Different Influences on Consumer Decision Making

ABSTRACT - Three papers dealing with diverse aspects of consumer behavior are discussed. While a few methodological problems are apparent, the more global concerns of lack of control/ comparison groups, uncontrolled anchoring, and labeling in consumer research are addressed.

Citation:

E. H. Bonfield (1981) ,"Different Influences on Consumer Decision Making", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 748-749.

Advances in Consumer Research Volume 8, 1981      Pages 748-749

DIFFERENT INFLUENCES ON CONSUMER DECISION MAKING

E. H. Bonfield, Temple University

ABSTRACT -

Three papers dealing with diverse aspects of consumer behavior are discussed. While a few methodological problems are apparent, the more global concerns of lack of control/ comparison groups, uncontrolled anchoring, and labeling in consumer research are addressed.

INTRODUCTION

The three papers presented for this discussion are diverse in subject matter. Each makes a contribution, and each leads to further questions for research. Two of the papers, those by Roberts, Wortzel, and Berkeley and by Burns and DeVere deal directly with family decision making. They contain primary research data. Meadow, Cosmas, and Plotkin present a literature review suggesting consumers described as elderly are different from other consumers.

If all three papers dealt with family consumer decision making, they would fit into the general model of intra-family influence Burns and DeVere briefly describe at the beginning of their paper. Burns and DeVere have explored two situational dimensions, while Roberts, Wortzel, and Berkeley could be categorized as exploring a mothers-children relational dimension. While Meadow, Cosmas, and Plotkin correctly do not place their research in the family decision process domain, the individuals in the group they study are frequently actors in family decision making.

The discussion of each paper which follows contains little concern for the shortcomings found with the specific research projects. The exploratory nature of all three studies leads instead to a concern for suggestions for improving future research based on the present experiences.

MOTHERS' PERCEPTIONS

The Roberts, et al research is a positive case of "massaging data." The data from a current, large-sample, commercial study was apparently made available to the authors for additional analysis. This is enlightened behavior on the part of Needham, Harper, and Steers, and is to be commended.

Additional analysis of data gathered for other purposes is rarely as suited for testing additional hypotheses as studies designed for that specific purpose. That situation exists for the present research. Therefore. as the authors point out, their analysis is exploratory, and any findings must be taken as tentative.

"Attitudes"

Because of the exploratory nature of the study, the sets of scales labeled "attitude" are useful as indicators of such constructs as "concern about nutrition" and "cosmopolitaness." It should be clear, however, there is no basis, other than anecdotal, that these scales are valid measures of any attitude. Furthermore, the reliability measures of the "attitudinal" constructs suggest consistency in the way people mark scales expected to measure the construct. If there is only face validity of a construct, however, the analyst must question the reasons for these high interscale correlations. For example, are "concern about sex" (typical among ACR members) and "concern about violence" the same thing? Do they co-vary because of the admitted "middle America" bias of the sample, because all respondents were women, because they were women with children,...? Is it possible the scales are marked the same, but the depth of concern--the involvement--is different? Some of these alternative hypotheses might be explored within the context of the present data set, while others would have to be tested with new research.

Labeling

After using factor analysis on the scales measuring perception of children's influence on brand choice, the author's labeled each factor despite the fact that only one, "gum," readily lends itself to such a label. Labeling can be a dangerous task in factor analysis, as well as in other areas, because it can be misleading. Are cheese and pasta children's foods? Are cold cereal and restaurants sweets or snacks?

A further problem stems from the sample used in factor analysis. If the authors used listwise deletion of cases as an option in factor analysis, only those respondents indicating use of all 35 products by marking the perceived influence scales could be included in the factor analysis. Considering pets alone, it is questionable whether a large proportion of the households owned, and therefore probably bought, both dog and cat food. The scales used forced respondents who did not use a product into a "least frequency" category. While there is legitimacy in utilizing this type of scale, inclusion of nonusers with users clouds interpretation.

Controlling for Children's Ages

Controlling for children's ages is difficult as the authors amply demonstrate. The problem is easily handled in one-child (and zero-child) households. Where more than one child is present, for any one product category, the mother might anchor on one child or somehow cognitively average her perception of the influence of her children on brand choice. An answer for this problem would be to instruct the mother to anchor on a specific child whenever more than one child is present. A randomization process for child selection could easily be developed to assure representativeness of the total sample, as well as adequate sample size for each major category of family composition.

SITUATION DIMENSION EFFECTS

Burns and DeVere have developed a "general model of intra-family influence" which was used to guide their selection of the study presented here. They utilized written scenarios as stimuli differentiating four treatment levels. While it can be argued the written scenarios would not be realistic to the couples in each experimental group, their approach is a viable compromise to more expensive and more difficult to physically control alternatives. A useful alternative the researchers might have considered as a manipulation check would have been to add a control situation. This addition would have added some 40 additional couples to the total sample, but, given the sampling frame and data collection methods, the increased time and dollar cost would have been small. No scenario would have been provided the control group.

There is no quarrel with the data analysis as given. The use of figures, particularly when interaction effects are involved, would make the results both easier for the researchers to interpret and for readers to follow.

Overstated Agreement

The use of means for husbands and wives are likely to increase the apparent level of agreement on the dependent variables. Since data on both spouses in the household is available, it is possible to compare responses on a house-hold-by-household basis. Davis and Rigaux (1974) and Bonfield (1978) have found considerably greater disagreement when spouses' responses are compared on nominal scales similar to the disagreement outcome variable used by Burns and DeVere. The constant sum and discussion time estimates would show even greater disagreement unless some reasonable nominal scale were superimposed. A "scientific" approach might be to utilized standard deviation estimates to create nominal boundaries. An ad hoc approach of trying 10 to 20 percent differences in constant sum estimates and five-minute differences in discussion times might be even more useful.

Uncontrolled Anchoring

A problem which appears common to most, if not all, family decision process research can be labeled, "uncontrolled anchoring." That is, there is a lack of specificity as to what stimulus arrays respondents are reacting. For example, Burns and DeVere nicely imply a temporal anchor in their four scenarios. The respondents probably predicted what they thought would happen. In other research, however, it is impossible to determine whether respondents are predicting or reporting perceptions of what has happened in the past. If the latter is the anchor desired by the researcher, respondents should be asked to think about their last purchase occasion for the product.

Controlled anchoring means specifically describing products. Life insurance can be purchased on husbands' lives, wives' lives, children's lives, and in the case of mortgage insurance, jointly on both husbands' and wives' lives. A decision to buy a set of pots and pans is different from buying a replacement pot or pan or an additional pot or pan. Encyclopedias might be bought for household use or for children's use.

Finally, decisions should be more specifically described. While need verification is a relatively clear construct among family decision making researchers, "the need to buy one" may imply need recognition to respondents. "What model or type" may well mean different things to different respondents. Controlled anchoring might lead to the specification, for example, of "whether to buy term or whole life insurance," and "what size pot or pan," on future questionnaires. The model/type question might simply be deleted as inappropriate for encyclopedias.

An important aspect of having a model to guide research is the provision of a framework for exploring questions of differential anchoring. The questions addressed above suggest further exploration in the situational block of the Burns and DeVere model.

"WHAT DO YOU MEAN, 'I'M ELDERLY'"

Meadow, Cosmas, and Plotkin provide a good literature revue on an age category segment of the United States population. Their review is a useful addition to the consumer research literature in that it brings to this group a set of annotated sources in a single place. They add evaluation which cannot be found elsewhere. Two observations about the research and literature reviewed seem appropriate.

Labeling People "Elderly"

A label is a convenient, short-hand code used to indicate a group. When researchers label, the intent is to be able to generalize about whatever has been labeled. An important problem associated with labeling is the potential for labeling to lead to stereotyping. Stereotyping is the process of ascribing an attribute possessed by some members of a group to all members of that group.

Clearly, those people who have been labeled "elderly" are chronologically older than other age groupings of people. Some, even many or most, of the people who have been labeled "elderly" use dentures, are retired, and use less long-term credit. Labeling people "elderly" in the stereotyping sense, may lead to restricted granting of long-term credit--which, in turn, would make the group's limited use of long-term credit a self-fulfilling prophecy.

Some labeling of sub-segments among older age groups may result in mislabeling. For example, Towle and Martin (1976) labeled a single psychographic sub-segment "Information Seeker" and "persuasible" apparently because the psychographic description of them was "kind" and "sincere." Whether one goes backward or forward in the logical process, no one label follows from any of the others.

Perhaps the most negative result of labeling is that some members of the group, in accepting the label, end up with lowered self images and reduced self esteem. For example, some individuals may decide they cannot exercise as much or continue in a job they enjoy and are capable of handling well simply because they have had a birthday.

Control/Comparison Groups

Consumer research, in general, has been characterized by nonuse of control groups. This generalization characterizes the research of those who have studied those in our population who are over 40, 45, or 65 depending on the study. Unless there is an appropriate comparison group, any characterization of a specific group as different is unsupported.

CONCLUSION

It was not the purpose of this paper to evaluate three exploratory studies. The discussion of each paper has lead to general statements about problems with uncontrolled anchoring and labeling in consumer research, as well as a need for control or comparison groups in this research. The specific discussions suggested ways of combating these problems.

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