Methodological Perspecives

Leonard Jon Parsons, Georgia Institute of Technology
[ to cite ]:
Leonard Jon Parsons (1983) ,"Methodological Perspecives", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 605-606.

Advances in Consumer Research Volume 10, 1983      Pages 605-606


Leonard Jon Parsons, Georgia Institute of Technology

This collection of papers had no common theme. However, underlying this potpourri is a message that advances in statistical methodology can add new insights in consumer research, provided that they are fully and properly employed.


This paper explored the relationship between the acceptability of risk for various products, services, and activities and nine postulated components of risk. The analysis-was conducted at the segment level.

The nine components of risk were originally measured on 7-point scale. The authors collapsed these seven into three categories. The proportions of a segment falling into each category were then further processed by a logit scaling method due to Bechtel and Wiley (1982). The authors claim that the advantage of all this is that the dependent and independent variables will possess a common metric. However, their procedure might increase scale error (Hubert and Lehmann 1975). In this particular application, benefits of massaging the variables may not outweigh the costs.

The main analysis consisted of estimating the three segment functions using seemingly unrelated equations regression. This technique has been reviewed in Parsons and Schultz (1976, pp. 71-2) and used previously in marketing in the estimation of sales response functions (Beckwith 1972, Clarke 1973, Houston and Weiss 1974, and Wildt 1974). The equations are seemingly unrelated because although each equation contains only one dependent variable, these dependent variables are disturbance related.

The authors found a major gain in efficiency by estimating the three segment equations together using the seeminglv unrelated equations technique in comparison to estimating each equation individually using ordinary least squares regression. Unfortunately they did not carry the implication of this gain to its logical conclusion. If there is a gain in efficiency, then there must be one or more common factors that are affecting the dependent variables, but that have been omitted from the analysis. Given the extensive list of risk components, what might these factors be? On questioning, the authors indicated their belief was that benefits should be included in future work. Thus, much as perceived quality cannot be understood in abstraction from price, perceived risk cannot be separated from perceived benefits.


This paper explores the relationships among consumer sentiment, purchase intentions, disposable income, and retail sales of appliances at the aggregate level using time series analysis. Despite the advanced methodology, the empirical results are not much different from what has been known for two decades.

Parsons and Schultz (1976, pp. 118-119) point out that a forecast and a prediction are not synonymous. An explanatory marketing model is comprised of theoretical marketing premises and justifiable factual statements of initial conditions. From such a model, a set of prediction statements that attribute definite probabilities to specified observable marketing events can be deduced. Important and pragmatic statements about future occurrences can be made without deducing these statements from initial conditions with the aid of a model. A forecast is an extrapolation from statistical estimates obtained in one historical period to observations generated in another historical period.

The authors of this paper are interested in forecasting consumer spending. Although they have no a priori logical model and, indeed, deliberately ignore significant contemporaneous relationships, they use the terminology of prediction when they should have been using that of forecasting. Among other things, consumer sentiment does not drive disposable income, but rather it is a leading indicator of disposable income. Consumers know about salary changes before they actually take place, are aware of unrealized capital gains and losses, and can usually anticipate layoffs. Moreover, the authors provide no information about how well their model forecasts. All in all, this is disappointing paper by two able methodologists.


In-store shoppers tasted and evaluated two unidentified products from a set of six products comprised of two h brands for each of three different grocery products. The order of the consumption was manipulated. The question is whether or not the shoppers' evaluations of a particular product are influenced by the other product with which it was consumed and the order in which the products were consumed.

The authors begin their analysis by assessing the relationship between demographics and preferred evaluative criteria. Significant differences were found for orange juice and hot dogs, but not potato chips, on age and sex, but not income. The sources of these differences can best be seen by rewriting their Table 1 putting the raw data as proportions.



Generally speaking, taste is not a major contributor to these differences. This was borne out by the MANOCOVA results in which the demographic covariates were not significant.

Their primary analysis was done using MANOCOVA. There were three main effects: type of product, type of accompanying product, and order of tasting. The important effects, however, are the interaction effects. The authors' most important finding is that the type of product consumed in conjunction with another product can effect the ratings of that product. This means managers must be careful about the choice of brands in any tie-in or group promotion involving two or more brands in an event that offers consumers an incentive to purchase each of the participating brands. The same warning applies to cross-ruff couponing in which coupons for one brand are placed in or on the package of another product. Moreover, some care must be exercised in showing the consumption of other products in the advertisements of any brand.

The type of product and the order of tasting interacted to affect product evaluations. The marketing manager would seem to have little control over the order of product consumption. A recommended order could be stressed in marketing communications.

The authors' analysis would have been strengthened if they had followed up their multivariate analysis with the use of simultaneous confidence intervals, rather than incorrectly using univariate F-tests, to identify which response led to rejection of the hypothesis of no interaction effect. Nonetheless, the authors have addressed an important issue for marketing managers. The next step would be to repeat the experiment but with brands identified. How much will the halo from brand image impact the results?


Bechtel, Gordon G. and Wiley, James B. (1982), "Probabilistic Measurement of Attributes: A Logit Analysis by Generalized Least Squares," Discussion Paper 67, Center for Econometrics and Decision Sciences, University of Florida.

Beckwith, Neil E. (1972), "Multivariate Analysis of Sales Responses of Competing Brands to Advertising," Journal of Marketing Research, 9 (May), 168-76.

Clarke, Darral G. (1973), "Sales-Advertising Cross-Elasticities and Advertising Competition," Journal of Marketing Research, 10 (August), 250-61.

Houston, Franklin S., and Doyle L. Weiss (1974), "An Analysis of Competitive Market Behavior," Journal of Marketing Research, 11 (May), 151-5.

Hubert, James, and Donald R. Lehmann (1975), "Assessing the Importance of the Sources of Error in Structured Survey Data," in Control of "Error" in Market Research Data, John U. Farley and John A. Howard, eds., Lexington, Massachusetts: Lexington Books.

Parsons, Leonard J. and Randall L. Schultz (1976), Marketing Models and Econometric Research. New York: North Holland.

Wildt, Albert R. (1974), "Multifirm Analysis of Competitive Decision Variables," Journal of Marketing Research, 11 (February). 50-62.