parenting practices and adolescents’ friendship networks

14
CHRIS KNOESTER,DANA L. HAYNIE, AND CRYSTAL M. STEPHENS The Ohio State University Parenting Practices and Adolescents’ Friendship Networks Using data on 11,044 adolescents from the National Longitudinal Study of Adolescent Health, we examine the relationship between parenting practices and the composition of an adolescent’s friendship network. In general, parent-child relationship quality, selecting a neighborhood because of its schools, and parents’ supervision are negatively associated with having a more delinquent friendship net- work but positively associated with having a more prosocial network. These results suggest that parents have the capacity to shape the characteristics of their child’s friendship net- works, even after taking into account the likeli- hood that adolescents have friends who are similar to themselves. Scholars have long investigated the extent to which parents and friends influence adolescent behavior (Simons, Simons, & Wallace, 2004; Warr, 2002), but few have considered whether the effects of parents and friends on adolescent behavior may be connected. Indeed, parents may seek to direct their child away from the influ- ence of delinquent friends and push them toward prosocial friends (Parke & Buriel, 1998; Warr, 2005). Although there is consensus that friends affect adolescent behavior, whether friends or parents are more influential is debated and there has been limited focus on parent-friend links. This study examines how and whether parents influence the composition of an adolescent’s friendship network. First, the relationship between parenting prac- tices and friendship networks matters because friends influence adolescent behavior (Giordano, 2003; Haynie, 2001; Haynie & Osgood, 2005; Warr, 2002). Furthermore, although some simi- larity between the characteristics of adolescents and those of their friendship networks can be attributed to the tendency of adolescents to select as friends those who are like themselves, parents may play an important role in this process (Giordano; Kandel, 1996). Second, it is useful to understand connections between parents and their adolescent child’s friends because of recent debates about the relative influence of parents and friends on adolescent behavior (Kandel, 1996; Warr, 2002). Friends are especially salient to children during adolescence. Some scholars claim that the influ- ence of friends exceeds that of parents during this developmental stage, although other evidence suggests that parents are more influential (Gecas & Seff, 1990; Harris, 1998; Thornburg, 1982). At present, connections between parenting practices and friendship formations are not well under- stood. Finally, sociologists have focused on pre- dicting adolescent outcomes and have been less concerned with the process of relationship for- mation (Giordano, 2003). Developmental psy- chologists have been largely concerned with friendships among children and predicting social Department of Sociology, 300 Bricker Hall, 190 N Oval Mall, The Ohio State University, Columbus, OH 43210 ([email protected]). Key Words: delinquency, deviance, education, friendship networks, parenting practices. Journal of Marriage and Family 68 (December 2006): 1247–1260 1247

Upload: chris-knoester

Post on 21-Jul-2016

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Parenting Practices and Adolescents’ Friendship Networks

CHRIS KNOESTER, DANA L. HAYNIE, AND CRYSTAL M. STEPHENS

The Ohio State University

Parenting Practices and Adolescents’

Friendship Networks

Using data on 11,044 adolescents from theNational Longitudinal Study of AdolescentHealth, we examine the relationship betweenparenting practices and the composition of anadolescent’s friendship network. In general,parent-child relationship quality, selecting aneighborhood because of its schools, andparents’ supervision are negatively associatedwith having a more delinquent friendship net-work but positively associated with havinga more prosocial network. These results suggestthat parents have the capacity to shape thecharacteristics of their child’s friendship net-works, even after taking into account the likeli-hood that adolescents have friends who aresimilar to themselves.

Scholars have long investigated the extent towhich parents and friends influence adolescentbehavior (Simons, Simons, & Wallace, 2004;Warr, 2002), but few have considered whetherthe effects of parents and friends on adolescentbehavior may be connected. Indeed, parentsmay seek to direct their child away from the influ-ence of delinquent friends and push them towardprosocial friends (Parke & Buriel, 1998; Warr,2005). Although there is consensus that friendsaffect adolescent behavior, whether friends or

parents are more influential is debated and therehas been limited focus on parent-friend links.This study examines how and whether parentsinfluence the composition of an adolescent’sfriendship network.

First, the relationship between parenting prac-tices and friendship networks matters becausefriends influence adolescent behavior (Giordano,2003; Haynie, 2001; Haynie & Osgood, 2005;Warr, 2002). Furthermore, although some simi-larity between the characteristics of adolescentsand those of their friendship networks can beattributed to the tendency of adolescents to selectas friends those who are like themselves, parentsmay play an important role in this process(Giordano; Kandel, 1996).

Second, it is useful to understand connectionsbetween parents and their adolescent child’sfriends because of recent debates about therelative influence of parents and friends onadolescent behavior (Kandel, 1996; Warr, 2002).Friends are especially salient to children duringadolescence. Some scholars claim that the influ-ence of friends exceeds that of parents during thisdevelopmental stage, although other evidencesuggests that parents are more influential (Gecas& Seff, 1990; Harris, 1998; Thornburg, 1982). Atpresent, connections between parenting practicesand friendship formations are not well under-stood.

Finally, sociologists have focused on pre-dicting adolescent outcomes and have been lessconcerned with the process of relationship for-mation (Giordano, 2003). Developmental psy-chologists have been largely concerned withfriendships among children and predicting social

Department of Sociology, 300 Bricker Hall, 190 N OvalMall, The Ohio State University, Columbus, OH 43210([email protected]).

Key Words: delinquency, deviance, education, friendshipnetworks, parenting practices.

Journal of Marriage and Family 68 (December 2006): 1247–1260 1247

Page 2: Parenting Practices and Adolescents’ Friendship Networks

competence (Giordano; Parke & Bhavnagri,1989; Sroufe, Egeland, & Carlson, 1999). Inaddition, conceptual frameworks typically stressthe mediating effects of child behaviors in therelationship between parents and peers (Brown,Mounts, Lamborn, & Steinberg, 1993; Dishion,Capaldi, Spracklen, & Li, 1995; Hartup & Laursen,1999). These approaches offer little consider-ation or evidence of the means through whichparents may affect the friendship network charac-teristics of their adolescent children, afteraccounting for the likelihood that adolescentsselect friends who are similar to themselves orhave been influenced by their friends already.

Our study goes beyond extant research on theparent-friend link. First, we integrate theoreticalperspectives from criminology (Hirschi, 1969;Nye, 1958) and developmental psychology(Parke & Bhavnagri, 1989; Parke & Buriel,1998) to consider whether and how parents influ-ence a child’s choice of friends, even after takinginto account self-selection into friendship net-works and the influence of friends on an adoles-cent’s characteristics. The use of statisticalcontrols for the similarity between the character-istics of adolescents and their friends allows us tofocus on more direct mechanisms through whichparents may affect friendship networks and alsoproduces conservative estimates of the extent towhich parenting practices are associated withadolescent friendship network characteristics.

Second, our study is more comprehensive thanprevious work because we assess the relation-ships between three dimensions of parentingpractices (i.e., neighborhood choice, monitoringand supervision, and parent-child relationshipquality) and the delinquent (i.e., fighting anddeviance) and prosocial (i.e., grade point average,extracurricular activities, and college expecta-tions) characteristics of friendship networks.

Finally, we use data from the National Longi-tudinal Study of Adolescent Health that offerthree unique benefits: (a) a large nationally repre-sentative sample of adolescents; (b) network datathat allow us to define comprehensive friendshipnetworks; and (c) self-reports of attitudes andbehaviors from the main adolescent respondent,the adolescent’s parents, and the adolescent’sfriends. These features of the data improve thevalidity and generalize ability of our findings byminimizing the likelihood that our results area function of a small sample, the result of anunrepresentative sample, or biased by reportsfrom a single source.

CONCEPTUAL FRAMEWORK

Social control theory is often used to describehow parents affect a child’s risk of delinquency(Hirschi, 1969; Nye, 1958). Nye emphasized thatparents influence a child through direct, indirect,and internalized controls. Direct controls involvesupervision and the threat of punishment. Indirectcontrols operate when children act so as not todisappoint their parents. Internalized controlsrefer to the role that a youth’s conscience playsin making decisions. Thus, parents reduce child-ren’s risk of engaging in delinquent behaviorsthrough close monitoring, attachment, andsocialization to disapprove of delinquent behav-iors, and to realize that their parents disapproveof delinquency as well (Hirschi; Nye). Nye’sframework of direct, indirect, and internalizedcontrols can also be applied to the process bywhich parents may influence the composition ofa child’s friendship networks.

Parke and colleagues’ model of parentingprocesses (Parke & Bhavnagri, 1989; Parke &Buriel, 1998) does not focus on delinquency orprosocial behaviors per se but suggests thatparents can have direct and indirect effects ona child’s friendship formations. Direct influencesare those that are meant to affect a child’s friend-ships explicitly. Indirect influences occur whenthe goal is not to affect friendships explicitly.Below, we identify five ways through whichparents may directly or indirectly affect the com-position of a child’s friendship network.

First, parents may directly influence a child’schoice of friends by manipulating the environ-ment (Parke & Bhavnagri, 1989). They mayselectively choose neighborhoods to live in andschools for a child to attend. Neighborhoods areimportant because their safety is negatively asso-ciated with having deviant peers (Brody et al.,2001). Further, school characteristics affect edu-cational outcomes beyond the influence of signi-ficant others (Buchmann & Dalton, 2002).

Second, parents may influence a child’s friend-ship network by supervising and monitoring thechild (Parke & Bhavnagri, 1989; Warr, 2005).In doing so, parents may implicitly or explicitlyencourage their child to seek out less delinquentand more prosocial friends. Time spent underthe watch of parents also reduces the amount oftime available to be alone with friends (Warr,1993). There is some evidence that parental mon-itoring is positively associated with having anacademically oriented friendship group and

1248 Journal of Marriage and Family

Page 3: Parenting Practices and Adolescents’ Friendship Networks

negatively associated with having a delinquentfriendship group (Brown et al., 1993; Warr,2005).

Parents may indirectly influence a child’sfriendship network by modeling desirable behav-ior and socializing the child to disapprove ofdelinquency and to embrace conventional values(Warr, 1993; Whitbeck, Simons, Conger, &Lorenz, 1989). Parents—even parents whothemselves violate the law—exhibit almost uni-versal disapproval of adolescent delinquencyand encourage prosocial orientations (Hirschi,1969). Socialization into altruistic values andat least moderate parental discipline are nega-tively associated with having delinquent friends(Whitbeck et al., 1989). In addition, a child whohas not learned prosocial skills at home may berejected by conventional peers and may fall intoa delinquent friendship group (Patterson, Reid,& Dishion, 1992; Simons et al., 2004).

Finally, the quality of parent-child relationsmay indirectly affect friendship network compo-sition (Parke & Bhavnagri, 1989; Warr, 2005).High-quality parent-child relations may encour-age a child to choose friends whom they knowtheir parents will like (Warr, 1993). Low-qualityparent-child relations may prompt a child to careless about parental preferences and to be moreinclined to seek out alternative sources of socialintegration, such as through membership ina delinquent friendship group. Indeed, ineffectiveparenting practices such as harsh discipline, lowlevels of involvement, family conflict, and a lackof nurturance may negatively influence parent-child relationship quality and may lead a childinto affiliations with delinquent friends (Brodyet al., 2001; Patterson et al., 1992; Simons et al.,2004). In fact, the effectiveness of socializationand modeling by parents may depend on parent-child relationship quality: A child may be morelikely to value parental input and modeling whenthe parent-child relationship is strong.

CONFOUNDING FACTORS

Family structure, socioeconomic status, an adoles-cent’s age, gender, and race are associated with par-enting practices, adolescent deviance, and adolescenteducational outcomes (Buchmann & Dalton, 2002;Simons et al., 2004; Wojtkiewicz & Donato,1995). Because adolescents and their friends havesimilar characteristics (Buchmann &Dalton;Simonset al.; Warr, 2002), these background factors andan adolescent’s own characteristics may affect

the relationship between parenting practices andthe composition of adolescent friendship net-works. Residential mobility and having friendswho are not schoolmates may also confound theassociation between parenting practices and in-school friendship networks (Haynie, 2001).Therefore, we control for the effects of back-ground factors, adolescent’s behaviors, residen-tial mobility, and the selection of friends fromoutside of school in our analysis.

HYPOTHESES

On the basis of our conceptual framework and pre-vious research, we expect that parents who selecttheir neighborhood because it has lower crime thanothers and because the quality of its schools arebetter are more likely to have a child with moreprosocial and fewer delinquent friendships. Weanticipate that parental monitoring and supervi-sion are positively associated with more prosocial,and negatively associated with more delinquent,friendships. Finally, we expect that high parent-child relationship quality and low levels of conflictare associated with having less delinquent andmore prosocial friendships.

Because positive parenting practices may befound in clusters and may be less influential inaffecting the characteristics of an adolescent’sfriends when an adolescent is extremely delin-quent or prosocial (Simons et al., 2004; Warr,2005), we also consider interaction effects inour analyses. For example, it may be that parent-ing practices are especially influential whenthey are consistent across multiple dimensions.Also, parenting practices may be less likely tobe associated with a child’s friendship networkcharacteristics when the adolescent is especiallydelinquent or prosocial. These adolescents maystrongly identify with their status, may be moreprone to select friends like themselves, and mayresist being influenced by their parents.

METHOD

Data

Data for this study come from the first wave ofdata collection in the National LongitudinalStudy of Adolescent Health (Add Health),a nationally representative sample of 7th to 12thgrade students from the 1994 – 1995 school year.The study involved an in-school questionnaire

Parenting and Adolescents’ Friendship Networks 1249

Page 4: Parenting Practices and Adolescents’ Friendship Networks

followed by an in-home survey. In the in-schoolsurvey, all students attending school on the dayof the self-administered questionnaire in each ofthe 132 high schools and middle schools weresurveyed (N ¼ 83,135). The information thatthis sample of adolescents provided allows forthe construction of friendship network charac-teristics. Respondents were given school rostersand were asked to list up to five male and fivefemale friends. Because all students in theschool were surveyed, it is possible to recreatethe friendship networks within schools.

Following the in-school questionnaire, a ran-domly sampled subset of 14,396 adolescentswho participated in the school survey were inter-viewed in their homes. In most cases, a parent wasinterviewed as part of the in-home survey. Ourfinal sample includes adolescents who partici-pated in the in-school survey, the in-home inter-view, had a parent participate in the survey, andfor whom we have network data and weights(N ¼ 11,044). The further reduction of the sam-ple resulted from missing network data (i.e.,identifying friends who were not interviewed,n ¼ 871 [6%]), parenting information (i.e., par-ent not interviewed, n ¼ 1,261 [9%]), andweights (n ¼ 1,220 [9%]) that are used to ac-count for clustering, oversampling, and nonres-ponses in the data (Chantala & Tabor, 1999).We use multiple imputation for the modestamount of missing data present in our controlvariables because of the shortcomings of list-wise deletion of missing data (Acock, 2005).

Missing data were more likely to exist amongsingle parent, Black, Hispanic, and low–socio-economic status families. It was also morelikely to occur among adolescents who recentlychanged schools and who identified more friendsfrom outside of school. School-level and student-level weights, however, attempt to account for theprobability of nonresponse (see Tourangeau &Shin, 1999).

Measures

Descriptive characteristics for our dependent,independent, and control variables are shown inTable 1. Our dependent variables are drawn fromself-reports of delinquent and prosocial charac-teristics reported by each friend of the adolescentrespondents. We consider each adolescent who isnominated by the respondent as a friend and eachadolescent who nominated the respondent as

a friend (i.e., the send and receive network).Two of our dependent variables measure delin-quent friendship behaviors: friends’ fightingand friends’ deviance. Friends’ fighting is theaverage amount of fighting reported by an adoles-cent respondent’s friends in the past 12 months.Friends’ deviance is the average amount of devi-ance each friend reported from getting drunk,smoking cigarettes, or skipping school (a¼ .76).

We consider three measures of the prosocialcharacteristics of friendship networks. Friends’grade point average is the mean grade point aver-age reported by friends. Friends’ college expect-ations is based on responses to a question aboutone’s expectations of graduating from college,where higher scores equal a greater perceivedlikelihood of graduating from college (0 ¼ nochance, 8 ¼ it will happen). We take the aver-age report of all friends. Friends’ participationin extracurricular activities indicates the averagenumber of after school extracurricular activitiesin which friends reported participating.

Three dimensions of parenting serve as ourindependent variables: choice of neighborhood,monitoring and supervision of children, and thequality of parent-child relationships. We usetwo dichotomous variables to consider the influ-ence of neighborhood choice. One item consistsof parents’ reports (1 ¼ yes) of whether theychose to live in their current neighborhoodbecause of its low crime relative to others. Thesecond item (1 ¼ yes) indicates whether theychose to live in their current neighborhood becauseit has better schools than other neighborhoods.

We examine the importance of monitoring andsupervision of children with three indicators.First, we use a supervision scale that consists ofitems that asked about a parent’s knowledge oftheir children’s friends (a ¼ .61). Parents wereasked if they had met their child’s best friend(1 ¼ yes), their child’s best friend’s parents(1 ¼ yes), and the total number of the parents oftheir child’s friends that they had actually met.We standardized each item and took the meanto form the scale. Second, we use an availabilityscale that sums adolescents’ responses to ques-tions regarding when their parents were home atvarious times of the day (a ¼ .58). Adolescentsreported how often their mother and father werehome when they left for school in the morning,came home from school in the afternoon, andwent to bed at night. We used the maximumvalue of adolescents’ responses for each item inreference to their mother and father, standardized

1250 Journal of Marriage and Family

Page 5: Parenting Practices and Adolescents’ Friendship Networks

each subsequent value, and used the mean asthe scale score. Finally, we use a scale that as-sesses adolescent autonomy (a ¼ .63). We stan-dardized and then took the mean of adolescents’responses regarding whether their parents allowthem to make their own decisions (1 ¼ yes)about their curfew on weekends, the amount oftime the adolescents watch television, whichshows the adolescents watch on television, whattime adolescents go to bed on weeknights, andwith whom they hang out.

Finally, we analyze parent-child relationshipquality with two scales that focus on (a) aspectsof a quality parent-child relationship and (b)parent-child conflict. The quality parent-childrelationship scale (a ¼ .65) integrates reportsof parent-child interaction, communication, andattachment. Adolescents reported on whetherthey had done the following things with theirmother and their father within the prior 4 weeks:

went shopping; played a sport; went to church/religious event; went to a movie, show, museum,or sporting event; and worked on a project forschool (1 ¼ yes). Adolescents also reportedwhether they had recently talked to their motherand father about a party or person they weredating, a personal problem, their grades, andother things going on at school (1 ¼ yes).Finally, our scale includes items in which ado-lescents rate their attachment to both theirmother and father. The adolescents reportedhow close they feel to their mother and fatherand how much they feel that each parent caresabout them (1 ¼ not at all, 5 ¼ very much). Wetook the maximum response for each item(which may be in reference to either parent),standardized each, and used the mean as thescale score. Our standardized scale (a ¼ .74) ofparent-child conflict uses both parent and ado-lescent measures of conflict in their relationship.

Table 1. Descriptive Statistics for All Variables (N ¼ 11,044)

M SD Minimum Maximum

Friends’ fighting 0.73 0.59 0 4

Friends’ delinquency 2.86 2.39 0 18

Friends’ grade point average 2.80 0.52 0 4

Friends’ extracurricular activities 2.22 1.21 0 9

Friends’ college expectations 6.44 1.36 0 8

Chose neighborhood for lower crime 0.13 0.34 0 1

Chose neighborhood for better schools 0.11 0.32 0 1

Parent-child relationship quality 0.00 1.00 2.12 9.53

Parent-child conflict 0.00 1.00 1.09 2.3

Parents’ supervision 0.00 1.00 3.59 2.36

Parents’ availability 0.00 1.00 1.08 3.04

Adolescent’s autonomy 0.00 1.00 2.12 0.67

Two parents (at least one nonbiological parent)a

0.19 0.39 0 1

Single parenta

0.24 0.43 0 1

Other family structurea

0.06 0.25 0 1

Black 0.23 0.42 0 1

Hispanic 0.16 0.38 0 1

Age 15.61 1.75 11 21

Female 0.51 0.50 0 1

New school 0.27 0.38 0 1

Family socioeconomic status 6.04 2.51 0 10

Friends from outside of school 1.38 1.77 0 10

Adolescent’s fighting 0.79 0.94 0 4

Adolescent’s delinquency 2.99 3.35 0 18

Adolescent’s grade point average 2.75 0.69 0 4

Adolescent’s extracurricular activities 2.31 2.40 0 9

Adolescent’s college expectations 6.32 1.86 0 8

aReference category is having two biological parents.

Parenting and Adolescents’ Friendship Networks 1251

Page 6: Parenting Practices and Adolescents’ Friendship Networks

Adolescents reported whether they had hada serious argument with their mother and fatherin the previous 4 weeks (1 ¼ yes). Parents re-ported how often they distrusted their child, didnot understand their child, did not get alongwith their child, and felt their child interferedwith their activities (1 ¼ never, 5 ¼ always).We standardized each of these measures indi-vidually and used the mean as the scale score.

Race, age, gender, family structure, familysocioeconomic status, and adolescent respond-ent’s characteristics are included as control vari-ables in all analyses. We use dichotomousvariables to indicate Black race-ethnicity andHispanic race-ethnicity. We also account for ado-lescent’s age (in years) and gender (1 ¼ female).Family structure consists of four categories:(a) adolescent lives with two biological parents,(b) some other two parent structure, (c) a singleparent only, and (d) nobody that the adolescentidentifies as a parent. Family socioeconomicstatus is based on the educational attainment andoccupational prestige of parents (see Moody,2001). Adolescent’s own characteristics are takenfrom responses to the same questions that weuse to construct our dependent variables. Wealso include controls for whether the adolescenttransferred to a new school in the past year (1 ¼yes) and the number of out of school nomina-tions the adolescent provided.

ANALYTIC STRATEGY

We examine whether parenting practices areassociated with friends’ fighting, delinquency,grade point average, involvement in extracurric-ular activities, and college expectations in twomodels for each dependent variable. In all mod-els, we control for the association between theadolescent child’s behavior and friendship net-work characteristics. Doing so allows us to isolatethe relationships between parenting practices andfriends’ characteristics after taking into accountthe average likelihood that children have friendswho are similar to themselves. First, we analyzethe additive relationships between parentingpractices and the characteristics of adolescent’sfriendship networks. Then, we create multiplica-tive interaction terms to test for significant inter-actions between our parenting practices, andbetween parenting practices and the adolescentchild’s behavior, in predicting the delinquentand prosocial orientations of the adolescent’sfriends. To aid in our interpretation of interaction

effects, if an interaction term is significant atp , .05 when it is entered into the basic modelalone, we graphed the equation (results notshown). In graphing the interaction equations,we use combinations of low (i.e., 1 SD , M),average (i.e., M), and high (i.e., 1 SD . M) val-ues of the variables involved in the interaction;all other variables in the equation are set to theirmean values. Finally, we present a full interac-tion model that contains the variables from ourbasic model plus each interaction term that wassignificant, when examined alone. This strategyallows us to present evidence of interaction ef-fects in a succinct manner.

Because we have dependent variables withunique distributions, we use Poisson, negativebinomial, and OLS regression techniques. Forour skewed dependent variables, OLS (ordinaryleast squares) is not appropriate for our analyses.We use negative binomial regression to analyzefriends’ delinquency because although a largenumber of friends report no or low levels of delin-quency, a small number report very high levels(i.e., the measure exhibits overdispersion). Weuse Poisson regression models to predict friends’fighting and expectations of college graduationbecause these variables have a nonnormal distri-bution but are not overdispersed. Finally, we ana-lyze friends’ grade point averages andinvolvement in extracurricular activities usingOLS regression because these measures approxi-mate normal distributions.

RESULTS

Below, we discuss the findings for each analysisof friends’ delinquent and prosocial behaviors.The coefficients from the negative binomial andPoisson models can be interpreted easily afterexponentiating them. A one-unit increase in theindependent variable multiplies the expectedlevel of the dependent variable by exp (Bj),where Bj is the parameter estimate for the jthvariable. The coefficients from the regressionmodels indicate the additive effect of a one-unitincrease in the independent variables on thedependent variable.

We first focus on predicting friends’ fighting.As shown in Model 1 of Table 2, when parentsreport having chosen their current neighborhoodbecause of its good schools (b ¼ �0.08, p ,.01), their adolescent children have friends whoengage in 8% less fighting (e

�.08 ¼ .92) thanthe friends of adolescents whose parents did not

1252 Journal of Marriage and Family

Page 7: Parenting Practices and Adolescents’ Friendship Networks

report choosing their neighborhood because ofits good schools. The coefficient for parent-child relationship quality (b ¼ �0.11, p , .001)indicates that adolescents who report levels ofparent-child relationship quality that are onestandard deviation above the mean have friendswho are involved in 10% less fighting than ado-lescents who report average levels of parent-child relationship quality (e�.11 ¼ .90). Similarly,a one standard deviation increase in parent-childconflict (b ¼ 0.06, p , .001) is associated withhaving friends who engage in 6% more fighting.

There is evidence of three significant interac-tions in predicting friends’ fighting and theseare displayed in Model 2 of Table 2. First, theassociation between autonomy and friends’ fight-ing depends on an adolescent’s own behavior, tosome extent. Autonomy is negatively associatedwith friends’ fighting, except when the adoles-

cent child fights frequently. When the adolescentfights frequently, increased autonomy is associ-ated with having friends who frequently engagein fighting. In addition, there is evidence thatparents’ supervision and availability conditionthe association between parent-child relationshipquality and friends’ fighting. At higher levels ofsupervision, the negative association betweenparent-child relationship quality and friends’fighting is enhanced. In other words, adolescentsare least likely to have friends who fight fre-quently if they experience high levels of parent-child relationship quality and supervision. Parents’availability moderates the relationship qualityfriends’ fighting association such that relation-ship quality seems to matter more (i.e., be morenegatively associated) in predicting friends’fighting when parents are not available. Theseinteraction effects are consistent with the

Table 2. Poisson Regression of Friends’ Fighting on Parenting Practices (N ¼ 10,915)

Model 1 Model 2

b Exp b b Exp b

Parenting variables

Chose neighborhood for less crime �0.03 0.97 �0.03 0.97

Chose neighborhood for better schools �0.08** 0.92 �0.08** 0.92

Quality parent-child relations �0.11*** 0.90 �0.10*** 0.90

Parent-child conflict 0.06*** 1.06 0.06*** 1.06

Parents’ supervision 0.01 1.01 0.01 1.01

Parents’ availability 0.00 1.00 0.00 1.00

Adolescent’s autonomy �0.03 0.97 �0.07* 0.93

Control variables

Two parents (at least one nonbiological)a

0.04 1.04 0.04 1.04

Single parenta

0.06** 1.06 0.06** 1.06

Other family structurea

0.04 1.04 0.04 1.04

Adolescent’s fighting 0.07*** 1.07 0.08*** 1.08

Age �0.10*** 0.90 �0.10*** 0.90

Female �0.13*** 0.88 �0.13*** 0.88

Black 0.08* 1.08 0.08* 1.08

Hispanic 0.08* 1.08 0.07* 1.07

New school 0.04 1.04 0.03 1.03

Family socioeconomic status �0.04*** 0.96 �0.04*** 0.96

Friends from outside of school 0.01 1.01 0.00 1.00

Interaction variables

Adolescent’s Fighting 3 Autonomy 0.03** 1.03

Quality 3 Supervision �0.07* 0.93

Quality 3 Availability 0.07** 1.07

Note: Interaction terms in Model 2 were significant at p , .05 when entered into Model 1, alone.aReference category is having two biological parents.

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

Parenting and Adolescents’ Friendship Networks 1253

Page 8: Parenting Practices and Adolescents’ Friendship Networks

expectation that the association between parent-ing practices and friends’ characteristics maydepend on the child’s own characteristics andthe combination of parenting practices that areemployed.

We now turn to our analysis of friends’ devi-ance, the results of which are displayed inTable 3. In Model 1, we find that parent-childconflict (b ¼ 0.06, p , .001) is positively asso-ciated with having friends who report higherinvolvement in deviance. Compared to adoles-cents who report average levels of parent-childconflict, adolescents who report levels of parent-child conflict that are one standard deviationabove the mean have friends who engage in 6%more deviance (e

06 ¼ 1.06). Four interaction ef-fects are evident in predicting friends’ devianceand they are displayed in Model 2. First, pa-rents’ availability has a modest negative associ-

ation with friends’ deviance but only when theadolescent child’s own deviance is not high(i.e., below one standard deviation above themean). Second, parents’ supervision has a slightnegative association with friends’ deviance butonly when parent-child relationship quality ishigh. Again, adolescents appear to be leastlikely to have delinquent friends if they havehigh levels of both parent-child relationshipquality and supervision. Third, there is evidencethat the relationship between friends’ fightingand parents’ supervision depends on parent-child conflict. Higher levels of supervision arenegatively associated with friends’ deviancewhen low parent-child conflict exists, but higherlevels of supervision are positively associatedwith friends’ deviance when parent-child con-flict is prevalent. Finally, there is a modest neg-ative association between parents’ availability

Table 3. Negative Binomial Regression of Friends’ Deviance on Parenting Practices (N ¼ 10,984)

Model 1 Model 2

b Exp b b Exp b

Parenting variables

Chose neighborhood for less crime 0.01 1.01 0.01 1.01

Chose neighborhood for better schools �0.06 0.94 �0.06 0.94

Quality parent-child relations �0.03 0.97 �0.03 0.97

Parent-child conflict 0.06*** 1.06 0.06*** 1.06

Parents’ supervision 0.01 1.01 0.01 1.01

Parents’ availability �0.02 0.98 �0.05* 0.95

Adolescent’s autonomy 0.00 1.00 0.01 1.01

Control variables

Two parents (at least one nonbiological)a

0.07** 1.07 0.07** 1.07

Single parenta

0.05* 1.05 0.05* 1.05

Other family structurea �0.02 0.98 �0.02 0.98

Adolescent’s deviance 0.07*** 1.07 0.07*** 1.07

Age 0.15*** 1.16 0.15*** 1.16

Female 0.05* 1.05 0.05* 1.05

Black �0.19*** 0.83 �0.19*** 0.83

Hispanic �0.01 0.99 �0.01* 0.99

New school 0.04 1.04 0.04 1.04

Family socioeconomic status �0.02*** 0.98 �0.02*** 0.98

Friends from outside of school 0.01 1.01 0.01 1.01

Interaction variables

Adolescent’s Deviance 3 Availability 0.01 1.01

Quality 3 Supervision �0.05 0.95

Conflict 3 Supervision 0.05 1.05

Availability 3 Autonomy 0.07** 1.07

Note: Interaction terms in Model 2 were significant at p, .05 when entered into Model 1, alone.aReference category is having two biological parents.

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

1254 Journal of Marriage and Family

Page 9: Parenting Practices and Adolescents’ Friendship Networks

and friends’ deviance but only when adoles-cents do not have high levels of autonomy.

The findings from our analysis of friends’grade point average are displayed in Table 4.As shown in Model 1, when parents reportchoosing their neighborhood because of its goodschools, their adolescent children have friendswith higher grade point averages (b ¼ 0.04, p, .05). Also, friends’ grade point average ispositively associated with parent-child rela-tionship quality (b ¼ 0.04, p , .05) andparents’ supervision (b ¼ 0.03, p , .001) butnegatively associated with parent-child conflict(b ¼ �0.04, p , .01). As shown in Model 2,the relationships involving our indicators ofparent-child relationship quality are moderatedby the adolescent child’s own grade point aver-age. The interactions indicate that higher levels

of parent-child relationship quality are morelikely to be associated with an adolescent childhaving friends with slightly higher grade pointaverages when the adolescent’s own grade pointaverage is higher than a value of one standarddeviation below the mean grade point average.Similarly, higher levels of parent-child conflictare more likely to be associated with an adoles-cent child having friends with slightly lowergrade point averages when the adolescent’s owngrade point average is above one standard devi-ation below the mean grade point average.Finally, there is evidence that the associationbetween supervision and friends’ grade pointaverages depends on parent-child relationshipquality. Higher levels of supervision are posi-tively associated with friends’ grade point aver-ages but only when parent-child relationship

Table 4. OLS Regression of Friends’ GPA on Parenting Practices (N ¼ 10,880)

Model 1 Model 2

b SE b SE

Parenting variables

Chose neighborhood for less crime 0.00 0.02 �0.01 0.02

Chose neighborhood for better schools 0.04* 0.02 0.04* 0.02

Quality parent-child relations 0.04* 0.02 �0.05 0.02

Parent-child conflict �0.04** 0.01 0.06 0.03

Parents’ supervision 0.03*** 0.01 0.04*** 0.01

Parents’ availability 0.00 0.01 0.00 0.01

Adolescent’s autonomy �0.01 0.02 �0.01 0.02

Control variables

Two parents (at least one nonbiological)a �0.03 0.02 0.04 0.02

Single parenta

0.00 0.02 0.06** 0.02

Other family structurea �0.04 0.04 0.04 0.04

Adolescent’s GPA 0.23*** 0.01 0.08*** 0.01

Age �0.02* 0.01 �0.10*** 0.01

Female �0.01 0.01 �0.13*** 0.01

Black �0.14*** 0.03 0.08* 0.03

Hispanic �0.11** 0.03 0.07* 0.03

New school �0.01 0.02 0.03 0.02

Family socioeconomic status 0.03*** 0.00 �0.04*** 0.00

Friends from outside of school �0.01 0.01 0.00 0.01

Interaction variables

Adolescent’s GPA 3 Quality 0.03** 0.02

Adolescent’s GPA 3 Conflict �0.07* 0.01

Quality 3 Supervision 0.07** 0.02

Note: Interaction terms in Model 2 were significant at p , .05 when entered into Model 1, alone. GPA ¼ grade point

average.aReference category is having two biological parents.

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

Parenting and Adolescents’ Friendship Networks 1255

Page 10: Parenting Practices and Adolescents’ Friendship Networks

quality is not low. Friends’ grade point averagesare highest when parent-child relationship qual-ity and supervision are both at high levels.

Friends’ involvement in extracurricular ac-tivities is the focus in Table 5. The coefficientsfrom Model 1 indicate that parent-child relation-ship quality (b ¼ 0.17, p , .001), parents’supervision (b ¼ 0.12, p , .001), and adoles-cent’s autonomy (b ¼ 0.07, p , .05) are posi-tively associated with friends’ involvement inextracurricular activities. Interaction effects inModel 2 suggest that the positive effects of pa-rents’ supervision depend on parent-child rela-tionship quality and parents’ availability. Thepositive association between supervision andfriends’ extracurricular activities is strengthenedwhen parent-child relationship quality and pa-rents’ availability is higher.

Table 6 identifies the parenting practices thatare associated with adolescents having friendswith college expectations. Model 1 indicates that

compared to adolescents whose parents did notreport choosing their neighborhood because ofbetter schools, adolescents who have parentswho report choosing their neighborhood on thebasis of the quality of its schools have friendswhose expectations of attending college arehigher (b ¼ 0.02, p , .05). Parent-child rela-tionship quality is positively associated withfriends’ college expectations (b ¼ 0.02, p ,.01), whereas parent-child conflict is negativelyassociated with friends’ college expectations(b ¼ �0.01, p , .05). Two interactions areapparent in Model 2. First, autonomy interactswith an adolescent’s own college expectationsin predicting friends’ expectations. The associa-tion between autonomy and friends expectationsis only positive when an adolescent child doesnot have low expectations of going to college(i.e., greater than one standard deviation belowthe average level of autonomy). Friends’ collegeexpectations are highest when the adolescent

Table 5. OLS Regression of Friends’ Extracurricular Activities on Parenting Practices (N ¼ 11,044)

Model 1 Model 2

b SE b SE

Parenting variables

Chose neighborhood for less crime �0.02 0.04 �0.02 0.04

Chose neighborhood for better schools 0.08 0.06 0.08 0.06

Quality parent-child relations 0.17*** 0.04 0.16*** 0.05

Parent-child conflict �0.05 0.03 �0.05 0.03

Parents’ supervision 0.12*** 0.03 0.13*** 0.03

Parents’ availability 0.01 0.03 0.00 0.03

Adolescent’s autonomy 0.07* 0.03 0.07* 0.03

Control variables

Two parents (at least one nonbiological)a �0.10* 0.04 �0.09* 0.04

Single parenta �0.03 0.05 �0.03 0.05

Other family structurea

0.03 0.08 0.03 0.08

Adolescent’s extracurricular activities 0.23*** 0.01 0.13*** 0.01

Age �0.07** 0.02 �0.07** 0.02

Female 0.04 0.03 0.04 0.03

Black �0.03 0.07 �0.03 0.07

Hispanic �0.30*** 0.08 �0.30*** 0.08

New school �0.12* 0.06 �0.12* 0.06

Family socioeconomic status 0.06*** 0.01 0.06*** 0.01

Friends from outside of school �0.03** 0.01 �0.03** 0.01

Interaction variables

Quality 3 Supervision 0.13** 0.03

Supervision 3 Availability 0.06** 0.05

Note: Interaction terms in Model 2 were significant at p, .05 when entered into Model 1, alone.aReference category is having two biological parents.

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

1256 Journal of Marriage and Family

Page 11: Parenting Practices and Adolescents’ Friendship Networks

child’s expectations are high and the child hashigh autonomy. Second, higher levels of super-vision are positively associated with friends’college expectations but only when relationshipquality is not low. Friends’ college expectationsare highest when both supervision and parent-child relationship quality are high.

DISCUSSION

In this study, we considered whether and how pa-rents influence the characteristics of their adoles-cent child’s friendship networks, after taking intoaccount their child’s own behaviors. The findingssuggest that parents can shape the network offriends that their child has, even after consideringthe extent to which their child may select, and beinfluenced by, friends who share similar attitudesand behaviors. Parent-child relationship quality,selecting a neighborhood because it has better

schools than other neighborhoods, and parentalsupervision are associated with friendship net-work characteristics.

The most consistent predictors of friendshipnetwork characteristics are the quality and con-flict in parent-child relationships. As expected,our findings are consistent with research that indi-cates that higher levels of parent-child relation-ship quality enhance child outcomes (Knoester,2003; Simons et al., 2004), if one assumes thathaving friends with more prosocial and fewerdelinquent characteristics is a positive outcomefor a child. The results also support the argumentthat parent-child bonds function as social controls(Nye, 1958).

There is also evidence that neighborhoodchoice can affect a child’s friendship network,as we anticipated. When parents report selectinga neighborhood because of its better schools, anadolescent is more likely to have friends whofight less, have higher grade point averages, and

Table 6. Poisson Regression of Friends’ College Expectations on Parenting Practices (N ¼ 10,902)

Model 1 Model 2

b Exp b b Exp b

Parenting variables

Chose neighborhood for less crime 0.00 1.00 0.00 1.00

Chose neighborhood for better schools 0.02* 1.02 0.02* 1.02

Quality parent-child relations 0.02** 1.02 0.02** 1.02

Parent-child conflict �0.01* 0.99 �0.01* 0.99

Parents’ supervision 0.00 1.00 0.00 1.00

Parents’ availability 0.00 1.00 0.00 1.00

Adolescent’s autonomy 0.01 1.01 �0.03 0.97

Control variables

Two parents (at least one nonbiological)a �0.02 0.98 �0.02 0.98

Single parenta �0.01 0.99 �0.01 0.99

Other family structurea

0.01 1.01 0.01 1.01

Adolescent’s college expectations 0.02*** 1.02 0.02*** 1.02

Age �0.01*** 0.99 �0.01*** 0.99

Female 0.01 1.01 0.01 1.01

Black 0.03** 1.03 0.03** 1.03

Hispanic �0.01 0.99 �0.01 0.99

New school �0.01* 0.99 �0.01 0.99

Family socioeconomic status 0.01*** 1.01 0.01*** 1.01

Friends from outside of school 0.00 1.00 0.00 1.00

Adolescent’s Expectations 3 Autonomy 0.01* 1.01

Quality 3 Supervision 0.02* 1.02

Note: Interaction terms in Model 2 were significant at p , .05 when entered into Model 1, alone.aReference category is having two biological parents.

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

Parenting and Adolescents’ Friendship Networks 1257

Page 12: Parenting Practices and Adolescents’ Friendship Networks

have higher college expectations. Of course,school and neighborhood choice are largely afunction of socioeconomic status. Yet our findingpersists even after controlling for the effects ofsocioeconomic status, regardless of whether weuse measures of occupational prestige, education,or income of parents (income results are not re-ported because of the high percent of missingdata). Future work should continue to explorethe extent to which neighborhood choice, on thebasis of the relative quality of its schools, is a valu-able predictor of friendship network characteris-tics and subsequent child outcomes.

Unexpectedly, we find that when parentsreport choosing their neighborhood because ithas lower crime than other neighborhoods, anadolescent has similar friendship network charac-teristics to those of an adolescent whose parentsdo not report choosing their neighborhood forthis reason. Even in supplementary analyses thatisolate this indicator of neighborhood choice, theresult persists. A likely explanation is that parentswho report choosing a neighborhood because oflow crime may be choosing from among lessdesirable neighborhoods and parents who reportselecting a neighborhood because of its betterschools may be choosing from among moredesirable neighborhoods.

There is more modest support for our hypoth-eses about parental supervision, availability,and provision of autonomy. Familiarity witha child’s friends and the parents of the friends ispositively associated with friends’ grade pointaverages and college expectations. Parental pres-ence at home and adolescent’s autonomy onlyappear relevant to predicting friends’ characteris-tics when they interact with other parenting prac-tices. Because moderate levels of autonomy maybe ideal, we also disaggregated our autonomymeasure into dummy variables with moderateautonomy as the reference category. Doing sodid not change our findings. Nor did we findcurvilinear effects of autonomy in predictingfriendship network characteristics.

Finally, we find some evidence of expectedinteraction effects. The effects of parenting prac-tices on friendship formations appear to dependon the characteristics of the adolescent children.For example, adolescents are more likely to selectfriends like themselves in terms of being in fightsand having college expectations, when they aregranted more autonomy. Also, parenting practi-ces sometimes seem to enhance or detract fromthe effects of other practices. In fact, in every

analysis, we find evidence that parent-child rela-tionship quality and parental supervision interactsuch that adolescents are least likely to havedelinquent friends and most likely to have proso-cial friends when both parent-child relationshipquality and parental supervision are high. Futureresearch should continue to explore these interac-tion effects.

There are limitations to our study. We rely oncross-sectional data and cannot adequately iden-tify the extent to which parents seek to affectan adolescent’s choice of friends. The cross-sectional data do not allow us to consider thecumulative or lagged effects of parenting priorto adolescence. For this reason, our results mayunderestimate the parent-friend link. The use ofcross-sectional data is also problematic becausethere is little evidence of the direction of causalityin the association between parenting practicesand friendship network characteristics. Thecomplexity of this process may provide a partialexplanation for the equivocal findings regardingaspects of parental monitoring and supervision,in particular. That is, in some instances, associa-tions found in our data may be the result of re-sponses to a child’s friendships. In other cases,parents may be more proactive in directing a childtoward suitable friends.

Our study is also limited in its inability to cap-ture the extent to which parents attempt to moldtheir adolescent’s friendship network. For exam-ple, we are unable to distinguish between parentswho desire to live in different neighborhoods forthe benefit of their child but are unable to do sobecause of economic restrictions, job responsibil-ities, family concerns, or other reasons. Continu-ing to evaluate the role of neighborhoodcharacteristics in shaping friendships and perhapsmoderating the effect of parenting practices onfriendship networks is an important focus forfuture research. Future work should analyzehow significant others besides parents (e.g., otherfamily members, neighbors, adult friends) influ-ence a child’s friendship networks.

Nevertheless, this study contributes to ourunderstanding of the extent to which the effectsof parents and friends on adolescents are linkedand how parents may influence child outcomes.When combined with the evidence that friend-ship network characteristics predict delinquentbehavior (Giordano, 2003), our results providean additional challenge to Harris’s (1998) claimsthat parents are relatively unimportant in affect-ing adolescent development.

1258 Journal of Marriage and Family

Page 13: Parenting Practices and Adolescents’ Friendship Networks

Parents appear to be most able to steer a childtoward friends who possess more prosocial andfewer delinquent characteristics when they aresuccessful in maintaining positive parent-childrelationships characterized by involvement,attachment, understanding, and low levels ofconflict. Parents also seem to be influential whenthey can select neighborhoods on the basis of thelocation of better schools and get to knowa child’s friends and the parents of the friends.Future work should seek to understand why asso-ciations exist between parenting practices andfriendship network characteristics, the extent ofparents’ influence on a child’s choice of friends,and other means by which parents may affectfriendship formation.

NOTE

This research uses data from Add Health, a program projectdesigned by J. Richard Udry, Peter S. Bearman, and KathleenMullan Harris and funded by a grant P01-HD31921 from theNational Institute of Child Health and Human Development,with cooperative funding from 17 other agencies. Specialacknowledgment is due Ronald R. Rindfuss and BarbaraEntwisle for assistance in the original design. Persons inter-ested in obtaining data files from Add Health should contactAdd Health, Carolina Population Center, 123 W. FranklinStreet, Chapel Hill, NC 27516-2524 ([email protected]).

REFERENCES

Acock, A. C. (2005). Working with missing values.

Journal of Marriage and Family, 67, 1012 – 1028.

Brody, G. H., Ge, X. J., Conger, R., Gibbons, F. X.,

Murry, V. M., Gerrard, M., & Simons, R. L.

(2001). The influence of neighborhood dis-

advantage, collective socialization, and parenting

on African American children’s affiliation with

deviant peers. Child Development, 72, 1231 –

1246.

Brown, B. B., Mounts, N., Lamborn, S. D., &

Steinberg, L. (1993). Parenting practices and group

affiliation in adolescence. Child Development, 64,

467 – 482.

Buchmann, C., & Dalton, B. (2002). Interpersonal

influences and educational aspirations in 12 coun-

tries: The importance of institutional context.

Sociology of Education, 75, 99 – 122.

Chantala, K., & Tabor, J. (1999). Strategies to per-form a design-based analysis using the Add Healthdata: Research design. National LongitudinalStudy of Adolescent Health. Retrieved February

21, 2005, from http://www.cpc.unc.edu/project/

addhealth/strategies.html

Dishion, T. J., Capaldi, D., Spracklen, K. M., & Li,

F. (1995). Peer ecology of male adolescent drug

use. Development and Psychopathology, 7, 803 –

824.

Gecas, V., & Seff, M. A. (1990). Families and ado-

lescents: A review of the 1980s. Journal of Mar-riage and the Family, 53, 700 – 706.

Giordano, P. C. (2003). Relationships in adolescence.

Annual Review of Sociology, 29, 257 – 281.

Harris, J. R. (1998). The nurture assumption. New

York: Free Press.

Hartup, W. W., & Laursen, B. (1999). Knowing the

sounds: Friendship as a developmental context. In

W. A. Collins & B. Laursen (Eds.), Relationshipsas developmental contexts (pp. 63 – 84). Mahwah,

NJ: Erlbaum.

Haynie, D. L. (2001). Delinquent peers revisited:

Does network structure matter? American Journalof Sociology, 106, 1013 – 1057.

Haynie, D. L., & Osgood, D. W. (2005). Reconsider-

ing peers and delinquency: How do peers matter?

Social Forces, 84, 1109 – 1130.

Hirschi, T. (1969). Causes of delinquency. Berkeley:

University of California Press.

Kandel, D. B. (1996). The parental and peer contexts

of adolescent deviance: An algebra of interper-

sonal influence. Journal of Drug Issues, 26, 289 –

315.

Knoester, C. (2003). Implications of childhood exter-

nalizing problems for young adults. Journal ofMarriage and Family, 65, 1073 – 1080.

Moody, J. (2001). Race, school integration, and

friendship segregation in America. AmericanJournal of Sociology, 107, 679 – 716.

Nye, F. I. (1958). Family relationships and delin-quent behavior. New York: Wiley.

Parke, R., & Buriel, R. (1998). Socialization in

the family: Ethnic and ecological perspectives. In

N. Eisenberg & W. Damon (Eds.), Handbookof child psychology (pp. 463 – 552). New York:

Wiley.

Parke, R. D., & Bhavnagri, N. P. (1989). Parents

as managers of children’s peer relationships. In

D. Belle (Ed.), Children’s social networks andsocial supports (pp. 241 – 259). New York:

Wiley.

Patterson, G. R., Reid, J. B., & Dishion, T. J. (1992).

Antisocial boys: A social interactional approach.

Eugene, OR: Castalia.

Simons, R. L., Simons, L. G., & Wallace, L. E.

(2004). Families, delinquency, and crime. Los

Angeles: Roxbury.

Sroufe, L. A., Egeland, B., & Carlson, E. A. (1999).

One social world: The integrated development

Parenting and Adolescents’ Friendship Networks 1259

Page 14: Parenting Practices and Adolescents’ Friendship Networks

of parent-child and peer relationships. In W. A.

Collins & B. Laursen (Eds.), Relationships asdevelopmental contexts (pp. 63 – 84). Mahwah,

NJ: Erlbaum.

Thornburg, H. D. (1982). Development in adoles-cence. Monterey, CA: Brooks/Cole.

Tourangeau, R., & Shin, H. (1999). Grand sampleweight. National Longitudinal Study of AdolescentHealth. Retrieved April 26, 2006, from http://

www.cpc.unc.edu/project/addhealth/data/using/

guides

Warr, M. (1993). Parents, peers, and delinquency.

Social Forces, 72, 247 – 264.

Warr, M. (2002). Companions in crime: The socialaspects of criminal conduct. Cambridge, UK:

Cambridge University Press.

Warr, M. (2005). Making delinquent friends: Adult

supervision and children’s affiliations. Criminol-ogy, 43, 77 – 106.

Whitbeck, L. B., Simons, R. L., Conger, R. D., &

Lorenz, F. O. (1989). Value socialization and

peer group affiliation among early adolescents.

Journal of Early Adolescence, 9, 436 – 453.

Wojtkiewicz, R., & Donato, K. (1995). Hispanic edu-

cational attainment: The effects of family back-

ground and nativity. Social Forces, 74, 559 – 574.

1260 Journal of Marriage and Family