An Analysis of the Presence, Stability, an Antecedents of Husband and Wife Purchase Decision Making Influence Assessment Agreement and Disagreement

ABSTRACT - Noting the absence of empirical study of husband and wife over- and underestimation of purchase decision making influence assessments, the authors utilize Davis and Rigaux's (1974) and Burns' (1977) conceptualizations. Results confirm presence of a wide variety of disagreement as well as agreement groupings of couples who differ greatly from overall husbands and wives group averages. Some generalized tendencies for stability of agreement and disagreement predispositions across influence dimensions is apparent for one product (family automobile) and financial decisions across two products. Antecedent variable analyses imply differential sensitivity to resource theory with wives overestimating their influence when they make greater contributions and underestimating if they do not. Implications for future research and decision making dynamics are discussed.

Citation:

Alvin C. Burns and Jo Anne Hopper (1986) ,"An Analysis of the Presence, Stability, an Antecedents of Husband and Wife Purchase Decision Making Influence Assessment Agreement and Disagreement", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 175-180.

Advances in Consumer Research Volume 13, 1986      Pages 175-180

AN ANALYSIS OF THE PRESENCE, STABILITY, AN ANTECEDENTS OF HUSBAND AND WIFE PURCHASE DECISION MAKING INFLUENCE ASSESSMENT AGREEMENT AND DISAGREEMENT

Alvin C. Burns, Louisiana State University

Jo Anne Hopper, Southeastern Louisiana University

ABSTRACT -

Noting the absence of empirical study of husband and wife over- and underestimation of purchase decision making influence assessments, the authors utilize Davis and Rigaux's (1974) and Burns' (1977) conceptualizations. Results confirm presence of a wide variety of disagreement as well as agreement groupings of couples who differ greatly from overall husbands and wives group averages. Some generalized tendencies for stability of agreement and disagreement predispositions across influence dimensions is apparent for one product (family automobile) and financial decisions across two products. Antecedent variable analyses imply differential sensitivity to resource theory with wives overestimating their influence when they make greater contributions and underestimating if they do not. Implications for future research and decision making dynamics are discussed.

INTRODUCTION

Researchers working with husbands and wives perceptions of their relative purchase decision making influence have long acknowledged that while influence assessments usually agree in the aggregate, there is often a lack of agreement on the individual couple level. Davis (1976), Burns and Granbois (1979), and Safilios-Rothschild (1969) all point out the need for research which clarifies the nature and reasons for these reporting discrepancies. Thus far, however, scant attention has been given to this problem, and empirical research is almost nonexistent.

In addition to the above authors' comments, some preliminary conceptual thinking has been devoted to the problem. For instance, Heer (1962) and Granbois and Willett (1970) have noted over- and underestimation occurrences. Davis and Rigaux (1974) noted "vanity" and "modesty" in matched spouses and alluded to the need for further research to identify, isolate, and study couples with these perceptual biases. A more detailed model of the comparison of husbands to wives influence assessments has been proposed by Burns (1977) who identified nine different types of outcomes possible in a comparison of husband to wife influence assessments. Both of these conceptualizations serve as frameworks for the research reported in this paper.

OBJECTIVES OF THIS STUDY

The primary purpose of this study was to investigate instances of spousal agreement and disagreement in purchase decision making influence assessments. In particular, the intent was to assess the efficacy of Burns' (1977) and Davis and Rigaux's (1974) notions as paradigms in identifying the presence and persistence of agreement and disagreement. Four specific research questions served as the objectives of the study.

1. Can groupings of couples be identified who are similar in agreement or disagreement yet dissimilar form others and the aggregate?

2. If so, do couples maintain stable agreement and disagreement dispositions across independent influence dimensions?

3. What, if any, are the antecedent variables associated with certain matched influence assessment predispositions?

The rationale for this study stems from corroborated findings in husband and wife influence assessment research (Davis, 1976). Namely, purchase influence distribution between husbands and wives varies as a function of personal attributes, product-specificity, and type of decision. It seems reasonable to expect that disagreements may obey some regularities as well, and if determined empirically, future research will be better equipped to separate out or otherwise adjust for systematic perceptual differences.

FIGURE 1

AGREEMENT AND DISAGREEMENT REGIONS FROM MATCHED HUSBAND AND WIFE RESPONSES

The Davis and Rigaux (1974) and Burns (1977) models are not competing; they simply differ in detail. Davis and Rigaux suggest that three groups can be fount: (1) those in agreement; (2) husbands overestimating relative to their wives; and (3) wives overestimating relative to their husbands. In contrast, the Burns motel compares a husband's to his wife's assessments of decision making influence in a consensus-nonconsensus framework mapped onto a two-dimensional grid. In the instance of a three-point response scale ("husband," "joint," or "wife"), the comparison generates nine cells, each of which is assigned a descriptive label. Figure 1 relates both motels applied to a 100% influence scaled-response measure. The dotted lines define Davis and Rigaux's agreement area and two disagreement areas. Also, it can be seen that Burns has identified the three different "regions" within each of Davis and Rigaux's areas. These two motels constitute basic paradigms for this study.

METHOD: SAMPLE AND INSTRUMENT

Data was collected as part of a larger study using a convenience sample of 85 married couples living in Baton Rouge, Louisiana. Respondents were asked to take part in a study on how husbands and wives make purchase decisions. The study involved filling out a series of questionnaires at home with an administrator present to give the questionnaires and provide instructions. Husbands and wives were prevented from comparing answers in two ways: (1) questionnaire segments were administered randomly to each spouse, and (2) the administrator, ostensively involved in administration paperwork, remained with respondents throughout the session.

Respondents were asked to estimate the amount of influence of each spouse, defined as "...the amount of persuasion..." on a 100-point constant sum scale. This scale was chosen as it has been claimed by Jenkins (1978) to be more reliable and valid than other scale types. The spouses were told to consider the purchase of two major durable products: (1) the family automobile and (2) family room furniture. These products were selected as they are customarily used in husband and wife purchase decision making influence studies (Burns and Granbois, 1979) and both are products in which over- and underestimates have been reported. For each product, several subdecisions were specified: (1) price to pay; (2) where to buy; (3) model/brand; (4) how to pay; (5) style; and (6) when to purchase. Each subdecision was separated into a search for information stage and a final decision stage as influence differences have been found across stages in the purchase decision (Davis and Rigaux, 1974). A final item pertaining to initial recognition and suggestion to replace the old product completed the set of thirteen response items for each product.

The sample exhibited a profile similar to those often reported in husband and wife purchase decision making studies: it was upscale in terms of education and income. Husbands and wives averaged 5 years and 4 years of schooling beyond high school, respectively. Forty percent of the husbands earned more than $35,000 per year. Husbands were 40 years old, and wives were 38 years old, on the average. Marriage length averaged 15 years.

ANALYSIS AND FINDINGS

Determination of Influence Dimensions

The raw data was comprised of 170 spouses' influence distribution responses to each of thirteen subdecision items for each of two products. For convenience> it was decided to analyze the influence attributed to the wife. (Since a constant sum scale was used, husband influence was the complement.) As it was highly probable that much multicollinearity existed in the responses, a principal components factor analysis was applied for each product to determine the dimensionality of influence perceived by husbands and wives. The eigenvalue "equal to or greater than one" rule was used to determine the number of factors, and varimax rotation was applied to assist in interpretation. Both the automobile and furniture influence assessments yielded two factors which were quite similar. In particular, for each product one factor defined Financial considerations (price, how to pay, when to buy) and the other pertained to Nonfinancial considerations (where, brand/model, style, and suggesting the need to buy). Rotated factor loadings for items associated with each factor were .60 or higher. Cumulative variance explained was 79% and 82% for Family Room Furniture and Family Automobile, respectively.

Cronbach's alphas were computed to assess the internal consistency of the items loading on each factor, and these were determined to be as follows: Automobile Financial (.93); Automobile Nonfinancial, (.96); Furniture Financial, (.83); and Furniture Nonfinancial (.97). These values indicated that sufficient reliability of measurement had been attained to warrant further analysis. Next, Pearson product moment correlation coefficients were computed on the sum scores of the items for each factor to confirm that the dimensions were independent. The highest correlation was found to be .64 for husbands Financial to Nonfinancial Automobile sum scores and .61 for wives in the same pairing. All other correlations ranged from -.18 to .54, signifying little covariation was present across the four perceived influence dimensions.

Identification of Agreement and Disagreement Groupings

Comparison of the husbands group and wives group means corroborated previous findings of agreement in the aggregate. These means are presented as part of Table 1, signified as "TOTAL" for each Influence Dimension, where the greatest difference between the husbands group and the wives group means was 2.82 in the case of the Furniture Financial decision.

TABLE 1

CENTROIDS OF CLUSTER ANALYSIS RESULTS FOR INFLUENCE DIMENSIONS

The next phase of the analysis was the application of cluster analysis to matched husbands-to-wife responses. Although not without its critics and problems, cluster analysis does reveal the presence of heterogeneous subgroups within a sample based on the ratio of within-group to between-group variance and seemed the most appropriate numerical taxonomy procedure available under the objectives of the research (Punt and Stewart, 1983). A technique with the potential to identify widely differing subgroups was especially desirable in light of Burns' model which presupposes several types of agreement and disagreement couples.

A separate cluster analysis was run for each of the four purchase influence dimensions. Ward's method was used, and the cubic clustering criterion (CCC) was graphed against cluster steps to determine the final number of clusters (SAS Institute. 1982, p. 420). Inspection revealed that the CCC peaked at 8 clusters for Nonfinancial decisions for both products and at 9 clusters for Financial decisions in both instances. Figure 2 presents graphs of the cluster centroid locations while Table 1 compares their descriptive statistics. Table 1 also contains designations of the Burns' model regions they occupy. Unfortunately, neither Davis and Rigaux (1974) nor Burns (1977) indicated a relevant criterion to judge the separation of agreement and disagreement, so we adopted the rule that any difference between husbands' and wives' cluster centroid means of 102 or greater constituted disagreement. In addition, it was necessary to define range limits for agreed wife-; joint-; and husband-influence regions. We decided on limits of 61%-100%; 40%-60%; and 0%-39%, respectively.

FIGURE 2

CENTROID LOCATIONS FOR AGREEMENT AND DISAGEREMENT CLUSTERS

The figures and Table 1 reveal the great degree to which the aggregate-level means mask disagreement as well as other forms of agreement. For instance, in the case of Automobile Nonfinancial decision, determined as an Agreed Husband influence area in the aggregate, only 2 of the 8 cluster centroids fell into this region, while 2 were found in the Agreed Joint Region. In other words, only 29% of the sample fell into clusters consistent with the aggregate result designation. Presumption and Concession by both spouses were also apparent. A great variety of agreement and disagreement groups was evident across all four of the purchase influence agreement dimensions.

Investigation of the Stability of Agreement and Disagreement

These independent descriptive result comparisons were insufficient, however, as they did not address the stability of couples across cluster results. If couples exhibit stable dispositions, they should be in similar regions from influence dimension to influence dimension. If stability is not found, the phenomena of agreement and disagreement are either random variables or highly specific to the influence dimensions studied. Consequently, an attempt was made to determine the degree of stability of couples across influence dimensions.

Since cluster membership is a nominal variable, the Chi Square test for independence was appropriate. It became apparent at this point that the 8- and 9-cluster solutions could not be used due to the small sample size relative to the number of cells. Consequently, it was decided to use collapsed cluster analysis results in the form of 4-cluster solutions to overcome the Chi Square cell-size problem. Table 2 reports these results and reveals that with the exception of Furniture Nonfinancial decisions, over- or underestimation disagreements were still prevalent.

TABLE 2

CENTROlDS OF CLUSTER ANALYSIS RESULTS FOR INFLUENCE DIMENSIONS

(4 GROUP SOLUTIONS)

Chi Square results determined significant associations between Automobile Nonfinancial and Financial decisions (Chi Square = 66.5; df = 9; prob. = .0001) and between Automobile Financial and Furniture Financial decisions (Chi Square - 19.2; df - 9; prob. - .02). Inspection of these two cross-tabulation tables revealed that agreement, overestimation and underestimation was relatively stable except in the case of wife overestimation relative to the husband for Financial decisions. Approximately one-third of the couples remained in agreement that the decision would be jointly decided in both comparisons. Another five percent agreed on husband autonomy in both Automobile influence dimensions. Other subsets of couples exhibited stable disagreement predispositions across the Family Automobile purchase. Fourteen percent maintained a posture of wife overestimation while nine percent retained a husband overestimation disposition. Husband overestimation of the wife's influence also held for nine percent of the sample in the Financial decision area across both products. Although seemingly inconsequential percentages, when compared to the relatively small cluster sizes for disagreement couples, these value take on more significance.

Associations of Antecedent Variables

The final analysis undertaken addressed the question of antecedent variables which might assist in identifying and/or understanding agreement or disagreement between spouses' purchase influence assessments. Two reasons compelled this analysis. First, Pun; and Stewart (1983) contend that a gauge of the usefulness of cluster analysis is the demonstration that the clusters are meaningfully related to variables other than those used to generate them.

The second reason stemmed from prior research which has relied on demographic, work status, and attitudinal variables as determinants of husband-wife role structure. Findings have supported the notion that wives tend to assume greater decision making influence as a function of greater resource contributions to the family. Also, more liberal sex role attitudes held by either spouse are often associated with more wife influence (Davis, 1976; Burns and Granbois. 1979).

Consequently, several demographic variables were selected with the expectations that agreed wife influence and wife overestimation relative to the husband would be positively associated with her work status, education, or liberal attitudes. In the latter case, the Brogan-Kutner sex role orientation scale (1976) was used. Cronbach alphas were found to be. .71 for husbands and .67 for wives. Correlations between the several candidate variables revealed the highest to be .38 between years married and husband income level, revealing overall low multicollinearity.

Table 3 presents the results of significance tests performed using the 4-cluster groupings. Individual ANOVA's were performed rather than MANOVA's due to low correlations. Duncan's multiple range analysis was the post hoc test applied to the cases of significant (p<s.10) computed F values. These are not presented to conserve space, but two patterns were clearly evident. First, corroboration of past findings of agreement occurred. That is, agreed joint decision making was associated with shorter marriages and more liberal wife sex role orientation attitudes for both spouses. Similarly, agreed decision making autonomy for either spouse was associated with longer marriages, higher income, and more traditional wife sex role orientations for both spouses.

The second pattern concerned spouses disagreeing on relative purchase decision making influence. Wife overestimation of her influence relative to the husband's assessment was associated with greater: (1) wife education level; (2) wife income; and (3) wife hours worked per week. Husband overestimation of the wife's influence relative to her assessment was associated with less: (1) wife income; (2) wife hours worked; and (3) husband education. Interestingly, with the single exception of husband education, the significant antecedent variables for disagreement were all wife attributes. They were also largely specific to the Furniture Financial dimension. The implication here is that both spouses credit amounts of decision making influence to the wife as a function cf two factors: (1) the product under consideration, and (2) her personal characteristics which are related to her contributions to the family socioeconomic status. However, the wife is seemingly more sensitized to these factors than is the husband

DISCUSSION

This study sought to clarify the nature, stability, and antecedents of agreement, over- and underestimation of husband and wife purchase decision making influence assessments. A brief summary of the findings is in order her. As predicted by the conceptualizations of Davis and Rigaux (1974) and Burns (1977), a substantial proportion of the sample of husbands and wives did not agree.

TABLE 3

RESULTS OF SIGNIFICANCE TESTS FOR ANTECEDENT VARIABLES

Moreover, there was a great variety of disagreement and agreement groupings for any single influence assessment dimension. Within certain couples, there appeared to be a generalized tendency for spouses to maintain agreement, over- or underestimation postures across decisions for the family automobile purchase and to some extent for financial considerations across products. As noted in other studies, agreement was related to length of marriage and sex role attitudes. On the other hand, resource contributions appeared to figure into wives' overstatements of wife decision influence relative to their husbands. At the same time, a reverse effect could be seen operating for wives with less resource contributions, for they tended to understate their influence. These occurrences were concentrated in the financial decision of furniture purchase.

There are two sets of implications evident from the results which require brief discussion. The first is methodological, and the second is specific to husband and wife purchase decision making dynamics. In the first case, this study has demonstrated that a comprehensive paradigm such as the one suggested by Burns (1977) is a viable means of separating out agreeing as well as disagreeing couples. The relatively stable predispositions found make clear that perceptual differences are not random variables and must be explicitly taken into account in future husband and wife purchase decision making research. The paradigms used in this research can assist in separating out individual differences, and the antecedent variable associations suggest covariate relationships which should be considered.

The second set of implications concerns the impact of wife resources on the decision making process. Resource theory interpretation of husband and wife decision making implicitly assumes that spouses perceive these contributions equally. However, the findings of significant associations of wife resource characteristics with wife overestimation point to the conclusion that she views her contributions more extremely than does her husband. Assuming that resource theory underlies husband and wife decision making, the differences between spouses' sensitivities to the wife's contributions provide ample opportunity for spirited interaction over financial decisions for certain household durables. This interaction might be especially interesting to study in couples where the wife has recently entered the workforce, acquired more education, or received a pay increase. Subsequent research would be well advised to consider these perceptual differences in resource contributions as possible determinants of influence strategies, distributions, and outcomes.

REFERENCES

Brogan, D. and Kutner, N. C. (1976), "Measuring Sex-Role Orientation: A Normative Approach," Journal of Marriage and the Family, (Feb.), 31-39.

Burns, A. C. (1977), "Husband and Wife Purchase Decision Making Roles: Agreed, Presumed, Conceded, and Disputed," Advances in Consumer Research, 4, 50-55.

Burns, A. C. and Granbois, D. H. (1979), "Advancing the Study of Family Purchase Decision Making," Advances in Consumer Research, 6, 221-226.

Davis, H. L. and Rigaux, B. P. (1974), "Perceptions of Marital Roles in Decision Processes," Journal of Consumer Research, 1, (2), 51-62.

Davis, H. L., (1976), "Decision Making Within the Family," Journal of Consumer Research, 2, 241-260.

Granbois, D. H., and Willett, P. R. (1970), "Equivalence of Family Role Measures Based on Husband and Wife Data," Journal of Marriage and the Family, 32, 68-72.

Heer, D. M. (1962), "Husband and Wife Perceptions of Family Power Structure," Marriage and Family Living, 24, 65-67.

Jenkins, R. (1978), "Minimizing Bias in the Measurement of Spousal Influence in Family Decision Making by Self-Report Methods," Research Frontiers in Marketing Dialogues and Directions, 1978 Proceedings. Chicago: AMA. 238-43.

Punj, G. and Stewart, D. (1983), "Cluster Analysis in Marketing Research: Review and Suggestions for Application," Journal of Marketing Research, May, 10, 2, 134-48.

Safilios-Rothschild, C. (1970), "The Study of Family Power Structure: A Review 1960-69," Journal of Marriage and The Family, 32, 539-552.

SAS Institute, SAS User's Guide: Statistics, 1982 Edition, SAS Institute Inc., Gary, North Carolina, 1982, p. 420.

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