Toward Improving Household Consumption Behavior Research: Avoidance of Pitfalls in Using Alternative Household Data Collection Procedures

Citation:

Alvin C. Burns and James W. Gentry (1990) ,"Toward Improving Household Consumption Behavior Research: Avoidance of Pitfalls in Using Alternative Household Data Collection Procedures", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 518-530.

Advances in Consumer Research Volume 17, 1990      Pages 518-530

TOWARD IMPROVING HOUSEHOLD CONSUMPTION BEHAVIOR RESEARCH: AVOIDANCE OF PITFALLS IN USING ALTERNATIVE HOUSEHOLD DATA COLLECTION PROCEDURES

Alvin C. Burns, Louisiana State University

James W. Gentry, University of Nebraska - Lincoln

BACKGROUND

This Special Session of the 1989 Association for Consumer Research Conference responds to a need expressed by the participants of the ACR Household Consumption Behavior Session held as part of the 1988 ACR Conference. The 1988 Forum Session relied on a Delphi investigation and revealed critical issues inhibiting research in this area. The only documentation of this session exists in the form of a summary prepared by James Gentry. To help the reader comprehend the issues, views, and general conclusions of this two-consecutive session forum, we have appended the summary to this article. (While parts of the summary are written in jocular fashion, its underlying concern for advancing this area should not be minimized.) During both sessions, and addressed more explicitly in the second session, participants agreed that the vast majority of household consumer behavior research suffers from cross-sectional survey, positivist methodology bias. It was agreed that since most participants had little experience with alternative methodologies, a sharing of knowledge about them was a logical first step toward resolution of this bias and the ultimate advancement of our understanding of household consumption behavior.

OBJECTIVES OF THE 1989 FORUM

To this general end, the current special session was organized with the following three specific objectives in mind:

1. To provide for knowledge exchange on alternative research methodologies;

2. To create a forum for debate on alternative methodologies; and ultimately,

3. To encourage a broadening of methodological approaches to overcome the positivist methodology bias inhibiting household consumption behavior research.

The general orientation of the special session was intended to be educational. Throughout, it was assumed that intimate experience with an alternative methodology had generated a wealth of invaluable knowledge which needed to be communicated to household consumer behavior researchers who might consider using the methodology. Moreover, because of the special problems of household research (e.g., multiple respondents, divergent perceptions, joint consumption, shared resources, etc.), it was reasonable to believe that pitfalls exist, and that potential users of a method should be made aware to them.

PARTICIPANTS AND THEIR PRESENTATIONS

While the session was intended to be a dialogue between those who have learned the pitfalls of an alternative methodology and those who would consider applying that approach to their own research agendas, this Advances in Consumer Research contains abstracts of the various presenters' major points. The participants are listed below as they presented-at the 1989 ACR Conference Program.

"Methodological Problems in Survey and Experimental Research on Family Choice Processes"

Kim P. Corfman, New York University

 

"Pitfalls in Obtaining Family Purchase Research Data"

Ellen R. Foxman and Patriya S. Tansuhaj, Washington State University

 

"A Simulation Game as A Family Research Paradigm"

James W. Gentry, University of Nebraska-Lincoln

Jeffery J. Stoltman, Wayne State University

 

"Information Integration Theory Approach to Husband-Wife Decision Making"

James Shanteau, Kansas State University

C. Michael Troutman, Charles, Charles and Associates

 

"Longitudinal Methods for Family Consumer Research"

Alladi Venkatesh, University of California, Berkeley

Integrators:

William J. Qualls, Massachusetts Institute of Technology

Michael D. Rielly, Montana State University

The following pages contain abstracts of the various presenters' points. There are some variations from the original slate of presenters apparent with these abstracts. In a few cases, the presenters added co-authors who may or may not have been present at the forum, but who made important contributions to the presentations and abstracts

APPENDIX

Summary of the 1988 ACR Household Consumption Behavior Sessions

October 14, 1988 Honolulu, Hawaii

Co-Chairs

Al Burns, University of Central Florida

Jim Gentry, University of Nebraska-Lincoln

SUMMARY OF SESSION ONE: DOMAIN ISSUES

The intent of the session was to broaden the research perspectives held by household/family researchers. Given the disparate views espoused by the audience (summarized below), this goal was accomplished. On the other hand, another goal was to reach a consensus as to whom we should study and what topics most need study -- no such closure was reached.

The session started with the question "Why do we have families?" and quickly evolved into discussion about the differences between family and household. Eric Arnould cited Netting, Wilk, and Arnould, Households as a source of the dimensions that delineate kinship relationships: co-residence, pooling of resources, transmission of values, reproduction, and production and consumption functions. Becky Holman stressed that the key underlying variable is "relationships," noting that there are dysfunctional families as well as extremely close non-kinship relationships. Liz Wilson (and later C. W. Park) suggested that we incorporate models form small group decision-making and organizational behavior areas. Rich Lutz noted that we should let the phenomena determine our conceptualizations. Mike Reilly pointed out that the basic premise of the family consumer behavior area is that family decision-making processes are quite distinct from those incorporated in general interpersonal influence settings.

Another topic raised is whether we should study "decision making" or "choice." Russ Belk (and later Rich Lutz) argued strongly that the focus should be "choice," as many outcomes are not the result of a decision making process. Russ gave the example of organ donation, which studies at the University of Minnesota indicated really involved "decision avoidance" -- as options for the loved one become more negative sequentially, the family becomes backed into a corner and finally agrees to donate the organ. Mike Reilly mentioned his hypothesis that the "rationality" of the decision process is curvilinearly related to involvement; as involvement becomes very high, the "emotional afterburners" turn on.

A point that followed the choice/decision discussion was that we need to study sequences of choices. Cross-sectional approaches looking at single decisions somehow need to capture prior interactions within the family. Jeff Stoltman earlier had maintained that the construct of role specialization would be sufficiently rich to be used across a wide variety of household types; Becky Holman mentioned that such an emphasis might help capture the household members' expectations and choice history leading to the current choice task.

Other points raised:

--C.W. Park mentioned that we rely on our spouses as an additional memory bank, thus making our internal information search within the family a joint rather than individual process.

--Russ Belk pointed out that joint decision making is becoming less prevalent, as choices within the family are being made more and more individually.

--Mike Reilly mentioned the importance of shared consumption (as one item differentiating families from other groups). In a truly surprising gesture, Rich Lutz acknowledged that Mike had made an excellent point and encouraged the study of shared consumption as a direction for future research.

--Sandy Grossbart stressed the changing nature of role players in choice events across one's life. These people are often outside the nuclear family and can have major influence on choices; however, the players change over time.

--Bill Qualls suggested that we develop a categorization scheme for households, using "relationship" as one dimension. Al Burns had developed a couple of such schemes, plus a delineation of types of households in the session handout. Sandy Grossbart pointed out that the homeless are a household type not covered in the Burns' diagrams.

Session Two: Methodological Issues

There seemed to be agreement as to the need for longitudinal research on multiple household respondents involving choice sequences using multiple methodological approaches. The focus of the discussion was on how to achieve some of these goals given time and financial constraints. Unfortunately, no quick and dirty yet sufficient approaches were generated. Of the participants, only Venky Venkatesh was currently conducting a longitudinal household study.

Other issues concerned how to obtain the type of observation data that was being promoted repeatedly in the sessions. In the first session, Russ Belk had suggested that respondents could be asked to "tell stories." Don Granbois' suggestion in the Delphi process that we consider using focus groups was raised as a possible means of collection data from the family about choice sequences. Jeff Stoltman and Kim Corfman raised cautionary notes, discussing the potential for conflict (and even physical violence) that can occur as issues which the family typically chooses to lie dormant suddenly surface.

The last part of the session focused on where we should go from here. Al Burns presented his Blue Skies ideas and discussion followed in terms of what might be feasible. Jim Gentry volunteered to continue to develop a list of CB-oriented family researchers in order to improve communication among people in the field. For example, Benny Regaux-Bricmont used the list to announce the Fall 1990 special issue of Recherche et Applications en Marketing--the official journal of the French Marketing Association--on family decision making. Ellen Foxman offered to initiate a communication network via BITNET. Jim and Ellen will develop a questionnaire about interest in the area so that we can provide a summary of the sub-areas of household consumer behavior which people in the field are currently studying. The intended outcome will be a crude newsletter going to people in the field and, eventually, the development of a BITNET network for the dissemination of information, questionnaires, data bases, a family/household bibliography, etc.

Al Burns volunteered to work with the ACR hierarchy to increase the likelihood of having a special session at next year's conference that will focus on those issues stressed as being most critical in the current sessions.

Venky Venkatesh offered to try to arrange a Household Consumption Behavior Conference/Workshop at UC-Irvine in early 1990.

 

METHODOLOGICAL PROBLEMS IN SURVEY AND EXPERIMENTAL RESEARCH ON FAMILY CHOICE PROCESSES

Kim P. Corfman, New York University

Research on families in sociology and marketing has a long history of experimentation with a variety of methods for collecting data. Numerous observations have been made concerning the weaknesses of many of these approaches. This paper is an attempt to gather together the pitfalls associated with using surveys and experiments in the investigation of family choice processes.

SURVEY METHODS

Survey research makes use of reputational methods of measuring influence in decision-making. One or more family members are asked for reports on aspects of their joint decision-making through interviews, questionnaires, or diaries.

Because choice processes vary from one occasion to the next due to variations in the topic, the timing and the people involved, researchers tend to restrict their focus to particular types of decisions (Ferber and Lee 1974; Munsinger, Weber and Hansen 1975). While this limits the ability to generalize it does help justify conclusions that are drawn about the chosen topic. When survey methods are used, limiting investigation to specific types of choices makes data collection more difficult. Only a small subset of the typical sample is likely to have engaged in the activities selected, recently enough that they can reasonably be asked to report on them.

Because families make many decisions over the course of their existence some researchers have taken a longitudinal approach and examined the impact of choices made in the past on subsequent choices (Corfman and Lehmann 1987). When this is attempted in the context of a survey, all of the usual panel problems of attrition and reporting error apply.

Most of the problems associated with survey research are related to the fact that subjects are providing influence perceptions which may not conform to reality. The norm until very recently, despite 20 years of warnings against it (Bokemeier and Monroe 1983; Safilios-Rothschild 1970; Wilkening and Morrison 1963), was to ask a single family member for information on family members' influence in decision-making. Usually the wife (Blood and Wolfe 1960; Green and Cunningham 1975; Habemman and Elinson 1967) and sometimes the child (Herbst 1952, Hoffman 1960, Straus 1962) was selected for convenience, on the assumption that this individual would provide reasonably accurate data. I4lore recent studies have demonstrated that when more than one family member is questioned there is often considerable disagreement among them (Bums and Hopper 1986; Davis 1970, 1971; Wilkes 1975). In a recent study, Monroe et al. (1985) demonstrated that 24 percent of the couples in their sample would have been substantially misrepresented had only one spouse been questioned. The degree of incongruence and inaccuracy depends upon many factors including the decision topic, the specificity of the question and the actual involvement of the subject, but it can generally be concluded that influence perceptions are biased and should be used with care. Bokemeier and Monroe (1983) reviewed 80 articles on family power drawn from 48 different professional journals from 1965 to 1978 and found that many researchers extended conclusions to family members other than the one(s) from whom the data were collected. At the very least caution should be exercised if generalizations are applied to family members not questioned in the survey.

Few studies have explicitly addressed the sources of error in reporting on relative influence, but several may be hypothesized (Corfman 1989; Davis 1970; Safilios-Rothschild 1970). First, it may be difficult to infer from past encounters and their outcomes who had more influence. Family members do not always enter into joint decisions with completely formed preferences or strong commitments to them. When you add this to the likelihood that both individual and joint learning has occurred in the process of making a choice, it may not be at all clear who was more influential. Second, perceptual and reporting biases may operate. Perceptual biases are likely to result from the desire to preserve a self-image (e.g., dominant, expert, flexible, supportive). Reporting biases occur when a family member wishes to give an inaccurate impression. For example, social norms that dictate right to authority, appropriate family roles and the importance of equality in the relationship may affect the way a subject perceives or chooses to represent his or her family's choice processes. Finally, even when perceptions are accurate and subjects wish to report them, memories are far from perfect.

Another problem with survey research is related to the specificity with which questions are asked. It has been shown that more specific questions that concern individual decision topics or aspects of those decisions produce more congruence and probably more accuracy than global questions (Davis 1971; Munsinger, Weber and Hansen 1975; Turk an Bell 1972; Wilkes 1975). This is probably related to the issues of validity with respect to global measures. Can families be categorized according to influence patterns or decision-making roles as the use of global measures implies, or is influence specific to the type of decision, the aspect of the decision, the family and other contextual factors such as timing and proximity of potential influencers? The bulk of research in the consumer behavior of families argues for the latter. This means that global measures are not indicators of who always has more influence, but of who, on average, is more influential. Further, abstracting from many specific instances to a global statement is not an exercise to which subjects are likely to be accustomed and it involves the compounding of subjective judgments from many past decisions. Both are likely to result in error.

Experimental Methods

While experimental methods avoid many of the pitfalls associated with survey research, they present another set of problems. Most of these problems fall into the following categories of issues: convenience, task realism, experimental behavior, and situation volatility.

Convenience. Experimental research is considerably less convenient to conduct than survey research. Because the subject burden is greater it is more difficult to conscript a representative sample of family units. Further, because the unit of observation includes more than one individual, the logistics of getting the relevant parties together with the experimenter are complex. (As with survey research, identifying these parties is also an important and non-trivial task.) Once these problems have been addressed, the experimental session imposes a greater burden of effort and time on both the subjects and the experimenter than most surveys.

Task realism. When a similar experiment is being performed on all family units, the appropriateness of the decisions to each unit must be assessed to ensure that conclusions are not drawn from a family's behavior in making a decision they would not normally make. For example, researchers who use interaction-based measures have tended to present contrived situations to their subjects. Kenkel's (1965) method asks couples to decide how to spend an imaginary $300 and Strodtbeck's (1951) method asks family members to resolve differences in values revealed through their answers to a questionnaire. Although studies of consumer behavior tend to use more concrete tasks, they also run the risk of asking subjects to make unrealistic decisions.

The timing of experimental decisions is also necessarily artificial. Therefore, even if the decision is one the family might reasonably make, the members have not chosen to make it, nor have they chosen the time at which to make it. When a family is asked to made a sequence of decisions the condensed time frame adds an additional element of artificiality. Also related to the issue of timing is the observation that, in their natural settings, families often face several problems concurrently (Davis 1976). When an experimenter presents one clearly defined task at a time, the choice situation is unrealistically simplified.

Further complicating this is the nature of the decisions that tend to be included in an experimental sequence. Even if some effort is made to present a variety of topics, they are not likely to represent the range of decisions families make in a typical day or week. If it is hypothesized that the outcomes of past decisions affect future processes and outcomes, it is important to ask what "counts" as part of a family's decision history. Do all decisions play future roles or are the only relevant to decisions on similar subjects?

Another element of artificiality is imposed by the requirement often made or inferred by subjects that they must make decisions. A very popular option in "real life" is not to decide. Decisions are often put off temporarily or indefinitely. Finally, families participating in experiments face the same unrealistic conditions faced by all experimental subjects. They are working under ideal conditions meaning, among other things, that they are likely to have higher energy levels and face fewer distractions than they would in their natural environments (Davis 1976).

Experimental behavior. A number of researchers have hypothesized that people behave differently when they are being observed than when they are alone (Corfman and Lehmann 1987; Karlsson 1964; Olson 1969; Safilios-Rothschild 1970). One clear reason for this is that in public people prefer conform to social norms and exhibit the behavior they believe is socially acceptable. This may result, for example, in fewer uses of threats and coercion, fewer displays of anger and frustration, and greater assumption of authority by traditional husbands. While some kinds of behavior may be socially acceptable in other settings, they are not appropriate in the presence of an unfamiliar third party (an experimenter or observer). For this reason such potentially influential intimate behavior as displays of affection and sexual negotiations are less likely to be exhibited in experimental settings. The gender of the experimenter may also influence the effect. Kenkel (1961) found that when observers were female, wives were more active and powerful. Conducting experiments in families' own homes may reduce the artificiality. O'Rourke (1963) reported more disagreement, more activity, less decision-making efficiency, and less emotionality when decisions were made in the lab than when they were made at home. Subjects may also behave differently in experiments because some influence techniques are not available to them within the lab or the limited time frame. Family members cannot wait for the right moment or mood to press for acquiescence. They cannot engage in many reward or bribe activities such as giving gifts and cooking meals. Nor can they nag intermittently over a long period of time. Finally, regardless of the potential importance of the choice topics presented to subjects, they rarely lose sight of the fact that they are participating in an experiment in which they cannot be held to their choices. The result is that preferences are often less intense and positions defended less vigorously.

Volatility of situation. A final problem that may arise in experiments with the potential for conflict is that the session may become more emotionally intense that the observer is equipped to handle. The presence of an observer and the experimental setting reduce this probability (and make responses less representative). However, experimenters should be prepared to redirect or terminate discussions.

PROBLEMS IN COMMON

In addition to those discussed above, survey and experimental methods have a number of problems in common. The following affect the legitimacy of generalizing from the results of one study to other families and situations.

Family researchers tend to use nonprobability samples, especially convenience samples. As observed by Kitson et al. (1982), convenience samples are inexpensive and convenient, but volunteers may have special characteristics or concerns about the topic which make them poor representatives of the population. For example, families who are willing to participate in studies of their choice behavior are likely to be more stable than the average for the population. Convenience samples also tend to omit isolated people who are not tied into social networks. While these samples have obvious shortcomings, they do have value if their limitations are kept in mind. There are ways to reduce the relationship between the topic being studied and subjects' interests and self-selection bias related to the topic. For example, while the task itself should not be misrepresented, the research questions can be obscured until the debriefing. Researchers can also change the reasons for self-selection so that they are not so closely related to the research subject. For example, if large enough financial or gift incentives are offered they will overcome the reluctance of some families to be examined.

Identifying the family members who are involved in the decisions being studied is another difficult task researchers encounter. Involvement varies according to type of decision, product category, decision phase, attribute, time of day, decision history, family characteristics, and probably other factors as well. Most studies focus on the couple as the relevant unit of observation, although some choices may be made by only one spouse and others may involve children, grandparents, and other members of the extended family.

Even if the involved parties are correctly identified for the particular survey or task, because relative influence (as well as involvement) varies according to the product category, decision phase, attribute, etc., care must be taken not to generalize from the findings to situations which may only appear to be similar.

While the tone of this review is undeniably negative, this does not mean that current methodologies are without value. Despite the many problems associated with doing survey and experimental research on families, awareness of the pitfalls, use of multiple methods, and integration of evidence accumulated over a period of time will undoubtedly lead to interesting and important insights.

REFERENCES

Blood, Robert O., Jr. and Donald M. Wolfe (1960), Husbands, Wives: The Dynamics of Married Living, Glencoe, IL, Free Press.

Bokemeier, Janet and Pamela Monroe (1983), "Continued Reliance on One Respondent in Family Decision-making Studies: A Content Analysis," Journal of Marriage and the Family, 45 (August), 645-652.

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

Corfman, Kim P. (1989), "Perceptions of Relative Influence in Families: Formation and Measurement," working paper, New York University, NY, 10003.

Corfman, Kim P. and Donald R. Lehmann (1987), "Models of Cooperative Group Decision-Making and Relative Influence: An Experimental Investigation of Family Purchase Decisions," Journal of Consumer Research, 14 (June), 1-13.

Davis, Harry L. (1970), "Dimensions of Marital Roles-in Consumer Decision Making," Journal of Marketing Research, 7 (May), 168-177.

Davis, Harry L. (1971), "Measurement of Husband-Wife Influence in Consumer Purchase Decisions," Journal of Marketing Research, 8 (August), 305-312.

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

Ferber, Robert and Lucy Chao Lee (1974), "Husband-Wife Influence in Family Purchasing Behavior," Journal of Consumer Research, 1 (June), 43-50.

Green, Robert T. and Isabella M. Cunningham (1975), "Feminine Role Perception and Family Purchasing Decisions," Journal of Marketing Research, 12 (August), 325-332.

Haberman, Paul W. and Jack and Elinson (1967), "Family Income Reported in Surveys: Husbands Versus Wives," Journal of Marketing Research, 4 (May), 191 - 194.

Herbst, P.G. (1952), "The Measurement of Family Relationships," Human Relations, S (February), 3 -35.

Hoffman, Lois W. (1960), "Effects of the Employment of Mothers on Parental Power Relations and the Division of Household Tasks," Marriage and Family Living, 22 (February), 2735.

Karlsson, Georg (1964), "Reliability of Data Obtained on Family Authority with a Culture Free Instrument," Transactions of the Fifth World Congress of Sociology, Vol IV, World Congress of Sociology, Washington, D.C.: International Sociological Association.

Kenkel, William F. (1965), "Observational Studies of Husband-Wife Interaction in Family Decision-Making," in Sourcebook in Marriage and the Family (2nd edition), ed., Marvin B. Sussman, Boston: Houghton-Mifflin, 144-156.

Kitson, Gay C., Marvin B. Sussman, Gwendolyn K. Williams, Randi B. Zeehandelaar, Barbara K. Shickmanter and Jeanne L. Steinberger (1982), "Sampling Issues in Family Research," Journal of Marriage and the Family, 44 (November), 965981.

Monroe, Pamela A., Janet Bokemeier, J. Morley Kotchen and Harlley McKean (1985), "Spousal Response Consistency in Decision-making Research," Journal of Marriage and the Family, 47 (August), 733-738.

Munsinger, Gary M., Jean E. Weber, and Richard W. Hansen (1975), "Joint Home Purchasing Decisions by Husbands and Wives," Journal of Consumer Research, 1 (March) 60-66.

Olson, David H. (1969), "The Measurement of Family Power By Self-Report and Behavioral Methods," Journal of Marriage and the Family, 31 (August) 545-550.

O'Rourke, John F. (1963), "Field and Laboratory: The Decision-making Behavior of Family Groups in Two Experimental Conditions," Sociometry, 26 (December), 422-435.

Safilios-Rothschild, Constantina (1970), "The Study of Family Power Structure: A Review 19601969," Journal of Marriage and the Family, 32 (November), 539-552.

Straus, M.A. (1962), "Conjugal Power Structure and Adolescent Personality," Marriage and Family Living, 24 (February), 17-25.

Strodtbeck, F.L. (1951), "Husband-Wife Interaction over Revealed Differences," American Sociological Review, 16 (August), 468-473.

Turk, James L. and Norman W. Bell (1972), "Measuring Power in Families," Journal of Marriage and the Family, 34 (May), 215-222.

Wilkening, Eugene A. and Denton E. Morrison (1963), "A Comparison of Husband and Wife Responses Concerning Who Makes Farm and Home Decisions," Marriage and Family Living, 25 (August), 349-351.

Wilkes, Robert E. (1975), "Husband-Wife Influence in Purchase Decisions -- A Confirmation and Extension," Journal of Marketing Research, 7, 224-227.

 

THE USE OF TRIAD DATA TO STUDY FAMILY PURCHASE DECISIONS

Patriya S. Tansuhaj, Washington State University

Ellen R. Foxman, Washington State University

INTRODUCTION

Families are the consumption units for many products purchased in the US and elsewhere, and family members clearly affect the purchase of numerous products by and for the sole use of individuals. Further, families are an enduring social unit, and it can be expected that both their influence on individuals and their own distinctive behaviors as consumption units will exist for the foreseeable future. These facts make knowledge of family purchase processes and behaviors a matter of considerable practical utility and theoretical import, and help explain researchers' continuing interest in studying family consumption. What we know (or think we know) about family purchase behaviors, however, is very much affected by how we try to study them. Family researchers often acknowledge potential methodological weaknesses in their studies. The purpose of this paper is to compare and contrast our approach to study design, data collection and analysis with more traditional approaches. The advantages and disadvantages of using triad data (i.e., responses from father, mother and child) to understand family purchase decisions are discussed. This information may help other researchers avoid problems we have encountered and perhaps lead to the development of alternative approaches to the conduct of family research that present more accurate, complete, and integrated pictures of family purchase decisions. The experience of using triad data also raises some interesting research questions for future research to answer.

AN EMPIRICAL STUDY USING TRIAD DATA

Most family studies have focused on purchase roles and influence relationships between husbands and wives. However, there is increasing evidence in the popular press that children now play a significantly more active role in family purchase decisions (Dagnoli 1987; Graham 1989). The broad purpose of our study was to assess children's influence in family purchase decision processes (Foxman, Tansuhaj and Ekstrom 1989). A series of questions were raised by the decision to include children in studying family purchasing:

1) Which children in the family should be questioned? It was felt to be practically impossible to question all children in the family and thus obtain a measure of all children's combined influence in purchase decisions. In the face of this practical limitation, it was decided to use an adolescent child as a respondent. This choice was supported by research on child development. Adolescents are more likely than younger children to have matured cognitively and to be active in a range of family purchase tasks (Elkind 1968; Mussen 1973; Mussen, Conger, and Kagan 1969).

2) Which family members should be asked to assess their perception of the adolescent's influence? Because there is ample evidence that spouses can disagree on purchase-related decisions and perceptions, it was felt to be necessary to obtain data from both mothers and fathers as well as the adolescent child

3) Which types of influence? Most family purchase studies have focused on particular products and services. Numerous studies have found that purchase influence perceptions differ depending on the product under consideration. It was felt, however, that perhaps a more normative dimension of product influence might also be assessed; that is, it might be useful to possible to tap perceptions of family member participation and influence in family purchasing as a whole. As a result, items assessing both product-specific purchase influence and general influence in family decision processes were included in the study.

WHAT WAS DIFFERENT ABOUT THIS STUDY?

The Foxman, Tansuhaj and Ekstrom (1989) study differed from most other marketing family purchase studies in its use of triad data and adoption of a Euclidean distance measure to represent triad influence relationships. These two aspects are discussed separately below.

Use of Triad Data

In a review of fifteen previous marketing studies related to children's influence in family decision making, it was found that mothers were usually asked to rate their children's influence. In two studies of older children (i.e., college students), children rated themselves (Converse and Crawford 1949; Perrault and Russ 1971). Berey and Pollay (1968) and Atkin (1978) included both mothers and small children in their observational studies. Jenkins (1979) was first to ask the fathers in a focus group of husbands and wives about their perception of children's purchase influence.

Belch. et al. (1980; 1985) and our study were the only studies which included influence perceptions from triads of fathers, mothers, and children. In the Belch studies, however, respondents were asked to rate their own and other family members' influence separately using a scale in which 1 meant "no input" and 6 meant "all of input" -- a measurement approach which permitted family members to overstate their influence. The study also combined the influence ratings as the mean of the father, the mother, and adolescent responses.

Our study avoided these problems by utilizing a relative influence scale (for the product influence items) and agree-disagree Likert scales (for the general influence items) -- and by not combining the influence perceptions of triad respondents. This approach made individual influence perceptions more easily discernible but presented analytic problems in summarizing triad influence relationships. These analytical problems are discussed in the next section.

Analytical Issues

The most critical problem stemming from using multiple respondents from each family (i.e., more than two) is data analysis. It is even more complicated when the aim is to arrive at a single household measure of perceived influence.

In our initial approach to understanding family influence relationships, we used a simple multitrait-multimethod analysis to identify agreement and disagreement between father-mother, father-child, and mother-child dyads. We later were able to obtain a single score representing the extent to which triad member influence perceptions converged by calculating the Euclidean distance among the family members' factor scores. It should be possible to construct an index of families based on this household measure of convergence. In our study, we-only tried to identify factors that might help explain convergence or a lack of it.

PROS AND CONS OF USING TRIAD DATA

From the aforementioned section, it is clear that the key advantage of using multiple respondents in family studies is that it potentially helps us obtain a more-accurate picture of family decision processes. The key disadvantages involve difficulty in collecting and analyzing the data.

Considering data collection for multiple respondents, our study encountered a severe nonresponse problem in attempting to obtain data from fathers. We asked over 500 junior high school students to fill out the adolescent version of the questionnaire. The actual data set employed in our study consisted of only 161 triads because, while over 400 mothers responded, less than 200 fathers did.

Although we gained additional information by including the father in attempting to more accurately assess children's influence, we actually gained that information from a much smaller number of families. As we consider including more respondents in family studies, we need to remember that the non-response problem is likely to increase as the size of the set of individuals to be analyzed as a family unit increases.

Considering data analysis, it is clear that the multivariate methods mostly favored in marketing family studies are poorly suited to the analysis of triad or larger-group data. Methods to deal with this type of data are largely undeveloped or unexplored. Consequently, researchers who surmount the problems associated with collecting data from larger numbers of family members still must deal creatively with the question of what to do with that data once it has been obtained.

FUTURE RESEARCH USING MULTIPLE RESPONDENTS

The obvious issues to be resolved in future multiple respondent family research are the data collection and analysis problems discussed above. However, there are additional matters that deserve concern. For example, including the relevant respondents in a particular family study is probably more important than including a lot of respondents. If it is possible to gain information somehow on which purchases a family member influences, then perhaps it will be possible to exclude them from those studies of decisions in which they do not participate .

The question of who needs to be considered as a family member is also important in both domestic and cross-national or cross-cultural family studies. Marketing studies for the most part consider the family to consist of a father, mother, and one or more children -- the "standard" U.S. nuclear family. Such a family is increasingly less standard in the U.S., and may never have been the standard in some U.S. subcultures or in other cultures. The point is that the relevant family members in family purchase studies may not be part of the nuclear family.

There is also the matter of possible gender effects associated with family purchase data collection. While such effects have been little investigated in marketing studies, there is some evidence that the sex of the interviewer, contact person, sponsor, or observer in a study can affect the data collected for that study in a systematic way (Warren 1980). While it could be relatively easy to control for such an effect in a single-respondent study, controlling for gender interaction effects between researchers and respondents will become increasing difficult as the number of respondents (presumably of different sexes) increases.

REFERENCES

Atkin, Charles K. (1978), "Observation of Parent-Child Interaction in Supermarket Decision-making," Journal of Marketing, 42 (October), 41-45.

Belch, George, Michael A. Belch, and Gayle Ceresino (1985), "Parental and Teenage Child Influences in Family Decision Making," Journal of Business Research, 13 (April), 163-176.

Belch, George, Michael A. Belch, and Donald Sciglimpaglia (1980), "Conflict in Family Decision making: An Exploratory Investigation," in Advances in Consumer Research, Vol. 7., ed. Jerry C. Olson, Ann Arbor, MI: Association for Consumer Research, 475-479.

Berey, Lewis A. and Richard W. Pollay (1968), "The Influencing Role of the Child in Family Decision Making," Journal of Marketing Research, 5 (February), 70-72.

Converse, Paul D. and Malcolm Crawford (1949), "Family Buying: Who Does It? Who Influences It?" Current Economic Comment, 11 (November), 38-f o.

Dagnoli, Judann (1987), "Study Finds Children Making Strong Impact on Purchases," Advertising Age, April 13, 2.

Elkind, David (1968), "Cognitive Development in Adolescence," Understanding Adolescence, ed. James F. Adams, Boston, MA: Allyn & Bacon, 128-158.

Foxman, Ellen, Watriya Tansuhaj and Karin Ekstrom (1989), "Family Members' Perceptions of Adolescents' Influence in Family Decision Making," Journal of Consumer Research, 15 (March), 48249.

Graham, Ellen (1989), "As Kids Gain Power of the Purse, Marketing Takes Aim at Them," The Wall Street Journal, January 19, 1.

Jenkins, Roger L. (1979), "The Influence of Children in Family Decision-Making: Parents' Perceptions," Advances in Consumer Research, Vol. 6, ed. William L. Wilkie, Ann Arbor, MI: Association for Consumer Research, 413-418.

Mussen, Paul H. (1973), The Psychological Development of the Child, Englewood Cliffs, NJ: Prentice-Hall.

Mussen, Paul H., John- J. Conber, and Jorome Kagan (1969), Child Development and Personality, New York: Harper and Row.

Perreault, William D., Jr. and Frederick A. Russ (1971), "Student Influence on Family Purchase Decisions," in Summer Educators' Conference Proceedings, Vol. 33, ed. Fred C. Allvine, Chicago, IL; American Marketing Association, 386-389.

Warren, Carol A. B. (1980), Gender Issues in Field Research, Newbury Park: Sage Publications.

 

A SIMULATION GAME AS A FAMILY RESEARCH PARADIGM

James W. Gentry, University of Nebraska-Lincoln

Jeffrey J. Stoltman, Wayne State University

Kevin Coulson, University of Nebraska-Lincoln

ABSTRACT -

This study discusses the simulation game as a research paradigm. It reviews the general literature on simulation gaming for research purposes, and then discusses a specific application in which the Bean Game (Yearns and Manning 1984) is used to investigate issues studied previously by Corfman and Lehmann (1987), Davis, Hoch, and Ragsdale (1986), and Park (1982). The game is found to impose an unrealistic structure on the data collection; nonetheless, the paradigm does provide insight into the process issues of interest.

INTRODUCTION

The use of a simulation game as a research paradigm is not a new phenomenon; in fact, Cohen and Rhenman (1961) stated that many of the developers of the early computer simulation games were as interested in using them for research purposes as they were for pedagogical purposes. There are a number of articles available (Bass 1964; Cohen and Rhenman 1961; Nees 1983; Rowland and Gardner 1973) which do a very thorough job of discussing the pros and cons of using a simulation game as a paradigm for investigating substantive issues.

The use of games for family research is relatively rare. While some of the early uses of games in family sociology were for research purposes (Haley 1962; Straus 1966), most of the games in the family area have been used in the area of family counseling/therapy (see Rabin 1983 for an annotated bibliography) or for pedagogical purposes in family economics classes. This paper will discuss the benefits and the methodological weaknesses associated with simulation games in terms of family research.

Advantages. To some extent, the game environment provides a middle ground between laboratory and field research. The simulation provides greater opportunity for control than does field research and yet allows one to study a sequence of decisions over a long period of (simulated) time. The simulation can insure that certain situations occur. The cost of collecting data is far lower in a simulation than in a field study, and the simulated environment removes much of the sensitivity associated with the problem area.

Decision history is extremely important in understanding family processes, with feedback from earlier decisions providing constraints for subsequent decisions. The longitudinal nature of most simulations can enable the researcher to manipulate experimentally a number of situational variables in order to explore the dynamic interaction and impact of these variables on group performance and satisfaction (Rowland and Gardner 1973). The game decision making can be interrupted, dissected, and reconstructed, allowing the researcher to obtain a wealth of information on the decision process itself. Bass (1964) notes that the use of a game provides the ability to replicate studies, a possibility which is lacking in field studies. The game can also provide the stimulus for focused interactive verbal protocols when spouses play the game together.

Disadvantages. A major criticism of games as research paradigms is their artificiality. Tedeschi, Schlenker, and Bonoma (1973) blamed this on the arbitrary rules and trivial outcomes which are provided by the experimenter, thus failing to tap the rich behavioral repertoires of subjects and to elicit powerful motives.

Another disadvantage is that "control" may be largely illusory, as respondent variation will result in vastly different wealth conditions at any given time. Any single manipulation later in the game will be seen differentially because of the different frames of reference generated from prior decision sequences. Consequently, a within household analysis may be required.

Evaluation of the "Bean Game" as a Family Research Paradigm

We used the "Bean Game" (Yearns and Manning 1984) to investigate spousal decision making in a house buying context. Subjects were required to allocate beans to various housing attributes (individually and then jointly) under varying financial conditions (budgets of 20, 15, and 12 beans).

Purpose. The intent for using the Bean Game in this study was to investigate spousal preference prediction (similar to Corfman and Lehmann 1987; Davis, Hoch, and Ragsdale 1986; Kenny and Acitelli 1988) and the house-buying decision process (similar to Park 1982; Park and Lutz 1982).

Process. Eighteen couples participated in the study. First each individual provided importance ratings for the 18 attributes included in the game and then provided ratings for his/her spouse. Each spouse was then instructed to play the game alone at the 20, 15, and 12 bean levels. The couples then played the game together, making any necessary tradeoffs to arrive at a conclusion satisfactory to both.

Brief Overview of the Analyses. We performed analyses both within and across couples. Within couples, we correlated the spouse's importance ratings with his/her partner's and with the bean allocations in the game. Across couples, we obtained pooled correlations among importance ratings and between the importance ratings and the actual allocations in order to make conclusions as to overall influence. The results support the anchoring and adjustment process suggested by Davis, Hoch, and Ragsdale (1986), as one's predicted ratings are closer to one's own ratings than to one's spouse's ratings. The results also indicate that one can better predict one's spouse's ratings with one's own ratings than with one's predicted ratings for the spouse. In aggregate, it appears that husbands can predict wives' ratings better than wives can predict husbands' ratings. By correlating the differences in the importance ratings with the bean allocations across couples, we were able to see if some were consistently "wife-dominated" while others were consistently "husband-dominated." The majority of the correlations indicated that husbands have relatively more influence.

Park (1982) discussed the relative ease of one spouse's determining the other spouse's preference levels on salient objective dimensions as opposed to salient subjective dimensions. Further, he suggested that couples would reach early agreement on the salient objective dimensions. Our process allowed us to investigate the relationships between spouses' ratings on the salient objective dimensions. Those objective dimensions rated greater than five on a seven-point scale by at least one spouse were investigated; in general, the correlations on these dimensions were not greater than the correlations on all dimensions. Thus, it would appear that the couples do not come to the task in greater agreement on the critical dimensions than on any other dimensions.

The Bean Game's structure allowed us to examine patterns of individual and joint behavior (in the form of bean allocations) across situations involving different levels of financial constraints. One outcome that could not be observed in the methodologies used by Corfman and Lehmann (1987), Davis, Hoch, and Ragsdale (1986), and Park (1982) is that the joint allocations often were more extreme than either of the individual allocations. For both the 20 and 12-bean conditions, joint allocations for five of the 18 dimensions were more extreme. Further, none of the extreme-shift patterns were the same across the two conditions. Thus the results provide little support for an averaging model, but do support the proposition that situational influences are very important.

Park (1982) also suggested that spouses would make concessions based on preference differences across spouses, a relationship found in the Corfman and Lehmann (1987) study. We investigated those dimensions in which there was a large divergence in the importance ratings within couples, and then looked at the joint allocations. We did not find a pattern of concessions toward the spouse with the higher importance ratings. Further, we found that most couples were largely unaware of the existence of the discrepancies.

OVERALL EVALUATION OF THE BEAN GAME

Our observation is that the Bean Game, while somewhat incomplete in its structure, can be used to provide insight. The data collection process for the game (where couples went through the task in the privacy of their own homes) required far less effort than that in the Park (1982) and Park and Lutz (1982) studies. Had we required subjects to do this in the laboratory and then recorded the process, we would have been able to develop decision nets as in the Park studies. An alternative to obtaining protocols for the whole process (each spouse individually and then the joint data collection) would be to have focused discussions of their predictions of each other's preferences, about those dimensions for which there were discrepancies between their importance ratings, and about the final joint bean allocations.

Tedeschi, Schlenker, and Bonoma (1973, p. 197) noted that "like other experimental paradigms, games constrain what the subjects can do." We found this to be true in terms of the attributes selected for consideration in the selection of a house. The game dealt with 18 such criteria, which did not include commonly considered attributes such as resale value, school district, or size of kitchen. The comprehensive nature of the 18 criteria included in the Bean Game was investigated by having 100 respondents (mostly older students) rate 40 housing attributes which had been compiled from a survey of studies dealing with the home purchase and from discussions with realtors. The attributes were factor analyzed; the results indicated that the game seemed to cover factors such as quality of construction and size quite well, but not factors such as neighborhood and extras. One positive note may be interjected, though. To the extent that the focus of the study is the investigation of the process involved, even a somewhat incomplete framework such as the Bean Game seems to provide insight.

The game's structure can also be questioned on other grounds. The game was developed by consumer economists for pedagogical purposes in mind. The resulting normative structure may be bothersome to subject The budget is set by the bean level, and subjects are required to allocate all of the beans. The implied notion is to Bet the best house possible given the constraints. Some subjects, upon realizing that they were going to end up with a disappointing house in any event, might have chosen to spend less than the total number of beans on the house and used the remaining beans for other consumer goods. In other words, tradeoffs outside the housing decision may exist in the real world. Allowing couples to choose between (1) a 20-bean house and their deteriorating car and (2) a five-bean car and a 15-bean house would provide a richer context and allow the study of non-comparable alternatives.

The preliminary study reported here does indicate that the game environment can be used to investigate issues studied earlier through the use of more elaborate procedures. The Park (1982) study required a tremendous amount of effort on the part of the subjects and the investigators. Adding the collection of protocols to the use of the game would have made data collection much more difficult, but it still would have been simpler than the procedures used by Park. Further, the game environment makes it feasible for the subjects to go through the process individually before doing it jointly; dealing with actual consumers did not allow this luxury. As in the study using the information integration methodology used by Troutman and Shanteau (1989 the individual- then-joint process allowed the discovery of instances where the joint decisions were more extreme than either of the two individual decisions, indicating that the simple averaging model may not explain the joint decision-making process.

The bean allocations provide an easily quantifiable dependent measure not available in the Park (1982) and Corfman and Lehmann (1987) studies. Concessions could be seen on an attribute basis, allowing analysis of more than just a dichotomous (choice-based) measure.

REFERENCES

Bass, B. M. (1964), "Business Gaming for Organizational Research," Management Science, 10 (April), 545-556.

Cohen, Kalman J. and E. Rhenman (1961), "The Role of Management Games in Education and Research," Management Science, 7 (January), 131 -166.

Corfman, Kim P and Donald R Lehmann (1987), "Models of Cooperative Group Decision-Making and Relative Influence: An Experimental Investigation of Family Purchase Decisions, Journal of Consumer Research, 14 (June), 1-13.

Davis, Harry L., Stephen J. Hoch, E. K. Easton Ragsdale (1986), "An Anchoring and Adjustment Model of Spousal Predictions," Journal of Consumer Research, 13 (June), 25-37.

Haley, Jay (1962), "Family Experiments: A New Type of Experimentation," Family Process, 1, 265-293.

Nees, Danielle B. (1983), "Simulation: A Complementary Method for Research on Strategic Decision-making Processes," Strategic Management Journal, 4 (March-June), 175-185.

Park, C. Whan (1982), "Joint Decisions in Home Purchasing: A Muddling-Through Process," Journal of Consumer Research. 9 (September), 151 -162.

Park, C. Whan and Richard J. Lutz (1982), "Decision Plans and Consumer Choice Dynamics," Journal of Marketing Research, 19 (February), 108-115.

Rabin, Claire (1983), 'Towards the Use and Development of Games for Social Work Practice," Social Work, 13, 175 196.

Rowland, Kendreth M. and David M. Gardner (1973), "The Uses of Business Gaming in Education and Laboratory Research," Decision Sciences, 4 (April), 268-283.

Straus, M. A. (1966), "Laboratory Experimental Studies of Family Roles," Indian Journal of Social Research, 7 (April), 25-36.

Tedeschi, James T., Barry R. Schlenker, and Thomas V. Bonoma (1973), Conflict, Power and Games: The Experimental Study of Interpersonal Relations, Chicago: Aldine Publishing Co.

Troutman, C. Michael and James Shanteau (1988), "Information Integration in Husband-Wife

Decision Making About Health-Care Services,' Dyadic Decision Making, eds. David Brinberg and James Jaccard, New York: Springer-Verlag.

Yearns, Mary H. and Carolyn Manning (1984), The Housing Game: High Prices/Hard Choices, Iowa State University: Cooperative Extension Service.

 

INFORMATION INTEGRATION THEORY APPROACH TO HUSBAND WIFE DECISION MAKING

James Shanteau, Kansas State University

C. Michael Troutman, Charles, Charles and Associates

ABSTRACT -

The approach described in this presentation addresses husband-wife decision making by focusing on the relation between spouses' individual and collective judgments. The purpose is to illustrate a new method for examining how spouses make joint decisions. Before considering the proposed approach, a methodological review is offered of prior approaches to dyadic decision making. This is intended to show the need for an alternative framework. A new perspective is then suggested to examine how spouses make joint decisions. This method is illustrated in three experiments of couples' collective decisions about medical services.

BACKGROUND

Research on family decision making has traditionally focused on the influence that each spouse has on consumer choices (Engel, Blackwell, & Miniard, 1986). Although various research procedures have been used, the measurement of social power has proved difficult (Turk & Bell, 1972). Moreover, there is a fundamental question about the role of social power in couple decision making. A discussion between husband and wife, for example, may involve both power issues and an exchange of information (Park, 1982). Thus, these methods may have been blind to important processes in collective decision making.

When spouses interact to make a decision, the outcome seems simple--they either agree or disagree. Even a casual analysis of husband-wife interaction, however, indicates that matters are more complicated than that. Whether the spouses agree or disagree is not so important in the present view as the sharing of information. To understand how collective decisions are made, it is necessary to examine simultaneously the husband's and the wife's individual decision processes as well as the collective decision processes. In this way, it becomes possible to analyze in detail how joint decisions relate to individual decisions

ALTERNATIVE APPROACH

The proposed approach derives from Anderson's (1981, 1982) information integration theory (IIT). This theory assumes that there are two fundamental processes involved in making decisions. The first is the evaluation process by which psychological value is attached to information. The second process involves the integration of separate pieces of information into a unitary judgment. The former is reflected in the estimation of weight and scale values, while the latter is reflected in an algebraic integration model.

One advantage of IIT is that it can be applied at the level of a single subject/couple. This means that both the parameter estimates and the model form can be determined and compared for each individual and/or couple. As shown by Troutman & Shanteau (1989), this provides important advantages over previous approaches (also see Anderson & Armstrong, 1989).

The research described here appears to contradict Sheth's (1974) assertion that the area of couple decision making has been researched sufficiently so that a comprehensive theory can be developed. The present findings suggest that the critical cognitive component remains to be understood.

REFERENCES

Anderson, N. H. (1981), Foundations of information integration theory, New York: Academic Press.

Anderson, N. H. (1982), Methods of information integration theory, New York: Academic Press.

Anderson, N. H. & Armstrong, M. A. (1989), Cognitive theory, and methodology for studying martial interaction. In D. Brinberg & J. Jaccard (Eds.), Dyadic decision making, New York: Springer-Verlag.

Engel, J. F., Blackwell, R. D., & Miniard, P. W. (1986) Consumer behavior, (5th ed.), Chicago: Dryden Press.

Park, C. W. (1982), Joint decisions in home purchasing: A muddling-through process. Journal of Consumer Research, 9, 151-162.

Sheth, J. N. (1974), A theory of family buying decisions. In J. N. Sheth (Ed.), Models of buyer behavior: Conceptual, quantitative, and empirical, New York: Harper & Row.

Troutman, C. M., & Shanteau, 1. (1989), Information integration in husband-wife decision making about health care services. In D. Brinberg & J. Jaccard (Eds.), Dyadic decision making, New York: Springer-Verlag.

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

 

LONGITUDINAL METHODS FOR FAMILY CONSUMER RESEARCH

Alladi Venkatesh, University of California, Berkeley

INTRODUCTION

Longitudinal analysis is particularly suited for studies that are designed to assess the effects of variables over lime. The value of such research addresses issues relating to changes, impacts and cause-effect relationships (Nicosia 1965; Bucklin and Carrnan 1967; Eckland, 1968; Crider et al., 1973; Winer 1983). In spite of their appeal, longitudinal designs present several problems which need careful attention (Nicosia 1965; Crider, et. al., 1973). In general, the problems may be classified as those relating to (a) attrition of the sampling units; (b) the diminishing relevance of certain research issues over time, (c) response biases and (d) high cost. These problems are in addition to those faced in the standard cross-sectional survey designs. In this paper, a review of longitudinal methods as appropriate to household research will be presented. The Paper will also include an empirical study of U.S. families conducted during 1984-86 using CATI (Computer Assisted Telephone Interviewing) technique and time dairy data collection procedures.

A GENERAL OVERVIEW

Longitudinal studies may be classified in two ways, according to the nature of sampling scheme, or according to the data collection procedures.

Nature of the Sampling Scheme: Based on a survey of different behavioral and social science disciplines, we identify four types of studies: (a) medium-term (1-3 years), qualitative studies of a limited number (about 25-50) of families, (b) longterm (5-10 years), qualitative studies of a limited number of families, (c) multivariate, quantitative studies of a small sample of families (about 50-150), and (d) multivariate, quantitative studies of medium to large sample of families. Types (a) and (b) research are usually undertaken by anthropologists, medical sociologists, applied researchers in family therapy, and development psychologists. Types (c) and (d) studies are found among demographers, family sociologists, applied economists, and marketing researchers.

Data Collection Procedures: We have identified different studies based on data collection procedures. Longitudinal research can utilize primary data or secondary data. Some researchers would add a third category, namely, field experiments or quasi- experiments, to identify the collection of longitudinal context where some type of intervention is examined via a pre-post, or multiperiod data collection procedures or measures.

In this paper our major emphasis is on primary data collection as it pertains to household research. Since much of the family research in Economics and Demography uses secondary data, a small section of the paper will be devoted to it towards that end. s

SELECTED ISSUES IN FAMILY RESEARCH

An Overview of Longitudinal Family Studies: We find that different disciplines have different orientations in the way they conduct family studies. Some examples are given below.

Family Development research theorists use life cycle models for their research. Most recently, there has been an emphasis on using family life events as markers of family life change. In this type of research, the researchers are less concerned with representativeness of their sampling schemes and more interested in abnormal samples as the basis for theory generation in family development.

Marketing researchers use panels of families for studying consumption patterns. Panel studies of purchase behavior of grocery products have undergone revolutionary changes in the last ten years. The Berkeley panel studies of the sixties used extensive time diaries for recording weekly grocery shopping. In the last ten years the scanner data technology has completely altered the nature of this type of research. The methodological considerations in these two types of research are quite different and there are other interesting issues of contract that need to be reviewed.

Anthropological studies of family behavior are based on developing family ethnographies over a period of time. Because of the intense nature of these studies and the fact that most anthropological research is carried out in cultures other than that of the researcher, there are extensive cost constraints in conducting this type of research. There are also other research issues which do not surface in the standard survey type of research. Ethonographic research involves the use of informants, recording of oral histories, and pursuing of observational techniques. It is not uncommon for anthropologists to return to their sites to observe changes etc., and continue where they left off.

Researchers of time-budget studies perform extensive analysis of family time allocation patterns across different activities. Time-budget studies are carried out in different parts of the world and much comparative data are now available for performing research using a multi-cultural perspective. An important issue in time-budget studies refers to the data collection procedures and the validity of the reported data.

Some Research Issues: In contrast to individual oriented research, research on families presents some special kinds of problems. In typical family research, we are concerned with both structural (e.g. family composition) variables and individual family member variables. Specifically, in longitudinal studies, both the structural issue in family research relates to the definition of a continuing or longitudinal family, taking into account the fact that a family's composition may change from one period to another. The paper will give an example of the Reciprocal Rule Model as a way to dete

A general problem in longitudinal analysis is the attrition bias, that is, bias due to the dropping out of families from one wave to the next. While replacement of families is a solution, we now have procedures available to estimate biases due to attrition which utilize a combination of probit/OLS techniques.

In typical survey type situations family data are collected from a single member within the family (e.g. husband or wife). Because of the inherent biases in such reporting, researchers have been arguing for collecting data from both husband and wife. Such data are still not accurate because we now have bias from two sources instead of a single source. We will illustrate a method of data collection to minimize dual bias.

AN EMPIRICAL STUDY

For this paper, data collected from a national sample of 614 households, between 1984 and 1986, will be used to provide examples of the issues discussed above. The data pertain to the adoption and use of personal computers at home. The project was funded by the National Science Foundation.

(Interested parties should contact the author regarding references cited in this paper.)

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