Methodological Considerations in Family Decision-Making Studies

[Marilyn M. Dunsing is Professor of Family and Consumer Economics and Jeanne L. Hafstrom is Assistant Professor of Family and Consumer Economics, School of Human Resources and Family Studies, University of Illinois, Urbana-Champaign Campus, Urbana.]

Studies of decision-making present numerous problems. The methodological problems encountered in data collection, in choice of participants, and in decision-making concept are examined. The idea presented is that dissimilarity in results among studies may be explained: 1) in terms of what method is used to collect data--the survey or self-report versus the behavior or observational method; 2) in terms of who is interviewed-husband, wife, children or some combination, thereof, as well as who does the interviewing--male or female; and 3) in terms of the concept itself--hypothetical versus real problems, what decisions are studied, and decision-making stages. Suggestions are made concerning additional research needed if the methodological problems involved in decision-making are to be solved.

Increasing recognition has been given to the methodological problems encountered in decision-making studies investigating "influence" or "power" in families (Hallenbeck, 1966; Olson, 1969; Safilios-Rothschild, 1969 and 1970; Davis, 1970; Turk & Bell, 1972; Davis & Rigaux, 1974). The purpose of the present paper is to examine the methodological problems encountered; first, in data collection; second, in choice of participants; and third, in the decision-making concept itself.

DATA COLLECTION

Methodological problems in data collection exist for both the survey or self-report method and the behavioral or observational method. Specific problems include those of cost, types of power, nature of the decision-making concept, measurement of the concept, and stages of the concept.

The survey method is much less costly than the observational method. This means that with a fixed amount of money a much larger sample can be obtained using the former method. However, cost should not be the deciding factor unless the two methods give equally good results. Then, of course, the less costly method would be used, all other things being equal. But all other things are not equal. "Other" things depend on the decision-making model being used.

How is decision-making measured? Who are the respondents? What are the independent variables? Which method is better to use depends-on the answers to these questions. For example, if the model calls for the respondent's perception of who usually wins when there is a disagreement as the measure of decision-making, if the model calls for only the wife being the respondent, and if the model calls for the use of "traditional" economic-demographic variables to explain influence in decision-making, then clearly the survey method is the one to use

On the other hand. if the model calls for a consensus on the course of action taken when a disagreement exists between respondents as the measure of decision-making, and for interaction variables to be used in addition to economic-demographic variables in explaining influence in decision-making, then clearly the behavioral method is the one to use. It should be apparent that if the model calls for the use of the behavioral method and if funds are insufficient for this method, switching to the survey method would not be appropriate until a reassessment of the model takes place.

Researchers critical of the survey method contend that predicted power, which is based on the respondent's perception of the situation, is not closely related to actual or real power. This contention needs to be evaluated. We know that people are influenced by their perceptions. For example, in a recent study dealing with satisfaction with level of living (Hafstrom & Dunsing, 1973) the findings indicated that the homemaker's perception of the adequacy of family income was a more important determinant of satisfaction than actual income. Since in this study perception was significantly related to satisfaction, why should it not also be related to action?

If the wife perceives erroneously that she has the authority to make a particular decision, such as the purchase of a dishwasher, and executes this decision, then from the point of view of market behavior her perception has become reality. If the husband does not object to her purchase, then family "power" with respect to this item may have changed, or her husband's perception of the situation may be that joint decision-making has taken place.

Those researchers critical of the survey method also contend that since it contains only the respondent's perception of who makes the decision, the observational method should be used to obtain the actual or real power. These critics need to recognize that not only are the respondents reacting according to their perceptions but also observers (if used) are making judgments about the respondents' behavior in terms of their own perception of the situation. The relevant question appears to be whose perceptions--respondents' or observers'--come closer to reality?

Those critical of the survey method argue that people are more likely to be influenced by socially-desired norms when they respond to questionnaires (in front of an interviewer) than when they are reacting with others to come to a consensus about a disagreement. The validity of this argument is questionable if the interaction among family members is taking place in front of an observer. Under these conditions, there is no a priori reason to assume that the behavior of respondents would be influenced by conventional norms any differently in the behavioral than in the survey situation. If behavior is in fact influenced, it appears more likely that influence would occur in the behavioral rather than in the survey situation. This is because interaction takes place in public in the behavioral situation.

A number of studies have compared the results based on self-reporting with those based on the observational method. In general, the results do not show much agreement (Kenkel, 1963; Olson, 1969; Turk & Bell, 1972). Why is this the case? Is it because one method is in fact superior to another method, or is it because the results are based on different concepts of decision-making, and, consequently, should not be compared? In other words, if two studies are compared, dissimilarity in results may be explained in terms of differences in decision-making concepts rather than in terms of differences in method used. We will return to this problem later.

Dissimilarity in results between two studies also may be explained in terms of differences in measurement of the same decision-making concept. For example, the survey method uses the question "Who mainly decides--husband, wife, or both--when a disagreement occurs in purchasing a refrigerator?" What measurement of the concept is analogous using the behavioral method? Should the concept be measured by the amount of talking done by each participant? Should it be measured by who first suggested buying the article or by who made more suggestions regarding the purchase? Should it be measured by comparing the husband's and wife's instrumental acts, by comparing their expressive acts, or by comparing the husband's instrumental acts with the wife's expressive acts? We will return to this problem later.

Perhaps studies that have used different methods had dissimilar findings because different stages of the concept were measured. For example, the survey method may have investigated who dominated in recognizing that a problem existed in savings objectives, while the observational method may have investigated who dominated in making the final decision. Results may differ mostly because of noncomparability of stages in the family decision-making process. Again, we will return to this problem later.

These examples indicate that whether the behavioral method or the survey method is preferable depends on a number of different factors, many of which have been examined. When studies, which have used different methods, have been compared in the past it appears likely that differences in addition to method, such as meaning or measurement of the decision-making concept, have also existed. These differences make a comparison between the methods very difficult.

It has been suggested that which method to use--self-reporting or observational--should depend on the dimensions of decision-making that are being studied (Safilios-Rothschild, 1970). If this is the case, research is needed to determine which dimensions are best measured by self-reporting techniques and which are best measured by observational techniques.

CHOICE OF PARTICIPANTS

Regardless of whether the survey or observational method is used, a major problem is the choice of participants. Does it matter who is interviewed or observed when investigating "influence" or "power" in families? Should it be the wife, or husband, or both? Should it be only the children or just one of them? Should it be both the children and their parents? The majority of self-reporting studies in the past have used the wife, primarily because of her availability and, therefore, reduced cost per schedule. However, comparisons of responses made by husbands and by wives indicate differing results.

When the responses of all husbands are compared with those of all wives (mean scores), the findings are similar for many studies (Wilkening & Morrison, 1963; Davis, 1970; Granbois & Willett, 1970). However, when the responses of wives are compared with those of their own husbands, the results vary considerably depending on what decisions are being investigated. For example, in a recent study (Davis & Rigaux, 1974, p. 58) while 82 percent of the husbands and wives agreed on who made the final decision about housing (in terms of location, purchase price, or rent), only 49 percent agreed on who made the final decision about other furnishings (rugs, draperies).

Inasmuch as the responses of husbands and their wives differ, what is the nature of the bias introduced by the use of one or the other? Some studies (such as Ferber, 1955; Heer, 1962) have found that individuals tend to attribute less influence to themselves than is attributed to them by others. Other studies (such as Heer, 1962; Safilios-Rothschild, 1969) have found that while the husband was more likely to attribute "influence" to his wife than she was, the wife was more likely to perceive that they had equal "influence." In yet another study (Granbois & Willett, 1970), couples who disagreed (about half) were about equally divided between those where the wife's estimate of "influence" exceeded that of her husband's and those where the husband's estimate exceeded that of his wife's. These conflicting findings point to the need for additional research on the consequences of husband-wife participation in decision-making studies.

In some studies using the self-reporting method only children were used. Again, this is primarily because of availability and cost considerations. The feeling is that the results could be misleading if only children's responses are relied upon to obtain information about who makes the decisions in the family. Some evidence indicates that the characteristics of the children will influence the results. For example, Hess and Torney (1962) found that younger and older children were likely to perceive the family "influence" structure differently. In addition, they (as did Bowerman & Elder, 1964; and King, 1969) found that sex influenced the children's responses. Boys tended to perceive more father "influence" than girls, whereas girls tended to perceive more mother "influence" than boys. These findings indicate the need for more research concerning the effect of children's participation in decision-making studies investigating "influence" in families.

Another participant in the survey or observational method is the interviewer or observer. Much has been written about interviewer bias. It is recognized that bias can occur in terms of asking questions, interpreting questions, recording respondents' answers to open-ended questions, and giving "cues" to respondents, particularly when open-ended questions are used. Proper interviewing training should do much to eliminate this type of bias. However, another type may exist. Does the sex of the interviewer or observer influence the results?

One study (Kenkel, 1961) found some evidence that the sex of the observer did influence the respondents, particularly their interactive behavior. In this study, when a female observer was used the wives tended to have more influence and to perform more in the areas of giving ideas and suggestions than when a male observer was used. In another study (Feldman & Rand, 1965), the sex of the interviewer did not influence which sex "won" the decision. Additional studies are needed to determine whether the sex of the interviewer or observer does or does not influence the results. If sex does influence the results, then additional studies are needed to distinguish the effect of the interviewer's or observer's sex from the effects of the method used and the decisions studied.

DECISION-MAKING CONCEPT

Methodological problems related to the concept of decision-making are many and varied. One question often raised in the past is whether decision-making is unidimensional or multidimensional in nature. Another question is can families be characterized as husband-dominated, wife-dominated, or equalitarian. One investigation (Farber, 1949) divided decisions into policy and routine. others (such as Kenkel, 1957; Turk & Bell, 1972) used instrumental and expressive acts as measures of "power." As a result of findings from numerous studies (Safilios-Rothschild, 1969; Davis, 1970; Turk & Bell, 1972), the multidimensional nature of decision-making has become apparent.

In studies using the survey method, the specific question or questions asked as a basis for measurement of decision-making have varied considerably. For example, in one study respondents were asked who usually won when there was a disagreement between the spouses (Heer, 1958), while in another study respondents were asked who is the real boss in your family (Hess & Torney, 1962). Turk and Bell (1972, p. 221) obtained responses to both questions from the same group of husbands and wives. The correlation between these two measures was .40.

Turk and Bell (1972) also obtained responses to a group of decision questions used by Blood and Wolfe (1960). The questions were concerned with jobs, cars, vacations, life insurance, wife working, doctors, dwelling unit, and food. Turk and Bell combined the answers to these questions, as Blood and Wolfe had done, to arrive at one over-all score. The correlation between the Blood and Wolfe measure of decision-making and the question who usually won was only .17, but between the Blood and Wolfe measure and the question who is the real boss it was .54 (Turk & Bell, 1972, p. 221). These results raise doubts about whether the three sets of scores were measuring the same dimension of decision-making. For example, does the question who is the real boss in your family indicate influence or authority?

Doubts have also been raised about combining answers to numerous questions in order to obtain a "global" or "over-all" score to measure dimensions in decision-making. In the case of Blood and Wolfe all questions were combined, which indicated that the researchers assumed that they all measured the same dimension of decision-making. In addition, Blood and Wolfe gave all statements the same weight. Is either of the assumptions appropriate? Do all statements measure the same dimension? Do all statements represent areas of equal importance to the family? Safilios-Rothschild (1969) suggests that factor analysis be used on a series of statements to determine which of a number of specific decisions to include in a study of decision-making.

In studies using the behavioral method, variation has also occurred in the situations used as a basis for measurement of decision-making. In one study (Kenkel, 1961), husbands and wives were asked to come to a consensus on how they would spend a hypothetical gift of $300. In another study (Olson & Ryder, 1970), husbands and wives were asked to come to a consensus on 18 vignettes which presented various types of marital conflicts (such as wife's purchase of new shoes, husband's forgetfulness about throwing out the garbage, decision to have a baby). While hypothetical situations were used in both studies, the situation applied to the family itself in the former study but to other families in the latter study.

In yet another study (Olson, 1969), couples (who were expecting their first child in a few months) were asked to come to a consensus when differences existed on those items referring to decisions which might be encountered by couples the first few weeks after the birth of their first child. In another study (Hollenbeck, 1972), husbands and wives were asked to come to a consensus when differences existed about the timing of purchases within the next three years on items costing $50 or more which had been suggested individually and independently by them.

Numerous questions are raised by the examples presented above. Of what significance are the results when hypothetical rather than real problems of the participants are used? Of what significance are the results when the problems used pertain to other families rather than to the families participating in the study? Are the situations presented to the participants such that they become sufficiently involved in them so that they will play their typical roles? Do the situations represent areas of equal frequency in enactment?

The commonly used measure of dominance in decision-making in behavioral studies is Strodtbeck's (1951) revealed preference technique (RPT) or a modification of it. In the RPT each participant fills out an individual form, the responses of the participants are compared, participants are asked to come to a consensus on the items, and "winning" is measured by the proportion of instances in which the responses of each participant prevail. Both the Kenkel (1961) and the Olson and Ryder (1970) studies use "winning" as their measure

No matter how dominance in decision-making is judged, the importance of the decision for each participant should be determined or any conclusions reached by the researcher may be erroneous. Of what importance is it to decision-making if the wife wins because her husband couldn't care less about the outcome? In Hollenbeck's study, influence was measured by finding out not only the proportion of instances in which -the opinion of each participant prevailed but also the relative strength of feeling of each participant about this decision.

Irrespective of whether the survey or behavioral method is used, the researcher is faced with the problem of deciding on the decision area or product category to be studied. To illustrate, the decision area could be very broad such as material and non-material components. Or the decision area could involve different types of non-material components such as child rearing and marital relations, or different types of material components such as spending and saving. The decision area could be confined to different kinds of saving such as savings accounts, government bonds, and stock, or to broad categories of spending such as durable goods, nondurable goods, and services. The decision area could be narrowed still farther to specific types of durable goods such as automobiles, washers. dryers. and refrigerators

If the wife is found to dominate in the area of child rearing, can she also be expected to dominate in the area of marital relations? If the husband is found to dominate in the purchase of an automobile, can he also be expected to dominate in the purchase of a refrigerator? If he is found to dominate in the decision to save, can he also be expected to dominate in the decision to open a savings account? Differences in decision areas used may explain, wholly or in part, conflicting findings between studies concerning dominance in decision-making.

Another problem concerns stages in the decision-making process. In the past, researchers have not tended to recognize that the decision-making process contains a number of stages. In addition to specifying which decision areas are to be studied, researchers also need to specify the stage or stages of each decision area to be studied. A recent study (Davis & Rigaux, 1974) obtained decision information for three stages--problem recognition, search for information, and final decision--for 25 decisions varying from specific consumer goods to the child's school and program of study. Their findings indicate that for certain types of decisions the results vary considerably depending on the stage used. For example, with respect to automobiles, in the search for information stage the decision was husband-dominant, whereas in the final decision stage it was syncratic (more than 50 percent of the families made the decision jointly). It would appear that more research is needed to identify relevant stages in the decision-making process for numerous types of decisions.

Still another problem concerns the independent variables used. Economic-demographic variables have tended to be emphasized in the past. perhaps more imagination is called for in the use of different types of independent variables in future studies. In the Hollenbeck study (1972, pp. 71-72), the model included interaction variables (Bales) and personality variables (Cattell), as well as economic-demographic variables. In terms of beta coefficients, seven of the personality variables were significant at or beyond the .05 level compared to two of the interaction variables and one economic-demographic variable.

Husband's total hours away from home, the significant economic-demographic variable, is not generally used in decision-making studies. More traditional variables, such as income (husband's, wife's, total), husband's occupation, wife's employment status, and education and age of husband and wife, were used in the Hollenbeck study but were not significant. Perhaps consideration should be given to economic-demographic variables other than those generally used in future decision-making studies. While interaction variables call for the use of the behavioral method, personality variables do not. Hollenbeck's findings suggest that more consideration be given to the use of personality variables in future studies regardless of the method used.

In the past, many studies have interpreted decision-making findings as indicating power in the family. Safilios-Rothschild (1970) feels that this interpretation is not warranted. She feels that decision-making is only one measure of power; authority and influence are two additional measures. If this is the case, what are the implications for decision-making research?

In summary, it would appear that studies of decision-making present numerous problems. These include such things as selecting the method to use, deciding on the dimensions to study in terms of products and stages, deciding on who will be interviewed, and deciding on what independent variables to use. Much research effort is needed in order to solve these problems. Hopefully, as a result of future research in the area of family decision-making, it will be possible to make generalizations about decisions, particularly in the consumer area.

REFERENCES

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

Bowerman, C. E., & Elder, G. H., Jr. Adolescent perception of family power structure. American Sociological Review, 1964, 29, 551-67.

Davis, H. L. Dimensions of marital roles in consumer decision making. Journal of Marketing Research, 1970, 7, 168-77.

Davis. H. L., & Rigaux, B. P. Perception of marital roles in decision processes. The Journal of Consumer Research, 1974, 1 (2), 51-62.

Farber, B. A study of dependence and decision-making in marriage. Unpublished doctoral dissertation, University of Chicago, 1949.

Feldman, H., & Rand, M. Egocentrism-altercentrism in the husband-wife relationship. Journal of Marriage and the Family, 1965, 27, 386-91.

Ferber, R. On the reliability of purchase influence studies. Journal of Marketing, 1955, 19, 225-32.

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.

Hafstrom, J. L., & Dunsing, M. M. Level of living: Factors influencing the homemaker's satisfaction. Home Economics Research Journal, 1973, 2, 119-32

Heer, D. M. Dominance and the working wife. Social Forces, 1958, 36, 341-47.

Heer, D. M. Husband and wife perceptions of family power structure. Marriage and Family Living, 1962, 24, 65-67.

Hess, R. D., & Torney, J. V. Religion, age and sex in children's perceptions of family authority. Child Development, 1962, 33, 781-89.

Hallenbeck, P. N. An analysis of power dynamics in marriage. Journal of Marriage and the Family, 1966, 28, 200-03.

Hollenbeck, C. L. Husband-wife influence in purchasing decisions. (Doctoral dissertation, University of Illinois), Ann Arbor, Mich.: University Microfilms, 1972. No. 72-19, 848.

Kenkel, W. F. Influence differentiation in family decision making. Sociology and Social Research, 1957, 42, 18-25.

Kenkel, W. F. Family interaction in decision-making on spending. In N. N. Foote (Ed.), Household decision making. New York: N.Y.U. Press, 1961. Pp. 140-164.

Kenkel, W. F. Observational studies of husband-wife interaction in family decision-making. In M. B. Sussman (Ed.), Sourcebook in marriage and the family. (2nd ed.) Boston, Mass.: Houghton-Mifflin, 1963. Pp. 144-156.

King, K. Adolescent perception of power structure in the Negro family. Journal of Marriage and the Family, 1969, 31, 751-55.

Olson, D. H. The measurement of family power by self-report and behavioral methods. Journal of Marriage and the Family, 1969, 31, 545-50.

Olson, D. H., & Ryder, R. G. Inventory of marital conflicts (IMC): An experimental interaction procedure. Journal of Marriage and the Family, 1970, 32, 443-48.

Safilios-Rothschild, C. Family sociology or wives' family sociology? A cross-cultural examination of decision-making. Journal of Marriage and the Family, 1969, 31, 290-301.

Safilios-Rothschild, C. The study of family power structure: A review 1960-1969. Journal of Marriage and the Family, 1970, 32, 539-52.

Strodtbeck, F. L. Husband-wife interaction over revealed differences. American Sociological Review, 1951, 16, 468-73.

Turk, J. L., & Bell, N. W. Measuring power in families. Journal of Marriage and the Family, 1972, 34, 215-222.

Wilkening, E. A., & Morrison, D. E. A comparison of husband and wife responses concerning who makes farm and home decisions. Journal of Marriage and Family Living, 1963, 25, 349-51.

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