Promotion in a Static Market

John Howard, Columbia University
[ to cite ]:
John Howard (1981) ,"Promotion in a Static Market", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 9-11.

Advances in Consumer Research Volume 8, 1981      Pages 9-11


John Howard, Columbia University

[Paper given before the Annual Meeting of the Association for Consumer Research in reply to the honor of being appointed a Fellow in Consumer Behavior, Washington, D.C. October 17, 1980. I am grateful to Professor Wilfried Vanhonacker, my colleague, for his splendid comments on the paper.]


I am deeply honored to receive this symbol of the esteem of my colleagues. It is one of the finest achievements that I could hope for. As my response in accepting it, I would first like to make a brief assessment of the state of knowledge in the field of consumer behavior.


Remarkable progress has been made since about 1955 in developing theory that has relevance for practice. Let me apply this pragmatic definition of progress--creation of theory relevant for practice--and briefly evaluate research on each of the stages of choice that we, economists and psychologists have found useful (Howard, 1977) and especially now that Jim Bettman and Michael Zinns (1977) have given some empirical legitimization to that typology.

In the most common case--a buyer confronted with an unfamiliar brand in a familiar product class (limited problem solving)--research has gone exceedingly well by my pragmatic definition of progress. I need only cite the recent splendid technologically-oriented review by Shocker and Srinivasan (1979) which was possible only because of theory development.

Research on the case of a buyer confronted with an unfamiliar product class (extensive problem solving)--innovation--is going well. We have broken out of the confines of the rural sociologist's work in diffusion, which emphasizes the social processes and individual differences among buyers. Instead, we are focusing upon information processing (Bettman 1979) and upon values as the source of choice criteria (Howard 1979). Also, especially encouraging is Peter Wright's (1980) fascinating recent experiment on product class advertising.

Research on the third case--a buyer facing a familiar brand in a familiar product class (routinized response behavior)--however, has been disappointing, even though economists under the stimulus of Stigler's seminal article have also been lending a hand (Rosen 1976). I would like to address it at greater length.


Our RRB theory seems to me to be beset with a serious contradiction. Stated most baldly, our behavioral theory predicts that in a stable market consumers will learn about the array of brands and zero in upon the most preferred brand. This loyal behavior, then, will continue ad infinitum unless a change in the market occurs. Our prediction implies there should be no advertising because there is nothing to communicate.

But if we look around at the real world, it seems to me that we see far more contradictions of this prediction than we see consistencies, e.g. cigarettes before health warnings triggered a spate of minor product innovations. This conclusion, of course, assumes that managers spend their advertising budgets wisely. From the decision net studies of a number of executive decision processes in the early 1960's, I believe that for short-term decision-making this assumption is reasonably valid (Clarkson 1962); Howard and Morgenroth 1968). Also, I am ignoring the possibility of "reminder promotion" which tells the consumer, in effect, that the product class is available, which can cause shifts among product classes such as in food where there is flexibility across product class boundaries.

This contradiction of our theory has troubled me for several years. Not only is it blatantly obvious, but it involves a considerable proportion of our economy. Hence, whether our theory fits the facts in this case makes substantial difference. Further, we must remember that the success of our theory in the most common case--LPS--came as really quite a surprise to most of us. I was not nearly as confident that consumers used information in their buying as my words indicated. Because of my general skepticism, our failure in the simplest case--RRB--has had a profound effect in creating serious doubts in my mind about our entire venture into understanding the consumer.

In fact, in my concern about the contradiction, I have felt somewhat akin to Professor Weyl, the great mathematician who, in 1946, wrote," .... we are less certain than ever about the ultimate foundations of (logic and) mathematics... Like everybody and everything in the world today we have our 'crisis'... Outwardly it does not seem to hamper our daily work, and yet I for one confess that it has had a considerable practical influence on my mathematical life; it directed my interests to fields I considered relatively 'safe' and has been a constant drain on the enthusiasm and determination with which I pursued my research work."

Now, having recited my tale of woe, let me take a more positive view and conclude that the field even in the area of routinized response behavior is in better condition that I had thought.


In the 1960's, I became interested in Daniel Berlyne's work on the effect of novelty in esthetics and other areas of human behavior. Out of it I developed the ambiguity-arousal hypothesis which some of you might remember from the book with Professor Sheth (1969, pp. 202-220).

This hypothesis predicted that once consumers have zeroed in on their preferred brand in a stable market, they would, after a period of brand loyalty, become bored and shop around trying some alternative brands that are more novel, more interesting, and then again zero in on a preferred brand. A cyclic pattern of behavior would result: stability of brand loyalty, instability, stability, etc.

Not being able to think of an effective way to test it, I had forgotten all about it. Then, three years ago, I received in the mail a monograph on testing stochastic choice models done as a dissertation at Rotterdam University in Holland.

The author, B, Wierenga (1974), creatively develops the concept of "pool size." In this way his work went beyond that characterizing the stochastic choice research prior to the late 1970's. By defining "pool size" as the number of different brands bought in the last 10 purchases, he makes pool size the critical measure of behavior.

He uses three product classes of low-priced, frequently-purchased items to determine if the cyclic patterns predicted by the ambiguity-arousal hypothesis could be said to characterize the behavior of consumers. The products were beer, margarine, and a third unspecified item called "fopro." At least three of the brands for each product were national brands distributed over Holland as a whole and promoted by newspapers, television and radio on a national scale. The information was collected from a 2000-member commercial purchase panel for the two year period 1967 and 1968.

To be more specific, I will describe his method first as applied to the purchasing of a single family, but a different family for each product, as shown in Figure 1. S is the poolsize measured on the vertical axis and it is the number of different brands bought by the family in its last 10 purchases. Of course, t is time. At purchase St for fopro, for example, the poolsize is 2, as you will note. In the purchase sequence t-9, t-8, t-7, t-6, t-5, t-4, t-3, t-2, t-l, t, two different brands were purchased by this family. S was immediately reduced, however, to one particular brand because at purchase St+1., over the purchase sequence t-8, t-7, t-6, t-5, t-4, t-3, t-2, t-l, t and t+l, the family bought the same brand. This single brand purchasing continues with fopro for 28 purchases, at which time another brand, other than that originally bought, was purchased, which brings the total to t=29. Figure 1 implies that during the next 9 purchases, the pool will contain at least two brands and it is possible that a consumer in one time period completely switches to the new brand. If he did, the pool size of 2 lasts for exactly 9 purchases. After 10 purchases (t=39), the pool size returns to 1, which indicates that no further purchase of the new brand was made while no other brands were tried in the meantime. After that, there is a period of no brand-switching which lasts for 23 purchases (t=62). Again, an incidental purchase of a different brand occurs followed by a short rest (t=80). Then a period of intensive brand-switching begins where the pool size increases to as much as 4. Following this turbulent period, the family seems to have made its choice and continues for a long time buying the same brand.

As you see in Figure 1, fopro and beer show each one period of extreme search. "Search" is defined as when, during a certain period, the buyer tries a number of different brands before settling on a particular brand as contrasted with a straightforward switch from one favorite brand to another favorite brand. Margarine exhibits two such periods. Also, fopro was much more inclined to single brand periods than the other two products.

In Figure 1 we have seen in single households evidence of cycles where, after periods of intensive search, the family alternates with periods of continuous buying of the same brand within a period but not necessarily across periods. A similar picture emerges when all 2000 households are thrown together.

Unfortunately purchase patterns cannot be shown diagrammatically for more than a single family. Instead, they can be described by using the range of difference between the maximum and minimum of pool size over all families. The results for families buying two or more brands are shown in Sable 1. Families that bought only one brand during the two-year period of the data are excluded because we want to show the variation in switching required to conform to the search-loyalty cycle. These continuous repeat purchasers constitute roughly a fourth (fopro = 28%, beer = 27%, margarine =16%) for all three brands.



As you can see in Table 1, where RANGPLZ means range of variation in pool size, a substantial difference in levels of switching activity occurred. In fopro, for example, 2.691 exhibited no variation, while 46.69% showed a variation of 1. This is consistent with the search-disloyalty cycle concept. Specifically, the pool size is not constant but has periods of low values and periods of high values.



But we should ask ourselves whether these patterns of cyclic behavior could have been the result of promotion or other marketing activity instead of the ambiguity-arousal hypothesis? If we examine the pool sizes of different consumers at a given point in time, we find they are different. They are not low and high together. If the source of the cyclic pattern was a common marketing activity, the lows and highs of different consumers would tend to occur together in time.

But once they have decided to shift, marketing effort does appear to make a difference in which way they shifted, to which brands. Specifically, the start of search is often associated with changes in price, deals, shop where bought and unit-size bought, which supports the notion that promotion is justified as implied by the ambiguity arousal hypothesis. The absolute number and proportion of each of these promotional events associated with brand switches are seen in Table 2.



Additional support was also given the cyclic hypothesis in this study in that the linear learning model, which implies cyclic behavior, gave the best fit to the data among the various stochastic models (Wierenga p. 184).

We must be careful, however, in generalizing the results of the Dutch study to the United States. My colleague, Professor Wilfried Vanhonacker, informs me that most American studies have typically found zero order probability and with a pool size of about 4, although with some exceptions.

What I conclude is that hidden within this high shifting in the American market is a substantial number of customers exhibiting cyclic behavior as predicted by the ambiguity-arousal hypothesis. Not as high as the roughly 80% in Holland, but enough to justify the marketing effort we see in stable markets in the United States. This will, of course, have to be verified and I throw it out as a challenge to fellow researchers.


This finding renews my faith in progress in our work now that I have been provided an answer for the one gnawing question that has plagued me for years. It has implications for both private and public policy. I have emphasized the private policy, but its public policy significance can be still greater. It can be at least a partial answer to charges of pseudo-differentiability and shared monopoly, for example.


Bettman, J. R., (1976). An Information Processing Theory of Consumer Choice. Reading, Mass: Addison-Wesley.

Bettman, J. R. and Zinns, M. A., (1977). "Constructive Processes in Consumer Choice," Journal of Consumer Research, 4, 75-85.

Clarkson, G. P. E., (1962). Portfolio Selection: A Simulation of Trust Investment. Englewood Cliffs, New Jersey, Prentice-Hall, Inc.

Howard, J. A., (1977). Consumer Behavior: Application of Theory. New York: McGraw-Hill Co.

Howard, J. A., and Morgenrath, W. M., (1968). "Information Processing Model of Executive Decision," Management Science,14, March, 416-428.

Howard, J. A., and Sheth, J. M., (1969). The Theory of Buyer Behavior. New York: John Wiley & Sons, 157-165.

Rosen, S., "Advertising Information and Product Differentiation," University of Rochester, mimeo, 1976.

Weyl, H., (1946). 'Mathematics and Logic," American Mathematics Monthly, 53, 1-13.

Wierenga, B., (1974). An Investigation of Brand Choice Processes, University of Rotterdam.