Peer Group Influence on Urban Preadolescents' Attitudes Toward Material Possessions: Social Status Benefits of Material Possessions

Abstract

This study explores peer influence on urban preadolescents' perceptions of social status benefits of material possessions. A longitudinal design is used. Natural, interaction-based peer groups are identified through the Social Cognitive Map procedure. Findings indicate that high-status rather than low-status peers in a group are influential on individuals. Strong influence of high-status peers is observed in both boys' and girls' groups. High-status peers are particularly influential on low-status individuals in girls' groups and on high-status individuals in boys' groups. Additionally, high-status peers' influence is stronger on African Americans than on Hispanic Americans and tends to be stronger on Hispanic Americans than on White Americans. These findings imply that special attention should be given to high-status youth in groups who highly endorse social benefits of material possessions. Characteristics of the target youth (e.g., gender, ethnicity and individual status) should be considered in future efforts for reducing the pervasiveness of materialism.

Materialism has negative impacts on individuals' well-being and social adaptedness (Kasser and Ahuvia 2002; Kasser and Ryan 1993). As concerns grow over rising levels of materialism among youth (Goldberg et al. 2003; Schor 2004), understanding the ontogeny of materialism has recently moved to the center stage in research. In this study, we aim to examine peer influence on youth's attitudes toward material possessions, specifically perceptions of social status benefits. Previous studies indicate that social success is an important theme in materialism (Richins 2004; Richins and Dawson 1992), in that persons with strong materialistic values use possessions as an indicator of their own and others' success. Social success for youth is often indexed by high levels of social status (for instance, popularity) among peers (LaFontana and Cillessen 2010; Lease, Musgrove, and Axelrod 2002). Hence, to understand the emergence of materialism among youth, it is important to investigate how youth form perceptions of social status benefits associated with material possessions.

Two important aspects of peer influence have been proposed in a recent conceptual model (Brown et al. 2008) and in several theories (Bandura 1986; Latané 1981): (1) the strength of peer influence on an individual and (2) the degree to which an individual is influenced by peers. We focus on these two aspects of peer influence in this study. Preadolescents are an especially appropriate age group in which to apply this model because materialism peaks in preadolescence (ages 12–13) for individuals in the developmental age span of 8 to 18 years (Chaplin and John 2007) and conformity to peers—which increases with age in childhood—also peaks in this period and decreases afterwards (Monahan, Steinberg, and Cauffman 2009; Steinberg and Silverberg 1986).

HYPOTHESIS DEVELOPMENT

A Natural Interaction-Based Peer Group

It should be noted that a peer group in this study differs from that in previous studies in the consumer literature (Moschis and Churchill 1978; Netemeyer, Bearden, and Teel 1992). In previous studies, a peer group literally represents an individual's broad peer network, for instance, other children in the same class or neighborhood with whom this individual may or may not interact. A peer group in this study is a social group with members who interact with each other frequently. Social ties for such a group are natural and interconnected interactions among members (Cairns and Cairns 1994; Cairns et al. 1995). To identify such a group, we apply a method of studying groups used in developmental psychology.

Group Norms: High-Status Peers in a Group

Research on social development makes it clear that peer groups are important in modifying and shaping individual members' development in preadolescence (Cairns et al. 1995; Rubin et al. 2006). According to group norm theory (Sherif 1936, p. 124), cognitive components such as attitudes and prejudices “are not the product of individual preferences acquired over the lifetime of this or that individual. They are the products of interacting with members of a group.” Through the interactions with group members, an individual is motivated to learn and to adhere to group norms, in part because of concerns over experiencing negative consequences such as embarrassment, awkwardness or even hostile rejections by others that may result from deviations from the norm (Abrams, Rutland, and Cameron 2003). Therefore, the formation of adolescent perceptions of social benefits of material possessions would be dependent, to a large extent, on the group normative perceptions of this aspect.

A key definitional problem, however, is to determine who establishes group norms. In a number of previous studies (Ellis and Zarbatany 2007; Espelage, Holt, and Henkel 2003), group norms were typically assessed as the average across each group member's scores (see the review, Kindermann and Gest 2009). This assessment assumes that the strength of influence of all members in a group is equal. In fact, a characteristic of a peer group emerging in preadolescence is that members often differ in their social status in a group (Adler and Adler 1998; Harris 1995). On the basis of social impact theory (Latané 1981) and social learning theory (Bandura 1986), group members with differing levels of social status will vary in the strength of their social influence on a target individual. High-status members in a peer group are proposed to have stronger social influence than low-status members. Similarly, in a conceptual model of peer influence proposed by Brown et al. (2008), the strength of peer influence is speculated to be affected by the salience of the influencer.

Consistent with this proposition, Crosnoe and Needham (2004) found that adolescents who were in positions of prominence among their own friends were more likely to reflect the norms of their friendship network than peripheral adolescents. In addition, findings in recent studies (Dijkstra, Lindenberg, and Veenstra 2008; Shi and Xie 2012) indicated significant influence of high-status rather than low-status peers in the social network. For instance, Shi and Xie (2012) found that high-status rather than low-status peers in a group influenced the development of preadolescent aggression over time. All these imply that group norms would be set by high-status peers in a group. Based on our discussion above, we expect that:

H1: High-status peers, as compared with low-status peers in a group, have stronger influence on individual members' perceptions of popularity benefits associated with material possessions.

Individual Differences in Conformity to High-Status Peers in a Group

On the other hand, individual members may vary in the extent of conformity to high-status peers in a group. A considerable number of previous studies show that an individual's conformity to peers is contingent on this individual's openness to peer influence (Bearden, Netemeyer, and Teel 1989; Monahan, Steinberg, and Cauffman 2009). For example, Bearden, Netemeyer, and Teel (1989) found that an individual who was more open to peer influence was more likely to conform to his or her friend's purchase. According to social impact theory (Latané 1981), a low-status person as compared with a high-status person will be more open to influence from high-status others. In addition, Brown et al. (2008) propose in their conceptual model that within a group, a low-status individual in a group may enhance his or her openness to peer influence in order to remain in the group. Both perspectives predict that a low-status individual in a social group will be more influenced by high-status peers than a high-status individual. This proposition is supported by findings in Shi and Xie's (2012) study showing that low-status youth in a group were more influenced by peers' aggressive behaviors than high-status youth. We therefore expect that:

H2: Low-status individual members in a group, as compared with high-status individual members, will be more influenced by high-status peers' perceptions of popularity benefits associated with material possessions.

Gender

Peer influence often occurs through social interactions (Cairns and Cairns 1994; Dishion et al. 1991). It has been shown that school-age children interact with peers in same-gender social groups (Cairns and Cairns 1994; Maccoby 1998; Rubin et al. 2006). Increasingly, as children move from preschool period to preadolescence, they spend the majority of their free play time exclusively with peers of their own sex (Maccoby 1998); therefore, same-gender peer groups are the influential social context in preadolescence. Moreover, socialization mechanisms for same-gender groups would differ from mechanisms for mixed-gender groups given that different types of relationships such as dating and romantic relationships may be involved in mixed-gender groups (Connolly et al. 2004; Hand and Furman 2009). To more clearly identify relevant peer groups, in this study, we examine socialization in same-gender groups.

In the extant literature, little attention has been directed toward the issue of gender differences in socialization processes of materialistic values and attitudes. Previous research has examined gender differences in the levels of materialism, but findings are not consistent. Some studies have found boys to be significantly more materialistic than girls (Goldberg et al. 2003; Kilbourne and LaForge 2010), while others have found no gender differences (Burroughs and Rindfleisch 2002; Richins and Dawson 1992), and one study has shown girls to be more materialistic than boys (Workman and Lee 2011). Considering that the empirical evidence for gender is not consistent enough for us to make predictions, we explore gender differences in high-status peers' influence and in individual members' conformity to high-status peers without any specific hypothesis.

H3: Peer group socialization patterns would differ across gender.

Ethnicity

There is a limited research literature examining socialization of consumption behavior and cognition among ethnically diverse youth. However, there are several findings of note. Moschis and Moore (1985) found that White adolescents appeared to be similar to African American adolescents in the likelihood of discussing consumption and observing peer consumption behavior. Shim's (1996) study found that, as compared with Whites, Hispanic adolescents were more likely to display social/conspicuous-oriented consumption decision styles and were less likely to display a utilitarian style. Moreover, Hispanic adolescents were more receptive to commercials and to receive consumer education than were White adolescents. These findings suggest potential ethnicity differences in peer influence on attitudes and values associated with objects. Given that, we include ethnicity in this study. Nonetheless, due to the relatively few findings available for which to base predictions, we examine ethnicity differences without any specific hypothesis.

H4: Peer group socialization patterns would differ across ethnicity.

A Longitudinal Design in This Study

To test the influence of peers in a peer group context, we select a longitudinal design over a cross-sectional design with regard to these two concerns. One concern is that although cross-sectional studies can provide evidence about concomitant variations (Hoyle, Harris, and Judd 2001), these studies do not allow for assessing directionality of the influence. In contrast, longitudinal studies can evaluate influence with direction (Moschis and Moore 1982). The other concern is that cross-sectional studies are problematic for inferring influence, given that the active selection of like-minded members is a critical part of peer group processes (Kindermann 1993; Urberg, Degirmencioglu, and Pilgrim 1997). This implies that peer influences will be confounded by effects of selection into the peer group if study data are gathered at only a single time point (Jaccard, Blanton, and Dodge 2005; Kindermann 1993). However, longitudinal data may identify how much of the similarity among group members stems from changes over time to conform to group members, after controlling for the selection of similar peers as members.

METHOD

Participants and Procedure

A total of 321 seventh graders (55% female, mean age = 13.01, SD = 0.45, range: 12.19–14.25) participated in this study. Data were collected in the fall (Time 1) and spring (Time 2) semesters of seventh grade. The two data points were spaced five to six months apart (early December and mid May). All participants were included for the measurement of popularity and the identification of peer group membership at both time points. The majority of participants (94.7%) were followed for measuring perceptions of popularity benefits associated with material possessions at Time 2.

All participants were recruited from three middle schools in an urban school district in the northeastern United States. Only students with signed parental consent were included. The overall participation rate for this study was 59% (321/547). A higher participation rate was found among girls (68%) than among boys (50%), χ²(1, 547) = 17.77, p < .001. About one half of participants were African American, 31% White, 19% Hispanic and 1% Asian or other ethnicity. Fifty percent of participants were eligible for free-lunch at school and an additional 8% of participants were eligible for reduced-price lunch. Participants and non-participants did not differ in their ethnicity background and eligibility for free or reduced-price lunch, χ²s < 6.95, ps > .13.

At both time points, participants completed questionnaire measures in a group-administered survey session. During the survey, a lead administrator read the instructions and questions aloud, while additional assistants provided mobile monitoring and assistance to students. Before the beginning of the survey, participants were assured of confidentiality, and they were asked not to discuss their responses with other students. Blank papers were provided so that all participants could cover their answers in the survey. Participants were also told they could stop their participation at any time, with no repercussions.

Measures

Perceptions of Popularity Benefits of Material Possessions

In this study, perceptions of popularity benefits associated with material possessions are measured by the degree to which an individual perceives that material possessions can make someone popular among peers. On the basis of Xie et al.'s (2006) findings that clothes and money are two types of material possessions related to popularity, we use these two items as specific material possessions for this measurement.

At Time 1, female participants were asked to rate these two items for “what makes a girl popular in school?” Male participants were asked to rate the same two items for “what makes a boy popular in school?” A Likert-type format was used with scores ranging from 1.0 (low: very unlikely to make someone popular) to 5.0 (high: very likely to make someone popular). Scores were standardized by school. The correlation between scores on the two items (i.e., clothes and money) was significant, r = .33, p < .001. Perceptions of popularity benefits of material possession were calculated by the average of standardized scores on these two items.

Peer Groups

The Social Cognitive Map (SCM) procedure was used to identify peer groups (Cairns and Cairns 1994; Cairns et al. 1995). In the fall semester, seventh-grade participants were asked to nominate peer groups within their grade level through the prompts: “Are there any students in your grade who hang around together a lot?” “Please write their names.” Following the procedure and previous studies on peer groups (Farmer and Rodkin 1996), a list of names was not provided. Using the nominations of peer groups from all respondents within the grade, we identified social groups. Details of the SCM procedure are available from the authors.

In this study, group membership was identified for 272 participants belonging to 69 peer groups (with three or more members). Group size ranged from 3 to 16 (mean = 6.58, SD = 3.09). The majority of identified groups were of the same gender: 32 all-girls groups, 29 all-boys groups and 8 mixed-gender groups. All-girls group did not differ from all-boys group in group size, t(59) = 1.14, p = .26. Among these groups, 25 groups were ethnicity-mixed, 11 were White groups, 22 were African American groups and 11 were Hispanic groups.

In addition, four mixed-gender groups with high-status members of both genders were dropped from the analyses. For the remaining four mixed-gender groups, a total of five opposite-gender member(s) were identified. These opposite-gender members had very low levels of social status in the groups. After discussions with another co-worker in this area, we agreed that these groups with only one or two opposite-gender members whose social status was very low may function like same-gender groups. We also conducted all hierarchical linear modeling (HLM) models in this study without these four mixed-gender groups and similar findings emerged. Therefore, these four mixed-gender groups were then “made into” same gender groups by dropping low-status members who were opposite gender to the high-status members. A total of 245 individual members belonging to 65 peer groups were thus included in the analysis. Participants included in the analysis did not differ from participants excluded from analysis, nor did they differ from non-participants in their ethnicity background and eligibility for free or reduced-price lunch, χ²s < 1.36, ps > .24.

High- and Low-Status Members in a Group

In the SCM procedure, individual status in a group is measured based on the number of times a child is nominated into his or her group. Group Nomination is determined as the average of individual nominations for the two most nominated members in a group. An individual is classified as nuclear in a group if his or her individual nomination is greater than or equal to 70% of the Group Nomination. For this study, nuclear members are high-status members in a group, and all the others are low-status members. It is important to note that not every group contained low-status members, because the individual nominations could be similar among group members. Indeed, in 17 groups all members were high-status members. In the remaining 48 groups, 52% of group members (ranging from 18% to 83%) were high-status members.

Group Members' and High- and Low-Status Members' Perceptions of Popularity Benefits of Material Possessions

An average score of each member's perceptions of popularity benefits associated with material possessions at Time 1 was calculated as group members' perceptions. In addition, an average score of high-status members' perceptions of popularity benefits associated with material possessions at Time 1 was calculated as high-status members' perceptions. And an average score across low-status members at Time 1 was measured as low-status members' perceptions.

Stability of Peer Affiliation

Peer groups in childhood and adolescence are often dynamic in organization (Brown 1990; Cairns et al. 1995). Group membership is not static over time. Because the strength of high-status members' influence will vary by the amount of time high-status members have spent with an individual child, we control for the stability of affiliation with high-status members in our analyses involving high-status members' influence. Using the same SCM procedure, we identified peer groups in the spring semester of seventh grade (Time 2). We determined a target individual's stability of affiliation with high-status members by comparing high-status members in a target individual's groups at two predetermined times (i.e., fall and spring semesters). First, we counted the number of high-status members who remained in the same group with a target child (excluding the target child if he or she was a high-status member). Then we divided this number by the total number of high-status members in the target child's group at Time 1 (again excluding the target child if she or he was a high-status member).

Also, we control for the stability of affiliation with low-status members by examining the influence of low-status members, and control for the stability of affiliation with group members by examining the influence of all group members. The calculation is similar to that of the stability of affiliation with high-status members, except that we counted the number of low-status members for calculating affiliation with low-status members (excluding the target child if he or she was a low-status member) and the number of group members for calculating affiliation with group members (excluding the target child).

Overview of Statistical Analysis

An HLM procedure is used to test hypotheses in this study. The use of HLM is appropriate due to the hierarchical structure of data in this study: individual members are nested within peer groups (Luke 2004; Raudenbush and Bryk 2002).

We examine the influence of group members, high-status members and low-status members on an individual's perceptions of popularity benefits of material possessions. Individuals' perceptions at Time 1 (fall semester) and the stability of peer affiliation from Time 1 to Time 2 (spring semester) are entered in all Level-1 (or individual level) models as control variables. The variance in individual members' Time 2 perceptions is modeled at Level 1. At Level 2 (or group level), the randomly varying intercept from the Level-1 model is entered as the dependent variable, and peer perceptions are entered as a predictor. For instance, to test the influence of high-status members, the following Level-1 model is used:

• (1)

The Level-2 model for testing high-status members' socialization effect is represented by the following equation:

• (2)

In these two models, PERT2 and PERT1 represent individuals' perceptions at Time 2 and Time 1, respectively. STAB represents the stability of affiliation with high-status members in a group, and HIGHPERT1 represents high-status members' perceptions at Time 1. The significance of the coefficient for HIGHPERT1 (γ₀₁) determines whether high-status members have significant influence on individuals' perceptions. Similar models are estimated to test the influence of the average of group members and low-status members.

To test gender differences in high-status members' influence, we use the same Level-1 model (i.e., Equation 1). The Level-2 model for testing gender differences is represented by the following equation:

• (3)

The significance of the coefficient for HIGHPERT1 × MALE (γ₀₃) determines whether the influence of high-status members varies across gender.

We also test ethnicity differences in high-status peers' influence. Because some peer groups are ethnicity-mixed, the effect of ethnicity is tested at both individual (Level-1) and group (Level-2) levels. In this study, only 1% of participants are Asian or other ethnicity; given that, we focus on African American, White and Hispanic preadolescents. Two dummy variables are used for ethnicity at individual level: White vs. African Americans and Hispanic vs. African Americans. We use one dummy variable (i.e., ethnicity-mixed vs. ethnicity-segregated groups) to code ethnicity at group level.

We first test whether high-status peers' influence varies across ethnicity-mixed and ethnicity-segregated groups. The models are similar to the models for testing gender difference (i.e., Equations 1 and 3), except that ethnicity at group level and the interaction between high-status members' perceptions and group ethnicity are entered as predictors in the Level-2 model. The significance of the coefficient for the interaction between high-status members' perceptions and group ethnicity determines whether the influence of high-status members varies across ethnicity-mixed and ethnicity-segregated groups.

Next, we test whether individual ethnicity affects the strength of high-status members' influence. The Level-1 model is represented by the following equation where HISPA represents Hispanic Americans:

• (4)

The Level-2 model is represented by the following equations:

• (5)

• (6)

Considering that these two Level-2 models are similar, we use the former (i.e., Equation 5) as an example. The varying slope of WHITE (β_2j) in the Level-1 model is entered as the dependent variable in this Level-2 model. The significance of the coefficient for HIGHPERT1 (γ₂₁) in the Level-2 model determines whether high-status members' influence varies across White and African American individuals.

As peer groups are not ethnicity-segregated, it is necessary to test whether the effect of individual ethnicity varies across ethnicity-mixed and ethnicity-segregated groups. The models are similar to the forgoing models for testing the effect of individual ethnicity (i.e., Equations 4–6), except that in addition to HIGHPERT1, group ethnicity and the interaction between HIGHPERT1 and group ethnicity are also entered as the predictors in the Level-2 models. The significance of the coefficient for the interaction between high-status members' perceptions and group ethnicity in Level-2 models determines whether the effect of individual ethnicity on high-status members' influence varies across ethnicity-mixed and ethnicity-segregated groups.

Additionally, the effect of individual status on high-status members' influence is tested in this study. STAB (i.e., stability of peer affiliation) and PERT1 (i.e., individual perceptions at Time 1) are entered in the Level-1 model as control variables, and STATUS (i.e., individual status in a group) is entered as the predictor. The Level-1 model is represented by the following equation:

• (7)

The Level-2 model is described below:

• (8)

The significance of the coefficient for HIGHPERT1 (γ₁₁) on the varying slope of STATUS (β_1j) in the Level-1 model determines whether individual status moderates high-status members' influence.

Significant interactions in HLM models are presented according to the guidelines delineated by Aiken and West (1991). Variables in the interaction are split into 1 SD above the mean (high level) and 1 SD below the mean (low level) to plot the effect.

RESULTS

Correlations among Variables

Zero-order correlations among variables in this study are listed in Table 1. Given that scores for measuring perceptions of popularity benefits of material possessions are standardized by school, and scores for variables such as gender and social status are categorical, we do not present the means of these variables. Preadolescents' perceptions of social status benefits of material possessions are significantly correlated across two time points (r = .38, p < .001). In addition, girls as compared with boys are more likely to perceive material possessions as a means of achieving popularity at Time 1 (r = −.13, p < .05), whereas no gender difference is found at Time 2 (r = −.09, p = .12). Only one type of statistically significant ethnicity difference is detected: African Americans endorse material possessions for popularity more than Hispanics at Time 1 (r = −.18, p < .01). The perceptions of high-status members in a group are negatively correlated to the perceptions of low-status members at Time 1 (r = −.21, p < .01).

Table 1. Zero-Order Correlations for Individual- and Group-Level Variables

Level-1 Variables	4	5	6
Notes: For preadolescent gender, males have the higher coded value. African Americans are the reference group for ethnicity. *p < .05, **p < .01, ***p < .001.
1. Male (vs. female)	0.10	−0.13*	−0.09
2. White (vs. African Americans)	0.10	−0.02	−0.10
3. Hispanic (vs. African Americans)	−0.002	−0.18^**	−0.08
4. Individual status in a group (high-status member = 1)		−0.10	−0.02
5. Perceptions at Time 1			0.38***
6. Perceptions at Time 2
Level-2 Variables	2	3
1. Average of group members' perceptions at Time 1	0.66***	0.64***
2. High-status members' perceptions at Time 1		−0.21**
3. Low-status members' perceptions at Time 1

High-Status Members' Influence in the Peer Group

Results from HLM analysis of the influence of group members, high-status members and low-status members on individuals' perceptions of popularity benefits of material possessions are presented in Table 2. The estimates of the coefficients for four separate models are reported in the table. For all four models, the intercept in the Level-1 model (i.e., Level-1 Intercept in the table) is the dependent variable in the Level-2 model. In Model 1 of the four models, the influence of group members' influence is tested. The intercept and the coefficient for group members' perceptions (i.e., Average Level of Group Members' Perceptions) in the Level-2 model are reported in column 1 of the table. The coefficients for control variables (i.e., Individual Time 1 Perceptions and Stability of Affiliation) in the Level-1 model are also reported. Estimation results show that the coefficient for Average Level of Group Members' Perceptions is not statistically significant (b = 0.22, t(58) = 1.51, p = .14). This means that the average of group members' endorsement of material possessions for popularity at Time 1 does not influence individual members' endorsement of material possessions for popularity at Time 2. In Model 2, the influence of high-status members is tested. The intercept and the coefficient for High-Status Members' Perceptions in the Level-2 model are reported in column 2. Results show the coefficient for High-Status Members' Perceptions is statistically significant (b = 0.26, t(58) = 2.55, p < .05). This means the higher are scores for high-status members' endorsement of material possessions for popularity at Time 1, the more likely it is that individuals in the group increased in the levels of their endorsement of material possessions for popularity at Time 2. In Model 3, the influence of low-status members is tested. The intercept and the coefficient for Low-Status Members' Perceptions in the Level-2 model are reposted in column 3. The coefficient for Low-Status Members' Perceptions is not statistically significant (b = −0.03, t(39) = −0.26, p = .79). This means low-status members' endorsement of material possessions for popularity does not have a statistically significant influence on individual members.

Table 2. Peer Influence on Individuals' Perceptions of Popularity Benefits of Material Possessions

Variables	Model 1 Coefficient (SE)	Model 2 Coefficient (SE)	Model 3 Coefficient (SE)	Model 4 Coefficient (SE)
^a Stability of affiliation with group members. ^b Stability of affiliation with high-status members. ^c Stability of affiliation with low-status members. *p < .05, **p < .01, ***p < .001.
Level-1 intercept
Intercept	−0.02 (0.06)	−0.02 (0.06)	0.004 (0.07)	−0.01 (0.06)
Average level of group members' perceptions	0.22 (0.15)
High-status members' perceptions		0.26^* (0.10)		0.33^** (0.11)
Low-status members' perceptions			−0.03 (0.13)	0.05 (0.12)
Control variables
Individual Time 1 perceptions	0.41^*** (0.08)	0.40^*** (0.07)	0.42^*** (0.08)	0.37^*** (0.07)
Stability of affiliation	−0.25a (0.12)	−0.10b (0.11)	−0.58c (0.13)	−0.27a (0.14)
Level-1 pseudo R^2	0.21	0.22	0.22	0.23
Level-2 pseudo R^2	0.21	0.22	0.19	0.22

Following the instructions described by Luke (2004), we calculated pseudo R² (i.e., the proportional reduction of prediction error as compared with that of the null model with no predictors). Pseudo R²s for Model 1 in Table 2 indicate that by including two level-1 variables (i.e., Perceptions at Time 1 and Affiliation Stability) and one Level-2 variable (i.e., Average Level of Group Members' Perceptions), we are able to improve the predictive ability of this model by approximately 21%. The improved proportion of prediction error for Models 2 and 3 is approximately 19% to 22%.

In Model 4, we test the influence of high- and low-status members at the same time. Estimation results are reported in column 4. The coefficient for High-Status Members' Perceptions is statistically significant (b = 0.33, t(42) =3.08, p < .01), whereas the coefficient for Low-Status Members' Perceptions is not (b = 0.05, t(42) = 0.46, p = .65). These results indicate that it is high-status members' endorsement of material possessions for popularity, rather than low-status members' endorsement, that influence individuals' perceptions over time. The improved proportion of prediction error for Model 4 is around 22% to 23%.

As anticipated in H1, findings show that high-status rather than low-status peers in a group exert statistically significant influence on a target individual's perceptions. These findings are in accord with extant theories such as social learning theory (Bandura 1986) and social impact theory (Latané 1981) and with Brown's conceptual model of peer influence (Brown et al. 2008), which propose that high-status children will be more likely than low-status children to have strong influences on their peers. Building on the work of Crandall (1988) and findings in related studies (Crosnoe and Needham 2004; Dijkstra, Lindenberg, and Veenstra 2008), our findings suggest that a group's norm is likely to be set by its high-status members, rather than by all group members.

The Moderating Effects of Gender and Individual Status

Furthermore, we test whether the strong influence of high-status members would vary by gender. Results of estimates examining whether influence effects differ by gender are presented in Table 3. In Model 1, we test gender difference in high-status members' influence. The intercept in the Level-1 model (i.e., Level-1 Intercept) is the dependent variable in the Level-2 model. The intercept and the coefficients for High-Status Members' Perceptions, Male and the interaction of High-Status Members' Perceptions × Male in the Level-2 model are presented in column 1. The coefficient for the interaction of High-Status Members' Perceptions × Male is not statistically significant (b = 0.01, t(56) = 0.03, p = .97). This means high-status peers have similar influence in girls' and boys' groups. The improved proportion of prediction error compared with a null model for Model 1 in Table 3 is approximately 18% to 21%.

Table 3. High-Status Members' Influence on Individuals' Perceptions by Individual Status and Gender

Variables	Model 1 Coefficient (SE)	Model 2 Coefficient (SE)	Model 3 Coefficient (SE)
Notes: For preadolescent gender, males have the higher coded value. ^† p < .10, *p < .05, **p < .01, ***p < .001.
Level-1 intercept
Intercept	−0.05 (0.09)	−0.06 (0.07)	−0.16^* (0.07)
High-status members' perceptions	0.27^† (0.17)	0.30^* (0.14)	0.57^** (0.18)
Male (vs. female)	0.07 (0.11)		0.11 (0.13)
High-status members' perceptions × male	0.01 (0.19)		−0.44 (0.24)
Level-1 individual status
Intercept		0.07 (0.09)	0.16^† (0.13)
High-status members' perceptions		−0.03 (0.19)	−0.41^† (0.24)
Male			−0.04 (0.17)
High-status members' perceptions × male			0.63^* (0.30)
Control variables
Individual Time 1 perceptions	0.40^*** (0.07)	0.40^*** (0.08)	0.42^*** (0.08)
Stability of affiliation with high-status members	−0.08 (0.10)	−0.12 (0.11)	−0.11 (0.11)
Level-1 pseudo R^2	0.21	0.22	0.21
Level-2 pseudo R^2	0.18	0.21	0.19

In Model 2, we test the effect of individual status on high-status members' influence. Two Level-2 models are involved. In one Level-2 model, the intercept in the Level-1 model (i.e., Level-1 Intercept) is the dependent variable, and in the other Level-2 model, the coefficient for individual status in the Level-1 model (i.e., Level-1 Individual Status) is the dependent variable. The predictor in both Level-2 models is high-status members' perceptions. Estimates are presented in column 2. The coefficient for the interaction of Individual Status × High-Status Members' Perceptions is not statistically significant (b = −0.03, t(58) = −0.17, p = .87). This means individual status does not moderate high-status members' influence on individuals' perceptions.

In Model 3, we test whether gender plays a role in the effect of individual status on high-status members' influence. Two Level-2 models are involved. The dependent variables in these Level-2 models are the same as those for Model 2. The predictors in both Level-2 models are High-Status Members' Perceptions, Male and the interaction of High-Status Members' Perceptions × Male. Estimates are presented in column 3. The coefficient for the three-way interaction of Individual Status × High-Status Members' Perceptions × Male is statistically significant (b = 0.63, t(56) = 2.13, p < .05). This means in girls' groups, high-status members who highly endorse material goods for popularity induce a larger increase in low-status individuals' endorsement of material possessions for popularity than in high-status individuals' endorsement of this aspect (Figure 1, left panel). In boys' groups, high-status members who highly endorse material goods for popularity induce a larger increase in high-status individuals' endorsement of this aspect than in low-status individuals' (Figure 1, right panel). The improved proportion of prediction error for Models 2–3 is approximately 19% to 22%.

Figure 1. The Influence of High-Status Members on Individuals' Time 2 Perceptions by Individual Status and Gender (Left: Girls' Groups; Right: Boys' Groups)

Findings reported in this study are contrary to H2. We fail to find a statistically significant difference in conformity to high-status peers' influence for individuals with various social statuses in a group. Note that gender plays a significant role. In girls' groups, high-status peers who highly endorse social status benefits of material possessions are found to induce a larger increase in low-status individuals' endorsement than in high-status individuals'. This pattern was consistent with our expectation. However, results from boy's groups reveal an unexpected finding—that in boys' groups, high-status peers who highly endorse the social benefits of material possessions induce a larger increase in high-status individuals' endorsement than in low-status individuals'. Results for boys' groups are opposite for girls' groups, wherein low-status individuals as compared with high-status individuals are more open to influence from high-status peers. Therefore, the combined effect of individual status on high-status peers' influence is weak for the overall sample. This explains the failure of detecting a statistically significant effect of individual status for the complete sample, as was expected in H2.

It is not clear why the unexpected pattern occurs in boys' groups, and this finding itself needs to be replicated in future studies. Here we speculate that material possessions may not be a salient factor associated with high social status among boys (Adler and Adler 1998; Closson 2009). Since material possessions are not salient, low-status boys may pay less attention to material possessions and related perceptions of high-status peers than other prominent factors such as athletic performance. In contrast, high-status boys in a group may already possess these prominent factors associated with high social status, therefore they may pay more attention to subtle factors such as material possessions than the low-status boys. As a result, a stronger peer influence may be observed among high-status boys on perceptions of material possessions.

The Moderating Effect of Ethnicity

In this section, we focus on ethnicity difference in high-status members' influence. Results are presented in Table 4. In Model 1 presented in the table, we test whether high-status members' influence differs across ethnicity-mixed and ethnicity-segregated groups. The intercept in the Level-1 model (i.e., Level-1 Intercept) is the dependent variable in the Level-2 model. The intercept and the coefficients for High-Status Members' Perceptions, Ethnicity-Mixed Groups and the interaction of High-Status Members' Perceptions × Ethnicity-Mixed Groups in the Level-2 model are presented in column 1 of the table. The coefficient of High-Status Members' Perceptions × Ethnicity-Mixed Groups is not statistically significant (b = −0.10, t(56) = −0.61, p = .55). This means that high-status members' influence does not vary across ethnicity-mixed and ethnicity-segregated groups. The improved proportion of prediction error for Model 1 is approximately 25% to 29%.

Table 4. High-Status Members' Influence on Individuals' Perceptions by Ethnicity

Variables	Model 1 Coefficient (SE)	Model 2 Coefficient (SE)	Model 3 Coefficient (SE)
Note: African Americans are the reference group for ethnicity. *p < .05, **p < .01, ***p < .001.
Level-1 intercept
Intercept	0.08 (0.07)	−0.05 (0.11)	0.13 (0.10)
High-status members' perceptions	0.31* (0.12)	0.71^*** (0.15)	0.60*** (0.16)
Ethnicity-mixed groups (vs. ethnicity-segregated groups)	−0.21 (0.11)		−0.34* (0.17)
High-status members' perceptions × ethnicity-mixed groups	−0.10 (0.17)		0.28 (0.31)
Level-1 White (vs. African Americans)
Intercept		−0.12 (0.13)	−0.33* (0.16)
High-status members' perceptions		−0.79^*** (0.23)	−0.56* (0.27)
Ethnicity-mixed groups			0.36 (0.21)
High-status members' perceptions × ethnicity-mixed groups			−0.50 (0.44)
Level-1 Hispanic (vs. African Americans)
Intercept		0.11 (0.15)	−0.07 (0.20)
High-status members' perceptions		−0.45^** (0.17)	−0.35 (0.23)
Ethnicity-mixed groups			0.32 (0.26)
High-status members' perceptions × ethnicity-mixed groups			−0.22 (0.35)
Control variables
Individual Time 1 perceptions	0.40*** (0.07)	0.34^*** (0.07)	0.33*** (0.07)
Stability of affiliation with high-status members	−0.09 (0.10)	0.01 (0.10)	0.01 (0.10)
Level-1 pseudo R^2	0.29	0.26	0.20
Level-2 pseudo R^2	0.25	0.29	0.21

In Model 2, the effect of individual ethnicity on high-status members' influence is tested. Three Level-2 models are used. The intercept (i.e., Level-1 Intercept) and the coefficients for two dummy variables for ethnicity (i.e., Level-1 White and Level-1 Hispanic) in the Level-1 model are the dependent variables in these three Level-2 models, respectively. The predictor in these Level-2 models is High-Status Members' Perceptions. Estimation results are reported in column 2 of the table. The coefficient for the interaction of White × High-Status Members' Perceptions is statistically significant (b = −0.79, t(56) = −3.45, p = .001). High-status peers are found to have stronger influence on African Americans than on White Americans. In addition, the coefficient for the interaction of Hispanic × High-Status Members' Perceptions is statistically significant (b = −0.45, t(56) = −2.71, p < .01). High-status peers are found to have stronger influence on African Americans than on Hispanic Americans. The improved proportion of prediction error for Model 2 in Table 4 is approximately 26% to 29%.

Additionally, we test whether the strength of high-status peers' influence varies between White and Hispanic Americans, we recode ethnicity with Hispanic as the reference. Similar HLM models are used. High-status peers tend to have stronger influence on Hispanic than on White preadolescents (b = −0.34, t(57) = −1.86, p = .07).

In Model 3, we test whether the effect of individual ethnicity varies across ethnicity-mixed and ethnicity-segregated groups. Three Level-2 models are constructed and the dependent variables in these three Level-2 models are the same as those for Model 2. The predictors in the Level-2 models for Model 3 are High-Status Members' Perceptions, Ethnicity-Mixed Groups and the interaction of High-Status Members' Perceptions × Ethnicity-Mixed Groups. Estimation results for intercepts and coefficients in Level-2 models are presented in column 3. Results show that the coefficient for the three-way interaction of White × High-Status Members' Perceptions × Ethnicity-Mixed Groups is not statistically significant (b = −0.50, t(55) = −1.14, p = .26). This means that group ethnicity does not affect the difference in high-status members' influence on White and African Americans. The coefficient for the three-way interaction of Hispanic × High-Status Members' Perceptions × Ethnicity-Mixed Groups is not statistically significant (b = −0.22, t(55) = −0.62, p = .54). This means that group ethnicity does not affect the difference in high-status members' influence on Hispanic and African Americans. The improved proportion of prediction error for Model 3 is approximately 20% to 21%.

Additionally, we test whether the effect of individual ethnicity on high-status members' influence varies across gender. The coefficient for the three-way interaction between White, high-status members' perceptions and gender was not statistically significant (b = 0.40, t(55) = 0.94, p = .35). This means gender does not affect the difference in high-status members' influence on White and African Americans. In addition, the coefficient for the three-way interaction between Hispanic, high-status members' perceptions and gender is not statistically significant (b = 0.08, t(55) = 0.23, p = .82). This means gender does not affect the difference in high-status members' influence on Hispanic and African Americans. Because the effect of individual ethnicity does not vary across gender, we do not report these results in detail.

Furthermore, we test whether the effect of individual ethnicity on high-status members' influence varies across individual status. The coefficient for the three-way interaction between White, high-status members' perceptions and individual status is not statistically significant (b = 0.26, t(56) = 0.67, p = .51). Individual status does not affect the difference in high-status members' influence on White and African Americans. Also the coefficient for the three-way interaction between Hispanic, high-status members' perceptions and individual status is not statistically significant (b = 0.38, t(56) = 1.10, p = .28). Individual status does not affect the difference in high-status members' influence on Hispanic and African Americans. As the effect of individual ethnicity does not vary across individual status, we do not report the detailed results here.

DISCUSSION

With the rising concerns about the dangers and ubiquity of materialism (Goldberg et al. 2003; Schor 2004), consumer educators and researchers are working on how to diminish and prevent materialism among youth. This study is important because it highlights the point of intervention and prevention for reducing the pervasiveness of materialism. Findings in this study suggest that one way to reduce materialism may involve changes in an individual's attitudes toward material possessions. Preadolescence is an important stage for forming materialistic values and attitudes (Chaplin and John 2007). In order to prevent and intervene in the prevalence of materialism among youth, natural, interaction-based peer groups should be a crucial social context. As this study implicates, high-status youth in a group who highly endorse the social benefits of material possessions should be the focus of future intervention and prevention efforts. Furthermore, the intervention and prevention efforts should be elaborated based on individual characteristics of the target youth (e.g., gender, ethnicity and individual status in a group). Given that our findings suggest that low-status boys, in particular, low-status White boys, are not highly open to high-status adolescents' influence, additional efforts should also be given to these boys.

In addition to the practical implications, findings in this study advance our knowledge on mechanisms of peer influence on youth's attitudes toward material possessions. They add to prior materialism research in several ways.

First, we introduce the idea that peers differ in their influence on the formation of individual attitudes toward material possessions. Developmental psychologists and consumer researchers have long acknowledged peers as a facilitator of learning and socialization to young people (see the review in Azmitia and Perlmutter 1989; Mangleburg, Grewal, and Bristol 1997). This study is among the first in materialism research to show that during preadolescence, variability in peer influence involves two aspects: (1) the strength of influences of different peers in a social group and (2) the degree to which each group member conforms to their peers. Variations in both aspects are related to social status in a group. Second, we introduce the primary peer context—natural, interaction-based peer groups—into the studies of peer influence. A unique position is adopted in this study to explore mechanisms for within-group transmission of perceptions related to objects. More studies are required for elaborating within-group processes of consumption cognition and behavior. For instance, this study demonstrates high-status peers' influence on individual members' perceptions of popularity benefits of material possessions. Future studies are needed in investigating the exact group processes underlying such influence. Third, we examine the effects of gender and ethnicity on socialization processes of individual attitudes toward material goods. Little attention has been directed toward the effects of gender and ethnicity before, resulting in a lack of conceptualization about what the differences might be in related areas such as materialistic attitudes, values and the like. In fact, these differences are important given that they may be likely to have an impact on social consumption and become quite salient as children enter adolescence. Findings in this study indicate that the issue of gender and ethnicity differences in socialization processes is complex. More studies are required in future research.

Several limitations need to be mentioned. First, in this study, only one group is allowed for each child as required by the nested relationship between an individual and group in the HLM procedure. Around 20% of group members (55/272) in this study belong to two groups and 3% (7/272) belong to three groups. As the requirements of multilevel analysis demand that each child should be placed into only one peer group, those participants who belong to multiple groups are assigned to their primary group where they receive the most nominations. Children who affiliate with two or more peer groups may experience different types of influence. Future research comparing socialization effects across children's multiple groups would help clarify this issue. Second, values and attitudes are different. For instance, although materialistic individuals often endorse material objects for social achievement, individuals who endorse material objects for social achievement may not necessarily be materialistic. Future research is required to further probe the connections and distinctions between these two constructs. Third, only two items (i.e., money and clothes) are used in this study. Despite the finding that these two items are highly related to youth's popularity (Xie et al. 2006) and the fact that money represents an individual's spending power, and clothes are the publicly displayed possession which could directly symbolizes an individual's social status or success (Harms 1938), other types of material goods especially those popular products emerging recently in the market need to be considered in future research. Finally, the sample in this study is preadolescents in an urban school district in United States, and around half of the sample is African American. This sample is more likely to represent urban preadolescents than all preadolescents in the United States. According to Bronfenbrenner's (1979) propositions, individuals' perceptions, attitudes and values would vary across social environments. It is not certain whether findings in this study would be applicable to youth in a different region (e.g., a rural school district). Replications of findings in this study are needed in future research.

CONCLUSION

The present research provides good evidence that preadolescents' peer groups are important in the formation of individual members' attitudes toward material possessions, specifically perceptions of popularity benefits of objects. It is among the first in materialism research to show that during preadolescence, peers in a social group differ in the strength of influence on individual members' perceptions of popularity benefits of material possessions, and individual members also differ in conformity to peer influence. Variations in both aspects could be predicted by levels of social status in a group. Specifically, high-status rather than low-status peers in a group are found to influence a preadolescent's perceptions of social benefits of material possessions. The strength of high-status peers' influence is similar across gender. Moreover, high-status peers are particularly influential on low-status members in girls' groups and on high-status members in boys' groups. Additionally, high-status peers are found to have stronger influence on African Americans than on Hispanic Americans, and tend to have stronger influence on Hispanic Americans than on White Americans. In this study, we make an introduction to the materialism literature in three perspectives: (1) the variability in the strength of peer influence and individuals' openness toward peer influence; (2) the primary social context—a natural, interaction-based peer group—in preadolescence; (3) gender and ethnicity differences in socialization processes through peers. Moreover, findings in this study have practical implications on how to diminish and prevent materialism among youth. These findings suggest that special attention should be given to high-status group members who highly endorse social benefits of material possessions in future intervention and prevention efforts for reducing the pervasiveness of materialism. Individual characteristics of the target youth (e.g., gender, ethnicity and individual status) should also be considered.