Intervening Processes in the Relationship Between Neighborhood Characteristics and Psychological Symptoms in Urban Youth

Abstract

The present study tested for mediators of the relation between neighborhood risk characteristics and internalizing and externalizing symptoms in a sample of 1,120 low-income urban adolescents. Results indicate that a composite of census tract variables predicted both internalizing and externalizing symptoms in this sample. Stressful life events and exposure to violence mediated the relation between neighborhood risk and aggression, delinquency, and somatic symptoms. Gender and age were examined as moderators of these mediational pathways. Results of moderated mediation analyses revealed that both gender and age moderated the pathways between stressors and somatic complaints and between exposure to violence and somatic complaints. Implications for intervention and areas for future research are discussed.

Keywords

neighborhood psychopathology stress violence/violent behaviors low income

Shaw and McKay laid the foundation for the investigation of neighborhood risk for behavioral and psychological problems in 1942 when they compared rates of juvenile delinquency across Chicago neighborhoods. Their theory of social disorganization proposes that three primary characteristics of neighborhoods can be potential risk factors for juvenile delinquency: physical status (e.g., changes in the population), economic status (e.g., median family income), and population composition (e.g., racial/ethnic and national breakdown of the neighborhood). Shaw and McKay (1942) found that neighborhoods’ physical status, such as mobility, was correlated with delinquency; presumably, this is because those with greater economic resources can move away from less desirable neighborhoods, leaving behind more destitute and more delinquent youth. Not surprisingly, income also was found to be inversely related to delinquency. Finally, changes in racial/ethnic composition of neighborhoods were associated with increased delinquency. Such changes may create a discontinuity in the community’s roles and values, which in turn gives rise to reduced social control (i.e., connection to society/social relationships that promote positive behavior through attachment and monitoring, Sampson & Laub, 1994).

Since Shaw and McKay’s (1942) early work, a number of studies have documented a link between neighborhoods characterized by urban poverty and racial/ethnic segregation and externalizing outcomes, such as delinquency (e.g., Epstein, Botvin, Diaz, Williams, & Griffin, 2000; Lahey, Gordon, Loeber, Stouthamer-Loeber, & Farrington, 1999; Sampson & Groves, 1989; Tolan & Henry, 1996). There is a dearth of research, however, examining whether structural neighborhood factors are predictive of psychological symptoms other than delinquency in urban youth (for an exception, see Seidman et al., 1998).

In particular, there have been limited studies with adolescents linking neighborhood risk to internalizing symptoms, and research findings in this area have been mixed. Whereas some extant research has found that living in an economically impoverished neighborhood is related to depression and anxiety in adolescents (Aneshensel & Sucoff, 1996), other studies have not corroborated such findings. For example, a recent prospective study with 13-year-olds (Natsuaki et al., 2007) failed to find a connection between neighborhood risk factors and depression. Furthermore, Leventhal and Brooks-Gunn (2003) found neighborhood poverty to be associated with depression and anxiety for boys, ages 8 to 13 years, but not for older adolescents, ages 14 and 18, suggesting age and gender may moderate neighborhood effects. These contradictory findings are representative of the scant literature on potential neighborhood risk factors for developing internalizing symptomatology. Testing for neighborhood risk on specific forms of internalizing distress is important as extant research indicates that low-income urban youth are especially vulnerable to developing particular forms of internalizing distress, such as somatic complaints, anxiety, and depression (Grant, Katz et al., 2004).

Another limitation of the current literature is that few studies have examined potential mediators of the relationship between neighborhoods and psychological symptoms in youth. Recent critiques of neighborhood research highlight the importance of moving beyond main effect analyses to investigate underlying processes that might explain those main effects (Furstenberg & Hughes, 1997; Leventhal & Brooks-Gunn, 2000; Roosa, Jones, Tein, & Cree, 2003). Identification of mediating processes is useful not only for understanding the ways in which neighborhoods affect youth from a conceptual standpoint but also for providing potential avenues and more “tangible” mechanisms that might be addressed through prevention, intervention, and/or policy efforts (Furstenberg & Hughes, 1997; Leventhal & Brooks-Gunn, 2000; Roosa et al., 2003).

Those studies that have attempted to investigate intermediary processes have examined variables such as social control, which emanate from the literature on risk factors for delinquency and other externalizing outcomes (Elliott et al., 1996; Sampson & Groves, 1989; Sampson, Raudenbush, & Earls, 1997). There has been little focus on processes, such as exposure to community violence and stressful life experiences, which might explain the relationship between neighborhood risk factors and both externalizing and internalizing symptoms.

There is substantial evidence that residing in poor urban neighborhoods is associated with exposure to community violence as well as a host of other stressful experiences (e.g., substandard housing, overcrowding, evictions, and disruptions to important services; Allison et al., 1999; Attar, Guerra, & Tolan, 1994; Grant et al., 2003; Henrich, Schwab-Stone, Fanti, Jones, & Ruchkin, 2004; Kliewer et al., 2006; Kupersmidt, Shahinfar, & Voegler-Lee, 2002; Morales & Guerra, 2006; Youngstrom, Weist, & Albus, 2003). From a theoretical perspective (i.e., social disorganization framework), exposure to community violence and stressful life events might be conceptualized as emanating from a lack of social control, linking them specifically to externalizing outcomes such as delinquency (Leventhal & Brooks-Gunn, 2000; Sampson & Groves, 1989; Shaw & McKay, 1942). On the other hand, exposure to violence and stressful experiences might also be conceptualized as threat or loss events, which would suggest a specific theoretical link with internalizing outcomes such as anxiety or depression (Brown, 1993; Mash & Barkley, 2003; Rudolph et al., 2006; Stueve et al., 1998). Empirical evidence suggests that stressful life events, in general, and exposure to community violence, in particular, are predictive of both internalizing and externalizing problems in young people (Cooley-Quille, Boyd, Frantz, & Walsh, 2001; Duckworth, Hale, Clair, & Adams, 2000; Grant, Compas, Thurm, McMahon, & Gipson, 2004; Grant et al., 2003; McMahon, Grant, Compas, Thurm, & Ey, 2003; Tolan, Gorman-Smith, & Henry, 2003).

Although stressful life experiences and community violence could be conceptualized as neighborhood characteristics, themselves, there are theoretical and empirical reasons for viewing them as mediators. First, both stressful life events, in general, and exposure to community violence, in particular, are typically conceptualized and measured at the most proximal level (i.e., individual self-report), whereas neighborhood risk has typically been conceptualized and measured at an aggregate distal level (i.e., census data; Frank, Cerda, & Rendon, 2007; Grant, Compas et al., 2004; Richters & Martinez, 1990). Second, extant empirical investigation has provided evidence that these variables mediate the relationship between poverty (which is often conceptualized as a neighborhood variable) and psychological symptoms (Conger et al., 2002; Grant et al., 2005). Together, theory and empirical evidence provide a basis for examining exposure to community violence and other stressful experiences as potential proximal mediators of the relationship between neighborhood risk and psychological symptoms in young people.

An important consideration in examining potential mediating processes of the association between neighborhood risk and adolescent mental health problems is potential age and gender differences in experiences of intervening risk factors and expressions of symptomatology. For example, there is evidence that neighborhood effects are stronger for boys and for older youth, perhaps, because they are exposed to a greater extent to intervening risk factors (Carlson & Grant, 2008; Elliott et al., 1996; Kroneman, Loeber, & Hipwell, 2004). For example, boys report greater exposure to community violence (Carlson & Grant, 2008). In addition, boys generally report more externalizing symptoms (e.g., Achenbach, 1991; Achenbach & Rescorla, 2001; Lahey et al., 2006) and girls report more internalizing symptoms (Achenbach & Rescorla, 2001; Carlson & Grant, 2008; Garber, Keiley, & Martin, 2002; Hankin & Abramson, 2001). Furthermore, psychological symptoms, particularly internalizing ones, are likely to increase with age (Abela & Hankin, 2008; Hammen & Rudolph, 2003; Hyde, Mezulis, & Abramson, 2008). These findings suggest the importance of understanding potential age and gender effects in order to inform specific prevention and intervention efforts. For these reasons, age and gender were examined as moderators of the mediation models in the present study.

In sum, the present study examined the influence of neighborhood risk characteristics on internalizing and externalizing outcomes in low-income urban youth and also tested for mediators of these relationships as well as moderators of mediational pathways. To accomplish these goals, a neighborhood composite risk variable was developed. That variable was then used to test the following specific hypotheses:

Hypothesis 1: Neighborhood risk will predict internalizing and externalizing symptoms in urban adolescents.

Hypothesis 2: Exposure to community violence and stressful life experiences will mediate the relationship between neighborhood risk and symptoms.

Additional exploratory analyses were conducted to examine the extent to which gender and age might moderate the hypothesized mediational pathways.

Method

Participants

Participants in the present study attended one of seven urban public schools, which were selected based on high percentages of low-income students. Students were classified as low income based on eligibility for free/reduced school lunch programs (Chicago Public Schools Office of Accountability, 1995). The average percentage of low-income students across the seven participating schools was 92%. The seven schools are representative of urban public schools in Chicago that have a high proportion of low-income students (Chicago Public Schools Office of Accountability, 1995).

The sample in the present study included all students for whom there was complete data (92% of items completed on each measure) for each of the measures examined in this study (i.e., stressful life events, exposure to violence, psychological symptoms) as well as valid addresses that could be geocoded for census tract information. The decision to use 92% as our cutoff for including measures in the analyses was based on practices recommended by the developers of our outcome variable measures (Achenbach, 1991). The sample included 1,120 adolescents who ranged in age from 11 to 15 years (mean age = 12.73; 56% female). Thirty-three percent of the students were enrolled in the sixth grade, 31.3% were enrolled in the seventh grade, and 35.9% were enrolled in the eighth grade. The sample was primarily African American (62.9%), with significant representation of Latino/a (13%), and Asian/Asian American (10.3%) students. Other groups represented included European American (5.9%), bi- or multiracial (3.6%), Native American (1.4%), and students who identified themselves as members of some other racial/ethnic group (1.3%).

Procedures

The present study is a part of a larger study examining the relationship between stressful life experiences and mental health problems among urban adolescents. Before collecting any data, approval for the study was obtained from the psychology department institutional review board of the university with which the authors are affiliated.

The schools that agreed to participate in the present study were recruited using a standard procedure. Initially, 29 schools were randomly selected from a list of all Chicago public schools with student bodies that included 75% or more low-income students. These schools were contacted about the study until a total of seven schools agreed to participate. Once seven schools were recruited, recruitment efforts ceased. School administrators were given the option of selecting active or passive consent procedures. Administrators for six of the seven schools that participated in the present study selected passive consent. Thus, parents were advised that their children would be invited to participate in the project if they did not decline for their children to participate by returning the consent form. For the school that selected an active consent procedure, only children whose parents returned consent forms providing permission to participate were eligible. Participation rates for students attending the one school in which active consent procedures were used (i.e., parents were required to return a consent form stating they gave permission for their child to participate in the study) were much lower, approximately 50%. It was not possible to compare nonparticipants (due to nonreturn of consent forms in active consent school or absence/refusal in passive consent schools) with participants, but analyses of variance (ANOVAs) were used to test for study differences in schools in which active versus passive consent were used. No differences were found in major life events, aggression, delinquency, or anxiety/depression. However, youth in the school that allowed passive consent reported higher levels of exposure to violence (77.81 compared with 74.80, F = 7.27, p < .01) and somatic symptoms (4.89 compared with 4.02, F = 6.24, p < .05). Students completed assent forms prior to data collection. Approximately, 90% of the total number of students eligible to participate in the study (i.e., 1,382 sixth-eighth graders) were present on the days of the survey administration (1,244 students), which is consistent with typical absence rates in participating schools (Chicago Public Schools Office of Accountability, 1995). Approximately, 90% of those students (1,120) agreed to participate. Surveys were administered in school classrooms during regular class time by clinical psychology doctoral students, and efforts were made to ensure that at least one research assistant assigned to each classroom identified as a member of the predominant racial/ethnic group represented in that classroom. Questionnaires were read aloud by research assistants to ensure that students with varying reading levels kept pace with the administration, and students were given assistance if they had difficulty understanding any of the questions. Completion of the measures examined in the present study took approximately 50 minutes. Ninety-four percent of participants had completed data on all study measures. Participants with complete data were compared with those with incomplete data using one-way ANOVAs. No significant differences were found between participants with and without complete data.

Measures

Demographics

A one-page questionnaire was used to obtain information regarding participants’ age, gender, grade, and race/ethnicity.

Neighborhood risk composite variable

A one-page questionnaire was used to obtain participants’ addresses. Each student’s address was geocoded to yield the census tract in which he or she lived. For each census tract, the following census tract data were extracted: population, percent unemployment, median family income, and percentage of individuals who are African American (non-Hispanic; U.S. Bureau of the Census, 1990a, 1990b). These individual variables were recoded, with higher scores representing higher risk. The 1,120 student addresses represented 221 distinct census tracts. Table 1 presents basic statistical data on the census tracts included in the present study. As summarized in the table, the neighborhoods varied on a number of dimensions.

Table 1.

Descriptive Statistics for Census Tracts

Census variables	M	SD	Range
Population	21,293	13,643	64-77,598
Median household income	US$17,007	8,879	US$4,999-US$57,416
Percent African American, non-Hispanic	66.0%	0.42	0%-100%
Percent unemployment	22.3%	0.16	0%-100%

Note: 221 census tracts included in analyses.

A neighborhood risk composite variable was created based on theory and research (Shaw & McKay, 1942), suggesting that a constellation of contextual risk factors is likely to cause the greatest damage to young people. Creation of this variable was further justified by empirical evidence of a significant correlation between individual and aggregate measures of neighborhood characteristics (Hadley-Ives, Stiffman, Elze, Johnson, & Dore, 2000). To create the neighborhood risk composite variable, each of the census tract variable was first converted to a standardized score (i.e., z score, with a mean of 0 and standard deviation of 1). The composite neighborhood score for each census tract was then calculated by taking the mean of the standardized scores of each of the census variables.

Exposure to violence

Lifetime exposure to violence was assessed using the Exposure to Violence Survey–Screening Version (Richters & Martinez, 1990), a 51-item true or false questionnaire developed with fifth and sixth grade low-income urban African American youth. The measure asks respondents to report whether they have witnessed or experienced 27 types of violence/crime including gang violence, drug trafficking, burglary, police arrests, assaults, physical threats, sexual assaults, weapon carrying, firearm use, and intentional injuries such as stabbings, gunshots, suicides, and murders. Although victimization may be a more severe form of exposure to violence, witnessing violence and victimization were combined in the present study because significant clinical distress is experienced not only by youth who are victimized but also by those who witness violence (Purugganan, Stein, & Johnson, 2003). Additionally, youth who witness violence are also likely to be victimized (Rosenthal, 2000). Richters and Martinez (1990) report good test-retest reliability for the Exposure to Violence measure (r = .90), and, in the present sample, overall internal reliability was good (α = .94). In addition, in the present sample, the range of reliability values across males (α = .91) and females (α = .87) was small, as was the range of reliability values across grades (.81 to .92), schools (.86 to .90), and racial groups (.90 to .93).

Stressful life experiences

Major life events and daily hassles were assessed using the Urban Adolescent Life Experiences Scale (UALES; Allison et al., 1999). The design of the UALES is based on the Adolescent Perceived Events Scale (APES; Compas, Davis, Forsythe, & Wagner, 1987), a well-established, valid and reliable measure of stressful life events developed on predominantly middle-class European American adolescents. The UALES items were generated by low-income urban, predominantly African American, youth (Allison et al., 1999). Respondents are asked to rate the frequency with which they have been exposed to each of a series of stressful experiences on a scale ranging from 1 through 5, with higher numbers indicating greater frequency of exposure. Sample major life event items include “A friend has died,” “I broke up with a boyfriend or girlfriend,” and “A friend goes to jail.” Sample daily hassle items include “I have poor school supplies” and “I have transportation problems.” Test-retest reliability of the UALES was .84 in a pilot study of 6th through 12th graders (UALES; Allison et al., 1999). The original measure includes positive and negative events. In the present study, the measure was shortened to include only negative events as positive events have not been shown to predict psychological problems (Siegel & Brown, 1988). Furthermore, eight items that overlapped with items on the Youth Self-Report (YSR; Achenbach, 1991) were omitted to ensure that artificially high correlations were not found between stressors and psychological symptoms. The modified version of the UALES used in the present study had a 2-week test-retest reliability of .80 and internal consistency reliability of .92. The range in alpha values was small across schools (from .88 to .94), across grades (.91 to .92), and across racial/ethnic groups (.91 to .93). The alpha value for both males and females on this measure was .91.

Psychological symptoms

Internalizing symptoms of anxiety-depression, somatic complaints, and withdrawal and externalizing symptoms of aggression and delinquency were assessed using the YSR sum scores (Achenbach, 1991)¹. The YSR includes 119 behavior items, which adolescents rate on a 3-point scale as 0 = not true, 1 = somewhat or sometimes true, or 2 = very true or often true of themselves during the past 6 months. The YSR broad-band internalizing scale comprises narrow-band scales of anxiety-depression, somatic complaints, and withdrawal (Achenbach, 1991). The YSR broad-band externalizing scale comprises narrow-band scales of aggression and delinquency (Achenbach, 1991). Broad-band outcomes were not examined in the present study as they represent composites of the narrow-band outcomes and may mask important distinctions in expressions of distress among urban youth. Sample items for the anxiety-depression, somatic complaints, and withdrawal scales, respectively, include “I cry a lot,” “I feel dizzy or lightheaded,” and “I would rather be alone than with others.” Sample items for the aggression and delinquency scales, respectively, include “I get in many fights” and “I hang around with kids who get in trouble.” Normative data for the YSR are based on a nationally representative community sample of children and adolescents with separate norms for boys and for girls (Achenbach, 1991). Reliability and validity are well established for the YSR (Achenbach, 1991). In the present sample, the YSR internal consistency values were good and limited in range across males (α = .95) and females (α = .93) as well as across grades (.93 to .95), across schools (.92 to .99), and across racial/ethnic groups (.92 to .98).

Missing data were handled in the following way: Measures were included in the analyses if 92% or more of the items were completed as recommended by Achenbach and colleagues (Achenbach, 1991; Achenbach & Rescorla, 2001). Mean substitution was used for items that were left blank on measures that had 92% of the items completed.

Results

Initial Analyses and Overview of Primary Results

Univariate normality assumptions were tested by computing the ratio of the skewness and kurtosis statistics divided by their respective standard errors. The distribution of skew was −1.02 to 21.43 and kurtosis was −5.86 to 34.05. Further model assumptions were examined for three characteristics of bivariate normality: functional form, normality, and homoscedasticity as recommended by Singer and Willett (2003). Visual inspections of the plots associated with each test revealed the shape to be linear and normal, with randomly distributed residuals. Table 2 presents a correlation matrix of study variables as well as descriptive statistics.

Table 2.

Correlations Among Study Variables

Variable	1	2	3	4	5	6	7	8
1. Census		.28**	.29**	.01	.13**	.01	.19**	.10**
2. Violence			.66**	.26**	.47**	.23**	.34**	.47**
3. Stressors				.39**	.54**	.34**	.38**	.52**
4. Anxious- depressed					.41**	.68**	.57**	.58**
5. Delinquency						.32**	.36**	.69**
6. Withdrawn							.48**	.43**
7. Somatic								.47**
8. Aggression
M	0.00	75.12	189.54	6.05	4.21	4.16	4.12	9.86
SD	0.79	11.50	35.70	5.26	3.50	2.54	3.58	6.62

Note: Stressors = urban adolescent life experiences. Violence = Exposure to Violence–Screening Version. Census = neighborhood composite risk variable.

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

The SPSS Mixed Program was used to fit the multilevel models. Main effects and mediation analyses were conducted using multilevel modeling in order to account for the nesting of participants within census tracts (Bryk & Raudenbush, 1992). Traditional tests for mediation (Baron & Kenny, 1986; Holmbeck, 2002) were applied to multilevel modeling using recommended procedures (Bauer, Preacher, & Gil, 2006; Kenny, Korchmaros, & Bolger, 2003).

Primary results are presented in four steps. First, results of analyses to assess the validity of the neighborhood composite risk variable are summarized. Second, main effects of neighborhood and demographic characteristics on psychological symptoms are reported. Third, results of analyses testing exposure to community violence and stressful life events as mediators of the relationship between neighborhood characteristics and psychological symptoms are presented. Fourth, gender and age were examined as moderators of the mediational pathways that were tested.

Validity of Neighborhood Risk Measure

To empirically test the validity of the neighborhood risk composite variable, k-means cluster analysis was performed on the census variables to identify homogenous groups of cases based on census variables. This procedure yielded a two-cluster solution. One of those clusters was characterized by a high percentage of African Americans, a low total population, a high unemployment rate, and a low median income (n = 595, 54.1%). In comparison, the second cluster of cases was characterized by a lower percentage of African Americans, a higher total population, a lower unemployment rate, and a higher family median income (n = 504, 45.9%). See Figure 1 for results of cluster analysis. When regressions were run using cluster membership to predict YSR outcomes, cluster membership significantly predicted the same YSR scales that were predicted by the neighborhood risk composite. The consistency of these findings provides empirical support for the theoretically based neighborhood risk composite variable.

Figure 1.

Cluster centers (z scores) on neighborhood risk variables

Main Effect Analyses—Hypothesis 1: Neighborhood Risk Will Predict Internalizing and Externalizing Symptoms

In the present study, the neighborhood risk composite, students’ age, and gender were entered as Level 2 variables and each of the narrow-band internalizing and externalizing scales was entered as an outcome. Students’ census tract was used as the within-subjects variable and random intercepts were modeled. Gender was the only significant predictor of the anxious-depressed and withdrawn symptoms models, and the associations were positive, indicating that girls reported more of these symptoms than boys. Both gender and neighborhood risk positively predicted somatic complaints, whereas both age and neighborhood risk positively predicted aggression and delinquency symptoms. Results of these analyses are presented in full detail (including pseudo R² statistics) in Table 3. The pseudo R² statistic does not directly translate to the “proportion of variance accounted for” R² statistic used in ordinary least squares regression because it is not possible to obtain this statistic in multilevel models, but it can be interpreted as the relative improvement in model fit when two models are compared (Kreft & de Leeuw, 1998).

Table 3.

Multilevel Modeling Results of Neighborhood Risk and Demographic Variables Predicting Youth Self-Report Psychological Symptoms—Main Effects and Step 1 of Mediation Analyses

Symptoms
Predictor	B	SE	−2 LL	Pseudo R²
Anxious-depressed
Gender	1.15***	0.31
Age	0.04	0.16
Census	−0.05	0.24	6,866.51	4.75
Withdrawn
Gender	0.74**	0.23
Age	0.04	0.11
Census	0.02	0.17	2,396.32	1.12
Somatic complaints
Gender	1.08***	1.44
Age	−0.11	0.11
Census	0.82***	0.15	5,648.25	2.86
Aggression
Gender	0.01	0.39
Age	0.44*	0.20
Census	0.87**	0.32	7,370.93	0.48
Delinquency
Gender	−0.29	0.21
Age	0.48***	0.10
Census	0.57**	0.16	5,910.80	2.07

Note: −2 LL = −2 log likelihood. Census = neighborhood composite risk variable. Gender (1 = male, 2 = female). Pseudo R² statistic is based on the formula comparing the unconditional means model (with no predictors of the outcome variable) with the model with the predictors using the formula: (unrestricted error – restricted error) / unrestricted error (Singer & Willett, 2003).

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

Mediator Analyses—Hypothesis 2: Exposure to Community Violence and Stressful Life Experiences Will Mediate the Relationship Between Neighborhood Risk and Symptoms

A mediating variable is one that explains or accounts for the relationship between two other variables (Baron & Kenny, 1986; Holmbeck, 1997). A series of regression equations are recommended to test for mediating effects, with the following conditions necessary to support a mediational hypothesis: (a) the independent variable (e.g., neighborhood risk) is related to the dependent variable (e.g., internalizing symptoms), (b) the independent variable is related to the mediator (e.g., exposure to violence), (c) the mediator is related to the dependent variable and this relationship remains once the independent variable is included in the equation, (d) the relationship between the independent variable and the dependent variable decreases once the mediator is included in the equation (Baron & Kenny, 1986; Holmbeck, 1997). Full mediation is indicated if the independent variable is no longer significant once the mediator is controlled (Baron & Kenny, 1986; Holmbeck, 1997).

In order to test for possible mediator effects of exposure to violence and stressful life events in the present study, multilevel modeling analyses were run while controlling for age and gender. The first set of analyses was run using the neighborhood risk variable to predict the outcome variables (YSR scales). Controlling for age and gender, the neighborhood risk variable was found to significantly predict symptoms of aggression, delinquency, and somatic complaints such that riskier neighborhoods were predictive of higher levels of symptoms (see Table 3). Second, controlling for age and gender, the composite variable was used to predict total stressful life events and exposure to violence, and significant results were found such that riskier neighborhoods were associated with higher levels of stressors and exposure to violence (see Table 4). Third, multilevel analyses were run again, with each of the mediators predicting symptoms of aggression and delinquency and somatic complaints controlling for age and gender. Results of these analyses were significant and in the positive direction (see Table 5).

Table 4.

Neighborhood Risk Predicting Mediators (Stressors and Exposure to Violence)—Step 2 of Mediation Analyses

	Stressors				Violence
Predictor	B	SE	−2 LL	Pseudo R²	B	SE	−2 LL	Pseudo R²
Gender	−2.47	1.99			−1.48**	0.63
Age	4.96***	1.01			2.45***	0.32
Census	12.13***	1.77	1,1027.24	5.30	4.30***	0.59	8,448.90	5.61

Note: −2 LL = −2 log likelihood ratio. Pseudo R² statistic is based on the formula comparing the unconditional means model (with no predictors of the outcome variable) with the model with the predictors using the formula: (unrestricted error – restricted error) / unrestricted error (Singer & Willett, 2003). Stressors = urban adolescent life experiences. Violence = Exposure to Violence–Screening Version. Census = neighborhood composite risk variable. Gender (1 = male, 2 = female).

p < .05. **p < .01.

Table 5.

Stressors, Exposure to Violence, and Neighborhood Risk Predicting Outcomes—Results of Steps 3 and 4 of Mediation Analyses

	Somatic				Aggression				Delinquency
Predictor	B	SE	−2 LL	Pseudo R²	B	SE	−2 LL	Pseudo R²	B	SE	−2 LL	Pseudo R²
Stressors as predictor of symptoms
Gender	1.14***	.20			0.26	.34			−0.17	.18
Age	−0.32**	.10			−0.01	.17			0.24**	.09
Stressors	0.04***	.01	5,506.28	15.20	0.10***	.01	7,047.56	24.34	0.05***	.01	5,577.155	26.02
Stressors and census as predictors of symptoms
Gender	1.12***	.20			0.28	.34			−0.17	.18
Age	−0.29**	.10			−0.03	.17			0.23*	.09
Census	0.34*	.14			−0.40	.25			−0.08	.13
Stressors	0.04****	.01	5,500.95	15.13	0.10***	.01	7,045.28	24.24	0.05***	.01	5,625.85	26.02
Exposure to violence as predictor of symptoms
Gender	1.24***	.20			0.42	.35			−0.09	.19
Age	−0.40***	.10			−0.21	.18			0.14	.10
Violence	0.11***	.01	5,532.42	13.00	0.27***	.02	7,163.95	19.39	0.14***	.01	5,686.50	18.76
Exposure to violence and census as predictors of symptoms
Gender	1.21***	.20			0.45	.35			−0.09	.19
Age	−0.37***	.11			−0.24	.18			0.14	.10
Census	0.34*	.14			−0.34	.27			−0.01	.14
Violence	0.11***	.01	5,527.20	12.92	0.28***	.02	7,120.33	19.36	0.14***	.01	5,686.49	18.76

Note: −2 LL = −2 log likelihood. Pseudo R² statistic is based on the formula comparing the unconditional means model (with no predictors of the outcome variable) with the model with the predictors using the formula: (unrestricted error – restricted error) / unrestricted error (Singer & Willett, 2003). Stressors = urban adolescent life experiences. Violence = Exposure to Violence–Screening Version. Census = neighborhood composite risk variable. Gender (1 = male, 2 = female).

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

For the final tests of mediation, the neighborhood risk variable and the mediator variables (stressors and exposure to violence) were entered as predictors of the YSR aggression, delinquency, and somatization scales. First, with respect to aggression, results revealed that both mediators remained significant predictors of aggression while neighborhood risk was no longer a significant predictor (see Table 5). Sobel tests confirmed that these mediation effects did reach statistical significance (z = 6.49, p < .001 for stress; z = 6.69, p < .001 for exposure to violence). Second, with respect to delinquency, both of the mediators, again, were significant predictors of symptoms, and, again, neighborhood risk was no longer a significant predictor. Sobel’s test also confirmed these mediation effects were significant (z = 6.50, p < .001 for stress; z = 6.57, p < .001 for exposure to violence). Third, with respect to somatic complaints, stressors and exposure to violence were significant predictors but neighborhood risk remained a significant predictor once the mediators were included in the equation. Sobel’s test confirmed that partial mediation effects were significant for somatic complaints for each mediator (z = 3.45, p < .001 for stress; z = 6.08, p < .001 for exposure to violence). Results of these final steps of mediation analyses are presented in Table 5.

In summary, exposure to violence and stressful life experience mediated the relationship between neighborhood risk and three types of symptoms, aggression, delinquency, and somatic complaints. Both stressors and exposure to violence fully mediated the relationship between neighborhood risk and delinquency and between neighborhood risk and aggression. Both stressors and exposure to violence partially mediated the relationship between neighborhood risk and somatic complaints.

Moderated Mediation Analyses—Hypothesis 3: Gender and Age Will Moderate Mediational Pathways

Moderated mediation analyses using multilevel modeling were conducted to determine whether the mediated relationships that were tested in the previous section varied as a function of gender or age. Moderation could occur in one or both of the paths comprising the indirect effect of neighborhood risk on psychological symptoms: the path from neighborhood risk to the mediator (Step 2 from previous section’s mediation analyses) or the path from the mediator to symptoms (Step 3 from previous section’s mediation analyses).

Moderated mediation analyses were conducted using the guidelines laid out by Muller, Judd, and Yzerbyt (2005) and van Aken and colleagues (2007). Based on these guidelines, four criteria had to be met in the current study to establish moderated mediation: (a) neighborhood risk must significantly predict symptoms; (b) there must not be any moderation of the direct effects of neighborhood risk on symptoms (e.g., no significant Neighborhood risk × Age effect in the prediction of symptoms); (c) the effect of neighborhood risk on either mediator and/or the effect of either mediator on psychological symptoms must be moderated by gender or age; and (d) only if the effect of neighborhood risk on the mediator is moderated, then must the path between the mediator and symptoms be significant. Conversely, only if the effect of the mediator on symptoms is moderated, then must the path between neighborhood risk and the mediator be significant.

Criterion 1 was assessed using the results from the mediation analyses described in the previous section. Neighborhood risk was a significant predictor of three types of psychological symptoms: aggression, delinquency, and somatic complaints. Thus, only these three types of symptoms were included in moderated mediation analyses.

Criterion 2 was assessed by using neighborhood risk, gender, age, and one of the interaction terms (Neighborhood risk × Gender or Neighborhood risk × Age) to predict aggression, delinquency, and somatic complaints. Neither of the interaction terms emerged as a significant predictor of any of the three types of symptoms, indicating that the direct effects of neighborhood risk on aggression, delinquency, and somatic complaints did not vary in magnitude as a function of either gender or age. This result fulfilled the second criterion.

Criterion 3 was assessed by examining both gender and age as moderators of the effects of neighborhood risk on each of the mediator variables and by examining both gender and age as moderators of the effects of each of the mediator variables on aggression, delinquency, and somatic complaints. Separate analyses were conducted to test the effects of each type of moderator on each type of outcome variable. Neither gender nor age moderated the effects of neighborhood risk on either stressors or exposure to violence, indicating that the path from neighborhood risk to the mediators did not vary as a function of gender or age.

Similarly, neither gender nor age emerged as significant moderators of the effects of stressors or exposure to violence on either aggression or delinquency. However, both gender and age did significantly moderate the effects of stressors and exposure to violence on somatic complaints (see Figure 2). Gender significantly moderated the relationship between stressors and somatic complaints (B = .01, SE = .01, p < .05) such that girls reported more somatic complaints than did boys at high levels of stressors, whereas there were no gender differences at low levels of stressors. A similar effect for gender was also found for exposure to violence (B = .05, SE = .02, p < .01).

Figure 2.

Graphical representations of the moderated mediation effects of age and gender on the relationship between stressors and exposure to violence in the prediction of somatic symptoms

Age also significantly moderated the effects of stressors and exposure to violence on somatic complaints (see Figure 2). With respect to stressors, older participants reported more somatic complaints than did younger participants at high levels of stress (B = −.01, SE = .01, p < .05). However, younger participants reported more somatic complaints than older participants at high levels of exposure to violence (B = −.02, SE = .01, p < .05).

Finally, Criterion 4 was assessed by examining the impact of neighborhood risk on each of the mediators controlling for age and gender. Results of these analyses were described in the previous section, indicating that neighborhood risk significantly predicted both stressors and exposure to violence. Thus, the conditions for the moderating role of gender and age on the mediational paths between neighborhood risk and somatic complaints were met.

Discussion

Results of the cluster analyses used to establish validity for the neighborhood composite variable revealed that neighborhood tracts with higher percentages of African Americans had significantly lower median incomes and greater unemployment (both of which have been found to predict maladaptive outcomes; Sampson & Groves, 1989; Shaw & McKay, 1942). This pattern of findings is consistent with prior research indicating that historical and contemporary racism have resulted in unemployment and concentrated poverty in inner-city, predominantly African American neighborhoods, which serve to maintain a de facto segregation for many low-income, urban, African American youth (Garrett, Ng’andu, & Ferron, 1994; Huston, McLloyd, & Garcia Coll, 1997; Wilson, 1987, 1996). Results of the present analyses provide further evidence that low-income urban adolescents of African descent are at greatest risk for exposure to significant neighborhood risk factors.

Results of analyses testing the first hypothesis indicate that neighborhood risk is predictive of symptoms of aggression, delinquency, and somatic complaints, as measured by the YSR. These findings are consistent with existing literature documenting a link between neighborhood factors and externalizing symptoms (Sampson & Groves, 1989; Shaw & McKay, 1942). They also provide new evidence to support the hypothesis that neighborhood-level variables are significant predictors of internalizing symptoms. It is noteworthy that neighborhood factors predicted aggression, delinquency, and somatic complaints but not anxiety-depression or withdrawal symptoms. Prior analyses with the present sample (Grant, Katz et al., 2004) have indicated that symptoms of aggression, delinquency, and somatization are the most common forms of distress in low-income urban youth and that these psychological outcomes are especially likely to co-occur, controlling for sample base rates (Grant, Katz, et al., 2004).

A possible explanation for this pattern is that aggressive and delinquent symptoms are a natural response to the stressors of urban poverty (Gorman-Smith & Tolan, 1998). The contrast between poor living conditions and surrounding wealth and the daily hassles associated with trying to make ends meet, in the context of aggressive role models, may lead to anger, frustration, and aggression. At the same time, a readily available avenue for expression of aggression, in this context, is delinquency. Delinquent activities may be especially appealing for angry and frustrated youth because they provide a “constructive” outlet for aggression that promises to reduce some of the causes of that frustration (e.g., through added income as a result of drug sales or theft; Oyserman & Saltz, 1993; Rosario, Salzinger, Feldman, & Ng-Mak, 2003). Involvement in delinquent activity, in turn, is likely to lead to heightened aggression through increased exposure to danger, increased modeling of violence, and increased pressure to engage in violent behavior (Tolan, Gorman-Smith, & Loeber, 2000). Several studies have reported findings consistent with this pattern (see McCart et al., 2007).

A neighborhood context in which aggression and delinquency are common may dictate that youth who do not wish to withdraw from their social environment must be prepared to fight to avoid victimization (Tolleson, 1997). Symptoms of anxiety and/or depression (e.g., low self-esteem) are likely to be counterproductive to the “show of force” necessary to “win” in an aggressive exchange (Fishkin, Rohrbach, & Johnson, 1997). For these reasons, youth who employ aggression to cope with their violent surroundings may need to express internalizing distress in ways that are less likely to be viewed as signs of weakness (i.e., somatic complaints).

Consistent with these hypothesized explanations, Cassidy and Stevenson (2005) have suggested that neighborhood risk leads to feelings of vulnerability for low-income urban youth and that youth strive to combat such feelings by developing a strong façade, which includes diminished expression of internalizing distress and exaggerated expression of externalizing distress (Cassidy & Stevenson, 2005). Somatic complaints may represent the single form of internalizing distress that does not compromise this strong façade. Youth might even use physical symptoms as an acceptable means for avoiding situations in which aggressive responses might be the most dangerous (Reynolds, O’Koon, Papademetriou, Szczygiel, & Grant, 2001).

Results of main effects analyses also indicate that exposure to violence, stressful life experiences, and gender were predictive of internalizing symptoms, whereas exposure to violence and stressful life experiences also predicted externalizing symptoms. In fact, study results showed that exposure to violence and stressful life experiences accounted for the greatest amount of variance in explaining youths’ psychological symptoms. The link between stressors and symptoms in this study is consistent with findings in other studies that have found exposure to community violence and other stressful life events to predict adolescent mental health problems (Grant, Compas, et al., 2004). The finding that girls in the present study reported more internalizing symptoms than boys also has been documented consistently in earlier research (Garber et al., 2002; Hankin & Abramson, 2001; Hyde et al., 2008).

The lack of gender differences in externalizing symptoms is not consistent with most prior literature (e.g., Achenbach, 1991; Achenbach & Rescorla, 2001; Lahey et al., 2006). There is growing evidence, however, that gender differences in externalizing symptoms are less pronounced in the context of urban poverty (Grant, Katz, et al., 2004). Exposure to community violence in this context may reduce gender differences in externalizing symptoms by increasing the expression of externalizing symptoms in adolescent girls through the mechanisms described above (Fishkin et al., 1997; Sanders, Merrell, & Cobb, 1999).

The final main effects to emerge were the positive relationships between age and aggression and between age and delinquency. This pattern has been reported in prior studies with similar samples (e.g., Farrell, Sullivan, Esposito, Meyer, & Valois, 2005). A possible explanation for these findings is that urban adolescents adopt more aggressive and delinquent behaviors with age as a response to increased exposure to community violence (Stein, Jaycox, Kataoka, Rhodes, & Vestal, 2003).

Before moving to discuss mediational and moderational findings, it is important to acknowledge that the amount of variance in psychological symptoms that was accounted for by neighborhood risk in this study was small. Small effect sizes such as these are consistent with typical neighborhood effects findings in the literature (e.g., Chung, 2006). As discussed further below, it is likely that neighborhood risk exerts an indirect influence on mental health outcomes through a number of intermediary processes (mediators) that account for the remaining variance.

Results of mediational analyses provide insight into the mechanisms and processes through which neighborhood risk negatively affects adolescents. In particular, results are consistent with the hypothesis that low-income urban neighborhoods increase youth’s exposure to community violence and other stressful life experiences and that these stressors, in turn, predict symptoms of aggression, delinquency, and somatic complaints.

These findings fit within the framework of social disorganization theory (Sampson & Groves, 1989; Shaw & McKay, 1942) and more recent theories (i.e., norms/collective efficacy; Leventhal & Brooks-Gunn, 2003), which stipulate that exposure to community violence and exposure to other stressful experiences are likely to be associated with low social control. Such experiences appear to be “proximal transmitters” of neighborhood risk (Roosa et al., 2003). These results suggest that preventive interventions that successfully target exposure to community violence and other stressful life experiences are likely to mitigate the negative effects of urban poverty on adolescent mental health.

As described above, the mediators that emerged in the present study also may help explain the particular manifestation of psychological distress exhibited in the present sample. In particular, the combination of chronic stressors and exposure to community violence in the context of urban poverty may pull specifically for symptoms of aggression, delinquency, and somatic complaints (Grant et al., 2011).

Although neither age nor gender moderated the effects of neighborhood risk on any of the outcome variables, both age and gender did influence two of the mediational pathways. Specifically, both age and gender moderated the pathways from community violence to somatic complaints and from stressful life experiences to somatic complaints. At high levels of stressors and exposure to violence, girls reported more somatic symptoms than boys. This finding is consistent with prior literature and highlights girls’ particular susceptibility to internalizing symptoms in response to stress (Carlson & Grant, 2008; Hankin & Abramson, 2001).

A less predictable pattern emerged for age. At high levels of exposure to violence, younger participants reported more somatic complaints than older participants; however, at high levels of stressful life experiences, the reverse was true: older adolescents reported more somatic complaints than younger adolescents. One possible explanation for this pattern of findings is that youth who live in low-income urban neighborhoods develop strategies for coping with exposure to community violence that minimize their internalizing distress over time (Grant et al., 2011). At the same time, adolescence brings with it a host of new stressors for which youth may be less prepared (e.g., exclusion from peer cliques, romantic rejections, sexual stressors, relationship break-ups). Such stressors may be the reason that stressful experiences were associated with more internalizing distress for older participants in this study. Future research is needed to test this hypothesized interpretation.

Limitations

An important issue in neighborhood research is the source of data that are selected for analysis. Census data have been one of the most commonly used measures of neighborhood effects (Gonzalez, Cauce, Friedman, & Mason, 1996; Seidman et al., 1998), and some researchers have found few significant differences between resident-defined neighborhoods and census tract data (Elliott & Huizinga, 1990). However, other scholars argue that studies of neighborhood effects should be conducted using residents’ self-reported neighborhoods and/or street blocks (Gonzalez et al., 1996; Sampson & Groves, 1989; Taylor, 1997). Future research on mediating processes should take advantage of such approaches. Nonetheless, although census data may not have the same face validity as ethnographic approaches to defining neighborhoods, census tracts are usually defined with input from community members and often are drawn to reflect the physical boundaries and social and ethnic divisions which define neighborhoods (Leventhal & Brooks-Gunn, 2000).

Another limitation of the present study is that the data are cross-sectional. Therefore, we were not able to determine direction of effects or infer causality. Although the mediational findings are consistent with a causal hypothesis (based on the theoretical models that served as a basis for this study), additional research with longitudinal data is needed to test this hypothesis.

Finally, there are factors that limit the generalizability of the findings from the current study. In particular, the use of a school-based sample is less likely to include students with poor school attendance and may limit the representativeness of the sample for low-income urban youth.

Implications and Conclusion

Results of the present study suggest that neighborhoods can place youth at risk for both internalizing and externalizing problems. The most important conclusion from the present study, therefore, is that public policy shifts are needed to reduce high rates of poverty and de facto segregation so that young people are protected from these negative environments. At the state and federal level, policy initiatives could allocate necessary economic resources to decrease neighborhood poverty and thereby reduce incidences of crime and delinquency. At the local level, policy initiatives could promote collective school and neighborhood action to increase social control and cohesion among neighbors. At both levels, these types of policy initiatives could help reduce future stressful experiences, including exposure to community violence, and attenuate the effects of neighborhood risk on urban youth. Until such policy strategies are enacted, the development of effective prevention and intervention programs that target youth’s exposure to violence and stressful life experiences as well as their ability to cope with such exposure may help reduce the negative effects urban poverty has on youth.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

Bios

Brian N. Katz is director of psychology and family services at The Centre for Child Development in Surrey, British Columbia, Canada. He received his PhD in clinical child psychology from DePaul University and completed his clinical internship at Children’s Memorial Hospital in Chicago. He completed a postdoctoral fellowship at Brown Medical School in Providence, Rhode Island. His current research interests include evaluating the role of neighborhood and school environments on the developmental trajectories of urban youth.

Patricia Esparza is an adjunct professor in the psychology and counseling department at Webster University in Geneva, Switzerland. She also serves as a consultant for the department of mental health and substance abuse at the World Health Organization. She received her PhD in clinical community psychology from DePaul University. Her primary research interests are in coping mechanisms and mental health outcomes.

Jocelyn Smith Carter is an assistant professor of clinical child psychology at DePaul University. She received her PhD from Vanderbilt University and completed her clinical internship at Children’s Memorial Hospital in Chicago. Her research focuses on mental and physical health outcomes in urban adolescents and their families.

Kathryn E. Grant is a professor of clinical child psychology at DePaul University. She received her PhD from the University of Vermont and completed her clinical internship at Cook County Hospital in Chicago. Her research focuses on stressful life experiences and psychological symptoms affecting adolescents.

David A. Meyerson is a PhD candidate in the clinical child psychology program at DePaul University. He received his MA from DePaul in 2010. His research interests include stress, coping, resilience, and posttraumatic growth among children and adolescents as well as mentoring interventions for youth.

References

Abela

J. R. Z.

Hankin

B. L.

(2008). Handbook of depression in children and adolescents. New York, NY: Guilford.

Achenbach

T. M.

(1991). Manual for the Youth Self-Report and 1991 Profile. Burlington: University of Vermont, Department of Psychiatry.

Achenbach

T. M.

Rescorla

L. A.

(2001). Manual for the ASEBA school-age forms & profiles. Burlington: University of Vermont, Research Center for Children, Youth, & Families.

Aguinis

(2003). Regression analyses for categorical moderators. New York, NY: Guilford.

Allison

K. W.

Burton

Marshall

Perez-Febles

Yarrington

Kirsh

L. B.

Merriwether-DeVries

(1999). Life experiences among urban adolescents: Examining the role of context. Child Development, 70, 1017-1029.

Aneshensel

C. S.

Sucoff

C. A.

(1996). The neighborhood context of adolescent mental health. Journal of Health and Social Behavior, 37, 293-310.

Attar

B. K.

Guerra

N. G.

Tolan

P. H.

(1994). Neighborhood disadvantage, stressful life events, and adjustment in urban elementary-school children. Journal of Clinical Child Psychology, 23, 391-400.

Baron

R. M.

Kenny

D. A.

(1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182.

Bauer

D. J.

Preacher

K. J.

Gil

K. M.

(2006). Conceptualizing and testing random indirect effects and moderated mediation in multilevel models: New procedures and recommendation. Psychological Methods, 11, 142-163.

10.

Bryk

A. S.

Raudenbush

S. W.

(1992). Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage.

11.

Carlson

G. A.

Grant

K. E.

(2008). The roles of stress and coping in explaining gender differences in risk for psychopathology among African American urban adolescents. Journal of Early Adolescence, 28, 375-404.

12.

Cassidy

E. F.

Stevenson

H. C. J.

(2005). They wear the mask: Hypervulnerability and hypermasculine aggression among African American males in an urban remedial disciplinary school. Journal of Aggression, Maltreatment, and Trauma, 11, 53-74.

13.

Cohen

West

S. G.

Aiken

L. S.

(2000). Applied multiple regression/correlation analysis for the behavioral sciences. Mahwah, NJ: Lawrence Erlbaum.

14.

Conger

R. D.

Wallace

L. E.

Sun

Simons

R. L.

McLoyd

V. C.

Brody

G. H.

(2002). Economic pressure in African American families: A replication and extension of the family stress model. Developmental Psychology, 38, 179-193.

15.

Crowder

South

S. J.

(2003). Neighborhood distress and school dropout: The variable significance of community context. Social Science Research, 32, 659-698.

16.

Cooley-Quille

Boyd

R. C.

Frantz

Walsh

(2001). Emotional and behavioral impact of exposure to community violence in inner-city adolescents. Journal of Clinical Child Psychology, 30, 199-206.

17.

Chicago Public Schools Office of Accountability. (1995). The state report cards for 1994-1995: An analysis of the first decade. Chicago, IL: Chicago Public Schools.

18.

Chung

H. L.

(2006). Relations between neighborhood factors, parenting behaviors, peer deviance, and delinquency among serious juvenile offenders. Developmental Psychology, 42, 319-331.

19.

Compas

B. E.

Davis

G. E.

Forsythe

C. J.

Wagner

B. M.

(1987). Assessment of major and daily stressful events during adolescence: The Adolescent Perceived Events Scale. Journal of Consulting and Clinical Psychology, 55, 534-541.

20.

Duckworth

M. P.

Hale

D. D.

Clair

S. D.

Adams

H. E.

(2000). Influence of interpersonal violence and community chaos on stress reactions in children. Journal of Interpersonal Violence, 15, 806-826.

21.

Elliott

D. S.

Huizinga

(1990, August). The mediating effects of social structure in high-risk neighborhoods. Paper presented at the 8th annual meeting of the American Sociological Association, Washington, DC.

22.

Elliott

D. S.

Wilson

W. J.

Huizinga

Sampson

R. J.

Elliott

Rankin

(1996). The effects of neighborhood disadvantage on adolescent development. Journal of Research in Crime and Delinquency, 33, 389-426.

23.

Epstein

Botvin

G. J.

Diaz

Williams

Griffin

(2000). Aggression, victimization and problem behavior among inner-city minority adolescents. Journal of Child and Adolescent Substance Abuse, 9, 51-66.

24.

Farrell

A. D.

Sullivan

T. N.

Esposito

L. E.

Meyer

A. L.

Valois

R. F.

(2005). A latent growth curve analysis of the structure of aggression, drug use, and delinquent behaviors and their interrelations over time in urban and rural adolescents. Journal of Research on Adolescence, 15, 179-204.

25.

Fishkin

S. A.

Rohrbach

L. A.

Johnson

C. A.

(1997). Correlates of youths’ fears of victimization. Journal of Applied Social Psychology, 26, 1601-1616.

26.

Frank

Cerda

Rendon

(2007). Barrios and burbs: Residential context and health-risk behaviors among Angeleno adolescents. Journal of Health and Social Behavior, 48, 283-300.

27.

Furstenberg

F. F.

Hughes

M. E.

(1997). The influence of neighborhoods on children’s development: A theoretical perspective and a research agenda. In Brooks-Gunn

Duncan

Aber

J. L.

(Eds.), Neighborhood poverty: Context and consequences for children (pp. 23-47). New York, NY: Russell Sage.

28.

Garber

Keiley

M. K.

Martin

N. C.

(2002). Developmental trajectories of adolescents’ depressive symptoms: Predictors of change. Journal of Consulting and Clinical Psychology, 70, 79-95.

29.

Garrett

Ng’andu

Ferron

(1994). Poverty experiences of young children and the quality of their home environments. Child Development, 65, 331-345.

30.

Gonzalez

N. A.

Cauce

Friedman

R. J.

Mason

C. A.

(1996). Family, peer, and neighborhood influence on academic achievement among African-American adolescents: One year prospective effects. American Journal of Community Psychology, 24, 365-387.

31.

Gorman-Smith

Tolan

P. H.

(1998). The role of exposure to community violence and developmental problems among inner-city youth. Development and Psychopathology, 10, 101-116.

32.

Grant

K. E.

Thomas

K. J.

Carlson

G. C.

Gipson

P. Y.

Mance

G. A.

Carter

J. S.

. . . Lambert

S. F.

(2011). Specific processes linking economic stressors, exposure to community violence, and psychological symptoms affecting low-income urban youth. Manuscript submitted for publication.

33.

Grant

K. E.

Compas

Stuhlmacher

Thurm

A. E.

McMahon

S. D.

Halpert

J. A.

(2003). Stressors and child/adolescent psychopathology: Moving from markers to mechanisms of risk. Psychological Bulletin, 129, 447-466.

34.

Grant

K. E.

Compas

B. E.

Thurm

A. E.

McMahon

S. D.

Gipson

P. Y.

(2004). Stressors and child and adolescent psychopathology: Measurement issues and prospective effects. Journal of Clinical Child & Adolescent Psychology, 33, 412-425.

35.

Grant

K. E.

Katz

B. N.

Thomas

K. J.

O’Koon

J. H.

Meza

C. M.

DiPasquale

. . . Bergen

(2004). Psychological symptoms affecting low income urban youth. Journal of Adolescent Research, 19, 613-634.

36.

Grant

K. E.

McCormick

Poindexter

Simpkins

Janda

C. M.

Thomas

K. J.

. . . Taylor

J. J.

(2005). Exposure to violence and parenting as mediators between poverty and psychological symptoms in urban African American adolescents. Journal of Adolescence, 28, 507-521.

37.

Hadley-Ives

Stiffman

A. R.

Elze

Johnson

S. D.

Dore

(2000). Measuring neighborhood and school environments: Perceptual and aggregate approaches. Journal of Human Behavior in the Social Environment, 31, 1-28.

38.

Hammen

Rudolph

K. D.

(2003). Childhood mood disorders. In Mash

E. J.

Barkley

R. A.

(Eds.), Child psychopathology (2nd ed.; pp. 233-278). New York, NY: Guildford.

39.

Henrich

C. C.

Schwab-Stone

Fanti

Jones

S. M.

Ruchkin

(2004). The association of community violence exposure with middle-school achievement: A prospective study. Journal of Applied Developmental Psychology, 25, 327-348.

40.

Hankin

B. L.

Abramson

L. Y.

(2001). Development of gender differences in depression: An elaborated cognitive vulnerability-transactional stress theory. Psychological Bulletin, 127, 773-796.

41.

Holmbeck

G. M.

(1997). Toward terminological, conceptual, and statistical clarity in the study of mediators and moderators: Examples from the child-clinical and pediatric psychology literatures. Journal of Consulting and Clinical Psychology, 65, 599-610.

42.

Holmbeck

G. N.

(2002). Post-hoc probing of significant moderational and mediational effects in studies of pediatric populations. Journal of Pediatric Psychology, 27, 87-96.

43.

Huston

McLoyd

V. C.

Garcia Coll

(1997). Poverty and behavior: The case for multiple methods and levels of analysis. Developmental Review, 17, 376-393.

44.

Hyde

J. S.

Mezulis

A. H.

Abramson

L. Y.

(2008). The ABCs of depression: Integrating affective, biological, and cognitive models to explain the emergence of the gender difference in depression. Psychological Review, 115, 291-313.

45.

Kenny

D. A.

Korchmaros

J. D.

Bolger

(2003). Lower level mediation in multilevel models. Psychological Methods, 8, 115-128.

46.

Kliewer

Parrish

K. A.

Taylor

K. W.

Jackson

Walker

J. M.

Shivy

V. A.

(2006). Socialization of coping with community violence: Influences of caregiver coaching, modeling, and family context. Child Development, 77, 605-623.

47.

Kroneman

Loeber

Hipwell

A. E.

(2004). Is neighborhood context differently related to externalizing problems and delinquency for girls compared with boys? Clinical Child and Family Psychology Review, 7(2), 109-122.

48.

Kupersmidt

J. B.

Shahinfar

Voegler-Lee

M. E.

(2002). Children’s exposure to community violence. In Greca

A. M. L.

Silverman

W. K.

Vernberg

E. M.

Roberts

M. C.

(Eds.), Helping children cope with disasters and terrorism (pp. 381-401). Washington, DC: American Psychological Association.

49.

Lahey

B. B.

Gordon

R. A.

Loeber

Stouthamer-Loeber

Farrington

D. P.

(1999). Boys who join gangs: A prospective study of predictors of first gang entry. Journal of Abnormal Child Psychology, 27, 261-276.

50.

Lahey

B. B.

Van Hulle

C. A.

Waldman

I. D.

Rodgers

J. L.

D’Onofrio

B. M.

Pedlow

. . . Keenan

(2006). Testing descriptive hypotheses regarding sex differences in the development of conduct problems and delinquency. Journal of Abnormal Child Psychology, 34, 737-755.

51.

Leventhal

Brooks-Gunn

(2000). The neighborhoods they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126, 309-337.

52.

Leventhal

Brooks-Gunn

(2003). Moving to opportunity: An experimental study of neighborhood effects on mental health. American Journal of Public Health, 93, 1576-1582.

53.

Mash

E. J.

Barkley

R. A.

(2003). Child psychopathology (2nd ed.). New York, NY: Guilford.

54.

Maton

K. I.

Zimmerman

M. A.

(1992). Psychosocial predictors of substance use among urban Black male adolescents. Drugs and Society, 6, 79-113.

55.

McCart

M. R.

Smith

D. W.

Saunders

B. E.

Kilpatrick

D. G.

Resnick

Ruggiero

K. J.

(2007). Do urban adolescents become desensitized to community violence? Data from a national survey. American Journal of Orthopsychiatry, 77, 434-442.

56.

McMahon

S. D.

Grant

K. E.

Compas

B. E.

Thurm

A. E.

(2003). Stress and psychopathology in children and adolescents: Is there evidence of specificity? Journal of Child Psychology and Psychiatry, 44, 107-133.

57.

Morales

J. R.

Guerra

N. G.

(2006). Effects of multiple context and cumulative stress on urban children’s adjustment in elementary school. Child Development, 77, 907-923.

58.

Muller

Judd

C. M.

Yzerbyt

V. Y.

(2005). When moderation is mediated and mediation is moderated. Journal of Personality and Social Psychology, 89, 852.

59.

Natsuaki

M. N.

Brody

G. H.

Simons

R. L.

Gibbons

F. X.

Cutrona

C. E.

(2007). African American children’s depressive symptoms: The prospective effects of neighborhood disorder, stressful life events, and parenting. American Journal of Community Psychology, 39, 163-176.

60.

Oyserman

Saltz

(1993). Competence, delinquency, and attempts to attain possible selves. Journal of Personality and Social Psychology, 65, 360-374.

61.

Purugganan

O. H.

Stein

R. E.

Johnson

E. S.

(2003). Exposure to violence and psychosocial adjustment among urban school-aged children. Journal of Developmental and Behavioral Pediatrics, 24, 424-430.

62.

Reynolds

L. K.

O’Koon

J. H.

Papademetriou

Szczygiel

Grant

K. E.

(2001). Stress and somatic complaints in low-income urban adolescents. Journal of Youth and Adolescence, 30, 499-514.

63.

Richters

J. E.

Martinez

(1993). The NIMH community violence project: I. Children as victims of and witnesses to violence. Psychiatry, 56, 7-21.

64.

Roosa

M. W.

Jones

Tein

J. Y.

Cree

(2003). Prevention science and neighborhood influences on low-income children’s development: Theoretical and methodological issues. American Journal of Community Psychology, 31, 55-72.

65.

Rosario

Salzinger

Feldman

Ng-Mak

D. S.

(2003). Community violence exposure and delinquent behaviors among youth: The moderating role of coping. Journal of Community Psychology, 31, 489-512.

66.

Rosenthal

B. S.

(2000). Exposure to community violence in adolescence: Trauma symptoms. Adolescence, 35, 271-284.

67.

Rudolph

K. D.

Hammen

Daley

S. E.

(2006). Mood Disorders. In Wolfe

D. A.

Mash

E. J.

Wolfe

D. A.

Mash

E. J.

(Eds.), Behavioral and emotional disorders in adolescents: Nature, assessment, and treatment (pp. 300-342). New York, NY US: Guilford Publications.

68.

Sampson

R. J.

Laub

J. H.

(1994). Urban poverty and the family context of delinquency: A new look at structure and process in a classic study. Child Development, 65, 523-540.

69.

Sampson

R. J.

Groves

W. B.

(1989). Community structure and crime: Testing social-disorganization theory. American Journal of Sociology, 94, 774-802.

70.

Sampson

R. J.

Raudenbush

S. W.

Earls

(1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277, 918-924.

71.

Sanders

D. E.

Merrell

K. W.

Cobb

H. C.

(1999). Internalizing symptoms and affect of children with emotional and behavioral disorders: A comparative study with an urban African American sample. Psychology in the Schools, 36, 187-197.

72.

Seidman

Yoshikawa

Roberts

Chesir-Teran

Allen

Friedman

J. L.

Aber

J. L.

(1998). Structural and experiential neighborhood contexts, developmental stage, and antisocial behavior among urban adolescents in poverty. Development and Psychopathology, 10, 259-281.

73.

Siegel

Brown

(1988). A prospective study of stressful circumstances, illness symptoms, and depressed mood among adolescents. Developmental Psychology, 24, 715-721.

74.

Shaw

C. R.

McKay

H. D.

(1942). Juvenile delinquency and urban areas. Chicago, IL: University of Chicago Press.

75.

Sobel

M. E.

(1982). Asymptomatic confidence intervals for indirect effects in structural equation models. In Leinhardt

(Ed.), Sociology and methodology 1982 (pp. 290-312). San Francisco, CA: Jossey-Bass.

76.

Song

(1994). The Youth Self-Report Inventory: A study of its measurements fidelity. Psychological Assessment, 6, 236-245.

77.

Stein

B. D.

Jaycox

L. H.

Kataoka

Rhodes

H. J.

Vestal

K. D.

(2003). Prevalence of child and adolescent exposure to community violence. Clinical Child and Family Psychology Review, 6, 247-264.

78.

Stueve

Dohrenwend

B. P.

Skodol

A. E.

(1998). Relationships between stressful life events and episodes of major depression and nonaffective psychotic disorders: Selected results from a New York risk factor study. In Dohrenwend

B. P.

Dohrenwend

B. P.

(Eds.), Adversity, stress, and psychopathology (pp. 341-357). New York, NY US: Oxford University Press.

79.

Taylor

R. B.

(1997). Social order and disorder of street blocks and neighborhoods: Ecology, microencology, and the systemic model of social disorganization. Journal of Research in Crime and Delinquency, 34, 113-155.

80.

Tolan

P. H.

Gorman-Smith

Henry

D. B.

(2003). Developmental ecology of urban males’ youth violence. Developmental Psychology, 39, 274-291.

81.

Tolan

P. H.

Gorman-Smith

Loeber

(2000). Developmental timings of onsets of disruptive behavior and later delinquency of inner-city youth. Journal of Child and Family Studies, 9, 203-220.

82.

Tolan

P. H.

Henry

(1996). Patterns of psychopathology among urban poor children: Comorbidity and aggression effects. Journal of Consulting and Clinical Psychology, 64, 1094-1099.

83.

Tolleson

(1997). Death and transformation: The reparative power of violence in the lives of young Black inner-city gang members. Smith College Studies in Social Work, 67, 415-431.

84.

U.S. Bureau of the Census. (1990a). Census File Summary Tape File 3-A (STF 3-A). Retrieved from http://www.census.gov/

85.

U.S. Bureau of the Census. (1990b). Census File Summary Tape File 3 (STF 3)—Sample data. Retrieved from http://www.census.gov/

86.

van Aken

Junger

Verhoeven

van Aken

M. A. G.

Dekovic

Denissen

J. J. A.

(2007). Parental personality, parenting and toddlers’ externalising behaviours. European Journal of Personality, 21, 993-1015.

87.

Wilson

W. J.

(1996). When work disappears: The world of the new urban poor. New York, NY: Alfred A. Knopf.

88.

Wilson

W. J.

(1987). The truly disadvantaged: The inner city, the underclass, and public policy. Chicago, IL: University of Chicago Press.

89.

Youngstrom

Weist

M. D.

Albus

K. E.

(2003). Exploring violence exposure, stress, protective factors, and behavioral problems among inner-city youth. American Journal of Community Psychology, 32, 115-129.