial replication of the Bearden and Teel (1983) complaining motel is reported. Third, complaining or non-complaining is related to reports of subsequent satisfaction. The first objective builds on the works of Bearden and Teel and Gilly (1981) who have provided estimates of the satisfaction-complaining relationship. Generally, they find that this parameter is on the order of -.4. The second and third objectives will be accomplished through a causal motel test of the expectancy disconfirmation framework used in Bearden and Teel. This conceptualization, based on the work of Oliver (1980), posits that satisfaction/dissatisfaction is a joint function of expectations and the disconfirmation of those expectations. Satisfaction, in turn, is posited to directly affect the incidence of complaining. Complaining, in turn, is thought to affect future satisfaction through its effect on satisfaction with complaint resolution (cf. Bearden and Oliver 1985). Each of these objectives and its relevance is further elaborated below.

The Satisfaction-Complaining Relationship

Because satisfaction and dissatisfaction can be viewed as mutually exclusive locations on a satisfaction continuum and because the act of complaining can be represented as a zero-one activity, the relation between the two constructs is easily displayed in a two-by-two table. Those cells of greatest interest are the dissatisfied complainer and the dissatisfied non-complainer categories. Note that, without base-rate data from the population, the former group gives no indication of the size of the latter. In addition, there is ewidence that complainers can include satisfied customers voicing various minor concerns as well as non-consumers who perceive benefits, some fraudulent, to contacting the manufacturer (Jacoby and Jaccard 1981).

Two statistics would be helpful in summarizing the (dis)satisfaction-complaint relationship. The first is the biserial correlation between expressed satisfaction and complaining which, unfortunately, masks the satisfied/dissatisfied proportions in the population. The contingency and phi coefficients (Siegel 1956) would be more suitable for this analysis because they capture the content of the four cells in the two-by-two breakdown through their chi-squared component. Thus, large proportions of the off-diagonal satisfied complainer and dissatisfied non-complainer categories would lower the correlation to a greater degree than they might if a biserial relation were computed.

The greatest volume of data has been reported for two of i the four satisfaction-by-complaint cells. These data, summarized in Day, Grabicke, Schaetzle, and Staubach (1981), show that the percentage of complaints and other channel-directed redress seeking among dissatisfied customers ranged from a low of 23% to a high of 39% over wide range of durables, nondurables, and services. Complaints directly to the manufacturer ranged from 2.7% to 5.4%. Other estimates (Warland, Herrmann, and Willits 1975; Office of Consumer Affairs 1976; A. C. Nielsen Company 1979) are equally low and suggest that manufacturers should not rely on voluntary complaints for product feedback. Interestingly, the most frequent response to dissatisfaction in the Day et al. summary was to -do nothing." The present study will seek to corroborate these figures in a context differing from that previously investigated.

A Satisfaction-Complaining Framework.

The structure tested here is based on the generally held belief that dissatisfaction is a necessary cause for complaining to occur. While many theorists would agree with the central concept that dissatisfaction is a predisposing factor for complaining, most would qualify this proposition to include other facilitating conditions involving personality, attitudinal, and demographic factors as well as situational effects (Jacoby and Jaccard 1981). Nevertheless, satisfaction can be shown to be correlated with complaining behavior in the absence of other measured factors (Bearden and Teel 1983).

The model to be tested here, following Bearden and Teel (1983) and Oliver (1980), is shown in the Figure. As is traditional in expectancy disconfirmation paradigms, initial satisfaction is shown as a joint function of expectations and the disconfirmation of those expectations. Complaining, in turn, is thought to be a function of (dis)satisfaction only. The model, as shown, deviates from prior works in that it suggests a second, post-complaining, level of satisfaction or "secondary satisfaction." This derives from Bearden and Oliver (1985) who argued that the act of complaining itself can enhance satisfaction because it increases the likelihood that redress will occur. Alternatively, it often acts as a catharsis for pent-up frustration. Bearden and Oliver were careful to point out, however, that this applies only to direct ( public") complaining and not to complaints directed at friends and acquaintances- Finally, intention to rebuy is shown as a function of secondary satisfaction.

These relationships, which constitute the hypotheses tested in this study, are important in assessing the nomological as opposed to the simple relationship between satisfaction and complaining. If complaining and satisfaction are related, complaining should be related to the network of constructs in which satisfaction is embedded. The discussion which follows elaborates on a field setting in which a test of these hypotheses was made.

FIGURE

MODEL OF HYPOTHESIZED LISREL EFFECTS

METHOD

Sample and Design

Subjects were all first year MBA candidates in a two-year accredited university program (n - 107). The students represented a diverse background with an age range of 21 to 39 and a mean of 25. One-third of the class was female and 35% had an average of 2.6 years of work experience. Of these subjects, 66 (62%) participated in the study, a response rate comparable to that of voluntary questionnaire administrations generally. The self-reported demographic data showed that there were no significant differences in the profile of respondents and the population statistics noted above.

Because of limitations imposed by the school administration, a retrospective single survey approach relating to students' experiences in the first semester of the first year was used at the beginning of the second semester. Retrospective research shares commonalities with life-history and clinical analysis approaches and is necessitated when certain pieces of data cannot be collected as they emerge. It differs from cross-sectional methods in that subjects are instructed not to respond as they feel at the present, but to reflect back on their cognitions and emotions at some prior time. Obvious drawbacks to this approach are memory fallibility and justification mechanisms where subjects distort past events to model today's outcomes. The first limitation is lessened by the fact that a four month time interval was the maximum required for any retrospective measure. The second drawback is a more serious problem; limitations are noted in the discussion.

Measures

Focus group discussions among both first and second year MBA students as well as interviews with administrative I staff connected with the program yielded a list of 27 t factors thought to be important to the satisfaction of MBA , candidates generally. The factors included various dimensions ranging from the quality of the faculty to the location of the university. This same list vas used for the attribute measures of expectation, disconfirmation, and satisfaction.

To obtain these measures in retrospective fashion, subjects were asked to reflect back to when they entered f the program and to score their expected satisfaction from each attribute on 7-point scales ranging from expected to be very satisfied-' to 'expected to be very dissatisfied.' The expected satisfaction approach is discussed in Oliver (1981) as a valid alternative to attribute rating scales when performance-level diagnostic data are not required. The attribute-specific expectation measure vas formed as the sum of the 27 individual item responses.

An overall measure of expectation was obtained on a 12-item semantic differential of evaluative adjectives. Subjects were asked to answer the lead-in, Last fall, I thought the MBA program would be, with responses to 7-point items such as "useful, useless,-- -good decision, bad decision," and "good for me, bad for me.

Attribute specific disconfirmations were measured on "better-worse than expected" scales (Oliver 1980). The 7-point descriptors ranged from "much worse than I thought | it would be through ' just as expected to much better than I thought it would be. As with expectation, the 27 disconfirmation items were summed to form the attribute index. An overall disconfirmation measure was comprised of four summary items in "better-worse than" format. One reflected the benefits and advantages of the program, a second concerned the problems and disadvantages, while the remaining two referred to an overall disconfirmation impression.

Attribute-specific satisfaction was obtained as the sum of satisfied-dissatisfied responses to the 27 dimensions as before. Subjects were asked to respond to these questions with reference to their first semester experience and not to their feelings at the present time. An overall retrospective measure was constructed of five items recommended in Westbrook and Oliver (1981). One was a direct seven-point overall satisfaction-dissatisfaction measure, two others reflected feelings of pleasure-displeasure and happiness-unhappiness about the program, one was a %-satisfied scale, and the last vas the Andrews and Withey (1976) pie scale containing varying numbers of positive (satisfying) and negative (dissatisfying) wedges. As with the attribute level scales, all overall items were worded in the past tense to reflect the prior semester's experience.

Complaining activity was obtained in a special section of the survey for those subjects who voiced complaints during the first semester. Pocus group results showed that complaints were predominantly directed to either an administrator (dean or associate dean), a faculty member, or the MBA student advisor (a second year MBA selected to provide counsel to the incoming first year class). Subjects indicated their complaining activities by responding to a series of questions concerning the outcomes of complaints to either of the three entities no ted above.

For a discrete set of complaint activities to be used as multiple indicators in maximum likelihood estimation, binary coding was used to represent the presence or absence of a complaint to each entity. Although no indication of the intensity or duration of the complaint or of the ease with which it could be made is reflected in this measure, subjective reports of this nature to product dissatisfaction are the most common form of obtaining this information (Day et al. 1981; Bearden and Oliver 1985). Readers will recognize this as a limitation of the study.

Secondary or concurrent satisfaction was measured in the later sections of the questionnaire where students were asked to respond in terms of their feelings about the program nov that the second semester was beginning. To insure further that students would view the second satisfaction measure as distinct from that used to reflect satisfaction during the first semester, a different scale was needed. To accomplish this purpose, Westbrook and Oliver's (1981) 12-item Likert-type instrument with reported alpha reliabilities of .75 to .96 vas used. Bearden and Teel (1983) used variations on this instrument and report similarly high reliabilities. Sample items include: "I'm sure it vas the right thing to enroll in this MBA program" and "This MBA program hasn't worked out as well as I thought it would."

Finally, intention was measured in a manner similar to Bagozzi's (1982) three-item "chances in ten" scale reflecting the student's willingness to re-enroll for the second year. The first item vas a direct "Will you re-enroll...?" question. The second asked "If you had it to do all over again, would you enroll...?" The third asked -If you could enroll in another MBA program without loss of credits. would you still re-enroll here?"

Alpha reliabilities were obtained for all scales as follows. Attribute and overall expectations yielded estimates of .91 and .87; attribute and overall disconfirmation yielded estimates of .89 and .91; attribute and overall satisfaction yielded estimates of .93 and .98; secondary satisfaction generated an estimate of .95; and intention resulted in a coefficient of .84. As can be seen, all estimates are in the acceptable range of .8 and above (Nunnally 1978). No estimate was calculated for the complaint index as the three complaint behaviors were discrete events and not coherent indicators .

Analysis

Joreskog and Sorbom's (1983, Bagozzi 1980) maximum likelihood estimation technique (LISREL) vas used to test the underlying causal process shown in the Figure. As opposed to OLS estimation, maximum likelihood techniques do not assume that variables are measured without error, that they are pure measures of their constructs, or that the structural equations are specified correctly.

Two problems were apparent, however, with regard to the use of this technique on the data and sample used here. The first concerned the binary coding for the complaint data. When observed variables have been measured on ordinal or nominal scales, it is recommended that polychoric and polyserial correlations be computed and estimated with unweighted least squares (ULS) in LISREL. However, standard errors are not available for this procedure. As the complaint data were binary, estimates from the ULS procedure were initially examined. Because these findings were very similar to the maximum likelihood estimates and because the latter provides standard errors and t-values, the maximum likelihood results are shown. The second problem concerned the small sample size obtained. In such cases, the chi-squared goodness of fit test in LISREL is not appropriate. Thus, as with all model tests, the fit of the model is subject to interpretation and to one's willingness to infer fit from the individual parameter estimates.

RESULTS

Elementary Statistics

Ewidence relating to the basic relationship between satisfaction and complaining is shown in Table 1. The complaint dimension vas defined as either a complaint to any of the three parties identified in the survey or no reported complaining behavior, while the satisfaction/ dissatisfaction split was defined by the theoretical point of indifference on the overall satisfaction scale. The simple correlation between the two variables vas -.38, the contingency coefficient was -.32, and phi was -.34. Note that all converge to approximately -.35 and that no estimate is particularly high.

TABLE 1

REPORTS OF COMPLAINING BY SATISFACTION/DISSATISFACTION

In addition, the data show the following conditionals: The probability of a voiced complaint if dissatisfied was 62%; the probability of a dissatisfied subject given a registered complaint was 59%; and the probability of a basically satisfied student complaining was 28%. Each of these percentages compares to 1, 1, and O respectively if one were to take a "purist's vie;" of the satisfaction-complaining relationship.

Model Relationships

The 78 subdiagonal correlations on which the model estimates are based are available on request and are not shown here due to space restrictions. Generally, the attribute and overall measures of the expectation, disconfirmation, and satisfaction measures are highly correlated as one would expect. The complaining index components were negatively and significantly correlated (p < .01) with satisfaction as hypothesized and were correlated to a lesser extent (1 < .05) with secondary satisfaction and two of the three intention indicators. Finally, the intention indicators were highly correlated (n < .01) with secondary satisfaction as hypothesized.

The LISREL results corresponding to the measurement model in the Figure are shown in Table 2 while the results for the structural motel are shown in Table 3. The data show that all non-constrained elements were significantly related to their constructs. (The constrained elements were selected to be the most significant.)

The LISREL path estimates in Table 3 show that all structural parameters in Figure 1 were supported with the exception of B3 indicating that secondary satisfaction was not statistically related to the extent of complaining behavior although the coefficient was in the expected direction. The remaining paths were supported as hypothesized. As shown in Bearden and Teel (1983) and Oliver (1980), satisfaction is a function of both expectation and disconfirmation. Very high coefficient estimates were also obtained for the satisfaction-secondary satisfaction and secondary satisfaction-intention paths.

The resulting chi-square value of 108 (df - 60, p < .01) does not permit one to reject significant differences in the residual matrix. However, Joreskog and Sorbom (1982) note that chi-squared is an appropriate statistic only when all observed variables are multivariate normal, when the analysis is based on the covariance matrix, and when the sample is fairly large. Because these conditions did not apply here, only the individual structural equations are examined. Specifically, the squared multiple correlations were very high for primary satisfaction (R2 = .917), secondary satisfaction (R = .880), and intention (R2 = .826). As vas ewident from the earlier analysis of the complaining relationships, the complaint construct was not predicted well (B2 = .162). Thus, it appears that the satisfaction elements of the model were adequately specified whereas complaining was not.

TABLE 2

PARAMETER ESTIMATES FOR THE MEASUREMENT SUBMODELS

TABLE 3

PARAMETER ESTIMATES FOR CAUSAL PATES

DISCUSSION

With regard to the relationship between satisfaction and complaining, the study findings reveal that the proportion of complainers (41%) was a fairly accurate estimate of those dissatisfied (39%), although these two groups were not comprised of the same subjects. The data revealed that the overlap between the two groups vas 60%. This lower than anticipated degree of overlap is mirrored in the correlations obtained between satisfaction and complaining. Depending on the metric assumptions used, they range between -.32 and -.38. This compares to -.39 and -.43 in Bearden and Teel (1983), and -.38 in Gilly (1981). If one loosely aggregates these estimates, it appears that correlations in the -.35 to -.40 range appear typical. Taken one step further, this would imply that something on the order of 152 of the variance in complaining is explained by satisfaction/dissatisfaction.

A second issue involved the nomological validity of the complaining construct. This is best answered in the path analysis reported here. The LISREL results demonstrated that complaining vas indeed related to primary satisfaction when viewed in terms of observable variables and multiple indicators. However, the findings also show that the act of complaining vas not related to secondary satisfaction nor was it related to intention in a second LISREL run incorporating a direct path from n2 to n4. This latter analysis is not reported here because the results were nearlY identical to those discussed.

Implications

The implications of the findings fall into the areas of theory and practice. Theoretically, much more work on the complaining construct is required before meaningful statements concerning this phenomenon can be mate. Two observations are apparent from the low degree of shared variance between satisfaction and complaining. First. satisfaction and dissatisfaction probably have many other forms of behavioral response which would not be classified within the complaining category. Day and his colleagues (1981) and Jacoby and Jaccard (1981) have explored this suggestion more fully. Day, in particular, has delineated the many variations (or agenda) of consumer response to dissatisfaction. His studies have shown consistently that complaining is not the modal response category.

Secondly, it is apparent that satisfaction/dissatisfaction is not a sufficient condition for complaining to occur. Indeed, Jacoby and Jaccard (1981) argue that it may not even be necessary. Day (1983) has recently presented a fairly exhaustive set of conditions for the complaint response. These include dissatisfaction or inequity, causal attributions, search for channels, economic and psychological costs and benefits along with their respective subjective probabilities, alternative evaluation, and an action decision. Thus, the ease with which a complaint could be made, although not investigated here, should be an important component of future works. Until a more extensive study involving these suggested antecedents of complaining is made, explained variance estimates of 15% will probably not be exceeded.

The practical implications are more problematic. Because of the wide variation in dissatisfaction and complaining rates across product categories, any estimate of normal complaining must be done on a product by product basis. Thus, the results reported here are only suggestive of what interpretations others may apply to complaining activity.

At the very least, it should be apparent that not all dissatisfied consumers complain not are all complainers dissatisfied. In fact, the available commercial data reported earlier (e.g., A.C. Nielsen 1979) suggest that most companies are confronted with a "tip of the iceberg" problem. and this should be more true to the extent that the market is homogeneous, the product category is low priced, channel members are willing to absorb and "bury" returns and complaints, and "third party" options are available.

Limitations

Although this study vas intended only to be suggestive of approaches to the dissatisfaction-complaining relationship, a number of cautionary observations and some very real limitations of the research must be discussed. Perhaps the overriding shortcoming vas the design imposed by the research environment. A retrospective investigation is suspect on a number of points. Problems of subject recall, rationalization, and distortion are somewhat unavoidable. While a design of this nature is seemingly different from a concurrent study in that subjects are asked to reconstruct a temporal ordering of events, the responses obtained are, in all likelihood, more logically connected than they would have been hat a true longitudinal design been employed. It should also be noted that the responses obtained here are time-dependent and perhaps may have varied hat another time of administration been used.

The validity of the complaining criteria are also suspect. Self-reported behaviors, particularly of a sensitive nature, must also be viewed with a degree of caution. Unfortunately, corroborative reports of actual complaints to faculty and administrators were not available for reasons familiar to most readers.

The response rate, although not unusually low, posed two problems which limit the generalizability of the findings. Specifically, no attitudinal nor complaining data were available for the 41 non-respondents. Thus, no analyses of the satisfaction-complaining relationship for this group could be performed. In fact, the true proportions of dissatisfied subjects and complainers could not be determined and the estimates reported here could seriously under- or overstate these levels. A second limitation resulting from the response rate obtained is the inability to perform a cross-validation. Split-sample techniques would have yielded very unstable estimates given the small n involved. As a result, further replications of the current approach will be needed.

Summary

The basic relationship between satisfaction and complaining was tested within an empirically based satisfaction framework. Although a moderate correlation between satisfaction/ dissatisfaction and complaining was fount, the complexity of the relationship was underscored by the fact that complaining vas not related to secondary satisfaction or intention in multivariate mote. This supports the contention of Day (1983, Day et al. 1981) and others that dissatisfaction has many manifestations other than complaining and that the antecedents of complaining are much more involved that dissatisfaction alone. Researchers are encouraged to approach the complaining construct with much more sophisticated representations than have been used to date.

REFERENCES

Andrews, Frank H. and Stephen B. Withey (1976), Social Indicators of Well-Being, Nev York: Plenum Press.

Ash, Stephen B. (1980), Consumer Satisfaction and Complaining Behavior: Major Findings and Directions for Action. Ottawa: Consumer and Corporate Affairs Bureau of Canada.

Bagozzi, Richard P. (1980), Causal Hotels in Marketing, Nev York: John Wiley & Sons.

Bagozzi, Richard P. (1982), "A Field Investigation of Causal Relations Among Cognitions, Affect, Intentions and Behavior," Journal of Marketing Research, 19 (November), 562-84.

Bearden, William O. and Richard L. Oliver (1985), "The Role of Public and Private Complaining in Satisfaction with Problem Resolution," Journal of Consumer Affairs, 19 (Winter), 222-240.

Bearden, William O. and Jesse E. Teel (1983), "Antecedents and Consequences of Consumer Satisfaction and Complaint Reports," Journal of Marketing Research, 20 (February), 21-28.

Day, Ralph L. (1983), "The Next Step: Col only Accepted Constructs for Satisfaction Research," in International Fare in Consumer Satisfaction and Complaining Behavior, Ralph L. Day and H. Keith Hunt, eds., Bloomington: Indiana University School of Business, 113-117.

Day, Ralph L. and Stephen B. Ash (1979), "Consumer Response to Dissatisfaction with Durable Products," in Advances in Consumer Research, Vol. VI, William L. Wilkie, ed., Ann Arbor: Association for Consumer Research, 438-444.

Day, Ralph L. and Muzaffer Bodur (1978), "Consumer Response to Dissatisfaction with Services and Intangibles," in Advances for Consumer Research, Vol. V, H. Keith Hunt, ed., Ann Arbor: Association for Consumer Research, 263-272.

Day, Ralph L., Klaus Grabicke, Thomas Schaetzle, and Fritz Staubach (1981), "The Hidden Agenda of Consumer Complaining," Journal of Retailing, 57 (Fall), 86-106.

Fornell, Claes and Robert A. Westbrook (1984), "The Vicious Circle of Consumer Complaints," Journal of Marketing, 48 (Summer), 68-78.

Gilly, Mary C. (1981), "A Study of Post-Purchase Attitudes and Behavior of Consumers with Different Complaining Tendencies," in The Changing Marketing Environment: New Theories and Applications, Kenneth Bernhardt, et al., eds.. Chicago: American Marketing Association, 166-69

Jacoby, Jacob and James J. Jaccard (1981), "The Sources, Meaning, and Validity of Consumer Complaint Behavior: A Psychological Analysis," Journal of Retailing, 57 (Fall), 4-24.

Joreskog, Karl G. and Dag Sorbom (1983), LISREL V: Analysis of Linear Structural Relationships by the Method of Maximum Likelihood, Chicago: National Educational Resources, Inc.

Joreskog, Karl G. and (1982), Recent Developments in Structural Equation Modeling," Journal of Marketing Research, 19 (November), 404-16.

LaBarbera, Priscilla A. and David Mazursky (1983), "A Longitudinal Assessment of Consumer Satisfaction/ Dissatisfaction: The Dynamic Aspect of the Cognitive Process," Journal of Marketing Research, 20 (November), 393-404.

Leigh, Thomas W. and Ralph L. Day (1979), "Satisfaction, Dissatisfaction and Complaint Behaviors with Nondurable Products," in Nev Dimensions of Consumer Satisfaction and Complaining Behavior, Ralph L. Day and H. Keith Hunt, eds., Bloomington: Indiana University, School of Business, Division of Research, 170-183.

Nielsen, A. C., Company (1979), "44th Annual Nielsen Review of Retail Grocery Trends," Chain Store Age/ Supermarkets (October), 81-116.

Nunnally, Jus C. (1978), Psychometric Theory, Second Edition, New York: McGraw Hill.

Oliver, Richard L. (1980), "A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions," Journal of Marketing Research, 17 (November), 460-469.

Oliver, Richard L. (1981), "Measurement and Evaluation of Satisfaction Processes in Retail Settings," Journal of Retailing, 57 (Fall), 25-47.

Richins, Marsha L. (1980), "Consumer Perceptions of the Costs and Benefits Associated with Complaining," in Refining Concepts and Measures of Consumer Satisfaction and Complaining Behavior, H. Keith Hunt and Ralph L. Day, eds., Bloomington: Indiana University, School of Business, Division of Research, 50-53.

Siegel, Sidney (1956), Nonparametric Statistics, New York: McGraw-Hill.

Warland, Rex H., Robert O. Herr ann, and Jane Willits (1975), "Dissatisfied Consumers: Who Gets Upset and Who Takes Action," Journal of Consumer Affairs, 9 (Winter), 148-63.

Westbrook, Robert A. and Richard L. Oliver (1981), "Developing Better Measures of Consumer Satisfaction: Some Preliminary Results," in Advances in Consumer Research, Vol. 8, Kent B. Monroe, ed., Ann Arbor: Association for Consumer Research.

----------------------------------------