Dr. Michael R. Hyman
New Mexico State University

(Last edited on 27 December 1999)

Models in General

• "The goals of marketing research are (1) to expand the body of marketing knowledge, and (2) to help marketing managers make more accurate and more cost­effective decisions. Thus marketing academics should present their research findings so that they are understandable to marketing managers." Agree or disagree, and why? (96)
  
  
• Marketers create marketing models for many reasons. What are the different reasons? Do problems arise when managers apply models in ways inconsistent with their raison d'etre? Discuss. (96)
  
  
• What are the qualities of a good marketing model? Give an example of such a model, and explain what "makes it good." (96)
  
  
• The formulation of homogeneous groups that differ significantly from one another and the reduction of data are the two basic goals of all quantitative analyses of marketing data. Comment. (Hint: Your answer should refer to multivariate statistical methods.) (96)
  
  
• "Models of consumer behavior are limited because consumer behavior theorists disregard the random element of consumer behavior. In part, theorists ignore this element because they believe that all human behavior is causal; if we could measure and include all of the relevant variables, then we could explain the behaviors of consumers perfectly (cf. Bass, "Analytical Approaches in the Study of Purchase Behavior and Brand Choice," 1977)." Agree or disagree, and why? (96)
  
  
• In Repeat Buying: Theory and Applications (1972), Ehrenberg says:

Traditionally, the task of understanding consumer behavior seems to have been approached along two parallel lines, the "practical" one of collecting facts, and the "academic" one putting forward theories. These two activities have seldom met. Thus various theories of consumer behavior have been developed with little recourse to organized facts... and a great variety of facts have been collected with little if any conceptualization (p.217)

He then suggests the following technology for examining buying behavior:

. . . first of all uncover regular patterns in empirical data, then synthesizing "theory", by first modeling, and then inter­relating, these regularities. Such work typically starts by examining some factual data. The data must be judged to be "relevant"...but apart from this there need be no very explicit theoretical concepts or any priori framework of hypotheses. The immediate results of such an examination of data are then a variety of separate relationships, each describing some more or less isolated aspect of the system. This is the lowest level of model­building, and there are two main criteria for assessing success. One is the degree of generalization achieved: has the empirical regularity been shown to hold under a wide enough range of conditions to make it worth exploring further? The second criterion concerns the simplicity of each result: is it simple enough to facilitate its further use? (pp.217­218)

Compare Ehrenberg's model and theory building orientation to those of other marketing modelers, especially researchers who study consumer behavior. (96)

• Naert and Leeflang (Building Implementable Marketing Models, 1978) write that a manager only accepts a model that reflects his/her view of the market and the way that it works. Obviously, each manager views the market and the way that it works differently. To what degree, then, is a model useful for generalizing and formulating the "Laws of Marketing."

• "Because decision­calculus­type (i.e., judgment­based) models heavily rely upon the subjective estimates/judgments of experts, the output of such models is essentially useless as an aid in managerial decision making." Agree or disagree, and why? (96)
  
  
• What are decision­calculus models? How are they developed and implemented? Under what conditions are they the most useful? How do they differ from empirically­based marketing models? (96)
  
  
• How should we investigate the effectiveness of judgment­based models (i.e., experimental lab versus real world)? (96)
  
  
• How might the increase in available information (i.e., the current information explosion) affect the selection of subjective versus empirical decision models in marketing? (96)
  
  
• Although decision­calculus models are deterministic, judgment is used to assign values to some of the variables in such models. Yet, the uncertainty associated with these assignments is rarely incorporated into such models. What, then, is the argument in favor of incorporating Baysian principles into these models?
  
  
• "The values assigned to some of the variables in a decision­calculus model simply reflect cognitive time series or cognitive cross­sectional analyses by managers; thus, such subjective estimates are less valid than objective estimates." Comment. (96)

• What steps can modelers take to insure the validity and reliability of their decision­calculus models?

• Discuss the literature on the implementation of marketing models. Discuss those aspects of the model and the model building approach that make for successful implementation.
  
  
• The evolutionary­model­building approach implies that modelers should create custom­built models (i.e., models should match the integrative complexity of users) (cf. Parsons and Schultz (1976), Marketing Models and Econometric Research). What should modelers do if users lack the integrative complexity required to understand a complex but complete model?
  
  
• (a) What are the advantages and disadvantages of increasing the complexity of a model? (b) Can simple models be complete and visa-versa? Explain. (96)
  
  
• Compare and contrast MDSS to expert systems. For what types of marketing problems is one preferred to the other? (96)
  
  
• "Expert systems are mere compilations of managerial decision rules in the form of computer programs. Furthermore, knowledge engineers (i.e., the developers of expert systems) cannot possibly extract the complete sets of decision rules used by decision makers. Thus, expert systems --at best--can speed decision making; at worst, expert systems grossly misrepresent the decision process." Agree or disagree, and why? (96)
  
  
• "We should accept or reject an econometric model in marketing solely on the fit of the model to the data which is being modeled." Agree or disagree, and why? (96)

• What is the difference between implementation and successful implementation?
  
  
• Model users and model builders have different viewpoints. How does this difference affect theory building in marketing?
  
  
• What quantitative and qualitative criteria indicate the success or failure of a Marketing Decision Support System (MDSS)?

• In what ways do models used for control differ from models used for forecasting?
  
  
• Should we concern ourselves with a model's robustness if it yields accurate answers in general and inaccurate answers only for extreme values, ones which are unlikely to occur in real use (cf. Naert and Leeflang, Building Implementable Marketing Models, 1978)?
  
  
• "Because the tremendous oversimplification of reality inherent to marketing models precludes their isomorphic relation to reality, marketing models should only be used for predictive purposes, and never for descriptive or prescriptive purposes." Agree or disagree, and why? (96)
  
  
• "If a model forecasts accurately, then managers should use it for all their relevant decisions." Comment.
  
  
• Briefly review the literature on increasing forecasting accuracy by combining multiple forecasts. (96)
  
  
• "Managers should only rely on statistically-based forecasting methods. Human judgment is irrelevant to making accurate forecasts." Agree or disagree, and why? (96)
  
  
• Bootstrapping codifies expert judgment into an expert system through if/then rules inferred from the decisions and actions of experts. How well does bootstrapping convert expertise into if/then statements? (Provide empirical evidence for your assessment.) Is bootstrapping limited because it must ignore intuitions that are incompatible with the related expert system? (96)

• Will including additional descriptive variables reduce the "noise" in segmentation models, i.e., can we increase R² substantially if we add enough variables? Discuss. (96)
  
  
• "Bass, Tigert and Lonsdale (Journal of Marketing Research, 1967) suggest that despite the low R²s, socioeconomic and demographic variables can effectively segment consumers into subgroups when the mean differences between subgroups is large. However, low R²s show that the segments are heterogeneous, thus failing the homogeneity condition of successful market segmentation." Is the above statement important from a managerial perspective? Explain. Is the above statement important from a theoretical perspective? Explain. (96)
  
  
• What are the major questions relevant to the proper problem definition, research design, data analysis, interpretation and implementation of a market segmentation study? (96)
  
  
• "If marketing segmentation is product­market specific, then we cannot develop marketing theory and/or generalizations about market segmentation." Agree or disagree, and why?
  
  
• Marketers use different statistical tools to segment markets: cross­classification, automatic interaction detection (AID), regression, cluster analysis, and discriminant analysis. Briefly discuss how marketers use each of these tools to segment markets, and under what circumstances each tool is the most effective?
  
  
• (a) "Most segmentation studies have been conducted for consumer goods. Yet the concept of segmentation and most of the segmentation research approaches are equally applicable to industrial market situations" (Webster and Wind 1972). Agree or disagree, and why? (b) "Segmenting the market for a service is similar to segmenting the market for a product." Agree or disagree, and why? (96)

• Consumers with similar demographic/socioeconomic/attitudinal profiles may have vastly different product attribute preferences, and visa versa. What are the implications for managers and marketing theoreticians?
  
  
• Massy and Frank (Journal of Marketing Research, 1965) proposed using price/deal sensitivity to segment markets. How useful is this approach? Furthermore, as price sensitivity is related to socioeconomic variables, how different is this approach from traditional segmentation approaches?
  
  
• Discuss the literature on market segmentation. Include in your discussion the R² controversy and its resolution. (Hint: Discuss why Ronald Frank and others (e.g., Frank, et al., Journal of Marketing Research, 1967) failed to find a statistically significant relationship between socioeconomic variables and managerially/theoretically­salient market segments.)
  
  
• How do the behaviorally­oriented approaches to market segmentation differ from the decision­oriented (normative) approaches (see Mahajan and Jain, Journal of Marketing Research, 1978)? Under a decision­oriented approach, should resource constraints and segmentation occur simultaneously? (Hint: Explain with the help of a normative segmentation model.

• What is wrong with the classical economic model of pricing? Discuss a more appropriate framework for determining pricing decisions. Be sure to discuss the relevant concepts (e.g., What are experience curves and how can they be used in pricing strategy? How do myopically optimal pricing policies differ from globally optimal pricing policies?) (96)
  
  
• (a) Define price elasticity. (b) What are the different types of price elasticity? Under what contexts do the different types of price elasticity apply? (c) Price elasticities evolve over the product life cycle. Discuss these shifts and their impact on long­term pricing strategies. (96)
  
  
• What factors should managers consider when setting the price of a durable good? What additional or different factors should managers consider when setting the price of a non-durable good?
  
  
• Discuss the pricing of pure components (each component is priced and offered separately), pure bundles (all components are priced and offered as a bundle), and mixed bundles (both the bundle and the components are priced and offered separately). What model(s) could you use to select a bundling pricing strategy and to access its efficacy? (96)
  
  
• Compare and contrast rational/economics-based conceptualizations of price to psychological/value-based conceptualizations of price. (96)
  
  
• The price-quality relationship is a classic marketing notion. Briefly discuss past research on this relationship. In addition, discuss possibilities for and impediments to future research. (96)
  
  
• "Researchers in fact have been exasperated in determining optimal product price by the fact that prices frequently are interdependent" (Urban 1969). Comment. (96)

• Briefly review the marketing literature about the relation between quality and price, starting with McConnell (1968) (see Journal of Marketing Research).

• What are the underlying assumptions inherent to many of the new product growth models for durable goods? Does the violation of these assumptions "in reality" significantly inhibit the performance of these models? How have marketing mix variables been incorporated into recent incarnations of these models? (96)
  
  
• Trace the history of new product growth models, starting with Bass' (1969) classical article in Management Science. (Hint: Begin with a discussion of the underlying assumptions of Bass' model.) Are the revised models really an improvement over Bass' original version?
  
  
• Is Bass' NPGM applicable to industrial products (see Tigert and Farivar, Journal of Marketing, 1981)? To services? In non-Western countries? Discuss. (96)
  
  
• Can Bass' NPGM (and its extensions) help managers to make better decisions about product capacity? About whether or not to launch a new to the world product? (96)
  
  
• Discuss the efforts of marketers to internationalize diffusion models. (96)
  
  
• At the end of their review article on NPGM (Journal of Marketing Research, 1990), Mahajan, Muller, and Bass list roughly ten promising areas for future research. Take any four of these research areas and (1) discuss the relevant literature, and (2) explain why these three areas seem the most promising. (96)

• Managers often lack sufficient sales data to estimate their diffusion models in a timely manner. Are judgment­based parameter estimates for diffusion models a viable alternative (see Hyman, Journal of Business Research, 1988)?

• Marketers have developed several multiattribute approaches for evaluating product concepts. Assess the strengths and limitations of any two such approaches.
  
  
• Compare and contrast pre­test­market forecasting models with test­market forecasting models. Under what circumstances is either type a better choice? (96)
  
  
• Compare and contrast the following new product models based on pre­test market data: N.W. Ayer (Claycamp and Liddy, Journal of Marketing Research, 1969); ASSESSOR (Silk and Urban, Journal of Marketing Research, 1978; Urban and Katz, Journal of Marketing Research, 1983); and TRACKER (Blattberg and Golanty, Journal of Marketing Research, 1978).
  
  
• "There is a general logic to the structures and assumptions of new product models based on test/pre-test market data." Agree or disagree, and why? (96)
  
  
• Hauser and Shugan's DEFENDER model (see Marketing Science, 1983) deals with defensive strategies in response to new competition. Describe the assumptions and limitations underlying their model.

• Discuss the literature concerning the long­term effect of advertising upon sales. In particular, compare the ways in which the long­term effects have been modeled (i.e., model specification) and discuss the "data interval bias." (96)
  
  
• How can simultaneous equation regression model be used to solve the problem of identification in advertising? Discuss "testing" versus "estimation" by referring to Bass' (Journal of Marketing Research, 1969) simultaneous equation study and its critique by Dominguez and Page (Journal of Marketing Research, 1971).
  
  
• Two types of advertising models, the distributed lag model and the current effect model, have seemingly similar structures yet different underlying assumptions. That is, the former claims that previous sales result from the cumulative effect of advertising, whereas the latter claims that previous sales result from all marketing and exogenous variables. Which assumption is more plausible, and why?
  
  
• Additive response models of advertising are simple and easy to interpret. Explain why multiplicative models may be more useful.
  
  
• "Advertising seems to cause sales, and sales seem to cause advertising; thus we cannot measure advertising effectiveness." Agree or disagree, and why? (96)
  
  
• "Even for a correctly specified model with appropriately estimated parameters, qualitative features of market environments might affect the advertising-sales relationship" (Assmus, Farley, and Lehmann, Journal of Marketing Research, 1984). Discuss. (96)
  
  
• Some advertising researchers argue that companies overadvertise their mature products. Discuss the relevant literature and the reasons why companies overadvertise. (96)

• The studies of the advertising-sales relationship have used different measures of the dependent variable, such as market share and sales volume. Are these the best measures under all circumstances? Are other measures, such as aided or unaided recall, more appropriate under some circumstances?

• A key issue in the evolution of stochastic brand choice models involves their underlying assumptions. Discuss the major assumptions of specific models, highlighting the evolution of these assumptions. Based on your discussion, assess the reasonableness of these assumptions. (96)
  
  
• One might argue that the debate between marketers with a deterministic worldview and marketers with a stochastic worldview is of a purely academic nature, with the resolution unforeseeable and the end result trivial. Comment. (96)
  
  
• In the "Theory of Stochastic Preference and Brand Choice" (see Journal of Marketing Research, 1974), Bass produces strong evidence that brand choice behavior is stochastic. This evidence suggests that studying the psychology of brand choice is useless. Do you agree or disagree, and why? (96)
  
  
• Compare and contrast purchase incidence models to brand choice models. (96)
  
  
• Describe the Hendry model. Specifically, how does the Hendry model link consumer preferences, competitive market structures, and marketing strategies.
  
  
• Why is the assumption of stationary transition probabilities crucial to Markov brand­switching theory? Discuss the problem of stationarity relative to (a) the role of stationarity as a theoretical assumption, and (b) the empirical evidence.
  
  
• The concept of brand loyalty is frequently references in the marketing literature. (a) Describe and discuss the various operational definitions of brand loyalty. (b) Discuss the measurement problems associated with brand loyalty. (c) What are the managerial implications of the brand loyalty concept given your discussion of (a) and (b). (96)
  
  
• In his review article entitled "Analytical Approaches in the Study of Purchase Behavior and Brand Choice," Bass (1977) wrote:

The stochastic modeling school and the behavioral science-oriented attitude-theory and psychometric schools have only rarely acknowledged the existence and potential contribution of the other tradition in the pursuit of knowledge of consumer brand-choice behavior. Stochastic models cannot, by themselves, explain why brand-choice probabilities vary from one consumer to another. On the other hand, the deterministic orientation of the behavioral sciences is such that attitude and psychometric studies have often assumed that it is sufficient to explain or predict the references of consumers.

Comment. (96)

• Evaluate the following claim: Kuehn's learning model (see Journal of Advertising Research, 1962) is unrealistic because the influence of marketing mix variables is ignored when we assume that the probability of purchasing a brand is a function of previous purchases.
  
  
• The Hendry model is based on the notion of entropy? What is entropy? What are the assumptions, benefits, disadvantages, and results of a stochastic model based on entropy?
  
  
• Discuss the limitations in using the "distribution of runs approach" (see Frank, "Brand Choice as a Probability Process, Journal of Business, 1962) to explain the extent that an observed pattern of brand choices over time support the hypothesis that the probability of purchasing a certain brand is constant during the period of study.
  
  
• Describe the different zero­order models of brand choice. Do the purchase incidence models and brand switching models stem from the same fundamental assumptions about consumer choice? Explain. How can decision variables may be added to stochastic choice models?

• Three basic notions about the relationship of market share to firm performance pervade the marketing literature: (a) high market share leads to high profitability, (b) more efficiency leads to lower costs and higher market share, and (c) market share and profitability are related to intervening variables, such as management quality and luck. Discuss the evidence for each notion. What does each notion imply regarding marketing strategy? (96)
  
  
• Compare and contrast attraction-type models, linear additive models, and multiplicative models of market share. What are the advantages and disadvantages of each type of model? (96)

• Discuss (a) the different methods for partitioning markets, (b) the different types of market structures, (c) the problems and benefits of partitioning models, and (d) the managerial implications derived from partitioning models. (96)

• Compare and contrast Gary Lilien's ADVISOR2 model (Management Science, 1979, 1980) with Farris and Buzzell's analysis of PIMS data (see Management Science, 1980) in terms of results, products selected, and modeling philosophy.
  
  
• How can marketing models be made more accessible to the financial specialists who also influence investment decisions?