You are on page 1of 9

Measuring influence in the family decision making process using an observational method

Choi Lee, Christina Kwai; Marshall, Roger. Qualitative Market Research 1. 2 (1998): 88-98.
Turn on hit highlighting for speaking browsers
Hide highlighting
Abstract (summary)
Translate [unavailable for this document]
Many consumer behavior researchers have concluded that using self reports in the measurement
of influenceand family decision-making may provide data which is not reflective of
actual influence among family members. This suggests the need for different methods of
obtaining influence data. This paper reports the developments and preliminary application of an
observed influence scale; a scale developed to measure actualinfluence using content analysis of
videotaped family interactions. This observational method determines the relative influence of family
members over three stages of the decision-making process.
1. Introduction
Family researchers have often commented on the fallibility of self-designating scales as a measure
of influence. Although observational methods are frequently used by sociologists and psychologists in
their studies of group behaviour and interactions, the practice is uncommon in marketing.
Research has provided considerable evidence that most information on family decision
making gathered from self reports is very susceptible to the biases of family members (Davis and
Rigaux, 1974; Douglas and Wind, 1978; Granbois and Willett, 1970; Shuptrine and Samuelson, 1976).
Furthermore, respondents are only providing perceptions of influence which may not conform to
reality (Corfman, 1990). An observational approach provides a closer representation of the
actual decision process than respondents' perceptions of that process (Douglas and Wind, 1978).
The primary objective of this research is to describe the development and initial testing of an
observedinfluence scale (OIS) to measure the influence of family members in a purchase situation. The
observed data are derived from a content analysis of taped family interactions. In order to provide a
preliminary test of validity of the OIS, this technique is applied to Singapore and New Zealand
families. Traditional stereotypes of the Chinese and European cultures suggest some differences in
the influence exerted by parents and children.
The paper begins with a literature review on issues surrounding the measurement of influence, such as,
the definition of influence, the number of stages to be taken into account during the decision process,
the inclusion of children in the research and the method of using observation to assess influence. The
research approach used in this exploratory study is then described, followed by the results which
implicate that the instrument has both face and predictive validity.

2. Literature review
2.1 Definition of influence
A fundamental problem still confuses research in this area, that of defining influence (Kohli and
Zaltman, 1988; Mangleburg, 1990). In particular, the distinction between direct and
indirect influence is problematic. Directinfluence represents an "active role based directly on
the decision maker's own needs, and indirect influencerepresents a passive role in which
the decision maker takes another family member's needs indirectly into account" (Rossiter 1978, p.
425). Rossiter, although not offering a specific solution, condemned much of the research related to the
role of children's influence in family decision making conducted to date on the grounds that this
distinction has not been made. Ultimately any individual's true motivation for making a particular
choice is known only to him or herself - the level of research sophistication required to discern
accurately the difference between direct and indirect influence is daunting.
In practice it is often necessary to determine both types of influence in order to both direct an
advertising message and to determine the most suitable copy. In this paper, active and
passive influence are differentiated through determining both the influence attempts and the outcome
of these influence attempts during a decision making process.
2.2 Stages in the decision making process
It is well documented that the decision making process follows a number of stages, although it is not
clear exactly how many stages an individual goes through before making a final choice. Research
using self-reports have included three stages (Davis and Rigaux, 1974), four stages (Mochis and
Mitchell, 1986) and nine stages (Woodside and Motes, 1979). The use of questionnaires to
determine influence across the decision makingprocess assumes that individuals and groups go through
these stages sequentially, so that the individual needs to go through an information search before
evaluating alternatives. Although stages in the decision making process are often seen as being linear,
i.e. problem recognition, information search, alternative evaluation and choice, a study of the videotaped observations in this study showed this to be unrealistic. Group development theorists (for
example, Fisher, 1970; Gersick, 1988; Scheidel and Crowell, 1964) contend that the decision
making process can proceed in iterative cycles and not necessarily in a linear order. An observational
approach can help clarify this.
Three stages were considered here - configuration, negotiation and outcome. Influence during the
problem recognition stage was not considered in this research because the scenario-based problem
presented to the family makes this stage irrelevant. Bales and Strodtbeck (1951) proposed that groups
move through three stages toward decision goals: orientation, evaluation and control. The term

"configuration" is similar to the term "orientation", where the group members gear themselves to
the decision situation and decide on the "rules of the game". In the present context, the term seems well
suited to the stage in the family decisionwhere the family decides the rules of the game, and thus sets
the scene for the consequent discussion. Configuration was deemed to have taken place if discussion or
activity occurred that was determinant to the way the decision was made. For example, one member
could seize the instrument and suggest that restaurants should first be listed before they are rated. It is
not difficult to imagine a situation where one family member configures the decision so that he or she
obtains greater influence; perhaps a father could assign his wife a recording or secretarial role that
would, to a point, keep her out of the discussion.
During the negotiation stage, the family members consider various alternatives suggested, discuss the
pros and cons of each suggestion and negotiate to either add it to the consideration set or ignore it.
Negotiation can involve a variety of strategies, such as bargaining, expertise, legitimate, emotive or
coalitions.
The outcome of the decision process represents the final choice of the family as a whole, the
culmination of the two earlier stages where family members attempt to exert their influence. Although
these stages are described sequentially, it must be noted that the stages may not occur in sequence in
the decision process. Although, in the present research, configuration usually occurred at the beginning
of the interaction, frequent occasions arose where a re-configuration took place during the course of
the discussion - indeed, some families had no formal, identifiable configuration at all.
In the research reported here, the number of stages was not determined until a detailed study of the
videotapes had been carried out to determine the decision making pattern used by the families.
Although previous researchers have often estimated stages by simply dividing the decision process into
time phases, this was not possible in this instance because each family structured its decision so
differently.
2.3 Including children in the research process
Until recently, research on families has concentrated on husband-wife dyads. Many studies only
interview one partner, usually the wife, and her responses are taken to represent that of the family's
(Davis, 1971). Childrenand their influence on family decision making have been largely ignored.
While some recent studies have included children, these have predominantly been studies of triads
composed of a father, a mother and one adolescent child (Foxman et al., 1989; Kim and Lee, 1997).
Our focus is on the development of an observational scale which accounts for the influence of
two children in family decisions. As the objective is to develop an observational scale, the product
category is controlled - one product category, family restaurants, was chosen as a subject of interest.

This choice is based on previous findings which note that children do have some influence in this
product selection process (Nelson, 1979; Syzbillo and Sosanie, 1976).
2.4 Measuring influence using observation
The purpose of taking observations is to get a more valid measure of actual influence than can be
gathered using self-declared scales (Kenkel, 1963; Larzelere and Klein, 1987; Olson, 1969). To further
enhance the validity of the observations, interviews should be conducted in the respondents' own
homes where thedecision can take place in a naturalistic setting (Douglas, 1983; Gordon, 1997).
Videotapes can capture immediate responses and family interactions in these situations, the researcher
can then later rate the family interactions in terms of influence (Callan and Noller, 1986; Cherulnik,
1983; Kerlinger, 1986).
Speaking to someone face-to-face involves behaviour which is both visual and paralinguistic; this
behaviour becomes an integral part of the message. The visible non-verbal behaviour which can be
analysed includes smiles, frowns, eyebrow raises, gestures such as pointing a finger or waving, and
postures such as leaning forward or away from the message recipient. There are also patterns of
spacing and gaze. On the paralinguistic side, verbal behaviours include speed of talking, pitch,
loudness, and voice quality (Hulbert and Capon, 1972; Knapp, 1978). Not all of these indicators could
be utilized here due to restraints imposed by filming in subjects' homes, yet sufficient detail of the
family interactions were gathered to make useful judgements about influence.
Content analysis is not only a method of analysis but also is a method of observation. Instead of
observing people's behaviour directly, by responses to scales or by interviewing them, the researcher
takes as raw data the communication that people produce, and then seeks answers to questions by
analysing that. Content analysis is systematic and objective, with the aim of quantifying elements of
interest in the observation data (Kerlinger, 1964; Kolbe and Burnett, 1991; Rust, 1993). The attempt to
put numbers into our judgements adds rigour to qualitative research.
The types of behaviour that are of interest here are those that indicate an individual is attempting to
gaininfluence or to dominate others in the group during the process of making a family decision.
Paralinguistic indicators of dominance suggested in previous work include the amount of talk
(Sorrentini and Boutillier, 1975), as well as pitch and loudness of voice (Apple et al., 1979). Bodily
movement, positions and postures also convey arousal, and differences in dominance may be inferred
across individuals (Harper, 1985). Facial expressions have long been recognized as signals of
dominance and submissiveness (Ekman, 1982). For example, Zivin (1975) described a "win" face as
consisting of a raised brow, wide open eyes, firmly jutting neck posture, and a raised chin, whereas a
"loss" face consists of a furrowed brow, squinted eyes, retracted neck, and lowered chin. In this

research, an overall measurement of influence attempts was gained by judging facial animation, body
attitude, voice pitch and voice loudness, mutual gaze and time spent talking.
The outcome of these influence attempts was assessed by the number of suggestions made by each
individual which were accepted. This reasoning is consistent with previous research where Kenkel
(1963) measured a spouse's power by the proportion of items "purchased" from an imaginary gift of
$300 that was initially suggested by that person; and Strodtbeck (1951) also
measured decision outcomes as a measure of power by noting who won a series of revealed differences
during discussions. Corfman (1989), in her research about measuring relative influence, suggests that
outcome is quite different from influence and that this distinction should be made in
an influence measurement scale.
Hence, the first six items in the coding sheet (facial expression, body attitude, voice pitch, voice
loudness, mutual gaze and time spent talking) measure influence attempts. The seventh item, the
number of suggestions accepted, measures decision power or outcome. Table I describes these items
and notes the literature supporting them as indicators of influence.
3. Research approach
3.1 Research objective
3.2 Data collection
Data were collected from videotaped observations of family interactions triggered by a simulated
familydecision. The experiment was conducted in the respondents' own homes and involved 89 New
Zealand families of European ethnic origin, and 24 Singaporean Chinese families. All respondent
families were nuclear families with two adolescent children between the ages of 12 and 19 years.
Families were selected from a pool of participants drawn from four high schools around the Auckland
area in New Zealand, and three in Singapore.
Each family in the study was presented with the chance to win $150 to spend at a family restaurant. In
the interaction situation the scenario was presented that the family had already won the prize money,
and were then required to determine the family's choice of five favoured venues.
The initial experiment, including the scale development and testing, was conducted in New Zealand. A
second experiment duplicated the first, but all instruments were translated into Chinese (using backtranslation to validate the original translation) and applied in Singapore. Families in Singapore had the
choice of either language, depending on the language spoken at home.

Two researchers were involved in the data collection process. After introductions to the family
members on arrival at the subject's home, one researcher was involved in setting up the video camera
(a relatively small camera to avoid intrusion) while the other researcher spoke to the family to avoid
attention on the camera. The camera was set up at a corner of the room at some distance away from
where the family was seated, usually around the dining table. Other attempts to detract the family
members from the camera included not mentioning when the camera was turned on, getting all family
members to fill in two lengthy questionnaires before the interaction activity and not having anyone
behind the camera or in the room during the family interaction. One of the questionnaires measured
interest in eating out as a family and the other measured their sex-role orientation to determine if the
family is traditional or contemporary. These are used in another part of the study. The family's
demographic details were filled in after the videotaped interaction was completed.
A test was conducted with a random sample of 25 respondent families to determine the realism of the
scenario situation presented. A three-item, five-point Likert scale produced a mean of 1.96, which
represents agreement that the situation was realistic (Cronbach's Alpha = 0.79).
3.3 Content analysis of videotapes
The coding sheet developed to analyse the tapes consists of the seven items previously noted (Table I).
Three trained, independent judges were required to listen to the oral language and observe the body
language of the participants and note the extent to which these factors indicated an attempt
to influence other family members in the decision process. The judges were aware that the objective of
the study was to determine the amount of influence each member displayed relative to the rest of the
family members, and that the non-verbal and verbal communication was deemed to
determine influence. The four specific aspects that were based on the subjective judgements of the
coders were body attitude, facial expression, voice pitch and voice loudness. These aspects were each
summed to a single score drawn from a seven-point scale for each of these four aspects, where 1 =
extremely not influential, and 7 = extremely influential.
The other three items in the coding sheet were more objectively determined as these
involved makingfrequency counts. These three aspects reflected an effort on the part of the observed
individual to gaininfluence or dominance. The first of these was mutual gazing. Judges noted the
number of times each member held the gaze of another family member. The second aspect was the
amount of time each member spoke during the decision process. This was measured by using a
stopwatch. The number of minutes and seconds were noted and later converted to a proportion of the
total discussion time. Finally, the restaurants actually selected as family choices were noted against the
name of the person who made the original suggestion. The fact that the suggestion of a particular
individual is accepted by the group was taken as an indication of theirinfluence within the group. This

aspect represented the decision outcome of the group decision makingprocess. There is an element of
judgement required to determine whether a suggestion adopted by the group is representative of
the influence of an individual or is coincidentally the choice of every member in the group. There is no
way that this can be distinguished through observation; however, the extent to which all three judges
agree may indicate that the observation is accurate.
3.4 Constructing the observed influence scale (OIS)
The objective behind the influence measure is to allow a comparison of the observed influence of
family role players (fathers, mothers, children) within each family. With this measure it is possible to
make a judgement of the influence of, say, New Zealand children in relation to their parents compared
to Singaporean children in relation to their parents.
To achieve this comparative measure, the scores made by judges for each of the seven items used were
changed to proportions. Thus, for instance, the raw score for the number of times a father mutualgazed with other members of his family is not significant in itself, because the family may normally
mutually gaze with each other far more or less than some other families. If the number of times a father
shares a mutual gaze with any member of his family is expressed as a proportion of the total number of
mutual gazes exchanged by the family during the decision period, however, then this proportionalised
score can readily be compared to other fathers in the sample to gauge their relative attempts
to influence through eye-contact. The same principle applies to the length of time an individual talks
within a family decision, and even to the number of final family choices that emanated from an
individual role player. For example, if three out of the total five restaurant choices were selected by the
father, and two were selected by the mother, the proportion accorded to the father is 60/100, to the
mother 40/100, and zero for each of the children.
The first four items of coding categories that had been judged on a seven-point scale were converted to
a 100-point scale. Thus, for instance, a mother's body attitude is judged for the influence displayed
relative to theinfluence displayed by all family members. The measurement is a comment on the
interest displayed by a mother in relation to other family members, that can be compared to other
mothers in relationship to their families.
This device, of proportionalising the raw influence scores, also allows the diverse measurements in the
scale to be combined more easily. Each of the seven categories, after this transformation, can be added
and treated as part of a seven-item scale of observed influence.
3.5 Coding the decision stages

Two problems emerged when judging the parts of the interaction that comprise the configuration stage.
The first is machine dependency. Identifying configuration when it occurs at the beginning of a tape
presents no problem, but repeatedly identifying a small section that occurs in the body of the tape
becomes a crude process because of the inaccuracies in the tape counter mechanisms. This problem
would be significant during the coding process of these small sections, where a one-second difference
could result in a large difference in proportionalised judgements. The second problem relates to the
difficulty of this particular judgement process itself. The latter problem can be minimized by careful
definition of exactly what configuration means. Bearing these two problems in mind, it was determined
that the most appropriate method of determining the stages was for the researcher and a senior
colleague to determine judgementally the configuration stage of thedecision process by observing the
interactions on the videotape in open conference.
4. Results
4.1
4.2
4.3 Empirical test
5. Discussion and conclusion
The study takes a novel approach to measure influence in family purchase decisions. There is support
for face and predictive validity, demonstrating recorded observations are a powerful tool in the study
of group behaviour.
Content validity is achieved by using categories which are based on the literature, as well as an
assessment of these categories on marketing academics. Efforts to ensure reliability were achieved by
employing three well-trained judges with strong cultural affinity in their respective cultures; significant
and strong inter-judge reliability was obtained. Furthermore, the items in the OIS were found to be
internally consistent and the factor analysis provided support with two factors; one
measuring influence attempts, the other measuring outcome.
The scale was applied across two cultures to offer further preliminary support of validity. Although not
all stages in the decision making are significantly different, an examination of the mean values reveals
that theinfluence structure is in the right direction and for those stages which are significant, are as
predicted.
5.1 Research directions

Further opportunities for the scale's application are legion. Indeed, the sample used here to develop the
scale will lend itself to an analysis of patterns of family influence over, inter alia, social classes, the
working status of mothers and sex-role orientations of family members. The opportunity exists to use
similar techniques to consolidate and test the generalisability of the rather fragmented literature on
family influence patterns through widening the sample base and considering other decision topics, and
thus to develop the body of theoretical knowledge. Within a cross-cultural context, the sample
restriction of this work is useful as a measure of control is exerted through having the family type
restricted to nuclear families of two parents and two children. However, in New Zealand (and many
other Western countries), different household structures are emerging with single persons and other
non-traditional forms. In Singapore, too, there are still many extended families, where parents live with
their married children and assist them with the running of their households. In all these cases, it is hard
to generalise the findings from the present study too far, so more empirical work is needed.
Another avenue for future research lies in identifying the non-verbal communication cues which best
predictinfluence in a group situation. In this research the six items were combined to provide one score
of relativeinfluence, and one item to determine influence outcome. There is scope to determine which
of the six is the best predictor of influence attempts for the different cultures so that researchers need
only concentrate on one or two particular non-verbal communication cues. Furthermore, each item in
this research was given an equal weighting. This is acceptable in this preliminary study, but may not be
so when comparing other cultures. For example those cultures where children show respect by not
looking at their elders when speaking, but may exert influence using other means of non-verbal
communication. In this case, a lower weighting should be given to mutual gazing.
A further weakness of the research is an odd reflection of one of its greatest strengths - the painstaking
in-home videotaping of the interactions. Although highly desirable in the theory-building stage, a
faster, easier, more commercially viable method would have immense value to market researchers.
Until such a scale has been provided, however, the OIS developed here could provide as useful a tool
for practitioners as for academics.

You might also like