Consumer Satisfaction/Dissatisfaction: an Outsider's View

J. Edward Russo, University of Chicago
[ to cite ]:
J. Edward Russo (1979) ,"Consumer Satisfaction/Dissatisfaction: an Outsider's View", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 453-455.

Advances in Consumer Research Volume 6, 1979      Pages 453-455

CONSUMER SATISFACTION/DISSATISFACTION: AN OUTSIDER'S VIEW

J. Edward Russo, University of Chicago

The study of consumer satisfaction/dissatisfaction is a very recent addition to consumer research. It derives its impetus from the consumerism movement of 1965-75. Intellectually it is rooted in the measurement of social indicators, combined with some traditional business-oriented analyses of complaint behavior. Recognition of these origins is not just of historical interest. The field is new enough that these roots still largely define the goals and methodology for research.

The goals of consumer satisfaction/dissatisfaction (CS/D) research have been mainly pragmatic, especially the reduction of consumer dissatisfaction both by governmental policy makers and by marketing managers. This applied goal carries a more "engineering" than theory-based flavor. The methodology of the field, adapted from the social indicators movement, is primarily survey techniques, especially mail and telephone questionnaires.

THREE OVERRIDING ISSUES

Before discussing the individual papers in this session, it is worth acknowledging three larger issues that I see as dominating the study of consumer satisfaction/ dissatisfaction. The first of these is the relative roles of engineering and theory building. Should our goal be limited to "engineering" a solution to an applied problem, or should we strive for theories that explain the causes and processes of CS/D? The second two issues are problems that existing research in consumer satisfaction/dissatisfaction is confronting, so far with only limited success. The first I call overtheorizing, the top heavy construction of large theories on small data bases. The second is the validity of the data base itself, i.e., of the phenomena that are to be explained and that direct policy formation.

Engineering Versus Theory

If one's goal is to build a theory of CS/D, one must consider at least three aspects: the type of theory, one explaining output or process; the range of phenomena to be accounted for; and the need for, or value of, such a theory. There is a trend toward process theories, e.g., of consumer decisions (Bettman, 1978), of product knowledge (Johnson and Russo, 1978) and of advertising (Harris, 1979). Unlike these latter areas, CS/D has a more pragmatic goal and, thus, is more output oriented. The concern is more with how to increase satisfaction, by both public and commercial policies, than with the perceptual, evaluative, learning, memorial, computational or other processes that combine to generate a rating of CS/D.

The range of phenomena that a general theory of CS/D would need to account for is very wide. The field is necessarily concerned with all goods and services. These exhibit great variation over several important attributes, among them purchase price, product knowledge/familiarity, vender characteristics and regulatory assurances. The construction of a theory this general is a formidable task.

We should also ask ourselves: Do we really need a comprehensive theory? Is a detailed explanation necessary to solve an applied problem? Sometimes the answer is yes, with the alchemists' task of turning lead into gold as the prototype. But more often than not significant progress can be achieved with a simpler understanding. Medicine exhibits many such examples, one current one being the successful treatment of many cancers by chemotherapy. My judgment is that a comprehensive process theory of CS/D is neither necessary nor possible. Much applied progress can be accomplished without it, and we do not yet have a data base strong enough to support such a theory (more on this below).

In a more positive vein, the question can be rephrased as: What level of explanation of consumer satisfaction/ dissatisfaction should we strive for? Let me suggest a middle ground that satisfies the engineering goals of the field without being antitheoretical. At this stage in the development of the field we should expose the primary determinants of consumer satisfaction/dissatisfaction. That is, we should seek to identify the main variables that control CS/D. Implicit in this is the notion of ecological validity. The most important variables are those that explain the largest proportion of the variance in the phenomena of CS/D.

Overtheorizing

Overtheorizing refers to too high a ratio of theory to evidence. The scientific heart of any new area of research is the set of relevant phenomena. Admittedly, the collection of appropriate data is costly and subject to numerous threats to validity. However it is the foundation upon which any valid theory must be built. The problem of overtheorizing is exacerbated by the greater professional glamour associated with theory construction. However, a closer look at most theories in this area belies any attribution of glamour. They are more creatively borrowed than created. Furthermore, borrowed theories like borrowed clothes tend to fit poorly. I am not suggesting that we should not clothe our data with the best theories available. I do suggest, however, that the low cost of borrowing and high cost of data collection have led to overtheorizing in CS/D. This problem is naturally most acute for a new field, which has a limited wardrobe of theory and which must still bear the development cost of data collection methods.

Validity of the Data Base

There are at least two major sources of threats to the validity of this research. First, much of the data derives from survey techniques, with its litany of problems caused by selective returns, interviewer interactions or biases, the weak level of the responses, etc. One pair of problems associated with survey data deserve explicit mention. The first is a tendency to sample only complainers. This is especially true when the data analyzed are from a government agency or the files of a company's complaint department. A related problem is the tendency to sample extreme experiences. Even if a representative sample of subjects is successfully queried, the dissatisfaction experiences that they recall will tend to be more extreme ones.

The oversampling of extreme experiences is confounded by the reliance on memory to recreate and evaluate these experiences. Ample evidence suggests that this reconstruction process is subject to stereotypical biases. D'Andrade (1974) has demonstrated that the rating of behavior traits that are supposedly based only on observed responses are subject to systematic distortion in memory. For example, although antagonism and cooperation are believed to be very dissimilar traits their actual frequencies of emitted interpersonal responses correlated +.17 in the data that he reports. (Both behaviors are related to the intimacy of the interaction.) However, observers' memory-based ratings of these two behaviors correlated -.50. This negative correlation reflects the expected incompatibility between antagonism and cooperation rather than the actual behavior. In responding to an interview on consumer dissatisfaction there may be a tendency for many of the recalled experiences to be reconstructed from general stereotypes rather than from memory of the actual events.

A second area of threats to validity is associated with social psychological experimentation. Krishnan and Valle's paper surveys many of these very convincingly, and more comprehensive reviews are available. I only want to acknowledge that this alternative to survey methods is not without its problems for the empirical researcher.

DISCUSSION OF THE PAPERS

The discussion of the three papers in this session has two goals. The first is to identify and evaluate what I believe the authors intended to accomplish. The second is to state what I believe they should have tried to accomplish. The papers will be judged against their own goals rather than mine. However, I believe it is important to recognize failures of goals as well as failures of execution.

LaTour and Peat

In their paper LaTour and Peat review the expectation based theories that have been applied to consumer satisfaction/dissatisfaction and present another such theory of their own contrivance. Let us consider the review first, which is generally excellent. LaTour and Peat's initial comments on survey research are perceptive and highly relevant, especially the dangers of "fishing" for demographic relationships and the small percentage of variance usually accounted for by whatever is statistically significant. Their critical review of experimental methodology is also thoughtful and informative. Besides drawing on general principles of science, such as the critical role of experimental realism in CS/D research, their criticisms also include problems from psychology ( demand artifacts) and from statistics (lack of independence among statistical tests). Overall, the review is perceptive and to the point. It should be valuable to researchers entering the CS/D field or to those engaged in work in related fields.

LaTour and Peat's theoretical contribution is a generalization of Thibaut and Kelley's (1959) comparison level theory to individual product attributes. Although the authors see comparison level theory as offering several advantages over other expectation based theories, I fail to be convinced by their arguments. The partition of the comparison level by Thibaut and Kelley into personal, social and situational components introduces considerable complication, e.g., of identification and measurement, for an undemonstrated increase in explanatory power. More important, LaTour and Peat's generalization of the comparison level (or expectation level) to individual attributes seems to me to be a small and somewhat obvious theoretical step for which a high price is paid. The number of parameters in their formulation is extremely large, and whatever advantages it may have in generality it loses in parsimony. This paper seems typical of the danger of overtheorizing, an elaborate theory built over an inadequate evidential base.

The quality of the presentation of this paper is unusually high and deserves special note. It was well written generally but the arguments were especially clearly presented to the reader.

What should have been accomplished. The paper reports no test of the proposed theory. The reader has little idea of the value of Thibaut and Kelley's componential analysis of a comparison level in the CS/D context. Given the combinatorial increase in parameters, an empirical test, even one that rejected the theory, would have assured the reader that the theory is testable, i.e., the measurement of all parameters is feasible.

The failure to test the theory also avoids a salutary confrontation with ecological validity. CS/D with what products and services can most benefit from the three component partition of a comparison level, or are most appropriate for a multiattribute representation? Neither of these issues is addressed, as eventually they must, if the theory is to make a useful contribution to the study and reduction of consumer dissatisfaction.

Day and Ash

Day and Ash report a survey of complaints with durable products. Their goal, namely to collect a broad data base and point out areas where consumer satisfaction/ dissatisfaction is particularly high, is much closer to engineering than to theory. There is no pretense to theory building or even to relating their collected data to existing theory. The paper must surely be judged on the quality of the data and the descriptive analysis that is presented. Such an evaluation is not particularly easy because the presentation is unusually skeletal.

The data reported here represent a major effort that deserves recognition. Furthermore, they are part of a larger program of research that includes a wider range of products and services and that intends to observe a larger sample of consumers. Because one of the authors' goals is testing and eventual revision of their interviewing procedure, several potential problems should be pointed out. It appears that those informants who report less dissatisfaction are rewarded by finishing this self-administered questionnaire more quickly. This probably leads to a bias toward underreporting of dissatisfaction. A second potential problem arises in their validity check. They report correlations in the nineties (for three of four cases) between two measures of dissatisfaction. However, this close agreement might be an artifact of self consistency. Such an artifact is especially likely because while completing the questionnaire subjects could not be monitored. They could easily check an earlier response in order to be assured that a later one is consistent with it.

As with many surveys their methodology is susceptible to demand artifacts. To respondents who did not complain, they asked the question, "Why didn't you do anything?" Their subjects responded almost exclusively with "It wasn't worth the time and effort" or "I didn't expect it would do any good." These answers are overly rational and considerably more attractive that admitting ignorance or forgetfulness, which were the other two possible replies. Other possible explanations, such as the consumer's own fault, were not even presented as alternatives to the subjects. Because all of these latter responses are less socially acceptable, one suspects that they occur more frequently than reported by Day and Ash's interviewees.

My major complaint with the paper is the descriptive analysis. It falls below even minimal standards for the reporting of such data. For example, although the percentage of dissatisfied responses is reported over a 115 different product categories, not a single mean has been calculated and reported, either in the tables (which are quite extensive) or in the test. When these means are calculated the percentage of complaints reported is quite low (roughly 6%) when compared with other studies. For example, Andreasen and Best (1977) report 20% complaint rates using a different survey technique. This low complaint rate enhances the possibility of a response bias caused by the relative ease of reporting satisfaction rather than dissatisfaction.

They also present results for small sample sizes without recognizing the limitations thereof. For example, they report that 28.6% (2 of 7) of the purchasers of used pickup or panel trucks were dissatisfied (Table 4). In addition to the irritation of dealing with three digits of accuracy for such small sample sizes, the dangers of ignoring the limitations of these sample sizes are quite real. Day and Ash conclude from the above data that "the highest rates of reported dissatisfaction for any of the 63 items occurred with products used by comparatively small proportions of the sample." It is difficult to know whether this result is genuine or merely an artifact of the higher variability associated with small samples. Certainly it should not be reported until a deeper statistical analysis is performed.

Finally, the authors make little effort to link their data to existing theories or other bodies of data. The former is not their intention, but certainly the reader should expect citation of the representative empirical literature. Entire bodies of work are conspicuous by their absence, such as the volume by Hunt (1977). The reader is continually frustrated by the failure of the authors to do much more than put the tables of data into print.

What should have been accomplished. As suggested above, the analysis of the data needs considerable elaboration. Presumably Day and Ash are waiting to piece together all parts of their research program. This is all well and good; but at the present this paper is being reported separately, not as one chapter of a larger volume. The authors should have provided more pertinent interpretations of the data reported and appropriate comparisons to other descriptive studies. Certainly such an exercise would be valuable in itself since the authors must eventually do this for their expected monograph.

In fairness to the authors, it must be noted that some of the problems with this paper stem from the format of this conference. Instead of a revision of each paper based on private reviews prior to publication, the paper was not permitted a revision and is now reviewed publicly. The reader would be better served by being able to read a revised version of Day and Ash's paper than by reading these comments.

Krishnan and Valle

Krishnan and Valle help to introduce attribution theory to the consumer satisfaction/dissatisfaction literature. In addition, they present data based on a mail questionnaire survey. Attribution theory offers an important perspective, and they provide relevant supportive data.

The attribution notion is valuable because it highlights a variable that is normally overlooked in CS/D research. In order to understand the link between dissatisfaction and overt complaint behavior it may well be important to consider attributions of responsibility or blame, It might be expected that consumers who blame themselves for dissatisfaction with a product are less likely to complain. Indeed, Krishnan and Valle's data confirm this hypothesis.

The experiment has several problems worth noting although some of them are minor. Their report that subjects who attributed responsibility for the dissatisfaction to themselves made fewer overt complaints may be confounded by an artifact. How many subjects are likely to admit that they were to blame yet they complained it was the company's fault. They also are not as careful as they should have been in some of their data presentations. For example they claim that Factor 1 in their analysis is "non-assertive" when one of its components is complaints to a salesperson. The only significance level that they report in their study is .0001. One wonders what interesting relationships, possibly inconsistent with attribution theory, this extremely stringent level is obscuring. Also they tend to overstate their conclusions, e.g., that "external attributions are necessary" if complaint behaviors are to result. I rather suspect, and I know businesspeople would agree, that many consumers will hassle the company in hopes of economic gain even if they themselves are to blame for the product failure. Finally, not all of the statistical methods are as strong as one would like. For example, multivariate analysis of variance might have been used instead of the many univariate analyses of variance that were performed.

Overall, however, the contribution easily outweighs the criticism. Their calling attention to attributions of blame is the appropriate level of theorizing. It does not overtheorize a process model of attribution generation or a multiattribute model of attributions or some other theory more elaborate than the data can bear. In addition to exposing a pertinent factor in controlling CS/D phenomena, Krishnan and Valle report empirical results that test the explanatory value of their construct. The only additional step I would ask is a measurement of the percentage of variance in complaint behavior that is accounted for by attributions of blame. How does this factor compare in importance to such other factors as product/service category, sociodemographics, product knowledge and expectation level? This is a likely question for future research.

ACKNOWLEDGEMENTS

I would like to thank Bill Wilkie for the opportunity to discuss three interesting papers in an area in which I recently knew very little. I hope that my recent and limited familiarity with CS/D research has been sufficient to enable a commentary of value. And I ask the indulgence of the authors whose papers I have just discussed if I have failed to appreciate some merit that a more knowledgeable reviewer might have.

REFERENCES

James R. Bettman, An Information Processing Theory of Consumer Choice, 1978, forthcoming (draft, August 1977).

Richard Harris, Information Processing Research in Advertising, 1979, in preparation.

H. Keith Hunt (ed.), Conceptualization and Measurement of Consumer Satisfaction and Dissatisfaction, (Cambridge, Mass: Marketing Science Institute, 1977).

Eric J. Johnson and J. Edward Russo, "What is Remembered After a Purchase Decision?", Carnegie-Mellon Information Processing Technical Report, November 1978.

J. W. Thibaut and H. H. Kelley, The Social Psychology of Groups, (New York: John Wiley, 1959).

R. G. D'Andrade. "Memory and the assessment of behavior." In H. M. Blalock, Jr. (Ed.), Measurement in the Social Sciences, Chicago: Aldine, 1974.

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