A Micro Analytic Simulation of Consumer Behavior: Economic and Behavior Theory Antecedents

Gerald Eskin, The University of Iowa
M. Venkatesan, The University of Iowa
[ to cite ]:
Gerald Eskin and M. Venkatesan (1974) ,"A Micro Analytic Simulation of Consumer Behavior: Economic and Behavior Theory Antecedents", in NA - Advances in Consumer Research Volume 01, eds. Scott Ward and Peter Wright, Ann Abor, MI : Association for Consumer Research, Pages: 247-256.

Advances in Consumer Research Volume 1, 1974    Pages 247-256

A MICRO ANALYTIC SIMULATION OF CONSUMER BEHAVIOR: ECONOMIC AND BEHAVIOR THEORY ANTECEDENTS

Gerald Eskin, The University of Iowa

M. Venkatesan, The University of Iowa

I. INTRODUCTION

Our principal task here is to present some relationships between a relatively new modeling approach, sometimes labeled "Micro Analytic Simulation" and relevant economic and behavioral concepts. The literature reveals no satisfactory formal definition of Micro Analytic Simulation. Therefore, we shall specify a set of attributes which seem to characterize this modeling approach. The principal features appear to be:

(1) Micro - They are micro in the sense that they describe the behavior of individual decision maker (households, buyers, or consumers) rather than macro (group or market level) decision processes.

(2) Analytic - They tend to involve a specification of causal structure which purports to "explain" why various decisions are made rather than simply claim ability to forecast outcomes given knowledge of environmental conditions.

(3) Simulation - They tend to involve simulation in both the noun and verb sense. In the first sense, they describe or simulate the decision making process. In the second (verb) sense, the procedure used for generating forecasts and searching for optima is to run the model as opposed to the more conventional process of solving models by analytical methods. The simulation approach is necessary in most cases due to the complex and interactive nature of the model's structure

(4) Stochastic - Uncertainty components usually play an important role in the models. First, for a given set of environmental conditions and consumer characteristics, decisions for like consumers may vary. Second, the environmental conditions facing the consumer may be generated by a stochastic mechanism. In some model event the characteristics of the consumers are internally constructed by probabilistic processes.

(5) Latent Attitude Structure - Decisions are defined in terms of observable behavior such as choice of product or brand. These decisions tend to be determined or at least preconditioned by some form of attitude construct which, in turn can be decomposed into more basic latent constructs sometimes called attributes or characteristics. Thus consumers are presumed to "prefer" or "need" various goods because of some particular set of attributes or benefits forthcoming from purchase or consumption.

(6) Quantitative and Operational Form - This type of model exists in the form of computer code, capable of producing numerical outputs such as forecasts of number of persons purchasing various brands under given conditions. This reflects the quantitative, operational nature of the model.

This attribute of Micro Analytic Simulation is of considerable interest in that any theoretical concepts embedded in such a model will be by its very nature an operationalized version. It is unambiguous, operational and susceptible to validation. Next, a particular Micro Analytic Simulation Model will be described.

II. THE MODEL

Situation

Construction of the model was prompted by managerial uncertainty related to a product mix question. A cake mix manufacturer had just completed the development and national introduction of a line of premium quality products. This line varied from other mixes in that its shortening agent was real butter. The product was to sell at a substantial price premium compared to regular cake mixes. Shortly after market introduction and prior to the time that the firm could obtain a clear share reading on the product, two major competitors each introduced new products which appeared to compete with Premium Cake. These other new products were priced at about the same level as most other cakes (well below Premium Cake) but had no shortening whatsoever. The consumer was expected to add her own butter and hence produce a better cake.

The managerial concern related to the desirability of a change in product line strategy. Alternatives considered included (i) adding an "Add Butter" product to the line; (ii) replacing Premium Cake with Add Butter; or (iii) keeping the line as is (that is--with Premium Cake). The advisability of each strategy depended primarily on the share of market forthcoming under each alternative and the degree of substitution with the current line. Modeling was undertaken with the primary objective of making forecasts conditional on each hypothesized product line configuration and estimating the source of share substitution.

The Model was constructed by the marketing manager for the product, the market research analyst assigned to the product, and a specialist in model building. Each devoted about half time to the project over a 10 week time period. The time frame was determined by managements' desire to make a product change decision by a specified date.

Model Structure

A flow chart describing the model is displayed in Exhibit I. Its main components are discussed below.

Select Use Category

The principal segmenting variable in the model is Use Category or Occasion. It was presumed that cakes were generally purchased with a particular use occasion in mind and that different types of cakes would be suitable for different occasions. Occasions considered were:

Snacks

Everyday Family Dinner

Special Occasions

Perceived Need

The product need or desires on each occasion was thought of in terms of various components or characteristics. Components considered were: Quality Level, Price, A convenience-Involvement index and a Taste-Texture index.

EXHIBIT 1

DECISION FLOW

Quality was related to the norm for the product class where most cakes were perceived by the consumer as being similar. This normal level of quality might be adequate for most consumers, most of the time, but on special occasions many consumers desired higher quality. On some occasions lower levels might be acceptable (such as an afternoon snack for the children).

The convenience-involvement index takes on two possible values, either a consumer is convenience-oriented in which case she (he) wishes the simplest, least time consuming product. Alternatively, she (he) may take positive pleasure from involvement, wishing to add something of herself in the cake preparation process. The same consumer may be convenience-oriented when preparing dinner or snack and involvement-oriented when preparing something for a special occasion.

A Taste-Texture index is included because the new Add Butter products possessed a heavier texture and stronger-richer taste than most of the other products in the market.

Perception of Products

Products were described in the same terms as perceived need. That is, they could differ in terms of Quality, Price, Convenience-Involvement or Taste-Texture.

Quality: Regular cake was taken as the standard. Add Butter and Premium Cake were perceived as superior quality by some consumers but not by others.

Price: The market included some brands of lower than average price (here called Price Brands). Consumers all perceived Premium Cake to be high priced. In the case of Add Butter the price perception depended on whether the consumer in question added the cost of the butter to the base price in determining her perception.

Convenience-Involvement: Using Add Butter involved more involvement while other products were more convenient.

Taste-Texture: All products were similar except for Add Butter which had a heavier texture and a stronger, richer taste. The majority of consumers preferred the "normal" light-fluffy cake.

Prices and Dealing

In the absence of deals, consumers looked for products that matched their perceived need. In general, consumers were willing to pay the normal price associated with the quality level being sought. To get a match on price means that the product was priced no higher than the amount the consumer was willing to pay.

When deals were offered, if the consumer became aware of the special price she would check to see if any deal items matched or came close to her perceived need. Standards for matches are somewhat lower in this situation, particularly when the consumer is not shopping for a special occasion.

When a product cannot be found that matches perceived needs, the consumer looks for a second best, something that is close to what is desired.

III. ECONOMIC THEORY OF CONSUMER CHOICE

In general, it was the economist that first studied consumer choice behavior, the earliest systematic treatment usually being attributed to Gossen (1854). The full empirical conclusion of the utility maximization model was due to Slutsky (1915).

In the classical economic model there is assumed to exist a single valued ordinal utility indicator defined over all possible commodity bundles, these bundles each containing a finite set of infinitely divisible goods.

The earlier view of utility was in terms of a cardinal measure. To Edgeworth (1881) utility was "as real as his morning jam" (Samuelson 1947). Gossen assumed utility to be represented by a sum of quadratic expressions. Even before the turn of the century Fisher (1892) and Pareto (1896) were questioning the cardinality assumption both in terms of realism and the necessity for deriving conclusions. Most subsequent work has been in terms of ordinal measures although recently cardinality has again come into vogue. Each consumer is presumed in the classical model to maximize an ordinal utility indicator subject to a budget restriction and the requirement that commodity bundles must involve non-negative quantities of each good. Usually nonsatiety in the neighborhood of the budget line is also assumed.

Choice Sets

In our model, interest concerns a particular purchase. Thus quantity purchased of each good within the product class will be zero or one and the sum of all goods purchased within the class will also be at most one. This implies the non-divisibility of goods. For a product such as cake mix where each purchase is minuscule part of the total budget and under the above assumptions, the imposition of a budget constraint seems unnecessary. In most cases consumers will be able to find "exactly" what they desire. In this case income effects will be absent in the model and only a limited form of substitution effect will be exhibited. Consumers will still attempt to minimize the cost of obtaining a product which contains the optimal combination of desired characteristics. Strong substitution may occur in situations involving price competition through deals. Here consumers will consider trade-offs between desired characteristics and price reductions. This substitution effect will be limited to those consumers and related occasions where quality desires do not override price considerations.

Preference Ordering

The operational specification of the preference ordering need only be given in the neighborhood of the satiety point because of the lack of budget constraint and the availability of goods containing optimal or near optimal configuration of characteristics. The preference ordering in this neighborhood is specified as a lexicographic ordering in characteristics. (The lexicographics nature of this order is discussed in Section IV below.)

In terms of the treatment of the components of preference, our model most closely follows that of Lancaster (1966). In his model goods are presumed to be desired for the attributes they contain. His model involves three key parts: (1) a utility indicator defined in terms of these attributes or characteristics;

(2) a transformation between goods and characteristic space (called the consumption technology) which describes the attribute configurations forthcoming from the consumption of various bundles, and (3) a budget restriction defined in goods space. Consumption choices are then viewed as an indirect process in which attributes or characteristics are obtained through the consumption of goods. Consumers can be thought of as going through a multi-stage process in which they first attempt to determine which goods are efficient providers of desired attribute configurations. That is, consumers determine the cost minimizing source of various attribute bundles. Next a choice is made among these various efficient bundles by a process that is similar to that in the classical economic model. That is, at this stage, conventional income and substitution effects prevail.

The process of matching perceived need and perception of products is analogous to the economic model described above. Perceived Need (Utility) is defined in terms of attributes (quality, conveniencies, etc.) while Perception of Products (the consumption technology) describes the attributes forthcoming from various products. Efficiency effects operate through the consumers' unwillingness to pay prices out of line with desired quality levels and the minimization of costs over the set of acceptable products.

The above economic theory provides an overview within which the decision process in the simulation model can be examined. Because it is so general in nature we shall look next to the behavioral sciences for more detailed ties.

IV. BEHAVIORAL CONCEPTS

Our microanalytic simulation model employs an interesting variation of the consumer decision-process model, called situation reaction model. (Sandell, 1968, 1969). Generally, most of the decision-process models in consumer behavior can be classified as "object-dependent," that is, action is conditional on the object(s) encountered but not on the situation. In this model, the object does not instigate the choice, rather the choice is elicited by the situation (Sandell, 1968). The underlying rationale of this model is that the situation is responsible for much of the variation in the choice behavior for cake mixes and that it is responsible for interaction with objects, viz., product-class, and the type of product chosen.

The model starts with consumers who are aware of the new product type as well as the existing types, and who buy cake mixes for specific occasions, which are categorized. Therefore, the model does not incorporate any elements with respect to the process of awareness and particularly media exposures that would seem to be relevant. It starts with the given situation and awareness of the product type alternatives, including the new products.

The situational factors interacting with the need (perceived need) of the individual and the perceived characteristics (attributes) of the alternatives are presumed to influence the choice process of the individual consumer in the simulated state. Accordingly, the simulated consumer is assigned to one of the three categories of serving occasions (situations). Thus the perceived need element in the consumer decision-process is primarily dependent on the serving occasion.

A second-distinguishing feature of this model is that it employs only the relevant psychological characteristics of the individual consumer, given the type of situation (serving occasion for the cake) that has already been established. Unlike other microanalytic models (e.g., see Amstutz, 1967) which have employed demographic, socioeconomic, behavioral and psychological characteristics to differentiate among consumers in the model structure, our model employs only a limited set of psychological variables.

The concept of perceived need is used to suggest that need varies with serving occasion (situation), and this changes as new inputs of information becomes available to the consumer. The construct is also used in the conventional sense that the need indicates to the simulated consumer that a condition exists which requires relief (action). In this sense, it has both the energizing and directive aspects, and that the simulated consumer will strive towards matching this need that will be satisfied by the goal objects (type of cake mixes).

The perceived need is operationalized by the use of four critical attributes (or concepts a la Hansen, 1972 ) which connect the perceived need and the perception of alternatives along these four dimensions. These four salient concepts are: quality, price (perception of price--not actual) convenience and taste. Instead of evaluating these salient concepts positively or negatively, each salient concept is categorized into two or three levels. Hence the cognitive structure of the simulated consumer can be described in terms of the particular levels on these four salient concepts.

Four different types of cake mixes viz., Regular, Premium, Private (Price Brands), Add Butter, comprised the alternatives. The simulated consumer expresses his (her) perceptions of these four alternatives on the same four salient concepts, but with appropriate level on each concept. Thus the construct of perceived need expresses the particular level(s) of the salient concepts, contingent upon the serving situation. The satisfaction of this need is related to the simulated consumer's perception of the alternatives. The dimensions along which the simulated consumer perceives these four' alternatives (types) can be classified into three categories, viz., price, functional aspects, and social aspects (Hansen, 1972, p. 362). Price image is reflected in the three levels of price perception provided for it and similarly the functional image reflecting the performance dimension relates to the quality and taste concepts. Finally, the social image is the consumer's perception regarding the degree of involvement afforded in the use of the product type.

The choice of which particular type of product is chosen is dependent upon the matching of the perceived need (with the four dimensions) that will be satisfied by one of the four product types. If the perceived need as expressed in the four salient concepts (dimensions) matches with the perception of the product-type on the same four dimensions, then the prechoice cognitive structure can be said to be balanced and less conflict will be aroused (Hansen, 1972). If the pre-choice structure is less balanced, change in the salient cognitive structure takes place.

The changing of the cognitive structure is presumed to follow certain orderly pattern, based on the criteria an individual consumer might apply. As stated by Wright (1973), the "individual is essentially asking himself (herself) what processing strategy will extract me from this odious state of multi-alternative conflict quickly, effortlessly, and as surely as possible?". He points out that two criteria are involved, viz., cognitive exertion and likelihood of the strategy to discriminate a choice. For this our computer simulation model uses a lexicographic processing model. Generally, lexicographic processing models ran be explained as follows:

Assumes the individual enters the situation with an ordering of the relevant dimensions cognitively established, and compares options only in terms of their status on the most important dimension. If one is noticeably better on this dimension, he chooses it. If he cannot discriminate a choice on that dimension alone, he attends to cues on the dimension weighted second in importance, repeats the comparison, and so forth until the choice is indicated (Wright, 1973, p. 9).

While our model uses a lexicographic processing, it does not compare directly the alternatives only in terms of their status on the most important dimension. The simulated consumer does not put weights, adds, or multiplies the four salient attributes. She (he) looks for exact match on all the four attributes. If there is a perfect (exact) match, then no further processing is necessary. In the event of any mismatch, she (he) uses the occasion as the preconditioner to determine which attribute is least important. Then she changes the level of that attribute in a prescribed way, viz., change to the second best level on that component. Next, she checks to see whether an exact match is obtained.

The processing may be illustrated in the following way: Suppose our simulated consumer is shopping for something to be put in the lunch box, and if she cannot find what she wants, she will be willing to give up quality but she won't give up convenience nor will she give up low price. There should first be a match on the important dimensions, which are contingent upon serving situations. Then and only then cognitive restructuring in terms of the least important dimension for that situation (occasion) will occur in the model.

The process resembles to a great extent "aspiration-level formulations." As in these formulations, the simulated consumer holds rather stable expectations as to what he (she) will accept, given the situation. When one or more of the critical attributes of the product alternatives deviates from expectation, the simulated consumer decides whether to reevaluate the importance of that attribute, so as to match with the perceived need or not compromise at all.

In looking at this from Hansen's (1972) perspective, the attractiveness of an alternative depends upon its relationship to the salient concepts. Where there is not a "match," cognitive restructuring occurs in such a way as to change the attractiveness of the most attractive alternative such that a more balanced structure results and consequently less conflict ensues.

The perceived need and the required actions, viz., to make a choice among the four product types, is conflict producing and such conflict is reduced by the choice of one of the four product-types. But in this model, before a perceived choice alternative is chosen, a conflict producing situation emerges due to an environmental factor, viz., the availability of the deal. The simulated consumer is either exposed to (seen) or not exposed (not seen) to this "deal." If the simulated consumer is exposed to this "deal" this aspect of the conflict is resolved first before the conflict relevant to normal choice alternatives is attended to. In the absence of any "deals," the choice is indicated directly by the "match" between the perceived need and the perceived image of the alternative.

V. EPILOGUE

Model Forecasts and Actual Outcomes [Forecasted and actual share data has been rescaled to protect proprietary information. A linear transformation was used in rescaling.]

The simulation model described in Section II was used to forecast share configurations under three Product line alternatives. The forecasts were:

Strategy 1: Only Premium Cake in Line

Results: Premium Cake Share 3.4

Add Butter Share (combined for two competitors) 5.6

Strategy 2: Keep Premium Cake plus Introduce new Add Butter Product

Results: Premium Cake Share 3.3

New Add Butter Share 2.1

Share change compared to strategy (1)

Own Brand Regular - .8

Other Brands Add Butter -1.5

Strategy 3: Replace Premium Cake with Add Butter

Results: New Add Butter Share 2.3

Share change compared to strategy (1)

Premium Cake -3.4

Other Brands Regular +1.0

The managerial interpretation was that although the forecasted premium cake share was not large, it was larger than the share would be for a third brand of Add Butter cake. This coupled with indications of strong substitution effect between Regular and Add Butter lead to the decision to keep Premium and not to introduce Add Butter.

Eventual market outcomes were:

Premium Cake reached a peak share of 3.5 within the first year in national markets.

Add Butter peaked at a share of 5.1 (2 brands) within 18 months of national introduction.

Concluding Comments

1. The micro analytic simulation discussed here differs in scale from much of the work undertaken in this area. Prototype examples of the more ambitious approach is the work by Claycamp and Amstutz (1968) simulating the prescription writing behavior of physicians and extensions to other areas by Amstutz (1967). Contrary to this large scale approach which attempts to produce models of great completeness and generality, we have discussed here a model built to aid in making one decision. It is admittedly less general, leaving out many aspects of the decision process of the simulated consumer. To its credit is its reasonable predictive accuracy (to the extent that this can be tested) and efficiency in the use of model building resources. Finally, we note that the model was actually used in a managerial decision-making process. We leave for further debate the scale issue. That is, are we better advised as model builders and behavioral scientists to now turn our attention to limited objective, small scale modeling, reintroducing "black boxes" for all those aspects that are not of primary interest and leave large scale, "global" modeling for a day when our level of knowledge will support-such undertakings.

2. In the above descriptions of a micro behavioral model and various theories and concepts from economic and the behavioral sciences no causal ties can be implied. All that can be claimed at this time is overlap. The simulation model was not built "to operationalize some concept"; it was built to solve a problem. In an ex post sense, other concepts or theories could have been used to "explain" the comPonents and processing mechanisms.

If theory is to have relevance for applied work (including applied model building), it must either indicate variables to include or exclude in the analysis or suggest the form of structural ties. If applied work is to aid theory, it must provide a mechanism for testing the implications of theory. We believe that the class of models discussed here--viz., micro analytic simulation--will turn out to be useful in these processes. We again leave for future research and discussion the obvious question: HOW?

REFERENCES

Amstutz, Arnold E. Computer Simulation of Competitive Market Response. Boston, Massachusetts: The M.I.T. Press, 1967.

Claycamp, Henry J. & Amstutz, Arnold E. Simulation Techniques in the Analysis of Marketing Strategy, in Bass, King and Pessemier (Eds) Applications of the Sciences in Marketing Management. New York: John Wiley and Sons, 1968.

Fisher, I. Mathematical Investigations in the Theory of Value and Prices. Transactions of Connecticut Academy of Arts and Sciences, 1892, 9(1), 1-124.

Gossen, H. H. Entwicklung der Gesetze des menschlichen Verkehrs, und der daraus fliessenden Regeln fur menschlic es Handeln. Berlin: Prager, 1854.

Hansen, Flemming. Consumer Choice Behavior. New York: The Free Press, 1972.

Lancaster, K. J. A New Approach to Consumer Theory. Journal of Political Economy, 1966, 74(2), 132-157.

Pareto, V. Cours d'economie politique. Lausanne: F. Gouge, 1896-97.

Samuelson, P. A. Foundations of Economic Analysis. Harvard University Press, 1957.

Sandell, Rolf G. Effectives of Attitudinal and Situational Factors on Reported Choice Behavior. Journal of Marketing Research, November 1968, V(4), 405-408.

Sandell, R. G. Att Forsta Konsumenten. Stockholm, Sweden: F. A. Norstedt and Soners forlag, 1969.

Slutsky, E. Sulla Teoria del Bilancio del Consomatore. Giornale degli Economisti, 1915, 51, 1-26.

Wright, Peter L. The Simplifying Consumer: Perspectives on Information Processing Strategies. Paper presented at American Marketing Association Doctoral Consortium, Michigan State University, East Lansing, Michigan, August 1973.

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