CHILDREN'S INFLUENCE IN PURCHASE DECISIONS: A REVIEW AND CRITIQUE Tamara F.

Mangleburg, Virginia Tech ABSTRACT Based upon a review of the literature, children's influence in family purchase decisions was found to vary by product- and decision-related factors, as well as by parental, child and family characteristics. A number of problems associated with previous research were also identified, including problems with construct validity, overreliance on the survey method, and lack of theoretical explanations for observed patterns of influence. Finally, some possible avenues for future research were also discussed. The issue of children's influence in family purchase decisions is beginning to attract the attention of researchers of family buying processes. In the past, research on family decision-making tended to concentrate on examining variations in spousal influence; the role of the child was often overlooked (Davis 1976; Ferber 1973; McDonald 1980; Miller et al. 1982; Scanzoni 1980). However, there has recently been an increasing recognition of the child's importance in family purchase decisions. Specifically, a number of studies have found children to have at least some influence in decisions for a wide array of products. Therefore, it seems that if researchers wish to fully understand family decision-making, children should be included in studies of family consumer behavior. The purposes of this paper are to review previous research on children's influence in purchase decisions, and to evaluate the existing state of knowledge in this domain. Our goal is to point out some of the attendant strengths and weaknesses that characterize this area of research, and to offer some possible avenues for future research. The paper will be organized into three sections. In the first section, we will provide a brief overview of the substantive findings of previous studies. The review will be confined to examining studies which treat children's influence (attempts) or parental yielding as a dependent measure. The second section will be devoted to a critique of the existing literature. Finally, in the third section, we will discuss some possible directions for future research. OVERVIEW OF SUBSTANTIVE RESULTS In reviewing past research on children's influence, the results indicate that influence varies by a number of different factors. In this section, we will attempt to synthesize previous findings to discover the sources of variation in influence. [In addition to the substantive areas reviewed in this paper, the author initially had intended to include discussions of mass and personal communication effects. However, due to space constraints, these sections were deleted from the paper prior to submission. The reader is referred to Table 1 for a summary of studies which included these variables.] Product Type. One important source of variation in children's influence is product type. In general, children seem to have significant influence in product decisions for which they will be the primary consumer (Table 1). For example, children have been found to have substantial impact in decisions for breakfast cereals, snack foods, toys, children's clothes and school supplies. Children also influence decisions about family leisure time activities (such as vacations, movie attendance, eating out and cable TV subscriptions), although their influence is less in these decisions than in decisions for products far their own use. One factor that may partially explain these results is the level of the child's product involvement. Children are likely to view products for their own use as the most personally relevant. Leisure activities should also be child-involving, but to a lesser extent than products that the child uses frequently. In contrast to the significant role played by children in the child-related product decisions, children have less influence for products that are used by the entire family. This is particularly true when the family products involve substantial financial outlays. For example, children have been found to have little influence in decisions for cars, furniture, televisions and life insurance. Due to the financial risk associated with these family products, it appears that parents prefer to make these decisions without permitting the child to influence them. It is also likely that children perceive products such as furniture as having low personal relevance; therefore, they may not be motivated to influence these decisions. Intuitively, it seems likely that children's influence for any one product will

depend on an interaction between the child's product involvement and the financial risks associated with that product. Decision Stage. Another factor affecting the degree of children's influence in purchase decisions is the stage of the decision process. With one exception (Moschis and Mitchell (1986)), all of the studies examining children's influence across decision stages have used a three-stage model of the decision process. The three stages include problem recognition, information search and choice (Moschis and Mitchell (1986) also included an alternative evaluation stage). In general, children's influence is greatest in the problem recognition stage and declines significantly by the choice stage (Belch et al. 1985; Nelson 1978; Szybillo and Sosanie 1977). TABLE 1 RESEARCH ON CHILDREN'S INFLUENCE These results should be regarded with some skepticism, as the range of products over which this pattern has been examined is limited. Specifically, Nelson (1978) and Szybillo and Sosanie (1977) used family restaurants as products. Belch et al. (1985) used a wider variety of products; however, this study included a number of products for which children's influence was low overall (i.e., television, car, furniture and appliances). Additionally, the one other study employing decision stages did not specifically compare influence variations over stages (Moschis and Mitchell 1986). Subdecisions. In addition to product type and decision stage, children's influence has also been found to vary according to product subdecisions. The subdecisions investigated are adaptations of Davis's (1970, 1971) specific, as opposed to global, index of purchase decisions. The pattern emerging from these studies is that children's influence is lowest in the subdecisions of where to purchase (Belch et al. 1985; Jenkins 1979), gathering information (Darley and Lim 1986) and how much to spend (Belch et al. 1985; Darley and Lim 1986; Jenkins 1979; Nelson 1978; Szybillo and Sosanie 1977). Similarly, Foxman et al. (1989) found that both parents and children perceived that children had low influence in selecting price ranges. Children are more involved in subdecisions regarding color, make/model, and brand choices (Belch et al. 1985; Darley and Lim 1986; Jenkins 1979; Nelson 1978; Szybillo and Sosanie 1977). Given that parents are the primary socialization agents of children, it appears that they attempt to set bounds on children's influence through decreasing children's roles in the choice stage of the decision process, and in the subdecision of how much to spend. Limiting children's influence in these areas may be one way for the parent to teach the child responsibility and appropriate consumer behavior. It also seems that parents reserve more instrumental (i.e., allocation and scheduling) activities for themselves, and permit more child's influence in the more expressive subdecisions (i.e., color and model decisions). Children may lack the experience necessary to make informed decisions for instrumental activities. Parental Characteristics. A few studies have investigated how various parental attitudes affect children's influence in decision-making. For example, Berey and Pollay (1968) found mothers' child-centeredness to be inversely related to purchasing the child's favorite cereal. One reason for this may be that child-centered mothers are more concerned with the child's nutrition than are other mothers (Berey and Pollay 1968). Concern over nutrition may reflect these mothers' role perceptions of "what a good mother should do". Along this vein, Roberts et al. (1981) found that children had less influence in decisions when mothers were more traditional and conservative. It appears that parental characteristics, particularly those related to child rearing attitudes, may provide high explanatory power in examining children's influence. Child's Age. Another group of factors having an impact on a child's influence in purchase decisions are characteristics of the child. The most commonly investigated variable here is the child's age. Most studies have found that older children have significantly more influence than younger children (Atkin 1978; Darley and Lim 1986; Jenkins 1979; Moschis and Mitchell 1986; Nelson 1978; Ward and Wackman 1972). This result is due partly to older children's greater cognitive ability, as compared to younger children. Older children also have more experience with products and have learned more about consumer roles. These factors, in addition to developmental level, probably contribute to older children's greater influence in family decision-making.

Family Characteristics. A number of family demographic characteristics may also affect children's influence, although the results are more conflicting here; Some studies have found children's influence to be greater with increased family income (Jenkins 1979) or higher socio-economic status (Moschis and Mitchell 1986). However, Atkin (1978) and Ward and Wackman (1972) found no statistically significant effect for socio-economic status on children's influence attempts. It seems intuitive that children will have more influence in higher socio-economic status families, given that such families are likely to make more purchases than lower class families. However, the literature is not clear on this point. Another demographic variable that seems likely to affect children's influence is family size. Results here are mixed as well. Jenkins (1979) found children's influence to increase with family size; however, Ward and Wackman (1972) found no significant effect for number of children on children's influence attempts. Speculatively, one might expect family size to have a negative effect on any one child's influence, but that children's influence overall would increase. CONCEPTUAL AND METHODOLOGICAL CRITIQUE Based upon a review of the literature, it appears that children's influence varies by product and decision-related factors, as well as by parental, child and family characteristics. However, there are also a number of problems associated with research in this domain. In this section, we will outline some of the more salient conceptual and methodological limitations of previous research. Lack of Theoretical Explanation. One of the most glaring oversights in studying children's influence is the failure to provide conceptual justification for the observed patterns of influence. Previous research has been descriptive/relational and a-theoretical. Consequently, we may know that children's influence varies with a number of factors, but, without adequate theoretical development, we cannot answer "why" these variations occur. If we wish to understand family decision-making, the need for sound theorizing cannot be over-emphasized. Construct Validity. In addition to a lack of theory, another problem characteristic of research on children's influence has been the lack of attention paid to construct validity. The most noteworthy example of this is the failure to adequately define 'influence'. As Rossiter (1978) notes, there are two aspects to influence, an active and a passive dimension. In this instance, passive influence is characterized by other family members taking the child into account when making a purchase decision. as opposed to the child directly influencing the decision itself. In spite of Rossiter's clarification, most studies have not distinguished between active and passive influence. For example, Jenkins (1979) notes that "the definition of 'influence' varied from one [respondent] to another. Some perceived only the 'active' dimension ... while others perceived the word to encompass both the 'active' and 'passive' dimensions." The fact that respondents can attribute varying definitions to a term is due to inadequate conceptual definitions in the first place. If the researcher is unclear as to the meaning of a construct, it is no surprise that respondents are unclear as well. It is noteworthy that most studies reviewed here failed to define "influence" conceptually. The net result of this oversight is that respondents assign their own meanings to the concept. As a result, their definitions may or may not be congruent with the researcher's notion of what a term means. In any event, construct validity is a serious issue under these conditions. The failure of most studies to carry out reliability analyses also supports this conclusion (Table 2). Measures. Another issue related to children's influence is the type of measures used to assess influence. Studies have employed a diversity of operationalizations for influence (Table 3). With one-exception (Jenkins (1979) used a constant sum scale), studies have used 3, 5, 6 or 7 point Likert scales to measure influence. Some studies have asked respondents to rate influence separately for family members (Belch et al. 1985; Darley and Lim 1986; Jenkins 1979; Nelson 1978; Roberts et al. 1981), whereas others have included all family members on a single scale (Foxman and Tansuhaj 1988; Foxman et al. 1989; Moschis and Mitchell 1986; Szybillo and Sosanie 1977). In this latter approach, respondents are essentially asked to think about a purchase decision and aggregate the influence of all family members into one global level evaluation in order to respond to the scale The former approach of having respondents rate influence separately for family members has the advantages of (1) simplifying the cognitive tasks of respondents and (2) collecting data at a more disaggregate level. It is likely that significant information about influence is lost when respondents are asked to make global level evaluations.

Variation by Respondent. Results on children's influence also vary according to who is the respondent. Many studies have used only parents as respondents (Table 2); therefore, these studies can only provide information about parental perceptions of children's influence. These perceptions may or may not be accurate. Those studies that have included children, as well as parents, as respondents have generally found that children attribute more influence to themselves than their parents attribute to them (Table 1). Differences in perceptions of influence may be due to perceptual biases and/or social desirability effects. Additionally, perceptual divergences may also be a methodological artifact. Surveys, in this instance, may reflect more subjective assessments, whereas observational studies may be more objective (i.e., an observer who is removed from the process independently evaluates "influence"). Clearly, the issue revolves around the interests of the researcher, whether one wishes to focus on subjective perceptions or behavior. Whatever the focus of the study, however, children should be included as they are the relevant unit of analysis in most studies of children's influence. Method of Data Collection. Another issue related to research on children's influence is the method of data collection. All studies have used mono-method approaches to studying children's influence (Table 2). [Berey and Pollay (1968) used different methods to assess different variables; only one method was used per variable.] With few exceptions (Atkin 1978; Brody et al. 1981), the method of choice has been the survey. The use of a single method within a research domain is problematic in that the results found may be confounded by methodological artifacts (McGrath 1982). The rival explanation that results are due to research method cannot be discounted when only a single method is used. Thus, the associations found between children's influence and various factors may be due to the use of surveys rather than to any "true" association. The observational study by Atkin (1978) mitigates against this conclusion; however, more observational studies need to be done before we can have confidence in the findings. The use of surveys has a number of advantages, such as time and cost reductions, and the potential for large sample sizes. However, there are also a number of problems associated with their use. As previously mentioned, surveys may be more inter-subjectively biased than are observational studies. In addition, the use of surveys assumes that respondents have adequate recall of phenomenon and can report this information (i.e., they are capable of performing the "mental algebra" necessary to respond to the instrument (Nisbett and Wilson 1977)). Finally, the use of surveys is problematic in dealing with young children who generally lack the cognitive ability to respond to test items. Although it is possible to circumvent this problem by using interviews, the usual approach has been to exclude children as respondents and use only parents. Thus, in some situations, other techniques such as interviews or observation may be more appropriate. TABLE 2 CHILDREN'S INFLUENCE: RELIABILITY, VALIDITY AND SAMPLE CHARACTERISTICS Statistical Techniques. A final consideration in this domain of research is the method of data analysis. Most of the studies have used chi-square analysis, ANOVA or MANOVA to test various associations. One problem that may occur in using these techniques is violation of the nonindependence assumption. The techniques are especially sensitive to this violation. When multiple respondents from the same family are used, one would expect their responses to be correlated with each other, and therefore, nonindependent. Yet studies examining divergences in family member perceptions use analyses of variance on these nonindependent samples in order to assess differences. It is not surprising that these studies find highly divergent perceptions as the type 1 error rate is seriously inflated when the nonindependence assumption is violated. Additionally, many studies test numerous relationships within the context of a single sample and study, while failing to adjust the alpha level accordingly. When one fails to adjust alpha in response to numerous tests, significant results may be due partly to chance. Therefore, it is small wonder that five or seven results are significant when the researcher has tested thirty or more relationships, a common occurrence in research on children's influence. Summary. In summary, there are a number of problems associated with research on children's influence in purchase decisions. Inadequate conceptual definitions and lack of reliability raise serious construct validity concerns. Children are often excluded from the analysis even though children's influence is the subject of study. The use of a single method, usually surveys, gives rise to the rival hypotheses that results are method related. Additionally, surveys have a number of limitations that are especially problematic in this domain, including the

inability of young children to respond to test instruments and potential subjectivity biases. Finally, this research domain is hampered by inflated alpha levels; consequently, some associations found may be due to chance. It appears that many of these problems stem from a lack of sound theorizing (i.e., problems with construct validity, reliability, and inflated alpha levels due to testing numerous relationships instead of a few theoretically relevant relations). Others are more methodological in orientation (i.e., exclusion of children, use of single methods and surveys, misuse of statistical techniques). In conclusion, research on children's influence needs to be improved on both theoretical and methodological grounds. DIRECTIONS FOR FUTURE RESEARCH Although there are a number of problems associated with previous research, there are also a number of possible means of addressing the issues raised. Our goal in this section is to highlight some possible directions for future research, in order to build on past work and to increase our understanding of children's influence. One means of contributing to our knowledge of children's influence is to incorporate stronger theoretical explanations in our research. There are a number-of potential candidates in this regard. Within marketing, Corfman and Lehmann (1987) developed a model of spousal decision-making that could be expanded to include children. There are also a number of sociological approaches to studying the family that may also prove fruitful in this respect. Role theory, particularly the age-graded approach of Parsons and Bales (1955) and the more dynamic role orientations of symbolic interactionists, seems especially promising. Social learning theory may also offer high explanatory power in some instances. Another direction for future research is concerned with issues related to construct validity. Clearly, construct validity will continue to be problematic until we conceptually define "influence". Rossiter's (1978) distinction between passive versus active influence seems to be a particularly useful one. Researchers must decide what dimension of influence they wish to probe and make this distinction clear to respondents. It also seems likely that the type of influence characterizing a purchase decision will depend in part on product type. Specifically, one might expect children's influence to be active in child-related product decisions, and passive in the more risky family-related product decisions. Making the conceptual distinction between active and passive influence will allow us to test these, and similar, propositions. TABLE 3 CHILDREN'S INFLUENCE: MEASURES AND SCALES Another promising avenue for research on children's influence is further investigation of the effects of parental and child characteristics on influence. Up to this point, these variables have been given only a cursory treatment in the literature. However, one would expect these factors to be of primary importance in studying children's influence. Past studies on parental characteristics point to the potential contribution of parental attitudes in explaining children's influence. It seems likely that parental attitudes related to child-rearing practices would be especially critical, particularly for younger children. These sorts of relationships should be further clarified and developed. Similarly, for child characteristics, there is a need to go beyond merely examining the child's age, and to include other theoretically relevant variables, such as the child's product involvement and his/her attitudes towards parents. The explanatory power of many of these variables has remained virtually untapped; clearly, more work needs to be done in these areas. Given that most of the studies reviewed here used surveys to investigate children's influence, another means of improving our understanding of the phenomenon is to make greater use of observational studies. This is not to say that surveys should be abandoned, rather that the limitations of this method can be compensated for by observation, and vice versa. Observational studies may reduce subjectivity biases, and offer additional insights into how families actually make decisions. In any event, multiple methods are desirable in and of themselves in that they permit increased confidence in findings that are consistent across methods. An additional means for improving our understanding of children's influence is to make more appropriate use of statistical techniques. As previously noted, inflated alpha levels have been problematic in past research on children's influence. When testing numerous relationships, researchers should control the experiment-wise error

rate to avoid capitalizing on chance. If family members' responses are expected to be positively correlated, multivariate analyses which treat each member's responses as repeated measures can be employed. If the researcher is interested in assessing divergences in family members' perceptions, it may be appropriate to test at a more stringent alpha level, as non-independence tends to result in more liberal tests in this situation (see Judd and Kenny (1981) for a more thorough discussion of independence issues). Finally, past research on children's influence has tended to use the nuclear family as the normative criterion of study. Only one study (Darley and Lim 1986) has examined children's influence under different family forms. This is surprising given a recent study's estimate that 60% of today's children will live in a single-parent household sometime during their childhood (Norton and Glick 1986). What implications do alternative family forms have for children's influence? By failing to consider different family structures, we have in essence confined our analyses to one subset of children, namely those from intact households. CONCLUSION In conclusion, previous research on children's influence has been a-theoretical and empirical. Based upon a review of the literature, children's influence was found to vary with a number of factors. However, due to the limitations associated with past studies, we cannot explain why these patterns of influence exist. It appears that research on children's influence needs to be improved on both theoretical and methodological grounds. Until this occurs, our understanding of the phenomenon will continue to be incomplete and fragmentary. It is hoped that the proposals offered here will serve to stimulate future research and debate in this area. REFERENCES Atkin, Charles K. (1978), "Observation of Parent-Child Interaction in Supermarket Decision-Making," Journal of Marketing, 42 (October), 41-5. Belch, George, Michael A. Belch, and Gayle Ceresino (1985), "Parental and Teenage Influences in Family Decision Making," Journal of Business Research, 13 (April), 163-76. Berey, Lewis A. and Richard W. Pollay (1968), "The Influencing Role of the Child in Family Decision Making," Journal of Marketing Research, 5 (February), 70-2. Brody, Gene H., Zolinda Stoneman, Scott Lane and Alice K. Sanders (1981), "Television Food Commercials Aimed at Children, Family Grocery Shopping and Mother-Child Interactions," Family Relations, 30, 435-9. Corfman, Kim P. and Donald R. Lehmann (1987), "Models of Cooperative Group Decision-Making and Relative Influence," Journal of Consumer Research, 14 (June), 1-13. Darley, William F. and Jeen-Su Lim (1986), "Family Decision Making in Leisure-Time Activities: An Exploratory Analysis of the Impact of Locus of Control, (Child Age Influence Factor and Parental Type of Perceived Child Influence," in Advances in Consumer Research, R. J. Lutz (ed), 13, Ann Arbor, MI: Association for Consumer Research, 370-4. Davis, Harry L. (1970), "Dimensions of Marital Roles in Consumer Decision Making," Journal of Marketing Research, 7 (May), 168-77. Davis, Harry L. (1970), "Measurement of Husband-Wife Influence in Consumer Purchase Decisions," Journal of Marketing Research, 8 (August), 305-12. Davis, Harry L. (1976), "Decision Making within the Household, Journal of Consumer Research, 2 (March), 24160. Ferber, Robert (1973), "Family Decision Making and Economic Behavior: A Review," Family Economic Behavior: Problems and Prospects, in E. B. Sheldon, ed., Philadelphia: J. B. Lippincott.

Foxman, Ellen and Patriya Tansuhaj (1988), "Adolescents' and Mothers' Perceptions of Relative Influence in Family Purchase Decisions: Patterns of Agreement and Disagreement," in Advances in Consumer Research, M. Houston (ed), 15, Ann Arbor, MI: Association for Consumer Research, 449-53. Foxman, Ellen, Patriya Tansuhaj and Karin Ekstrom (1989), "Family Members' Perceptions of Adolescents' Influence in Family Decision Making," Journal of Consumer Research, 15 (March), 482-91. Jenkins, Roger L. (1979), "The Influence of Children in Family Decision-Making: Parents' Perceptions," in Advances in Consumer Research, 6, Ann Arbor, MI: Association for Consumer Research, 413-18. Judd, Charles M. and David A. Kenny (1981), Estimating the Effects of Social Interventions, New York: Cambridge University Press. McDonald, Gerald (1980), "Family Power: The Assessment of a Decade of Research, 1970-1979," Journal of Marriage and the Family, 42 (November), 841-54. McGrath, Joseph E. (1982), "Dilemmatics: The Study of Research Choices and Dilemmas," in Judgment Calls in Research, J. E. McGrath, J. Martin and R. Kulka (eds), Beverly Hills, CA: Sage Publications, Inc. Mehrotra, Sunil and Sandra Torges (1977), "Determinants of Children's Influence on Mothers' Buying Behavior," in Advances in Consumer Research, W. D.Perreault (ed), 4, Ann Arbor, MI: Association for Consumer Research, 56-60. Miller, B. C., B. C. Rollins and D. L Thomas (1982), "On Methods of Studying Marriages and Families," Journal of Marriage and the Family, 44, 851 -73. Moschis, George P. and Linda G. Mitchell (1986), "Television Advertising and Interpersonal Influences on Teenagers' Participation in Family Consumer Decisions," Advances in Consumer Research, R. J. Lutz (ed), 13, Ann Arbor, MI: Association for Consumer Research, 181-6. Nelson, James E. (1979), "Children as Information Sources in Family Decision to Eat Out," in Advances in Consumer Research, W. L. Wilkie (ed), 6, Ann Arbor, MI: Association for Consumer Research, 419-23. Nisbett, Richard E. and Timothy D. Wilson (1977), "Telling More than We Can Know: Verbal Reports on Mental Processes," Psychological Review, 84 (May), 231-59. Norton, Arthur J, and Paul C. Glick (1986), "One Parent Families: A Social and Economic Profile," Family Relations, 35, 9-17. Parsons, Talcott and Robert Bales (1955), Family, Socialization and Interaction Process, Glencoe: The Free Press. Roberts, Mary Lou, Lawrence H. Wortzel and Robert L. Berkeley (1981), "Mothers' Attitudes and Perceptions of Children's Influence and Their Effect on Family Consumption," Advances in Consumer Research, J. C. Olson (ed), 8, Ann Arbor, MI: Association for Consumer Research, 730-5. Rossiter, John R. (1978), "Children's Consumer Research: A Call for Rigor," in Advances in Consumer Research, W. L. Wilkie (ed), 6, Ann Arbor, MI: Association for Consumer Research, 424-6. Scanzoni, John (1980), "Social Processes and Power in Families," in Contemporary Theories about the Family, W. Burr, R. Hill, I. Nye and I. Reiss (eds), New York: The Free Press. Szybillo, George J. and Arlene K. Sosanie (1977), "Family Decision Making: Husband, Wife and Children," in Advances in Consumer Research, W. D. Perreault (ed), 4, Ann Arbor, MI: Association for Consumer Research, 46-9.

Ward, Scott and Daniel B. Wackman (1972), "Children's Purchase Influence Attempts and Parental Yielding," Journal of Marketing Research, 9 (November), 316-19.