P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

C 2003) Journal of Youth and Adolescence, Vol. 32, No. 2, April 2003, pp. 115–128 (°

The Importance of Family and School Domains in Adolescent Deviance: African American and Caucasian Youth Alexander T. Vazsonyi1 and Lloyd E. Pickering2 Received October 8, 2001; revised December 20, 2001; accepted March 7, 2002

Previous work has documented the similar importance of developmental domains in accounting for adolescent deviance in different racial/ethnic groups (e.g., Vazsonyi A. T., and Flannery, D. J., 1997, J. Early Adolesc. 17(3): 271–293). The current investigation is a replication and extension of this line of work. It examined the importance of the family (closeness, monitoring, and conflict) and school (grades, homework time, educational aspirations, and commitment) domains on a sample of adolescent (mean age = 16.4 years) African American and Caucasian youth (N = 809). The following important findings were made: (a) developmental processes including family and school domain variables and deviance were very similar for African American and Caucasian youth; (b) both developmental domains revealed independent predictive relationships with a number of different measures of adolescent deviance in both groups; and (c) the 2 domains uniquely accounted for 25% and 37% of the variance explained respectively in African American and Caucasian adolescent total deviance. KEY WORDS: family processes; parenting; academic competence; delinquency; race/ethnicity.

INTRODUCTION

Sickmund, 1999). At the same time, “nonofficial” data such as the National Youth Survey have indicated few important differences in levels of deviance between various ethnic groups in the United States (e.g., Elliott and Ageton, 1980). Explanations reconciling these apparent differences have been varied. Some researchers have suggested that these observed differences in official data may be due to ecological differences rather than due to racial group membership, so that being a member of 1 racial group versus another has little or no direct impact on subsequent delinquent behavior. In other words, whether an adolescent engages in norm-violating conduct and enters the criminal justice system is thought to be largely a function of where a youth resides and how economically privileged he/she may be, for instance, in an urban versus a rural area or in a poor versus a middle-class neighborhood (Hawkins et al., 1998; Peeples and Loeber, 1994). Others have suggested a racial disparity exists because different ethnic/racial groups may experience unique socialization, and perhaps unique etiology, which accounts for subsequent observed differences (e.g., McGoldrick, 1993). Finally, still others have implied based on the basis of local studies that there may exist real differences

The Problem Approximately 30% of all cases in the criminal justice system involve African American youth, even though the African American juvenile population constitutes 15% of all youth in the United States. This means that African American youth are much more likely than both Caucasian adolescents or youth from other ethnic/racial groups to be involved in both person and property offenses (Snyder and 1 Human Development and Family Studies, College of Human Sciences,

Auburn University, Auburn, Alabama. Received PhD from the University of Arizona. Research interests include the etiology of adolescent problem behavior, deviance, delinquency, criminological theory, and the comparative approach in the study of human development and behavior. To whom correspondence should be addressed at Department of Human Development and Family Studies, Auburn University, 284 Spidle Hall, Auburn, Alabama 36849; e-mail: [email protected]. 2 Doctoral candidate, Department of Human Development and Family Studies, College of Human Sciences, Auburn University, Auburn, Alabama. Research interests include adolescent school-to-work transition, religiosity, deviant behaviors, leisure activities, and parent– adolescent relationships.

115 C 2003 Plenum Publishing Corporation 0047-2891/03/0400-0115/0 °

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

116 in how delinquents are handled by the criminal justice system (e.g., Snyder and Sickmund, 1999). The level of violence in the United States far exceeds that found in other industrialized countries (Snyder and Sickmund, 1999), despite recent decreases in overall rates of juvenile crimes in the United States. For example, between 1990 and 1995, the number of annual homicides per 100,000 children who were under the age of 15 was 5 times greater in the United States than that in a combination of 15 Western European countries. Similarly, the rate of child and adolescent homicides involving a firearm was 16 times greater during a 1-year period in the United States than that in these same 15 European countries combined (Krug et al., 1998). These data suggest that, as a society, especially in comparison to other industrialized countries, we continue to face severe problems due to youth violence and delinquency. In fact, Hamburg (1998) has suggested that youth violence in the United States is a public health concern and as such, requires the application of a public health model to its “treatment.” This includes 3 main steps: (1) tracking of trends in youth violence, (2) identification of risk factors and correlates, and (3) development of and rigorous evaluation of interventions. In this study, we were interested in further addressing Hamburg’s second point, namely the identification of risk factors and correlates in both Caucasian and African American youth. This is an important and necessary step in trying to address the problem of youth violence and deviance as well as in attempting to develop subsequent prevention and intervention efforts. The current investigation is a partial replication and extension of previous work on the importance of the family and school domains in adolescent deviance in different racial/ethnic groups. On a sample of early adolescents, Vazsonyi and Flannery (1997) previously established that (1) family and school domains accounted for about 40% of the total variance explained in Caucasian and Hispanic deviance; (2) both the family and school domains independently accounted for variability in deviance; (3) while the family domain appeared to be more salient for Caucasian adolescents in explaining deviance (16%), the school domain was more salient for Hispanic deviant conduct (17%); and (4) despite mean level differences in amount of involvement in deviant behavior between Caucasian and Hispanic youth, developmental processes, namely the relationship between the predictor variables and the outcome, were highly similar for both groups. The current investigation extends this previous work in that it also tests the relative salience of the family and school domains in various forms of adolescent deviance, from more trivial norm-violating conduct (e.g., school misconduct) to more serious interpersonal acts of

Style file version July 26, 1999

Vazsonyi and Pickering violence (e.g., assault). Also, while the previous study compared Caucasian and Hispanic early adolescents residing in the American Southwest, this study focuses on a population sample of African American and Caucasian adolescents residing in the American Southeast. In the following sections, a brief overview of the literature on the importance of both the family and school contexts as well as potential ethnic/racial similarities and differences follows.

Developmental Contexts: The Family and School In a comprehensive review of the literature on the correlates and predictors of juvenile crime and delinquency based largely on longitudinal studies in the United States and abroad, Lipsey and Derzon (1998) identified both the family and the school domains as the most salient predictive domains of serious and violent offending. Similarly, Hawkins et al. (1998) concluded that, in addition to individual characteristics, previous problem behaviors, aggression, and antisocial tendencies, family and school factors were the most predictive of serious and violent offending. As the authors point out themselves, these latter 2 domains also contain the factors that have the greatest potential to be malleable in prevention and intervention efforts. Therefore, it follows that the importance of both the family and school domains require further study. Perhaps one of the most influential criminological theories attempting to explain juvenile delinquency is social control theory (Hirschi, 1969). It posits that youth who are weakly attached to society’s institutions such as the family or school are more likely to deviate from norms and to engage in norm-violating conduct. In effect, Hirschi suggests that, because these youth never learned to conform to rules and norms during their early socialization experiences, they are free to deviate. Recent theoretical and empirical work has continued to identify the necessity of studying adolescent development and negative behavioral outcomes from a contextual perspective, one that integrates not just a single, but multiple contexts of individual development (Bronfenbrenner, 1979; Steinberg and Avenevoli, 1998). There is general consensus that the family is perhaps the most important factor in predicting and understanding juvenile delinquency and even subsequent adult criminality. A large number of studies have focused on the impact of parenting practices or child-rearing styles on the development of delinquency (e.g., Dishion et al., 1991; Gorman-Smith et al., 2000; Loeber and Dishion, 1983; McCord, 1979; Patterson, 1982; Patterson and Dishion, 1985; Stouthamer-Loeber and Loeber, 1988). These have

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

The Importance of Family and School Domains in Adolescent Deviance found that a lack of emotional closeness, involvement, or support, in combination with inconsistent or lax discipline and supervision or monitoring are the most predictive of delinquency and later criminality. Recent work done on officially delinquent youth (Scholte, 1999) also found that parenting dimensions during the adolescent years may contribute additionally to whether delinquent behaviors of youth at the age of 15 progress towards more serious criminal behavior during late adolescence or not (see also Moffitt, 1993). Similarly, a number of studies have documented that greater attachment to school as evidenced by greater educational commitment (Herrenkohl et al., 2000; Hirschi, 1969), greater educational aspirations (Elliott and Voss, 1974; Vazsonyi and Flannery, 1997), spending more time on homework (Cernkovich and Giordano, 1992), getting better grades (Herrenkohl et al., 2000; Hirschi, 1969; Vazsonyi and Flannery, 1997), and simply greater school involvement (Cernkovich and Giordano, 1992; e.g., being a member of a school organization or club) promotes socially conforming behaviors. This is consistent with the recognition that the school operates as a secondary socializing agent that reinforces social norms, values, and mores (Durkheim, 1956; Hirschi, 1969).

Race/Ethnicity and Deviance Similar to work done on parenting practices and styles, some studies have also examined the importance of family processes (for a description of this term, see Vazsonyi and Flannery, 1997) on deviance and delinquency. Cernkovich and Giordano (1987) suggested that, similar to a gender gap in delinquency research, there also exists a racial gap, namely a dearth of studies that examine similarities and differences in the correlates and predictors of delinquency for different racial/ethnic groups. Consequently, they compared 7 types of family processes (caring and trust, supervision, intimate and instrumental communication, identity support, peer approval, and conflict) on Caucasian and African American youth. They found differences on the importance of family process dimensions in predicting adolescent deviant behavior by race/ethnicity. For example, family processes accounted for more variance in Caucasian youth than in African American youth (18% vs. 11%). Furthermore, while the communication variables were highly predictive of delinquency for Caucasian males, the discipline dimension was more predictive for African American males. This latter finding was also true for African American females, while conflict best predicted delinquency in Caucasian females. These findings were partially replicated in Vazsonyi and

117

Flannery (1997), where the unique contribution of family process variables accounted for 16% of total variance explained in Caucasian early adolescents. At the same time, the finding that certain family variables were more predictive of deviance than others suggested differences in developmental processes between African American and Caucasian youth (cf. Vazsonyi and Flannery, 1997). In a similar subsequent effort, Smith and Krohn (1995) compared the role of family processes (parent– child attachment, parent–child involvement, and parental control) in male delinquency across 3 different racial/ ethnic groups, namely African Americans, Caucasians, and Hispanics. They found that family variables accounted for about 10% in delinquency in African American and Caucasian males, while they accounted for about 20% for Hispanics. Both findings were not consistent with the two other studies reviewed here. In addition, the results in this study also suggested differences in developmental processes, similar to Cernkovich and Giordano’s findings. In other words, different family process measures predicted deviance differently by race/ethnicity. Cernkovich and Giordano (1992) also examined the importance of “school bonding” on adolescent delinquency across different ethnic/racial groups. They found that school domain variables accounted for 12 and 15% of the total amount of variance explained in adolescent delinquency for African American males and females, respectively. For Caucasian males and females, the school bonding variables accounted for 13 and 16%, respectively. This latter finding was not consistent with one other effort which found that school domain variables accounted for less than 10% in early adolescent deviance. It is also worth noting that the final predictive model in Cernkovich and Giordano’s study included risk of arrest, parental communication, socioeconomic status, and age as predictor variables. Given these at times inconsistent and mixed findings regarding the importance of family processes and school context variables in understanding and predicting adolescent deviance across different ethnic/racial groups, additional work needs to be completed in this area. Furthermore, while a number of investigations have been completed in urban or metropolitan areas or on high-risk populations (i.e., males), other studies need to investigate these questions in samples which include both males and females and samples drawn from “normal” or average populations.

Research Goals The current investigation further examined the unique importance of both the family and school domains in

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

118

Vazsonyi and Pickering

adolescent deviance. More specifically, the current investigation extended and replicated previous work by (1) examining the importance of family processes (closeness, monitoring, and conflict) in African American and Caucasian youth; (2) examining the importance of school variables in African American and Caucasian youth; (3) testing a predictive model of adolescent delinquency incorporating variables from both the family and school domains and testing the unique contributions of each domain as well as shared contributions while controlling for demographic variables; (4) testing this predictive model for different forms of deviance ranging from less serious to more serious deviant behaviors; and (5) comparing developmental processes, namely the interrelationship between predictor variables and deviance measures between African American and Caucasian youth. For this purpose, instead of employing pairwise comparisons of regression weights and then evaluating similarity as in Cernkovich and Giordano or Smith and Krohn, we employed a model-free LISREL approach (see e.g., Rowe et al., 1994; Vazsonyi and Flannery, 1997). This approach is a much more thorough and efficient evaluation of similarities/differences of 2 or more groups .

the school population, 95% of the surveyed sample) high school students. Only African American and Caucasian youth were included in the current analyses, resulting in a final study sample of n = 809 (mean age = 16.4, SD = 1.2) students. Four hundred and three participants were female (49.9%, mean age = 16.3, SD = 1.2) and 404 were male (50.1%, mean age = 16.4, SD = 1.2). The sample consisted of n = 627 Caucasian students (73.7%, mean age = 16.3, SD = 1.1) and n = 182 African American students (22.5%, mean age = 16.5, SD = 1.3). The majority of sample participants were from 2 biological parent homes (65.3%), while 16.7% were living with 1 biological parent only and 11.9% were in a stepfamily arrangement. The families of these participants were mostly middle and upper-middle class and their parents were well-educated (63.4% of fathers and 59.3% of mothers had college and/or graduate degrees). This high level of education was largely due to the fact that data were collected in a university town (see Table I).

METHOD

Age was measured by asking participants to indicate the month and year in which they were born. The 15th day of each respective month was used to calculate subjects’ specific ages. Subjects were asked to indicate their sex on a single item: “What is your gender?” Responses were given as 1 = male and 2 = female. A measure of social class was also included by asking subjects to indicate their family’s total annual income. Responses were given as (a) $20,000 or less; (b) $20,000–35,000; (c) $35,000–60,000; (d) $60,000–100,000; and (e) $100,000 or more. Finally, participants were asked to indicate their family structure by answering the following question: “Which of the following ‘home situations’ applies best to you? ‘I live with my. . . .’” Responses to this item were given as (a) biological parents, (b) biological mother only, (c) biological father only, (d) biological mother and stepfather, (e) biological father and stepmother, (f) biological parent and significant other, and (g) other.

Procedure Data were collected as part of the International Study of Adolescent Development (ISAD; Vazsonyi et al., 2001). A standard data collection protocol was approved by a university Institutional Review Board (IRB) and consisted of a self-report data collection instrument which included instructions on how to complete the survey, a description of the ISAD project, and assurances of anonymity and confidentiality. These instructions were read verbatim in each classroom before the surveys were distributed. Sample Data for this study were collected from adolescents who attended a high school in the southeastern region of the United States. The total student population at the high school was N = 1,134. All students in the school were invited to participate and 920 (81% of the school population) did so. The remainder of the students declined participation or were absent on the day of data collection. Forty-three (4% of the school population, 5% of the surveyed sample) of the completed surveys were excluded from the sample because they were incomplete (less than 50% completed), leaving a sample of n = 877 (77% of

Measures Demographic Variables

Family Domain The family process measure included a total of 13 items in 3 subscales part of the Adolescent Family Process instrument (Vazsonyi et al., 2002); items were rated on 2 different 5-point Likert-type response scales for both mothers and fathers. Responses for the closeness (6 items; sample item: “My mother gives me the right amount of

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

The Importance of Family and School Domains in Adolescent Deviance

119

Table I. Sample Characteristics Total sample Total sample size Age (SD) Sex Male Female Family structure Biological parents Biological mother only Biological father only Biological mother and stepfather Biological father and stepmother Biological parent and significant other Other Social class Less than $20,000 $20,000–35,000 $35,000–60,000 $60,000–100,000 More than $100,000

African Americans

Caucasians

809 16.4 (1.19)

182 (22%) 16.5 (1.32)

627 (78%) 16.3 (1.14)

404 (50%) 403 (50%)

89 (49%) 92 (51%)

315 (50%) 311 (50%)

526 (65%) 118 (15%) 17 (2%) 76 (9%) 20 (3%) 12 (2%) 37 (5%)

78 (43%) 57 (32%) 2 (1%) 22 (12%) 1 (1%) 5 (3%) 15 (8%)

448 (72%) 61 (10%) 15 (2%) 54 (9%) 19 (3%) 7 (1%) 22 (4%)

56 (8%) 93 (12%) 218 (29%) 245 (33%) 137 (18%)

33 (20%) 45 (27%) 51 (31%) 27 (16%) 9 (6%)

23 (4%) 48 (8%) 167 (29%) 218 (37%) 128 (22%)

Statistical test

Significance

t = −1.29 χ 2 = 0.07

p = 0.20 p = 0.79

χ 2 = 79.10

p = 0.00

χ 2 = 118.59

p = 0.00

Note. Discrepancies between totals when summed and total sample size reflect missing data.

affection.”) and monitoring (4 items; sample item: “When I am not home, my mother knows my whereabouts.”) subscale items were given as A = strongly disagree, B = disagree, C = neither disagree nor agree, D = agree, and E = strongly agree, while the conflict (3 items; sample item: “How often do you have disagreements or arguments with your mother?”) subscale items were answered as A = never, B = occasionally, C = sometimes, D = often, and E = very often. All mother and father items were prefaced with instructions which read, “In the next section, we would like to find out more about your relationship with your mother/stepmother or female caretaker and your father/stepfather or male caretaker.” Thus, for example, responses would include self-report ratings of paternal family process even though a participant may have indicated living in a single mother home.3 Reliability coefficients for the entire sample ranged from α = 0.78 to α = 0.85 for maternal scales and from α = 0.86 to α = 0.89 for paternal scales. For Caucasian students, reliabilities ranged from α = 0.77 to α = 0.86 for maternal scales and from α = 0.87 to α = 0.88 for paternal scales, 3 Upon closer examination of this, we found that 29.6% of African Ameri-

can and 8.1% of Caucasian youth responded to paternal family process items even though they indicated residing with a “biological mother only.” Similarly, 0.6% of African American and 2.1% of Caucasian adolescents rated maternal family process items in “biological father only” homes. On the basis of this, and the fact that parents not residing with youth likely also have an important influence in the socialization of these adolescents, we decided to include ratings of nonresidential parents as part of our analyses.

while for African American students, reliabilities ranged from α = 0.80 to α = 0.83 for maternal scales and from α = 0.84 to α = 0.92 for paternal scales. School Domain Four measures were used to assess the school domain. School grades were measured by 1 item which asked, “What are the average grades you usually get in your classes at school?” Responses were as follows: A = mostly A’s, B = mostly A’s & B’s, C = mostly B’s, D = mostly B’s & C’s, E = mostly C’s, F = mostly C’s & D’s, and G = mostly D’s and lower. This item was reverse coded for subsequent analyses. Self-reported homework time was assessed with the question, “What is the average amount of time you spend on a school night doing homework?” and utilized the following response categories: A = 0–30 min, B = 30 min to 1 h, C = 1–2 h, D = 2– 3 h, and E = 3 or more hours. Educational aspirations measured subjects’ desires to continue their schooling by asking, “How long do you plan to go to school?” Response categories included A = I would like to quit school as soon as I can, B = I plan to finish high school, then stop, C = I plan to go to trade (vocational) school when I graduate, D = I plan to get some college (but no degree), E = I plan to get a college degree (e.g., associate’s or bachelor’s), and F = I plan to get an additional degree after college (e.g., master’s or doctorate). Finally, an 11-item scale was used to assess educational commitment (Thornberry et al., 1991). Items, such as “I like school,” “Getting good

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

120 grades is important to me,” and “I try hard at school,” were answered on the following 4-point Likert-type scale: A = strongly disagree, B = disagree, C = agree, D = strongly agree. Reliabilities for this scale were α = 0.79 for the total sample, α = 0.81 for Caucasians, and α = 0.74 for African Americans. Deviance Lifetime deviance was measured by the 55-item Normative Deviance Scale (NDS) (Vazsonyi et al., 2001; Vazsonyi and Killias, 2001; Vazsonyi and Pickering, 2000). The scale was conceptualized to measure deviance in general adolescent populations and to provide epidemiological data consistent with Hamburg’s idea (Hamburg, 1998), and, therefore, examined a broader spectrum of deviant activities than just status and index offenses. This conceptualization of deviance is consistent with results from nationally representative data sets (e.g., the National Youth Survey; Huizinga et al., 1989) which report that over 90% of sampled males and females indicate having committed at least 1 delinquent act at some time in their life. Very few such self-report scales that include multi-item subscales with psychometric properties have been developed (single-item crime measures are more common in criminological work; e.g., 1 item measuring vandalism). The current investigation examined 7 subscales of the NDS (vandalism, alcohol, drugs, school misconduct, general deviance, theft, and assault). Responses for all items in the NDS were given on a 5-point Likert-type scale and identified lifetime frequency of specific deviant behaviors (A = never, B = 1 time, C = 2–3 times, D = 4–6 times, and E = mor e than 6 times). Reliability coefficients on the deviance subscales for the entire sample ranged from α = 0.82 to α = 0.92. For Caucasian students, reliabilities ranged from α = 0.81 to α = 0.92, while for African American students, reliabilities ranged from α = 0.83 to α = 0.92. In addition, all 55 items from the 7 subscales were summed into an overall measure of total deviance. The total deviance scale was highly reliable (α = 0.97 for both Caucasians and African Americans as well as the total sample). PLAN OF ANALYSIS In a first step, initial descriptive statistics were computed for several demographic variables and then compared by ethnic/racial group. Next, one-way ANOVAs with post hoc Scheffe contrasts were utilized to examine differences in deviance, family processes, and school variables by race. Third, in anticipation of predictive analyses, a correlation matrix including family process and school variables as well as the total deviance measure was

Vazsonyi and Pickering computed. To further visually examine the relationship between the family and school domains and deviance, correlations between each individual predictor variable (family and school) and the total deviance measure were plotted by race. Fourth, set hierarchical regression analyses were employed using both family process and school variables as predictors of different types of deviant behavior by race and key background variables as controls. These regression analyses utilized an inversely repeated 2-step procedure. In the first analysis, all family variables were entered in the first step of the regression followed by all school variables in the second step. The second part of this analysis involved repeating this 2-step procedure in reverse order with school variables being entered first and then family variables. This procedure allowed an examination of the unique variance explained by each set of predictors. In addition, variance which was shared between the 2 sets could then be identified. And fifth, developmental processes for both the family and school domains were compared by race in an attempt to establish whether they were similar or different in these groups. RESULTS Initial Analyses In preparation for subsequent regression analyses, potentially key demographic variables were examined for differences by ethnic group. Table I shows the results of these analyses. A t test revealed no significant differences in age (t = −1.29, p = 0.20) between African American (M = 16.5, SD = 1.32) and Caucasian adolescents (M = 16.3, SD = 1.14). In addition, a chi-square test indicated no differences between these groups on observed and expected frequencies for sex (χ 2 = 0.07, p = 0.79). However, significant differences between the groups were observed for both family structure (χ 2 = 79.10, p = 0.00) and social class (χ 2 = 118.59, p = 0.00). One-Way ANOVAs Table II shows results of a series of one-way ANOVAs on deviance, family, and school variables by race. Caucasian youth reported using both alcohol (M = 2.31, SD = 1.21) and drugs (M = 1.86, SD = 1.11) significantly more (F = 9.89, p = 0.00, for alcohol; F = 3.80, p = 0.05, for drugs) than African American youth (M = 2.00, SD = 1.01; M = 1.68, SD = 0.98). However, African American youth (M = 1.69, SD = 0.83) reported significantly higher levels of assaultive behaviors (F = 5.23, p = 0.02) than did Caucasian youth (M = 1.54, SD = 0.75). For the remaining measures of deviance

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

The Importance of Family and School Domains in Adolescent Deviance

121

Table II. Analysis of Variance on Deviance, Family, and School Variables by Race African Americans

Deviance Vandalism Alcohol use Drug use School misconduct General deviance Theft Assault Total deviance Family Mother items Closeness Monitoring Conflict Father items Closeness Monitoring Conflict School Grades Homework time Educational commitment Educational aspirations a Because

Caucasians

na

×

SD

n

×

SD

F

p

180 179 179 177 178 178 177 181

1.63 2.00 1.68 2.08 1.77 1.49 1.69 1.79

0.85 1.01 0.98 1.02 0.87 0.78 0.83 0.83

622 622 622 622 620 620 613 623

1.63 2.31 1.86 2.07 1.80 1.47 1.54 1.81

0.83 1.21 1.11 0.98 0.81 0.76 0.75 0.78

0.01 9.89 3.80 0.00 0.27 0.08 5.23 0.13

0.93 0.00 0.05 0.97 0.60 0.78 0.02 0.72

180 179 179

3.91 3.87 2.81

0.91 1.01 1.01

620 616 614

3.98 3.81 2.76

0.85 0.89 0.98

0.73 0.62 0.39

0.39 0.43 0.53

172 171 172

3.15 3.00 2.55

1.20 1.29 1.16

607 604 597

3.77 3.33 2.50

0.93 1.05 1.04

51.95 11.63 0.25

0.00 0.00 0.62

166 165 166 163

4.23 2.45 2.89 4.74

1.68 1.07 0.49 1.39

602 599 606 602

5.37 2.14 2.81 5.35

1.53 1.05 0.48 0.92

68.69 11.74 3.22 45.24

0.00 0.00 0.07 0.00

of pairwise deletion, sample sizes vary slightly by analysis.

(vandalism, school misconduct, general deviance, theft, and the total deviance sum score), mean level scores did not differ by race. No statistically significant mean level differences were found for any of the maternal family process subscales, but Caucasian youth reported significantly higher levels of both paternal closeness (M = 3.77, SD = 0.93) and monitoring (M = 3.33, SD = 1.05; F = 51.95, p = 0.00, for closeness; F = 11.63, p = 0.00, for monitoring) in comparison to African American adolescents (M = 3.15, SD = 1.20; M = 3.00, SD = 1.29). On school measures, African Americans (M = 2.45, SD = 1.07; M = 2.14, SD = 1.05) spent significantly more time doing homework (F = 11.74, p = 0.00) but reported significantly lower grades (F = 68.69, p = 0.00) than did Caucasians (M = 4.23, SD = 1.68; M = 5.37, SD = 1.53). Caucasian youth (M = 5.35, SD = 0.92) reported significantly higher levels of educational aspiration (F = 45.24, p = 0.00) than did African American youth (M = 4.74, SD = 1.39).

and Caucasian youth, 3 maternal family process variables, 3 paternal family process variables, and 4 school variables were correlated with the total deviance score. These correlations were then plotted to visually depict the relationship between family and school measures and total deviance. The absolute value of correlations ranged from r = 0.09 to r = 0.45 for African American adolescents and r = 0.25 to r = 0.50 for Caucasian adolescents. The resulting plot indicated a strong relationship between family and school measures and total deviance by race (r = 0.96). The strength of this correlation showed that

Correlational Analyses Figure 1 plots the correlations of each family and school predictor variable with the total deviance measure by race. In other words, for both African American

Fig. 1. Correlations of individual predictor variables with total deviance by race.

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

122 the pattern of relationships among these correlates was very similar for these 2 ethnic groups. Table III includes the correlation matrix for all predictor variables (family and school) as well as the total deviance measure. A rank ordering of correlations (absolute value) of predictor variables with total deviance by race showed that educational commitment had the strongest relationship with deviant behavior for both African Americans (r = 0.45) and Caucasians (r = 0.50). Other similarities in this rank-ordered list included grades (ranked third for both, African American (r = 0.39, Caucasian r = 0.44) and paternal conflict (ranked last for both, African American r = 0.09, Caucasian r = 0.25). Other than these 3 similarities, however, this rank-ordered list was different for the 2 races of adolescents. Some were only 1 rank different (e.g., maternal closeness, 5th for African Americans, 4th for Caucasians), while others showed larger discrepancies (e.g., educational aspirations, 2nd for African Americans, 5th for Caucasians). For African American adolescents, 2 of the predictor variables were unrelated to the others as well as to deviance. Both maternal and paternal conflict were not significantly correlated with total deviance for African Americans r = 0.13 and r = 0.09, respectively). However, given a larger sample size, it is likely that these would have reached statistical significance. Set Hierarchical Regressions Table IV presents the results of set hierarchical regression analyses involving the use of family and school variables as predictors of adolescent deviant behavior. These analyses were completed separately by racial/ethnic group.4,5 We controlled for sex,6 age, family structure, 4 To

determine whether it was necessary to do these analyses separately by race, regression analyses utilizing both race as a main effect and sets of interaction terms (Race × Family and Race × School) were conducted to test whether race added any explanatory power beyond the independent variables (family and school) already included in the model. First, 10 product terms involving race were calculated, 1 for each independent variable (6 family variables, 4 school variables). Second, a series of set hierarchical regressions were run separately for each deviance subscale and for total deviance. In these analyses, the 4 control variables and the 10 independent variables were entered on the first step, race was entered as a main effect on the second step, and then the set of interaction terms which applied to that particular analysis was entered on the third step. The test of both the main effect and the interaction terms was necessary to thoroughly exhaust the possibility that race may add explanatory power which should be explored through separate analyses. For example, in the analysis involving assault as a dependent variable and the school domain as the independent variable, the main effect of race accounted for only two tenths of a percent of variance, a nonsignificant finding (R 2 = 0.002, p = 0.204). However, this did not mean that race was unimportant in the explanation of assault since the subsequent entry of the 4 Race × School interaction terms (entered

Style file version July 26, 1999

Vazsonyi and Pickering and social class in these analyses (see Appendix A for the amount of variance explained by each demographic variable). Using the previously described procedure, we found that the family and school domains together explained from 18% (alcohol) to 29% (school misconduct) of the variance in the deviance subscales for African Americans and 25% (alcohol/assault) to 32% (school misconduct) for Caucasians. These 2 domains also accounted for 25% of the variance in the total deviance score for African American youth and 37% for Caucasians. When the amount of variance explained by control variables was included in the model (see Appendix A), the predictor variables explained 23% (alcohol) to 43% (vandalism) of the total variance in deviance subscales for African American and 29% (alcohol) to 41% (vandalism) for Caucasian youth. This model, including control variables, accounted for 39% of the variance in the total deviance measure for African American and 45% for Caucasian adolescents. Figure 2 presents the unique variance in total deviance explained by each domain (family and school) in the outside portion of each respective circle, while the amount of variance shared by both the family and school domains is found in the center portion of the diagram. The figure indicates that for African American adolescents the family domain uniquely explained 5% of the variance and the school domain uniquely explained 12% of the variance. In addition, the center portion of the diagram shows that another 8% of the variance in total simultaneously as a set on the third step of the regression after the control and independent variables as well as race had been partialled out) explained an additional 2.5% of the variance (R 2 = 0.025, p = 0.000). This finding indicated that race played an important role in the relationship between deviance and the family and school domains, though not always as a main effect. The fact that several of these tests revealed a significant amount of variance explained by either race or the interaction term sets or both (see Appendix B for a detailed summary of the results of these analyses) was an indication that it was necessary to complete subsequent regression analyses separately by ethnic groups. At the same time, although the statistical significance of the findings indicates a need to do separate analyses by race, the amount of variance explained by race and interaction terms is not meaningful substantively (R 2 range for race and interaction sets combined = 0.011–0.041). It is worth noting that race and interaction terms had the largest explanatory power for alcohol and drug use, and for assault. In general, these results support the LISREL findings in that, although there are slight significant variations by race, these differences are not substantial; the overall pattern of relationships among these variables is very similar for African Americans and Caucasians. 5 Of a total of 14 unique variances reported for African Americans in Table IV, 3 are nonsignificant and 4 are statistical trends. This is a result of lack of power due to low sample size after listwise deletion in the multiple regression analyses. However, because more than half (9) of these relationships are significant and 4 more approach significance, we decided to report these numbers as substantive findings. 6 A nonsignificant Race × Sex interaction indicated that analyses were not necessary by sex.

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

The Importance of Family and School Domains in Adolescent Deviance

123

Table III. Correlations of Individual Predictors and Total Deviance by Race 1 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

Total deviance Maternal closeness Maternal monitoring Maternal conflict Paternal closeness Paternal monitoring Paternal conflict Grades Homework time Educational commitment Educational aspirations

2

3

−0.29 −0.36 −0.38 0.48 −0.46 0.39 0.30 −0.51 −0.12∗∗ −0.35 0.44 0.29 −0.28 0.18 0.38 0.25 −0.15 −0.14 −0.44 0.27 0.23 −0.28 0.17 0.21 −0.50 0.43 0.33 −0.37 0.26 0.24

4

5

6

−0.21∗∗ −0.26 0.34 0.26 0.22∗∗ 0.26 −0.06ns −0.03ns −0.20 0.63 −0.12∗∗ 0.42 0.34 −0.42 −0.03ns −0.22 0.29 0.10∗ −0.04ns 0.16 0.12∗∗ −0.27 0.38 0.20 −0.11∗ 0.15 0.08ns 0.13ns −0.33 −0.09ns

7

8

0.09ns −0.39 −0.10ns 0.26 0.02ns 0.25 0.23∗∗ −0.01ns −0.02ns 0.22∗∗ 0.14ns 0.17∗ 0.03ns −0.22 −0.12∗∗ 0.22 −0.22 0.51 −0.06ns 0.39

9

10

−0.24∗∗ −0.45 0.36 0.16∗ 0.13ns 0.36 −0.05ns −0.15∗ 0.16∗ 0.33 0.11ns 0.29 0.08ns −0.09ns 0.22∗∗ 0.37 0.41 0.44 0.28 0.46

11 −0.44 0.25 0.30 0.05ns 0.22∗∗ 0.20∗ −0.01ns 0.37 0.31 0.55

Note. All correlations are significant at p < 0.001 unless otherwise noted: ns = nonsignificant. ∗ p < 0.05; ∗∗ p < 0.01. Correlations for African American adolescents are found in the top half of the matrix while those for Caucasian adolescents are in the bottom half.

deviance was shared by both the family and school domains. An examination of the figure and the results in Table IV indicates that both the family and school domains explained a similar amount of variance in deviance for Caucasians (averages, 8.6 and 8.5%, respectively) and African Americans (averages, 6.4 and 7.4%, respectively). In addition, the school and family domains seemed to account for a slightly greater amount of variance for Caucasian youth (average, 11.6%) than for African American youth (average, 7.3%). Model-Free LISREL Analyses In a final step, model-free LISREL analyses were completed to compare developmental process by racial group (Rowe et al., 1994; Vazsonyi and Flannery, 1997). For this purpose, we compared entire covariance matrices

from each group that include the antecedents (in this case, family and school variables) and outcomes (in this case, measures of deviance). This approach is superior to a large number of pairwise comparisons for each association. For example, for two (Caucasian and African American) 13 × 13 matrices (6 family measures and 7 deviance scales), each containing 78 unique covariances, 156 pairwise comparisons would have to be computed. Not only is such a “piecemeal” approach of pairwise difference testing extremely tedious (not to mention impossible to comprehend), but it is also likely to increase the risk of Type I error (inferring relationships where there are really none). In short, such an approach would be statistically unsound. As an alternative, the model-free LISREL approach provides an omnibus test of all intercorrelations among predictor and outcome variables involving the comparison of an entire covariance matrix from each group simultaneously.

Fig. 2. Unique and shared effects of family and school domains on adolescent total deviance. The amount of variance explained in total deviance did not differ significantly by race for either the Race × School interaction set (R 2 = 0.00; p = 0.689) or the Race × Family interaction set (R 2 = 0.00; p = 0.728); see footnote 6 for a further explanation.

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

124

Vazsonyi and Pickering Table IV. Set Hierarchical Regressions on Deviance by Race

Analysis 1 Step 1: Familyc Step 2: Schoold Analysis 2 Step 1: School Step 2: Family Total modele Overall model f

Vandalism

Alcohol

Aa

Cb

Drug use

School misconduct

General

Theft

Assault

A

C

A

C

A

C

A

C

A

C

A

C

0.16 0.08∗∗

0.22 0.08

0.13∗∗ 0.05†

0.19 0.06

0.12∗∗ 0.10∗∗

0.20 0.08

0.13∗∗ 0.16

0.19 0.13

0.10∗∗ 0.10∗∗

0.23 0.06

0.15 0.05†

0.19 0.09

0.12∗∗ 0.08∗∗

0.16 0.09

0.18 0.06∗ 0.24 0.43

0.22 0.09 0.31 0.41

0.10∗∗ 0.08∗ 0.18 0.23

0.15 0.10 0.25 0.29

0.16 0.06† 0.22 0.34

0.19 0.09 0.28 0.32

0.23 0.06† 0.29 0.36

0.26 0.06 0.32 0.39

0.15 0.05ns 0.20 0.33

0.19 0.11 0.29 0.38

0.11∗∗ 0.09∗ 0.20 0.29

0.20 0.07 0.27 0.35

0.15 0.05ns 0.21 0.29

0.17 0.08 0.25 0.40

Note. Figures in this table represent R 2 values; all R 2 values significant at p < 0.001 unless otherwise noted; ns = nonsignificant. Americans, n = 148. b Caucasians, n = 558. c All family variables (maternal and paternal closeness, maternal and paternal monitoring, maternal and paternal conflict) were entered together in 1 set on this step. d All school variables (grades, homework time, educational commitment, educational aspirations) were entered together in 1 set on this step e Figures in this row reflect the amount of variance accounted for exclusively by family and school variables, slight differences between sums of Steps 1 and 2 and total model R 2 are due to rounding error. f Age, sex, family structure, and social class were all controlled in these analyses, and the figures in this row reflect the amount of variance accounted for when they are included in the model. † p < 0.10; ∗ p < .05; ∗∗ p < .01.

a African

This test assesses similarities or differences in the relationships between predictor and outcome variables or, in other words, examines developmental process. Findings were evaluated employing standard fit indices that included the chi-square statistic, CFI, RMSEA, and the chi-square/df ratio (Browne and Cudeck, 1993; Loehlin, 1992; Rigdon, 1996). Results of the model-free LISREL analyses indicated strong similarity in developmental process for both the family and school domains by race. Fit statistics revealed that developmental process was similar for African American and Caucasian youth in analyses involving all variables from the family domain and deviance (χ 2 (91) = 254.13, CFI = 0.97, RMSEA = 0.05, χ 2 /df = 2.8) as well as analyses with the variables from the school domain and deviance (χ 2 (66) = 170.93, CFI = 0.98, RMSEA = 0.05, χ 2 /df = 2.6). These results indicated a highly similar pattern of associations by race and excellent fit for the model-free comparisons; this suggested that developmental processes (i.e., patterns of association between antecedents/correlates and outcome variables) involving the family and school domains and deviance were very similar for both African American and Caucasian adolescents. DISCUSSION The current investigation extends our understanding of the importance of family and school contexts in deviance of African American and Caucasian youth in a number of important ways: First, relatively few differ-

ences were found between the 2 groups on measures of deviance. Similarly, on variables from the family context, no differences were found by race on self-reported maternal closeness, monitoring, or conflict; in contrast, Caucasian youth reported higher levels of closeness and monitoring by fathers than did African American youth. On the other hand, measures of the school context indicated consistent differences across 3 of the 4 measures. African American youth reported spending more homework time and receiving lower grades, while Caucasian youth reported greater educational aspirations. Our findings appear counterintuitive because, despite the fact that African American youth spent more time doing homework, they reported lower grades and lower academic aspirations. At the same time, our findings are also consistent with previous work. For example, a number of investigations, have documented that minority youth generally report lower levels of education aspirations (e.g., Kao and Tienda, 1998) and lower grades in comparison to Caucasian youth. In fact, Taylor and colleagues (1994) have suggested that minority youth are less engaged in schoolwork due to a perception of a “job ceiling;” more specifically, in their study, they found that African American adolescents reported lower grades than Caucasian youth. These findings are consistent with what sociologists have termed strain, namely when African American youth lower their academic aspirations due to perceived frustration from an apparent inability to compete in a majority culture (Agnew and Jones, 1988; Agnew et al., 1996). It is worth noting that many comparision studies

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

The Importance of Family and School Domains in Adolescent Deviance have focused on low SES African American youth; yet, based on a largely middle class sample, our findings were very similar. Socioeconomic status may have contributed to our counterintuitive finding, namely that African American youth reported spending more time doing homework visa` -vis Caucasian peers even though they had lower academic aspirations and grades. It is quite possible that middle class African American youth have a stronger work ethic in comparison to lower SES adolescents which would explain our finding. In addition, previous work by Stevenson and colleagues (1990) has documented that minority parents of elementary schools children believed more strongly in homework, and therefore, both child selfreports as well as maternal ratings of homework time were significantly higher in comparision to Caucasian children and mothers (cf., Larson et al., 2001). In predictive analyses, findings were very similar to previous work with some exceptions. Most important, the predictive model accounted for 23 and 36% of the total variability in total deviance for African American and Caucasian adolescents, respectively (39 and 45%, including demographic variables). This further supports the importance of these 2 domains in understanding adolescent deviant behavior in different racial groups. Indeed, since all adolescents spend the vast majority of their time in these 2 contexts, we would expect substantial explanatory power by them (for a recent study, see also Dornbusch et al., 2001, although the importance of race was not examined as race was used as a control variable). Previous research (Vazsonyi and Flannery, 1997) established that family domain variables accounted for 16% of the variance in adolescent deviance. The family context variables in the current investigation accounted for 11% of the variance in total deviance for Caucasian youth, but only for 5% in African American youth. The pattern and direction of these findings are consistent with earlier efforts by Cernkovich and Giordano (1987), for example, although they found slightly larger estimates (18 and 11%).7 Together, these findings suggest that family process variables uniquely explained between 5 and 11% of the variance in both ethnic groups and across all deviance measures. Interestingly, in the study by Smith and Krohn (1995) on males only, the estimates were more similar by race, namely about 10% for both groups. This may be an indication that influences on male deviant behavior, particularly those of a family nature, do not vary by race. More

125

specifically, according to Smith and Krohn, African American and Caucasian males did not differ though there was a substantial difference between both of these groups and Hispanic males. In addition, Wilkinson (1980) found that family processes were equally important among males of various ethnicities, but not among females. Further work should be done to identify possible differences in familial influence on male deviant behavior across various races; in this study, we did not complete analyses by sex. For the school context variables, the total amount of variance explained in deviance was identical, namely 12% for African American and 10% for Caucasian youth. These numbers were very consistent with previous work by Cernkovich and Giordano (1992). Despite some mean level differences on predictor and outcome variables, and despite some moderate differences in the predictive relationships of the variables by race, comparisons of developmental processes suggest great similarity in the patterns of associations between predictors and outcomes for the 2 groups. This study adds to previous work (Rowe et al., 1994; Vazsonyi and Flannery, 1997) which suggests that there are very small differences in the relationships between behavioral antecedents and developmental outcomes by racial and ethnic groups (see also Gorman-Smith et al., 2000). This is not to say that mean level differences between racial groups do not exist, nor does it imply that they are not important. In addition, it does not mean that the search for the causes of such observed differences should cease. Rather, it simply suggests that the mechanisms that effectively socialize children and adolescents to conform to society and to refrain from norm- or law-violating conduct may be very similar for adolescents of different racial and ethnic backgrounds. In other words, socialization processes may operate in a very similar fashion across ethnicity, but despite the similarity in the mechanism, we still need to search for answers regarding differences in the levels of socialization within each domain. For example, why is it that African American youth report systematically lower levels of closeness and monitoring with their fathers in comparison to Caucasian youth? Why do we find consistent differences between the 2 groups on measures of home and school influence? Despite these unanswered questions, this study provides evidence of the importance of both contexts which is consistent with ongoing prevention and intervention efforts (e.g., Sherman et al., 1997). LIMITATIONS

7 It

is unclear whether Cernkovich and Giordano’s regression analyses included any background variables (Cernkovich and Giordano, 1987). If they were included, it is possible that the amount of variance explained by family and school variables would have been reduced.

A number of shortcomings and weaknesses in the current investigation require some discussion. First, it would be premature to draw conclusions on the importance

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

126

Vazsonyi and Pickering by sex. Initial regression analyses testing the impact of a Sex × Race interaction did not suggest that it had a significant relationship with deviance. Indeed, previous work has documented that the associations between predictor and outcome variables did not vary as a function of sex (Rowe et al., 1994). Because the African American sample was relatively small, completing analyses by sex would have further reduced statistical power; in turn, conclusions based on such analyses would have been more tentative. Finally, this study was crosssectional in nature and based on adolescent self-reports. Although this means that our findings cannot be considered evidence of causality and are based on single informants (cf. Farrington, 1988; Hindelang et al., 1979, 1981; Junger-Tas and Marshall, 1999; Moffitt, 1988), selfreport data can capture a wider variety of deviant behaviors not identified by official records (Wilkinson, 1980).

of the family and school domains for each racial group on the basis of this study alone. Unlike some previous studies which have used census-based probability samples (e.g., Cernkovich and Giordano, 1987, 1992) or stratified samples with overrepresentation of high-risk youth (e.g., Agnew and Jones, 1988; Smith and Krohn, 1995), the current investigation is based on a convenience sample that was collected in a small college town located in the southeastern region of the United States (see Wilkinson, 1980, which was also based on a convenience sample). Though large and representative of the local population, the sample included fewer than 200 African American adolescents. Thus, findings (e.g., African American youth reporting more homework time) could be related to unique sample characteristics. Therefore, future efforts need to replicate the current findings in other, diverse samples. Another potential weakness is that the current investigation did not examine the correlates of deviance

APPENDIX A Variance Explained by Demographic Variables by Race Vandalism

Family structure Social class Age Sex

Aa

Cb

0.02

0.01

Alcohol A

Drug use

C

A

0.00 0.01

0.00

School misconduct

C

A

C

0.02∗∗

0.01 0.01 0.04∗ 0.02

0.01 0.00 0.00 0.00 0.00 0.00 0.07∗∗ 0.00 0.02 0.03∗∗∗ 0.04∗ 0.02∗∗ 0.09∗∗∗ 0.10∗∗∗ 0.03 0.00 0.08∗∗ 0.01∗

General

Theft

Assault C

A

0.02∗∗

0.01

C

Total deviance

A

C

A

A

0.02∗∗

0.00

0.00

0.00

0.00 0.01∗ 0.04∗∗∗

0.01 0.01 0.04∗ 0.01∗∗ 0.01 0.00 0.03∗ 0.00 ∗∗ 0.06 0.01 0.02 0.00 0.01 0.00 0.05∗∗ 0.01∗ 0.04∗ 0.07∗∗∗ 0.07∗∗ 0.05∗∗∗ 0.03∗ 0.11∗∗∗ 0.07∗∗ 0.05∗∗∗

0.02∗∗∗ 0.01

C 0.02∗∗

Note. Figures in this table represent R 2 values. Americans, n = 148; b Caucasians, n = 558. ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.

a African

APPENDIX B Variance Explained by Race (Controlling for Demographic and Independent Variables)

Race (main effect)a Race × School interaction term sets Race × Family interaction term sets

Vandalism

Alcohol

Drug use

School misconduct

General

Theft

Assault

Total deviance

0.000 0.012∗

0.024∗∗∗ 0.012∗

0.015∗∗∗ 0.010

0.007∗∗ 0.009∗

0.002 0.005

0.006∗ 0.013∗

0.002 0.025∗∗∗

0.007∗∗ 0.002

0.006

0.005

0.016∗

0.011

0.004

0.006

0.004

0.003

Note. Figures in this table represent R 2 values. direction of the regression coefficient (beta) for each of the statistically significant outcomes in this row indicates that Caucasian youth are more likely to participate in these deviant behaviors than are African-American youth. ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.

a The

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

Style file version July 26, 1999

The Importance of Family and School Domains in Adolescent Deviance REFERENCES Agnew, R., Cullen, F. T., Burton, V. S., Jr., Evans, T. D., and Dunaway, R. G. (1996). A new test of classic strain theory. Just. Quarter. 13: 681–704. Agnew, R., and Jones, D. H. (1988). An examination of inflated educational expectations. Sociol. Q. 29: 315–337. Bronfenbrenner, U. (1979). The Ecology of Human Development: Experiments by Nature and design. Harvard University Press, Cambridge, MA. Browne, M. W., and Cudeck, R. (1993). Alternative ways of assessing model fit. In Bollen, K. A., and Long, J. S. (eds.), Testing Structural Equation Models. Sage, Newbury Park, CA, pp. 136–162. Cernkovich, S. A., and Giordano, P. C. (1987). Family relationships and delinquency. Criminology 25(2): 401–427. Cernkovich, S. A., and Giordano, P. C. (1992). School bonding, race, and delinquency. Criminology 30(2): 261–291. Dishion, T. J., Patterson, G. R., Stoolmiller, M., and Skinner, M. L. (1991). Family, school, and behavioral antecedents to early adolescent involvement with antisocial peers. Dev. Psychol. 27(1): 172– 180. Dornbusch, S. M., Erickson, K. G., Laird, J., and Wong, C. A. (2001). The relation of family and school attachment to adolescent deviance in diverse groups and communities. J. Adolesc. Res. 16(4): 396– 422. Durkheim, E. (1956). Education and Sociology. Free Press, Glencoe, IL. Elliott, D. S., and Ageton, S. S. (1980). Reconciling race and class differences in self-reported and official estimates of delinquency. Am. Sociol. Rev. 45: 98–110. Elliott, D., and Voss, H. (1974). Delinquency and Dropout. Heath, Lexington, MA. Farrington, D. P. (1988). Self-reported and official offending from adolescence to adulthood. In Klein, M. W. (ed.), Cross-National Research in Self-Reported Crime and Delinquency. Kluwer, Dordrecht, The Netherlands, pp. 929–964. Gorman-Smith, D., Tolan, P. H., and Henry, D. B. (2000). A developmental–ecological model of the relation between family functioning to patterns of delinquency. J. Quantit. Criminol. 16(2): 169–198. Hamburg, M. A. (1998). Youth violence is a public health concern. In Elliott, D. S., Hamburg, B. A., and Williams, K. R. (eds.), Violence in American Schools Cambridge University Press, New York, pp. 31– 54. Hawkins, D. F., Laub, J. H., and Lauritsen, J. L. (1998). Race, ethnicity, and serious juvenile offending. In Loeber, R., and Farrington, D. P. (eds.), Serious and Violent Juvenile Offenders: Risk Factors and Successful Interventions. Sage, Thousand Oaks, CA, pp. 30–46. Herrenkohl, T. I., Maguin, E., Hill, K. G., Hawkins, J. D., and Abbott, R. D. (2000). Developmental risk factors for youth violence. J. Adolesc. Health 26(3): 176–186. Hindelang, H. J., Hirschi, T., and Weis, J. G. (1979). Correlates of delinquency: The illusion of discrepancy between self-report and official measures. Am. Sociol. Rev. 44: 995–1014. Hindelang, M. J., Hirschi, T., and Weis, J. G. (1981). Measuring Delinquency. Sage, Beverly Hills, CA. Hirschi, T. (1969). Causes of Delinquency. University of California Press, Berkeley. Huizinga, D., Menard, S., and Elliott, D. (1989). Delinquency and drug use: Temporal and developmental patterns. Just. Q. 6: 419– 455. Junger-Tas, J., and Marshall, I. H. (1999). The self-report methodology in crime research. Crime Just. 25: 291–367. Kao, G., and Tienda, M. (1998). Educational aspirations of minority youth. Amer. J. Educat. Res. 106: 349–384. Krug, E., Mercy, J., Dahlberg, L., and Powell, K. (1998). Firearmand non-firearm-related homicide among children: An international comparison. Homicide Stud. 2(1): 83–95.

127

Larson, R. W., Richards, M. H., Sims, B., and Dworkin, J. (2001). How urban African American young adolescents spend their time: Time budgets for locations, activities, and companionship. Amer. J. Comm. Psychol. 29(4): 565–597. Lipsey, M. W., and Derzon, J. H. (1998). Predictors of violent and serious delinquency in adolescence and early adulthood: A synthesis of longitudinal research. In Loeber, R., and Farrington, D. P. (eds.), Serious and Violent Juvenile Offenders: Risk Factors and Successful Interventions. Sage, Thousand Oaks, CA, pp. 86–105. Loeber, R., and Dishion, T. J. (1983). Early predictors of male delinquency: A review. Psychol. Bull. 94: 68–99. Loeber, R., and Stouthamer-Loeber, M. (1986). Family factors as correlates and predictors of juvenile conduct problems and delinquency. In Tonry, M., and Morris, N. (eds.), Crime and Justice (Vol. 7). University of Chicago Press, Chicago. Loehlin, J. (1992). Latent Variable Models: An Introduction to Factor, Path and Structural Analysis. Erlbaum, Hillsdale, NJ. Matsueda, R. L., and Heimer, K. V. (1987). Race, family structure, and delinquency: A test of differential association and social control theories. Am. Sociol. Rev. 52: 826–840. McCord, J. (1979). Some child-rearing antecedents of criminal behavior in adult men. J. Pers. Soc. Psychol. 37: 1477–1486. McGoldrick, M. (1993). Ethnicity, cultural diversity, and normality. In Walsh, F. (ed.), Normal Family Processes Guilford Press, New York, pp. 331–360. Moffitt, T. E. (1988). Accommodating self-report methods to a lowdelinquency culture: A longitudinal study from New Zealand. In Klein, M. W. (ed.), Cross-National Research in Self-Reported Crime and Delinquency Kluwer Academic, Dordrecht, The Netherlands, pp. 43–66. Moffitt, T. E. (1993). Adolescence limited and life course persistent antisocial behavior: A developmental taxonomy. Psychol. Rev. 100: 674–701. Patterson, G. R. (1982). Coercive Family Process. Castalia, Eugene, OR. Patterson, G. R., and Dishion, T. J. (1985). Contributions of families and peers to delinquency. Criminology 23: 63–79. Peeples, F., and Loeber, R. (1994). Do individual factors and neighborhood context explain ethnic differences in juvenile delinquency? J. Quantit. Criminol. 10: 141–158. Rigdon, E. E. (1996). CFI versus RMSEA: A comparison of two fit indexes for structural equation modeling. Struct. Eq. Model. 3(4): 369–379. Rowe, D. C., Vazsonyi, A. T., and Flannery, D. J. (1994). No more than skin deep: Ethnic and racial similarity in developmental process. Psychol. Rev. 101(3): 396–417. Scholte, E. M. (1999). Factors predicting continued violence into young adulthood. J. Adolesc. 22: 3–20. Sherman, L. W., Gottfredson, D., MacKenzie, D., Eck, J., Reuter, P., and Bushway, S. (1997). Preventing Crime: What Works, What Doesn’t What’s Promising? Office of Justice Programs, U.S. Department of Justice, Washington, DC. Smith, C., and Krohn, M. D. (1995). Delinquency and family life among male adolescents: The role of ethnicity. J. Youth Adolesc. 24(1): 69–93. Snyder, H. N., and Sickmund, M. (1999). Juvenile Offenders and Victims: 1999 National Report. Office of Juvenile Justice and Delinquency Prevention, Washington, DC. Steinberg, L., and Avenevoli, S. (1998). Disengagement from school and problem behavior in adolescence: A developmental–contextual analysis of the influences of family and part-time work. In Jessor, R. (ed.), New Perspectives on Adolescent Risk Behavior Cambridge University Press, New York, pp. 392–424. Stevenson, H. W., Chen, C., and Uttal, D. D. (1990). Beliefs about achievement: A study of Black, White, and Hispanic children. Child Develop. 61(2): 508–523. Stouthamer-Loeber, M., and Loeber, R. (1988). The use of prediction data in understanding delinquency. Behav. Sci. Law 6: 333– 354.

P1: GDX Journal of Youth and Adolescence

PP617-joyo-451053

October 1, 2002

18:11

128 Taylor, R. D., Casten, R., Flickinger, S. M., Roberts, D., and Fulmore, C. D. (1994). Explaining the school performance of African American adolescents. J. Res. Adolesc. 4(1): 21–44. Thornberry, T. P., Lizotte, A. J., Krohn, M. D., Farnworth, M., and Jang, S. J. (1991). Testing interactional theory: An examination of reciprocal causal relationships among family, school, and delinquency. J. Crim. Law Criminol. 82: 3–35. Vazsonyi, A. T., and Flannery, D. J. (1997). Early adolescent delinquent behaviors: Associations with family and school domains. J. Early Adolesc. 17(3): 271–293. Vazsonyi, A. T., Hibbert, J., R., and Snider, J. B. (2002). Exotic Enterprise No More? Adolescent Reports of Family and Parenting Process in Adolescents From Four Countries. Manuscript submitted for publication.

Style file version July 26, 1999

Vazsonyi and Pickering Vazsonyi, A. T., and Killias, M. (2001). Immigration and crime among youth in Switzerland. Crim. Just. Behav. 28(3): 329– 366. Vazsonyi, A. T., and Pickering, L. E. (2000). Family processes and deviance: A comparison of apprentices and non-apprentices. J. Adolesc. Res. 15(3): 368–391. Vazsonyi, A. T., Pickering, L. E., Junger, M., and Hessing, D. (2001). An empirical test of A General Theory of Crime: A four-nation comparative study of self-control and the prediction of deviance. J. Res. Crime Delinq. 38(2): 91–131. Wilkinson, K. (1980). The broken home and delinquent behavior: An alternative interpretation to contradictory findings. In Gottfredson, M., and Hirschi, T. (eds.), Understanding Crime. Sage, Beverly Hills, CA, pp. 21–42.