Consumer Judgment Strategies: Beyond the Compensatory Assumption

Citation:

Peter L. Wright (1972) ,"Consumer Judgment Strategies: Beyond the Compensatory Assumption", in SV - Proceedings of the Third Annual Conference of the Association for Consumer Research, eds. M. Venkatesan, Chicago, IL : Association for Consumer Research, Pages: 316-324.

Proceedings of the Third Annual Conference of the Association for Consumer Research, 1972      Pages 316-324

CONSUMER JUDGMENT STRATEGIES: BEYOND THE COMPENSATORY ASSUMPTION

Peter L. Wright, University of Illinois

[Peter Wright is Assistant Professor of Business Administration at the University of Illinois, Urbana.]

An important problem which is receiving increasing attention among consumer researchers focuses on consumer judgment of competing alternatives. While historically there has been interest in the processes of judgment and choice in the marketplace, the acceleration of social and product innovations which will characterize tomorrow's marketplace (which has provoked one contemporary social commentator (Toffler, 1970) to predict a society characterized by "over-choice") has heightened the need for rigorous examination of different models of consumer judgment strategies. Recent consumer research attention has centered on a linear additive "attitude" model as providing insight into consumer choice behavior. The models proposed by Fishbein (1964) and Rosenberg (1956) have been most popular; these models belong to a class of models which may collectively be labeled "compensatory-processing" models.- These models have in common the identification of two variables as important in mediating a consumer's behavior toward a product; beliefs about the-product characteristics and the evaluative importance of the characteristics. The basic assumption is that individual consumers affectively evaluate alternatives along a set of dimensions. The consumer is then seen as combining these unidimensional cues according to a simple linear rule. The "index" resulting from such a process is expected to guide the individual's choice according to a decision rule which says "choose the alternative with the highest index." Attention has been directed at demonstrating that this two-component linear model satisfactorily predicts behavior (Bass and Talarzyk, 1972, Hansen, 1972), and in operationally refining measurement of the two variables (Cohen and Houston, 1970; Schendal, Wilkie, and McCann, 1971).

There has been little interest thus far in critical examination of the assumption about the strategy an individual consumer may use in combining information about a multidimensional product which is implicit in the additive compensatory model (however, see Bettman, 1971, and Russ, 1971). One reason for this undoubtedly is that this assumption has been the dominant one across disciplines where human choice behavior has been a topic of empirical inquiry. The general success of predictions based on such linear, compensatory models cannot be disputed, and the search for ways to detect the actual operation of some choice strategy other than the linear one among a variety of choice alternatives has not been overwhelmingly successful.

Two comments should be made however in order to place the continued pursuit of non-compensatory choice strategies in perspective. First, it is generally agreed that the formal linear model is statistically extremely powerful, and is capable of providing close approximations to data which is not itself strictly linear (Green, 1968). We must keep in mind the important distinction, however, between successful prediction of consumer preferences and successful representation of the actual cognitive processes which individual consumers are bringing to bear on the choice task. Prediction is relevant to certain of the objectives of the publics we serve (managers and consumer advocates), but prediction alone frequently is accompanied by only a low level of understanding of the psychological behaviors involved. Second, it will probably prove more useful in the long run to begin to directly attack the question of when an individual might use different available strategies rather than trying to prove the general superiority of one or another model.

ALTERNATIVE CONSUMER JUDGMENT STRATEGIES

The linear compensatory model has been verbally described above; its formal presentation takes the form:

U = E w_i s_i

where:

U is the overall rating or ranking of the alternative

w is the weight assigned to characteristic i

s is the evaluative (affective) rating of characteristic i

Other models of the judgmental process have been proposed by social-psychologists (e.g., Coombs, 1964; Dawes, 1964). One such model, labeled the "conjunctive" model, suggests that an individual uses a minimum evaluation procedure in which the multidimensional object is judged on its minimum performance on all characteristics. The contrast between this strategy for judging the product and that assumed by compensatory strategy should be understood. When using the compensatory strategy, the consumer focuses equal attention across the entire range of possible ratings for product characteristics, and allows a better than average performance on one characteristic to effectively cancel out a poorer than average performance on another characteristic. If the individual is using a conjunctive strategy, he focuses a disproportionate amount of attention on the negative end of the evaluative continuum. Assume the individual cognitively arrays product characteristics on seven-point bipolar scales where 1 = the worst rating possible and 7 = the best rating possible. In terms of this scale, when using the conjunctive strategy he tends to ignore ratings falling at the number 5, 6, or 7 points relative to those falling at 1, 2, or 3.

An alternate strategy, the "disjunctive" model, suggests that the consumer may judge the product on its best characteristic regardless of the other attributes. This strategy is essentially the converse of the conjunctive strategy, since now a disproportionate amount of attention is focused on the upper (positive) end of the continuum for evaluation. Points 7, 6, and 5 are more influential in the overall judgment than the remainder of the scale.

Until recently suitable functions to represent mathematically the conjunctive and disjunctive strategies were not available. Einhorn (1970) has noticed that two fairly familiar geometric surfaces can serve as reasonable approximations of these models. He argues that the conjunctive model can be represented by the parabolic response surface:

EQUATION

Here the consumer's judgment of the multidimensional product will be highest where there are no wide discrepancies in ratings across the separate product dimensions.

The disjunctive strategy is approximated by a hyperbolic response surface, and the following mathematical equation is appropriate:

EQUATION

where:

U = overall product rating or utility

s_i = evaluative ratings of characteristic i

w_i = weight of characteristic i

a_i = some arbitrary value set above the highest si rating (i.e. above the asymptote) so that U does not become infinite

Here the consumer will find the product quite attractive if he can rate it extremely high on a single attribute (of course if he can rate it extremely high on more than one attribute, he is ecstatic).

Einhorn points out that although the linear model appears to be mathematically simpler, its actual use by the individual may in fact require considerably more cognitive exertion than either of these two alternate strategies which require more complex mathematical equations. There is considerable precedence, it should be noted, throughout the study of behavior for researchers to adopt as their working model of their subjects' behavior one which is easy to analyze from the perspective of the researcher, regardless of whether the model would in practice be simple and efficient for the subjects to use.

INDIVIDUAL USE OF THE STRATEGIES

One purpose of this paper is to present data demonstrating that individuals do indeed appear to adopt strategies in judging alternative products which can be better represented by the non-linear, non-compensatory models than by the traditional linear model. The method of inquiry was to require individuals, in this case males, ages 19-22, who did not at the time personally own an automobile and were approaching a change in financial status which made them likely purchasers during the upcoming year, to rank order two sets of fifteen hypothetical automobiles. These cars were described for the subject in terms of their rating on four attributes: selling price, style, ease of handling, riding comfort. The information about each attribute was given on a seven-point scale which varied from "greatly above average" to "greatly below average," with the middle point "average." All the cars in both stimulus sets were described as belonging to the "under $4000" class, and the ratings referred to this particular class of cars.

Subjects were presented with the first set of hypothetical cars with each one described on a separate card. They were asked to rank them in order of personal preference as products which they themselves would purchase upon graduation. Having ranked the first set, they repeated the judgments for a second validation set.

For each individual subject, a linear regression analysis was used to determine the weighting parameters. The use of regression coefficients as approximations of the weighting of the product dimensions in the judgment exercise is, it might be noted, more clearly justified when, as was the case here, the individual is the unit of analysis then when the weights derive from the more common aggregated analysis. In the case of the conjunctive and disjunctive models, the weights were found by using multiple regression on the log-transformed variables. The weights obtained from the first set of automobile stimuli were then applied to the second set in order to obtain a predicted ranking. The actual ranking of the second set by each subject was then correlated with the derived predicted ranking by means of the Spearman rho procedure. These second set rankings were used as the criterion for comparisons with the predicted rankings from the linear, conjunctive, and disjunctive models.

The data for two of the subjects will be discussed. For Subject A the rank ordering predicted by the linear model correlated .68 with the actual ranking; the ranking predicted by the conjunctive model correlated .89 with the actual ranking; and that predicted by the disjunctive model correlated .26 with the actual ranking. The difference between the linear and conjunctive models is significant at the .01 level, as is the difference between the linear and disjunctive models. It appears that this subject's judgment strategy was best represented by the parabolic function (conjunctive model), and that the disjunctive model was distinctly inappropriate.

Subject B appeared to approach the judgment task differently than Subject A. In his case, the ranking predicted by the linear model correlated .55 with the actual ranking; the ranking predicted by the conjunctive model correlated .57; and the ranking predicted by the disjunctive model correlated .82. The difference between the correlation of the disjunctive model and that of either of the other models is significant at the .05 level. The strategy of Subject B was apparently more closely approximated by the hyperbolic function associated with the disjunctive model than by the linear or parabolic functions.

For both these subjects, examples can be found in the data to illustrate where either the conjunctive or disjunctive models picked up fairly dramatic errors in the predicted rankings obtained using the linear model. These data have been selected in order to clearly demonstrate that individual consumers may use a variety of rules for psychologically combining separate pieces of information to arrive at an overall product judgment.

THE NEED FOR SYSTEMATIC STUDY

There appear to be compelling reasons for expanding the study of consumer judgment and choice to include the probability that individual consumers employ various strategies other than the monolithic linear compensatory strategy: (1) it can be successfully demonstrated, as the data presented show, that psychometric functions which are theoretically justifiable representations of the conjunctive and disjunctive strategies provide a better fit to the observed judgments of some individual consumers than does the linear model; (2) the subjects themselves, when asked to verbally recapture the nature of their own just-completed judgment process, will frequently describe a conjunctive or disjunctive strategy quite clearly (Goldberg, 1968). This was observed in the present study even in subjects for whom the formal conjunctive or disjunctive models afforded no predictive advantage relative to the linear model; and (3) the importance of the point raised earlier about the amount of cognitive exertion required of an individual consumer in applying these strategies to process product information cannot be minimized. Much of what we know about cognitive behavior leads us to expect that people will use devices for simplification of their information handling tasks wherever possible; this is an efficient adaptation to their complex environment which enables them to buy time for a host of other activities. Perceptual categorization is one example of this. Realistically, for a consumer to actually do what the linear, compensatory model says he does would be much more difficult and take much more time than doing what the alternate models, with their "focusing-on-one-end-of-the-evaluative-dimension" property, suggest.

RELEVANCE TO RESEARCH ON DYNAMICS

If in fact individual consumers do use various judgment strategies, then another important reason for systematically studying this area springs from what we are ultimately about as consumer researchers. Much marketing activity is concerned with influencing, rather than merely predicting, consumer behavioral preferences. Thus while researchers may start with structural models of established predictive validity, the ultimate value of such models appears to be in bringing order to the study of how and why preferences are changed by changes in product features, advertising, distribution, etc. Currently the study of attitude change is beset by problems of inaccurate predictions and difficult to interpret findings. Consider the relevance of the question of alternative judgment strategies. Predictions as to what happens to a person's overall judgment of a product when a belief about some single characteristic is changed are entirely dependent on which judgment strategy the person is using relative to that product. An example may clarify this idea. Assume Consumer X has formed a judgment of a product based on his beliefs about three characteristics of the product. We now wish to predict the effect on his summary judgment of a slight change in his rating of the product on one of the three dimensions. (Assume he thinks each dimension to be equally important; this will simplify the discussion although it is not necessary to the argument.) If Consumer X is indeed using a linear, compensatory judgment strategy in this case, then a change of one scale-unit on any of the three belief dimensions will have an equivalent effect on the overall judgment. This is true regardless of what his prior beliefs were, i.e., where he previously located the product on each of the belief dimensions. However, if Consumer X is instead applying a conjunctive strategy, our expectations about the outcome of such changes becomes contingent on both the prior belief and the specific dimension where change is accomplished. Since the conjunctive strategy implies that the judgment will not improve until the product reaches satisfactory cutoffs on each dimension, a positive change in an already satisfactory belief will not make any difference. The overall judgment will show a shift only when the change is accomplished along the dimension where prior beliefs fell below the minimum cutoff. Similarly, if Consumer X is applying a disjunctive judgment rule here, then changing previously negative beliefs to neutral or neutral beliefs to mildly positive will not have a noticeable effect on the overall judgment. Only by creating the belief that the product is superior on one (or more) dimensions can the judgment improve.

For the linear model, all locations on the belief dimension are equivalent. For the conjunctive or disjunctive models, certain locations are markedly more important than others to the consumer. These differences become extremely important in setting objectives for an advertising program, in stating researchable hypotheses, and in trying to explain why a certain manipulation was successful at the belief level but not at the attitudinal level. Considerations of the possibility of alternate judgment rules can no longer be ignored in research and theory on attitude dynamics.

MODELING THE ENVIRONMENT FOR CONSUMER JUDGMENT

As has been seen, individuals differ in their use of strategies in the same task. There are also undoubtedly differences within the same individual across various tasks. One further purpose of this paper is to outline a conceptual model of the task environment, specifying certain dimensions of the situation in which the judgment occurs which may influence which model a consumer applies.

Perception of Penalties and Rewards

Consumers may be aware that the situation they are facing places particularly heavy penalties on making either a Type I or Type II error in their judgment of the product. If such is the case, they will probably adapt their judgment conventions to accommodate this analysis much as researchers have done in their setting of conservative significance levels to protect against certain errors in judging empirical evidence. If the consumer perceives that the probability that he will be drastically penalized for committing an error of inclusion is high, he should favor the conjunctive strategy. If the housewife expects to be blamed for her mistake in choosing a product (and not rewarded significantly for especially good choices), the conjunctive strategy is quite reasonable. The same applies to the industrial buyer's perception of reward system in his organizational role. Conversely, if the individual believes that he will be penalized for excluding a product which is extraordinarily satisfying, he may favor a disjunctive strategy. Such penalties will likely be in the form of opportunity costs: praise or tangible rewards foregone.

The individual may perceive these penalties to apply only for the specific product, for all product decisions made in a particular role, or for most decisions he has and will encounter in life. When a person tends to develop a chronic strategy for making judgments based on this latter type of perception, then his judgment style takes on the properties of a personality trait. It would not be surprising if the general environment itself is set up to favor certain strategies; for example it may be that penalties for errors of inclusion are simply much more prevalent in society than penalties for exclusion. Or, if not more prevalent, than perhaps more uncomfortable psychologically, especially in the short run.

Exclusiveness of the Behavioral Expression of the Judgment

Presumably a person's judgment of a product will carry with it certain behavioral ramifications. The futility of measuring predispositions which are not linked to specified behaviors has been noted. Sometimes the context of the judgment will be such that the person must choose one and only one of the alternative products available, and the selection is not revocable. In other cases, the behavioral choice may not be mutually exclusive of the choice of other alternatives. For example, a person may have a given set of needs to satisfy with a fixed resource. If he can only acquire a single product alternative, given this limit, he will tend to favor a conjunctive or linear strategy since these arrive at their final judgment through some consideration of performance on all dimensions. If he has, however, the luxury of being able to select more than one of the alternatives, he may use the disjunctive strategy. Here he could look for outstanding performance on one attribute of Product A to satisfy a certain need quite well, outstanding performance on another attribute of Product B to satisfy another need quite well, and so forth. Again, the question of penalties is relevant, and this factor can be considered a subset of the above discussion.

Range of Alternatives Being Judged

It is quite likely that the strategies may be used sequentially. The aim of all product judgments is eventual selection of some appropriate behavior toward the products. At a certain stage in selecting appropriate behaviors to satisfy certain needs, the consumer will find that there are a large number of candidate products which he must judge. It would be efficient at this stage to apply the conjunctive strategy to whittle down the domain of alternatives (hopefully) to a more manageable group. It was noted earlier that application of the linear model is quite complex, and its application to each product in a large group is even more exhausting. Having eliminated many of the candidates as worthy of positive behavioral action, the consumer may then be able to apply the linear model as the discriminator. Or he may first use the conjunctive to eliminate all products not meeting the cutoff, then the disjunctive to eliminate all not having at least one superior characteristic, then the linear. This corresponds to an attempt at cognitive simplification.

Of course, the use of differential weighting of the attributes involved is another such device for simplification. In the case above, one weights different parts of the dimensions differently to make judgments easier, but assigning greater importance to specific dimensions as a whole is also a commonly recognized method for accomplishing the same thing. (This is another case where the subject complicates the researcher's formal model in order to make things less complicated for himself.) Thus, by weighting one or two dimensions very heavily and others so lightly as to virtually ignore them, a choice may become apparent. If the individual is not able, however, to really reduce the set of evaluative product dimensions he feels worth considering to one or two, then this type of simplification doesn't work for him.

Does It Work?

An ultimate question which the consumer confronts in situationally applying one or the other judgment strategies is whether or not that strategy does indicate for him the product alternative which he should choose. After all, that is his goal in going through the process of integrating all this information. If, in judging a number of alternatives, he can not discriminate among them using one strategy he will undoubtedly try using another until at last some choice is indicated. If this is impossible (and there are certainly product domains where there exists sufficient homogeneity among the alternatives such that all of these schemes are frustrated), then additional information must be gathered to enable a choice. In terms of the sequential pattern of use discussed above, if the consumer faced with a lot of alternatives to evaluate applies any of the models and, lo and behold, finds that one discriminates so successfully that no further evaluation is necessary, then the process stops right there.

WORKING FORWARD VS. WORKING BACKWARD

One final distinction might be made which is perhaps more closely related to the conduct of research on this topic than to the behavior of the consumer. Two research paradigms used in the study of judgment and attitude models must be contrasted. One approach--that illustrated here--has the subject respond to hypothetical stimuli described for him by the researcher. In such a case, the subject presumably does not already have any attitude formed toward the stimulus prior to receiving the experimental information, and his beliefs about the stimulus characteristic are controlled by the researcher. His task is clearly to integrate the information and arrive at a judgment. The contrasting approach is to utilize as the stimulus some real and identifiable object with which the consumer individual has had previous experience. Here, the subject enters the research situation with his judgment already formed and the latent structure of beliefs is extracted from him by the researcher. In both cases, different models may be fit to the resulting data and their relative predictive validity examined.

These two paradigms are actually studying different phenomena. The former is clearly focusing on the development of a judgment from its component parts, whereas in the latter this creative aspect is missing. It may very well be that these differences in the actual behavior being studied do have a biasing effect on the probability that a certain model will appear superior. It would appear, for example, that the situation in which the subject works his way backward from his existing attitude to his separate beliefs will tend to promote a linear strategy. Just as when, in originally integrating the information, it may have been simpler to adopt a non-linear strategy, when describing the structure behind an already developed attitude it may be simpler for the subject to cognitively run a linear gradient through the belief-scales he is filling out.

The whole point in a person developing an attitude in the first place is that it frees him from constantly having to go through the process of reintegrating all the information he carries about the product whenever some potential for acting toward the product occurs. Thus the individual asked to describe his belief system may often be forced to recreate it in part. In any case it is clear that the two research situations do differ in terms of the direction of the dynamics involved, and that there is no necessary reason for the processes of someone working backward to mirror the processes involved in integrating new information.

REFERENCES

Bass, F. W. and Talarzyk, W. W. An attitude model for the study of brand preference. Journal of Marketing Research, 1972, 9, 93-96.

Cohen, J. B. and Ahtola Houston, M. The structure of consumer attitudes. Working paper, University of Illinois, 1971.

Bettman, J. R. The structure of consumer choice processes. Journal of Marketing Research, 1971, 8, 465-471.

Coombs, C. H. A Theory of Data. New York: Wiley, 1964.

Dawes, R. M. Social selection based on multidimensional criteria. Journal of Abnormal and Social Psychology, 1964, 68, 104-109.

Einhorn, H. J. The use of nonlinear, noncompensatory models in decision making. Psychological Bulletin, 1970, 75, 221-230.

Fishbein, M. An investigation of the relationships between beliefs about an object and attitude toward that object. Human Relations, 1964, 16, 233-240.

Green, B. Descriptions and explanations: a comment on papers by Hoffman and Edwards. In B. Kleinmutz (Ed.), Formal Representation of Human Judgment. New York: Wiley, 1968, 91-98.

Goldberg, L. Simple models or simple processes? some research on clinical judgements. American Psychologist, 1968, 23, 483-496.

Hansen, F. Consumer Choice Behavior: A Cognitive Theory. New York: The Free Press, 1972.

Russ, F. Evaluation process models and the prediction of preference. Proceedings, Second Annual Conference, Association for Consumer Research, 1971, 456-461.

Scendal, D., Wilkie, W., and McCann J. An experimental investigation of attribute importance. Proceedings, Second Annual Conference, Association for Consumer Research, 1971, 404-416.

Toffler, A. Future Shock. New York: Random House, 1970.

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