Predicting Behavior: More Theory Needed

Richard W. Olshavsky, Indiana University
[ to cite ]:
Richard W. Olshavsky (1982) ,"Predicting Behavior: More Theory Needed", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 465-466.

Advances in Consumer Research Volume 9, 1982      Pages 465-466


Richard W. Olshavsky, Indiana University


The paper presented by Paul W. Miniard, Carl Obermiller, and Thomas Page is concerned with a topic quite different from the topic covered in Stephen M. Goldberg's paper. Therefore, each paper will be discussed separately.


The stated objective of Miniard, Obermiller, and Page's study was to determine if Warshaw's (1980) results could be attributed to "inequities in the measures' contextual correspondence." It is perhaps unfortunate that they failed to replicate fully Warshaw's results (i.e., in their noncontextual direct vs. conditional condition). In my opinion, however, their outcome and their inability to explain their outcome (to quote them, "No explanation is readily available.") is a fortunate development because it raises further doubts about the utility of the traditional theoretical relationship between intentions (BI) and behavior (B). If the measurement of BI is to prove managerially useful then it is essential that a valid theory concerning the relationship between BI and B be used. In this paper, I propose a quite different theoretical relationship between BI and B, a relationship that has quite different implications for intentions measurement than that pursued by Warshaw (1980) and by Miniard, Obermiller, and Page. But first, the important aspects of the traditional view must be stressed.

The Traditional Theoretical View of the BI-B Relationship

As pointed out by Miniard, et al and by Warshaw (1980), BI is viewed by most theorists as a variable that intervenes between cognitive variables (like beliefs (B') and attitudes (A)) and B (e.g., Engel, Kollat, and Blackwell 1978; Howard and Sheth 1969). And, as Warshaw (1980) pointed out, the specific nature of this relationship evolves from the Fishbein extended model (e.g., Ajzen and Fishbein 1973).

That is,

B = BI = Wo(Aact) + W1(NB).(MC).

The first important aspect of this theory of the BI-B relationship to note is that BI comes later in the chain o cognitive events which are hypothesized to determine behavior. This implies, as Warshaw (1980) states, that "intentions outperform attitudes, beliefs, and other cognitive measures as behavioral correlates" (p. 26). A second important aspect of this theory to note is that the relationship between BI and B is not precisely stated. The conditions which influence the degree of relationship between BI and B are outside the formal theory and take the form of verbal statements concerning the "degree of correspondence" that exists between BI and B. Correspondence is defined in terms of target, action, contest, and time (Ajzen and Fishbein 1980).

Because the observed empirical relationship between BI and B is frequently low and because the A:zen and Fishbein theory does not explicitly predict B (only BI), Warshaw (1980) and Miniard, et al made an effort to incorporate some of the variables Ajzen and Fishbein (1980) state are important into the measurement of BI. Warshaw (1980) argue that since "direct measures of BI confounds intentions and situational contexts" then "behavior-specific intentions should be derived from condition-specific measures" (p. 27).

As I see it, enormous difficulties arise from this solution to the problem. In particular, the difficulties arise from the large number of contingencies which may be determinant (i.e., target, action, context, and time). Even in the relatively simple purchase situation selected by Warshaw (1980), i.e., soft-drinks, the number of contingencies that may determine outcome (brand choice) is quite large. For instance, the brand of soft-drink purchased may depend not only on the general nature of the acquisition site (i.e., vending machine, store, restaurant, bar) and the number of brands purchased, as pointed out by Warshaw (1980) and by Miniard, et al. but also on several other factors such as the specific type of store involved (e.g., supermarket, discount, convenience) or restaurant (e.g., fast-food or gourmet) the specific type of other foods purchased (e.g., hamburger or Beef Wellington), the time of day, the presence or absence of others, the nature of other people in the shopping party, and the intended use of the product (e.g., beverage or mixer). Clearly this approach to the problem will fail if only because of the large number of variables and the lack of a comprehensive theory to guide the selection of likely determinant variables in each shopping condition. The only reason Warshaw (1980) avoided this problem is because he limited, quite arbitrarily and without any rationale, the number of contingent variables. Even so, the conditional format required nine times as many measures as the direct format in these studies.

A New View of the BI-B Relationship

Another solution to this problem is to develop a better theory. One drastically different theoretical view of the BI-B relationship I would like to propose is to assume that BI is not a variable that intervenes between the cognitive variables (like B' and A) and B. For instance, according to one version of information processing theory (Olshavsky 1975), B is determined directly by an interaction between characteristics of the consumer, the consumer's goal, and the characteristics of the task environment. According to this theory BI does not intervene between cognitive variables and B; indeed BI may play no role in determining B. Instead, BI is viewed as a prediction mate by the consumer concerning his or her own future behavior. As such, the formulation of an intention is itself a complex cognitive process akin to "attribution." The inputs to this prediction process may include those same consumer and task environment variables that determine behavior directly but in addition may include other inputs such as retrieved information concerning past behavior in this context.

Therefore, according to this theory, the degree of relationship between BI and B will vary with the inherent predictiveness of the behavior and the degree of sophistication of the consumer in recognizing and understanding which variables determine his/her behavior and the manner in which these variables interact to determine behavior. The degree of predictiveness of behavior is expected to vary with many of the some variables described by Ajzen and Fishbein, namely target. action. context. and time.

This new view of the BI-B relationship suggests that less effort should be directed at developing a better measure of BI and more effort should be directed at the development of a more comprehensive theory of behavior (not BI). Such a theory will explicitly incorporate those variables which Warshaw (1980) is presently trying to built into a measure of BI. Further, such a theory may provide a better explanation as to why BI and B are frequently observed to display a weak correspondence. Finally, this theory may help to explain why Miniard, et al. (1981) failed to replicate fully Warshaw's results.

In summary, the low correspondence frequently observed between BI and B and the incompleteness of the traditional theories of the BI-B relationship has lead Warshaw (1980) to seek a better measure of BI. I am arguing that this approach is doomed to failure if only because of the large number of contingencies involved and because of the lack of a comprehensive theory to determine, a priori, which of these contingencies will be important. Another solution lies in the development of a better, more comprehensive theory of consumer behavior (not intentions). And in particular, I propose a theory that eliminates BI as a variable that determines behavior; instead, intentions are viewed as the outcome of a separate cognitive process, namely the prediction of one's own consumer behavior.


The stated objective of Goldberg's study was to determine if "differences in consumer lifestyles provide a useful basis for predicting differences in brand loyalty." No specific hypotheses concerning the nature of these differences were presented. The only rationale provided for performing this study is that, "little research has attempted to relate consumer lifestyles to brand loyalty, in spite of the current emphasis on lifestyle variables as a basis for market segmentation and promotional objectives."

Although studies that are strictly empirical are generally defensible, the atheoretical nature of this study is a very serious shortcoming, because it is not intuitively clear just what relationship, if any, holds between lifestyle and brand loyalty. If one simply accepts the author's operationalization of "brand loyalty" and "lifestyle," there is still considerable ambiguity concerning the relationship between these two variables. For instance, what is the theoretical relationship between a lifestyle characterized by "passive sports enthusiast" and brand loyalty? According to the author, the positive relationship observed was consistent with "management's feelings," that is, those beer drinkers whose lifestyles have a closer fit to this description should be more brand loyal. The problem is, without a theory, just the opposite relationship could easily be hypothesized; that is, "passive sports enthusiasts" should be less brand loyal. Or, it could just as easily be hypothesized that being a "passive sports enthusiast" is unrelated to brand loyalty!

Furthermore, the atheoretical nature of this study prevents any clear expectations concerning the relationship between "attitude toward beer in general" and the lifestyle - brand loyalty" relationship.

As a consequence of the atheoretical character of this study it is very difficult to provide a meaningful interpretation of the results either for the development of theory or for guiding management strategy. The author interprets his findings as providing evidence that "lifestyle has a differential effect on brand loyalty for different consumer segments." Given his bias that such a relationship exists, the very low R2 (0.058) and the very low correlation coefficients (which ranged from 0.010 to 0.031) were interpreted as providing support for the expectation. The author attributes the low percentage of variance explained and low correlation coefficients to the intermediate factor analysis and to the individual as opposed to group level of aggregation involved. Even so, the absolute magnitude of the overall R2 was so small as to be of dubious value for managerial policy. Moreover, given that brand loyalty was measured both in terms of an attitudinal as well as a behavioral aspect, unambiguous implications for market share from the size and direction of the correlation coefficients cannot be de.

In summary, more theoretical effort is required to develop specific hypotheses between lifestyle and brand loyalty before managerially useful research can be conducted. Hopefully, such theoretical efforts will also serve to resolve many of the definitional problems surrounding these two concepts.


Ajzen, I., and Fishbein, M. (July 1973), "Attitudinal and Normative Variables as Predictors of Specific Behaviors," Journal of Personality and Social Psychology, 27, pp. 41-57.

Ajzen, I., and Fishbein, M. (1980), Understanding Attitudes and Predicting Social Behavior, Englewood Cliffs, NJ: Prentice-Hall.

Engel, J. Blackwell, R., and Kollat, D. (1978), Consumer Behavior, 3rd ed., Hinsdale, IL: Dryden Press.

Howard, J., and Sheth, J. (1969), The Theory of Buyer Behavior, New York: John Wiley and Sons. Inc.

Olshavsky, Richard W. (1975), "Implications of an Information Processing Theory of Consumer Behavior," in Edward M. Mazze (ed.), 1975 Combined Proceedings, Marketing in Turbulent Times and Marketing: The Challenge and the Opportunities. American Marketing Association, pp. 151-155.

Warshaw, Paul (1980), "Predicting Purchase and Other Behaviors from General and Contextually Specific Intentions," Journal of Marketing Research, 17, (February) pp. 26-33.