A Multi-Level Approach to Family Role Structure Research


Donald H. Granbois (1971) ,"A Multi-Level Approach to Family Role Structure Research", in SV - Proceedings of the Second Annual Conference of the Association for Consumer Research, eds. David M. Gardner, College Park, MD : Association for Consumer Research, Pages: 99-107.

Proceedings of the Second Annual Conference of the Association for Consumer Research, 1971     Pages 99-107


Donald H. Granbois, Indiana University [Professor of Marketing, Indiana University]

Since researchers in several disciplines have become interested in family role structure, reviews of the topic have revealed no shortage of empirical findings. Synthesis of these findings is impaired, however, by the variety of conceptual approaches and methods chosen by researchers, the lack of standardization of independent variables and the scarcity of methodological studies intended to assess the impact of research setting and method. Efforts at theorizing about the determinants of variations in family role structure have been further hampered by the variety of dependent variables used by researchers and by the tendency to stop at "two-variable propositions" instead of considering relationships between dependent variables. All of these points are persuasively illustrated by Davis (1970). The objectives here are to identify four major issues raised by an attempt to synthesize role structure research, and to offer a simple conceptual model that appears to be helpful in organizing the generalizations and hypotheses reported in published studies. As the title of the paper implies, an important feature of the model is its recognition of more than one concept of family roles in decision making; its levels, in fact, represent three linked dependent variables that must be considered.


(1) Individual vs. group orientation. The current popularity of individual decision models in the marketing literature on consumer behavior is reflected in the orientation that focuses on the individual viewing other family members only as potential influences. The model of Engel, Kollat and Blackwell (1968) typifies this approach. In contrast is the orientation (most common in family sociology) that views the family as a group, as a set of interacting roles. The proposed model treats the "individual vs. group" issue as a research question rather than one of researcher preference or training. It assumes that the degree of joint involvement in family decision making is measurable, and that every family can be expected to exhibit a distribution of decisions in terms of individual vs. joint decision making.

(2) Structural vs. process dependent variables. Family member roles are often summarized in terms of single measures of influence or role performance. Other research views family decision making as ongoing interaction. The choice is more than a methodological issue, but it may also be resolved empirically. The validity of simple summary measures of role structure is here assumed to be directly related to the degree of preference agreement that either exists prior to the initiation of the decision process or that occurs as the process unfolds. When agreement exists, the usual model of the decision process should provide an adequate framework, together with summary measures of "who did what" in implementing the process. More complex "interaction process" measures are called for only when disagreement exists or is revealed. The process evoked in the presence Of disagreement is appropriately called conflict resolution, and modes or strategies for achieving an acceptable outcome should be identified empirically.

(3) Survey vs. observation; real vs. hypothetical task. Wide methodological variation occurs in the literature; retrospective, cross-sectional surveys and longitudinal surveys comparing family member predictions with actual decisions contrast sharply with actual observations of decisions or tasks in home or behavioral laboratory settings. Observation studies, further, have involved beth "real" and hypothetical decisions and tasks. Methodological research, comparing results when the same issue is studied in two or more settings (sometimes using the same subjects) is clearly required.

(4) Generalizability over decisions, families and time. Researchers have varied a great deal in their willingness to generalize across decision areas. Most have reported variations among families related to socio-economic characteristics, education, wife's employment, etc. (without agreeing upon the effects of these determinants); little has been attempted empirically in terms of trends over time, although many speculate about such changes. The model includes consideration of the impact of decision type, family characteristics, and stage in family life cycle on each of the three dependent variables.

Since the same independent variables appear at each level, it is c.ear that their impact on role structure is expected to be complex and not easily summarized in simple two-variable propositions. Thus, some of the apparently conflicting results of previous research on the impact, say, of life cycle result from attempts to synthesize projects tapping different "levels" of the model and projects using dependent variables that are ambiguous as to the level involved.

Figure 1 summarizes the proposed model in flow chart form. The rationale for each level, comments on methodology at each level and suggested hypotheses dealing with the proposed determinants follow.


Much of the existing theoretical discussion of family role structure is most applicable at this level, although research has tended to use measures tapping more general concepts of involvement and role performance than the one intended here. The relevant concept proposed here is willingness to commit family resources without conferring; individuals presumably sometimes act unilaterally as individuals, making decisions without concern for the preference structures of other family members. The commonly used categories of overall role structure such as joint, husband dominant, wife dominant and autonomic (Wolfe, 1959) may be useful in identifying patterns at this level, but the empirical basis for categorizing families must be selected with care. The underlying variable has two components: Specific decisions may be jointly or individually made, and individually made decisions may be made by either husband or wife. An autonomic pattern thus reflects two dimensions: a high proportion of individually made decisions, and roughly equal division of these between husband and wife. The difficulty of categorizing families on the basis of responses to a small set of simple questions about past decisions (where a one-dimension response set is provided) should be obvious, although this categorization is commonly attempted.

If retrospective questions are to be used at this level, several requirements seem called for:

(1) Questions should be clearly limited to the measurement of the involvement of one or more than one preference structure in each past decision asked about.

(2) Two-step response categories should be provided, so that joint vs. individual involvement is clearly separated from the question of whose involvement occurred in the individual decisions.

(3) Extensive pretesting of independent responses of husbands and wives should take place before assuming that a single spouse accurately reports for the family. There is evidence that this is not the case when more inclusive questions are asked about the relative importance of husband and wife in past decisions. (Granbois & Willett, 1970).

(4) Either a large and varied group of decisions should be covered or rather general summary questions about the usual pattern in each decision area should be used. Neither alternative is a totally satisfactory solution to the problem of selecting a truly representative sample of decisions on which to base a general classification scheme.

(5) Questions tapping various phases of the decision process should be included, since involvement may vary within decisions. For example, problem recognition (beginning to consider a generic product purchase), strategies for identifying alternatives, deciding on evaluative criteria, evaluating alternatives and implementing selection should all be represented.



Observational and experimental methods seem to offer promise at this level, although existing research of this type seems more applicable at the other two levels. Couples in a behavioral laboratory (or conceivably an in-home situation) could be given a series of hypothetical (but realistic) decision tasks to work through independently, with the option of conferring or reaching agreement wherever it was felt necessary. Perhaps time limits or rewards for number of decisions completed would be necessary to insure some reasonable proportion of independent decisions. In any case, subjects should be asked to behave "normally"--to confer in situations where they usually would in real decisions, and to act independently when this alternative seemed realistic Retrospective questions could, of course, be asked of subject couples so that comparisons of the two methods could be easily made.

Hypotheses at this level must clearly distinguish between the two dimensions of involvement; the proportion of joint decision making and the allocation of independent decisions may each show somewhat different patterns of relationship with such determinants as social class and life cycle.

Space does not allow documentation or elaborate discussion, but the classification in Table 1 reflects illustrative hypotheses relevant at Level 1. Most of the relationships implied in Table 1 have received some empirical support, although it is clear that most of the relevant determinants are complexly interrelated and adequate multi-variate research taking these interactions into account has not been undertaken. In addition, the dependent variables in studies suggesting these relationships have been more inclusive than the one proposed here.



The effects of life cycle at this level have perhaps the most persuasive rationale: A shift away from joint participation over the course of the marriage may be expected for several reasons. Joint participation is costly in terms of time and may be specifically avoided to the extent that the often unpleasant process of conflict resolution is seen as a likely result. Intense needs for "togetherness" may dictate early joint participation in nearly every activity. Product use structures may become more separate, as increased income allows multiple automobiles, houses with more rooms, etc. The effects of life cycle, ironically, may be the most difficult to assess, particularly in cross-sectional research where it is usually impossible to separate generational differences between older and younger couples from true life cycle differences.


A key branch in the proposed model is agreement in preference structures, which is thought to determine the type of decision process evoked when both husband's and wife's preferences must- be considered. The flow chart visualization dichotomies agreement for ease of presentation, but it is likely that gradations may need to be considered. Experts in preference analysis may well argue that appropriate concepts and methodology are difficult enough to produce for the individual without the complication of developing agreement measures between spouses, but the need for such measures seems clear.

Both objective (researcher-scored) and subjective (perceived) measures of agreement need exploration, and patterns of relationships between objective and subjective agreement may be worth investigating. Objective measures can be obtained by administering identical instruments to husbands and wives and comparing responses; for example, a dissertation underway at Indiana University will explore correlations between spouses' responses to a multi-dimensional scaling instrument. Subjective measures include direct questions on perceived agreement and subjects' predictions of their spouses' responses to instruments administered to both husbands and wives. Such studies might well replicate the methodology used in a study of married couples' political attitudes, which found, as predicted, that perceived similarities between spouses were greater than actual. (Byrne &- Blaylock, 1963). This tendency for respondents to overstate similarities, while consistent with the social psychological theory that predicts a tendency toward cognitive symmetry in interpersonal relationships, may lead the researcher toward an overstatement of the agreement between spouses. A second potential measurement problem grows out of a likely tendency for married people, even when asked their own preferences, to be influenced by those expected of their spouses, since preferences may be conceived only in the context of a future "real" choice situation where spouse's preferences would be relevant.

Both questionnaire and behavioral laboratory choice techniques may be useful in gathering preference data, and there seems to be little basis at this time for advocating one over the other. Methodological research appears to be called for at this level. In every case, agreement measures must be sought only for decision areas where preferences of both husband and wife are seen as relevant.

Despite the apparent applicability of a variety of theoretical approaches predicting similarity in values, attitudes and beliefs among members of reference groups of all types, there is a somewhat surprising lack of empirical evidence on the degree of preference similarity between husband and wife. Perhaps the most important prediction relates degree of preference similarity to length of marriage; over the life cycle, preferences would be expected to grow more similar as the number of shared experiences and influences grows. Several complicating variables (or competing explanations) need to be accounted for in assessing evidence relating to this hypothesis, however.

(1) Cross-sectional research may reveal greater similarity between long-married spouses simply because these people represent earlier generations where agreement was more highly valued, and these spouses consciously try to reflect each others' preferences.

(2) Couples who violently disagree divorce, leaving only more compatible older couples to be studied.

(3) Length of marriage may be closely but not perfectly related to the number of times a particular product has been purchased; the latter may, in fact, be the true determinant of preference similarity.

(4) Longer-married respondents may tend more to reflect their spouses' expected preferences even when asked to report their own preferences because of the possibility (mentioned earlier) that preferences can't be articulated in the abstract but, rather, reflect expected future choice situations.

It is interesting to note that one consequence of the discussion above is that joint purchasing behavior (as distinguished from decision making) may be expected to decrease over the life cycle for two reasons: Fewer occasions arise where both husband's and wife's preferences must be satisfied, and where preferences of both are relevant but similar, purchasing can easily be delegated to one family member.

Empirical evidence at least indirectly supports the conclusion that preferences will grow more similar over time; other hypotheses must be based on sheer speculation. For example, variations in degree of preference similarity may occur across product groups because of variations in use patterns; husbands and wives may use the family car in different ways and as a consequence develop divergent preferences, whereas joint (and often simultaneous) television viewing may enhance similar preference patterns for television sets. The reader can supply many other hypotheses that merit empirical test.


Discussion of family role structure often revolves around the concept of power, and it can be argued that power is relevant only when husband and wife differ over desired outcomes. This level of the model is in many respects the most interesting (and methodologically challenging), yet practically no empirical research in consumer behavior has dealt specifically with conflict resolution. While predictions at both Level 1 and Level 2 imply a decline in the need for conflict resolution over the life cycle, it seems safe to predict at least occasional conflict in every family, although the process of resolution may vary considerably among families.

In addition to its importance in understanding family role structure, the process of conflict resolution should be studied as a potential determinant of the outcome of decision processes. Intentions fulfillment research in two quite different settings consistently reveals important proportions of purchase plans not fulfilled. That decisions not to buy ("Halt" outcome) may result from inability to resolve conflicts is indirectly suggested both in super drug store research (Granbois, 1968) and investigation of the automobile buying process (Brown, 1961). Intentions research indicates husbands and wives do not always agree on their level of intentions to purchase goods, nor on brands or stores to be selected. Thus, it is not surprising that intentions (typically reported by one family member) often result in decisions not to buy at all.

The role of empirical research here is to establish categories for classifying strategies of conflict resolution as well as to discover how these categories are distributed across decisions, in families and over the life cycle. Despite some efforts to secure data by asking respondents who "usually wins" when conflict arises, questionnaire methods seem least applicable at this level, and experimental and observational methods seem most promising.

A behavioral laboratory study at Indiana University of married couples' decision making demonstrates the empirical study of the process of conflict resolution. For each of four hypothetical decisions--new automobile, color television set, a $600 windfall and intentions for major purchases for the next year--husbands and wives individually completed highly structured decision forms before interacting, thus committing themselves to solutions (and incidentalLy enabling later comparison of individual and joint responses). As expected, numerous conflicts were revealed as joint solutions were sought. Discussion was tape recorded, and has been analyzed in terms of two variables: length of speaking time for each partner and a coding scheme of conflict resolution modes suggested by Blood (1962). The outcome at each decision point was coded as consensus (immediate agreement), wife concession, husband concession, sequential compromise, halfway compromise, creative compromise (new solution created), or arbitrary criterion (such as flipping a coin). Despite the limitations of small sample size and the disproportionate number of highly educated, upper middle income subjects, two patterns emerged in the analysis of modes of conflict resolution. Tables 2 and 3 show these results (Longman, 1970).





Future research of this kind may well be structured around the provocative model of family decision making offered by Pollay (1968), which seems relevant at the conflict resolution level. Briefly, Pollay hypothesizes determinants of the long-run division of utilities that will be maintained between spouses and advances the hypothesis that decision outcomes will tend to reflect this division. Individual decisions may deviate from this long-run division to the extent that the partners are willing to absorb "utility debt." Laboratory studies involving long sequences of decisions might investigate the tendency for outcomes to approximate the (independently established) division of utilities, the magnitude of utility debt spouses seem willing to absorb and the process of bargaining and negotiating Pollay predicts will occur as spouses attempt to seek acceptable outcomes in conflict situations.

Hypotheses about the distribution of conflict resolution modes across decision types, family types and stages in the family life cycle cannot at this point be documented with empirical evidence, but all three sets of determinants seem logical to explore. For example, partners with expertise in certain areas might be expected to win concession more often in those areas when conflict arises (perhaps illustrating French and Raven's concept of "expert power" (1959). The relative importance of husband and wife concessions may vary with their relative "coercive power" (French & Raven, 1959), which in turn may be a function of the comparative resources (education, income, etc.) each contributes to the marriage (Blood & Wolfe, 1960), and there may be a pattern related to social class, with wives in the lower classes and husbands in the upper classes winning concessions most frequently (Komarovsky, 1961). Over the life cycle, couples may develop a tendency to avoid outright attempts to win concession through coercion and to rely more heavily on the less damaging modes of compromise. Although it is difficult to document objectively, listening to married couples work out joint decisions leaves one with the distinct conclusion that young husbands tend much more than older husbands to achieve concession through coercion and that young wives display overt concern for maintaining a "friend]y" relationship even at the expense of giving in. Older married couples seem to have learned to handle conflict much more smoothly. Future research needs to explore all three kinds of determinants of modes of conflict resolution as well as to develop better coding schemes for analyzing the process.


Blood, R. O. Jr. Marriage. New York: The Free Press, 1962.

Blood, R. O. Jr. & Wolfe, D. M. Husbands and wives: The dynamics of married living. Glencoe: The Free Press, 1960.

Brown, G. H. The automobile buying decision within the family. In N. N. Foote (Ed.), Household decision-making. New York: New York University Press 1961. pp. 193-199.

Byrne, D. & Blaylock, B. Similarity of attitudes between husbands and wives. Journal of Abnormal and Social Psychology, 1963, 67, 636-640.

Davis, H. L. Some problems in building a propositional inventory: The case of family decision making. Paper presented at the Conference of the American Marketing Association, Denver, September 1968.

Engel, J. F., Kollat, D. T., & Blackwell, R. O. Consumer behavior. New York: Holt, Rinehart, 1968.

French, J. R. P. Jr. & Raven, B. The bases of social power. In D. Cartwright (Ed.), Studies in social power. Ann Arbor: Institute for Social Research, 1959. pp. 150-167.

Granbois, D. H. Improving the study of customer in-store behavior. Journal of Marketing, 1968, 32, 28-33.

Granbois, D. H. & Willett, R. P. Equivalence of family role measures based on husband and wife data. Journal of Marriage and the Family, 1970, 32, 68-72.

Komarovsky, M. Class differences in family decision-making on expenditures. In N. N. Foote (Ed.), Household decision-making. New York: New York University Press, 1961. pp. 255-265.

Longman, D. S. A laboratory study of husband-wife interaction in consumption decision making. Unpublished doctoral dissertation, Graduate School of Business, Indiana University, 1970.

Pollay, R. W. A model of family decision making. British Journal of Marketing, 1968, 2, 206-216.

Wolfe, D. M. Power and authority in the family. In D. Cartwright (Ed.), Studies in social power. Ann Arbor: Institute for Social Research, 1959, pp. 99-117.



Donald H. Granbois, Indiana University [Professor of Marketing, Indiana University]


SV - Proceedings of the Second Annual Conference of the Association for Consumer Research | 1971

Share Proceeding

Featured papers

See More


Thanks for Nothing: Expressing Gratitude Invites Exploitation by Competitors

Kelly Kiyeon Lee, Georgetown University, USA
Jeremy A. Yip, Georgetown University, USA
Cindy Chan, University of Toronto, Canada
Alison Wood Brooks, Harvard Business School, USA

Read More


Do You Trust the System? Interaction Effect between Perceived Economic Mobility and Socioeconomic Status on Fair Market Ideology and Consumer Responses

Chun-Ming Yang, Ming Chuan University, Taiwan
Chia-Chi Chang, National Chiao Tung University

Read More


E2. Donation versus Adoption: How the Mode of Helping Moderates the Effect of Emotions on Helping

Ziqi Shang, Renmin University of China
Xiuping Li, National University of Singapore, Singapore
aradhna krishna, University of Michigan, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.