The Link Between Mental Health Problems and Youth Violence in Adolescence

Abstract

The early identification of mental illness in youngsters is an important goal for youth, their families, and society. This study utilized continuous indicators of DSM-oriented psychopathology to explore the link between adolescent mental health and physical violence. Relying on data from the Project on Human Development in Chicago Neighborhoods (PHDCN) and controlling for various community, friend, family, and individual risk factors of violence, the role of various mental health problems on self-reported violence is examined. Both violence prevalence and frequency outcomes were studied. Results indicated that oppositional defiant problems was a weak predictor of violence prevalence but stronger predictor of violence frequency, controlling for other indicators. Other individual-level predictors of violence included prior violence, deviant peers, family criminality and mental health problems, and poor family relations. Community-level predictors were neighborhood ties, neighborhood decline, neighborhood organizations, and anomie, though the latter variable reduced offending.

Keywords

mental health oppositional defiant delinquency violence PHDCN multilevel

A surprising proportion of people suffer from mental health problems throughout their lifetimes, and some people with mental health problems are at risk of becoming involved in criminal offending. The nature and severity of mental health problems examined in research vary from general composite indicators to specific diagnostic indicators. Much of the literature has relied on specific operational criteria from the Diagnostic and Statistical Manual of Mental Disorders (DSM; American Psychiatric Association [APA], 2000) to diagnose psychiatric disorders among adults as well as children and adolescents. DSM-oriented mental health problems studied include externalizing problems such as attention-deficit/hyperactivity disorder (ADHD), conduct disorder (CD), and oppositional defiant disorder (ODD) and internalizing problems such as anxiety, depression, and somatic complaints. In prospective studies, onset of DSM-oriented mental health problem prevalence has been shown to occur at rates greater than 40% in middle to late adolescence (Copeland, Shanahan, Costello, & Angold, 2011; Costello, Mustillo, Erkanli, Keeler, & Angold, 2003; Jaffee, Harrington, Cohen, & Moffitt, 2005; Kim-Cohen et al., 2003), with some of these problems remaining stable over time and others transitioning into different disorders (e.g., Kim-Cohen et al., 2003).

Often, studies of DSM-oriented problems rely on categorical classifications of mental illness. Some research suggests, however, that continuous or dimensional measures of mental health problems may serve as better predictors for outcomes such as delinquency among adolescents (e.g., Fergusson & Horwood, 1995). Indeed, in meta-analyses comparing categorical and dimensional measures of psychopathology, Markon, Chmielewski, and Miller (2011) demonstrated dimensional measures of psychopathology were more reliable and valid than categorical measures. The present study employs dimensional measures of DSM-oriented mental health problems to examine the link between mental health and violence among a nonforensic sample of adolescents.

Mental Health and Delinquency and Violence

Literature has demonstrated an association between certain mental health problems and delinquency, and weak, inconsistent, and nonexistent associations with others. For externalizing problems such as ODD and ADHD, the literature finds a more consistent association with delinquency. ODD, a disorder marked by consistent defiant, hostile, and disobedient behavior toward authority figures (APA, 2000), has been shown to be a precursor to adult deviance and youth violence, especially when there is an early age of onset (Farrington, 2005; Langbehn, Cadoret, Yates, Troughton, & Stewart, 1998). Children diagnosed with ADHD, characterized by persistent inattention, impulsivity, and hyperactivity (APA, 2000), have higher levels of delinquent and deviant behaviors in adolescence and adulthood, especially when these disruptive actions begin in childhood (Moffitt & Silva, 1988). A minority of all ADHD children (at most 20%-25%) have been shown to persist in their antisocial offending patterns into adulthood (Rapport & Chung, 2000). Those persons who persisted in their ADHD-related antisocial behavior were more likely to develop criminal career trajectories and committed serious acts, including violence (Broidy et al., 2003; Moffitt, 1990).

For internalizing behaviors such as anxiety, depression, and somatic problems, studies have revealed less consistent findings with regard to delinquency and violence. Youths with anxiety disorders have been found in some studies to be less likely to commit violent acts, with scholars opining that anxiety mediates other forms of disruptive problem behaviors, especially in boys (Walker et al., 1991). As such, anxiety disorders may serve as a potential protective factor against aggressive behaviors, especially when found to co-occur with oppositional behaviors (Connor, 2002). Depressive disorders have been linked empirically with disruptive behaviors in clinical populations of youth (Biederman, Mick, Faraone, & Burback, 2001; Puig-Antich, 1982). Forms of depression in children and adolescents have been associated with violent behaviors, such as suicide ideation and attempts (Capaldi, 1992; Goldstein, Walton, Cunningham, Trowbridge, & Maio, 2007), delinquency in adolescence (Teplin, 2001), and homicidal ideation in adulthood (Birmaher et al., 1996). Violence research has been extremely limited in its inclusion of somatic problems, but recent studies have begun investigating the link among maltreatment, stress, community violence, and somatic pain as it relates to quality of life issues (see, e.g., Bailey et al., 2005; Crofford, 2007).

The association between mental health and delinquency is particularly noteworthy among forensic and deviant populations. Youths entering the criminal justice system have disproportionally higher rates of mental illness than the general population, with two thirds of males and three quarters of females in detention meeting a clinical diagnosis and 46% of boys and 57% of girls having two or more forms of psychopathology (Teplin et al., 2006; also see Cocozza, 2002). Although still lacking, literature on comorbidity in mental health problems and offending among offending populations has also revealed inconsistent findings. In a meta-analysis of adolescent offenders, Cottle, Lee, and Heilbrun (2001) reported nonsevere pathology, a combination of anxiety and stress disorder, and conduct problems predicted recidivism, but they did not specifically examine distinctions in mental health symptoms. Plattner et al. (2009) revealed ODP was a significant predictor of reincarceration among incarcerated male adolescents, whereas general anxiety disorder and dysthymia predicted reincarceration for incarcerated adolescent girls. In a study of Belgium adolescent offenders, Colins, Vermeiren, Schuyten, and Broekaert (2009) examined the influence of comorbid DSM-oriented problems on offending. Violent offenders were more likely to have depressive problems associated with dysthymia but not major depressive disorders. Violent offenders were not more likely to have ADHD, anxiety, or ODP problems than other offender types.

Other Risk Factors for Delinquency

In addition to mental health problems, various familial, peer, and community factors have been connected to violent behavior among youth. Familial influences are important predictors of antisocial behavior, particularly for younger children. Parenting plays an important part in the socialization process and the etiology of delinquency. In a recent meta-analysis, Hoeve et al. (2009) reported a moderate association between unsupportive parents and delinquency. Youth who experience high levels of conflict within the family setting are more likely to become involved in delinquency. In particular, a substantial body of research has long supported the hypothesis that physical maltreatment, or abuse, leads to delinquency (e.g., Brezina, 1998; Stouthamer-Loeber, Loeber, Homish, & Wei, 2001). Victims of abuse have also been shown to engage in violent offending (Lansford et al., 2007; Maas, Herrenkohl, & Sousa, 2008; Widom & Maxfield, 2001; but see Yun, Ball, & Lim, 2011).

Research also recognizes that having criminal family members increases the risk of delinquency (e.g., Farrington, Coid, & Murray, 2009; van de Rakt, Nieuwbeerta, & Apel, 2009), even when controlling for other relevant factors, including mental health problems (e.g., Farrington, Jolliffe, Loeber, Stouthamer-Loeber, & Kalb, 2001). In addition to criminality, psychopathology can be intergenerationally transmitted (e.g., Goodman et al., 2011; Pfiffner, McBurnett, Rathouz, & Judice, 2005; Weissman et al., 2006). Youth who come from families where the members have mental health issues are also at increased risk of aggressive behavior (e.g., Nigg & Hinshaw, 1998).

For adolescents especially, peer factors are key in the etiology of antisocial behavior, with delinquent peer association consistently found as one of the most robust predictors. Whether by socialization or selection processes (see Haynie, 2001; Knecht, Snijders, Baerveldt, Steglich, & Raub, 2010), youth who associate with deviant peers are more likely to engage in delinquency (e.g., Haynie, 2001; Rebellon, 2006; Warr & Stafford, 1991). Furthermore, peer rejection and conflict have been shown to increase the likelihood of aggressive and delinquent behavior among youth (e.g., Dodge, Price, Coie, & Christopoulos, 1990; Kreager, 2004; Kupersmidt & Coie, 1990).

Neighborhood conditions can also affect delinquency and violence. Families living in socially segregated and disorganized areas are much less likely to develop strong ties with their neighbors or form critical social networks, which in turn leads to a lack of collective attitudes and goals. This inability to foster and participate in routine, positive neighborhood interactions contributes to low collective efficacy and adversely affects families since these neighborhoods are unlikely to have resources to improve conditions (Jencks & Mayer, 1990). Research examining the influence of neighborhood characteristics on crime rates has found that neighborhoods with higher social cohesion, support networks, and numbers of organizations and lower disorder, decline, and perceived violence tend to have lower crime rates (e.g., Markowitz, Bellair, Liska, & Liu, 2001; Morenoff, Sampson, & Raudenbush, 2001; Sampson & Groves, 1989; Sampson & Raudenbush, 1999). The growing body of multilevel research examining the effects of neighborhood characteristics on individual delinquent behavior has corroborated these findings (e.g., Grunwald, Lockwood, Harris, & Mennis, 2010; Molnar, Cerda, Roberts, & Buka, 2008).

The Present Study

This current study extends the literature on the association of mental health problems and violence by examining the risk of violence prevalence and frequency associated with specific DSM-oriented problems, controlling for other individual, family, and community risk factors in a large representative community sample of socioeconomically and culturally diverse adolescents. In addition, this study explores the dependency of mental health problems with other risk factors in their association with violence. Specifically, it is hypothesized that DSM-oriented mental health problems will significantly affect youth self-reported violence, while controlling for comorbidity in mental health problems. It is also hypothesized that mental health problems will moderate the effects of significant family, peer, and neighborhood factors on violence.

Method

Participants

For the individual-level measures, this study relies on secondary data from the first two waves of the Longitudinal Cohort Study from the Project on Human Development in Chicago Neighborhoods (PHDCN). The Longitudinal Cohort Study collected extensive in-home and assessment data from primary caregivers (referred to henceforth as parents) and youth for seven 3-year age cohorts from birth to age 18. Participants were selected utilizing a three-stage stratified sampling design, based on racial/ethnic and socioeconomic status (SES) composition of neighborhood clusters (i.e., homogeneous and contiguous census tract boundaries). Much has been written about the sampling design of the PHDCN and therefore, will, not be reiterated here (for a description, see Earls & Visher, 1997). The result was a sample representative of the general population in the Chicago area with respect to racial/ethnic and socioeconomic distribution at the neighborhood level.

The present study examined data from Waves 1 and 2 of the PHDCN. Wave 1 data were collected between 1995 and 1997 (75% average response rate); Wave 2 data were collected between 1997 and 2000 (86% average response rate). Specifically, Cohorts 12 and 15 were used from these data. Youths were on average 13 years old from Cohorts 12 and 15 at Wave 1. Among these participants, 1,201 provided responses to the items used to create the dependent variables for Wave 2.

For the community-level measures, this study relied on secondary data from the Community Survey from the PHDCN. Using the same sampling design as the Longitudinal Cohort Study, the Community Survey collected data from adult Chicago residents in randomly selected households in 1994. The community version of the Community Survey contains aggregate data at the neighborhood cluster level derived from individual residents’ responses. These data can be linked to the youth data for multilevel analyses.

Dependent Variable

Two dependent variables for self-reported physical violence at Wave 2 were created. Physical violence prevalence was measured as having reported involvement in any of six behaviors within the past year: hitting someone you live with (10.2% reported committing), hitting someone you do not live with (23.1% reported committing), attacking someone with a weapon (4.0% reported committing), throwing objects at people (11.3% reported committing), being in a gang fight (5.8% reported committing), and threatening to hurt someone (6.7% reported committing). Since the sample is representative of the general population of Chicago youth, low prevalence rates were expected.

Frequency of violence was also examined in the study. An additive index (α = .673) was created for the number of times in the past year youths reported having engaged in the same six self-reported violence behaviors used for prevalence at Wave 2. (Although the alpha coefficient of this index, and others below, falls just below a .70 threshold typically used to indicate a good level of internal consistency, most of the indexes have been well established in the literature. Therefore, the indexes, although arguably less consistent, are used in analyses.) This index was highly skewed with the frequency of offending ranging from 0 to 552; hence, the frequency measure was logarithmically transformed base 10 with 1 added to scores before log transformation to permit log of 0 (skewness for raw frequency = 20.26; skewness for log frequency = 1.99).

Individual-Level Independent and Control Variables

Several control variables and independent variables were included from Wave 1 as predictors of Wave 2 violence. Table 1 contains descriptive statistics for each of the variables used in this study. With two noted exceptions, parental maltreatment and poor family relations, the independent variables were sufficiently normally distributed.

Table 1:

Descriptive Statistics for Individual and Community Measures

	n	%		n	%
Individual-level variables (N = 1,201)
Sex			Race
Female	620	51.6	Non-White	181	15.1
Male	581	48.4	White	1,020	84.9
Total	1,201	100.0	Total	1,201	100.0
Socioeconomic status			Ethnicity
Middle or low	866	72.1	Hispanic	646	53.8
High	335	27.9	Non-Hispanic	555	46.2
Total	1,201	100.0	Total	1,201	100.0
Family depression			Family anxiety
No	903	75.2	No	943	78.5
Yes	298	24.8	Yes	258	21.5
Total	1,201	100.0	Total	1,201	100.0
Family criminality			Prior violence
No	822	68.4	Never	642	53.5
Yes	379	31.6	Ever	559	46.5
Total	1,201	100.0	Total	1,201	100.0
Future violence prevalence
No	774	64.4
Yes	427	35.6
Total	1,201	100.0
	M	SD	Min	Max
Individual-level variables (N = 1,201)
Age	13.01	1.57	10	16
Affective Problems	5.21	6.09	0	35
Anxiety problems	6.65	6.68	0	26
Somatic problems	9.07	8.87	0	50
Attention deficit problems	2.74	3.94	0	19
Oppositional defiant problems	5.82	7.04	0	30
Delinquent peers	11.51	2.58	8	21
Family conflict	2.74	1.96	0	9
Parental maltreatment (ln)	0.99	0.90	0	3.14
Poor friendships	7.11	1.96	4	14
Poor family relations (ln)	2.02	0.22	1.10	2.83
Future violence frequency (log)	0.24	0.41	0	2.74
Neighborhood-level variables (N = 78)
Neighborhood ties (EB)	3.14	0.26	2.71	4.16
Anomie (EB)	2.50	0.12	2.21	2.83
Social cohesion (EB)	3.35	0.25	2.78	3.96
Decline (EB)	1.92	0.17	1.47	2.40
Organizations (EB)	−0.27	0.58	−1.80	1.15
Perceived violence (EB)	2.00	0.33	1.41	2.74

Note. EB = measures reflect Bayesian estimation of average person-level scores, adjusted for missing, of Chicago neighborhood clusters.

Sociodemographic characteristics

Dichotomous measures were created for sex (0 = female, 1 = male), race (0 = White, 1 = non-White, mostly Black or African American), ethnicity (0 = non-Hispanic, 1 = Hispanic), and family SES at Wave 1. The categorical measure of family SES was created by dichotomizing the distribution of an imputed maximum SES measure based on the Duncan Socio-Economic Index (Reiss, 1961) of social status, which ranged from 0 to 100, with 100 being the highest social stratification level. The distribution was split at the 75th percentile with 1 = middle and low SES and 0 = high SES. A measure of youths’ age in number of years at Wave 1 was also included in analyses.

Prior delinquency

A dichotomous indicator of prior violent behavior was created from the same items from Wave 1 as the dependent variables for violence at Wave 2, except Wave 1 did not contain an item for threatening to hurt someone (prevalence rates reported: 18.0% for hitting someone they lived with, 34.1% for hitting someone they did not live with, 4.3% for attacking someone with a weapon, 19.5% for throwing objects, and 8.1% for gang fights). The prior violence indicator was scored 0 = never and 1 = ever offended before Wave 1.

DSM-oriented scales

At Wave 1, Cohort 12 and Cohort 15 youths were administered the Youth Self-Report (YSR; Achenbach, 1991), a 112-item instrument containing comparable items to the Child Behavior Checklist (Achenbach, 1992) and appropriate for older children. Youths were asked to indicate how true certain characteristics were about themselves in the past 6 months. For each item, responses were 0 = not true, 1 = somewhat true, and 2 = very true. Achenbach and Rescorla (2001) provide raw score conversions to T scores to allow for comparisons with normalized populations; that is, the scales are standardized for age and gender population differences. For each of the DSM-oriented measures, we created an adjusted T score (T score −50); hence, the scores ranged from 0 to 50, rather than from 50 to 100. Achenbach and Rescorla provide a classification of normal-, borderline-, and clinical-level mental health problems with respect to each DSM-oriented scale. Scores of 0 to 14, 15 to 19, and 20 to 50 were within Achenbach’s normal, borderline, and clinical ranges, respectively. These classifications are used to inform the results of this study. The YSR is widely recognized as a reliable and valid instrument that includes age- and gender-appropriate measures of emotional and behavioral problems among children and young adults (see Achenbach & Edelbrock, 1983; Achenbach, McConaughy, & Howell, 1987; Achenbach & Rescorla, 2001).

The five DSM-oriented scales were as follows. Affective problems was created from 12 items of the YSR (α = .702), including frequent crying, suicide attempts, sleeping problems, and feelings of worthlessness. Anxiety problems was created for 6 items from the YSR (α = .565), including problems with unusual dependence on adults, nervousness, and worrying. Somatic problems was created from 7 items from the YSR (α = .654) that describe physical problems with unknown medical causes such as headaches, nausea, and rashes. Attention-deficit/hyperactivity problems (ADHP) was created from 5 items of the YSR (α = .692), such as restlessness, lack of concentration, and impulsivity. Finally, a DSM-oriented scale for oppositional defiant problems (ODP) was created from 5 items from the YSR (α = .752) for problems such as arguing, disobedience, and having a temper. (For the PHDCN, a few items from Achenbach’s original scales were missing from the survey instrumentation. Specifically, 1 out of 13 items for affective, 2 out of 7 items for anxiety, 1 out of 9 items for somatic, and 2 out of 7 items for ADHP problems were missing. Scale scores, therefore, reflect a more conservative measure of mental health problems.) The YSR also contains a scale for CD problems; however, this scale was excluded from analyses because of the tautology of CD with offending behavior.

Family variables

Six measures of family relations were included in this study. Family conflict is an index from the Family Environment Scale (FES; Moos & Moos, 1994) protocol measuring conflict in family functioning. The family conflict scale has demonstrated moderate internal consistency (α = .70 to .72; Skeer et al., 2011, and Boyd, Gullone, Needleman, & Burt, 1997, respectively), though the reliability was weaker in the present sample (α = .650). The family conflict index is a summary of nine true–false items administered to parents regarding family relations: “We fight a lot in our family,” “Family members rarely become openly angry” (reverse), “Family members sometimes get so angry they throw things,” “Family members hardly ever lose their tempers” (reverse), “Family members often criticize each other,” “Family members sometimes hit each other,” “If there’s a disagreement in our family, we try hard to smooth things over and keep the peace” (reverse), “Family members often try to one-up or out-do each other,” and “In our family, we believe you don’t ever get anywhere by raising your voice” (reverse). Mean substitution was used to replace missing values (n = 27). Higher scores reflected greater family conflict issues.

Three dichotomous (0 = no, 1 = yes) measures of family mental health and criminal involvement history were also included in the study. Parents were asked to complete extensive information about the family, identifying who members of the family were and providing information about each family member’s history of mental health problems and legal issues (Janca, Bucholz, & Janca, 1992). Family member depressed was a dichotomous measure indicating a family member had “ever suffered from depression, that is, they have felt so low for a period of at least two weeks that they hardly ate or slept, or couldn’t work or do whatever they usually do.” Family member anxiety was a dichotomous measure indicating that a family member had “ever had problems with their nerves or had a nervous breakdown.” Family member criminal was a dichotomous measure indicating that a family member “had trouble with the police or been arrested.” Missing values were replaced with the mode, or zero (n = 12, n = 12, n = 20, for depressed, anxiety, and criminal, respectively).

A measure of physical maltreatment from the parent toward the child was developed from parent responses to the Conflict Tactics Scale for Parent and Child (CTSPC; Straus, Hamby, Finkelhor, Moore, & Runyan, 1998). In general, the CTSPC has demonstrated weak to moderate reliability (Straus et al., 1998) for various composite indexes. Parental maltreatment reflected an additive index of five items (α = .657) from the CTSPC. Primary caregivers were asked how many times in the past year they had done the following to the youth: “push, grab, or shove,” “slap or spank,” “kick, bite, or hit,” “hit or try to hit [youth] with something,” and “beat [youth] up.” Responses were 0 = never, 1 = once, 2 = twice, 3 = 3–5 times, 4 = 6–10 times, 5 = 11–20 times, and 6 = more than 20 times. Mean substitution was used to replace missing values (n = 20). Because of high skewness, the index was natural log transformed, with higher scores indicating greater problems with parental abuse or maltreatment.

Youths were also administered the Provision of Social Relations (Turner, Frankel, & Levin, 1983) protocol to evaluate social support available to the youth from family and friends. The family and friend scales for positive and negative support demonstrate good reliability (Turner et al., 1983; Turner, Grindstaff, & Phillips, 1990). Poor family relations reflected an additive index of six items (α = .661) relating to family support. This index included responses to the following items: “No matter what happens, I know that my family will always be there for me should I need them,” “Sometimes I’m not sure if I can completely rely (count) on my family” (reverse), “My family lets me know they think I’m a worthwhile (valuable) person,” “People in my family have confidence in me,” “People in my family help me find solutions to my problems,” and “I know my family will always stand by me.” Responses were 1 = very true, 2 = somewhat true, and 3 = not true. Mean replacement was used for missing values (n = 1). Because of high skewness, this measure was natural log transformed. Higher scores reflected poor family relations.

Friend variables

Two measures of friend or peer relations were included in the study. Poor friendships was a measure created from youths’ responses to the Provision of Social Relations (Turner et al., 1983) protocol to evaluate social support available to the youth from friends. The poor friendships variable was an additive index of five items (α = .636), including “When I’m with my friends I feel completely able to relax and be myself,” “When I want to go out to do things, I know that many of my friends would enjoy (like) doing these things with me,” “I have at least one friend that I could tell anything to,” “I feel very close to some of my friends,” and “My friends would take the time to talk about my problems, should I ever want to.” Responses were 1 = very true, 2 = somewhat true, and 3 = not true. Mean replacement was used for missing values (n = 1). Higher scores reflected poor friend relations.

Youths were also administered the Deviance of Peers (Huizinga, Esbenson, & Weihar, 1991) protocol where they were asked to report on the deviance of their peers. Youths were asked to indicate how many of their friends during the past year had engaged in certain activities, where 1 = none, 2 = some of them, and 3 = all of them. Delinquent peers was an additive index of eight items (α = .788), including “Purposely damaged or destroyed property that did not belong to them,” “Stolen something worth $5 or less,” “Stolen something worth more than $5 but less than $500,” “Stolen something worth more than $500,” “Gone into or tried to go into a building to steal something,” “Gotten into physical (fist) fights with schoolmates/co-workers or friends,” “Hit someone with the idea of hurting them,” and “Attacked someone with a weapon with the idea of seriously hurting them.” No cases were missing for this measure. Higher scores reflected greater involvement with delinquent peers.

Community-Level Variables

The restricted PHDCN data that are publicly available are limited to measures created directly from the Community Survey. Unfortunately, these data do not provide crime and census population (e.g., poverty, race distribution, residential mobility) indicators for the Chicago neighborhood clusters. The scientific directors of the Community Survey have created, however, several community indicators of neighborhood conditions, social support, and perceived crime or disorder. These indicators were created from mean scores for individuals residing in certain neighborhood clusters, adjusted for missing data. Each of the community measures was created by the PHDCN scientific directors using empirical Bayes (EB) estimation based on the distribution of the data (for details about EB scale creation from the Community Survey, see Raudenbush & Sampson, 1999).

Six community measures were included in the regression analyses. Anomie was a summary index of five items reflecting residents’ attitudes about abiding by the law and being goal oriented. Residents were asked to indicate how much they agreed (1 = strongly disagree to 5 = strongly agree) with the following statements: “Laws were made to be broken,” “It’s okay to do anything you want as long as you don’t hurt anyone,” “To make money, there are no right and wrong ways anymore, only easy ways and hard ways,” “Fights between friends or within families is nobody else’s business,” and “Nowadays a person has to live pretty much for today and let tomorrow take care of itself.”

Cohesion was an index of social cohesion within neighborhoods. This index was created by combining residents’ responses to how well they agree with five items: “This is a close-knit neighborhood,” “People around here are willing to help their neighbors,” “People in this neighborhood generally don’t get along with each other” (reverse), “People in this neighborhood do not share the same values” (reverse), and “People in this neighborhood can be trusted.” Responses ranged from 1 = strongly disagree to 5 = strongly agree.

Neighborhood decline was an index that measured whether residents felt certain conditions in their neighborhoods had changed over the past 5 years with respect to “Personal safety,” “The way the neighborhood looks,” “The people living in the neighborhood,” and “The level of police protection in the neighborhood.” Responses for each item were 1 = better, 2 = same, 3 = worse.

Neighborhood organizations was a summary index of the presence of specific social programs, activities, and services within neighborhoods. This index was created by combining residents’ affirmative responses (1 = yes) to eight items about the presence of the following: “A park, playground, or open space within walking distance of your home,” ". . . neighborhood [has] a community newspaper, newsletter, or bulletin,” ". . . neighborhood [has] a crime prevention program or a neighborhood watch,” “A family health service in this neighborhood,” “A block group, tenant association, or any other group dealing with local issues,” “An alcohol or drug treatment program in neighborhood,” “A family planning clinic in the neighborhood,” and “A mental health center in the neighborhood.”

Number of ties was a summary measure reflecting the total number of friends and relatives living in residents’ neighborhoods. Residents were asked to report how many (a) relatives or in-laws and (b) friends lived in their neighborhood. Responses ranged from 1 = none to 5 = 10 or more.

Perceived violence was an additive index of five items for how often residents perceived the following problems occurring in their neighborhoods within the past 6 months: “A fight in this neighborhood in which a weapon was used,” “A violent argument between neighbors,” “Gang fights,” “A sexual assault or rape,” and “A robbery or mugging.” Responses ranged from 1 = never to 4 = often.

Analysis

In a study of neighborhood and individual effects on offending, it is important to account for the nested nature of individuals within neighborhoods. If single-level regression analyses are conducted, Type I errors may be inflated because the effects of youth within neighborhoods depend on neighborhood context. Hierarchical linear modeling (HLM) accounts for nonindependence of youths’ observations (Level 1) nested within neighborhood clusters (Level 2; Hox, 2002; Raudenbush & Bryk, 2002). HLM was used to perform the analyses in this study.

In HLM, analyses begin with the estimation of an unconditional model for the dependent variable, which is used to calculate the amount of variance, referred to as the intraclass correlation (ICC), in the outcome variable between Level 2 units. Significant variation justifies the need for multilevel modeling. Respectable ICC values and design effects (1 + [mean within group sample size − 1] × ICC) greater than 2 (Sorra & Dyer, 2010) indicate multilevel regression is necessary for nested data. In the present study, unconditional HLM models were estimated for violence prevalence and violence frequency. Then, if significant variation between neighborhood clusters existed, full HLM models were estimated. Finally, interaction effects for significant DSM-oriented problems were examined for the models.

Results

Bivariate Results

As shown in Table 2, an examination of bivariate correlations between the individual-level measures indicated that a few of the variables demonstrated moderately strong correlations. The correlations between anxiety problems and affective or depressive problems (r = .55) and ADHP and ODP (r = .54) were strongly positive. The strongest associations with future violent behavior were prior violent behavior, followed by both ODP and delinquent peer associations.

Table 2:

Bivariate Correlations for Individual-Level Measures With YSR Measures for Youths (N = 1,201)

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21
1	Sex	—
2	Age	.01	—
3	Race	−.01	−.03	—
4	Ethnicity	.02	.02	.39*	—
5	SES	−.05	−.03	.22*	.27*	—
6	Prior violence	.10*	.24*	.03	−.19*	−.09*	—
7	Affective	.03	.07*	.04	.10*	−.02	.21*	—
8	Anxiety	.08*	.03	.06*	.08*	−.02	.13*	.55*	—
9	Somatic	.04	−.03	.04	−.06*	−.03	.21*	.42*	.32*	—
10	ADHP	−.07*	.12*	−.04	−.01	−.08*	.27*	.46*	.35*	.27*	—
11	ODP	−.03	.25*	−.01	−.01	−.09*	.44*	.42*	.29*	.22*	.54*	—
12	Delinquent peers	.15*	.12*	.12*	−.09*	−.01	.39*	.23*	.16*	.19*	.23*	.34*	—
13	Family conflict	.00	.06*	−.07*	−.13*	−.01	.14*	.07*	−.01	−.01	.06*	.11*	.10*	—
14	Family depressed	−.05	.07*	−.04	.03	−.01	.08*	.06*	−.02	.03	.05	.10*	.05	.10*	—
15	Family anxiety	−.07*	.02	−.02	.08*	.03	.04	.06*	.02	.04	.04	.06*	.01	.11*	.38*	—
16	Family criminal	−.03	.02	−.05	−.15*	−.05	.15*	.00	−.01	.05	.06*	.10*	.10*	.18*	.24*	.14*	—
17	Maltreatment	.06	−.08*	.08*	−.12*	−.04	.12*	.09*	.07*	.11*	.10*	.16*	.12*	.25*	.07*	.04	.11*	—
18	Poor friendships	.20*	−.06	.11*	−.01	−.02	.08*	.14*	.14*	.10*	.02	.01	.18*	.06*	−.00	.03	.01	.11*	—
19	Poor family	−.05	.16*	.04	.02	.06*	.16*	.30*	.17*	.16*	.19*	.29*	.23*	.17*	.06*	.04	.07*	.08*	.23*	—
20	Violence prev.	.12*	.06*	.06	−.14*	−.03	.34*	.07*	.05	.09*	.12*	.24*	.25*	.12*	.10*	−.03	.16*	.11*	.04	.12*	—
21	Violence freq.	.13*	.07*	.05	−.14*	−.02	.34*	.08*	.08*	.10*	.14*	.25*	.26*	.07*	.09*	−.05	.16*	.12*	.06*	.15*	.94*	—

Note. YSR = Youth Self-Report; ADHP = attention deficit/hyperactivity problems; ODP = oppositional defiant problems.

p < .05.

Table 3 reports the bivariate correlation results for the six community-level, or neighborhood cluster-level, measures. The strongest associations were between social cohesion and perceived violence. Subsequent regression analyses were completed using social cohesion and perceived violence in separate models to avoid multicollinearity problems. Perceived violence was also strongly, positively associated with anomie, but this relationship did not suggest multicollinearity problems. A review of case influence and collinearity diagnostics, including residuals, did not indicate problems with outliers or multicollinearity once social cohesion and perceived violence were separated.

Table 3:

Bivariate Correlations for Neighborhood Cluster Measures (N = 78)

		1	2	3	4	5	6
1	Neighborhood ties	—
2	Anomie	.30*	—
3	Social cohesion	.41*	−.47*	—
4	Perceived violence	−.07	.55*	−.82*	—
5	Decline	−.11	.33*	−.24*	.06	—
6	Organizations	.09	−.19	.19	.05	−.33*	—

p < .05.

Unconditional Model Hlm Results

An unconditional model was estimated to examine the variance in self-reported future violent prevalence among the PHDCN youths across Chicago neighborhood clusters. The HLM prevalence models were estimated with the Bernoulli distribution specified because of the dichotomous outcome variable. The results for the unconditional model are reported in the top half of Table 4. For a neighborhood cluster with a “typical” violence prevalence rate, the expected log-odds of violence was −0.62, corresponding to an odds of EXP[−0.62] = 0.54, or a probability of 1/(1 + EXP[0.62]) = 0.35. This typical probability, associated with a neighborhood cluster-level random effect of zero, was roughly equivalent to the population-wide violence rate estimate (see Table 1 for the mean) of 0.35. This suggests that the outcome was approximately normally distributed. The ICC was .31, but this statistic is not informative in nonlinear models where the individual-level variance is heteroscedastic (Raudenbush & Bryk, 2002). However, the chi-square results indicated that significant variation did exist among the neighborhood clusters in violence prevalence, and thus a hierarchical linear model is best for examining the violence prevalence outcome. Examination of the confidence intervals for the log-odds indicated that 95% of the neighborhood clusters fell between 18.5% and 55.8% with respect to probability of violence. It appears that very few of the neighborhood clusters had violence rates near zero, and in others more than half of the youths were involved in minor violent behavior at Wave 2.

Table 4:

Unconditional Model Results for Future Violence Across Neighborhood Clusters

Violence Prevalence
Fixed Effect		Coeff.	SE
Average log-odds neighborhood mean, γ₀₀		−0.62	0.08*
Random Effect	Variance Component	df	χ²	p Value
Neighborhood mean, u₀ _j	0.20	77	130.08	.000
Level 1 effect, r_ij	0.44
Reliability estimate	0.38
Violence Frequency
Fixed Effect		Coeff.	SE
Average neighborhood mean, γ₀₀		0.24	0.01
Random Effect	Variance Component	df	χ²	p Value
Neighborhood mean, u₀ _j	0.00	77	101.14	.034
Level 1 effect, r_ij	0.17
Reliability estimate	0.22

Note. Model coefficients and standard errors are based on the population-average models.

p < .05.

An unconditional model was also estimated for future violence frequency among the PHDCN adolescents. The HLM frequency unconditional model was estimated for a normal or continuous distribution. The results for the unconditional frequency model are reported in the bottom half of Table 4. The ICC for the unconditional frequency model was .019, and the chi-square results were significant. Examination of the design effects revealed an effect statistic of 1.27 (1 + [15 − 1] × 0.019) for the frequency model, which is less than the expected value of 2 that provides a rule of thumb for justifying the need for multilevel analyses (Sorra & Dyer, 2010). Consequently, the neighborhood measures were attributed to the PHDCN youths and ordinary least squares (OLS) regression analyses were estimated for the frequency of violence model.

Final Regression Model Results

HLM regression results for violence prevalence

In the HLM regression analyses, the individual-level predictors were group-mean centered around the Level 2 (i.e., neighborhood cluster) mean and community-level predictors were grand-mean centered. Group- and grand-mean centering allows for a better estimation of contextual effects in the models, especially considering the stratified nature of the data. The group-mean centered estimation allows for the examination of the change in violence prevalence that occurs to a youth by virtue of residing in one neighborhood versus another, or the contextual effect (see Raudenbush & Bryk, 2002, pp. 134-149). Although centering is less important for Level 2 predictors, centering the community predictors around their grand means for the neighborhood can improve numeric stability in the models (p. 35). The HLM regression results reported below are for fixed effects models.

Models 1 and 2 in Table 5 report the findings of the full multilevel models for violence prevalence, including social cohesion and perceived neighborhood violence, respectively. In both multilevel models, prior violence, sex, race, family member depressive problems, family member criminal involvement, and ODP increased the odds of future violence. Furthermore, in both models, family member anxiety problems decreased the odds of future violence. The effect sizes of the individual-level predictors remained relatively stable across the models. The greatest individual-level predictor of violence prevalence, controlling for all else, was prior violence. The odds of future violence were 2.6 times greater for youths who self-reported prior violent behavior, compared to those without prior violence. The weakest significant predictor of future violence prevalence was ODP, which increased the odds a meager 5%. It is interesting that the magnitude of the effect of ODP was comparable to that of delinquent peer associations.

Table 5:

Regression Results for Community and YSR Youth Informant Measures and Violence

	HLM Prevalence				OLS Frequency (log)
	Model 1		Model 2		Model 3		Model 4
	B	Odds Ratio	B	Odds Ratio	b	β	b	β
Intercept	−0.73*	0.48	−0.73*	0.48	0.33	—	0.08	—
Individual-level variables (N = 1,201)
Sex (male)	0.64*	1.90	0.64*	1.89	0.08*	0.11	0.09*	0.11
Age	−0.09	0.92	−0.09	0.92	−0.01	−0.03	−0.01	−0.03
Socioeconomic status (middle/low)	0.05	1.05	0.05	1.05	0.02	0.02	0.03	0.03
Race (non-White)	0.60*	1.82	0.60*	1.81	0.06	0.05	0.07	0.06
Ethnicity (Hispanic)	−0.15	0.86	−0.14	0.87	−0.08*	−0.10	−0.08*	−0.10
Prior violence (ever)	0.96*	2.62	0.96*	2.61	0.13*	0.16	0.13*	0.16
Affective problems	−0.02	0.98	−0.02	0.98	−0.01	−0.07	−0.01	−0.07
Anxiety problems	−0.00	1.00	−0.00	1.00	0.00	0.02	0.00	0.02
Somatic problems	−0.00	1.00	−0.00	1.00	0.00	0.01	0.00	0.01
Attention deficit problems	0.01	1.01	0.01	1.01	0.02	0.02	0.00	0.02
Oppositional problems	0.04*	1.05	0.04*	1.05	0.01*	0.14	0.01*	0.14
Delinquent peers	0.06*	1.06	0.06*	1.07	0.02*	0.11	0.02*	0.11
Family conflict	0.06	1.06	0.06	1.06	−0.00	−0.02	−0.00	−0.01
Family member depressed	0.39*	1.48	0.39*	1.48	0.05	0.06	0.05	0.06
Family member anxiety	−0.49*	0.61	−0.49*	0.61	−0.09*	−0.09	−0.09*	−0.09
Family member criminal	0.38*	1.47	0.39*	1.47	0.08*	0.09	0.08*	0.09
Parental maltreatment (ln)	−0.00	1.00	−0.00	1.00	0.02	0.04	0.02	0.04
Poor friendships	−0.05	0.95	−0.05	0.95	−0.00	−0.01	−0.00	−0.01
Poor family relations (ln)	0.65	1.92	0.65	1.92	0.14*	0.08	0.14*	0.08
Neighborhood-level variables (N = 78)
Neighborhood ties	0.34	1.40	0.66*	1.94	0.07	0.05	0.02	0.01
Anomie	1.49	4.42	1.02	2.77	−0.24*	−0.08	−0.19	−0.06
Social cohesion	0.71	2.03	—	—	−0.09	−0.06	—	—
Perceived violence	—	—	−0.24	0.78	—	—	0.02	0.01
Decline	1.35*	3.84	1.38*	3.97	0.01	0.01	−0.02	−0.01
Organizations	0.31*	1.37	0.35*	1.41	−0.01	−0.01	−0.01	−0.01
R ²					.17		.17
Adj. R²					.15		.15

Note. YSR = Youth Self-Report; HLM = hierarchical linear modeling; OLS = ordinary least squares. For HLM, model unstandardized coefficients are based on the population-average models, individual-level variables are group centered, and neighborhood-level variables are grand-mean centered.

p < .05.

It is interesting that several of the community-level predictors were significantly related to future violent prevalence among the PHDCN adolescents. Youths who resided in neighborhoods where the residents perceived increased conditions of neighborhood decline (e.g., safety, neighborhood condition, and policing problems) were significantly more likely to engage in future violent behavior. This measure was the strongest predictor in both models, increasing the odds of future violent behavior by almost 4 times compared to youths living in neighborhood clusters with few of the residents reporting such problems. Youths who resided in neighborhoods that contained more neighborhood organizations were also more likely to engage in future violence. For the model including perceived violence, rather than social cohesion, youths residing in neighborhoods where residents reported a higher number of friends and relatives living in the area were also significantly more likely to engage in future violence.

Dependency between the ODP measure and the other significant variables in the prevalence models were examined by including multiplicative interaction variables at the individual level and multiplicative slope interactions at the cross level (i.e., between the individual and community levels). None of the interaction terms for the prevalence models were significant. (Results for the interaction effects are available from the authors on request.)

OLS regression results for violence frequency

Models 3 and 4 in Table 5 report the findings of the OLS regression models for violence frequency, including social cohesion and perceived neighborhood violence, respectively. Similar to the HLM prevalence models, a history of violence, being male, having family members with criminal involvement, and having greater oppositional defiant problems increased the likelihood of engaging in more future violence. Moreover, having family members with anxiety problems significantly reduced future violent behaviors. Unlike the prevalence models, Hispanic ethnicity, rather than race, significantly affected future violence. Hispanic youth reported significantly lower levels of future violent behavior than non-Hispanic youth, controlling for all else. Although family member depression predicted future violence prevalence, it did not predict future violent frequency. In addition, youths who reported poor family relations engaged in significantly more violent acts. Similar to the prevalence models, prior violence was the strongest individual-level predictor of future violent frequency. It is interesting that this effect was greater than the neighborhood measures. Also of import is the magnitude of the ODP measure in the violence frequency models. ODP was the second strongest predictor of future violent frequency, followed closely by delinquent peer associations. The only neighborhood measure to significantly affect violence frequency was anomie. Youths residing in neighborhoods where residents espoused negative attitudes toward the law and unconventional goal orientations had significantly lower levels of violent behavior at Wave 2.

Dependency between the ODP measure and the other significant variables in the frequency models was also examined by including multiplicative interaction variables in OLS regression analyses. Two of the interaction terms for the frequency models were significant. Youths with greater delinquent peer involvement and ODP engaged in significantly more future violence (for social cohesion model: b = 0.001, β = .346, p = .012; for perceived violence model: b = 0.001, β = .347, p = .012). Youths with family members who had anxiety problems and lower ODP engaged in significantly fewer violent behaviors (for social cohesion model: b = −0.010, β = −.101, p = .009; for perceived violence model: b = −0.009, β = −.101, p = .010). (Results for remaining interaction effects are available from the authors on request.)

The utility of using a continuous measure of DSM-oriented problems that accounts for clinical as well as subclinical levels of problems can be demonstrated by examining the difference in predicted violence across the normal, borderline, and clinical thresholds of mental health problems. Figure 1 illustrates the relationship between ODP and future violence frequency controlling for family, peer, and community effects. The graph in Figure 1 illustrates the predicted number of violent behaviors for the observed range of ODP scores, where the remaining predictors in the regression equation are held constant at the mean (using regression coefficients reported in Model 3 in Table 5). Three probability lines are presented: one for the average youth, one for youths with no prior history of violence, and one for youths with a history of violence. Depending on youths’ history of violence, youths with “normal” levels of ODP committed between 1.37 and 2.38 violent behaviors at Wave 2. Youths with “borderline” ODP engaged in 1.80 to 2.61 violent acts. Youths with “clinical” levels of ODP reported involvement in 1.97 to 3.20 violent behaviors. (The results were comparable for the model replacing social cohesion with perceived violence.)

Figure 1:

Predicted Frequencies of Future Violence for ODP Scores by Average, No Prior Violence, and Prior Violence

Discussion

The present study has expanded our knowledge of how DSM-oriented problems affect physical violence among a nonforensic sample of adolescents. In particular, this study has addressed a gap in the literature about how mental health problems relate to violence once risk factors associated with peer, family, and community domains are controlled. To achieve this goal, two hypotheses were proposed. First, it was hypothesized that DSM-oriented mental health problems would be significantly related to violence prevalence and frequency among the adolescents, controlling for individual, family, peer, and neighborhood factors. Second, it was hypothesized that mental health problems would moderate the effects of the other significant risk factors on violent behavior.

The results revealed support for the first hypothesis and partial support for the second hypothesis. Among the adolescents participating in Cohorts 12 and 15 of the PHDCN, ODP was the only significant predictor of self-reported violent behavior while controlling for comorbidity in the DSM-oriented mental health problems. This finding is consistent with that of Boots and Wareham (2009, 2010) in their examination of DSM-oriented problems across multiple cohorts of the PHDCN. Although there was support for the first hypothesis, it is noteworthy that the effect size of ODP on violence prevalence was quite weak, increasing the odds of violence by only 5%; however, the effect of ODP on violence frequency was much more robust.

It is interesting that the effect size of ODP on violence prevalence was comparable to that of delinquent peers for the PHDCN youths. Family risk factors had greater effects on violence prevalence than the mental health problems and peer factors. Youth who reported residing in family settings where one or more family member had a history of criminal involvement or depressive problems were more likely to engage in violence, whereas youth who resided with one or more family member who had anxiety issues were significantly less likely to engage in violence. The influence of family factors on violence frequency, however, was weaker than the ODP and delinquent peer measures. Furthermore, experiencing poor family relations significantly affected violence frequency, unlike prevalence, and family member depression became nonsignificant in the frequency model.

The findings regarding the salience of family factors was not unexpected. As previously mentioned in our literature review, the predictive effect of familial criminality on next generation offending is well established. Moreover, the impact of family-related factors influencing mental health and violence pathways is consistently observed across a multitude of studies. It is therefore noteworthy that our analysis found no significant effects between child maltreatment and youth violence, which was somewhat unanticipated given the robust empirical evidence of this relationship (see, e.g., Maas et al., 2008; Widom & Maxfield, 2001).

It is interesting that adolescents with family members who reported anxiety problems were significantly less likely to commit violence over time. In other words, the presence of anxiety within the family became a protective factor for youths’ self-reported violence, both prevalence and frequency. It is speculated that worrying and concern of a family member may translate into closer supervision, monitoring, and contact with the child that may keep youngsters from having opportunities to engage in risky behaviors. The significant interaction between ODP and family member anxiety seems to reflect this notion. There is also research to support this contention. For instance, Dubrow and Garbarino (1989) reported that mothers living in lower socioeconomic neighborhoods with high rates of violence developed a number of strategies to protect their children from the dangers lurking within their communities, such as informing their children about dangerous places in their communities, enforcing and developing rules to avoid problems and get assistance when necessary, and providing close guardianship and supervision of their children.

The findings with respect to the neighborhood effects were also quite surprising. Consistent with the literature, neighborhood decline increased the risk of violence prevalence. Contrary to theory and prior research, organizations and neighborhood ties (for the prevalence model including perceived violence, not social cohesion) also increased the odds of violent behavior and anomie decreased the frequency of violence. In a study of PHDCN youth from Cohorts 9, 12, and 15, Molnar et al. (2008) found neighborhood organizations and services predicted lower aggression. The authors measured aggression as clinical range problems within a YSR aggression scale rather than using self-reported violent behavior. It is also noteworthy that Molnar et al. found no significant effects for a measure of collective efficacy, which was composed of the social cohesion scale and an informal social control scale in the PHDCN data. Studies of the impact of social networks have also revealed inconsistent findings. For example, Warner and Rountree (1997; Rountree & Warner, 1999) demonstrated that social ties depend on gender and the racial composition of neighborhoods. Such interaction effects are worthy of consideration in future research. Furthermore, Haynie, Silver, and Teasdale (2006) found that neighborhood disadvantage indirectly influenced adolescent youth violence by providing opportunities to get involved in violent peer networks, which may also explain our findings with respect to neighborhood ties. It is unclear why youth residing in neighborhoods attitudes consistent with anomie or normlessness regarding laws and goal orientation would commit fewer, rather than greater, incidents of violence. Future studies might include measures of individual-level strain or anomie as well as neighborhood anomie to examine the influence of one versus the other in predicting violent behavior.

To address the second hypothesis, we examined the interaction effects between the significant DSM-oriented ODP measure and other significant measures across the models. This was done to explore the dependent nature of DSM-oriented problems on sociodemographic, peer, family, and neighborhood factors. The individual-level and cross-level interactions for the violence prevalence model indicated no dependence of DSM-oriented ODP problems on any variables included in the models. There were, however, two significant interaction effects for the violence frequency model. For violence frequency, the influence of ODP on violence depended on delinquent peer associations and family member anxiety. As mentioned above, having a family member with anxiety may serve as a protective factor for youths with ODP and reduce their likelihood of future violent behavior. On the other hand, having delinquent peers exacerbates the impact of ODP on violent offending. It may be that delinquent peer association increases defiant and incorrigible attitudes of youth.

From a public policy perspective, the multidimensional component of our study has important practical and policy-oriented implications for intervention and prevention methods calling for more holistic treatment strategies. In conjunction with other studies, our work contributes insights into which mental health difficulties may be most problematic among urban adolescents when considering comorbidity and how other individual, family, peer, and neighborhood risk factors affect the magnitude of mental health consequences. The value of using dimensional DSM-oriented problem indicators was shown in Figure 1 with the examination of the predicted frequencies for violence in normal, borderline, and clinical thresholds of ODP. The frequency of violence increased approximately 10% between the maximum normal ODP score and the maximum borderline ODP score and approximately 22% between the maximum borderline ODP score and the maximum clinical (observed) ODP score. A continuous indicator of mental health problems that captures borderline problems areas may be particularly relevant from a practical perspective when studying covariates of mental health problems. Such findings highlight the salience of borderline and subclinical scores that are ignored in categorical testing. Still, some caution is urged when considering the present findings, as replication with other data sets and populations is needed.

This study has a number of additional limitations worth mention. Namely, we were limited to using only those measures that were available with the Community Survey rather than the crime and census measures that are available to the principal research team. The restricted data measures did not allow for a fully specified model that considers variables such as crime rates, poverty, racial segregation, and residential mobility that might be significantly related to youth violence. Because the publicly available data did not allow for the examination of crime or census indicators, we were unable to explore neighborhood factors as they may uniquely contribute to violence. Thus, it is possible that our findings are picking up on the neighborhood clusters that have higher crime rates and concentrated disadvantage issues. Furthermore, there were some limitations related to the measures. The violence measures capture relatively minor physical forms of aggression. Because of the nonforensic nature of the sample, incidents of more severe forms of violence among the youth were quite rare. This study should be replicated with samples of more criminally involved youth as well as similar populations. Many of the composite indicators possessed weak internal consistency, as measured by Cronbach’s alpha scores. Weak internal consistency may have affected the validity of these findings and deserves replication to confirm our results. Finally, although the focus on DSM-oriented scales in particular allowed for an innovative method of exploring mental health issues, these measures were unable to account for other forms of serious psychopathology empirically linked to aggression and violence, such as substance use. In addition, other forms of risk and protective factors that may be contributing to violence were not included here. Future research should examine the impact of additional comorbid, risk, and protective indicators on violence.

Footnotes

Authors’ Note:

The authors wish to thank the National Institute of Justice (NIJ) for funding a secondary data analysis grant (2009-IJ-CX-0007) for this project and the researchers from the Project on Human Development in Chicago Neighborhoods (PHDCN) and the Inter-University Consortium for Political and Social Research for access to these data. The findings and opinions expressed in this article do not imply any policy or research endorsement by NIJ or PHDCN. Any mistakes or shortcomings are those of the authors.

Jennifer Wareham is an associate professor in the Department of Criminal Justice at Wayne State University. Her research specializes in understanding the causes and correlates of delinquency and assessing the efficacy of delinquency prevention.

Denise Paquette Boots is an associate professor in the Program in Criminology at the University of Texas at Dallas. Her recent research focuses on life-course criminology, mental health and youth violence, and domestic violence policy.

References

Achenbach

T. M.

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

Achenbach

T. M.

(1992). Manual for the Child Behavior Checklist/2-3 and 1992 profile. Burlington: University of Vermont, Department of Psychiatry.

Achenbach

T. M.

Edelbrock

C. S.

(1983). Manual for the Child Behavior Checklist and Revised Child Behavior Profile. Burlington: University of Vermont, Department of Psychiatry.

Achenbach

T. M.

McConaughy

S. H.

Howell

C. T.

(1987). Child/adolescent behavioral and emotional problems: Implications of cross-informant correlations for situational specificity. Psychological Bulletin, 101, 213-232.

Achenbach

T. M.

Rescorla

L. A.

(2001). Manual for the ASEBA School-Age Forms and Profiles: An integrated system for multi-informant assessment. Burlington: University of Vermont.

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text revision). Washington, DC: Author.

Bailey

B. N.

Delaney-Black

Hannigan

J. H.

Ager

Sokol

R. J.

Covington

C. Y.

(2005). Somatic complaints in children and community violence exposure. Developmental and Behavioral Pediatrics, 26, 341-348.

Biederman

Mick

Faraone

S. V.

Burback

(2001). Patterns of remission and symptom decline in conduct disorder: A four-year prospective study of an ADHD sample. Journal of the American Academy of Child and Adolescent Psychiatry, 40, 290-298.

Birmaher

Ryan

N. D.

Williamson

D. E.

Brent

D. A.

Kaufman

Daul

R. E.

. . . Nelson

(1996). Childhood and adolescent depression: A review of the past 10 years. Part I. Journal of the American Academy of Child and Adolescent Psychiatry, 29, 914-918.

10.

Boots

D. P.

Wareham

(2009). An exploration of DSM-oriented scales in the prediction of criminal offending among urban American youth. Criminal Justice and Behavior, 36, 840-860.

11.

Boots

D. P.

Wareham

(2010). Does controlling for comorbidity matter? DSM-oriented scales and violent offending in Chicago Youth. Aggressive Behavior, 35, 1-17.

12.

Boyd

C. P.

Gullone

Needleman

G. L.

Burt

(1997). The Family Environment Scale: Reliability and normative data for an adolescent sample. Family Process, 36, 369-373.

13.

Brezina

(1998). Adolescent maltreatment and delinquency: The question of intervening processes. Journal of Research in Crime and Delinquency, 35, 71-99.

14.

Broidy

L. M.

Nagin

D. S.

Tremblay

R. E.

Bates

J. E.

Brame

Dodge

K. A.

. . . Vitaro

(2003). Developmental trajectories of childhood disruptive behaviors and adolescent delinquency: A six-site, cross-national study. Developmental Psychology, 39, 222-245.

15.

Capaldi

D. M.

(1992). Co-occurrence of conduct problems and depressive symptoms in early adolescent boys: II. A 2-year follow-up at Grade 8. Development and Psychopathology, 4, 125-144.

16.

Cocozza

(2002). Youth with mental health disorders in the juvenile justice system: National trends, critical issues and emerging responses. Juvenile justice and mental health: Proceedings of the Robert Lee Sutherland Seminar XII. Austin, TX: Hogg Foundation for Mental Health.

17.

Colins

Vermeiren

Schuyten

Broekaert

(2009). Psychiatric disorders in property, violent, and versatile offending detained male adolescents. American Journal of Orthopsychiatry, 79, 31-38.

18.

Connor

D. F.

(2002). Aggression and antisocial behavior in children and adolescents: Research and treatment. New York, NY: Guilford.

19.

Copeland

Shanahan

Costello

E. J.

Angold

(2011). Cumulative prevalence of psychiatric disorders by young adulthood: A prospective cohort analysis from the Great Smoky Mountains Study. Journal of the American Academy of Child and Adolescent Psychiatry, 50, 252-261.

20.

Costello

E. J.

Mustillo

Erkanli

Keeler

Angold

(2003). Prevalence and development of psychiatric disorders in childhood and adolescence. Archives of General Psychiatry, 60, 837-844.

21.

Cottle

C. C.

Lee

R. J.

Heilbrun

(2001). The prediction of criminal recidivism in juveniles: A meta-analysis. Criminal Justice and Behavior, 28, 367-394.

22.

Crofford

L. J.

(2007). Violence, stress, and somatic complaints. Trauma, Violence, & Abuse, 8, 299-313.

23.

Dodge

K. A.

Price

J. M.

Coie

J. D.

Christopoulos

(1990). On the development of aggressive dyadic relationships in boys’ peer groups. Human Development, 33, 260-270.

24.

Dubrow

N. F.

Garbarino

(1989). Living in the war zone: Mothers and young children in a public housing development. Child Welfare, 58, 3-20.

25.

Earls

F. J.

Visher

C. A.

(1997). Project on Human Development in Chicago Neighborhoods: A research update (NIJ research in brief; NCJ 163603). Washington, DC: U.S. Department of Justice, Office of Justice Programs, National Institute of Justice.

26.

Farrington

(2005). Integrated developmental & life-course theories of offending: Advances in criminological theory (Vol. 14). New Brunswick, NJ: Transaction.

27.

Farrington

D. P.

Coid

J. W.

Murray

(2009). Family factors in the intergenerational transmission of offending. Criminal Behaviour and Mental Health, 19, 109-124.

28.

Farrington

D. P.

Jolliffe

Loeber

Stouthamer-Loeber

Kalb

L. M.

(2001). The concentration of offenders in families, and family criminality in the prediction of boys’ delinquency. Journal of Adolescence, 24, 579-596.

29.

Fergusson

D. M.

Horwood

L. J.

(1995). Predictive validity of categorically and dimensionally scored measures of disruptive childhood behaviors. Journal of the American Academy of Child and Adolescent Psychiatry, 34, 477-487.

30.

Goldstein

A. L.

Walton

M. A.

Cunningham

R. M.

Trowbridge

M. J.

Maio

R. F.

(2007). Violence and substance use as risk factors for depressive symptoms among adolescents in an urban emergency department. Journal of Adolescent Health, 40, 276-279.

31.

Goodman

S. H.

Rouse

M. H.

Connell

A. M.

Broth

M. R.

Hall

C. M.

Heyward

(2011). Maternal depression and child psychopathology: A meta-analytic review. Clinical Child and Family Psychology Review, 14, 1-27.

32.

Grunwald

H. E.

Lockwood

Harris

P. W.

Mennis

(2010). Influences of neighborhood context, individual history and parenting behavior on recidivism among juvenile offenders. Journal of Youth and Adolescence, 39, 1067-1079.

33.

Haynie

D. L.

(2001). Delinquent peers revisited: Does network structure matter? American Journal of Sociology, 106, 1013-1057.

34.

Haynie

D. L.

Silver

Teasdale

(2006). Neighborhood characteristics, peer networks, and adolescent violence. Journal of Quantitative Criminology, 22, 147-169.

35.

Hoeve

Dubas

J. S.

Eichelsheim

van der Laan

P. H.

Smeenk

Gerris

J. R. M.

(2009). The relationship between parenting and delinquency: A meta-analysis. Journal of Abnormal Child Psychology, 37, 749-775.

36.

Hox

(2002). Multilevel analysis: Techniques and applications. Mahwah, NJ: Lawrence Erlbaum.

37.

Huizinga

Esbenson

Weihar

A. W.

(1991). Are there multiple paths to delinquency? Journal of Criminal Law and Criminology, 82, 83-118.

38.

Jaffee

S. R.

Harrington

Cohen

Moffitt

T. E.

(2005). Cumulative prevalence of psychiatric disorder in youths. Journal of the American Academy of Child and Adolescent Psychiatry, 44, 406-407.

39.

Janca

Bucholz

Janca

(1992). Family history and assessment module. St. Louis, MO: Washington University of Medicine.

40.

Jencks

Mayer

(1990). The social consequences of growing up in a poor neighborhood. In Lynn

L. E.

McGeary

M. F. H.

(Eds.), Inner-city poverty in the United States (pp. 111-186). Washington, DC: National Academy Press.

41.

Kim-Cohen

Caspi

Moffitt

T. E.

Harrington

Milne

B. J.

Poulton

(2003). Prior juvenile diagnoses in adults with mental disorder: Developmental follow-back of a prospective-longitudinal cohort. Archives of General Psychiatry, 60, 709-717.

42.

Knecht

Snijders

T. A. B.

Baerveldt

Steglich

C. E. G.

Raub

(2010). Friendship and delinquency: Selection and influence processes in early adolescence. Social Development, 19, 494-514.

43.

Kreager

D. A.

(2004). Strangers in the halls: Isolation and delinquency in school networks. Social Forces, 83, 351-390.

44.

Kupersmidt

J. B.

Coie

J. D.

(1990). Preadolescent peer status, aggression, and school adjustment as predictors of externalizing problems in adolescence. Child Development, 61, 1350-1362.

45.

Langbehn

D. R.

Cadoret

R. J.

Yates

W. R.

Troughton

E. P.

Stewart

M. A.

(1998). Distinct contributions of conduct and oppositional defiant symptoms to adult antisocial behavior. Archives of General Psychiatry, 55, 821-829.

46.

Lansford

J. E.

Miller-Johnson

Berlin

L. J.

Dodge

K. A.

Bates

J. E.

Pettit

G. S.

(2007). Early physical abuse and later violent delinquency: A prospective longitudinal study. Child Maltreatment, 12, 233-245.

47.

Maas

Herrenkohl

Sousa

(2008). Review of research on child maltreatment and violence in youth. Trauma, Violence, & Abuse, 9, 56-67.

48.

Markon

K. E.

Chmielewski

Miller

C. J.

(2011). The reliability and validity of discrete and continuous measures of psychopathology: A quantitative review. Psychological Bulletin, 137, 856-879.

49.

Markowitz

F. E.

Bellair

P. E.

Liska

A. E.

Liu

(2001). Extending social disorganization theory: Modeling the relationships between cohesion, disorder, and fear. Criminology, 39, 293-319.

50.

Moffitt

T. E.

(1990). Juvenile delinquency and attention deficit disorder: Boys’ developmental trajectories from age 13 to age 15. Child Development, 61, 893-910.

51.

Moffitt

T. E.

Silva

P. A.

(1988). Self-reported delinquency, neuro-psychological deficit, and history of attention deficit disorder. Journal of Abnormal Child Psychology, 16, 553-569.

52.

Molnar

B. E.

Cerda

Roberts

A. L.

Buka

S. L.

(2008). Effects of neighborhood resources on aggressive and delinquent behaviors among urban youths. American Journal of Public Health, 98, 1086-1093.

53.

Moos

(1994). Family Environment Scale manual: Development, applications, research (3rd ed.). Palo Alto, CA: Consulting Psychologist Press.

54.

Morenoff

J. D.

Sampson

R. J.

Raudenbush

S. W.

(2001). Neighborhood inequality, collective efficacy, and the spatial dynamics of homicide. Criminology, 39, 517-560.

55.

Nigg

J. T.

Hinshaw

S. P.

(1998). Parent personality traits and psychopathology associated with antisocial behaviors in childhood attention-deficit hyperactivity disorder. Journal of Child Psychology and Psychiatry, 39, 145-159.

56.

Pfiffner

L. J.

McBurnett

Rathouz

P. J.

Judice

(2005). Family correlates of oppositional and conduct disorders in children with attention deficit/hyperactivity disorder. Journal of Abnormal Child Psychology, 33, 551-563.

57.

Plattner

Steiner

The

S. S. L.

Kraemer

H. C.

Bauer

S. M.

Kindler

. . . Feucht

(2009). Sex-specific predictors of criminal recidivism in a representative sample of incarcerated youth. Comprehensive Psychiatry, 50, 400-407.

58.

Puig-Antich

(1982). Major depression and conduct disorder in prepuberty. Journal of the American Academy of Child and Adolescent Psychiatry, 21, 118-128.

59.

Rapport

M. D.

Chung

K.-M.

(2000). Attention deficit hyperactivity disorder. In Hersen

Ammerman

R. T.

(Eds.), Advanced abnormal child psychology (pp. 413-440). Mahwah, NJ: Lawrence Erlbaum.

60.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage.

61.

Raudenbush

S. W.

Sampson

R. J.

(1999). Econometrics: Toward a science of assessing ecological settings, with application to the systematic social observation of neighborhoods. Sociological Methodology, 29, 1-41.

62.

Rebellon

(2006). Do adolescents engage in delinquency to get the social attention of peers? An extension and longitudinal test of the social reinforcement hypothesis. Journal of Research in Crime and Delinquency, 43, 387-411.

63.

Reiss

A. J.

(1961). Occupations and social status. New York, NY: Free Press.

64.

Rountree

P. W.

Warner

B. D.

(1999). Social ties and crime: Is the relationship gendered? Criminology, 37, 789-813.

65.

Sampson

R. J.

Groves

W. B.

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

66.

Sampson

R. J.

Raudenbush

S. W.

(1999). Systematic social observation of public spaces: A new look at disorder in urban neighborhoods. American Journal of Sociology, 105, 603-651.

67.

Skeer

M. R.

McCormick

M. C.

Normand

S.-L. T.

Mimiaga

M. J.

Buka

S. L.

Gilman

S. E.

(2011). Gender differences in the association between family conflict and adolescent substance use disorders. Journal of Adolescent Health, 49, 187-192.

68.

Sorra

J. S.

Dyer

(2010). Multilevel psychometric properties of the AHRQ hospital survey on patient safety culture. BMC Health Services Research, 10, 199-211.

69.

Stouthamer-Loeber

Loeber

Homish

D. L.

Wei

(2001). Maltreatment of boys and the development of disruptive and delinquent behavior. Development and Psychopathology, 13, 941-955.

70.

Straus

M. A.

Hamby

S. L.

Finkelhor

Moore

D. W.

Runyan

(1998). Identification of child maltreatment with Parent–Child Conflict Tactics Scale. Development and psychometric data for a national sample of American parents. Child Abuse & Neglect, 22, 249-270.

71.

Teplin

L. A.

(2001). Assessing alcohol, drug, and mental disorders in juvenile detainees (Office of Juvenile Justice and Delinquency Fact Sheet #02). Washington, DC: U.S. Department of Justice.

72.

Teplin

L. A.

Abram

K. M.

McClelland

G. M.

Mericle

A. A.

Dulcan

M. K.

Washburn

J. J.

(2006). Psychiatric disorders of youth in detention. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

73.

Turner

R. J.

Frankel

Levin

(1983). Social support: Conceptualization, measurement, and implications for mental health. In Greenley

J. R.

(Ed.), Research in community and mental health (Vol. 3, pp. 67-111). Greenwich, CT: JAI.

74.

Turner

R. J.

Grindstaff

C. F.

Phillips

(1990). Social support outcome in teenage pregnancy. Journal of Health and Social Behavior, 31, 43-57.

75.

van de Rakt

Nieuwbeerta

Apel

(2009). Association of criminal convictions between family members: Effects of siblings, fathers, and mothers. Criminal Behaviour and Mental Health, 19, 94-108.

76.

Walker

J. L.

Lahey

B. B.

Russo

M. F.

Frick

P. J.

Christ

M. A. G.

McBurnett

. . . Green

S. M.

(1991). Anxiety, inhibition, and conduct disorder in children: I. Relations to social impairment. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 187-191.

77.

Warner

B. D.

Rountree

P. W.

(1997). Local social ties in a community and crime model: Questioning the systemic nature of informal social control. Social Problems, 44, 520-536.

78.

Warr

Stafford

M. C.

(1991). The influence of delinquent peers: What they think or what they do. Criminology, 29, 851-865.

79.

Weissman

M. M.

Wickramaratne

Nomura

Warner

Pilowsky

Verdeli

(2006). Offspring of depressed parents: 20 year later. American Journal of Psychiatry, 163, 1001-1008.

80.

Widom

C. S.

Maxfield

M. G.

(2001). An update on the “cycle of violence.” Washington, DC: U.S. Department of Justice, Office of Justice Programs.

81.

Yun

Ball

J. D.

Lim

(2011). Disentangling the relationship between child maltreatment and violent delinquency: Using a nationally representative sample. Journal of Interpersonal Violence, 26, 88-110.