Consumer Research and Marketing Science

John R. Rossiter, University of Technology, Sydney
ABSTRACT - It is the thesis of this paper that consumer researchers have drifted away from marketing science and that this trend is extremely dangerous. Evidence is marshalled to illustrate this trend. The danger of losing touch with marketing science is exemplified for consumer research by examining three major topics: the emergence of scanner data in marketing science which threatens to empirically rewrite consumer behavior theory; ignorance of the behavioral facts of consumer purchase behavior discovered by marketing scientists; and failure to consider the marketing science concept of stochasticity as one of the causes of consumer brand choice. We conclude that marketing science must be reintegrated with consumer behavior theory and research.
[ to cite ]:
John R. Rossiter (1989) ,"Consumer Research and Marketing Science", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 407-413.

Advances in Consumer Research Volume 16, 1989      Pages 407-413

CONSUMER RESEARCH AND MARKETING SCIENCE

John R. Rossiter, University of Technology, Sydney

ABSTRACT -

It is the thesis of this paper that consumer researchers have drifted away from marketing science and that this trend is extremely dangerous. Evidence is marshalled to illustrate this trend. The danger of losing touch with marketing science is exemplified for consumer research by examining three major topics: the emergence of scanner data in marketing science which threatens to empirically rewrite consumer behavior theory; ignorance of the behavioral facts of consumer purchase behavior discovered by marketing scientists; and failure to consider the marketing science concept of stochasticity as one of the causes of consumer brand choice. We conclude that marketing science must be reintegrated with consumer behavior theory and research.

INTRODUCTION

Consumer research is losing touch with marketing science. This trend has been qualitatively apparent, to regular attenders of ACR conferences, from the time since ACR and The Institute of Management Sciences (TIM )) ceased to hold their conferences adjacently in the same location. The separation has become overwhelmingly evident in the last three years with the "new directions" that ACR has taken (see, for example, Holbrook 1987). The impending divorce of marketing science from consumer research seems quantitatively apparent, too, in an informal count of the proportion of "marketing science-type" articles published in the Journal of Consumer Research (Figure 1). There has been a marked decline in such articles in the last couple of years. As a further indicator of disciplinary separation, only 4.6 percent of names in the 1987 ACR Membership Directory are people known by their publications to have an interest in marketing science. [For obvious reasons, these data are not available from the author, but others can do their own count.]

It will be argued in this paper that the now almost complete divorce from marketing science by consumer behavior theorists and researchers is extremely dangerous. It is symbolically appropriate in Hawaii, which is perhaps the most qualitative of ACR conference venues that we have visited to date, that we should pause and reflect on the consequences of this trend.

The danger in ignoring marketing science is illustrated in this paper by examining three main topics that have major implications for consumer behavior theory and research:

1. The potential of the burgeoning amount of scanner data and results, which is primarily the province of marketing scientists, to empirically test our previous speculative consumer behavior theories and indeed to rewrite consumer behavior theory itself.

2. The large blind spot in the U.S. consumer behavior textbooks regarding the behavioral facts of consumer behavior, especially as represented by the work of the British marketing scientist, Ehrenberg (e.g., 1972, 1987), and by behavioral norms published by market research companies.

3. Inattention to the possibility that stochasticity may play a causal role in consumer brand choice rather than being merely a statistical descriptive device that marketing scientists use to account for inexplicable variations in purchase behavior.

The three topics will be expounded and then some conclusions will be suggested for the "new" ACR.

SCANNERS: BOOTSTRAPPING CONSUMER BEHAVIOR THEORY?

Until recently, consumer researchers who have developed theories about how consumers buy in response to advertising, sales promotion, price, and other marketing stimuli have had to test these theories by either small-scale laboratory research (e.g., Jacoby's 1974, 1977 studies of information overload theory) or by inference from aggregated sales data (rare among ACR researchers but common among managers-as-consumer-researchers in industry, who use Nielsen, SAMI, and other aggregate sales measures). However, with the emergence of scanner data on individual household purchases and, to some extent, earlier, household diary panel data, consumer researchers for the first time have the opportunity to measure real consumer behavior in response to marketing stimuli in the actual marketplace. Scanner data can be used at the store level to record responses to store-mediated stimuli such as price or in-store promotions (e.g., Guadagni and Little 1983) or in a household panel (particularly Information Resources, Inc.'s BehaviorScan service) whereby individual household receipt of marketing stimuli can be experimentally varied by accessing a whole town or community via cable TV, for TV advertising, or by random mail or door-to-door delivery, for other media advertising or promotions, and then arranging with the town or community's retailers to accept electronic I.D. cards to record purchases from a panel of residents. Obviously, never before have consumer researchers had such an ideal "real-world laboratory" with which to test their theories.

The results of scanner-panel tests are starting to pour out of the marketing science literature. It is indeed fortunate that the originators of some of the scanner panel services were academics (notably John Little of M.I.T. and Len Lodish of Wharton who began Management Decisions Systems which later merged with Information Resources, Inc., but also others) who realized the value of releasing their data and results to the academic audience. The main journal in the discipline, Marketing Science, is beginning to carry academic reports of scanner panel experiments (e.g., Abraham and Lodish 1987; and Blattberg and Levin 1987 who provide a direct test of Blattberg, Eppen and Lieberman's 1981 "inventory transfer" theory of trade promotions). More recently, the TIMS/ORSA conference in Seattle (March 1988) featured over 20 papers on scanner data results and scanner panel experiments. For a consumer researcher interested in real-world behavioral responses to marketing stimuli, what a goldmine!

FIGURE 1

TREND IN PROPORTION OF "MARKETING SCIENCE-TYPE" ARTICLES

The scanner-researchers are primarily pursuing an "empirical" or inductive approach: they look at the data first, and then try to extract generalizable principles, that is, theories, especially theories that can serve as the basis for expert systems that managers can use. Essentially, this empirical approach to theory development is going to have the effect of "bootstrapping" the theories developed by the more deductive approach of consumer researchers, which is to formulate a general theory and then test it by particular observations in the laboratory. This bootstrapping is going to take time, because the empirical field settings and experiments (inductive observations) differ in many respects from one study to another. But you can't argue with the data, because they are real. As the inductive observations accumulate over many diverse instances with known dimensions of diversity, the consumer researchers' laboratory-tested theories about consumer behavior are in many cases going to be radically revised. To pick two examples: Scott's (1976) theory about the effects of promotion-aided trial on repeat purchase behavior is being tested in many scanner panels right now although the panel researchers may not know it; similarly Dodson, Tybout and Sternthal's (1978) theory that smaller face-value promotions lead to more brand loyalty than larger face-value promotions when the promotion is withdrawn is also implicitly the subject of many scanner panel studies.

Herein lies the critical importance of consumer researchers getting back into collaboration with the marketing scientists who have access to scanner data. Consumer researchers can help to explain the empirical results and can suggest new experiments to test consumer behavior theories in the deductive mode. Consumer researchers can guide marketing scientists who are "mining" scanner data by telling the goldminers "where to dig." Numerous instances of the potentiality of this collaboration were witnessed by the few consumer researchers who attended the TIMS/ORSA conference in Seattle earlier this year. Collaboration between consumer researchers and marketing scientists must be encouraged to maximize the value of this new-found data source for consumer behavior theory.

IGNORING THE BEHAVIORAL FACTS

A second major danger in consumer researchers splitting from marketing science is the appalling ignorance, or ignoring, of the fundamental behavioral facts which our discipline seeks to explain. This is especially true of U.S. consumer researchers, who seem to exhibit a large blind spot when it comes to looking at the empirical realities of purchase behavior.

To support the accusation of a blind spot, let's look at a basic example, first from a simple perspective and then from a more elaborate perspective. Suppose that a marketing manager or even a student doing a consumer behavior project wants to introduce a new brand of a typical supermarket product in an existing product category. What level of trial or "market penetration" and what market share can the new brand hope to attain? The leading (U.S.) textbooks in consumer behavior provide no help at all. [For this purpose, I surveyed 10 leading U.S. textbooks that had Consumer Behavior as their title: Alphabetically, the texts were: Assael (1987); Engel, Blackwell, and Miniard (1986); Hawkins, Best, and Coney (1986); Howard (1977); Loudon and Della Bitta (1984); Peter and Olson (1987) which is supposed to be 50 percent behavioral in content; Robertson, Zielinski, and Ward (1984); Schiffman and Kanuk (1983); Wilkie (1986); and Zaltman and Wallendorf (1983).]

The marketing scientists supply the answers. As to trial, the new brand can expect 15 percent to try it; as to market share, the likely maximum share attainable is 18 percent. NPD Research, Inc., which operates two national purchase panels, publishes normative data like this quite regularly- (e.g., Johnson 1984; Rubinson 1986). For example, the following data from Rubinson 1986) are based on over 80 supermarket product categories (Table 1). The declining norms for trial and (first) repeat purchase in recent years pose an interesting trend for consumer researchers to explain. The explanation probably lies in the increasing number of brand alternatives per category which are not real alternatives but are simply line extensions (Rubinson 1986) that attract fewer triers. The norms in Table 1 are of course simplified averages, although NPD has norms for various product categories that are occasionally made public in limited form.

A more elaborate answer to the basic consumer behavior question posed by our hypothetical marketing manager or marketing student is provided in the brilliant work of British marketing scientist Andrew Ehrenberg. Again, and incredibly, a search of the same 10 U.S. consumer behavior textbooks finds no reference to Ehrenberg in eight of them, a half-page summary (no data) in Howard (1977), and a minor footnote in Engel et al. (1986). Whereas this omission may be due partly and inexcusably to American geocentrism, it's also due to the failure of U.S. (academic) consumer researchers to thoroughly study the very behavior in which they are supposed to be experts.

Ehrenberg and his colleagues (e.g., Ehrenberg 1972; Goodhardt, Ehrenberg and Chatfield 1984; Keng and Ehrenberg 1984; Wellan and Ehrenberg 1988) have shown, on the basis of more data sets studied than perhaps any other research team, the remarkable aggregate predictability and "lawfulness" of consumer behavior--for store patronage, consumer durables, and not only supermarket products. Table 2, for example, from Ehrenberg (1987), shows the penetration, repeat rates, and market shares in the U.S. instant coffee market circa 1985. Look at the regularity of consumer behavior with which any consumer researcher must contend. There is clearly a direct relationship between penetration (trial) and market share, with repeat rates being relatively constant across brands and playing little part in market share determination. [The data in Table 2 illustrate the widespread phenomenon that Ehrenberg has called the "Double Jeopardy" effect: that smaller penetration brands also tend to be bought less often, thus doubly jeopardizing their market performance.] The more sophisticated answer, therefore, to our new brand introduction question is that the new brand's likely market share depends on the penetration attained, given that the repeat rate in the product category is known. [An alternative estimate is given by the Hendry system, based on order of entry (Rossiter and Percy, 1987). A new brand, for the average category, can expect to attain market share = 100 x (.43)n-1, where the new brand is the nth brand. For qualification of the order-of-entry effect, see Urban, Carter, Gaskin, and Mucha (1986).] A new instant coffee brand, for instance, hoping to attain a 10 percent market share, would have to gain trial by at least 18 percent of households (cf. Folger's in the table) given that its repeat rate is normal.

TABLE 1

TRIAL AND (FIRST) REPEAT NORMS FOR THE AVERAGE NEW SUPERMARKET BRAND

The same sort of overall regularity of consumer behavior holds for store patronage. And it is very predictable. Table 3 shows store patronage for instant coffee from a U.K. panel (hence the store names may be unfamiliar) in terms of the observed "store penetration" and "store repeat rates" compared with the theoretical predictions of these from the Dirichlet stochastic model employed by Keng and Ehrenberg (1984). Again the regularity and predictability are remarkable.

Why haven't these types of behavioral facts of consumer behavior been reported in the U.S. consume behavior textbooks? Moreover, one rarely, if at all, hears or sees reference to them at ACR conferences or in JCR. Their importance is unarguable and central: the behavioral regularities of purchase behavior in the product category constitute the "norms" that we--as consumer researchers--must "beat," using our psychological theories of how to better produce responses to marketing stimuli. Managers-as-consumer-researchers, of course, have the same goal. But academic consumer researchers have the additional task of having to explain these facts. [For instance, the Dirichlet model used by Ehrenberg, like most stochastic models, is purely descriptive: there is no consumer behavior theory or explanation behind it.]

STOCHASTIC CAUSES?

The final illustration of the danger of our impending divorce from marketing science is seen in the concept of stochasticity. Almost 15 years ago, Bass (1974, p. 1) advanced the radical proposition that "brand choice behavior is substantially stochastic." In the zero-order extreme version of stochastic or probabilistic brand choice, the consumer, each time she faces a brand choice in the store, metaphorically rolls a multi-faced die to make her choice; the die's faces are "loaded" proportional to each brand's market share. [In the best-fitting stochastic models (e.g., the BBD or NBD-type models, including the Dirichlet), each consumer has her own die weighted by personal (previous brand choice) purchase.] There is no cognition, no memory of past choices, just a simple series of metaphorical dice throws.

Stochastic models describe consumer brand choice behavior extremely well. At the aggregate or total market level, over the course of a year, it is not uncommon to find a correlation of greater than .90 between model-predicted and actual brand-switching patterns in the product category. But at the individual consumer (or more usually individual household) level, too, it is possible to be able to predict over 70 percent of the repeat purchases and switching occasions that an individual will experience in purchases over the year (e.g., Winter and Rossiter 1988) by using a model that has at least some degree of stochasticity in it. There is no fully deterministic (precise causal relationships) model developed by consumer researchers--for example, multiattribute attitude models--that can come any where near this degree of predictive accuracy in the real world.

TABLE 2

PENETRATION, REPEAT RATES, AND MARKET SHARES IN THE U.S. INSTANT COFFEE MARKET (CA. 1985)

TABLE 3

U.K. INSTANT COFFEE: STORES (CHAINS) OF PURCHASES OVER 24 WEEKS, OBSERVED AND THEORETICAL

The evidence for zero-order brand choice behavior, in which there is no cognitive carryover whatsoever from one purchase to the next, is mounting. In a particularly damaging (for consumer researchers) empirical tour de force, Bass, Givon, Kalwani, Reibstein, and Wright (1984) demonstrated that, for nine frequently-purchased supermarket product categories, the hypothesis of a zero-order brand choice process could not be rejected at the 10 percent significance level for an amazingly high 70 percent of households. [Our computations based on Bass et al.'s (1984) data revealed that the average across the nine products for "stationary" households, whose probabilities of purchase for each brand remained constant over the year of the panel, was 74 percent; for "nonstationary" households, it was 63 percent. The weighted average was 70 percent of households.] This high level was observed also for products such as white bread, margarine, and sugar, where it is difficult to argue that the "randomness" or zero-order nature of choice was spuriously due to the interspersing of different family members' choices within the household's overall purchase record (Kahn, Morrison, and Wright 1986).

Of course, "zero-order" does not necessarily mean "zero-order stochastic." As Bass (1974, p. 2) pointed out, the choice process could be fully determined by a multitude of variables such as recent advertising impressions, point-of-purchase promotions, out-of-stocks, and so forth, that make a fully-caused choice appear to be random, and which co-occur with unpredictable frequencies on each purchase occasion so that choice becomes zero-order. However, it's hard to believe that conditions change that much from one purchase occasion to the next. Rather, it seems that Bass' alternative hypothesis (1974, p. 2) of a "stochastic element in the brain," or at least some sort of stochastic causal process, must be equally entertained.

In a recent paper, Winter and Rossiter (1988) have proposed a consumer brand choice model that contains a substantial stochastic component. This model postulates, not that there is a random element in the brain, but that consumers voluntarily and deliberately engage in random (stochastic) choice behavior from time to time to satisfy a need for variety. This model states that consumers develop an individual purchasing "pattern" for the product category which indicates when to repeat in a brand loyal sense and when to "go stochastic" and simply select a brand probabilistically off the shelf. At the individual consumer level, this model predicts repeats and stochastic selections extremely well. And, note, it postulates stochasticity as one major cause of consumer brand choice, moving stochasticity from descriptive to explanatory status.

This (not so) new perspective of stochastic causality is emanating from marketing science. Again, it has been 15 years since Bass (1974) criticized consumer researchers' models such as those of Howard and Sheth (1969) or Engel, Kollat, and Blackwell (1968) for overemphasizing deterministic causes of brand choice. After 15 years of trying, it is abundantly clear that traditional consumer researchers' theories just cannot cope if they continue to overlook stochasticity. It is the marketing scientists who will force us to revise our theories to accommodate the now undeniable empirical challenge posed by stochasticity that we have heretofore ignored. Stochasticity must be regarded as a likely cause of consumer behavior.

CONCLUSIONS AND IMPLICATIONS

The theme of this year's ACR conference is that "multiple purposes, philosophies, and methods guide the generation^of knowledge in the field of consumer behavior" (ACR Conference 1988 Call for Papers). The contention in this paper, based on observing what is going on ag,_he now-separate ACR and TIMS conferences and in their respective journals, is that consumer researchers are in grave danger if they continue to divorce themselves from marketing scientists.

Three topics were examined to demonstrate the danger of not reintegrating consumer research with marketing science:

1. Empirical results from the marketing scientists' greater access to consumer purchase data from scanner panels threatening to inductively rewrite our largely deductive consumer behavior theories.

2. The incredible blindness, shown by U.S. consumer behavior textbook writers in particular, towards the fundamental facts of consumer behavior as discovered by marketing scientists--the very behavior which our theories hope to explain.

3. Failing to address the mounting evidence for -Frank Bass' 15-year-old claim that "brand choice behavior is substantially stochastic" by not recognizing that stochasticity may be a cause of consumer behavior.

Reconciliation of the marriage breakdown between consumer researchers and marketing scientists is imperative if consumer researchers are to meet the fundamental challenge of the question "Can you explain consumer behavior?" being asked of us by our previous partners. Consumer researchers can and must help to explain the facts of consumer behavior that are increasingly being revealed to us by the marketing scientists .

Practical solutions for consumer researchers are quite evident: invite our marketing science colleagues back to our ACR conferences to insure marketing science input to our discipline; establish an interim policy of favoring more marketing science-based articles in JCR [By my count, there are at least 13 of the 56 editorial board members of JCR who have a primary or secondary research interest in marketing science. This approaches the incidence to which JCR's publication of marketing science-type articles has fallen, as suggested in Figure 1 earlier.]; personally, keep current with the marketing science literature; and, above all, start "dating" our former marriage partners lest they leave us "single" and out of touch with what's going on.

REFERENCES

Abraham, M.N., and L.M. Lodish (1987), "Promoter: An Automated Promotion Evaluation System," Marketing Science, 6 (Spring), 101-123.

Assael, H. (1987), Consumer Behavior and Marketing Action (3rd ed.), Boston, MA: Kent.

Bass, P.M. (1974), "The Theory of Stochastic Preference and Brand Switching," Journal of Marketing Research, 11 (February), 1-20.

Bass, F.M., M.N. Givon, M.U. Kalwani, D. Reibstein, and G.P. Wright (1984), "An Investigation into the Order of the Brand Choice Process," Marketing Science, 4 (Fall), 267-287.

Blattberg, R.C., and A. Levin (1987), "Modelling the Effectiveness and Profitability of Trade Promotions," Marketing Science, 6 (Spring), 124-146.

Engel, J.F., D.T. Kollat, and R.D. Blackwell (1968), Consumer Behavior, New York: Holt, Rinehart and Winston.

Engel, J.F., R.D. Blackwell, and P.W. Miniard (1986), Consumer Behavior (5th d.), Chicago, IL: Dryden.

Dodson, J.A., A.M. Tybout, and B. Sternthal (1978), "Impact of Deals and Deal Retraction on Brand Switching," Journal of Marketing Research, 15 (February), 72-81.

Ehrenberg, A.S.C. (1972), Repeat-Buying, Amsterdam, Netherlands: North-Holland.

Ehrenberg, A.S.C. (1987), "Buyer Behavior and NPD," Working Paper, London, England: London Business School.

Goodhardt, G.J., A.S.C. Ehrenberg, and C. Chatfield (1984), "The Dirichlet: A Comprehensive Model of Buying Behavior," Journal of the Royal Statistical Society A, 147, 621-655.

Guadagni, P., and J.D.C. Little (1983), "A Logit Model of Coffee Choice Calibrated on Scanner Data," Marketing Science, 2 (Summer), 203-238.

Hawkins, D.I., R.J. Best, and K.A. Coney (1986), Consumer Behavior (3rd ed.), Plano, 1X: Business Publications, Inc.

Holbrook, M.B. (1987), "What is Consumer Research?" Journal of Consumer Research, 14 (June), 128-132.

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

Howard, J.A., and J.N. Sheth (1969), The Theory of Buyer Behavior, New York: Wiley.

Jacoby, J., D.E. Speller, and C.A. Kohn (1974), "Brand Choice Behavior as a Function of Information Load: Replication and Extension," Journal of Consumer Research, 1 (June), 3342.

Jacoby, J., G.J. Szybillo, and J. Busato-Schach (1977), "Information Acquisition Behavior in Brand Choice Situations," Journal of Consumer Research, 3 (March), 209-216.

Johnson, T. (1984), 'The Myth of Declining Brand Loyalty," Journal of Advertising Research, 24 (January), 9-17.

Kahn, B.E., D.G. Morrison, and G.P. Wright (1986), "Aggregating Individual Purchases to the Household Level," Marketing Science, 5 (Summer), 260-268.

Keng, K.A., and A.S.C. Ehrenberg (1984), "Patterns of Store Choice," Journal of Marketing Research, 21 (November), 399-409.

Loudon, D.L., and A.J. Della Bitta (1984), Consumer Behavior (2nd ed.), New York: McGraw-Hill.

Peter, J.P., and J.C. Olson (1987), Consumer Behavior, Homewood, IL: Irwin.

Robertson, T.S., J. Zielinski, and S. Ward (1984), Consumer Behavior, Glenview, IL: Scott, Foresman.

Rossiter, J.R. and Percy (1987), Advertising and Promotion Management, New York: McGraw-Hill.

Rubinson, J. (1986), "Sales Simulation--New Marketing Realities Produce New Testing Needs," Marketing Review, New York Chapter of the American Marketing Association, 42 (November), 13-15.

Scott, C.A. (1976), "The Effects of Trial and Incentives on Repeat Purchase Behavior," Journal of Marketing Research, 13 (August), 263-269.

Schiffman, L.G., and L.L. Kanuk (1983), Consumer Behavior (2nd ed.), Englewood Cliffs, NJ: Prentice-Hall.

Urban, G.L., T. Carter, S. Gaskin, and Z. Mucha (1986), "Market Share Rewards to Pioneering Brands, Management Science, 32 (June), 645-659.

Wellan, D.M., and A.S.C. Ehrenberg (1988), "A Successful New Brand: Shield," Journal of the Market Research Society, 30 (January), 35-44.

Wilkie, W.L. (1987), Consumer Behavior, New York: Wiley.

Winter, P.L., and J.R. Rossiter (1988), "Pattern-Matching Purchase Behavior and Stochastic Brand Choice: A Low Involvement Product Category Model," Working Paper, Graduate School of Management, University of California, Los Angeles.

Zaltman, G., and M. Wallendorf (1983), Consumer Behavior (2nd ed.), New York: Wiley.

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