A Reinterpretation of Farley and Ring's Test of the Howard-Sheth Model of Buyer Behavior

James Taylor, California State University at Fullerton
Jonathan Gutman, University of Southern California
[ to cite ]:
James Taylor and Jonathan Gutman (1974) ,"A Reinterpretation of Farley and Ring's Test of the Howard-Sheth Model of Buyer Behavior", in NA - Advances in Consumer Research Volume 01, eds. Scott Ward and Peter Wright, Ann Abor, MI : Association for Consumer Research, Pages: 438-446.

Advances in Consumer Research Volume 1, 1974    Pages 438-446

A REINTERPRETATION OF FARLEY AND RING'S TEST OF THE HOWARD-SHETH MODEL OF BUYER BEHAVIOR

James Taylor, California State University at Fullerton

Jonathan Gutman, University of Southern California

One of the orientations in research in consumer behavior in recent years has been to develop broad theoretical constructs to explain consumer behavior. The scope of these models makes empirical testing of the model as a whole difficult because of the difficulty of operationalizing the concepts and the relationships between them. Nevertheless, there has been one notable exception, the Farley and Ring (1970) test of the Howard and Sheth Model of Buyer Behavior (Howard and Sheth, 1969). The scope alone, of Farley and Ring's undertaking, would make it a significant piece of research. It deserves even more special attention because of their implication that it supports the Howard and Sheth theory.

The importance of such support, or lack of lt, is difficult to overstate because it points to a direction for development by the entire consumer research community. This importance demands that Farley and Ring's work receive the closest possible critical evaluation. Although two comments (Hunt and Pappas, 1972; Lutz and Resek, 1972) have been published, we do not believe that the real issues involved have yet been brought out fully. We agree with Farley and Ring, in their reply to the two published comments, that: "The major questions involved in progress in applying these types of models, in our view, are not related to technological niceties but to fundamental conceptual and operational issues." (Farley and Ring, 1970, p. 353)

It is exactly the operational and conceptual issues raised by their work that we intend to address. The authors' experience with consumer research in the field has developed a healthy appreciation for the technical "trade-offs" involved in the actual execution of field research. It should be clearly understood that the thrust of this article does not lie in the area of "niceties."

It is exactly that appreciation of the difficulties inherent in translating theoretical constructs into field research that leads us to insist upon a rigorous confrontation with the problems, rather than dismissing them as Farley and Ring do, because: "...it is not convenient to work on these definitional problems in the context of such a large and relatively clumsy formulation of the model..." (Farley and Ring, 1967. D. 350)

Since the questions raised here may have very broad implications, lt seems important to specify the limitations that we have placed upon the content of this paper. It is, therefore, our intention to develop the four following points:

1. Illustrate the considerable difficulties involved in actually operationalizing the variables in broad theoretical models of consumer behavior by contrasting Farley and Ring's actual variables with Howard and Sheth's theoretical variables. We believe that the discrepancies in this one aspect of the test alone, are sufficient grounds to question whether the Howard and Sheth model was, in fact, tested.

2. Whereas operationalism is important in market testing, in developing theory equal weight must also be given to how these concepts relate to one another. We believe that the interrelationships between the hypothetical constructs in the Howard-Sheth model are equally as important to an adequate test of the model as are the definitions of those constructs. Provision must be made in a test of the model for uncovering these relationships. The manner in which these problems were dealt with will be examined.

3. Present an alternative analysis of Farley and Ring's results, which we believe is consistent with the realities of actual field research, in an attempt to point out the risks faced by the analyst when variables and relationships are not rigorously specified in advance.

4. Suggest an approach to dealing with the very real problems of conceptual and operational issues which may not be "convenient" but is pragmatically realistic as well as necessary for the advancement of knowledge in this area.

OPERATIONAL DEFINITIONS

The problem here is that the operational definitions used by Farley and Ring simply do not reflect the conceptual definitions of Howard and Sheth. Lutz and Resek (1972) caught a piece of this problem in their analysis of Farley and Ring's work.

Farley and Ring's first endogenous variable is "Attention" (Y1). Howard and Sheth define "Attention" as follows:

Attention refers to the degree of "openness" of the buyer's sensory receptors for a particular feature of a specified stimulus display and a consequent narrowing of the range of objects to which the buyer is responding in relation to the range to which he is exposed. (Howard and Sheth, 1969, p. 154)

Farley and Ring define "Attention" as: "Recollection of seeing ads for the product." (Farley and Ring, 1970, p. 436)

What Farley and Ring have done is to define away all of the buyer's environment and stimuli except for advertising, and only that advertising which can be recalled, at that. The restrictiveness of the Farley and Ring definition of attention makes comparison with the richness of the Howard-Sheth definition difficult.

Farley and Ring's second variable is "Perceptual Bias" (Y2). Howard and Sheth (1969, p. 168) define "Perceptual Bias" as:

The quality or meaning of the information the buyer receives is the subject of interest here...the quality of the message is determined by Perceptual Bias.

But Farley and Ring (1970, p. 436) define "Perceptual Bias" as: "A filter variable, related to a tendency to use new products and rate them well." The two definitions have nothing whatsoever to do with each other.

Farley ant Ring's third variable is "Stimulus Ambiguity" (Y3). Howard and Sheth (1969, p. 156) define this variable as follows:

The ambiguity of the Stimulus Display is another influence on the flow of information into the buyer's mental process. It is the lack of clarity of the neural messages incited by the physical object or event (Stimulus Display) and which are permitted to enter the buyer's body via Attention.

About "Stimulus Ambiguity," Farley and Ring (1970, p. 436) say, "The ability to describe the product accurately. Results from the first telephone interview. Characteristics chosen were 'powdered, in a box, many flavors, drunk, and individual portions.' The variable is the total number of correct identifications among the five categories."

Suppose for a moment that it might be possible to agree that a new food product could adequately be described in terms of physical features as "powdered, in a box, many flavors, drunk and individual portions," and that having done so, the physical object of the product would be unquestionably distinguished from all other products. It would still be necessary to specify information about the tactile sensations of the product itself (taste, flavor, texture, etc.), and the satisfactions associated with using the product (satisfy hunger, stick to the ribs, children receiving nutrition, etc.), and the imagery of the container itself (rich colors? luxurious type faces?, etc.),and about the transaction itself (good value? what must be foregone? etc.), and more. At best. Farley and Ring have defined a small part of "Stimulus Ambiguity.

The next Farley and Ring variable is "Motive" (Y4). Howard and Sheth (1969, p. 199) say that:

Motives are the biogenic or psychogenic needs, wants or desires of the buyer in purchasing and consuming an item in a product class. They include the consciously sought goal, which is considered to determine behavior.

Farley and Ring (1970, p. 436) say that "Motive" is: "A scale variable collected on the first mail questionnaire on interest in differences among brands of the Product."

It is extremely difficult to imagine what Farley and Ring could have had in mind when they arrived at their definition.

The next variable is "Overt Search" (Y5) and Howard and Sheth (1969, p. 155) describe "Overt Search" as:

Overt Search is the 'process' by which the buyer selects a particular element of his environment as the Stimulus Display in order to clarify the saliency of motives in a given situation. It involves at the least the buyer shifting his head, for example, and extends, in the other extreme, to seeking out and talking with particular people in order to obtain the information wanted, as well as to formal search effort. Hence it extends to extensive thinking, to manipulating the symbols received in the process of exerting the physical movement to acquire the information.

"Overt Search" to Farley and Ring (1970, p. 436) means: "Endogenous, experience-induced search for information by talking to someone else."

Note that this comparison has not been selective. The progression started with Y1 and has moved directly on to Y5. The same sorts of comparisons with Y6 through Y11 reveal the same sort of between the concepts of the model and the operational definitions, but limitations of space will end the comparisons here. It should be stressed that the questions raised here are not whether Howard and Sheth's concepts are correct or not. It is whether those concepts have, in fact, been measured.

Whatever Farley and Ring have tested, it seems unlikely that it was the Howard and Sheth Model of Buyer Behavior. Rather, they have responded to the considerable difficulty of the problem by simplifying it to manageable limits. This simplification, however understandable it may have been, is of questionable value in determining the utility of the model.

AN ALTERNATIVE INTERPRETATION

The likelihood that Farley and Ring have not tested the Howard-Sheth Model has been demonstrated. Farley and Ring's operational definitions bear little, if any, resemblance to Howard and Sheth's definitions of the model's variables.

Unanswered, however, is the question, "What did they accomplish?" The authors believe that Farley and Ring's lack of rigor in developing operational definitions led them into the twin traps of common method variance and "halo effect." Understanding how this happened leads directly into an alternative analysis of their results.

The first thing they seem to have done is to have collected a whole series of the kind of misleading coefficients that are frequently encountered when data are collected from the same individuals about the same subject, i.e., Common Method Variance. For example, it is logically almost impossible for a consumer to have an "Intention to Buy" some product without having some degree of "Attitude toward the Product."

The low values of the coefficients on that run through much of the Farley and Ring data is symptomatic, in our experience, of the Common Method problem. While such relatively small, statistically significant, relations may be valuable to econometricians in their model testing, consumer researchers would be well advised to treat them with caution.

In any case, we believe that Common Method is only part of the explanation of Farley and Ring's results. Table 1, on the next page, recasts Farley and Ring's (1970, p. 434-435) Table 2 data purposively without regard to the numerical value shown. Any dependent variable that was found to relate significantly with any other variable, regardless of level of significance, has been indicated.

Examination of Table 1 makes two things immediately clear. One is that the majority of the variables relate with only one other variable. The other is that three variables have large clusters of related variables. Two factors are probably at work here. One is the Common Method problem. But that factor should be expected to run through all of the dependent variables in roughly the same way. The second factor is that not every endogenous variable was included in every equation. While this factor accounts for some of the clustering, or lack of it, it is not sufficient to explain this particular clustering by itself.

TABLE 1

RELATIONSHIPS BETWEEN STRUCTURAL PARAMETERS IN LINEAR FORMULATION OF HOWARD-SHETH MODEL AS DETERMINED BY FARLEY AND RING

The three dependent variables with clusters of related variables are "Motive," "Attitude" and "Intention." Let's look at exactly what Farley and Ring used as operational definitions when they measured these three variables.

Motive: ...interest tn differences among brands of the product (Farley and Ring, 1970, 0. 436)

Attitude: ...overall evaluation of the product...(Farley and Ring. 1970, p. 436)

Intention: ...likelihood of buying the product in the next month (Farley and Ring, 1967, p. 436)

Interest in brand differences certainly should lead to more sensitivity in the consumer's overall evaluation of the product and both dimensions should inprrove the sensitivity of the "likelihood" of purchase. We believe that Farley and Ring have measured the "halo" effect that is customarily found around specific products.

Our strong suspicion is that Farley and Ring have simply measured the "consumer profile" for the product (Household Size, City Size, Education, Income, Working Wife, Homemaking Skill, etc.) along s single dimension. If the product has a reasonably identifiable market, or "consumer profile," then those consumers shouLd be most interested, know the most about the product, and be most interested in buying the product. Hence, the "halo" effect.

Note also that all three of these measurements were taken at the same time and before the test began. This certainly raises questions about the "flow through time" aspect of the test.

RELATIONSHIPS BETWEEN THE VARIABLES

Not only is it necessary that rigorous planning be given to operationalizing variables, the same kind of attention must be paid to the nature of the relationships.

For instance, it is interesting to note that the three measures discussed above (motive, attitude, and intention) were all taken at the same time and before the test began. The authors appreciate the difficulty of collecting data at frequent intervals over long periods of time. Nevertheless, it seems that Farley and Ring have again opted for conveniences as opposed to rigor. As noted, this certainly raises questions about the "flow through time" aspect of the test. Furthermore, collecting data at five or ten-week intervals doesn't really allow a test of interrelationships between variables .

For example, the attitude-attention relation is measured in terms of attitude being assessed in the pre-mail questionnaire (scaled overall evaluation of the product) and attention being measured 11 weeks later in terms of recollection of seeing ads for the product. The authors are not sure of what this tells us in terms of an empirical test of the model.

Also, consider the relationship between stimulus ambiguity and attention (sensitivity to information as a function of uncertainty and/or lack of meaningfulness of information received from the environment).

This relationship is assessed by the respondent's ability to give an accurate description of the product during the fifth week as related to whether he can recall ads for the product during the tenth week. As Howard and Sheth suggest, (1969, p. 37) stimulus ambiguity may change in a single communication. Therefore, not only the operational definitions stand in the way of our learning enough from the test of this relationship but the time interval involved also serves to confound our attempts at correspondence between the theory and the empirical test.

ANOTHER APPROACH TO STUDYING CONSUMER BEHAVIOR

The following is an approach for researching the Howard and Sheth model that the authors deem more likely to advance knowledge in this area than the Farley and Ring approach. The essence of our approach is not new, but useful nevertheless. It simply consists of taking a "micro" look at a construct or constructs and trying to discover the form of the linkages these constructs have with other related variables.

As just one example, let us choose still another construct included in the Farley and Ring "model"--satisfaction. This construct has been defined by them as: "Post-purchase scaled satisfaction with the product, measured in second telephone interview." (Farley and Ring, 1970, p. 436) The corresPonding Howard and Sheth (1969, p. 145) definition is:

...the buyer's cognitive state of being adequately or inadequately rewarded in a buying situation for the sacrifice he has undergone. The adequacy is a consequence of matching actual past purchase and consumption experience with the reward that was expected from the brand in terms of its anticipated potential to satisfy the Motives served by the particular product class.

There are a number of elements included in this definition that do not emerge from Farley and Ring's definition. Most notably lacking are "sacrifice" and "matching consumption experience with expected reward." Ignoring these complications doesn't solve any problems in the long run. Three studies have been published that are particularly useful in pointing up the difficulties that arise in this area.

Cardozo (1965) conducted a study on satisfaction and expectation based on dissonance theory and contrast theory. His two hypotheses were: (1) a negative disconfirmation of expectancy (results were worse than expectation) would produce lower satisfaction; and (2) increased effort would increase satisfaction. His study confirmed both of his hypotheses. The reasoning was that contrast theory would make a product that did not live up to expectations less satisfying, but that effort expended would lessen this effect (as expected from dissonance theory).

A subsequent study by Olshavsky and Miller (1972) has obtained results that contradict Cardozo's main finding. Olshavsky and Miller's hypothesis is based on consistency theory. Their hypothesis was that high expectation-low performance would result in higher perceived quality (and hence, satisfaction) than would low expectation-low performance. In other words, the "subject's evaluations of performance tended to be assimilated toward manipulated expectations."

Anderson (1973) in a careful study, extended Olshavsky and Miller's approach by extending the limits of difference between expected quality and delivered quality. In doing so, he observed both assimilation and contrast effects.

The linkages that satisfaction is postulated to have with other constructs also have to be considered. A chain of events such as the following is suggested by Howard and Sheth (1969, p. 147).

FIGURE

Actual satisfaction is difficult to measure because expected satisfaction affects and distorts it. Expected Satisfaction also, of course, affects Final Satisfaction because it is one of the major inputs in the comparison process determining Final Satisfaction. And, as Final Satisfaction feeds back into Attitude, creating a revised Expectation, it too is difficult to measure.

An ideal situation for studying these complexities would involve tracking a new product through time. Through initial contact with panel members, one could obtain information about actual satisfactions (in this case, if product were "new" enough, it might be very close to final satisfaction). This, then, would translate directly into Attitude (forming an expected satisfaction). This process could be tracked through time. The essential nature of such a process is to show the form of the generalized relationship A = f (S, A) -- or as Howard and Sheth (1969, p. 147) have shown it: At + z = f(St +1 - At) + At.

By forming two or more such panels, one could try out various marketing mixes that would serve to manipulate expectation and hence satisfaction. The limits of the extent to which expectation affects final satisfaction could be studied. One of the big problems in conducting such a research program (and differentiating it from market testing as currently carried out) is discovering and providing for the time dimensions involved in closing either or both of the loops shown above.

In following these concepts through their cycles, it becomes evident that one rapidly loses the ability to measure Actual Satisfaction--it becomes confounded with Expected Satisfaction. And when we look closely at Final Satisfaction, defined as a comparison between consumption experience and expectation, we can see that both concepts are affected by prior attitude.

What we have to come to terms with is the fact that there is a circularity between Attitude and Satisfaction. Each helps to determine the other. And, in measuring each, we are really measuring part of the other.

The authors believe that the complexity present in studying Satisfaction is present in understanding all of the other variables in the model. Hence similar experiments would have to be conducted to determine relationships between other constructs. In closing, our position would be that the model as a whole provides a good basis for hypothesis formulation and testing. However, we-continue to believe that the whole model cannot be tested simultaneously, given the present state.

SUMMARY

We believe that there can be no conclusion other than that Farley and Ring have not tested a theoretical construct of buyer behavior and that their work, no matter how impressive its scope, should not be viewed as support for a workable, overall theory of buyer behavior. At this "state of the art" it seems unlikely to us that much useful progress will be made toward understanding consumer behavior by working "in the context of such a large and relatively clumsy formulation..." and Farley and Ring's example provides excellent evidence to back-up this view.

The operational and conceptual problems here are both large and real. Ignoring them because it is convenient can lead into the tenderest of analytical traps. As the pressures of consumerism and government regulation converge on marketing, particularly consumer goods marketing, we believe that it would be the greatest of disservices to promise any easy way out of the very thorns and complex problem of "why buyers behave as they do."

REFERENCES

Anderson, Rolph E., Consumer Dissatisfaction: The Effect of Disconfirmed Expectancy on Perceived Product Performance, Journal of Marketing Research, 1973, 10, 38-44.

Cardozo, "An Experimental Study of Customer Effort, Expectation, and Satisfaction," Journal of Marketing Research, 1965, 2, 244-249.

Farley, John V. and Ring, L. Winston, "An Empirical Test of the Howard-Sheth Model of Buyer Behavior," Journal of Marketing Research, 1970, 7, 427-438.

Howard, John A. and Sheth, Jagdish N., The Theory of Buyer Behavior, New York: John Wiley & Sons, Inc., 1969.

Hunt, Shelby D. and Pappas, James L., "A Crucial Test for the Howard Sheth Model of Buyer Behavior," Journal of Marketing Research, 1972, 9, 346-348.

Lutz, Robert J. and Resek, Robert W., "More on Testing the Howard-Sheth Model of Buyer Behavior," Journal of Marketing Research, 1972, 9, 344-345.

Olshavsky, Richard W. and Miller, John A., "Consumer Expectations, Product Performance, and Perceived Product Quality," Journal of Marketing Research, 1972, 9, 19-21.

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