the influence of friendship groups on intellectual self-confidence and educational aspirations in college




Anthony Lising Antonio

Stanford University






direct correspondence to:

anthony lising antonio

school of education

stanford, ca 94305-3084











         Research on the influence of peer groups on socialization and development in college has been of continual interest since Davis’ classic paper describing relative deprivation effects in 1966.  The bulk of this college peer group research, however, has focused on the reference group of the student body and neglected the more proximal influence of interpersonal membership groups.  In addition, race has rarely been incorporated as a salient characteristic of peer groups.  In this study, the author examines the influence of college friendship groups on students’ intellectual self-confidence and educational aspirations.  Findings indicate that at the interpersonal level, the competing effects of relative deprivation and environmental press are present among white students.   Racial diversity in the friendship group was found to have positive effects, but only for students of color.



         Over the past 30 years, research on how college impacts student development has continually pointed to the peer group as perhaps the dominant change agent during the college years (Feldman and Newcomb 1969; Pascarella and Terenzini 1991).  A college student’s peers act as a reference group, or an environmental source of sociocultural norms in the midst of which a student grows and develops (Clark and Trow 1966).  A large body of empirical evidence has been collected over the years to support this conclusion (Astin 1977, 1993a; Feldman and Newcomb 1969; Pascarella and Terenzini 1991). 

         A review of the research on the impact of college peer groups reveals an interesting trend.  The earliest work on peer groups (primarily in the 1950s and early 1960s) focused on peer associations that were structured organizationally by either residential circumstances or formal group affiliations (Feldman and Newcomb 1969).  Most of this work was conducted at single institutions.  Furthermore, there was recognition that while the student body characteristics of individual colleges may accentuate initial differences between students attending different institutions, student subcultures and friendship groups within institutions probably mediate the developmental impact of the student body (Feldman and Newcomb 1969).  Despite general consensus on the important role that friendship groups play in socialization in college, however, the majority of published peer group studies since the mid-1960s have been multi-institutional and have operationalized the student peer group as a singular, campus-wide entity (Pascarella and Terenzini 1991).  Peer group effects, in other words, were studied as between-institution effects (e.g., Astin 1977; Bassis 1977; Davis 1966; Thistlewaite and Wheeler 1966; Werts and Watley 1969).  Our current understanding of these institutional effects is that they are likely mediated through interactions with students within a variety of interpersonal environments (Astin 1993a; Kuh 1995; Pascarella and Terenzini 1991).   Given concurrent research underscoring the importance of student interaction and engagement on campus for development and retention (Astin 1984; Pascarella 1985; Tinto 1975; Weidman 1989), it is surprising that little current work on peer group influence in college focuses on interpersonal environments such as friendship groups and cliques.

         The campus environment itself has changed greatly since the 1950s and 1960s.  Colleges and universities are rapidly becoming ethnically and racially diverse student communities (Justiz 1994), and increasing campus diversity has been accompanied by a rise in racial tension on campus, battles over free speech and the curriculum fought across racial lines, and social self-segregation by race (Altbach 1991).  These troubling patterns are less than gentle reminders that issues of racial and ethnic difference pervade many corners of the university, and questions regarding student experiences and student development on today’s campuses must include the role of racial diversity in their formulation.  The general purpose of this study is to conduct a contemporary examination of peer group influence in college that focuses on interpersonal environments and also addresses the role of racial diversity in those environments.

Peer groups and peer group influence

         Researchers in the fields of sociology and social psychology have tended to view student peers as a determinant of school context, which acts as a referent against which students evaluate themselves (Alwin and Otto 1977).  The vast majority of the work that has drawn conclusions on the influence of college peer groups reflects this view, if not explicitly so, in the manner in which the peer group is operationalized methodologically.   In these studies, the peer group was thought of as a reference group encompassing the entire student body.  Early work, for example, likened the campus to a frogpond within which students formed judgments of their abilities and aspirations.  Such studies typically measured the relationship between an individual characteristic and the aggregate characteristics of a sample of a school’s student body to infer peer group effects (e.g., Bassis 1977; Davis 1966; Drew and Astin 1972; Pascarella, Smart, Ethington, and Nettles 1987; Thistlethwaite and Wheeler 1966; Werts and Watley 1969).  The most recent research on college peer group effects continues to follow this conceptual and methodological model.  For example, a number of studies continue to use the average freshman class SAT scores of an institution to characterize the peer academic context (e.g., Astin 1993a; Hurtado and Carter 1997) and institutional aggregates of individual-level variables such as social attitudes and political views to characterize the peer social context (e.g., Astin 1993a; Dey 1996, 1997; Milem 1998).

         Astin’s (1993a) multi-institutional, comprehensive study of college impact contained many institutionally-based peer group measures such as peer socioeconomic status, materialism, and feminism.  These and other peer group measures developed by Astin are multi-item composite variables and represent an improvement in the characterization of the peer environment with respect to the range of characteristics measured as well as in the reliability of the measures.  Several researchers recently have illustrated peer group effects using similarly constructed composite variables (e.g., Chang 1995; Milem 1994; Pascarella, Edison, Nora, Hagedorn, and Terenzini 1996; Sax 1994, 1996).  While the use of multiple composite measures has allowed researchers to more extensively capture the college peer environment, they remain contextual measures that are distal to actual student experiences. 

         As Feldman and Newcomb (1969) have noted, peer groups can also be thought of as membership groups.  Within such social groups, shared and consensual sets of norms are developed through interpersonal interaction.  Individuals then change under the pressure of direct approval (or disapproval) of valued, trusted peers.  This process of peer influence is theoretically distinct from that occurring via reference groups.  Reference group peers influence students through school-level, macro-social processes.  Researchers assert, however, that micro-social processes, particularly interpersonal interactions within membership groups, mediate these institutional-level influences (Alexander and Eckland 1975; Alwin and Otto 1977).  A separate line of research focusing on the effects of student involvement on development has, in fact, shown that interpersonal interactions are a primary contributor to overall development in college (Astin 1977, 1993a; Pascarella and Terenzini 1991).

         Many studies of middle and secondary schools have taken this more microscopic approach and have focused on circles of friends and acquaintances, not as contextual influences, but as proximal environments of interpersonal interaction and influence.  These studies have demonstrated the influence of friends in school on academic performance, aspirations for college, and educational attainment (Hallinan 1983).  In a study of peer effects on standardized achievement scores, for example, Epstein (1983) found that both initially low- and high-scoring students who had high-achieving friends had higher scores a year later than similar students with low-scoring friends.  With the use of such locally-defined environments, Epstein is able to offer interpretations in terms of interpersonal dynamics that provide deeper insight into the social experiences of children in schools – friends communicate expectations, exchange information, provide social reinforcement, or act as models for observational learning.  Research of this type at the college level generally has not been undertaken since Wallace’s landmark study of student culture in 1966.

         Wallace’s book, Student Culture, clearly illustrated the importance of the college student's interpersonal peer group in influencing members' attitudes towards the attainment of high grades, academic achievement, and aspirations for graduate study (Wallace 1966).  To show these effects, he measured the relationship between the student’s interpersonal environment (Rossi 1966) and change in views and aspirations during the first year of college.  The interpersonal environment for each student was determined not by the researcher, but by each individual respondent.  That is, each student responding to a questionnaire examined a list of names of all students at the college, and beside the name of each student recognized, indicated a degree of like or dislike for that person and the number of hours per week he/she spent time with them.  Clearly, Wallace’s method allows the researcher to access the most proximal of students’ social environments in college and provides a model in which to study interpersonal peer groups.  The downside to this method is that it is methodologically difficult to carry out given the size and complexity of many postsecondary institutions, and correspondingly, may realistically be limited to the study of single institutions.  Perhaps it is because of the greater efficiency of gathering institutionally-based peer data and the increasing interest in between-institution effects that little research along this vein has followed Wallace.  The result, however, is that we know much more about the influence of reference groups on campus and tend to rely on that knowledge to understand the role played by interpersonal peer groups in student learning and development. 

Intellectual self-concept, Educational aspirations, and the “Frog pond” Effect

         Researchers contend that college attendance is a period of psychosocial moratorium (Erikson, 1968), a time when students continually act and react to the environment in the struggle to establish their adult role or identity (Chickering 1969; Sanford 1962).  It is not surprising, then, that studies of peer group influence in college have focused primarily on the effect of the peer environment on academic or intellectual self-concept and educational aspirations (Feldman and Newcomb 1969).  Much of the effort to understand how college affects academic self-concept and educational aspirations has been devoted to peer effects and what has become known as the “frog pond” effect (Pascarella and Terenzini 1991).

            The frog pond effect is based upon Reference Group Theory, which asserts that self-evaluations are based upon an individual’s relative position in comparison to others.  The effect essentially refers to the specific role played by the campus environment in a student’s evaluation of his or her abilities, competencies, and potential.  Two theories have received the most attention in studies of academic self-concept, educational aspirations, and the role of the peer group.  “Environmental press” states that more selective schools will enhance self-concept via the context created by high achieving students (Thistlewaite and Wheeler 1966).  In other words, students will take into account the academic selectivity of their local environment in forming judgements of their abilities.  “Relative deprivation” states that selective schools will detrimentally affect self-concept and aspirations via evaluative standards that are limited to the student’s local peer environment.  Davis (1966) found a negative relationship between career aspirations and a high-achieving student body and invoking relative deprivation, declared that the aphorism, “It’s better to be a big frog in a small pond than a small frog in a big pond,” may hold true in the college environment. 

            Although research generally supports the relative deprivation model, particularly within a high school context (Marsh, 1991), findings have revealed conflicting results.  For example, Drew and Astin (1972), in a national study of academic self-concept, found support for both environmental press theory and the relative deprivation model.  Bassis (1977) found evidence generally supportive of relative deprivation theory, but in a study published the same year, Astin (1977) found that institutional selectivity (a measure of peer ability) had no effect on academic self-concept in either direction.  Although a number of studies at both the college and high school level have continued to explore the relationship between the institutional peer context and self-concept (Astin 1993a; Bachman and O’Malley 1986; Marsh and Parker 1984; Pascarella, Smart, Ethington, and Nettles 1987; Sax 1994) and degree aspirations (Astin 1993a; Heath 1992; Tsui 1995), findings remain somewhat inconclusive (Pascarella and Terenzini 1991).  Contradictory and inconclusive findings may be due, in part, to the almost exclusive focus on the role of institutional reference groups and consequent inattention given to the mediating influence of interpersonal membership groups on student development. 

Racial Diversity and Peer group Research

         Recently, a handful of studies have investigated the role of racial diversity in the student body on development in college.  A primary objective of these studies has been to understand the effects of interacting with someone of another race or ethnicity.  Astin (1993a, 1993b) included interracial interaction among a number of student involvement activities in his multi-institutional studies of student development and found cross-race socialization to be associated with increases in cultural awareness, commitment to racial understanding, and commitment to the environment, as well as higher levels of academic development and satisfaction with college.  Villalpando (1996) and Tanaka (1996) reported similar findings for Chicanos and white students.  In another multi-institutional quantitative study, Chang (1996) found that interracial interaction in college is associated with discussing racial issues, taking ethnic studies courses, and attending racial/cultural awareness workshops.  Furthermore, he demonstrated that these behaviors associated with interracial interaction also enhance student retention, college satisfaction, intellectual self-concept, and social self-concept.

         Scant attention, on the other hand, has been given to racial diversity in peer groups.  A few studies (Astin 1993a; Chang 1996; Hurtado 1990, 1992; Milem 1994; Tanaka 1996) have incorporated percentages of various racial groups in the student body as environmental measures in their models of college impact.  In each of these studies, however, the authors generally interpret these measures as structural characteristics of institutions rather than as measurements of peer group characteristics.  Chang (1996), for example, explored the between-institution effects of the level of racial diversity in the student body or what he called, “structural diversity.”  He found that the higher the degree of structural diversity at an institution, the more likely a student will socialize across race and discuss racial issues.  In using such a distal measure of the environment, Chang’s results could best be interpreted as a structural characteristic that affects opportunities to interact across race rather than as a peer group characteristic that encourages or validates interracial behavior.  Given the long-recognized importance of the peer group in college student development, the centrality of identification and affiliation in group formation and influence, and the continuing interest in racial diversity in higher education, it is surprising that race and ethnicity have surfaced only as structural characteristics of institutions in the area of college impact research.  Again, the dominant analytical strategy of focusing on multi-institutional studies of institutional-based peer groups may have precluded researchers from pursuing more micro-level studies of peer influence processes, whether centered on race or other group characteristics.

         Given the importance of understanding interpersonal peer environments in the context of racial diversity in student development, this study focuses on the college friendship group[1] – a student’s best friends on campus – and its effect on students over time.  The specific questions addressed in this study are:

  (1) To what extent does the interpersonal environment created by the academic abilities and aspirations of the friendship group affect intellectual self-confidence and degree aspirations in college? 

  (2) What role, if any, does the racial diversity of students’ best friends affect the development of intellectual self-confidence and degree aspirations?


Conceptual Framework

         Weidman’s (1989) model of socialization in college is perhaps the most appropriate theoretical model with which to investigate and interpret peer group effects.  My use of Weidman’s model follows similar studies of peer effects by Dey (1996, 1997) and Milem (1998).  Weidman conceptualizes the major influences on student change in college to be pre-college or student background characteristics, the academic and social normative context of an institution, and the impact of parental and non-college reference groups.  Normative contexts are particularly important in Weidman's model for influencing change in personal orientations during college.  However, Weidman also makes three points about the role of the interpersonal environment and interpersonal processes in socialization.  First, he cites Homans (1950, 1961) and argues that the socialization process is quite dependent on interpersonal interaction and the sentimental intensity of the relationship associated with interaction.  Second, he notes that frequency of interaction is also critical.  Lastly, he underscores a conclusion made by a number of researchers, that the long-term academic impacts of college are not the result of classroom experiences, but of informal forms of social interaction with students and faculty. 

         By focusing on friendship groups, this study concentrates on two parts of Weidman’s model, the normative context of informal peer groups and implicitly, the socialization process of interpersonal interaction.  To isolate these elements of the socialization process in college, I borrow from the conceptual and methodological models of college impact of Astin (1984, 1993a), models that are also implicit in Weidman’s (1989) framework.  Astin’s (1993a) model of college impact emphasizes the intercorrelated nature of student pre-college characteristics (inputs) and environmental elements of the college experience.  This relationship becomes problematic when trying to isolate the unique contribution of the educational environment on student outcomes because student inputs are frequently related to both environments and outcomes.  In other words, qualities of the student may explain their eventual outcome (smart students will get high grades) and may also determine the types and nature of their educational experiences (math majors will take more math courses).  In the statistical implementation of Weidman’s socialization model, therefore, I make an effort to properly control for the confounding relationship of inputs to friendship group measures.

Data and Methodology

         Data for this longitudinal, quantitative study were collected during the 1996-1997 academic year at the University of California, Los Angeles (UCLA), a racially and ethnically diverse, public research university[2].  A sample of 2222 third-year students who were previously surveyed as freshmen in 1994 (using a general freshman survey[3]) were surveyed again in the 1996-97 school year with an instrument specifically designed for the study.  A two-wave administration of the survey yielded a response rate of 31 percent.  Four surveys were dropped from the sample because of insufficient or suspect data, yielding a final sample for analysis of 677 respondents.  Because a variety of pre-college data was available for both respondents and non-respondents (taken from the 1994 freshman survey), it was possible to predict the probability of responding to the follow-up survey based on those data and calculate a weighting factor to partially adjust for non-response bias.  Using a procedure described by Astin and Molm (1971), weights were calculated such that the data of respondents who appeared most similar to non-respondents in terms of pre-college characteristics (i.e., respondents with lower probabilities of responding) received larger weights relative to those students whose characteristics resembled those most likely to respond.  Weighted data was used in all analyses[4].

         The follow-up instrument collected demographic data, measures of behavior and involvement in activities in college, and data on various outcome measures including self-rated abilities and highest level of degree aspirations.  Most significant to this study, the names of fellow students whom students identified as members of their friendship group were also collected.  Respondents were asked to name up to seven UCLA students with whom they spent most of their time and who they considered to be their “best friends” on campus.  They also identified the racial/ethnic composition of their friendship group.  The written names were used to retrieve data on friendship group members collected by the annual freshman surveys.  Aggregates of the friends’ freshman survey data for each identified friendship group were computed and operationalized as measures of actual friendship group characteristics.  Because it was not possible to obtain freshman data for all friendship group members, only respondents with sufficient friendship group data were retained for analyses involving friendship group measures[5].  In these analyses, the sample size was reduced to 426 students.

         The two dependent variables are single item measures taken from the follow-up survey.  Academic self-concept was measured with a traditional self-rated ability question that asked the student to rate her “self-confidence (intellectual)” as compared to “the average person your age.”  The rating was made on a 5-point scale (“lowest 10%” to “highest 10%”).  A separate question on the survey asked students to report the highest academic degree they intend to obtain and was scored on a 4-point scale (“none” to “Ph.D/Ed.D, M.D., J.D”).  Both variables were pretested prior to college entry in 1994 with similar measures.

         The independent variables derived from the surveys are listed in the Appendix in Table A.  The pre-college data were collected by the freshman survey in 1994.  These measures include the relevant pretest measure for each analysis, gender (female), race/ethnicity, socioeconomic status, and a measure of academic ability, the student’s SAT score.  Table 1 contains the means and standard deviations.


insert table 1 about here



         Five friendship group measures were chosen for the model.  Three variables are aggregate measures and include group averages of intellectual self-confidence in 1994, SAT composite scores, and degree aspirations in 1994.  Note that it was necessary to include the corresponding individual-level variables for these measures in the regression model to reduce self-selection effects that confound relationships between group-level measures and the dependent variables (Burstein 1980).  Appropriately, each of the pretest measures, in addition to SAT score, was included in the analysis of both dependent variables.

         The racial composition of each student’s friendship group was collected with the follow-up survey and used to calculate a measure of the racial diversity of the friendship group.  Racial diversity of the friendship group was measured on a four-point scale.  The degree of racial diversity was defined by the percentage of the largest racial or ethnic group represented in the friendship group:

(1) Homogeneous -- the largest racial/ethnic group makes up 100% of the friendship group

(2) Predominantly one race/ethnicity -- the largest racial/ethnic group makes up 75 - 99% of the friendship group

(3) Majority one race/ethnicity -- the largest racial/ethnic group makes up 51 - 74% of the friendship group

(4) No majority -- the largest racial/ethnic group makes up 50% or less of the friendship group


These definitions were applied only to friendship groups consisting of two or more students.    Note that in the particular sociopolitical context at UCLA, racial diversity was defined by students at the level of ethnicity (Author 1998).  Appropriately, the various Asian American and Latino ethnic groups were not collapsed into broad, singular categories for the purposes of defining racial diversity within friendship groups.  Pan-Asian groups consisting of one Chinese American, one Filipino, and one South Asian student, for example, were not defined as homogeneous but as having no majority racial/ethnic group.

         The final three variables in the model incorporate one of Weidman’s primary mechanisms of socialization, interaction among students.  A composite variable of three “time diary” items (studying, partying, and talking with students) provides a general measure of student interaction.  Two additional variables measure the frequency of one specific type of interaction hypothesized to be related to both intellectual self-confidence and educational aspirations, having conversations about homework or classwork (with best friends and with other students).

         The primary set of analyses featured blocked multiple regression procedures to estimate the relationship between the outcome measures and the five friendship group characteristics while holding constant pre-college characteristics and 1994 pretests of intellectual self-confidence and degree aspirations.  Independent variables were entered in three discrete blocks for all equations, in accordance with the college impact and socialization models of Astin (1984, 1993a) and Weidman (1989).  Pre-college characteristics were entered into the regression equation first, followed by the block of friendship group measures and subsequently, the measures of college involvement.  Since preliminary analyses indicated a strong, statistically significant interaction between friendship group diversity and race, separate analyses were conducted for white students (n=151) and students of color[6] (n=285).  Because both individual-level and group-level variables were modeled simultaneously, potential multicollinearity problems in the regression analyses were monitored by reviewing the intercorrelations among independent variables, tolerance indices, and corresponding variance inflation factors[7].


         There are a few key limitations that the reader should keep in mind when reviewing the results of this study.  First, the sample drawn for this study represents a single campus with a large enrollment (~36,000), a particularly diverse student body, and other specific characteristics.  It is not known whether the relationships found in this study generalize to other four-year institutions.  An anonymous reviewer familiar with UCLA, for example, noted that a large proportion of third-year students and older live off-campus and therefore may retain off-campus friendships as well.  These non-college reference groups are not considered here and may also be influential.  Second, because freshman survey data were used to generate friendship group measures, the design of the study also did not allow for transfer students as members of students' friendship groups, which may eliminate members of some students' interpersonal environments and bias friendship group measures.  As noted above, freshman data was not available for every best friend named by respondents, which also may introduce error into the calculation of friendship group measures.  Lastly, the single-item measures for degree aspirations and intellectual self-confidence used in this analysis are not easily generalized to higher-order constructs and therefore should be viewed with caution. 



         Before reporting the results of the multivariate analyses, it is instructive to examine a number of bivariate relationships between the dependent variables and key independent variables.  Table 2 displays distributions of the two dependent variables with two student characteristics, gender and race.  Statistically significant differences were found for intellectual


insert table 2 about here



self-confidence.  While a large majority of men (81 percent) rate themselves highly in intellectual self-confidence, a smaller proportion of women (64 percent) rate themselves similarly.  At the lower end of the scale, women are more than twice as likely than men to report themselves among the lowest in terms of intellectual ability, though this difference is not significant at the 0.05 level.  Similarly, white students are much more likely than are students of color to rate themselves highly on intellectual self-confidence.  No significant race or gender differences were found with respect to student’s highest degree aspirations.

         The major premise of this study is that elements of the interpersonal environment are important influences on socialization in college.  Table 3 provides a simple illustration of the nature of these influences.  For the purposes of comparison, three dichotomous friendship group


insert table 3 about here



variables were created.  For both intellectual self-confidence and degree aspirations, friendship groups were classified as “high” or “low,” based upon whether the score for each group measure was above or below the sample mean for each respective variable.  Columns 1 through 4 in Table 3 clearly exhibit the same kind of patterns found among high school students by Epstein (1983).  Students who have best friends with relatively high levels of intellectual self-confidence tend to be more self-confident intellectually after two years of college compared to students with less confident friendship groups.  A similar relationship between individual and group characteristics is evident with respect to degree aspirations.  Interpersonal environments that are high or strong in a particular quality, characteristic, or trait, appear to enhance that same quality among students over time.  Note that these patterns, while compelling, do not take into account entering student characteristics and may mask the effects of self-selection into particular kinds of interpersonal environments (Cohen 1985; LeVine 1966). 

         The remaining two columns in Table 3 test to see whether a group characteristic other than one directly analogous to the outcome also can be a potentially important socializing influence.  Comparing students with a low level of diversity in their friendship group ("homogeneous" groups) to their counterparts with relatively higher levels ("no majority" groups), we find no statistically significant differences in intellectual self-confidence or degree aspirations.  Some relationship is implied, however, with degree aspirations.  While about 11 percent of students who have diverse friendship groups restrict their educational aspirations to the baccalaureate degree, a larger proportion (18 percent) have similarly low aspirations among students with homogeneous friendship groups.

         As noted above, preliminary regression analyses of the two outcome measures indicated a statistical interaction between race and diversity of the friendship group.  Figure 1 depicts the interaction graphically.  The graphs below show how diversity in the friendship group essentially has the opposite relationship with intellectual self-confidence and highest degree aspirations for white students and students of color.  For white students, those who have a higher degree of diversity in their friendship group tend to be less self-confident and have lower educational aspirations than do those with homogeneous groups.  For students of color, diversity is associated with enhanced self-confidence and aspirations.  The extent to which these differential patterns persist after taking into account initial differences in the dependent variables is discussed next.


insert figure 1 about here



         Table 4 contains the standardized regression coefficients for the regression models of intellectual self-confidence.  Three regression models each were computed for white students and students of color.  Model 1 coefficients reflect the effects associated with each independent variable after controlling for pre-college characteristics only.  For the variables not in the model, the coefficient each variable would have received if added to Model 1 are also provided in brackets.  Friendship group measures are added in Model 2, providing estimates of their effects independent of each other.  Model 3 coefficients additionally take into account measures of student involvement.


insert table 4 about here



         Elements of the interpersonal environment of the friendship group exhibit significant relationships with intellectual self-confidence for both white students and students of color.  The effects, however, are quite distinct between the two groups of students.  Among students of color, the group level of intellectual self-confidence has the positive effect consistent with the notion of environmental press; they appear to benefit psychologically from interaction within a highly confident set of best friends.  For white students, a positive influence on intellectual self-confidence appears to emanate more from high group levels of educational aspirations than with high group self-confidence.  The simple correlations indicate that group levels of both intellectual self-confidence and degree aspirations are associated with intellectual self-confidence midway through college.  Controlling for individual-level variables diminishes the association with group intellectual self-confidence (evident by comparison to the bracketed coefficient in model 1), leaving group degree aspirations as the only positive group effect.  Unlike the case of students of color, white students’ friendship groups also exhibit negative effects.  The depressive effect of group SAT score is indicative of Davis’ (1966) classic relative deprivation interpretation, in which students are likely depressing their self-evaluations in the presence of high-achieving friends.

         As suggested by figure 1 above, racial and ethnic diversity in the friendship group has a positive effect on intellectual self-confidence for students of color.  Among white students, friendship group diversity is negatively correlated with intellectual self-confidence (partial correlation=-.21, p<.05, after controlling for the pretest only) but fails to gain significance in any of the regression models.  For students of color, a diverse interpersonal environment of friends appears to enhance intellectual self-confidence regardless of the academic ability, educational trajectories, or degree of self-confidence possessed by themselves or by their closest friends.  For white students, friendship group diversity, at best, has no bearing on their intellectual self-confidence. 

         It is worth noting that the role of SAT scores as a predictive characteristic is markedly different as well for the two groups of students.  At both the individual and group levels, SAT scores are closely associated with white students’ sense of their intellectual abilities.  High scores at the individual level enhance self-confidence and as we have seen, group level scores appear to have a relative deprivation type of effect.  The same variables show no effects, positive or negative, among students of color.

         Lastly, measures of involvement have no significant effect on intellectual self-confidence for either group of students, holding constant pre-college variables and friendship group characteristics.

         Table 5 tells a similar story with respect to educational aspirations.  Among white students, high aspirations are associated with initially high aspirations as freshmen and by having a highly self-confident friendship group.  Despite a negative value for the bivariate correlation (r = -0.17, p< 0.05), racial diversity in the friendship group again appears to be unrelated to the outcome for white students.  Again, a slightly different pattern of effects is suggested by the data for students of color.  No relationships were found between any of the three academically-oriented friendship group characteristics and educational aspirations among students of color.  Diverse interpersonal environments also appear to have beneficial effects on aspirations, although this relationship should be regarded with caution since the coefficients do not quite reach significance at the conventional 0.05 level and the change in explained variance (adjusted R2) from models 1 to 2 is negligible.


insert table 5 about here




         In studying the interpersonal environment of the friendship group, I believe this study serves as a meaningful call to re-focus empirical and theoretical treatments of college peer group influence.  The peer group effects found in this study are convincing evidence that the micro-level interpersonal environments of a college campus are important sites of influence on socialization and student development.  The supposition by researchers that interpersonal environments mediate institutional-level peer group effects is strongly supported by this research, and further, the complexity of the findings underscore a need for researchers and administrators to better understand the role of micro-environments in socialization in college.

         With regard to theory, evidence of both relative deprivation and environmental press was found to operate simultaneously at the interpersonal level.  However, different aspects of the interpersonal environment accounted for each type of influence.  In the analysis of intellectual self-confidence, group SAT scores had a depressive effect while educational aspirations had an enhancing effect.  The inclusion of multiple measures of the interpersonal environment reveals that different but related aspects of the peer environment can have opposite effects.  In addition, these results may explain why previous research has been somewhat inconclusive on this theoretical point.  First, variation of peer effects at the interpersonal level may cancel out overall peer effects at the institutional level and yield no net effect.  This explanation is consistent with Marsh’s (1984) frame of reference interpretation on the effects of ability grouping on the academic self-concept of school children.  Second, reliance on a single aspect of the peer environment at any level may neglect the effects of other peer characteristics and essentially reveals only one of many distinct yet interrelated processes.  Marsh (1991) demonstrated this point within a high school context where he found school-average SES and school-average ability to have opposite effects on both educational aspirations and academic self-concept.  He speculated that the positive effects of school-average SES were due to group identification processes while the negative effects of school-average ability were the result of social comparison (relative deprivation) processes.  The results of the current study suggest that both types of processes operate at the interpersonal level.  It is important to note that these opposing peer effects were found for the white student sample only.  Evidence of relative deprivation in the interpersonal context was not found among students of color, and the extent to which this conclusion applies across race is unknown.  Finally, while the current study does not make an attempt to compare the relative influence of membership groups (best friends) to reference groups (the campus peer group), the variation of effects found within the white student sample and between white students and students of color suggests that membership groups may not merely mediate campus peer influence, they may serve to isolate members from more distal institutional influences as well. 

         These findings, coupled with the positive effects of racial diversity evident for students of color only, suggest that the peer factors that influence students’ intellectual self-confidence and degree aspirations operate differentially by race.  In their comprehensive study of self-concept development among college students, Pascarella, Smart, Ethington, and Nettles (1987) found few racial differences among factors influencing self-concept.  The few differences found by the researchers were in the differential impact of various behavioral activities while in college by white and black students.  Furthermore, they noted that there were no patterns in these differential effects by race.  The variations found between white students and students of color in this study suggest that the factors which produce differential patterns of effects on self-concept may originate in the frequently unmeasured interpersonal environment of students.

         The methodological focus of this study validates the applicability and feasibility of conducting quantitative, longitudinal studies of the interpersonal environment on large, sociologically complex campuses.  Studies of the influence of friends are typically limited by cross-sectional designs and/or simultaneous measurement of student and friendship group characteristics (Cohen 1983).  In taking advantage of a national freshman survey administered prior to college (during orientation), I have been able to avoid many of such inferentially confounding pitfalls for most group measures.  Most significantly, with the use of student-defined friendship group measures, we are assured of capturing actual environments of interaction, and consequently, more accurately assessing the impact of the environment.

         The results also raise interesting questions with regard to diversity.  The assessment of the influence of racial diversity in the interpersonal environment showed that diversity is an important peer characteristic to consider along with traditional measures of peer ability and self-concept.  A previous study showed that racial diversity in the friendship group is important for increasing a student’s commitment to racial understanding and is associated with interracial interaction outside of the friendship group (Author 1998).  The present study indicates that racial diversity is also important when examining academically related cognitive outcomes.  While it is important to recognize that diversity does have an effect on academically oriented outcomes, what is missing from this discussion is a theory of how diversity operates in the context of academics.  In the case of interracial interaction and racial understanding, the mechanism appears to be the exposure and dealing with issues of racism, discrimination, and cultural difference (Author 1998).  The connection that interracial interaction and friendships has to academic outcomes is less clear.

         The positive effect of friendship group diversity on intellectual self-confidence and (more tentatively) educational aspirations was found for students of color only, and the absence of a similar effect among white students can help us think about the relationship between diverse friendships and academic outcomes.  In the realm of self-concept and aspirations, diversity may simply provide students – students of color – a normative context which contains more varied reference points from which to evaluate themselves.  Under this interpretation, diversity in the friendship group presents students with multiple referents with respect to academic ability, and the presence and tacit acceptance of cultural diversity supports the legitimacy of adhering to multiple norms while remaining a cohesive group.  The standard deviation in SAT scores among best friends who are members of the more homogenous friendship groups (s=121), for example, is smaller than among best friends in the more diverse groups (s=131) in this sample.  Alternatively, a racially diverse comparative context may reduce a devaluation of ability among students of color due to “stereotype threat” (Steele 1997) that may be triggered in predominantly white settings and in this manner, function to enhance self-esteem.  In this interpretation, racially diverse friendship groups act as enclaves of safety against threats to self-esteem in the greater environment.  Finally, perhaps there is simply an environmental press effect for students of color because they are validated by interacting closely with non-white students with high (relative to stereotypical assumptions) aspirations and competencies.  The combination of this validation with the re-framing of their psyche in a non-white frame may make group diversity as influential, and in some cases, more influential than academic competencies or self-esteem in the group, as the findings indicate.

         The results for white students also raise questions and suggest directions for future study.  Are white students’ academic self-beliefs and aspirations unaffected by racial diversity?  Results of the current study imply that racial diversity is not a salient environmental characteristic in academic domains for white students.  In fact, the data suggest a negative effect for diversity on intellectual self-confidence.  This result contradicts the findings of Chang (1996) who found interracial interaction among white students to enhance intellectual self-concept.  These discrepant findings indicate a need to probe deeper into the friendship groups of white students and understand the differences in interaction within racially diverse groups compared to more homogeneous ones.  The differences may lay in the nature and quality of relationships among best friends in different types of friendship groups.  For example, Author (1998) found descriptions of mutual trust and emotional closeness within friendship groups to be more common among students with more homogenous friendship groups than those with relatively diverse groups.  

           Finally, why do white students appear more susceptible to the effects of relative deprivation on intellectual self-confidence than do students of color?  The differential effects of SAT scores at both the individual and group levels suggest that SAT scores may carry heavier psychological weight for constructing self-concept among white students as compared to students of color.  At the group level, it is difficult to exactly determine what group average SAT scores represent.  For white students, higher group SAT scores may be a measure of academic ability, competitiveness, or perhaps, academic stress.  Future research that “unpacks” the operational meaning of this classic peer measure in the context of the friendship group will help us to further understand the mechanism of relative deprivation and the differential effects observed in this study.


Altbach, Philip.G.  1991.  “The racial dilemma in American higher education.”  Pp. 3-17 in The Racial Crisis in American Higher Education, edited by Philip G. Altbach and Kofi Lomotey.   Albany, NY:  State University of New York Press.

Alwin, Duane F., and Luther B.Otto. 1977.  “High school context effects on aspirations.” Sociology of Education 50:259-273.

Author.  1996.  “Racial balkanization, multiculturalism, and student development.”  Unpublished manuscript.

------.  1998.  “The Impact of Friendship Groups in a Multicultural University.  Unpublished doctoral dissertation, University of California, Los Angeles.

Astin, Alexander W.  1977.  Four Critical Years.  San Francisco:  Jossey-Bass.

------.  1984.  “Student involvement:  A developmental theory for higher education.”  Journal of College Student Personnel 25:297-308.

------.  1993a.  What Matters in College?  Four Critical Years Revisited.  San Francisco:  Jossey-Bass.

------.  1993b.  “Diversity and multiculturalism on campus:  How are students affected?”  Change 25:44-49.

Astin, Alexander W., and Linda D. Molm. 1971.  “Correcting for response bias in follow-up surveys.”  Unpublished manuscript, Graduate School of Education, University of California, Los Angeles.

Bassis, Michael S.  1977.  “The campus as a frog pond:  A theoretical and empirical reassessment.”  American Journal Sociology 82:1318-1326.

Burstein, Leigh.  1980.  “The analysis of multilevel data in educational research and evaluation.” Review of Research in Education 8:158-233.

Clark, Burton R. and Martin Trow.  1966.  “The organizational context.”  Pp. 17-70 in College Peer Groups, edited by Theodore M. Newcomb and Everett K. Wilson.  Chicago:  Aldine.

Chang, Mitchell J.  1995.  “Race identity and race relations in higher education:  Fraternity and sorority membership among students of color.”  Presented at the annual meeting of the Association for the Study of Higher Education, Orlando, FL, November.

-------.  1996.  “Racial diversity in higher education:  Does a racially mixed student population affect educational outcomes?”  Unpublished doctoral dissertation, University of California, Los Angeles.

Cohen, Jere.  1983.  “Commentary:  The relationship between friendship selection and peer influence.”  Pp. in Friends in School.  Patterns of Selection and Influence in Secondary Schools, edited by Joyce L. Epstein and Nancy L. Karweit.  New York:  Academic Press.

Davis, James.  1966.  “The campus as a frog pond:  An application of the theory of relative deprivation to career decisions of college men.”  American Journal of Sociology 72:17-31.

Dey, Eric L.  1996.  “Undergraduate political attitudes:  An examination of peer, faculty, and social influences.”  Research in Higher Education 37:535-554.

------.  1997.  “Undergraduate political attitudes: Peer influence in changing social contexts.”  Journal of Higher Education 68:398-413.

Drew, David E., and Alexander W Astin.  1972.  “Undergraduate aspirations:  A test of several theories.”  American Journal of Sociology 77:1151-1164.

Epstein, Joyce L.  1983.  “The influence of friends on achievement and affective outcomes.”  Pp. in Friends in School.  Patterns of Selection and Influence in Secondary Schools, edited by Joyce L. Epstein, and Nancy L. Karweit.  New York:  Academic Press.

Erikson, Erik.  1968.  Identity:  Youth and Crisis.  New York:  Horton.

Feldman, Kenneth A., and Theodore M. Newcomb.  1969.  The Impact of College on Students.  San Francisco:  Jossey-Bass.

Hallinan, Maureen T.  1983.  “Commentary:  New directions for research on peer influence.”  Pp. in Friends in School.  Patterns of Selection and Influence in Secondary Schools, edited by Joyce L. Epstein, and Nancy L. Karweit.  New York:  Academic Press.

Heath, Tamela.  1992.  “Predicting the educational aspirations and graduate plans of black and white college and university students: When do dreams become realities?”  Presented at the annual meeting of the Association for the Study of Higher Education, Minneapolis, MN, October. 

Homans, George C.  1950.  The Human Group.  New York:  Harcourt, Brace, and World.

------.  1961.  Social Behavior: Its Elementary Forms.  New York:  Harcourt, Brace, and World.

Hurtado, Sylvia, and Deborah F. Carter.  1997.  “Effects of college transition and perceptions of the campus racial climate on Latino college students’ sense of belonging.”  Sociology of Education 70:324-345.

Justiz, Manual J.  1994.  “Demographic trends and the challenges to American higher education.”  Pp. in Minorities in Higher Education, edited by Manual J. Justiz, Reginald Wilson, and Lars G. Björk.  Phoenix, AZ:  Oryx Press.

Kerr, Clark.  1991.  “Foreward.”  Pp. vii-x in The Racial Crisis in American Higher Education, edited by Philip G. Altbach and Kofi Lomotey.  Albany, NY:  State University of New York Press.

LeVine, Robert A.  1966.  “American college experience as a socialization process.”  Pp. 107-132 in College Peer Groups, edited by Theodore M. Newcomb and Everett K. Wilson.  Chicago:  Aldine.

Marsh, Herbert W.  1991.  “Failure of high-ability high schools to deliver academic benefits commensurate with their students' ability levels.”  American Educational Research Journal 28:445-480.

Marsh, Herbert W.  1984.  “Self-concept, social comparison, and ability grouping:  A reply to Kulik and Kulik.”  American Educational Research Journal 21:799-806.

Milem, Jeffrey F.  1994.  “College, students, and racial understanding.”  Thought and Action 9: 51-92.

-------.  1998.  “Attitude change in college students:  Examining the effect of college peer groups and faculty normative groups.”  Journal of Higher Education 69:117-140.

Pascarella, Ernest T.  1985.  “College environmental influences on learning and cognitive development:  A critical review and analysis.”  Pp. in Higher Education:  Handbook of Theory and Research.  Volume 1, edited by John C. Smart. New York:  Agathon.

Pascarella, Ernest T., Marcia Edison, Amaury Nora, Linda S. Hagedorn, and Patrick T. Terenzini.  1996.  “Influences on students’ openness to diversity and challenge in the first year of college.”  Journal of Higher Education 67:174-195.

Pascarella, Ernest T., John C. Smart, Corina A. Ethington, and Michael T. Nettles.  1987.  “The influence of college on self-concept:  A consideration of race and gender differences.”  American Educational Research Journal 24: 49-77.

Pascarella, Ernest T., and Patrick T. Terenzini.  1982.  “Contextual analysis as a method for assessing residence group effects.”  Journal of College Student Personnel 23:108-114.

------.  1991.  How College Affects Students.  San Francisco:  Jossey-Bass.

Rossi, Peter H.  1966.  “Research strategies in measuring peer group influence.” Pp. 190-214 in College Peer Groups, edited by Theodore M. Newcomb and Everett K. Wilson.  Chicago:  Aldine.

Sax, Linda J.  1994.  “Mathematical self-concept:  How college reinforces the gender gap.”  Research in Higher Education 35:141-166.

------.  1996.  “The dynamics of tokenism:  How college students are affected by the proportion of women in their major.”  Research in Higher Education 37:389-425.

Steele, Claude M. 1995.  “Stereotype Threat and the Intellectual Test Performance of African Americans.” Journal of Personality and Social Psychology 69:797-811.

Tanaka, Gregory K.  1996.  “The Impact of Multiculturalism on White Students.”  Unpublished doctoral dissertation, University of California, Los Angeles.

Thistlethwaite, Donald L., and Norman Wheeler.  1966.  “Effects of teacher and peer subcultures upon student aspirations.”  Journal of Educational Psychology 57:35-47.

Tinto, Vincent.  1975.  “Dropout from higher education:  A theoretical synthesis of recent research.”  Review of Educational Research 45:89-125.

Tsui, Lisa.  1995.  “Boosting female ambition: How college diversity impacts graduate degree aspirations of women.”  Presented at the annual meeting of the Association for the Study of Higher Education, Orlando, FL, November.

Villalpando, Octavio D.  1996. “ The long-term effects of college on Chicano and Chicana students' "other-oriented" values, service careers, and community involvement.”  Unpublished doctoral dissertation, University of California, Los Angeles.

Wallace, Walter L. 1966.  Student Culture.  Chicago:  Aldine.

Weidman, John C.  1989.  “Undergraduate socialization:  A conceptual approach.”  Pp. 289-322 in Higher Education:  Handbook of Theory and Research.  Volume V, edited by John C. Smart. New York:  Agathon.

Werts, Charles E., and Donivan J. Watley.  1969.  “The student's dilemma:  Big fish-little pond or little fish-big pond.”  Journal of Counseling Psychology 16:14-19.

Table 1

Means and standard deviations

                                                                                                               Mean (std dev)                

Variable                                                                                White Students       Students of Color

n                                                                                                   151                        285

Pre-college characteristics

Intellectual self-confidence (pretest)                                               4.08 (0.67)             3.78 (0.71)

Highest degree aspirations (pretest)                                               5.47 (0.69)             5.46 (0.67)

Gender-female                                                                              1.52 (0.50)             1.50 (0.50)

SES                                                                                           23.02 (4.43)           18.60 (7.02)

SAT composite score                                                                 1209 (94)               1162 (128)


Friendship Group Characteristics

Group intellectual self-confidence                                                  4.03 (0.53)             3.82 (0.42)

Group degree aspirations                                                              4.48 (0.43)             4.52 (0.37)

Group SAT composite score                                                      1202 (72)                1178 (102)

Friendship Group Racial Diversity                                                 2.49 (1.17)             3.18 (1.04)


College involvement variables

Student-student Interaction                                                         14.71 (3.58)           13.01 (3.84)

Conversations about classwork (w/friendship group)                     2.65 (0.53)          2.57 (0.54)

Conversations about classwork (outside friendship group)             2.32 (0.65)          2.08 (0.62)


Table 2

Intellectual self-confidence and highest degree aspirations in 1996-97 by gender and race

Percent among

                                                                                                                        t-test         White           Students      t-test

                                                                                    Men      Women        p-val         Students      of Color        p-val

Intellectual Self-Confidence

“Above Average” or “Highest 10%”         80.5     63.9    <.001          77.8          68.8        .039

“Below Average” or “Lowest 10%”           1.9       4.8      .089            3.0            3.6        .726


Degree Aspirations

Doctorate, Medical, or Law Degree         50.4     50.2     .969          48.6          51.2        .605

Bachelor’s Degree                                   10.3     12.4     .467          13.5          10.4        .343


n                                                               210      227                        151           285


Table 3

Intellectual self-confidence and highest degree aspirations in 1996-97

by friendship group characteristics

Percent among

                                                                 Grp Intel S-C        Grp Deg Asp       Grp Diversity

                                                                 Low       High        Low       High        Low       High

Intellectual Self-Confidence

“Above Average” or “Highest 10%”          67.0       77.2a       69.9       73.9        74.7       72.9

“Below Average” or “Lowest 10%”             4.0         2.7          3.6         3.1          3.7         4.4


Degree Aspirations

Doctorate, Medical, or Law Degree           50.4       50.3        45.6       55.3a       59.2       54.1

Bachelor’s Degree                                     12.9       10.0        10.8       12.5        18.0       10.9


n                                                                229        207         235        198           79        184

aStatistically significant t-test indicating difference between low and high groups, p<.05


Figure 1

                        Friendship Group Diversity                                                            Friendship Group Diversity



Table 4

Regression of Intellectual Self-Confidence on Pre-College Characteristics, Friendship Group Characteristics, and Involvement Variables

                                                                                                     White Students                                 Students of Color           

                                                                                          r     model 1  model 2  model 3          r     model 1  model 2  model 3

Pre-college characteristics and Control variables

Gender:  female                                                            -.30      -.13          -.16          -.15            -.18      -.08          -.06        -.06

SES                                                                                 .27       .18*         .17*         .16*           .02       .05           .05         .05

Intellectual self-confidence in 1994                           .49       .37***    .37***    .38***      .44       .41***    .40***  .41***

Highest degree aspirations in 1994                           .33       .14           .05           .03             .05       .02           .03         .01

SAT score                                                                     .27       .15*         .17*         .17*           .19       .08           .05         .05


Friendship Group characteristics

Group Intellectual self-confidence                            .27      [.03]          .03           .02             .19      [.12*]       .12*       .13*

Group degree aspirations                                           .17      [.10]          .15*         .17*           .05      [.00]        -.02        -.01

Group SAT score                                                        -.04    [-.15]        -.21*       -.20*           .08      [.03]          .03         .02

Group Racial/ethnic diversity                                   -.28    [-.14]        -.13          -.12             .12      [.12*]       .12*       .12*


Involvement variables

Student-student interaction                                       .19      [.09]        [.08]          .08            -.07    [-.07]       [-.06]      -.07

Classwork conv'ns w/friendship group                    .07      [.05]        [.03]          .00            -.06    [-.00]       [-.01]      -.01

Classwork conv'ns outside of friendship group     .05    [-.01]        [.00]        -.01            -.01      [.04]        [.05]        .06


R2                                                                                                                                               .354         .413         .419                       .209         .238       .244

Adjusted R2                                                                                                                    .326         .366         .354                       .193         .209       .206

Model 1: standardized coefficients after controlling for pre-college characteristics

Model 2: standardized coefficients after controlling for pre-college characteristics and friendship group characteristics

Model 3: standardized coefficients after controlling for pre-college characteristics, friendship group characteristics, and involvement variables

***p<.001, **p<.01, *p<.05


Table 5

Regression of Highest Degree Aspirations on Pre-College Characteristics, Friendship Group Characteristics, and Involvement Variables

                                                                                                     White Students                                 Students of Color           

                                                                                          r     model 1  model 2  model 3          r     model 1  model 2  model 3

Pre-college characteristics and Control variables

Gender:  female                                                            -.07       .02           .09           .09            -.01      -.02          -.02        -.01

SES                                                                                 .03      -.05          -.07          -.08            -.10      -.10          -.08        -.07

Intellectual self-confidence in 1994                           .29       .20*         .16           .17             .19       .10           .10         .09

Highest degree aspirations in 1994                           .42       .41***    .37***    .37***      .40       .39***    .39***  .40***

SAT score                                                                    -.03      -.14          -.17          -.17             .08       .11           .11         .13


Friendship Group characteristics

Group Intellectual self-confidence                            .25      [.26**]     .25*         .23*         -.04      [.01]          .01         .00

Group degree aspirations                                           .24      [.16]          .14           .14             .09      [.02]          .01         .00

Group SAT score                                                        -.03      [.00]        -.11          -.11             .00    [-.02]        -.02        -.03

Group Racial/ethnic diversity                                   -.17    [-.08]        -.03          -.03             .15      [.11]        .11        .11


Involvement variables

Student-student interaction                                       .12      [.03]        [.02]          .08            -.12    [-.04]       [-.03]      -.02

Classwork conv'ns w/friendship group                    .11      [.11]        [.04]          .00            -.03      [.03]        [.05]        .07

Classwork conv'ns outside of friendship group     .02    [-.06]       [-.02]        -.01            -.08    [-.12*]     [-.11]      -.12


R2                                                                                                                                               .232         .303         .305                       .196         .209       .224

Adjusted R2                                                                                                                    .198         .246         .228                       .180         .180       .185

Model 1: standardized coefficients after controlling for pre-college characteristics

Model 2: standardized coefficients after controlling for pre-college characteristics and friendship group characteristics

Model 3: standardized coefficients after controlling for pre-college characteristics, friendship group characteristics, and involvement variables

***p<.001, **p<.01, *p<.05, p=.06




Table A

Variables in the regression model

Pre-college characteristics

Intellectual self-confidence pretest                                       5 pt scale, “lowest 10%” to “highest 10%”

Highest degree aspirations pretest                                       6 pt scale, “none” to “Ph.D/Ed.D., M.D., J.D.”

Gender-female                                                                           1-male, 2-female

SES                                                                                             3 item composite (a=.803) composed of:

   mother’s education (self-report)                                         8 pt scale, “grammar school or less” to

                                                                                                                      “graduate degree”

   father’s education (self-report)                                           8 pt scale, “grammar school or less” to

                                                                                                                      “graduate degree”

   family income (self-report)                                                   14 pt scale, “less than $6000” to “over $200,000”

SAT composite score (self-report)                                        continuous


Friendship group measures

Group intellectual self-confidence in 1994                           continuous (group average)

Group degree aspirations in 1994                                          continuous (group average)

Group SAT composite score                                                  continuous (group average)

Racial diversity of friendship group                                     4 pt scale, “homogeneous” to “no majority/mixed”


College involvement measures

Student-student Interaction                                                   3 item composite (a=.660) composed of:

   studying with other students                                             8 point scale, "None" to ">20" hours per week

   partying with other students                                              8 point scale, "None" to ">20" hours per week

   talking with students outside of class                               8 point scale, "None" to ">20" hours per week


Conversations about classwork:                                          

   w/students in friendship group                                          3 pt scale, “Not at all” to “Frequently”

   w/students outside of friendship group                           3 pt scale, “Not at all” to “Frequently”


[1] The term, “friendship group” is used here to mean the interpersonal environment composed of a student’s best friends on campus.  With this definition, best friends may form a singular, cohesive group or a more diffuse friendship network.  The term is used in lieu of the somewhat more cumbersome, "interpersonal environment of best friends."

[2] The undergraduate student body at the time of the study was approximately 40% white, 35% Asian American, 16% Latino, and 6% African American.

[3] UCLA participates in the Cooperative Institutional Research Program, a national freshman survey administered annually prior to the beginning of the school year.

[4] Analyses were conducted on both weighted and unweighted data.  Since the results of the two analyses were quite similar and substantively consistent, the weighted analyses are presented because they reflect an attempt to correct for non-response bias.

[5] Parametric analyses were conducted to determine sufficiency with regard to friendship group data wherein the total number of friends with freshman data and the percentage of named friends with freshman data were varied independently and the correlations between a number of group-level and individual-level variables of the same measure (e.g., student SAT score and group SAT score) were examined for stability.  Stability was observed when at least 50% of the named students or at least three friends in a respondent's friendship group had freshman data.  See Author (1998) for a complete description of the development of friendship group aggregates.

[6] In this study, students of color were defined as students not self-identifying as “white/caucasian.”

[7] Correlations among independent variables were all less than 0.35 except for two exceptions:

(1) For students of color, student's SAT score correlates with SAT score of the friendship group, r=0.44.  This moderate correlation is not unexpected given friendship theories of homophilic selection.

(2) For white students, being a woman is negatively correlated with friendship groups' intellectual self-confidence, r = -0.46

In both of these cases, large suppressor effects due to intercorrelated independent variables were not apparent.


Most tolerance indices were in the .80 to .95 range.  The lowest indices were calculated to be 0.67 (corresponding VIF=1.49) for student SAT score among students of color and 0.56 (corresponding VIF=1.78) for group intellectual self-confidence among white students.  These values are both exceptions in the analyses, and are easily outside of the commonly accepted range that is indicative of a multicollinearity problem.  Details of these analyses are available from the author.