Abstract
Objectives. We sought to identify neighborhood-level resources associated with lower levels of aggression and delinquency among youths aged 9–15 years at baseline after accounting for risk factors and other types of resources.
Methods. Data were derived from the Project on Human Development in Chicago Neighborhoods, which focused on 2226 ethnically diverse, urban youths, their caregivers, and the 80 neighborhoods in which they resided at baseline.
Results. Living in a neighborhood with a higher concentration of organizations or services serving young people and adults was associated with lower levels of aggression (odds ratio [OR]=0.9; 95% confidence interval [CI]=0.8, 1.0); living in such a neighborhood also moderated family, peer, and mentor resources. For example, the presence of well-behaved peers was associated with lower levels of aggression among youths living in neighborhoods where the concentration of organizations and services was at least 1 standard deviation above the mean; the association was less strong among youths living in neighborhoods with organizations and services 1 standard deviation below the mean or less.
Conclusions. Certain family, peer, and mentoring resources may confer benefits only in the presence of neighborhood resources. Increasing neighborhood resources should be considered in interventions designed to reduce urban youths’ involvement in violence.
Aggressive and delinquent adolescent behavior continues to be a significant public health concern, despite 3 decades of prevention efforts.1 According to estimates from the 2005 Youth Risk Behavior Surveillance System, 28% of girls and 43% of boys in US high schools had engaged in a physical fight during the previous year, 7% of girls and 30% of boys had carried a weapon in the previous 30 days, and 8% of girls and 12% of boys had driven a car after drinking alcohol in the previous 30 days.2
In addition to more-traditional epidemiological studies that have identified risk factors, scientists have focused recently on “protective factors” that are associated with resiliency in the presence of risks. A separate but overlapping body of work has identified factors associated with lower levels of aggression and delinquency, whether or not risks are present; such factors are typically labeled resources (positive factors external to individuals) or assets (positive factors residing within individuals).3 Recent research has shown that poor minority students with greater numbers of assets engage in fewer risk behaviors than do those with no or fewer assets.4,5 Similarly, in a study involving a large suburban sample of mostly White youths, 38% of the variance in violence was explained by assets, primarily positive peer relationships and conflict resolution skills.6
Prevention scientists have primarily focused on individual assets or family, peer, and mentoring resources in efforts to prevent aggression and delinquency; to date, only limited data are available on neighborhood-level preventive interventions. The Moving to Opportunity experiment is one example that shows that neighborhood prevention has promise; results showed that youths who moved from poverty into more-affluent neighborhoods were less likely than control participants to commit violent or property crimes 2 years later. Further follow-ups showed persistent effects among girls, whereas rates of violent crime but not property crime decreased among boys.7,8
The impact of neighborhood conditions on child development is a long-standing area of research concern; in the past decade, development of multilevel techniques has made it possible to statistically estimate relationships between neighborhood-level characteristics and individual outcomes.9,10 A review of this literature identified low socioeconomic status and residential instability as 2 neighborhood structural factors that are consistently associated with delinquency and aggression.11
Recently, researchers have examined social resources as well,9 described here as neighborhood resources. Much of this work has focused on collective aspects of communities, which Mayer and Jencks postulated to be most likely to confer protection against negative social outcomes in poor neighborhoods.12 One collective resource that has been studied is collective efficacy, a construct with roots in Bandura’s concept of perceived self-efficacy.13 Bandura postulated that a community’s strengths and efforts reside at least partially in members’ beliefs that they can solve collective problems by working together.13
In various studies, collective efficacy has been associated with reduced levels of violence,14–16 including lower rates of concealed gun possession among youths in Chicago.16 Similar collective behaviors associated with reduced behavior problems include collective socialization (i.e., assessments of adults’ interactions with other families’ children), shown to predict reduced conduct problems among African American children,17 and neighborhood-level social support, shown to predict fewer behavior problems among children at high risk for abuse and neglect.18
Less is known about the association between availability of services and organizations in a neighborhood and individual aggression or delinquency on the part of its young residents. Although it is plausible that rates of youth violence may be lower in neighborhoods with playgrounds, community newspapers, health services, substance abuse treatment services, after-school programs, and the like than in neighborhoods without such resources (but with similar structural circumstances), this has yet to be shown empirically in a multilevel statistical framework.
Evidence has also emerged that neighborhood-level factors may moderate individual-level predictors of aggression. For example, a longitudinal study of girls showed that violent victimization was associated with twice the odds of subsequent perpetration of violence; however, effect sizes differed substantially by neighborhood. In safer, more affluent neighborhoods, victimized girls were 4 times more likely than nonvictimized girls to subsequently engage in violence, whereas this association was not significant in impoverished, violent neighborhoods.19
Similarly, in another study, an association between early maturation and violent behavior among adolescent girls was significant only in neighborhoods with greater concentrated disadvantage.20 A study of African American caregivers and children showed that the effects of authoritative parenting on decreased levels of delinquency were more pronounced in neighborhoods with higher collective efficacy.21
Although neighborhood structural risk factors for aggression and delinquency among young people have been identified,11 epidemiological investigations of positive features of neighborhoods are limited. We sought to contribute to this work by examining the influence of neighborhood-level resources on aggression and delinquency and assessing whether these neighborhood resources moderate the effects of previously identified individual assets and family, peer, and mentoring resources.
METHODS
We derived the data for this study from the Project on Human Development in Chicago Neighborhoods (PHDCN), a multidisciplinary, multilevel longitudinal study of mental health and the development of antisocial behavior among urban youths.22
Study Design
The PHDCN was a longitudinal cohort study embedded within a larger community study. The neighborhood-level data used in the present study were obtained from a 1995 community survey of randomly selected Chicago residents and from US census data. Data on individual youths and their families enrolled in the PHDCN longitudinal cohort study were collected between 1995 and 2002; 6226 young people and their caregivers were enrolled.
Sampling
Chicago neighborhoods were initially grouped into 343 neighborhood clusters, each comprising approximately 8000 residents. Stratified probability sampling was used to select a sample of 80 neighborhood clusters diverse in terms of race/ethnicity and socioeconomic composition.
In the PHDCN, dwelling units were enumerated within the 80 neighborhoods, and approximately 35 000 randomly sampled households were screened for the presence of children and adolescents of eligible ages. Children and youths who were within 6 months of 7 target cohort ages (0 [in utero through 6 months], 3, 6, 9, 12, 15, and 18 years) were invited to participate. A primary caregiver was enrolled for all participants with the exception of the oldest cohort. The participation rate was 75%.
Baseline assessments (conducted during 1995–1997) and 2 additional follow-up interviews (1999–2002) were conducted at 24-month intervals. The individual-level data used in the present analyses were gathered from participants and caregivers enrolled in the 9-, 12-, and 15-year cohorts. Trained interviewers administered structured interviews in respondents’ homes.
In the 1995 community survey, members of a random sample of 8782 adults residing in Chicago were interviewed in their homes regarding the social and organizational features of their neighborhoods. The response rate was 75%, and the sample, independent of the families taking part in the PHDCN, was representative of adult residents of all 343 Chicago neighborhood clusters.
The sample included in the present analyses consisted of participants who had taken part in the PHDCN at baseline and who had complete data for all covariates and outcomes. From the original sample of 2345 participants in the 9-, 12-, and 15-year cohorts, 48 caregivers did not have data available on baseline outcomes, and 71 participants were missing data on at least 1 covariate. Thus, the final sample included 2226 participants. As a result of missing data, the resource analyses conducted here involved between 11 and 24 further deletions. At the neighborhood level, the original 80 neighborhood clusters provided data for our multilevel analysis. The members of the final sample differed slightly from the participants excluded from the analyses: excluded participants were more likely to be of low family socioeconomic position (defined as a composite of parental income, educational level, and occupational code), more likely to be from larger families, and less likely to be White.
Measures
Two outcomes, delinquency and aggression, were assessed at all 3 interviews. Resources, assets, and covariates were measured at baseline.
Outcomes.
Aggression and delinquency were treated as 2 separate outcomes given previous research showing that they relate differently to key covariates.23 At each interview, age-appropriate items from scales of the Achenbach System of Empirically Based Assessments—the Child Behavior Checklist,24 the Youth Self-Report,25 and the Young Adult Self-Report26 (each with high reliability and validity27)—were used to assess behaviors youths had engaged in within the past 6 months. At baseline, primary caregivers rated youths’ aggressive behaviors (20 items) and delinquent behaviors (13 items) using the Child Behavior Checklist. Youths who scored at or above clinical cutoffs (i.e., in the top decile) based on published norms24 were categorized as highly aggressive or delinquent.
Subsequently, as a means of reducing the burden on respondents, shortened versions of the study measures were created through factor analysis; these shortened scales, which included 13 aggression and 8 delinquency items, were dichotomized at the 90th percentile to correspond to clinical borderline cutoffs. At subsequent interviews, caregivers of participants who were aged 9 years at baseline completed the revised version of the Child Behavior Checklist; at the second interview, participants who were aged 12 or 15 years at baseline completed the shortened version of the Youth Self-Report; and, at the third interview, participants who were aged 15 years at baseline completed the revised version of the Young Adult Self-Report.24–26
Neighborhood-level resources.
Neighborhood-level resources were assessed through participants’ responses to the 1995 community survey, aggregated by neighborhood cluster. Collective efficacy was the sum of the scores obtained on 2 subscales focusing on social cohesion and informal social control.14 Social cohesion assessed trust and collaboration between neighbors (e.g., “people in this neighborhood can be trusted” and “people are willing to help”) on a 5-point Likert scale of agreement. Informal social control assessed the degree to which neighbors could be counted on to intervene if (for example) “fights broke out in front of their homes” or “children were skipping school and hanging out on street corners.”
The 10 items of these scales were summed and averaged. Most of the participants responded to all of the items; however, an unconditional 3-level hierarchical linear item response model28 was used to calculate person-specific standard errors so that respondents with missing data on particular items could be included in the analyses. The scale was constructed in the form of a continuous measure, individually standardized to a mean of 0 and a standard deviation of 1.
Organizations and services was a 14-item scale focusing on the presence of local organizations, resources, and programs where 8 focused on general items (parks and playgrounds, community newsletters, neighborhood watch programs, family health services, block group and tenant associations, substance abuse treatment centers, family planning clinics, and mental health centers) and 6 focused specifically on youth services (youth centers, recreational programs, after-school programs, mentoring and counseling programs, mental health services, crisis intervention services). Again, the scale was constructed as a continuous measure and individually standardized to a mean of 0 and a standard deviation of 1. The procedures just described were used to adjust for missing responses.
Additional neighborhood asset scales included intergenerational closure, representing how well neighbors knew the parents of their children’s friends; reciprocal exchange, assessing favors and advice exchanged29; and a social network index of the number of friends and family members living nearby.
Individual-level assets and resources.
Guided by the Search Institute’s framework of developmental assets,4 we constructed continuous individual asset scales and family, peer, and mentoring resource scales through an exploratory factor analysis of half of the study sample. We then refined these scales through deleted-item reliability analyses. Scales with acceptable internal consistency were reassessed through confirmatory factor analyses conducted with the other half of the sample. Scales were individually standardized with a mean of 0 and a standard deviation of 1.
We included items from the Provision of Social Relations,30 a social support measure. Three scales emerged from this measure: family support (6 items; e.g., “I know my family will always stand by me”), friend support (8 items; e.g., “I feel very close to some of my friends”), and nonparental mentors (4 items; e.g., “I have a teacher or coach who I can rely on and talk to”). Ratings were made on a 3-point scale (very, somewhat, or not true), and means were calculated. The prosocial peers scale (9 items) focused on the number of peers attending school regularly and obeying rules; again, ratings were made on a 3-point scale (none, some, or all). Individual assets included hours spent in activities, a sum of the time youths spent each week in school-based and after-school programs, and perceptions about the harmfulness of substances (13 items),31,32 which included 4 possible responses ranging from no risk to great risk.
Covariates.
The neighborhood-level covariate, concentrated poverty, was derived from 1990 US census data through a principal components analysis of the proportion of residents who were (1) living below the poverty level, (2) unemployed, and (3) receiving public assistance. In terms of individual-level covariates, family socioeconomic position was a composite of parental income, educational level, and occupational code; 2-parent maximum values were used. Age at each interview, race/ethnicity, and gender were included in all models as well. Models estimating associations between individual assets; family, peer, and mentor resources; and outcomes also adjusted for family size and caregivers’ marital status; models that included neighborhood-level resources did not include these 2 covariates.
Statistical Analyses
We used generalized estimating equations with a logit link function (using SAS33) to estimate log odds of high levels of aggression and delinquency associated with availability of resources.34 Unstructured within-participant correlations of responses across 3 interviews were modeled using pairwise log odds ratios (ORs; i.e., by estimating pairwise associations for binary responses between the baseline and first follow-up interviews and between the first and second follow-up interviews). The matrix structure of the longitudinal statistical model allowed incorporation of participants’ responses from any interviews for which data existed (assuming that data were missing at random); thus, if an individual participated at baseline but not at either follow-up, his or her baseline responses were incorporated into outcome estimations.
We initially modeled marginal probabilities of high levels of aggression or delinquency over the 3 interviews as a logistic function of individual-level covariates: associations between the selected asset and a high level of aggression or delinquency were estimated in models that controlled for age, age squared, gender, and family socioeconomic position. Race/ethnicity, caregiver marital status, family size, and gender–resource interaction terms were introduced next.
Associations between the odds of reports of high levels of aggression or delinquency over the 3 interviews and neighborhood-level resources were estimated, 1 resource at a time as a main effect, after we controlled for individual-level covariates and neighborhood-level concentrated poverty. Then we evaluated models with both neighborhood- and individual-level assets and family, peer, and mentoring resources and again modeled separately each resource and resource pair because of collinearity between them. Using cross-level interactions, we assessed moderation of individual-level assets and family, peer, and mentoring resources according to neighborhood resources. All of the independent variables and their cross-level interactions were tested, but only significant associations are described here and included in the tables.
RESULTS
Table 1 ▶ shows that highly aggressive or delinquent youths were significantly (P < .05) more likely than their peers to have witnessed violence. Highly aggressive or delinquent girls had less family support and fewer prosocial peers, and they were less likely than their peers to perceive substance use as harmful. In addition, highly aggressive girls had less support from friends and fewer nonparental mentors than their peers; highly delinquent boys had fewer prosocial peers than their peers. Highly aggressive or delinquent boys were more likely than their peers to be members of families of low socioeconomic status and to live in neighborhoods with less collective efficacy. Surprisingly, baseline prevalence rates of both high aggression (15.7% vs 11.9%) and high delinquency (11.3% vs 9.0%) levels were higher among girls than among boys.
TABLE 1—
Selected Sample Characteristics, by Gender and Level of Problem Behavior at Baseline Interview: Project on Human Development in Chicago Neighborhoods, 1995–2002 (n = 2226)
| Aggression Levela | Delinquency Levela | |||||||
| Girls (n = 1110) | Boys (n = 1116) | Girls (n = 1110) | Boys (n = 1116) | |||||
| Below 90th Percentile | 90th Percentile | Below 90th Percentile | 90th Percentile | Below 90th Percentile | 90th Percentile | Below 90th Percentile | 90th Percentile | |
| Total, no. | 936 | 174 | 983 | 133 | 985 | 125 | 1015 | 101 |
| Individual characteristics | ||||||||
| Race/ethnicity, % (no.) | ||||||||
| Hispanic/Latino | 34.6 (324) | 37.4 (65) | 37.5 (369) | 28.6 (38)b | 36.0 (355) | 27.2 (34) | 37.2 (378) | 28.7 (29) |
| White | 12.8 (120) | 8.6 (15) | 13.7 (135) | 7.5 (10)b | 11.9 (118) | 13.6 (17) | 13.4 (136) | 8.9 (9) |
| Black | 29.1 (272) | 29.9 (52) | 25.03 (246) | 35.3 (47)b | 28.4 (280) | 35.2 (44) | 25.7 (261) | 31.7 (32) |
| Other | 9.7 (91) | 10.9 (19) | 10.9 (107) | 13.5 (18) | 9.9 (98) | 9.6 (12) | 11.13 (113) | 11.9 (12) |
| Age at baseline, y, mean (SD) | 11.9 (2.4) | 12.4 (2.4)b | 11.9 (2.4) | 11.9 (2.4) | 12.02 (2.4) | 12.4 (2.7) | 11.9 (2.4) | 11.8 (2.6) |
| Family socioeconomic position score, mean (SD) | −0.1 (1.4) | −0.3 (1.3) | −0.03 (1.4) | −0.3 (1.3)b | −0.1 (1.4) | −0.2 (1.3) | −0.04 (1.4) | −0.3 (1.2)b |
| Witnessed violence on more than 1 occasion, % (no.) | 60.7 (568) | 75.3 (131)b | 67.04 (659) | 75.9 (101)b | 61.4 (605) | 75.2 (94)b | 67.2 (682) | 77.2 (78)b |
| No. of family members, mean (SD) | 5.3 (2.0) | 5.7 (2.1) | 5.3 (1.9) | 5.4 (2.2) | 5.4 (2.0) | 5.4 (2.2) | 5.3 (1.9) | 5.6 (2.5) |
| Family support rating, mean (SD) | 0.09 (0.9) | −0.5 (1.4)b | 0.03 (0.9) | −0.09 (0.9) | 0.05 (0.9) | −0.5 (1.4)b | 0.04 (0.9) | −0.17 (0.9) |
| Prosocial peers scale score, mean (SD) | 0.09 (1.0) | −0.2 (1.1)b | −0.007 (0.9) | −0.2 (1.1)b | 0.07 (1.01) | −0.2 (1.1)b | −0.02 (0.9) | −0.15 (0.9) |
| No. of hours spent in prosocial activities, mean (SD) | 0.002 (1.04) | −0.04 (1.2) | 0.01 (0.9) | −0.003 (0.9) | −0.01 (1.05) | 0.04 (1.2) | 0.004 (0.9) | 0.09 (1.1) |
| Friend support rating, mean (SD) | 0.1 (0.9) | −0.08 (1.04)b | −0.09 (0.9) | −0.2 (1.1) | 0.1 (0.9) | 0.03 (1.0) | −0.1 (0.9) | −0.2 (1.1) |
| Perception of harmfulness of substances score, mean (SD) | 0.05 (0.9) | −0.2 (1.2)b | −0.02 (0.9) | −0.09 (1.03) | 0.04 (0.9) | −0.2 (1.1)b | −0.009 (0.9) | −0.2 (1.1) |
| Nonparental mentor rating, mean (SD) | 0.02 (0.9) | −0.2 (1.1)b | 0.009 (0.9) | 0.05 (1.04) | −0.001 (0.9) | −0.1 (1.2) | 0.02 (1.0) | −0.009 (1.0) |
| Neighborhood characteristics | ||||||||
| Collective efficacy rating, mean (SD) | −0.02 (0.9) | −0.09 (0.8) | −0.001 (0.9) | −0.2 (0.8)b | −0.03 (0.8) | 0.003 (0.8) | −0.008 (0.9) | −0.2 (0.8)b |
| Organizations and services score, mean (SD) | −0.2 (1.03) | −0.3 (1.05) | −0.2 (1.01) | −0.2 (1.04) | −0.2 (1.03) | −0.2 (1.03) | −0.2 (1.0) | −0.2 (1.04) |
| Concentrated poverty score, mean (SD) | −0.07 (0.7) | 0.05 (0.7) | −0.07 (0.8) | 0.03 (0.8) | −0.06 (0.7) | 0.02 (0.8) | −0.07 (0.8) | 0.05 (0.8) |
Note. All asset and resource scales were treated as continuous and were individually standardized to have a mean of 0 and an SD of 1. For details on measures, see “Methods” section.
aAt baseline, primary caregivers rated youths’ aggressive behaviors (20 items) and delinquent behaviors (13 items) using the Child Behavior Checklist. Youths who scored in the top decile were categorized as highly aggressive or delinquent.
bBetween-group difference statistically significant at P < .05.
Multilevel Models of Aggression
Table 2 ▶ presents associations between individual-level resources measured at baseline and the odds of high levels of aggression over time, after we had controlled for baseline covariates. We present results separately when interactions between gender and resources were significant. Increases of 1 standard deviation in the levels of prosocial peers or nonparental mentors had a protective effect against aggression among boys but not among girls; however, in the case of both boys and girls, 1-standard-deviation increases in levels of support from family and friends were associated with lower odds of aggression. Perceptions regarding the harmfulness of substance use and prosocial activities were not significant for either gender.
TABLE 2—
Associations Between Selected Individual-Level Resources Measured at Baseline and Odds of High Levels of Aggression: Project on Human Development in Chicago Neighborhoods (n = 2226), 1995–2002
| Model 1, OR (95% CI) | Model 2, OR (95% CI) | Model 3, OR (95% CI) | Model 4, OR (95% CI) | Model 5, OR (95% CI) | Model 6, OR (95% CI) | |
| Intercept | 0.3*** (0.2, 0.6) | 0.3*** (0.2, 0.5) | 0.3*** (0.2, 0.6) | 0.3*** (0.2, 0.6) | 0.3*** (0.2, 0.6) | 0.3*** (0.2, 0.6) |
| Gender (reference: male) | 0.7** (0.6, 0.9) | 0.8** (0.6, 0.9) | 0.8** (0.6, 0.9) | 0.8** (0.6, 0.9) | 0.8** (0.6, 0.9) | 0.8** (0.6, 0.9) |
| Support from friends | 0.9** (0.8, 0.9) | |||||
| Male prosocial peersa | 0.7*** (0.7, 0.8) | |||||
| Female prosocial peersa | 0.9 (0.6, 1.3) | |||||
| Male nonparental mentorsa | 0.8** (0.7, 0.9) | |||||
| Female nonparental mentorsa | 0.9 (0.7, 1.4) | |||||
| Support from family | 0.8*** (0.7, 0.8) | |||||
| Perceptions about substance use | 0.9 (0.9, 1.04) | |||||
| Hours spent in prosocial activities | 0.9 (0.8, 1.0) |
Note. OR = odds ratio; CI = confidence interval. At baseline, primary caregivers rated youths’ aggressive behaviors (20 items) and delinquent behaviors (13 items) using the Child Behavior Checklist. Youths who scored in the top decile were categorized as highly aggressive or delinquent. Analyses controlled for gender, race/ethnicity, marital status of primary caregiver, family socioeconomic position, family size, age, and age squared. All assets and resources were continuous measures standardized to a mean of 0 and a standard deviation of 1. The ORs shown are estimates of the odds of highly aggressive behavior associated with a 1 standard-deviation increase in the resources shown after we had controlled for covariates. Generalized estimating equations with a logit link function were used to estimate the log odds of high levels of aggression over the 3 interviews associated with the indicated individual-level resources. For details on measures, see “Methods” section.
aInteraction between asset and female gender significant at P < .05.
**P < .01; ***P < .001.
Table 3 ▶ presents both main effects of neighborhood parameters and the odds of aggression associated with individual resources that exhibited significant interactions with neighborhood resources. We calculated significant interactions between individual-level resources and neighborhood-level resources at prototypical “high” (1 standard deviation above the mean) and “low” (1 standard deviation below the mean) levels of neighborhood-level resources. There was a main effect of living in a neighborhood in which organizations and services were available: this asset was associated with lower odds of aggression, as indicated in model 1. A 1-unit increase in concentration of organizations or services was associated with 0.9 times lower odds of aggression (95% confidence interval [CI] = 0.8, 1.0).
TABLE 3—
Associations Between Selected Neighborhood- and Individual-Level Resources Measured at Baseline and Odds of High Levels of Aggression: Project on Human Development in Chicago Neighborhoods (n = 2226), 1995–2002
| Model 1, OR (95% CI) | Model 2, OR (95% CI) | Model 3, OR (95% CI) | Model 4, OR (95% CI) | Model 5, OR (95% CI) | Model 6, OR (95% CI) | |
| Organizations and services | 0.9* (0.8, 1.0) | |||||
| Interactions between availability of organizations/services and prosocial peersa | ||||||
| Presence of prosocial peers in neighborhoods with services 1 SD below mean | 0.9* (0.8, 0.9) | |||||
| Presence of prosocial peers in neighborhoods with services 1 SD above mean | 0.7* (0.7, 0.8) | |||||
| Collective efficacy | 0.9 (0.8, 1.0) | |||||
| Interactions between collective efficacy and presence of family supporta | ||||||
| Presence of family support in neighborhoods with collective efficacy 1 SD below mean | 0.9 (0.7, 1.0) | |||||
| Presence of family support in neighborhoods with collective efficacy 1 SD above mean | 0.6* (0.5, 0.7) | |||||
| Interactions between collective efficacy and presence of prosocial peersa | ||||||
| Presence of prosocial peers in neighborhoods with collective efficacy 1 SD below mean | 0.9 (0.8, 1.1) | |||||
| Presence of prosocial peers in neighborhoods with collective efficacy 1 SD above mean | 0.7* (0.5, 0.8) | |||||
| Interactions between collective efficacy and presence of nonparental mentorsa | ||||||
| Presence of nonparental mentors in neighborhoods with collective efficacy 1 SD below mean | 1.0 (0.8, 1.2) | |||||
| Presence of nonparental mentors in neighborhoods with collective efficacy 1 SD above mean | 0.7* (0.6, 0.8) | |||||
Note. OR = odds ratio; CI = confidence interval. At baseline, primary caregivers rated youths’ aggressive behaviors (20 items) and delinquent behaviors (13 items) using the Child Behavior Checklist. Youths who scored in the top decile were categorized as highly aggressive or delinquent. In the analyses, we controlled for gender, race/ethnicity, family socioeconomic position, age, age squared, and neighborhood-concentrated poverty. Although we estimated the associations between all of the hypothesized neighborhood-level resources and aggression, as well as all of the possible cross-level interactions, only significant resources and interactions are included here. Generalized estimating equations with a logit link function were used to estimate the log odds of high levels of aggression over the 3 interviews associated with the indicated individual- and neighborhood-level resources; cross-level interaction terms were introduced to test for moderation of individual-level resources by neighborhood-level resources. For details on measures, see “Methods” section.
aORs represent the odds of highly aggressive behavior associated with a 1 SD increase in each individual-level asset within each type of neighborhood (either high or low on each neighborhood-resource scale).
*P < .05.
Availability of organizations and services also reinforced the positive effects of prosocial peers, as indicated by the significant cross-level interaction. Although a 1 standard-deviation increase in the level of prosocial peers was significantly associated with lower odds of aggression among boys in the earlier analyses (Table 2 ▶), this association differed by neighborhood concentration of organizations and services for both boys and girls. In neighborhoods with lower-than-average levels of organizations and services (1 standard deviation below the mean or lower), the presence of prosocial peers was associated with 0.9 times lower odds of aggression (95% CI=0.8, 0.9). However, a 1 standard-deviation increase in the level of prosocial peers was associated with 0.7 times lower odds of high aggression (95% CI=0.7, 0.8) among both boys and girls living in neighborhoods with high levels of organizations and services (1 standard deviation above the mean or higher).
The main effect of collective efficacy on aggression was only marginally significant (OR=0.9; 95% CI=0.8, 1.0), but several cross-level interactions were significant at P<.05. Several resources that were significant in individual-level models were not associated with aggression in neighborhoods with low collective efficacy. However, a 1 standard deviation increase in each of the following resources was significantly associated with lower aggression in neighborhoods with high collective efficacy: family support in model 4 (OR=0.6; 95% CI=0.5, 0.7), presence of prosocial peers in model 5 (OR=0.7; 95% CI=0.5, 0.8), and availability of nonparental mentors in model 6 (OR=0.7; 95% CI=0.6, 0.8).
Multilevel Models of Delinquency
Table 4 ▶ shows the associations between resources measured at baseline and the odds of high levels of delinquency over interviews 1 through 3. In contrast to aggression, a 1 standard-deviation increase in the level of prosocial peers or supportive friends was significantly associated with lower odds of delinquent behaviors among both girls and boys. As with aggression, 1-standard-deviation increases in availability of nonparental mentors and family support were associated with lower odds of delinquency among boys; a 1-standard-deviation increase in availability of nonparental mentors but not family support was associated with lower odds of delinquency among girls. Perceptions of the harmfulness of substance use were associated with lower odds of delinquency, whereas participation in prosocial activities was not.
TABLE 4—
Associations Between Selected Neighborhood- and Individual-Level Resources Measured at Baseline and Odds of High Levels of Delinquency: Project on Human Development in Chicago Neighborhoods (n = 2226), 1995–2002
| Model 1, OR (95% CI) | Model 2, OR (95% CI) | Model 3, OR (95% CI) | Model 4, OR (95% CI) | Model 5, OR (95% CI) | Model 6, OR (95% CI) | Model 7, OR (95% CI) | |
| Intercept | 0.6* (0.3, 0.9) | 0.5** (0.3, 0.8) | 0.5** (0.3, 0.9) | 0.5** (0.3, 0.9) | 0.5* (0.3, 0.9) | 0.5** (0.3, 0.8) | |
| Gender (reference: male) | 1.2* (1.0, 1.4) | 1.2* (1.0, 1.4) | 1.2* (1.0, 1.4) | 1.2** (1.1, 1.4) | 1.2* (1.0, 1.4) | 1.2* (1.0, 1.4) | |
| Support from friends | 0.9* (0.8, 0.9) | ||||||
| Prosocial peers | 0.9** (0.8, 0.9) | ||||||
| Nonparental mentors | 0.9** (0.8, 0.9) | ||||||
| Male family supportb | 0.7*** (0.7, 0.8) | ||||||
| Female family supportb | 0.9 (0.7, 1.1) | ||||||
| Perceptions about substance use | 0.9* (0.8, 0.9) | ||||||
| Hours spent in prosocial activities | 1.1 (0.9, 1.1) | ||||||
| Collective efficacy | 0.9c (0.8, 1.0) | ||||||
| Interaction between presence of prosocial peers and neighborhood collective efficacy 1 SD below mean | 0.9 (0.8, 1.2) | ||||||
| Interaction between presence of prosocial peers and neighborhood collective efficacy 1 SD above mean | 0.7* (0.6, 0.7) |
Note. OR = odds ratio; CI = confidence interval. At baseline, primary caregivers rated youths’ aggressive behaviors (20 items) and delinquent behaviors (13 items) using the Child Behavior Checklist. Youths who scored in the top decile were categorized as highly aggressive or delinquent. In the analyses, we controlled for race/ethnicity, gender, marital status of primary caregiver, family socioeconomic position, family size, age, and age squared. All resources were continuous measures standardized to a mean of 0 and an SD of 1. In the models of individual-level resources without interactions with neighborhood-level resources, ORs represent the odds of highly delinquent behavior associated with a 1 standard-deviation increase in each individual-level asset or resource; in the model with the cross-level interaction, the ORs represent the odds of highly delinquent behavior associated with the presence of prosocial peers for neighborhoods both high in collective efficacy and low in collective efficacy. Generalized estimating equations with a logit link function were used to estimate the log odds of high levels of delinquency over the 3 interviews associated with the indicated individual- and neighborhood-level resources; cross-level interaction terms were introduced to test for moderation of individual-level resources by neighborhood-level resources. For details on measures, see “Methods” section.
aWe controlled for gender, race/ethnicity, family socioeconomic position, age, age squared, and neighborhood-concentrated poverty. Although associations were estimated between all neighborhood-level resources and delinquency and with all possible cross-level interactions, only significant associations are included.
bInteraction between selected asset and female gender were marginally significant at P < .10.
cOR for main effect of collective efficacy, before addition of the interaction with presence of prosocial peers.
* P < .05; **P < .01; ***P < .001.
Unlike the case with aggression, a 1 standard-deviation increase in the level of neighborhood organizations and services was not associated with lower odds of delinquency, either on average or in cross-level interactions with individual resources. One significant cross-level interaction was found with collective efficacy (Table 4 ▶, model 7): in neighborhoods with high collective efficacy, a 1 standard-deviation increase in the level of prosocial peers was associated with 0.7 times lower odds of delinquency (95% CI=0.6, 0.7). This association was not significant among boys and girls living in neighborhoods with low collective efficacy.
DISCUSSION
Our goal was to assess the role of neighborhood-level resources in aggressive and delinquent behaviors among youths and the ways in which these resources interact with other previously identified individual assets and family, peer, and mentoring resources. We found that 2 neighborhood social resources were most pronounced in their effects, especially with regard to moderation of other resources. Organizations and services targeting young people and adults were protective against highly aggressive youth behavior. In addition, the significant interaction between availability of organizations and services and the presence of prosocial peers suggests that these 2 resources in combination were especially protective against aggressive behavior.
Although there was not a statistically significant main effect of collective efficacy on either highly aggressive or delinquent behaviors, there were significant interactions with several individual-level resources, including family support, presence of prosocial peers, and availability of supportive nonparental adults. Such results suggest that neighborhood collective efficacy is important if these other resources are to confer positive effects on young people.
We had hypothesized that neighborhood social networks would also be associated with reduced aggression and delinquency and would moderate other resources. A previous study of PHDCN caregivers showed that, after adjustment for individual-level social support, Hispanic families living in neighborhoods with large social networks engaged in less corporal punishment and physical abuse in disciplining youths than did Hispanic families living in neighborhoods without such networks.35 However, we did not find this neighborhood-level resource to be protective in the present study.
We assessed a pair of additional neighborhood collective behaviors that were not associated with aggression or delinquency: intergenerational closure, a measure of how well neighbors knew the parents of their children’s friends, and reciprocal exchange of goods and services between neighbors. We hypothesized that these behaviors might be important because they have been shown to be positively associated with residential stability and negatively associated with population density,29 2 structural factors associated with youth violence.11
A study involving a nationally representative sample of adults showed that respondents valued activities supporting youths in their community who were not their own children. For example, “encouraging success in school” was endorsed by 90% of the respondents as important; “teaching respect for cultural differences” was endorsed by 77%; and “providing service opportunities for youths” was endorsed by 48%.36 However, the majority of the adults had not actually been involved in efforts in any of these areas. Our evidence that neighborhood-level resources enhanced protective effects of family, peer, and mentoring resources may encourage adults to engage in community-building efforts with young people in their neighborhoods, activities for which strong social norms may already be in place.
Strengths and Limitations
This study involved several limitations. First, our data were derived from a single city, and the results may not be generalizable to nonurban areas. Second, we relied on youths’ self-reports and parent reports, which could have been subject to recall or social desirability bias. Also, the neighborhood data were collected in 1995, and there was no accounting for potential changes in neighborhood resources across the study years. The analytic sample consisted of participants who had complete data on the variables of interest; exclusion of those who were missing data on key exposure variables may have introduced additional biases. Unmeasured confounders may have also biased the results.
A major strength of this study was its longitudinal and multilevel design. We were able to examine independent as well as moderating effects of neighborhood-level resources, a key next step in studies of neighborhood effects on youth behavior. Future research should attempt to replicate these findings in nonurban areas and identify additional neighborhood resources with the goal of developing efficacious and effective preventive interventions.
Conclusions
Our results suggest that previously identified resources such as social support, availability of nonparental mentors, and presence of well-behaved peers may confer advantages to young people only in the presence of neighborhood resources. Thus, in addition to working with individual youths, teachers, families, and nonparental mentors, those involved in violence prevention programs should find ways to enhance neighborhood resources as well.
Acknowledgments
This study was supported the Centers for Disease Control and Prevention (grant R49/CCR118602 and R49/CCR115279). Funding for the Project on Human Development in Chicago Neighborhoods (PHDCN) was provided by the John D. and Catherine T. MacArthur Foundation, the National Institute of Mental Health, and the National Institute of Justice.
We extend our gratitude to David Hemenway, director of the Harvard Youth Violence Prevention Center, and Angela Browne, former associate director of the center, for their support and ideas; to the field staff who tirelessly collected the data in Chicago; to Deborah Azrael for comments on drafts of the article; and to Kathy McGaffigan for assistance with data preparation and guidance on analyses. We also acknowledge the contributions of the scientific directors of the PHDCN: Jeanne Brooks-Gunn, Felton Earls, Stephen Raudenbush, and Robert Sampson. We are especially grateful to the families participating in the PHDCN, who gave us their energy and time to help us better understand their lives and communities.
Note. This article is solely the responsibility of the authors and does not represent the official views of the Centers for Disease Control and Prevention.
Human Participant Protection This study was reviewed and approved by the institutional review board of the Harvard School of Public Health. A parent or guardian provided written consent before each assessment, and each young person assented as well.
Peer Reviewed
Contributors B. E. Molnar originated the study, supervised its implementation, assisted with the analyses, and led the writing. M. Cerda conducted the analyses and contributed to the writing of the article. A. L. Roberts and S. L. Buka participated in the planning and writing of the article. All of the authors helped to conceptualize ideas, interpret findings, and review drafts of the article.
References
- 1.Catalano RF, Berglund ML, Ryan JAM, Lonczak HS, Hawkins JD. Positive youth development in the United States: research findings on evaluations of positive youth development programs. Ann Am Acad Political Soc Sci. 2004;591:98–124. [Google Scholar]
- 2.Eaton DK, Kann L, Kinchen S, et al. Youth risk behavior surveillance—United States, 2005. MMWR CDC Surveill Summ. 2006;55:1–108. [PubMed] [Google Scholar]
- 3.Fergus S, Zimmerman MA. Adolescent resilience: a framework for understanding healthy development in the face of risk. Annu Rev Public Health. 2005;26: 399–419. [DOI] [PubMed] [Google Scholar]
- 4.Scales PC. Reducing risks and building developmental assets: essential actions for promoting adolescent health. J Sch Health. 1999;69:113–119. [DOI] [PubMed] [Google Scholar]
- 5.Scales PC, Foster KC, Mannes M, Horst MA, Pinto KC, Rutherford A. School-business partnerships, developmental assets, and positive outcomes among urban high school students: a mixed-methods study. Urban Educ. 2005;40:144–189. [Google Scholar]
- 6.Leffert N, Benson PL, Scales PC, Sharma AR, Drake DR, Blyth DA. Developmental assets: measurement and prediction of risk behaviors among adolescents. Appl Dev Sci. 1998;2:209–230. [Google Scholar]
- 7.Katz LF, Kling JR, Liebman JB. Moving to Opportunity in Boston: early results of a randomized mobility experiment. Q J Econ. 2001;116:607–654. [Google Scholar]
- 8.Kling JR, Ludwig J, Katz LF. Neighborhood effects on crime for female and male youth: evidence from a randomized housing voucher experiment. Q J Econ. 2005;120:87–130. [Google Scholar]
- 9.Rajaratnam JK, Burke JG, O’Campo P. Maternal and child health and neighborhood context: the selection and construction of area-level variables. Health Place. 2006;12:547–556. [DOI] [PubMed] [Google Scholar]
- 10.Sampson RJ, Morenoff JD, Gannon-Rowley T. Assessing “neighborhood effects”: social processes and new directions in research. Annu Rev Sociol. 2002;28: 443–478. [Google Scholar]
- 11.Leventhal T, Brooks-Gunn J. The neighborhoods they live in: the effects of neighborhood residence on child and adolescent outcomes. Psychol Bull. 2000; 126:309–337. [DOI] [PubMed] [Google Scholar]
- 12.Mayer SE, Jencks C. Growing up in poor neighborhoods—how much does it matter? Science. 1989; 243:1441–1445. [DOI] [PubMed] [Google Scholar]
- 13.Bandura A. Exercise of personal and collective efficacy in changing societies. In: Bandura A, ed. Self-Efficacy in Changing Societies. Cambridge, England: Cambridge University Press; 1995:1–45.
- 14.Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science. 1997;277:918–924. [DOI] [PubMed] [Google Scholar]
- 15.Browning CR. The span of collective efficacy: extending social disorganization theory to partner violence. J Marriage Fam. 2002;64:833–850. [Google Scholar]
- 16.Molnar BE, Miller MJ, Azrael D, Buka SL. Neighborhood predictors of concealed firearm carrying among children and adolescents: results from the Project on Human Development in Chicago Neighborhoods. Arch Pediatr Adolesc Med. 2004;158:657–664. [DOI] [PubMed] [Google Scholar]
- 17.Simons LG, Simons RL, Conger RD, Brody GH. Collective socialization and child conduct problems: a multilevel analysis with an African American sample. Youth Soc. 2004;35:267–292. [Google Scholar]
- 18.Runyan DK, Hunter WM, Socolar RRS, et al. Children who prosper in unfavorable environments: the relationship to social capital. Pediatrics. 1998;101:12–18. [DOI] [PubMed] [Google Scholar]
- 19.Molnar BE, Browne A, Cerda M, Buka SL. Violent behavior by girls reporting violent victimization: a prospective study. Arch Pediatr Adolesc Med. 2005;159: 731–739. [DOI] [PubMed] [Google Scholar]
- 20.Obeidallah D, Brennan RT, Brooks-Gunn J, Earls F. Links between pubertal timing and neighborhood contexts: implications for girls’ violent behavior. J Am Acad Child Adolesc Psychiatry. 2004;43:1460–1468. [DOI] [PubMed] [Google Scholar]
- 21.Simons RL, Simons LG, Burt CH, Brody GH, Cutrona C. Collective efficacy, authoritative parenting and delinquency: a longitudinal test of a model integrating community- and family-level processes. Criminology. 2005;43:989–1029. [Google Scholar]
- 22.Earls F, Buka SL. Project on Human Development in Chicago Neighborhoods. Washington, DC: National Institute of Justice; 1997.
- 23.Cheong YF, Raudenbush SW. Measurement and structural models for children’s problem behaviors. Psychol Methods. 2000;5:477–495. [DOI] [PubMed] [Google Scholar]
- 24.Achenbach TM. Manual for the Child Behavior Checklist/4-18 and 1991 Profile. Burlington: University of Vermont; 1991.
- 25.Achenbach TM. Manual for the Youth Self Report and 1991 Profile. Burlington: University of Vermont; 1991.
- 26.Achenbach TM. Young Adult Self Report. Burlington: University of Vermont; 1997.
- 27.Achenbach TM, Rescorla LA. Manual for the ASEBA School-Age Forms & Profiles. Burlington: University of Vermont; 2001.
- 28.Raudenbush SW, Johnson C, Sampson RJ. A multivariate, multilevel Rasch model with application to self-reported criminal behavior. Sociol Methodology. 2003;33:169–211. [Google Scholar]
- 29.Sampson RJ, Morenoff JD, Earls F. Beyond social capital: spatial dynamics of collective efficacy for children. Am Sociol Rev. 1999;64:633–660. [Google Scholar]
- 30.Turner R, Frankel B, Levin D. Social support: conceptualization, measurement, and implications for mental health. In: Greeley JR, ed. Research in Community and Mental Health. Greenwich, CT: JAI Press; 1983: 67–111.
- 31.Johnston LD, O’Malley PM, Bachman JG. National Survey Results on Drug Use From the Monitoring the Future Study, 1975–1994: Volume I, Secondary School Students. Rockville, Md: National Institute on Drug Abuse; 1995. Report 95-4026.
- 32.National Household Survey on Drug Abuse: Population Estimates, 1991. Rockville, Md: National Institute on Drug Abuse; 1991.
- 33.SAS Procedures Guide, Version 8. Cary, NC: SAS Institute Inc; 1999.
- 34.Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal Analysis. Hoboken, NJ: Wiley-Interscience; 2004.
- 35.Molnar BE, Buka SL, Brennan RT, Holton JK, Earls F. A multilevel study of neighborhoods and parent-to-child physical aggression: results from the Project on Human Development in Chicago Neighborhoods. Child Maltreat. 2003;8:84–97. [DOI] [PubMed] [Google Scholar]
- 36.Scales PC, Benson PL, Roehlkepartain EC, Hintz NR, Sullivan TK, Mannes M. The role of neighborhood and community in building developmental assets for children and youth: a national study of social norms among American adults. J Community Psychol. 2001; 29:703–727. [Google Scholar]