journal of family issues - sage publications
TRANSCRIPT
http://jfi.sagepub.com
Journal of Family Issues
DOI: 10.1177/0192513X03258307 2004; 25; 571 Journal of Family Issues
Robert Crosnoe and Glen H. Elder, JR. Family Dynamics, Supportive Relationships, and Educational Resilience During Adolescence
http://jfi.sagepub.com/cgi/content/abstract/25/5/571 The online version of this article can be found at:
Published by:
http://www.sagepublications.com
On behalf of: National Council on Family Relations
can be found at:Journal of Family Issues Additional services and information for
http://jfi.sagepub.com/cgi/alerts Email Alerts:
http://jfi.sagepub.com/subscriptions Subscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
http://jfi.sagepub.com/cgi/content/refs/25/5/571SAGE Journals Online and HighWire Press platforms):
(this article cites 17 articles hosted on the Citations
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
10.1177/0192513X03258307 ARTICLEJOURNAL OF FAMILY ISSUES / July 2004Crosnoe , Elder / FAMILY DYNAMICS
Family Dynamics, SupportiveRelationships, and EducationalResilience During Adolescence
ROBERT CROSNOEUniversity of Texas at Austin
GLEN H. ELDER, JR.University of North Carolina at Chapel Hill
If problematic relationships with parents are an academic risk factor during adolescence,then nonparental sources of support (e.g., friends, siblings, and teachers) may be arenas ofcomfort that promote educational resilience in the face of such risk. In a series of structuralmodels using the National Longitudinal Study of Adolescent Health, the authors found thatnonparental relationships are more likely to be directly associated with academic behaviorthan to interact with parent-related risk. Protective interactions occur only among certainsubgroups. For example, close relationships with teachers and involvement with friends pro-tect against parent-related academic risk among Asian American adolescents, whereas sup-port from friends operates similarly for younger girls. In other subgroups, parental andnonparental relationships interact but not in a protective way. These patterns demonstrate thecomplex interplay of developmental ecology and larger social structures during the adoles-cent stage of life as well as the context-specific nature of resilience.
Keywords: resilience; life course; education; race; gender
Contemporary developmental research has cultivated a greater theoreticalrecognition of the complexity of the adolescent stage of life, drawing at-tention to the influence of interpersonal ties on adjustment, the interde-pendence of developmental settings, and the overarching role of socialstructure (Bronfenbrenner & Morris, 1998; Elder, 1998). This theoreticalrichness has, in turn, engendered empirical research that bridges multiple
571
Authors’Note: The authors acknowledge support by the National Institute of Mental Health(MH 00567, MH 57549) and a Spencer Foundation Senior Scholar Award to Elder. Thisresearch is based on data from the National Longitudinal Study of Adolescent Health(J. Richard Udry, principal investigator), which was funded by Grant PO1-HD31921 fromthe National Institute of Child and Human Development to the Carolina Population Center,with cooperative funding participation by 17 other agencies. The authors would like to thankJeylan Mortimer for her helpful comments on earlier drafts of this article.
JOURNAL OF FAMILY ISSUES, Vol. 25 No. 5, July 2004 571-602DOI: 10.1177/0192513X03258307© 2004 Sage Publications
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
domains on multiple levels to more fully gauge how young people comeof age. This study follows this trend by exploring the potential function-ality of the overlap among familial and extrafamilial relationships acrossdiverse social groups. Specifically, it examines whether support fromfriends, siblings, and teachers protects against the academic risk of emo-tionally distant relationships with parents and whether this risk-protectioninteraction differs by developmental stage, race/ethnicity, and gender.
Essentially, this study centers on the linkage between the parent-ado-lescent relationship and academic adjustment. This linkage has long beena research focus, but it can be studied in a more nuanced way. We viewit here within the developmental ecology of adolescence, the larger so-cial structure, and the intersection of the two. The motivation to pursuethis topic in this way is gleaned from life course theory (Elder, 1998;Settersten, 1999, which is usually applied to more temporally orientedquestions but is relevant to studies within one stage of life. This paradigmhighlights the importance of interconnected relationships with significantothers (linked lives), timing, and macro-structural context in organizingand shaping developmental trajectories. Drawing on this paradigm, rela-tionships that occur within major settings (e.g., the family, school, peergroup) overlap to influence adolescent adjustment, but the nature of thisoverlap differs by when it occurs in life and cannot be divorced from thelarger social structure (e.g., gender and race/ethnicity).
This study is also structured by a risk-protection framework borrowedfrom epidemiological research. Risk refers to individual or social factorsthat are associated with a greater likelihood of poor developmental out-comes, whereas protective factors decrease the association between riskfactor and outcome (Garmezy & Masten, 1986). In this study, the poten-tial risk factor is emotionally distant parent-adolescent relationships,whereas support from friends, siblings, and teachers serve as potentialprotective factors that may decrease the negative academic impact of thefamily risk factor. To embed this in more substantive terms, we introducetwo concepts that will undergird our presentation. An arena of comfort is asupportive interpersonal context that enhances the ability to cope withchallenges in other settings or, in other words, an interpersonal example ofa protective factor (Simmons & Blythe, 1987). The process by which pro-tective factors buffer against risk factors is the heart of educational resil-ience—success at school despite difficult circumstances. Bringing theseconcepts together, resilient youth do well despite distant relationshipswith parents, possibly because of nonparental arenas of comfort.
The empirical analyses derived from these conceptual frameworks ex-tend the work of Call and Mortimer (2001), who explored arenas of com-
572 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
fort for several social psychological outcomes among Midwestern youth.Here, we apply this framework explicitly to the educational realm, exam-ine how these processes play out across developmental stages, and locatethese processes within the contexts of race/ethnicity and gender. Further-more, we conduct this research with a nationally representative sample ofAmerican youth—the ongoing National Longitudinal Study of Adoles-cent Health.
In the following sections, we discuss the importance of social ties foradolescent academic outcomes, paying attention to age, race/ethnic, andgender differences. After describing our sample and methods, we presentresults from a series of structural models on the overlapping and sociallyembedded nature of adolescent relationships.
PARENTS, ADOLESCENTS,AND EDUCATIONAL EXPERIENCES
At the center of this study is the linkage between adolescent academicadjustment and emotional support from parents. In other words, we do notdelve into parents instrumental support or educational involvement but in-stead explore the academic implications of the affective quality of the par-ent-adolescent relationship. Such an approach is drawn from life coursetheory, which emphasizes how behavioral trajectories are intertwinedwith relationship trajectories and recognizes the connections among vari-ous social contexts.
The linkage between emotional support from parents and adolescentacademic performance is well-established. Warm and supportive relation-ships have been found to promote academic achievement and positive atti-tudes about education, whereas more distant or conflictual relationshipscan be tremendous stressors that disrupt proper functioning in school(Amato & Gilbreth, 1999; Call & Mortimer, 2001; Demo & Acock, 1996;Grotevant, 1998). Thus, in epidemiological terms, parent-adolescentemotional distance is an academic risk factor, its presence increasing theprobability of academic problems.
Sociological interpretations of the academic implications of parent-adolescent dynamics often draw on the concept of social capital, assertingthat close ties improve academic prospects by facilitating the transmissionof important resources from adult to child in the form of instrumentalassistance or pro-school attitudes, whereas distant ties block this trans-mission. Certainly, evidence supports this explanation (Coleman, 1988;Furstenberg, Cook, Eccles, Elder, & Sameroff, 1999). Although evidence
Crosnoe, Elder / FAMILY DYNAMICS 573
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
does support this scenario, a more social psychological explanation bettermatches our focus on affective bonds and emotional support. Close rela-tionships with parents serve as the secure foundation for adolescents’nav-igation of the external world. With such support, young people have moresecurity and confidence to meet challenges in other domains—negativepeer influences, school changes, academic pressures—and to success-fully complete the developmental tasks of adolescence—identity forma-tion, learning of responsibility, formulation of mature relationships(Dornbusch 1989). Without this secure base, however, adolescents areless able to cope with the rapid changes of their lives and adapt to newroles and environments (Csikszentmihalyi & Schneider, 2000; Simmons& Blythe, 1987).
Crucial to the life course approach is the appreciation of issues of tim-ing and macro-context, which suggest a potential variability in the familydynamics just described. For example, the influence of parents on adoles-cent behavior declines with age, as young people attempt to establish au-tonomy from the family (Crosnoe, 2000). Some evidence suggests thatAfrican American and Hispanic American youth benefit more academi-cally from supportive parenting, whereas Asian American adolescents aremore immune to harsh parenting (Deater-Deckhard, Dodge, Bates, &Pettit, 1996; Steinberg, Dornbusch, & Brown, 1992). Finally, girls tend tohave a greater emotional stake in their relationships with parents and, con-sequently, seem more reactive to parenting in positive and negative ways(Call & Mortimer, 2001; Windle, 1992). Thus, the parent-adolescentrelationship should be examined in relation to timing and context.
PERSONAL RELATIONSHIPS ASARENAS OF COMFORT
Prior research suggests that a lack of emotional support from parentshampers adolescent academic functioning. This risk factor is the given inthis study. Our main objective is to examine the arenas of comfort thatcounterbalance this risk. To do so, we look at relationships from three pri-mary settings of adolescent life: the peer group, family, and school. Canrelationships in these settings mitigate the potential consequences ofproblems in the parent-adolescent relationship?
This focus derives from the life course principle that young people de-velop within a system of social ties. When these ties are generally positive,they amplify social redundancy, creating a secure, interconnected base ofsupport for adolescents to live their lives and face new experiences in mul-
574 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
tiple domains. When supportive relationships mitigate the effects of un-healthy relationships, such overlap is developmentally functional in thatthe secure base necessary for successful development is not underminedby problems in any one domain (Elder & Conger, 2000; Rutter, 1985). Inthe case of this study, adolescents who grow up in the midst of family dis-cord might not be as academically successful as those from more func-tional families, but they will do better than expected if their lives are builton other sources of support. We define success of this kind in difficultcircumstances as educational resilience—nonparental arenas of comfortprotecting against the academic risk of problematic relationships withparents.
What interpersonal contexts promote educational resilience? We focuson relationships with friends, siblings, and teachers as potential arenas ofcomfort. All three loom large in the social worlds of adolescents, andemotional support from all three has been found to be associated with aca-demic behavior (Buhrmester, 1992; Crosnoe, 2000; Sanders & Jordan,2000). Again, the focus of this study is not on the main effects of these re-lationships on academic performance but instead on how they interactwith parent-adolescent relationships to influence academic outcomes.Past research has rarely taken such an approach, and so we have to basemuch of our argument on related topics.
Friendships become more prominent during adolescence, often sup-planting parents as significant others (Crosnoe, 2000). Because of this,support from friends may buffer against a lack of parental support. Indeed,evidence suggests that friendships help young people cope with lifestressors, such as divorce (Hetherington, 1989; Windle, 1992). On theother hand, Call and Mortimer (2001) reported that in their special sam-ple, support from friends did not moderate the impact of family problemson grades.
Sibling relationships are embedded in the same family system as par-ent-adolescent relationships and may be important for resilience (definedin terms of problems with parents). In general, we know less about sib-lings than other relationships. Although young people do not automati-cally turn to siblings in the absence of parental support (Dunn, 1992),those who rely on siblings during problematic family situations (e.g., di-vorce) are better adjusted (Hetherington, 1998; Jenkins & Smith, 1991).This dynamic could generalize to other family problems and to academicoutcomes.
Teachers can play an important role in the academic lives of youngpeople (Muller, Katz, & Dance, 1999). Because of their power to guide,support, and set standards and expectations, teachers may promote educa-
Crosnoe, Elder / FAMILY DYNAMICS 575
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
tional resilience in the face of problems between parents and adolescents.Muller (2001) found that support from teachers can protect against the ac-ademic problems associated with nonemotional family disadvantages,whereas Call and Mortimer (2001) found that such support can bufferagainst problems with parents in nonacademic ways. Such research sug-gests that close relationships with teachers may, to some extent, replacesupport that is missing at home.
Thus, past research suggests that emotionally supportive relationshipswith friends, siblings, and teachers may promote educational resilience byserving as arenas of comfort. This phenomena, however, requires a moresystematic treatment, such as recognizing that it may be highly contextspecific. For this reason, we draw on the life course principles of timingand macro-context and examine whether this phenomenon varies by de-velopmental stage, race/ethnicity, and gender.
If we know little about how personal relationships moderate family-related academic risks, then we know next to nothing about how suchmoderation differs by these three factors (or any other). Past research onrelated topics does little to guide us. For example, older adolescents tendto be more oriented to nonparental relationships (Crosnoe, 2000). Thissuggests that they have greater access to sources of emotional support,which increases the protective potential of nonparental relationships. Atthe same time, this suggests that older adolescents may be less affected byproblems with parents, which reduces academic risk and therefore lessensthe need for protection.
A similar confusion surrounds race/ethnicity and gender. White ado-lescents tend to have the most advantages (e.g., resources, status) that pro-mote school success, so that the power of any one domain to derail orenhance their educational trajectories is likely to be lower than for adoles-cents from more disadvantaged groups. Yet past studies have found thatthe ability of nonparental relationships to promote academic success var-ies sharply, and somewhat inconsistently, across minority groups com-pared to Whites (Ainsworth-Darnell & Downey, 1998; Steinberg et al.,1992). Turning to gender, girls tend to need greater emotional support todo well in school, but they also have closer relationships across the board(Call & Mortimer, 2001; Kuttler, La Greca, & Prinstein, 1999). This sug-gests that the negative impact of distant parent-adolescent relationshipsmay be greater for girls but that girls may be more likely than boys to drawsupport from other sources that counterbalances this negative impact.
Because of the alternative expectations that arise from different read-ings of past research relevant to this topic, we treat our additional analysesas exploratory. Rather than looking at main effects of social relationships
576 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
on academic success within different groups, we study whether the inter-action of parental and nonparental relationships differs by developmentalstage, race/ethnicity, and gender.
FAMILY RISK, ARENAS OFCOMFORT, AND EDUCATIONAL RESILIENCE
To summarize, this study has two main objectives, both of which arerelated to the proximate developmental settings of adolescence, themacro-structural context in which these settings exist, and the linkage be-tween these two levels. First, we seek to determine whether emotionallysupportive relationships with friends, siblings, and teachers serve as are-nas of comfort that promote educational resilience in problematic familyenvironments. Second, we explore whether this overlap of parental andnonparental relationships is more or less protective in certain populationsubgroups.
METHOD
SAMPLE
This research uses data from the National Longitudinal Study of Ado-lescent Health (Add Health), a nationally representative study of adoles-cents in Grades 7 through 12. A sample of schools was selected from a listof American high schools provided by the Quality Education Database.To ensure diversity, sampling was stratified by region, urbanicity, schoolsector, racial composition, and school size. Each high school in the samplewas matched to one of its feeder schools, with the probability of the feederschool being selected proportional to its contribution to the high school.More than 70% of the selected schools agreed to participate, with replace-ments selected from each community. This multistage design resulted in afinal sample of 132 schools in 80 communities.
All students in this population completed the In-School questionnairein the 1994-1995 school year. Of these, a subgroup of students, selectedevenly across high school/feeder school pairs, was selected to participatein two waves of in-home interviews in 1995 and 1996. A total of 14,736adolescents participated in both in-home interviews. Our study sampleconsists of Add Health adolescents who meet four requirements: Theyparticipated in both waves of in-home interviews, had one parent inter-
Crosnoe, Elder / FAMILY DYNAMICS 577
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
viewed (parent data set, collected at Wave 1), are members of the fourmain ethnic groups (White, African American, Hispanic American, AsianAmerican), and were in Grades 7 through 11 in Wave 1 (to ensure that allyouth in the panel would be in school at Wave 2).
These criteria result in a study sample of 11,788 youth. Sample charac-teristics are presented in Table 1. A majority of the sample is White (56%)and female (51%). Sample adolescents are, on average, about 15 and 16years old and have a B+ grade point average. Table 1 also includes infor-mation on the same factors for the full Wave 1 sample. Comparisons be-tween the two groups reveal some bias due to attrition and selection crite-ria. Compared to the Wave 1 adolescents, the sample adolescents areyounger (recall that we excluded all Wave 1 seniors), better students, andcome from more advantaged backgrounds.
MEASURES
The dependent variable is drawn from the Wave 2 in-home question-naire, whereas all other measures are based on items in the Wave 1 in-home and parent questionnaires.
Off-track academic behavior. The scale for our dependent variableconsists of five items: whether the student has repeated the last grade inschool (1 = yes), whether the student has had trouble in the past schoolyear getting homework done (0 = never to 4 = everyday), the sum ofwhether the student has been expelled or suspended from school in thepast school year (1 = yes for each), whether the student has skipped school
578 JOURNAL OF FAMILY ISSUES / July 2004
TABLE 1Comparative Statistics for Study Sample
Versus All Add Health Adolescents in Wave 1
Mean (Standard Deviation)
Study Sample Wave 1 Adolescents
Female .51 (.50) .51 (.50)Age 16.01a (1.49) 16.16 (1.72)Intact family .53a (.50) .50 (.50)White .56a (.50) .50 (.50)Grade point average 2.78a (.77) 2.75 (.77)n 11,788 20,745
a. t tests indicate that two means differ significantly across groups (t > 1.96, p < .05).
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
in the past year, and the reverse-coding of the student’s grade point aver-age (the average of self-reported grades in English, math, social studies,and science in the past year, 1 = F/D and 4 = A). The correlations amongthese items range from .10 (p < .001) to .28 (p < .001). The five items arestandardized and summed to create the scale (M = –.16, SD = 2.63, α =.65). Statistics for each item in this measure are presented in Table 2.
We constructed this scale to move beyond mere measures of achieve-ment (e.g., school grades) and better gauge the social psychological expe-rience of schooling. Low scores on this scale group together a wide varietyof students, including high achievers and those who are just getting by.The high end, however, is more meaningful, identifying the group of stu-dents whose educational careers are in clear trouble.
Emotional distance between parent and adolescent. Our primary inde-pendent variable consists of five composite measures (all coded so thathigher values represent more problematic relationships). One, bondingwith adolescent, is based on parent report, whereas four, bonding tomother and father, communication with parents, activities with parents,and general family cohesion, are based on adolescent report. See Appen-dix A for a complete description of the construction of each of these five
Crosnoe, Elder / FAMILY DYNAMICS 579
TABLE 2Univariate Statistics for Off-Track Academic Behavior Measure
Frequency (%) Mean (Standard Deviation)
Held back in schoolNo 95.0Yes 5.0
Suspended/expelled from schoolNeither 85.1Either 10.6Both 1.3
Skipped classesNo 66.8Yes 33.2
Had trouble with homeworkNever 30.0Just a few times 41.7About once/week 16.3Almost everyday 8.3Everyday 3.7
Low grade point average 2.20 (.76)
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
composites. The composites are positively correlated (p < .001), and withthe exception of bonding to adolescent with cohesion, these correlationsare moderate to strong. After standardizing all five measures, we take theirmean for the final scale (M = –.04, SD = .52, α = .66).
Based on the work of Furstenberg et al. (1999), we combine severalcomposites together in this scale in order to tap, more broadly, the overallemotional tone of the parent-adolescent relationship. Some parent-ado-lescent relationships may be high on some factors and low on others, sothat the middle values on this scale may be somewhat ambiguous. Yet thehigh and low ends of the scale represent clear extremes. Theoretically, thismeasure draws on three well-established dimensions of parenting: affec-tive ties, shared activities, and security (Coleman, 1988; Hetherington,1989). Statistically, it combines adolescent and parent reports, which in-creases construct validity and reduces shared method variance (Conger,Reuter, & Conger, 1994).
Friendship. Three components of friendship are considered in thisstudy. Each adolescent was asked a battery of questions about specificfriends, a maximum of five female and five male friends. Involvement isthe sum of four items (1 = yes, 0 = no): whether, in the past week, the ado-lescent had gone to the friend’s house, hung out somewhere with thefriend, and talked on the telephone with the friend, and whether the ado-lescent had spent time with the friend during the past weekend (M = 2.10,SD = 1.07, α = .62). Support is a single item: whether, in the past week, theadolescent talked to the friend about a problem (M = .49, SD = .41).
For each item, scores are averaged across all listed friends. Therefore,if only one friend is named, then the support score for that friend serves asfriends support, but if 10 friends are named, then the average of supportacross the friends serves as friends’ support. The impact of friendshipmight differ depending on the number of friends that an adolescent has, oralternatively, there might be a critical threshold, where having one friendis the crucial distinction. To account for this, we code as zero on involve-ment and support all respondents who name no friends and include a thirdfriendship measure, the number of friends, which is the count of thefriends listed by the respondent (M = 3.05, SD = 2.59). These three com-ponents reflect Hartup’s (1993) call for studying both the qualities andquantity of friendships.
Sibling relationships. Adolescents were also asked questions aboutspecific siblings (maximum = 7). Sibling support refers to emotionalcloseness and is based on a single item: how often the adolescent feels
580 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
love for the specific sibling. Responses, ranging from 1 (never) to 5 (veryoften), are averaged across siblings (M = 1.17, SD = 1.88). Slightly lessthan half of the sample has no siblings. To avoid excluding these adoles-cents, we code them as zero on support and include a measure for presenceof siblings (1 = yes, 0 = no), based on the count of siblings listed by the re-spondent. This measure is included only as a statistical control. Its inclu-sion converts the sibling support variable into an interaction term, mea-suring support if siblings are present.
Teacher relationships. This scale consists of three items: the extent towhich the adolescent has trouble getting along with teachers, believes thattheir teachers treat students fairly, and feels that teachers care about him orher. Responses to each item range from 1 to 5: almost every day to neverfor the first item, strongly disagree to strongly agree for the second, andnever to very much for the third. The average of the three serves as thecomposite measure (M = 3.72, SD = .76, α = .68). Like several past stud-ies, this measure does not refer to relationships with specific teachers butinstead taps the adolescent’s general feelings about the teachers in his orher school (Sanders & Jordan, 2000; Steinberg, Lamborn, Dornbusch, &Darling, 1992).
Sociodemographic controls. Seven variables serve as controls: gender,age, parent education, race/ethnicity, parents educational expectations,adolescents’ educational expectations, and prior off-track behavior. SeeAppendix B for descriptive information on these control variables.
PLAN OF ANALYSIS
For hypothesis testing, we estimated a series of structural models(without latent constructs) in Amos 4.0. We used this package because itallowed us to account for measurement error, estimate missing data(through full information maximum likelihood), and compare parametersacross groups (Arbuckle & Wothke, 1999). All models are fully saturatedand can be interpreted like standard regressions.
In our basic modeling plan, we first included Wave 1 parent-adolescentdistance (along with the controls) as a predictor of Wave 2 off-track be-havior. Next, we added Wave1 friend, sibling, and teacher measures aspredictors. Finally, we included interaction terms for parent-adolescentdistance and all nonparental relationships. If an interaction term is signifi-cant (and in the proper direction), then that relationship can be viewed as
Crosnoe, Elder / FAMILY DYNAMICS 581
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
an arena of comfort (Garmezy & Masten, 1986; Simmons & Blythe,1987).
The second main question of this study was whether these processesdiffer by developmental stage (defined by school level), race/ethnicity,and gender. We do this with group modeling (Arbuckle & Wothke, 1999).For example, we first estimated a model in which the protective power ofteacher bonding (the interaction of bonding with parenting) is freely esti-mated for boys and girls. In the next step, we constrained this effect to beequal across genders. If the change in χ2/df between these steps was statis-tically significant, then we could conclude that this association differsacross the two groups.
RESULTS
RELATIONSHIPS FROM VARIOUSECOLOGICAL SETTINGS AND ACADEMIC BEHAVIOR
To begin, we offer a preliminary look at the overlap of relationships inthe ecology of adolescent development. Table 3 presents the correla-tions among the important relationship measures and off-track academicbehavior.
Parent-adolescent emotional distance is significantly correlated withother relationships: positively with support from friends and negativelywith support from siblings and teachers. It is also positively correlatedwith off-track behavior, meaning that problems at home coincide with ac-ademic problems. Two other patterns in these data are important. Friend-
582 JOURNAL OF FAMILY ISSUES / July 2004
TABLE 3Correlations Among Relationships and Academic Behavior
1 2 3 4 5 6
1. Parent-adolescent distance2. Number of friends .023. Friend involvement .01 –.14*4. Friend support .02* –.11* .32*5. Sibling support –.03* .05* –.00 .006. Teacher-bonding –.32* –.02* –.09* –.09* .03*7. Off-track behavior .23* .04* .09* .01 .01 –.31*
*Correlations significant at p < .05.
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
ship elements seem to go along with problems at home and school.Teacher bonding has the most consistent pattern, negatively correlatedwith all other relationships (except support from siblings) and with off-track academic behavior.
OVERLAPPING RELATIONSHIPS ANDEDUCATIONAL RESILIENCE
Having given a glimpse of the key relationships that constitute adoles-cent life, we now turn to the main objective of this study: to explore rela-tionship overlap by examining whether supportive nonparental relation-ships can serve as arenas of comfort. Table 4 presents results from a seriesof structural models relevant to this question.
Of course, this objective rests on the core idea that emotional distancein the parent-adolescent relationship is an academic risk factor. Our analy-ses indicate that it is (β = .08, p < .001 in Model 1). The strength of thisassociation is second in magnitude only to initial off-track behavior, butsome caution is needed when interpreting the size of the effect. One stan-dard deviation increase in parent-adolescent distance is associated with anincrease in off-track behavior of only 3% of a standard deviation. Giventhe severity of the behaviors cataloged in this academic scale, any increaseis problematic, but we stress that these effects are by no means large.
Inclusion of relationships from other ecological settings reduces thisrisk of parent-adolescent distance even more (β = .06, p < .001 in Model2), although it remains statistically significant. Of the potential arenas ofcomfort, teacher bonding has the strongest impact on off-track behavior(β = –10, p < .001; one standard deviation increase in bonding associatedwith 4% of a standard deviation decrease in off-track behavior). Whereasteacher bonding reduces off-track behavior, friendship-related factors allincrease academic problems. Sibling support has no relation to academicbehavior.
Our main focus is on conditional effects (see Model 3) or interactionsamong relationships. Significant and negative interactions would indi-cate that nonparental relationships are arenas of comfort that provide ref-uge from problems at home (in other words, they promote educationalresilience).
Only one interaction term is significant, but interestingly, it is positivein sign rather than negative. For this interaction between parent-adolescentdistance and teacher bonding, we can sum the main effect of parent-adolescent distance on off-track behavior with the interactive effect ofparent-adolescent distance and teacher-bonding (.30 + .18 = .48, b coeffi-
Crosnoe, Elder / FAMILY DYNAMICS 583
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
584 JOURNAL OF FAMILY ISSUES / July 2004
TABLE 4Results From Regressions Predicting
Off-Track Academic Behavior
Model 1 Model 2 Model 3
b β b β b β
Individual characteristicsFemale –.28*** –.05 –.34*** –.06 –.06 –.34***
(.04) (.04) (.04)Age –.11*** –.05 –.12*** –.07 –.12*** –.07
(.01) (.02) .02Parent education –.03*** –.02 –.04*** –.04 –.04*** –.04
(.01) (.01) (.01)African American .19*** .03 .23*** .04 .23*** .04
(.06) (.05) (.05)Hispanic American .21*** .03 .27*** .04 .27*** .04
(.09) (.09) (.09)Asian American .18* .01 .26*** .02 .25*** .02
(.09) (.09) (.09)Parents educational –.02 –.00 –.03 –.01 –.03 –.01
expectations (.03) (.03) (.03)Adolescents educational –.14*** –.05 –.13*** –.05 –.13*** –.05
expectations (.02) (.02) (.02)Prior off-track behavior .43*** .49 .40*** .45 .40*** .45
(.01) (.01) (.01)Relationships
Parent-adolescent distance .42*** .08 .31*** .06 .30*** .06(.04) (.04) (.11)
Number of friends .03*** .03 .03*** .03(.01) (.01)
Friend involvement .10*** .04 .10*** .04(.02) (.02)
Friend support .18*** .03 .19*** .03(.06) (.06)
Presence of siblings .23 .04 .24 .04(.15) (.16)
Sibling support –.02 –.01 –.02 –.01(.04) (.04)
Teacher bonding –.34*** –.10 –.35*** –.10(.03) (.03)
Interaction termsNumber of friends –.01 –.01
(.02)Friend involvement .00 .01
(.04)
(continued)
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
cients used rather than β) (see Jessor, Van Den Bos, Vanderryn, Costa, &Turbin, 1995). For the sake of interpretation, this sum can stand as the co-efficient for the association between parent-adolescent distance and off-track behavior for adolescents high in teacher bonding (e.g., for thisgroup, one unit increase in parent-adolescent distance is associated with18% of a standard deviation increase in off-track behavior). In otherwords, a supportive relationship with a teacher does not serve as an arenaof comfort but instead appears to exacerbate risk in the family setting.
Thus, emotional distance from parents is a significant academic riskfactor for adolescents. This effect is small in magnitude, although it stillexceeds the effects of family background, race/ethnicity, and academic at-titudes. Nonparental relationships can influence academic behavior, inpositive and negative ways, but do not protect against this parent-relatedrisk. Life course theory suggests an extension of these analyses. This theo-retical paradigm guides research on individual adjustment toward empha-sizing timing and macro-context. In this vein, we now turn to the questionof whether patterns of interactions among parental and nonparental rela-tionships differ by developmental stage (defined by school level), race/ethnicity, and gender.
Crosnoe, Elder / FAMILY DYNAMICS 585
Friend support .08 .01(.11)
Presence of siblings .04 .01(.27)
Sibling support –.04 –.01(.07)
Teacher bonding .18*** .03(.05)
R2 .30 .33 .33
NOTE: Model 1 is baseline model. Model 2 includes main effects for relationship variables.Model 3 includes interaction terms for all relationship variables (with parent-adolescent dis-tance). b coefficients are unstandardized (with standard errors in parentheses), and β coeffi-cients are standardized.*p < .05. **p < .01. ***p < .001.
TABLE 4 (continued)
Model 1 Model 2 Model 3
b β b β b β
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
EDUCATIONAL RESILIENCE AND SCHOOL LEVEL
The developmental nature of adolescent relationships and academicadjustment requires that we take into account issues of timing. The role ofvarious relationships, and the interactions among these roles, may differas adolescents move through this stage of life. A true life course studywould trace these patterns across time, a procedure that is not possiblewith the current data. Instead, we attempt to capture the spirit of this lifecourse principle in our study of educational resilience by breaking downour basic model for high school and middle school students.
The top portion of Table 5 presents descriptive statistics, by schoollevel, for the key variables in the model. High school students have ahigher mean level of parent-adolescent distance than do middle schoolstudents, and they also seem to be more peer oriented than their middleschool counterparts. The bottom portion of Table 5 presents results fromgroup modeling of our basic model by school level. For the most part, thetwo groups are quite similar. Initially, parent-adolescent distance was asignificant academic risk factor for both, although this risk becomesnonsignificant for middle school students once other factors are taken intoaccount. Turning to conditional effects, the pattern seen for high schoolstudents replicates the pattern for the full sample, but the finding from thefull sample on teacher bonding (the interaction suggesting that teacherbonding strengthens the association between parent-adolescent distanceand off-track behavior) does not hold for the middle school students.
Thus, for the most part, the overlap between parental and nonparentalrelationships, when it does occur, does not seem to promote educationalresilience at either stage. Timing is important, however, for the unex-pected overlap, discussed above, between relationships with parents andteachers, which occurs only in later adolescence.
EDUCATIONAL RESILIENCEAND ADOLESCENT RACE/ETHNICITY
Macro-contexts are important to understanding the role of the inter-change between parent-related risk and nonparental arenas of comfort.According to life course theory, the relation between individual develop-ment and linked lives is embedded in macro-contexts. The social struc-tural elements of race/ethnicity and gender are representative of macro-contexts in American society. Again, these data are not equipped tocontextualize developmental or relationship trajectories within race/ethnicity or gender, but they do allow the exploration of how overlap in
586 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
developmental settings varies by structural location. To pursue this, weperform race/ethnic-specific analyses of our main model (see Table 6).
The top portion of Table 6 contains mean differences by race/ethnicity.As expected, White and Asian American youth are less likely to be offtrack in school than are African American and Hispanic American youth.
Crosnoe, Elder / FAMILY DYNAMICS 587
TABLE 5Results From Regressions Predicting
Off-Track Academic Behavior, by School Level
High School (n = 8,532) Middle School (n = 3,045)
Mean Mean
Descriptive statisticsOff-track behavior –.11a –.37bParent-adolescent distance –.01a .13bNumber of friends 3.19a 2.55bFriend involvement 2.17a 1.91bFriend support .52a .41 bSibling support 1.16 1.10Teacher bonding 3.71 3.74
Group Modeling b β b β
Main effectsParent-adolescent distance .24*** (.11) .05 .33 (.22) .06Number of friends .02* (.01) .03 .03 (.02) .06Friend involvement .05* (.02) .03 .16*** (.04) .03Friend support .15* (.17) .02 .08 (.11) .01Presence of siblings .18 (.19) .01 .53 (.31) .09Sibling support –.02 (.05) –.03 –.08 (.05) –.06Teacher bonding –.38*** (.05) –.11 –.37*** (.06) –.11
Interaction termsNumber of friends –.01 (.02) –.00 –.04 (.04) –.02Friend involvement –.00 (.04) –.00 .03 (.08) .01Friend support .17 (.12) .01 –.16 (.21) –.01Presence of siblings .14 (.32) .02 –.38 (.55) .04Sibling support –.08 (.08) –.03 .14 (.14) .05Teacher bonding .22***a (.06) .04 .01b (.09) .00
R2 .33 .34
NOTE: Controlling for gender, age, parent education, race/ethnicity, parents educational ex-pectations, adolescents educational expectations, and prior off-track behavior. b coefficientsare unstandardized (with standard errors in parentheses), and β coefficients are standardized.Coefficients with different subscripts differ significantly (p < .05) across school level, ac-cording to one-way ANOVA for means an ∆χ2/df for interactions.*p < .05. **p < .01. ***p < .001.
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
588
TA
BL
E 6
Des
crip
tive
Sta
tist
ics
and
Sele
cted
Res
ults
Fro
m G
roup
Mod
elin
g of
Off
-Tra
ck B
ehav
ior,
by
Rac
e/E
thni
city
Afr
ican
His
pani
cA
sian
Whi
teA
mer
ican
Am
eric
anA
mer
ican
(n=
6,5
95)
(n=
2,5
98)
(n=
1,9
54)
(n=
641
)
Des
crip
tive
Sta
tist
icM
ean
Mea
nM
ean
Mea
n
Off
-tra
ck b
ehav
ior
–.42
c–.
15b
.37 a
–.40
cPa
rent
-ado
lesc
ent d
ista
nce
–.04
c–.
07c
–.01
ab.0
1 aN
umbe
r of
fri
ends
3.18
b2.
63c
3.00
b3.
51c
Frie
nd in
volv
emen
t2.
15a
2.02
b2.
05b
1.94
cFr
iend
sup
port
.50
.48
.49
.48 a
Sibl
ing
supp
ort
1.20
ab1.
16b
1.07
c1.
24a
Teac
her
bond
ing
3.73
b3.
65c
3.73
b3.
88a
Gro
up M
odel
ing
bβ
bβ
bβ
bβ
Mai
n ef
fect
sPa
rent
-ado
lesc
ent d
ista
nce
.31*
**(.
14)
.06
.09
(.26
).0
2.4
9*(.
24)
.10
.91
(.51
).1
7N
umbe
r of
fri
ends
.03*
(.01
).0
3.0
5*(.
02)
.04
.02*
(.02
).0
2.0
2(.
03)
.02
Frie
nd in
volv
emen
t.0
9*(.
03)
.04
.11*
(.05
).0
4.1
4**
(.05
).0
6.1
3(.
09)
.05
Frie
nd s
uppo
rt.2
2*(.
07)
.03
.19
(.12
).0
3–.
03(.
15)
–.00
.30
(.26
).0
2Pr
esen
ce o
f si
blin
gs–.
05(.
19)
–.01
1.74
***
(.44
).2
9.0
7(.
45)
.01
.47
(.71
).0
8Si
blin
g su
ppor
t.0
4(.
05)
.03
–.33
***
(.11
)–.
24.0
1(.
11)
.01
–.13
(.17
)–.
09Te
ache
r bo
ndin
g–.
34**
*(.
04)
–.10
–.35
***
(.06
)–.
10–.
41**
*(.
08)
–.12
–.12
(.15
)–.
03
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
589
Inte
ract
ion
term
sN
umbe
r of
fri
ends
–.00
(.02
)–.
00–.
04(.
04)
–.02
–.02
(.04
)–.
01–.
12(.
06)
–.06
Frie
nd in
volv
emen
t–.
03b
(.05
)–.
01.2
0*a
(.16
).0
7–.
02b
(.10
)–.
01–.
47* c
(.15
)–.
16Fr
iend
sup
port
.09 b
(.13
).0
1–.
48c
(.24
)–.
04.6
2*a
(.27
)–.
05.7
6 a(.
48)
.05
Pres
ence
of
sibl
ings
.21
(.33
).0
2–.
66(.
72)
–.06
–.24
(.68
)–.
031.
71(1
.30)
.17
Sibl
ing
supp
ort
–.11
(.08
)–.
04.0
2(.
17)
.07
–.01
(.16
)–.
00–.
36(.
33)
–.12
Teac
her-
bond
ing
.17*
* b(.
06)
.03
.24*
b(.
11)
.04
.20 b
(.13
).0
3–.
54* a
(.24
)–.
07R
2.3
5.2
8.2
8.3
6
NO
TE
:Con
trol
ling
forg
ende
r,ag
e,pa
rent
educ
atio
n,pa
rent
sed
ucat
iona
lexp
ecta
tions
,ado
lesc
ents
educ
atio
nale
xpec
tatio
ns,a
ndpr
ioro
ff-t
rack
beha
vior
.bco
effi
cien
tsar
eun
stan
dard
ized
(with
stan
dard
erro
rsin
pare
nthe
ses)
,and
βco
effi
cien
tsar
est
anda
rdiz
ed.C
oeff
icie
nts
with
diff
eren
tsub
scri
pts
diff
ersi
gnif
i-ca
ntly
(p
< .0
5) a
cros
set
hnic
ities
, acc
ordi
ng to
one
-way
AN
OV
Afo
r m
eans
an
∆β2 /d
ffor
inte
ract
ions
.*p
< .0
5. *
*p<
.01.
***
p<
.001
.
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
Asian American and Hispanic American youth tend to be more distantfrom parents, with African American youth having the least distant re-lationships. Other personal relationships show no clear pattern of race/ethnic differences. Asian American adolescents have the most friends andreceive the most support from teachers. Whites spend the most time withfriends. African American adolescents lag behind the others in student-teacher relationships. Hispanic American youth lag behind the others insupport from siblings.
Does the ability of nonparental relationships to serve as arenas of com-fort differ by race/ethnicity? The bottom portion of Table 6 contains re-sults from group models relevant to this question. Results for Whites rep-licate those of the full sample; parent-adolescent distance is a slight riskfactor, no nonparental relationship reduces this risk, and teacher bondingappears to increase this risk.
Results from three minority groups, however, offer some intriguingdifferences. Before adding the interaction terms, parent-adolescent dis-tance was a significant academic risk factor for all groups, although thisrisk became nonsignificant in subsequent steps of analysis. Friendshipshave a dualistic interaction with parental distance among African Ameri-can adolescents. The negative interaction term for friends support in thisgroup (b = –.48, p < .05, 1 unit increase in support associated with a 18%standard deviation decrease in off-track behavior) suggests that emotion-ally supportive friendships promote educational resilience. On the otherhand, the positive interaction for involvement in this group (b = .20, p <.05, 7% standard deviation increase in off-track behavior) suggests thatsuch involvement is associated with the academic influence of parent-adolescent distance being even more problematic.
Hispanic American youth also have a significant interaction betweensupport from friends and parent-adolescent distance, but unlike AfricanAmerican youth, it is positive in sign and larger in magnitude (b = .62, p <.05, 24% standard deviation increase in off-track behavior for every oneunit increase in support). Like involvement with friends for AfricanAmerican youth and teacher bonding for African American and Whiteyouth, support from friends does not counterbalance the parent-relatedrisk but instead appears to strengthen this risk. For Asian American youth,we see a significant interaction term for teacher bonding (b = –.54, p < .05,20% standard deviation decrease in off-track behavior) and involvementwith friends (b = –.47, p < .05, 17% standard deviation decrease in off-track behavior). Unlike all other groups, these interactions, which aresomewhat moderate in size, suggest that support from nonparental adults
590 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
and friends can be arenas of comfort when emotional support is lacking athome.
Of course, issues of timing, as well as linked lives, may be contex-tualized within the larger social structure, and so we also performed by-race/ethnic analyses within school levels. The results of these analyses arefar too voluminous to present in tabular form, but we can summarize thegeneral pattern. Analyses for high school students show basically thesame pattern of results as seen in Table 6. Analyses of middle school stu-dents, however, show some differences.
Among White youth in middle school, no nonparental relationshipsinteracted with parent-adolescent distance in a positive or negative way,unlike White high school students, for whom we saw an interaction forteacher bonding. For African American adolescents in middle school, theprotective role of friends’ support is the same as for their counterparts inhigh school, but the problematic role of involvement with friends andteacher bonding, seen among older students, does not occur. In addition,the interaction of friends’ support and parent-adolescent distance for His-panic American students, seen in the full model, only held in high school.Unfortunately, we could not break up Asian American youth by schoollevel, due to a lack of statistical power.
To summarize, nonparental relationships can protect against parent-related risk, but this protection depends, to some extent, on timing, themacro-context of race/ethnicity, and the interaction between relation-ships. For African American youth, friends’ support protects againstproblems at home. For Asian American youth only, close relationshipswith teachers are protective. At the same time, in some circumstances,nonparental relationships are associated with parent-adolescent distancebeing an even stronger risk factor, as in the case of friendship involvementamong older African American youth, teacher bonding among olderWhite and African American youth, and friends’ support among olderHispanic American youth.
EDUCATIONAL RESILIENCE ANDADOLESCENT GENDER
A second element representative of the larger social structure, gender,could also be an important macro-context for the interactions betweenparental risk and nonparental protection. Consequently, we performed thesame sets of analyses for gender as for race/ethnicity. Table 7 contains theresults of these analyses. As expected, girls have fewer academic prob-lems. Although boys and girls do not differ in emotional distance from
Crosnoe, Elder / FAMILY DYNAMICS 591
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
parents, girls receive more emotional support from friends, siblings, andteachers.
Multivariate analyses show some interesting gender differences. Be-fore adding the interactions, parent-adolescent distance was a risk factorfor both groups, but this risk became nonsignificant for girls in the final
592 JOURNAL OF FAMILY ISSUES / July 2004
TABLE 7Descriptive Statistics and Selected Results From Group Modeling
of Off-Track Behavior, by Gender
Boys (n = 5,767) Girls (n = 6,021)
Descriptive Statistics Mean Mean
Off-track behavior .23a –.54bParent-adolescent distance –.04 –.03Number of friends 3.05 3.04Friend involvement 2.12a 2.07bFriend support .37b .61aSibling support 1.12b 1.23aTeacher-bonding 3.68b 3.76a
Group Modeling b β b β
Main effectsParent-adolescent distance .25 (.17) .05 .40** (.15) .09Number of friends .04* (.01) .03 .03* (.01) .03Friend involvement .10*** (.03) .04 .11* (.03) .05Friend support .16* (.08) .02 .21 (.05) .03Presence of siblings .36 (.22) .06 .10 (.23) –.02Sibling support –.05 (.05) –.03 .00 (.05) .01Teacher-bonding –.31*** (.04) –.09 –.39*** (.04) –.12
Interactions termsNumber of friends .03 (.02) –.01 –.01 (.02) –.00Friend involvement .01 (.06) .01 .01 (.05) .00Friend support .38*a (.17) .03 –.21*b (.10) .03Presence of siblings .25 (.38) .03 –.15 (.32) –.02Sibling support –.11 (.10) –.03 .02 (.09) .01Teacher bonding .28*a (.07) .05 .08b (.06) .02
R2 .32 .30
NOTE: Controlling for age, parent education, race/ethnicity, parents educational expecta-tions, adolescents educational expectations, and prior off-track behavior. b coefficients areunstandardized (with standard errors in parentheses), and β coefficients are standardized.Coefficients with different subscripts indicate significant gender difference (p < .05), accord-ing to one-way ANOVA for means an ∆χ2/df for interactions.*p < .05. **p < .01. ***p < .001.
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
model. Turning to interactions, support from friends offers the greatestcontrast between boys and girls. For girls, the negative interaction term forsupport from friends (b = –.21, p < .05) suggests that this factor is an arenaof comfort that counterbalances parent-related risk, but the opposite istrue for boys (b = .30, p < .05). The problematic role of teacher bondingdiscussed above only holds among boys (b = .28, p < .001), but teacherbonding does not interact significantly with parent-adolescent distancefor girls.
Again, we reestimated this group model by school level. The resultsjust described for boys hold only in high school. Analysis of middleschool boys reveals no meso-system level interactions, protective or oth-erwise. For girls, we see significant interactions in middle school only.The protective interaction of friends’ support seen in the full sample alsooccurs among younger girls, but a nonprotective interaction (e.g., increas-ing risk) also occurs in this group for support from siblings.
To summarize, girls tend to enjoy more support than boys in non-parental relationships, but in general, they do not necessarily derivegreater protection from these relationships. Only supportive friendshipsserve as arenas of comfort for girls and only in middle school. No personalrelationship serves as an arena of comfort for boys at either school level,and some actually operate in the opposite direction. Thus, boys and girlsare more alike than different in the overlap of relationships in their lives,but their differences also depend on their developmental stage (if schoollevel is a proxy for stage).
CONCLUSION
A key developmental task of adolescence is to establish independencefrom parents, but this independence should develop within a supportivefamily environment. The absence of close relationships with parents is ac-ademically problematic for adolescents, just as it is for children, whetherit results from the traits and behaviors of the parent, adolescent, or both.Yet adolescents are also active participants in a larger social world that ex-tends far beyond their families. This expanded interpersonal world pro-vides potential relationships to counterbalance what is lacking at homeand to reinforce what is right at home. In applying the life course perspec-tive to educational resilience, our goals have been to explore the nature ofthese overlapping relationships during adolescence, to determine whethersuch overlap promotes educational resilience, to examine the role of tim-ing in conditioning the protective function of this overlap, and to con-
Crosnoe, Elder / FAMILY DYNAMICS 593
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
textualize this constellation within the larger structures of race/ethnicityand gender.
Like past studies, we have shown that families can be a source of aca-demic risk. Such risk involves more than socioeconomic disadvantage butalso a lack of emotional bonding between parents and adolescents. Youngpeople who lack support at home, in the form of an absence of emotionalties and positive interactions, are more likely to get off track in school, inthe form of lower engagement, attachment, and achievement, even con-trolling for differences in family background, parents’ and adolescents’attitudes about education, and prior academic behavior. Thus, the aca-demic importance of this home-school connection encompasses the veryfoundation of the parent-adolescent relationship—its ability to serve as apsychological base for the navigation of extrafamilial contexts. If thisbase is undermined, then adolescents may lack the confidence and consci-entiousness to meet challenges, pursue tough goals, and cope with adver-sity. Lacking these abilities, they might be overwhelmed by school and, inturn, disengage.
Unlike most studies, we have also explored how this parent-related riskcan be counterbalanced by alternate sources of emotional support. In gen-eral, supportive nonparental relationships are related to academic out-comes in positive (e.g., teacher bonding) and negative (e.g., friends) ways,but they do not serve as arenas of comfort. These findings mirror those ofCall and Mortimer (2001) on school grades in a special sample of adoles-cents. Yet this pattern of findings is not the final word. Unlike Call andMortimer, we worked with nationally representative data (AddHealth)that allowed a more nuanced exploration of these processes that can bettercapture the guiding principles of the life course perspective. For our pur-poses, this nuance refers to variability in these processes that is related totiming and macro-context. This nuance is at the heart of the phenomenonof resilience during the adolescent stage of the life course.
In some cases, nonparental personal relationships are directly relatedto academic outcomes. These associations do not mean that they are are-nas of comfort, no matter what their positive effects may be. To be anarena of comfort, a nonparental relationship has to reduce the impact of arisk factor (parent-adolescent distance) on a developmental outcome (aca-demic behavior). For the most part, nonparental relationships do not serveas arenas of comfort, except in specific examples. These examples arehighly context specific—related to the intersection of timing, macro-con-texts, and type of relationship. For example, the overlap among parentaland nonparental relationships is most functional for Asian Americanyouth and for girls in middle school.
594 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
Among Asian American youth, a supportive relationship with a teach-ers can be an arena of comfort that mitigates the academic risk of an emo-tionally distant relationship with parents. If, as Chao (1994) suggested,the training components of Asian American parenting is crucial to the suc-cess of Asian American youth, then the absence of such disciplined guid-ance and management may derail academic trajectories among theseyouth. Teachers are authority figures who can offer support and guidanceand are likely to stand in when parent-adolescent relationships breakdown. In other words, due to the unique nature of parenting in this popula-tion, teachers may be more capable of filling voids in the lives of AsianAmerican adolescents. In this group, involvement with friends also ap-pears to be an arena of comfort for these adolescents. According toSteinberg et al. (1992), peer involvement is often more positive, in aca-demic terms, among Asian American youth because their peer groups areoften organized around academic activities. Such a phenomenon couldexplain how just spending time with friends, not necessarily receivingsupport from friends, could lessen the academic risk of problems at home.
For younger girls, a supportive friendship can also be an arena of com-fort. Girls tend to be more oriented toward interpersonal relationships andto develop stronger emotional ties with friends (Crosnoe, 2000). For thesereasons, they may be more likely to draw on one source of support whenanother is lacking and to derive more emotional sustenance from relation-ships with age mates in troubled times than boys. As they get older, thefunctionality of this overlap may decrease as their vulnerability to prob-lems at home decrease or as their peer relationships develop into romanticrelationships.
The main premise of this study was that parental and nonparental rela-tionships would condition each other in a functional way. The phenomenajust described capture such functionality. Such functional overlap wasrare; overlap itself was more common. Instead, we often saw what ap-peared to be dysfunctional overlap—potential protective factors related toa stronger association between risk and the academic outcome. In otherwords, we also saw examples of arenas of discomfort. We are hard-pressed to explain such examples, but we can offer some speculation.
The most glaring example of this arena of discomfort phenomenoncenters on teacher bonding. Supportive relationships with teachers weredirectly related to lower off-track behavior in practically every groupstudied. We hypothesized that this source of support would also reducethe association between the parent-related risk and academic behavior,but it did not. In fact, teacher bonding appeared to increase this associa-tion. How could this be? One possibility is that boys, especially older
Crosnoe, Elder / FAMILY DYNAMICS 595
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
boys, may receive greater attention and support from teachers becausethey have a history of academic problems, which, of course, are probablyrelated to their family problems. Such support may translate into morepositive feelings about teachers. Our analyses cannot tease out thebidirectional nature of relations in such a scenario. We should stress againthat although these interactive effects contradict our expectations, thedirect effects of teacher bonding are in line with past research.
Other unexpected findings (e.g., arenas of discomfort) concernsfriendship. Among African American youth, spending more time withfriends exacerbates the impact of parent-related risk, whereas having sup-portive friendships assuages its impact. Heavy involvement with friendsmay take away from school-related activities, whereas emotional supportfrom friends can bolster youth in facing challenges, including academicones. Such findings add to the expanding literature on the role of peers inthe educational experiences of African American youth, suggesting thatthe emotional tone of peer dynamics needs to be considered along withoft-studied peer values. For older Hispanic American youth, however,support from friends is related to a greater association between risk andoutcome. Because Hispanic American families tend to be more orientedtoward family relationships, their drawing of support from friends mightreveal a degree of interpersonal conflict that is related to their overallfunctioning.
Each ethnic group has its own patterns of relationship overlap, andthese patterns may be embedded in the different experiences of eachgroup and the meaning that they attach to the family and to academics. Ingeneral, the risk of emotional distance between parents and adolescents ismore intractable among White and Hispanic American youth and moremalleable among African American and Asian American youth.
The lack of malleability of parent-related risk among Whites may berelated to their economic advantages. In this group, the disadvantage ofweak bonds with parents may lie more in the instrumental support that islost (e.g., parents’ investments and knowledge of education), which maybe harder to replace than psychological resources. For Hispanic Americanyouth, the reasons for this lack of protective overlap are probably quitedifferent. Hispanic American youth are often economically disadvan-taged and, of all groups studied here, have the most academic difficulties.They, like Asian American youth, also represent an incredibly heteroge-neous population, including many immigrants with varying languageskills. The cultural divide that separates many Hispanic American youthfrom other groups and from the American school system may be even
596 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
harder to bridge when they lack a firm foundation of instrumental andemotional support at home.
The greater malleability of parent-related academic risk, in positiveand negative ways, among African American youth largely centers ontheir peer relationships. Such patterns of arenas of comfort and discom-fort suggest that that the importance of their relationships with parents liesin social support—parents making adolescents feel secure and worthy inthe face of a school system that often alienates them. Friends may providealternate sources of support that mirror this ability of parents, even thoughexcessive peer orientation can be problematic. Among Asian Americans,relationship overlap is more often functional than not. The unique orienta-tion of Asian American youth, their friends, and parents to schooling andschool success may increase the likelihood of ecological interactions pro-moting healthy outcomes.
Of course, the interactions presented here, although statistically signif-icant, are small in magnitude. No interaction explained more than onefourth of a standard deviation in the dependent variable of off-track be-havior. The magnitude of these associations begs the question of whetherthey are substantively meaningful. We contend that they are. These asso-ciations arise from a conservative analytical framework that measures achange in academic behavior—behavior that may very well be firmly en-trenched by adolescence—over a 1-year period. Still, they are strongerthan the associations for more status-related factors, such as family back-ground. We also argue that in light of the problematic nature of the behav-iors cataloged in our dependent variable, any significant reduction in off-track behavior is meaningful.
This study has built on past research, especially that of Call andMortimer (2001), by focusing on educational resilience (as opposed togeneral resilience or resilience in other domains) and by drawing on alarger, more representative sample that allows for intragroup analyses ofdevelopmental stage, race/ethnicity, and gender. We encourage others tobuild on this research in several ways.
One clear avenue for future research is to explore other dimensions ofpersonal relationships. Here, we examined the affective tone of parent-adolescent relationships and the supportive features of three types of non-parental relationships. Although we certainly did not measure the parent-adolescent relationship simplistically, we could have also examined otheraspects of this relationship, such as parental hostility and parental non-residency, or other aspects of the family system, such as conflict betweenparents. Other nonparental relationships could also be important to resil-ient pathways, such as those with mentors (e.g., ministers, coaches, Big
Crosnoe, Elder / FAMILY DYNAMICS 597
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
Brothers or Sisters), romantic partners, and grandparents. In looking atsuch relationships, researchers should also take into account the interac-tions between the qualities of the relationship and the characteristics andbehaviors of those involved in the relationship.
Furthermore, our treatment of resilience could certainly be expanded.We examined resilience here by proxy, as the reduction of risk. Resiliencecan be measured more directly by, for example, creating a category ofhigh-achieving youth from problematic family environments and thenpredicting membership in this category. Moreover, the importance ofschool level demonstrated here reinforces the benefits of longitudinal ap-proaches to resilience. The nature of relationships change with age, asdoes the role of young people in school. Modeling trajectories across mul-tiple time points, especially in relation to social context, is the best way tounderstand resilience.
Resilience during the early stages of the life course is a complex phe-nomenon. With few exceptions, risk and protective factors are not univer-sal but vary in relation to different domains, proximate ecological con-texts, and social structural factors. This study only hints at the challengesof studying how some young people cope with risk. The complexity ofresilience, as captured here, illustrates the more general complexity of theadolescent life course. We can only understand the resilient pathways ofyoung people by viewing them as dynamic, intertwined with the experi-ences of significant others, and embedded in the structure of the largersociety.
598 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
AP
PE
ND
IX A
Fiv
e C
ompo
site
Mea
sure
s U
sed
in P
aren
t-A
dole
scen
t D
ista
nce
Con
stru
ct
Fiv
e C
ompo
site
Var
iabl
esIt
ems
Pare
nt-r
epor
ted
lack
of
bond
ing
with
ado
lesc
ent
(M =
1.8
1, S
D =
.64,
α=
.74)
Mea
n of
par
ents
ass
essm
ent o
f ho
w w
ell h
e or
she
get
s al
ong
with
ado
lesc
ent,
the
exte
nt to
whi
ch h
e or
she
mak
es d
e-ci
sion
s ab
out t
he a
dole
scen
ts li
fe w
ith th
e ad
oles
cent
, the
deg
ree
to w
hich
he
or s
he c
an tr
ust t
he a
dole
scen
t, an
d sa
t-is
fact
ion
with
his
or
her
rela
tions
hip
with
the
adol
esce
nt (
1 to
5 =
alw
ays
to n
ever
for
firs
t thr
ee it
ems,
str
ongl
y ag
ree
to s
tron
gly
disa
gree
for
fou
rth)
.A
dole
scen
t-re
port
ed la
ckof
bon
ding
with
par
ents
(M =
1.7
5, S
D =
.66,
α=
.88)
For
fath
ers,
ado
lesc
ents
rat
e ho
w c
lose
they
fee
l to
thei
r fa
ther
s, h
ow lo
ving
thei
r fa
ther
s ar
e, h
ow s
atis
fied
they
are
with
the
com
mun
icat
ion
with
ado
lesc
ent,
and
how
sat
isfi
ed th
ey a
re w
ith th
e re
latio
nshi
p (a
= .8
9). F
or m
othe
rs, a
do-
lesc
ents
ans
wer
sam
e ite
ms,
plu
s ho
w m
uch
they
fee
l tha
t the
ir m
othe
rs c
are
abou
t the
m a
nd h
ow o
ften
thei
r m
othe
rsta
lk w
ith th
em w
hen
thin
gs g
o w
rong
(α=
85)
. Tak
e th
e m
ean
for
each
par
ent,
and
then
the
mea
n ac
ross
par
ents
, if
have
info
rmat
ion
for
both
(1
= v
ery
muc
h, 5
= n
ot a
t all)
.A
dole
scen
t-re
port
ed la
ck o
fco
mm
unic
atio
n w
ith p
ar-
ents
(M
= 2
.16,
SD
=1.
18,α
= .7
0)
For
each
par
ent,
adol
esce
nts
rate
how
oft
en th
ey ta
lked
with
par
ents
, in
the
past
mon
th, a
bout
som
eone
the
adol
esce
ntw
as d
atin
g, a
per
sona
l pro
blem
the
adol
esce
nt w
as h
avin
g, s
choo
l or
grad
es, a
nd th
ings
that
wer
e go
ing
on a
t sch
ool.
Take
the
sum
for
eac
h pa
rent
, and
then
the
mea
n ac
ross
par
ents
, if
have
info
rmat
ion
for
both
(0
= y
es, 1
= n
o).
Ado
lesc
ent-
repo
rted
lack
of s
hare
d ac
tiviti
es w
ithpa
rent
s (M
= 3
.58,
SD
=1.
04,α
= .6
4)
For
each
par
ent,
adol
esce
nts
resp
ond
whe
ther
, in
the
last
mon
th, t
hey
had
gone
sho
ppin
g, p
laye
d a
spor
t, go
ne to
a r
eli-
giou
s ev
ent,
gone
to a
mov
e or
oth
er c
ultu
ral e
vent
, or
wor
ked
on a
pro
ject
with
thei
r pa
rent
s. T
ake
the
sum
for
eac
h,an
d th
en th
e m
ean
acro
ss p
aren
ts, i
f ha
ve in
form
atio
n fo
r bo
th (
0 =
yes
, –1
= n
o).
Ado
lesc
ent-
repo
rted
lack
of f
amily
coh
esio
n (M
=2.
00, S
D =
.70,
α=
.78)
Ado
lesc
ents
ass
ess
the
degr
ee to
whi
ch th
ey f
eel t
hat t
he p
eopl
e in
thei
r fa
mili
es u
nder
stan
d th
em, t
heir
fam
ily h
as f
unto
geth
er, a
nd th
eir
pare
nts
pay
atte
ntio
n to
them
(1
= v
ery
muc
h, 5
= n
ot a
t all)
.
599
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
AP
PE
ND
IX B
Des
crip
tion
s an
d D
escr
ipti
ve S
tati
stic
s fo
r C
ontr
ol V
aria
bles
Con
trol
Var
iabl
eD
escr
ipti
on
Gen
der
(51%
fem
ale,
49%
mal
e)Se
lf-r
epor
ted
gend
er (
1 =
fem
ale,
0 =
mal
e)
Age
(M
= 1
6.01
,SD
= 1
.49)
Self
-rep
orte
d ag
e, in
yea
rs
Pare
nt e
duca
tion
(M =
5.4
8, S
D =
2.2
3)Pa
rent
al r
epor
t of
educ
atio
nal a
ttain
men
t (1
= e
ight
h gr
ade
or le
ss; 2
= m
ore
than
eig
hth
grad
e, le
ss th
an h
igh
scho
olgr
adua
tion;
3 =
voc
atio
nal i
nste
ad o
f hi
gh s
choo
l; 4
= h
igh
scho
ol g
radu
ate;
5 =
com
plet
ed a
GE
D; 6
= v
ocat
iona
lsc
hool
aft
er h
igh
scho
ol; 7
= s
ome
colle
ge; 8
= c
olle
ge g
radu
ate;
9 =
pro
fess
iona
l tra
inin
g). T
he in
terv
iew
ed p
aren
ts(m
ostly
mot
hers
) ga
ve in
form
atio
n fo
r se
lf a
nd s
pous
e. W
e to
ok th
e m
ean
of b
oth
if b
oth
avai
labl
e or
the
valu
e of
one
if o
nly
one
avai
labl
e.R
ace/
ethn
icity
(56
% W
hite
,22
% A
fric
an A
mer
ican
,17
% H
ispa
nic
Am
eric
an,
5% A
sian
Am
eric
an)
Self
-rep
orte
d ra
ce/e
thni
city
(fo
ur d
umm
y va
riab
les)
Pare
nts
educ
atio
nal e
x-pe
ctat
ions
(M
= 1
.30,
SD =
.71)
Pare
nt r
espo
nse
to: H
ow d
isap
poin
ted
wou
ld y
ou b
e if
ado
lesc
ent d
id n
ot g
radu
ate
from
col
lege
? (1
= n
ot d
isap
poin
ted
to 3
= v
ery
disa
ppoi
nted
).
Ado
lesc
ents
edu
catio
nal
expe
ctat
ions
(M
= 4
.46,
SD =
1.0
1)
Ado
lesc
ent r
espo
nse
to: O
n a
scal
e of
1 to
5, w
here
1 is
low
and
5 is
hig
h, h
ow m
uch
do y
ou w
ant t
o go
to c
olle
ge?
Prio
r of
f-tr
ack
beha
vior
(M =
–.3
0, S
D =
3.0
0)
See
expl
anat
ion
for
Wav
e 2
off-
trac
k be
havi
or in
text
.
600
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
REFERENCES
Ainsworth-Darnell, J., & Downey, D. (1998). Assessing racial/ethnic differences in schoolperformance. American Sociological Review, 63, 536-53.
Amato, P. R., & Gilbreth, J. G. (1999). Nonresident fathers and childrens well-being: A meta-analysis. Journal of Marriage and the Family, 61, 557-573.
Arbuckle, J., & Wothke, W. (1999). Amos 4.0 users guide. Chicago: SmallWaters.Bronfenbrenner, U., & Morris, P. (1998). The ecology of developmental processes. In
W. Damon (Ed.), Handbook of child psychology (Vol. 1). New York: John Wiley.Buhrmester, D. (1992). The developmental courses of sibling and peer relationships. In
F. Boer & J. Dunn (Eds.), Childrens sibling relationships: Developmental and clinical is-sues (pp. 19-39). Hillsdale, NJ: Lawrence Erlbaum.
Call, K. T., & Mortimer, J. T. (2001). Arenas of comfort in adolescence: A study of Adjust-ment in context. Mahwah, NJ: Lawrence Erlbaum.
Chao, R.K. (1994). Beyond parental control and authoritarian parenting style: Understand-ing Chinese parenting through the cultural notion of training. Child Development, 65,1111-1119.
Coleman, J. S. (1988). Social capital and the creation of human capital. American Journal ofSociology, 94, S95-120.
Conger, R., Reuter, M., & Conger, K. (1994). The family context of adolescent vulnerabilityand resilience to alcohol use and abuse. Sociological Studies of Children, 6, 55-86.
Crosnoe, R. (2000). Friendships in childhood and adolescence: The life course and new di-rections. Social Psychology Quarterly, 63, 377-391.
Csikszentmihalyi, M., & Schneider, B. (2000). Becoming adult: How teenagers prepare forthe world of work. New York: Basic Books.
Deater-Deckhard, K., Dodge, K., Bates, J. E., & Pettit, G. S. (1996). Physical disciplineamong African American and European American mothers: Links to childrensexternalizing behaviors. Developmental Psychology, 8, 320-325.
Demo, D. H., & Acock, A. C. (1996). Family structure, family process, and adolescent well-being. Journal of Research on Adolescence, 6, 457-488.
Dornbusch, S. M. (1989). The sociology of adolescence. Annual Review of Sociology, 15,233-259.
Dunn, J. (1992). Sisters and brothers: Current issues in developmental research. In F. Boer &J. Dunn (Eds.), Childrens sibling relationships: Developmental and clinical issues(pp. 1-17). Hillsdale, NJ: Lawrence Erlbaum.
Elder, G. H. (1998). The life course as developmental theory. Child Development, 69, 1-12.Elder, G. H., & Conger, R. D. (2000). Children of the land: Adversity and success in rural
America. Chicago: University of Chicago.Furstenberg, F., Cook, T., Eccles, J., Elder, G. H., Jr., & Sameroff, A. (1999). Managing to
make it: Urban families and adolescent success. Chicago: University of Chicago.Garmezy, N., & Masten, A. (1986). Stress, competence, and resilience: Common frontiers
for therapist and psychopathologist. Behavior Therapy, 17, 500-521.Grotevant, H. D. (1998). Adolescent development in family contexts. In W. Damon (Ed.),
Handbook of child psychology (pp. 1097-1147). New York: John Wiley.Hartup, W. (1993). Adolescents and their friends. New Directions in Child Development, 60,
3-22.Hetherington, E. M. (1989). Coping with family transitions: Winners, losers, and survivors.
Child Development, 60, 1-14.
Crosnoe, Elder / FAMILY DYNAMICS 601
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from
Hetherington, E. M. (1998). Social capital and the development of youth from nondivorced,divorced and remarried families. In W. A. Collins & B. Laursen (Eds.), Relationships asdevelopmental contexts. The Minnesota symposia on child psychology (Vol. 30, pp. 177-209). Mahwah, NJ: Lawrence Erlbaum.
Jenkins, J. M., & Smith, M. A. (1991). Factors protecting children living in disharmonioushomes: Maternal reports. Journal of the American Academy of Child and AdolescentPsychiatry, 29, 60-69.
Jessor, R., Van Den Bos, J., Vanderryn, J., Costa, F., & Turbin, M. (1995). Protective factorsin adolescent problem behavior: Moderator effects and developmental change. Develop-mental Psychology, 31, 923-933.
Kuttler, A.F., La Greca, A., and Prinstein, M. (1999). Friendship qualities and socio-emo-tional functioning of adolescents with close cross-sex friendships. Journal of Researchon Adolescence, 9, 339-366.
Muller, C. (2001). The role of caring in the teacher-student relationship for at-risk students.Sociological Inquiry, 71, 241-255.
Muller, C. L., Katz, S. R., & Dance, L. J. (1999). Investing in teaching and learning: Dynam-ics of the teacher-student relationship from each actors perspective. Urban Education,34, 292-337.
Rutter, M. (1985). Resilience in the face of adversity: Protective factors and Resistance topsychiatric disorder. British Journal of Psychiatry, 147, 598-611.
Sanders, M. G., & Jordan, W. J. (2000). Student-teacher relations and academic achievementin high school. In M. G. Sanders (Ed.), Schooling students placed at risk: Research, pol-icy, and practice in the education of poor and minority adolescents (pp. 65-82). Mahwah,NJ: Lawrence Erlbaum.
Settersten, R. A. (1999). Lives in time and place: The problems and promises of developmen-tal science. Amityville, NY: Baywood.
Simmons, R. G., & Blythe, D. A. (1987). Moving into adolescence: The impact of pubertalchange and school context. Hawthorne, NY: Aldine de Gruyter.
Steinberg, L. D., Dornbusch, S. M., & Brown, B. B. (1992). Ethnic differences in adolescentachievement: An ecological perspective. American Psychologist, 47, 723-729.
Steinberg, L. D., Lamborn, S., Dornbusch, S. M., & Darling, N. (1992). Impact of parentingpractices on adolescent achievement: Authoritative parenting, school involvement, andencouragement to succeed. Child Development, 63, 1266-1281.
Windle, M. (1992). A longitudinal study of stress buffering for adolescent problem behav-iors. Developmental Psychology, 28, 522-530.
602 JOURNAL OF FAMILY ISSUES / July 2004
© 2004 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at SAGE Publications on March 25, 2008 http://jfi.sagepub.comDownloaded from