Neighborhood Structure,Immigrant Status,and Youth Violence

Abstract

Research indicates that children of immigrants are less likely to engage in violence than children of native-born parents, even when they live in high-risk neighborhoods, suggesting that foreign-born parents employ strategies that buffer children from delinquency. Parental supervision is important for adolescent well-being, and some scholars suggest it is especially important for adolescents residing in disadvantaged communities. Others argue supervision is more critical for youth residing in advantaged contexts, where parental involvement is normative. To date, evidence on the interplay between supervision and neighborhood characteristics is mixed, suggesting a more complex relationship. This study considers whether immigrant status further conditions the interplay between supervision and neighborhood characteristics, using data from a sample of adolescents residing in Chicago neighborhoods. Findings indicate that less supervised, first-generation adolescents are more likely to perpetrate violence in low-risk neighborhoods, while less supervised, second- and third-generation adolescents are more likely to perpetrate violence in high-risk settings.

Keywords

immigration adolescents violence supervision neighborhood effects

Introduction

Prior research offers mixed evidence on the ways that family contexts mediate (and/or moderate) the effects of neighborhood structural characteristics on youth outcomes. Although effective parenting is believed to have a uniformly positive effect on child and adolescent well-being, some scholars suggest that parenting strategies are of greatest import in high-risk communities, where they serve as a buffer from the risks associated with residence in a disadvantaged area (Rankin & Quane, 2002; Sampson, 1992). Other scholars, however, argue that parenting deficits are more critical to the well-being of youth growing up in low-risk, more advantaged contexts, in which involved parents are the norm (Bámaca, Umaña-Taylor, Shin, & Alfaro, 2005). In these settings, adolescents with uninvolved parents may be more likely to engage in problem behavior, including violence, because they interpret parenting deficits as lack of care. In contrast, adolescents residing in high-risk, disadvantaged communities may view a lack of parental involvement as inevitable, given the many life challenges facing parents in these settings.

Yet, the interplay between parenting and neighborhood characteristics may be further conditioned by immigrant status. Immigrant family and neighborhood contexts have been shown to be protective for adolescents, findings that some scholars have suggested are paradoxical, since immigrant families and neighborhoods more often experience above-average levels of economic disadvantage (Markides & Coreil, 1986; Sampson, 2008). Researchers suspect that family social processes may explain part of this advantage (Marsiglia, Parsai, & Kulis, 2009); however, the precise mechanisms remain uncertain, perhaps because the immigrant experience shapes the meaning of family social processes. Immigrant parents are thought to place a higher value on the family, yet many are operating under constraints as they strive to make their way while residing in high-risk, economically disadvantaged communities (Horowitz, 1983; Marsiglia et al., 2009). Thus, immigrant status may alter expectations for the nature and extent of parenting, with implications for the relationship between parenting strategies and violence among children of foreign-born (vs. native-born) parents.

This study examines the interplay between parental supervision and neighborhood structural characteristics. Is the relationship between parental supervision and individual adolescent violence mediated and/or moderated by neighborhood structural characteristics? Second, to what extent does the interplay between parental supervision and neighborhood structure differ by the adolescent’s immigrant status? Children of foreign-born parents may differ from children of U.S.-born parents in their expectations for the level of family social support in ways that further condition the relationships between and among parental supervision, neighborhood structure, and adolescent violence.

Family Social Processes and Youth Violence

Scholars have long identified the importance of family structure and social processes as predictors of adolescent delinquent and problem behavior outcomes. Children who are raised in intact families, with close, supportive family relationships, and who are subject to parental monitoring/supervision have better outcomes, on average, than children who grow up in less advantageous developmental circumstances (Farrington, 2002; Loeber & Stouthamer-Loeber, 1986; Smith & Stern, 1997). These effects are enduring, as early life family deficits have been found to be associated with problem behavior well into adulthood (McCord, 1979; Sampson & Laub, 1993).

Much of the research on family factors has drawn on individual-level theories of social control, which specify that adolescents who maintain attachments to conventional parents and other prosocial individuals refrain from crime and delinquency because they are concerned about the social costs of violating conventional normative expectations (Hirschi, 1969; Nye, 1958; Reiss, 1951). From a social control perspective, parental supervision is viewed as having both direct and indirect social control effects on adolescent behavior. Parents who monitor their children directly inhibit delinquency by keeping a watchful eye on children and adolescents, thereby limiting opportunities for delinquent behavior. Yet the practice of regular monitoring has indirect effects that are influential even when the adolescent is out of sight; adolescents are more likely to refrain from behaviors that would not meet with their parents’ approval (Hirschi, 1969). The lasting impact of monitoring may help to explain why parental supervision has the strongest and most consistent negative association with delinquency of the family social processes that scholars typically examine (Farrington & Loeber, 1999; Smith & Stern, 1997).

Immigrant Status, Delinquency, and Family Social Processes

The role of the family has also been pinpointed as a potential key to understanding the seemingly negative relationship between immigration and crime. Immigrants are often popularly depicted as crime prone (while immigrant communities are depicted as crime ridden) (Martinez, 2002). In the United States, these beliefs have stoked antipathy toward immigrants (both undocumented and documented) and have led to calls for more restrictive immigration and border-control policies, despite a lack of empirical evidence linking immigration to crime (Martinez, 2002; Sampson, 2008). To address this gap between opinion and evidence, scholars have increasingly turned their attention to empirically examining the association between immigrant status and crime. Across a wide range of studies (Butcher & Piehl, 1999; Harris, 1999; Martinez, 2002; Sampson, 2008; Xie & Greenman, 2005), the findings are, for the most part, consistent and counter to popular belief: The likelihood of individual delinquency is lower among immigrant youth, and the protective nature of immigrant status is progressively eroded with each succeeding generation. Moreover, at a macro level, immigrant communities are protective for residents’ health, well-being, and involvement in crime.

The importance of the association between family influences and delinquency has led scholars to propose that the association may be explained, at least in part, by considering the ways that immigrant families differ from native-born families (Estrada-Martínez, Padilla, Caldwell, & Schulz, 2011; Marsiglia et al., 2009). The aspirations that underlie parents’ decisions to immigrate mean that their children’s activities are of paramount importance, and immigrant parents may shepherd their children carefully via supervision/monitoring, with an emphasis on attainment and the avoidance of trouble (Germán, Gonzales, & Dumka, 2009; Harker, 2001; Portes & Fernández-Kelly, 2008).

More traditional family structures may also play a role; observed ethnic differences in adolescent violence were explained by individual immigrant status, neighborhood immigrant concentration, and a variable measuring whether the adolescent’s parents were married (Sampson, Morenoff, & Raudenbush, 2005). Further, familism—a term encompassing a set of values that treat family relationships, and shared family goals, as being of central importance—prevails in many immigrant families (Cuéllar, Arnold, & González, 1995; Portes & Rumbaut, 2001; Updegraff, McHale, Whiteman, Thayer, & Delgado, 2005). Indicators of familism and family support/cohesion have been shown to have a protective effect on conduct problems among Mexican-origin youth (Germán et al., 2009; Marsiglia et al., 2009).

Family Social Processes in Context

For all families, parental supervision and monitoring are premised on the notion that there are other, social contextual threats to successful child development that could prevail if children and adolescents are left untended or without guidance (Furstenberg, Cook, Eccles, Elder, & Sameroff, 1999; Sampson, 1992). Thus, it is important to consider family processes in context. Ecological approaches to the study of human development emphasize the necessity of considering the various social contexts, such as the family and the residential neighborhood, that shape child and adolescent behavioral outcomes (Booth & Crouter, 2001; Bronfenbrenner, 1979; Cook, 2003; Elliott et al., 2006; Leventhal & Brooks-Gunn, 2000; Steinberg, Mounts, Lamborn, & Dornbusch, 1991).

Certainly, in recent decades, there has been a surge in research considering the role of the residential neighborhood in child and adolescent outcomes (Leventhal & Brooks-Gunn, 2000; Sampson, Morenoff, & Gannon-Rowley, 2002). Youth delinquency is shaped in part by sociospatial determinants, as identified by Shaw and McKay (1942) and specified in their articulation of the classic theory of social disorganization. Building on earlier insights from the Chicago School of sociology (Thomas & Znaniecki, 1984 [1918-1920]), Shaw and McKay (1942) determined that youth delinquency tended to occur in the most economically disadvantaged neighborhoods in Chicago, those that saw a substantial “churn” of diverse residents due to migration.

In recent years, scholars have built on the Chicago School tradition and drawn on advances in statistical methods and survey data that permit them to simultaneously consider individual- and neighborhood-level determinants of individual outcomes while controlling for issues of neighborhood selection and clustering of survey respondents (Sampson et al., 2002). Many such studies have examined neighborhood effects on adolescent violence; these studies suggest a mixed (positive or null) association between neighborhood economic disadvantage and individual youth violence and a consistent, negative association between neighborhood immigrant concentration and individual youth violence (Desmond & Kubrin, 2009; Maimon & Browning, 2010; Sampson et al., 2005; Zimmerman & Messner, 2010).

These findings indicate that neighborhoods with higher levels of economic disadvantage are higher risk contexts for adolescents, while neighborhoods with greater proportions of immigrant residents are lower risk contexts for youth. Ecological theories of child development suggest that family social processes may be of greater import in high-risk contexts such as disadvantaged neighborhoods than in low-risk contexts, as they may serve as buffers, protecting youth from delinquent influences in the immediate environment (Furstenberg et al., 1999; Sampson, 1992). Although both family and neighborhood determinants are associated with adolescent violence, however, there are surprisingly few studies that examine the interactive effects of family and social contextual determinants on serious youth delinquency.

Several qualitative studies document the strategies families use to protect youth from threats in disadvantaged neighborhoods, including careful monitoring and resource seeking (Anderson, 1999; Jarrett, 1997; MacLeod, 2008; Smith, 2008). Using survey data from the Chicago Youth Development Study, Gorman-Smith, Henry, and Tolan (2004) found that family functioning moderated the impact of neighborhood exposure to violence on violence perpetration among minority male adolescents. Similarly, in a quantitative study using survey data from mothers and their adolescent children residing in poor and mixed-income Chicago neighborhoods (but limited to Black families), Rankin and Quane (2002) found evidence that the negative effect of parental monitoring on adolescent problem behavior was greater in high-risk neighborhoods (those with lower levels of collective efficacy).

Yet other studies indicate that supervision is not more protective for youth residing in high-risk contexts than it is for those residing in lower risk settings. Levine-Coley and Hoffman (1996) found that well-supervised children who resided in high-crime neighborhoods received lower teacher ratings on behavior than did less supervised children who resided in low-crime areas (but noted that their correlational analyses limited their ability to draw conclusions about causal direction). Furstenberg, Cook, Eccles, Elder, and Sameroff (1999) found somewhat mixed results in a study of adolescents and their families residing in Philadelphia. Their study highlights the importance of cumulative advantage. Youth from effective families who resided in lower risk contexts had the best outcomes of the youth in the study. Effective families buffered youth in high-risk contexts but to a lesser extent than those in low-risk contexts. Further, youth in low-risk contexts with ineffective families were unable to take advantage of the resources in their local environments and had poor outcomes as a result.

Finally, in a more recent study (limited to a sample of Latino adolescents residing in the Midwest), Bámaca, Umaña-Taylor, Shin, and Alfaro (2005) found that greater maternal monitoring was associated with higher self-esteem for girls residing in lower risk settings but was not related to self-esteem for girls in higher risk neighborhoods. The authors noted the possibility that deficits in family functioning are more critical to adolescents residing in neighborhoods in which well-functioning families are the norm—in other words, low-risk contexts (Bámaca et al., 2005). In these settings, adolescents may more keenly perceive the gap in effective family social processes, as the contrast between their circumstances and those of their peers is one of relative disadvantage (potentially explaining why the authors found an effect primarily for self-esteem, and only among girls, who are typically more supervised than boys). On the other side of the coin, the authors argued, in high-risk, disadvantaged neighborhoods, adolescents may have lower expectations for the level of supervision that parents can provide, given the many stressors in their parents’ lives. As a result, deficits in monitoring are less critical than other factors in determining adolescent well-being.

Immigrant Youth, Family Supervision, and Neighborhood Contexts

Comparisons may also alter the ways that family social processes are associated with problem behavior among immigrant youth living in lower and higher risk contexts. The path to acculturation is one of shifting sands, for both first- and second-generation immigrant adolescents and their foreign-born parents (Marsiglia et al., 2009; Portes & Rumbaut, 2001). On average, immigrant families are more likely to reside in economically disadvantaged neighborhoods (Sampson, 2008), and it may be that immigrant youth in these contexts (particularly first-generation adolescents) avoid delinquency even when they are not well supervised by parents, as an acknowledgment of the challenges their parents face, and/or out of concern that poor behavior will subvert family goals of attainment. For example, Morenoff and Astor (2006) found a puzzling negative association between neighborhood disadvantage and violence but only for first-generation immigrant adolescents. For second-generation youth, however, the picture may differ. Scholars have noted two concepts relevant to their adjustment: segmented assimilation and acculturative dissonance. First, segmented assimilation theory emphasizes the match between an individual and the characteristics of the context into which the individual is assimilating (Portes & Zhou, 1993). If assimilation occurs in a high-risk context, it is thought that (absent other protective factors, such as a local immigrant population) the adolescent will be more likely to take on behaviors, including violence, common in the local milieu. For second-generation youth residing in disadvantaged neighborhoods, parental supervision may be of even greater import than it is for first-generation adolescents in these same contexts.

The second important factor unique to immigrant families is that of acculturative dissonance, a form of generational conflict stemming from gaps in acculturation between a foreign-born parent and child that can highlight their very different ideas regarding norms for behavior and the extent of “free rein” adolescents should receive. In such cases, high levels of parental monitoring may be viewed by adolescents as excessively restrictive and redolent of an “Old World” that they must cast off to assimilate into the local context (Portes & Rumbaut, 2001; Waters, 1999; Le & Stockdale, 2008).

Current Study

This study builds on prior research by considering the interrelationships between and among immigrant status, parental supervision, and neighborhood structural context in predicting a violence outcome for a sample of adolescents residing in Chicago neighborhoods. It considers the following research questions:

Research Question 1: Are there significant mean differences in parental supervision by immigrant generation? It is expected that there will be significant mean differences in supervision by immigrant generation, with the highest mean among first-generation adolescents and the next highest mean among second-generation youth.

Research Question 2: Does the relationship between parental supervision and violence differ by immigrant generation, net of controls? As most of the research and theory regarding the immigrant experience emphasizes the importance of the local context, the expectation is that supervision has a similar association with violence across immigrant generations when immigrant status and supervision are not further conditioned by neighborhood poverty and immigrant concentration.

Research Question 3: Does the association between parental supervision and violence vary by neighborhood structural characteristics, net of controls? Consistent with prior research, the relationship between supervision and violence is expected to be moderated by neighborhood poverty and immigrant concentration; however, given mixed findings on the nature of the moderating effect, it is unclear whether supervision is more protective in a high- (vs. low-) risk neighborhood.

Research Question 4: Do parental supervision, immigrant status, and neighborhood structural characteristics interact in their associations with adolescent violence, net of controls? Drawing on theory and prior research, the expectation is that the magnitude and direction of the association between supervision and violence will be conditioned by both immigrant status and neighborhood structural characteristics. Again, however, the likely direction of the association is unclear from prior research and theory.

Data and Method

Data

This study uses multi-level data from Waves 1 and 2 of the Project on Human Development in Chicago Neighborhoods (PHDCN; Earls, Raudenbush, Reiss, & Sampson, 1995) linked to data from the 1990 U.S. Decennial Census. The PHDCN sampling design began by combining the 847 census tracts in Chicago (as of the 1990 Census) into 343 neighborhood clusters (NCs) of approximately 8,000 people each. (Census tracts combined into NCs are contiguous.) The NCs were drawn to be homogeneous in terms of residents’ socioeconomic and family structure characteristics, race/ethnicity, and housing and are based on natural boundaries, such as freeways and railroad tracks. Definition of the NCs was also guided by knowledge of Chicago’s neighborhoods (Sampson, Raudenbush, & Earls, 1997).

The 343 NCs formed the basis for a two-stage sampling design in which 80 of the 343 NCs were selected on the basis of a cross-classification scheme that matched seven categories of ethnic composition with three categories (high, medium, and low) of socioeconomic status (SES). This resulted in 21 strata (e.g., “medium-income primarily Latino”). Three cross-classified categories were not observed, namely low-income White, high-income primarily Latino, and high-income Latino/African American, reflecting the consequences of long-term residential segregation in Chicago. Nonetheless, the observed strata show considerable variation in the socioeconomic circumstances in which minorities reside (Raudenbush, Johnson, & Sampson, 2003; Sampson et al., 1997).

Households were randomly selected within each of the 80 NCs, and children within seven age categories (birth, 3, 6, 9, 12, 15, and 18) were sampled from these households. Survey questions and assessments were administered to the children and their primary caregivers in two waves of the PHDCN Longitudinal Cohort Study (PHDCN-LCS). Wave 1 was administered during 1995–1996, while Wave 2 was administered in 1998–1999. Individual-level measures used in the analyses are derived from child and primary caregiver responses in the PHDCN-LCS for the 9-, 12-, and 15-year-old cohorts. The violence outcome in this study is derived from Wave 2; all other individual variables are measured at Wave 1. Neighborhood measures of poverty and immigrant concentration are from the 1990 U.S. Decennial Census. The sample is limited to adolescents in the 9-, 12-, and 15-year-old cohorts who have valid data for both waves. To address missing data at Level 2 across independent variables (which just exceeded 15%), 10 multiply imputed individual/family-level (Level-2) files were generated using the ICE program for Stata version 9 (Royston, 2005; StataCorp, 2005) and were read into HLM version 7 software (Raudenbush et al., 2011) for analyses (the Level-1 outcome item file and Level-3 neighborhood files are single, nonimputed files). The final imputed sample consists of 1,908 adolescents from the 9-, 12- and 15-year-old cohorts, residing in 79 neighborhoods.

Dependent Variable

The dependent variable is a measure of the adolescent’s self-reported violence at Wave 2, based on five binary indicators of whether an adolescent committed the violent act in the year prior (1 = yes, 0 = no). These items are whether the adolescent hit someone he or she does not live with, threw objects at someone, attacked someone with a weapon, was in a gang fight, or carried a hidden weapon. Descriptive statistics for these components are set forth in Table 1. The descriptive statistics for all other variables in the models are set forth in Table 2 (by immigrant generation), which is discussed in greater detail in the Results section.

Table 1.

Descriptive Statistics for Components in the Violence Scale.

Violence Items: In last year,	All		First		Second		Third +		Min	Max
Violence Items: In last year,	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Min	Max
Hit someone do not live with	.18	.39	.08	.27	.11	.32	.25	.44	.00	1.00
Thrown objects at someone	.09	.29	.05	.22	.07	.25	.12	.33	.00	1.00
Attack with weapon	.03	.16	.01	.09	.01	.11	.04	.20	.00	1.00
Been in a gang fight	.04	.19	.03	.18	.04	.19	.04	.20	.00	1.00
Carried a hidden weapon (referent)	.06	.24	.04	.20	.06	.23	.07	.26	.00	1.00

Note. N = 1,908. Min = minimum; max = maximum.

Table 2.

Descriptive Statistics for Independent Variables in the Violence Analysis, by Immigrant Generation.^a

Immigrant Generation	First (.14)				Second (.33)			Third + (.53)		Min	Max
Independent Variables	Mean	Std. Dev.	vs. 2	vs. 3+	Mean	Std. Dev.	vs. 3+	Mean	Std. Dev.	Min	Max
Demographic characteristics
Race/ethnicity
Latino	0.80	—	***	***	0.87	—	***	0.14	—	0.00	1.00
African American	0.06	—	***	***	0.03	—	***	0.62	—	0.00	1.00
White/other (referent)	0.15	—	***	***	0.11	—	***	0.24	—	0.00	1.00
Male	0.53	—	**	**	0.49	—	—	0.50	—	0.00	1.00
Age	12.20	2.36	***	***	11.79	2.45	***	12.00	2.41	7.77	16.38
SES	−0.85	1.23	***	***	−0.42	1.32	***	0.29	1.40	−3.16	3.69
Family
Two biological parents	0.67	—	—	***	0.68	—	***	0.32	—	0.00	1.00
Family attachment and support	0.11	0.69	—	***	0.10	0.68	**	−0.07	0.77	−3.62	1.61
Supervision	−0.17	1.01	***	***	0.00	1.05	**	0.06	0.93	−3.84	3.48
Unstructured socializing	3.79	1.24	—	***	3.80	1.30	***	4.07	1.16	1.00	5.00
Individual characteristics
Verbal aptitude^b	92.68	20.98	***	***	98.47	19.21	***	97.20	18.36	39.81	156.00
Impulsivity	2.50	0.59	—	***	2.53	0.57	***	2.82	0.57	0.56	4.85
Prior problem behavior	−5.20	0.89	***	***	−5.04	0.95	***	−4.75	1.06	−7.81	−0.67
Neighborhood characteristics
Poverty^c	−0.04	0.61	**	***	−0.09	0.75	***	0.16	0.90	−3.07	1.51
Immigrant concentration^c	0.95	0.84	***	***	1.04	0.83	***	−0.10	0.90	−1.54	3.07

Note. SES = socioeconomic status. ^a N = 1,908 adolescents residing in 79 neighborhoods. ^bWide Range Achievement Test (Wilkinson 1993). ^cMeasure used in models is centered on the neighborhood-level mean.

*p < .05. **p < .01. ***p < .001 (Wald’s tests).

Independent Variables

Supervision

“Supervision” is measured at Wave 1 and is derived from a set of 24 items comprising the supervision subscale adapted from the Home Observation for Measurement of the Environment instrument (HOME; Caldwell & Bradley, 1984). The 24 items are dichotomous measures of the primary caregiver’s reports in response to questions such as whether the adolescent has a curfew on school nights, whether the adolescent is permitted to wander in public places without adult supervision for more than 2 hours, and whether the adolescent goes to a place supervised by adults after school. The measure is an empirical Bayes (EB) residual from a two-level Rasch (item-response) model. The reliability of the scale is .60.

Other family factors

To better isolate the relationship between the key family social process variable, supervision, and violence, the analyses include measures of other family factors that operate as social controls on behavior and are thus predicted to have a negative association with violence (Hirschi, 1969). The first of these is a self-reported measure of the extent to which the adolescent receives social support from his or her family. “Family support” is derived from a set of 5-item Likert-type responses at Wave 1 indicating the extent to which the adolescent agrees that “no matter what happens, my family will always be there,” “my family lets me know I am a valuable person,” “people in my family have confidence in me,” “people in my family help me find solutions to problems,” and “I know my family will always stand by me.” The reliability of the scale is .80. Also, in light of prior research on the mediating role of traditional family structure on the relationship between ethnicity and violence (Sampson et al., 2005), the analyses include a dichotomous measure indicating whether the adolescent resides with both biological parents (1 = yes).

Immigrant status

Binary variables indicate first- and second-generation immigrant status (with third generation and above the omitted referent). First generation is coded as “1” if the adolescent was born outside the United States, and second generation is coded as “1” if the adolescent was born in the United States but at least one parent is foreign born.

Other demographic characteristics

Control variables for other important individual- and family-level demographic characteristics include sex (1 = male), age (along with a squared term for age, to account for the known curvilinear relationship between age and crime; Gottfredson & Hirschi, 1990), and race/ethnicity (two binary variables indicating Latino/Latina and African American, with White/other as the omitted category). Family socioeconomic resources may offset the level of risk present in local contexts. Family SES is a measure comprised of the first principal component of income, occupational prestige, and educational attainment for the respondent’s primary caregiver.

Unstructured socializing

Drawing on routine activities perspectives that emphasize the role of delinquent opportunity (Cohen & Felson, 1979), studies have shown that youth who spend time socializing in the absence of authority figures have a greater likelihood of involvement in delinquent behaviors, including violence perpetration and substance use (Haynie & Osgood, 2005; Maimon & Browning, 2010; Osgood & Anderson, 2004). Importantly (and in contrast to measures of peer delinquency), unstructured socializing is positively associated with delinquency even for youth who associate with nondelinquent peers. The goal of controlling for unstructured socializing in these analyses is to isolate the more enduring aspects of parental supervision that serve as indirect social controls on adolescent behavior, above and beyond differences in delinquent opportunity. The measure of unstructured socializing is the adolescent’s response to the question, “How often do you hang out with friends?” Responses ranged from a value of 1 (never) to 5 (every day).

Prior problem behavior, impulsivity, and verbal aptitude

Additional controls include measures of an adolescent’s participation in prior problem behavior, impulsivity, and verbal aptitude. The prior problem behavior measure was constructed using a set of 29 dichotomous items from Wave 1 that capture both violent and nonviolent problem behavior, such as whether the adolescent has “used force to rob someone,” “stolen from a car,” or “damaged property” within the last year. The final measure is the EB residual from a two-level Rasch model. The reliability of the measure is .62. (Due to the very strong association between prior problem behavior and the violence outcome used in the analyses, its inclusion in statistical models tends to obscure other important associations. For this reason, the prior problem behavior control is included only in the very final model as a way to assess the stability of the other predictors.)

The measure of impulsivity is derived from a subscale of the Emotionality and Sociability Inventory (EASI; Buss & Plomin, 1984). The EASI instrument was administered to primary caregivers (and, in the case of older adolescents, the adolescents themselves) who indicated how characteristic each behavior was of the adolescent subject (from 1 = uncharacteristic, to 5 = characteristic). Item responses were summed and divided by the number of items to derive the scale. The reliability of the measure is .75. Verbal aptitude was assessed by the respondents’ performance on the reading component of the Wide Range Achievement Test (WRAT-3; Wilkinson, 1993), a standardized test that evaluates reading achievement. Adolescent respondents were permitted to take the WRAT-3 in languages other than English.

Neighborhood characteristics

Neighborhood measures are derived from the 1990 U.S. Decennial Census. Poverty is the proportion of residents in the neighborhood in poverty (the measure used in models is standardized by taking the natural log of the neighborhood proportion in poverty and then further transforming it into a z-score). Immigrant concentration is the first principal component derived from a factor analysis of two measures: the proportion of residents who were foreign born and the proportion of Hispanic residents. (The z-score and factor analyses were based on the full sample of 343 neighborhoods in the PHDCN Community Survey.)

Analytic Strategy

The analyses proceed in two stages. First, descriptive statistics are shown for the sample to determine the extent to which there are significant mean differences in supervision by immigrant generation (Research Question 1). Then, the analyses proceed to multivariate logistic regression to specifically examine the interplay between and among family supervision, immigrant generation, and neighborhood structural characteristics. These multi-level logistic regression models progressively address each of Research Questions 2 through 4, after first presenting a baseline model (one that includes parental supervision and immigrant status along with the other individual and neighborhood covariates but without interaction effects).

The multivariate analytic strategy fits a set of three-level Rasch models with random intercepts to the five dichotomous indicators of violent behavior (Raudenbush et al., 2003). For the baseline model, the equation at Level 1 models the log odds of a respondent in a given neighborhood endorsing an item on the violence scale, adjusted by indicators of the other violence items (with one omitted referent), and includes an error term indicating item severity. At Level 2, the equations model variation between individual respondents in violence perpetration. The main Level-2 equation thus includes the measures for immigrant generation and supervision, and the other individual-level demographic, individual, and family characteristics and social processes, and models each respondent’s adjusted violence score (the intercepts from the level-1 equation). Finally, the equations at Level 3 model adjusted intercepts from the Level-2 equation, which represents the violence score for each neighborhood, adjusted for the individual and family covariates included at Level 2. The main Level-3 equation includes the neighborhood predictors of interest in this study—neighborhood poverty and immigrant concentration—and an independent, normally distributed error term.

To address Research Question 2 (whether the role of supervision differs by immigrant generation), a set of two-way interaction terms between immigrant generation and supervision is added to the model at Level 2. For Research Questions 3 and 4, the roles of parental supervision and immigrant status relative to neighborhood context are assessed through cross-level interaction terms. The models for these two research questions thus add equations in which neighborhood poverty and immigrant concentration measures predict the coefficients for the variables thought to vary by neighborhood contextual factors. For Research Question 3, which considers whether the relationship between supervision and violence varies by neighborhood poverty and immigrant concentration, an additional Level-3 equation is included in which neighborhood poverty and immigrant concentration predict the Level-2 coefficient for parental supervision. Finally, to address Research Question 4, the two-way interaction terms are reintroduced at Level 2, and a set of additional Level-3 equations are added to estimate each of the Level-2 coefficients for first- and second-generation immigrant statuses, family supervision, and the Level-2 interactions between immigrant status and family supervision.

Results

Descriptive Results

Table 2 provides descriptive statistics for the independent variables in the analyses, by immigrant generation (first, second, and third-plus). As shown at the top of the three immigrant generation columns, 14% of the sample are first-generation immigrants, 33% are second-generation immigrants, and 54% are third-generation or higher immigrants. Within each of these immigrant generation categories, there are statistically significant mean differences by comparison to the other two categories. Significant differences between the first- and second-generation means and the first- and third-generation means are denoted in the final two columns under “First,” and significant differences between the second- and third-generation means are denoted in the last column under “Second.”

First, turning to Research Question 1: Are there significant mean differences in supervision by immigrant generation? Interestingly, there are significant differences across all three immigrant-generation pairings in the extent of family supervision. Mean supervision increases with immigrant generation, indicating that foreign-born parents supervise their children less, on average, than U.S.-born parents. This finding is somewhat counter to expectations, particularly because foreign-born parents are thought to take a strong interest in shepherding their children to ensure that familial goals of immigration, such as attainment, are fulfilled. On the other hand, it may be that foreign-born parents are more taxed in terms of the ability to supervise their children and/or are less equipped to identify the need for oversight. Alternately, it may be that the settings in which children of foreign-born parents live are, on average, qualitatively different than the settings in which children of U.S.-born parents reside, in ways that suggest less vigilance is needed (such as less disadvantaged settings or contexts with a greater proportion of immigrant residents).

Turning to the other variables in the analyses, there are significant mean differences between first- and second-generation adolescents on all demographic characteristics, including race/ethnicity, age, and family SES. By comparison to second-generation youth, greater proportions of first-generation adolescents are African American and White, and fewer first-generation adolescents are Latino/Latina; however, both the first and second generations are predominantly of Latino/Latina ethnicity. Unsurprisingly, a much greater proportion of third-generation adolescents (vs. first and second) are African American or White. By comparison to both second- and third-generation-plus youth, the first-generation immigrant category includes more respondents who are male, are slightly older, and are from families with lower SES. Further, there are significant mean differences by immigrant generation in verbal aptitude and prior problem behavior. Second-generation adolescents have the highest mean verbal aptitude score, and the mean for prior problem behavior increases with immigrant generation. There are no significant differences between first- and second-generation adolescents in impulsivity, but the means for each of these groups are significantly lower on this variable than the third-generation-plus impulsivity mean. Finally, first- and second-generation adolescents report similar levels of unstructured socializing, but these groups differ significantly from third-generation-plus youth who are more likely to engage in unstructured socializing.

There are notable similarities and differences in the means by immigrant generation for the family structure and support variables. First, there are no significant differences between first- and second-generation immigrants in the proportion living with two biological parents and in the extent of family attachment and support; however, both differ significantly from third-generation-plus adolescents, who report lower levels of family attachment and support and are less likely to live with both biological parents.

Turning, then, to neighborhood characteristics, on average, first-generation adolescents live in neighborhoods with lower levels of poverty than do second- or third-generation youth, and third-generation adolescents live in the most impoverished neighborhoods. Although both first- and second-generation adolescents, on average, live in neighborhoods with higher levels of immigrant concentration than third-generation-plus adolescents, the highest mean on this variable is found in the second-generation category and differs significantly from both the first- and third-generation-plus means.

These descriptive results suggest that family supervision may be a key variable for delineating among immigrant generations in the likelihood of violence perpetration. Further, the differences in the types of neighborhood contexts in which different immigrant generations reside may be important, as characteristics in the local environment can provide cues for parents in determining the level of supervision a child may require. Thus, there may be an interaction between supervision and the local environment that helps to explain violence perpetration among adolescents, and this mechanism may differ as well by immigrant generation.

Logistic Regression Results

Multivariate logistic regression results are set forth in Table 3. For ease of interpretation, the results shown have been converted from log odds to odds ratios (log odds results are available from the author upon request). Neighborhood variables are centered on the grand mean.

Table 3.

Multilevel Logistic Regression Models (Odds Ratios) of Violence (Including Neighborhood Characteristics).^a

Independent	Model
Variables	1	2	3	4	4a
Individual and family characteristics
Latino	1.581***	1.592***	1.574***	1.646***	1.548***
African American	1.862***	1.867***	1.828***	1.799***	1.651**
Male	1.910***	1.907***	1.904***	1.881***	1.731***
Age	1.274***	1.273***	1.274***	1.275***	1.144***
Age squared	0.967***	0.967***	0.967***	0.968***	0.958***
SES	0.949	0.948	0.949	0.955	0.964
Two biological parents	0.695***	0.694***	0.694***	0.705***	0.757**
Unstructured socializing	1.304***	1.305***	1.304***	1.296***	1.268***
Verbal aptitude	1.001	1.001	1.001	1.002	1.000
Impulsivity	1.392***	1.393***	1.398***	1.418***	1.222**
Prior problem behavior	—	—	—	—	1.853***
Immigrant status
First generation	0.535***	0.510***	0.535***	0.444***	0.513***
Second generation	0.796*	0.795*	0.796*	0.897	0.903
Family support
Family support main effect	0.946	0.947	0.945	0.948	0.937†
Supervision
Supervision main effect	0.923*	0.901†	0.931^†	0.881*	0.919
1st Gen. × Supervision	—	0.886	—	0.963	1.011
2nd Gen. × Supervision	—	1.116	—	1.012	0.998
Neighborhood main effects
Poverty	1.071	1.069	1.067	1.315***	1.352***
Immigrant concentration	0.855**	0.856**	0.847***	0.970	1.017
Neighborhood interactions
1st Gen. × Poverty	—	—	—	0.419***	0.396***
1st Gen. × Imm. Conc.	—	—	—	0.897	0.866
2nd Gen. × Poverty	—	—	—	0.750*	0.777*
2nd Gen. × Imm. Conc.	—	—	—	0.735***	0.699***
Supervision × Poverty	—	—	0.917*	0.895*	0.850**
Supervision × Imm. Conc.	—	—	0.952†	0.841**	0.828**
Three-way interactions
1st Gen. × Supervision × Poverty	—	—	—	1.503*	1.679**
1st Gen. × Supervision × Imm. Conc.	—	—	—	1.190	1.174
2nd Gen. × Supervision × Poverty	—	—	—	0.845†	0.927
2nd Gen. × Supervision × Imm. Conc.	—	—	—	1.280*	1.273**
Intercept	0.005***	0.005***	0.005***	0.004***	0.154***

Note. SES = socioeconomic status. ^aNeighborhood level N = 79; Person level N = 1,908.

^† p < .10. *p < .05. **p <.01. ***p < .001 (two-tailed tests).

The models shown in Table 3 proceed as follows. First, Model 1 is the baseline model that includes all control variables (except prior problem behavior) and the main (noninteraction) terms for the key variables of interest—immigrant status, family supervision, and neighborhood characteristics (poverty and immigrant concentration).

Then, Model 2 considers Research Question 2: Does the relationship between supervision and violence differ by immigrant generation? This model includes a single set of interactions—the two-way interactions between individual-level immigrant statuses and family supervision—to assess whether family supervision differs in its association with violence by immigrant generation. Model 3 considers Research Question 3: Does the relationship between supervision and violence vary by neighborhood poverty and immigrant concentration? This model thus includes another single set of interactions—the two-way, cross-level interactions between family supervision and neighborhood poverty and neighborhood immigrant concentration—to determine whether the effect of family supervision is of greater or lesser import in impoverished neighborhoods and those with higher levels of immigrant concentration.

Next, Model 4 turns to Research Question 4: Do family supervision, immigrant status, and neighborhood structural characteristics interact in their associations with a violence outcome, net of controls? This model includes interactions between and among immigrant status, family supervision, and the neighborhood structural characteristics (the three-way, cross-level interactions). Finally, Table 3 shows a supplemental Model (4a) that includes a control for prior problem behavior, to determine whether the results in Model 4 are robust to this control. This is of particular importance, given this study’s focus on family supervision.

As shown in the baseline model, Model 1 of Table 3, a number of factors are significantly associated with violence. First, by comparison to non-Latino White adolescents, Latino and African American youth are more likely to engage in violent behaviors. Male adolescents are also substantially more likely to engage in violence than female adolescents (nearly twice as likely). Age is also positively associated with violence at younger ages, but the relationship begins to reverse later in adolescence. (The oldest respondents in the sample are in late adolescence at Wave 2, the time point at which the violence items were measured.) Finally, impulsivity is positively associated with violence perpetration.

With regard to the key predictors in the models, and consistent with prior research, immigrant status is significantly associated with violence; first-generation immigrants are half as likely as third-generation-plus adolescents to engage in violence, while second-generation immigrants are over 20% less likely to endorse a given item on the violence scale than third-generation-plus youth. Looking to family structure and social processes, youth who reside with two biological parents are less likely to engage in violence than those who do not (consistent with prior findings). Although family support/attachment is not significantly associated with violence, parental supervision has (as expected) a significant, negative association with violence (at p < .05). Finally, at the neighborhood level, immigrant concentration is negatively associated with violence (p < .01), but neighborhood poverty is not significantly associated with violence.

Model 2 includes two-way interactions between immigrant statuses and family supervision to assess whether the relationship between family supervision and violence differs depending on immigrant generation (Research Question 2). As shown, these interaction terms are not statistically significant. Moreover, their inclusion does not materially affect other variables in the analysis (all of which retain effects similar to those shown in Model 1), with the exception that the main family supervision term becomes only marginally statistically significant (at the p < .10 level).

Model 3 examines Research Question 3 and includes a different set of interaction terms to assess whether family supervision differs in its association with violence depending on the structural characteristics of the neighborhood context (cross-level interactions). As shown, inclusion of these interactions has little effect on the control variables in the analysis, all of which remain similar in magnitude, direction, and statistical significance to the results shown for Model 1. With regard to the interaction terms, neighborhood poverty moderates the association between supervision and violence (significant at p < .05). The supervision measure remains negatively associated with violence (although it is only marginally statistically significant at p < .10), and the neighborhood poverty measure is not significantly associated with violence. With regard to the association between neighborhood immigrant concentration and violence, the coefficient for immigrant concentration remains negative but increases slightly in magnitude as well as in statistical significance (p < .001). Further, the interaction between family supervision and immigrant concentration is negative and marginally significant (at p < .10).

Finally, Model 4 turns to Research Question 4. Three-way cross-level interaction terms are added to the analyses to assess the interplay between and among family supervision, immigrant status, and neighborhood structural characteristics. (Although the two-way lower order interaction terms between supervision and immigrant status were not significant in Model 2, they are included here to ensure proper specification of the three-way interaction model; see Aiken & West, 1991.) The results illustrate the complexity of the interrelationships between and among the three sets of key variables in the analyses (immigrant status, parental supervision, and neighborhood structural characteristics). First, there are few marked changes in the coefficients for the control variables in the analyses, by comparison to Model 3. The most notable difference is that the positive relationship between Latino ethnicity and violence increases in magnitude by roughly 8% over the prior model.

Turning to the conditional effects for the predictors of interest, the inclusion of the three-way interaction terms results in an increase in the magnitude of the negative coefficient for first-generation immigrant status (which remains significant at p < .01). In contrast, the magnitude of the negative association between second-generation immigrant status and violence decreases, and the second-generation term is no longer statistically significant. The supervision coefficient remains negative, increases in magnitude by an additional 5% by comparison to Model 3, and becomes statistically significant at p < .05. The neighborhood measures also vary markedly in their associations with violence by comparison to Model 3. Neighborhood poverty becomes significantly positively associated with violence (at p < .001), and the magnitude of the relationship is quite strong. In contrast, neighborhood immigrant concentration decreases substantially in magnitude and is no longer statistically significant.

As in Model 2, the two-way interactions between immigrant generation indicators and supervision are not statistically significant. In contrast, three of the four two-way interactions between immigrant generation and neighborhood characteristics are significant and negatively associated with violence as follows: First Generation × Poverty (at p < .001), Second Generation × Poverty (at p < .05), and Second Generation × Immigrant Concentration (at p < .001). Further, and importantly, family supervision interacts significantly with both neighborhood poverty (at p < .01) and immigrant concentration (at p < .01).

In addition, there are significant three-way interactions. The effect of First Generation × Supervision × Poverty is positive and significant at p < .05. In contrast, Second Generation × Supervision × Poverty is negative (and is marginally statistically significant, at the p < .10 threshold). Finally, the coefficient for Second Generation × Supervision × Immigrant Concentration is positive and statistically significant at p < .05.

Model 4 illustrates a number of significant interaction effects, suggesting that it is important to consider the interplay between and among immigrant status, parental supervision, and neighborhood characteristics in predicting adolescent violence. It is difficult, however, to interpret the meaning of any individual coefficient for the lower order and interaction terms included in Model 4 without considering the net effects of these measures. For this reason, I provide Figure 1, which is a graph illustrating predicted probabilities of violence by immigrant generation; low and high supervision; and low, mean, and high neighborhood poverty. Predicted values are based on the Model 4 results. Low and high supervision are specified as 2 SDs below and above the parental supervision mean. Low and high poverty are specified as 1 SD below and above the mean (a narrower range to represent a plausible range of values for the neighborhoods in which first- and second-generation adolescents reside). All other variables in the analyses are set at their sample mean values, with the exceptions of age (which is set at 14 years) and neighborhood immigrant concentration (which is set at a level that is the average of the mean values for first- and second-generation immigrants, to better approximate the neighborhoods in which immigrant youth actually reside).

Figure 1.

Predicted probability of violence at age 14, by immigrant status, family supervision, and neighborhood poverty.

As shown in Figure 1, the association between neighborhood poverty and adolescent violence is conditioned by both immigrant status and parental supervision. First, looking across levels of neighborhood disadvantage for less supervised adolescents, second- and third-generation-plus youth who are less supervised are at considerably greater risk of violence when they reside in higher poverty contexts. The relationship is particularly marked for third-generation-plus youth who are less supervised; in a low-poverty setting, these adolescents have a probability of violence of just over .09. In high-poverty neighborhoods, the risk jumps to over .20, the highest plotted value. Similarly, low supervised, second-generation adolescents have a probability of violence of less than .05 in low-poverty neighborhoods, a figure that more than doubles (to over .11) in high-poverty settings. In contrast, less supervised first-generation adolescents are at greatest risk in low-poverty neighborhoods. The lowest predicted probability of violence across all plotted categories shown is for less supervised first-generation youth residing in higher poverty contexts (under .02). Less supervised first-generation youth residing in lower poverty contexts have the second highest predicted probability of violence of all of the immigrant generation family supervision combinations—a value of just over .14.

Looking across neighborhoods at highly supervised youth, there is little variation in predicted violence across levels of neighborhood poverty for highly supervised first- and third-generation-plus youth. For highly supervised second-generation youth, however, the pattern is somewhat different and is best understood by examining patterns within the three categories of neighborhood poverty shown in Figure 1. Perhaps surprisingly, more highly supervised second-generation adolescents residing in low-poverty neighborhoods have twice the likelihood of violence of less supervised second-generation adolescents residing in these same contexts (.092 vs. .045). Thus, although supervision is protective for second-generation youth in impoverished neighborhoods, it is associated with a greater likelihood of violence in settings that are more affluent and nearly equals the probability of violence of less supervised third-generation-plus adolescents in these settings.

To further illustrate the relationships shown in Model 4, I provide a separate Figure 2, which again depicts predicted probabilities of violence, this time across levels of neighborhood immigrant concentration (based on Model 4 results). For Figure 2, the values for all variables are set as in Figure 1, with neighborhood poverty set at its mean. Immigrant concentration values are based on the average of the descriptive statistics for neighborhood immigrant concentration for the first- and second-generation adolescents in the sample; low and high immigrant concentration are 1.5 SDs above and below the mean. (To orient the reader, the values for the middle set of columns are the same in both Figures 1 and 2, since both immigrant concentration and poverty are at their means in this set of columns.)

Figure 2.

Predicted probability of violence at age 14, by immigrant status, family supervision, and immigrant concentration.

As shown in Figure 2, looking across levels of immigrant concentration, the probability of violence for less supervised third-generation-plus adolescents is greater at higher levels of immigrant concentration, and this group has the greatest likelihood of violence of the groups depicted (nearly .20 in high immigrant concentration neighborhoods). In contrast, for all other groups (including highly supervised third-generation adolescents), the predicted probability of violence is lower at higher levels of immigrant concentration. Looking within categories of immigrant concentration, higher levels of supervision are associated with a lower probability of violence for all groups, with the exception of second-generation youth residing in neighborhoods with higher levels of immigrant concentration. Within this category, more supervised, second-generation youth are slightly more likely to engage in violence than their less supervised, second-generation peers.

To examine the robustness of the results shown in Model 4, one additional Model (4a) introduces the control for prior problem behavior (including violence). The inclusion of this variable decreases the magnitude of many of the individual and family characteristics in the model, and the African American, two biological parents, and impulsivity coefficients decrease in statistical significance to the p < .01 threshold. Among the variables of interest, the first-generation coefficient decreases slightly in magnitude but remains significant at p < .05. The supervision term is no longer statistically significant. The neighborhood poverty term increases slightly in magnitude and remains statistically significant at p < .05. Turning to the two-way interactions, all remain significant (with the Supervision × Neighborhood Poverty term increasing in statistical significance to p < .01), and several increase slightly in magnitude. Finally, among the three-way interaction terms, the effect of First Generation × Supervision × Neighborhood Poverty increases substantially in both magnitude and statistical significance (to p < .01), the Second-Generation × Supervision × Immigrant Concentration coefficient increases in statistical significance (to p < .01), and the Second Generation × Supervision × Poverty coefficient decreases and is no longer statistically significant. Thus, most of the variables of interest remain significant, with the exceptions of the supervision term and the three-way Second Generation × Supervision × Poverty term. This suggests that the association between supervision and adolescent violence is manifested through reference to the characteristics of the local neighborhood context: poverty and immigrant concentration for third-generation-plus adolescents, poverty alone for first-generation adolescents, and immigrant concentration alone for second-generation youth.

Discussion

Summary of Key Findings

With regard to Research Question 1, this study finds significant mean differences in supervision by immigrant generation; however, counter to expectations, mean supervision increases with immigrant generation. Moving to Research Question 2, the results affirm that there is not a significant two-way interaction between supervision and immigrant generation.

Yet, this study highlights the importance of examining the interaction between family supervision and neighborhood characteristics in their effects on adolescent violence (Research Question 3) and interactions between and among immigrant status, family supervision, and neighborhood characteristics (Research Question 4). First, for Research Question 3, findings indicate that neighborhood poverty moderates the association between parental supervision and violence. Turning to Research Question 4, this study again confirms prior research findings: In most contexts, first- and second-generation immigrant youth are less likely to perpetrate violence than third-generation-plus adolescents. Similarly, neighborhoods with higher levels of immigrant concentration appear to be protective for most youth (although the extent of this protective effect varies by immigrant generation), with the exception of less supervised third-generation-plus adolescents (see Figure 2). The finding with regard to less supervised third-generation-plus adolescents suggests that immigrant neighborhoods may be risk environments for this group uniquely, due perhaps to their inability to tap into neighborhood resources that promote child and adolescent well-being (if the neighborhood is an immigrant enclave) or because they are the “other” in a neighborhood with a greater proportion of foreign-born residents.

Overall, deficits in parental supervision are associated with a greater likelihood of violence in higher poverty contexts for second- and third-generation adolescents. In contrast, for first-generation adolescents, deficits in supervision are associated with a greater likelihood of violence in lower poverty contexts as shown in Figure 1. Finally, for second-generation youth, higher levels of supervision are associated with a lower probability of violence in high-poverty settings, and a higher probability of violence in low poverty settings, by comparison to less supervised second-generation adolescents in those same contexts (although the association for second-generation youth is partly explained by prior problem behavior in Model 4a).

The patterns shown in Figures 1 and 2 thus show mixed support for the notion that supervision is a strategy that can uniformly offset the effects of high-risk contexts on problem behavior, supporting some (but not all) of the findings of prior research (Bámaca et al., 2005; Furstenberg et al., 1999). For second- and third-generation-plus adolescents, the probability of violence is much lower in high-risk contexts when they are highly supervised. For first-generation adolescents, the probability of violence is also lower if they are more supervised—but only in low-poverty contexts. These seemingly inconsistent results suggest that there is something about more advantaged neighborhoods that makes them particularly high risk for foreign-born adolescents who are less supervised. Yet the puzzle extends to the other end of the neighborhood poverty spectrum as well: Why are unsupervised first-generation adolescents so unlikely to be violent in economically disadvantaged contexts?

One possibility is related to the mechanism suggested by Bámaca and colleagues (2005): Perhaps first-generation adolescents are more likely to perceive and internalize deficits in supervision in more advantaged contexts, by comparison to their peers, and act out in response. In disadvantaged contexts, on the other hand, they may understand that low supervision is due to the other challenges their parents are facing, and is not a “deficit,” as such. As a result, they may refrain from violence out of a sense of familial duty.

An alternate explanation is related to the “otherness” of first-generation immigrants in comparison to their native-born peers (even those who are children of immigrants themselves). Foreign-born adolescents may be less able to assimilate to the behaviors that dominate in their local environments or may be blocked from doing so. Due to cultural barriers, first-generation youth may be less likely to become enmeshed with native-born adolescent groups. In settings in which the perpetration of violence is more common, such as economically disadvantaged areas, this may protect them from delinquent involvement (particularly because it is unwise to engage in violence in such settings without the backing of peers). In more advantaged settings, on the other hand, first-generation youth may be excluded from groups that offer access to prosocial activities and resources (Rodriguez, 1993). They may engage in violence, even on an individual basis, out of frustration or strain (Agnew, 1992; Merton, 1938). Perhaps some foreign-born parents in more affluent settings do not highly supervise their children because they want their children to avail themselves of resources and opportunities in the local environment. If these opportunities are foreclosed (as for a first-generation adolescent vs. a second-generation immigrant), the adolescent may experience a unique strain, owing in part to their inability to fulfill their parents’ aspirations for them, despite apparently abundant resources.

Policy Implications

Of the findings in this study, perhaps the most discouraging is the greater likelihood of violence among less supervised first-generation adolescents residing in more advantaged settings, by comparison to their native-born (and more supervised first-generation) peers residing in the same contexts. Yet this finding highlights the potential for intervention programs to ensure that immigrant adolescents are able to tap into the more abundant resources available in advantaged areas and to bolster the efforts of foreign-born parents to shepherd children in these settings. As one example, bilingual mentors could be made available to assist first-generation youth with homework and accompany their parents to school events. These mentors could also serve as sources of social and cultural capital and provide linkages to the wider community, to help level the playing field and ensure that the prospects for attainment among the foreign born in advantaged settings resemble those of their native-born peers.

Relatedly, it is evident that immigrant neighborhoods are protective against violence for most adolescents—with the notable exception of less supervised third-generation-plus youth. Greater understanding of the social processes that underlie the protective nature of immigrant neighborhoods is warranted, but policies directed at inclusion for adolescents whose native-born parents are less involved may help them to gain entrée to immigrant social networks (and tap into resources associated with residence in an immigrant enclave).

Limitations

This study does, however, have limitations that should be noted. Given the complexity of the statistical models and the need to preserve statistical power to support the analyses, I was unable to fully examine a number of other factors that likely condition these findings. First, I was unable to examine subcategories of immigrant generation, disaggregated by race/ethnicity/nativity or (for foreign-born youth) the year of arrival in the United States, without sacrificing statistical power and model stability. Certainly, racial/ethnic characteristics and identity, cultural factors (such as language use), and “newness” are all important considerations; these aspects add further layers to the immigration process, and their inclusion would yield a more complete picture of the “match” between an individual adolescent and his or her neighborhood context. Further, theory and research on familism suggest that there are culturally distinct parenting practices that prevail across different races and ethnicities (Updegraff et al., 2005), and distinctions between and among immigrant families on this basis are likely very important.

These potential differences highlight a second limitation of this research—some components in the measure of parental supervision used in this study reflect American standards and include items related to children’s schooling. As a result, they are potentially less valid for foreign-born parents who may not be able to assist their children with homework or visit the school to speak to teachers, particularly if there are language barriers (Turney & Kao, 2009). The findings for Research Question 1 (which showed significantly lower means in supervision for first- and second-generation adolescents compared to third-generation-plus adolescents) could be due in part to the limitations of some of the scale items for measuring supervision among foreign-born parents. The current study is intended as an initial effort to examine interactions between and among immigrant status, supervision, and contextual factors, using a well-established scale for supervision. Future research is warranted to fully explore the nature of supervision in immigrant families to better understand how supervision interacts with community factors in shaping adolescent delinquency.

Third (and again related to statistical power), I was unable to examine whether (and how) gender further conditions the interrelationships among immigrant status, supervision, and neighborhood characteristics. Prior research findings highlight the importance of gender and gender roles, particularly when family supervision is a variable of interest and the outcome is violence (see Fagan & Wright, 2012; Zimmerman & Messner, 2010). Given substantial cultural variation in norms and expectations for girls and boys, it is likely that the processes described here are further conditioned by gender.

Finally, the analyses and conclusions are based on data from a single urban context, the city of Chicago. Although the PHDCN data have been used widely and are well regarded, the findings may differ in newer immigrant destinations or in contexts with rich, multigenerational histories of immigration, such as Los Angeles or the Southwest United States. Relatedly, contextual measures such as “immigrant concentration” cannot fully delineate between concentration due to residential segregation (constraints on residential mobility) versus the potentially prosocial effects of immigrant “enclaves,” neighborhoods that are desired destinations for immigrant families and are resource rich.

Conclusion and Future Research

Nonetheless, these findings suggest that it is important to consider how family and neighborhood factors work together in moderating the risks that adolescents face. Further, this study suggests that a “one-size-fits-all” approach may be inappropriate when considering variation in delinquency by immigrant generation. In particular, these findings suggest a theoretical shift: Why are seemingly low-risk contexts not low risk for foreign-born adolescents? Further, why are neighborhoods with higher levels of immigrant concentration not protective for less supervised third-generation-plus adolescents?

Future research on the interplay between and among immigrant status, family supervision, and neighborhood characteristics may help to tease out the precise mechanisms that underlie these differences, possibly by identifying the factors that make more affluent contexts “risky” for foreign-born youth. One potential avenue is an examination of measures of acculturation and assimilation to better understand the barriers that block foreign-born youth from access to resources in more advantaged settings.

Further (and as noted in the Limitations section, above), future research should consider the many layers of immigrant identities, particularly intersections of race/ethnicity, gender, and immigrant status. In many immigrant families, parenting practices are shaped by cultural legacies, and the linkages between and among supervision, immigrant status, and neighborhood characteristics cannot be fully explored without taking these factors into account. The interrelationships are undoubtedly complex, but continued research on the contexts in which immigrant family social processes “play out” is of great import. As illustrated by the findings of this study, the family/community nexus may be the key to understanding the “paradoxical” protective aspects of immigrant status and immigrant communities.

Footnotes

Appendix

Table A1.

Items in Parental Supervision Scale.

Item	Supervision Subscale, PHDCN Wave 1, Primary Caregiver (Cohorts 9–15)
1.	Subject has a set time (curfew) to be home on school nights.
2.	Subject routinely obeys curfew on school nights.
3.	Subject has a set time (curfew) to be home on weekend nights.
4.	Subject routinely obeys curfew on weekend nights.
5.	Primary caregiver has established rules about homework and checks to see if homework is done.
6.	Primary caregiver assisted subject with homework and school assignments (at least) every other week during current or most recent school year.
7.	Primary caregiver requires subject to sleep at home on school nights.
8.	When primary caregiver is not available to subject at home, reasonable procedures have been established for him/her to check in with primary caregiver, or their designee, on weekends and after school.
9.	After school subject goes somewhere that adult supervision is provided.
10.	Primary caregiver establishes rules for subject’s behavior with peers and asks questions to determine whether they are being followed.
11.	Subject is not allowed to wander in public places without adult supervision for more than 2 hours.
12.	Primary caregiver has had contact with two of the subject’s friends in the last 2 weeks.
13.	Primary caregiver talks daily with subject about his/her day.
14.	Primary caregiver has visited the school or talked to the teacher or counselor within the last 3 months.
15.	Family has TV, and it is used judiciously, not left on continuously.
16.	Primary caregiver has discussed television programs with subject during the past 2 weeks.
17.	Primary caregiver has discussed current events with subject during the past 2 weeks.
18.	Primary caregiver has discussed the hazards of drug and alcohol abuse with subject during the past year.
19.	Primary caregiver denies subject access to alcohol (including beer or wine) in the home.
20.	Primary caregiver knows signs of drug usage and remains alert to possible experimentation.
21.	Subject is regularly taken to a doctor’s office or clinic for check-ups and preventive health care (once a year).
22.	Family has a fairly regular and predictable daily schedule for subject.
23.	Primary caregiver sets limits for subject and generally enforces them.
24.	Primary caregiver is generally consistent in establishing or applying family rules.

Note. PHDCN = Project on Human Development in Chicago Neighborhoods; TV, television.

Acknowledgments

The author wishes to thank the Center for Family and Demographic Research at Bowling Green State University and members of the Racial Democracy, Crime and Justice Network based at Ohio State University for their support of this project.

Author’s Notes

The findings reported here do not necessarily represent the views of the funders or scientific directors of the Project on Human Development in Chicago Neighborhoods.

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Project on Human Development in Chicago Neighborhoods was funded by the John D. and Catherine T. MacArthur Foundation, the National Institute of Justice, and the National Institute of Mental Health. This research was supported in part by the Center for Family and Demographic Research, Bowling Green State University, which has core funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R24HD050959).

References

Agnew

(1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30, 47–87.

Aiken

L. S.

West

S. G.

(1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks, CA: Sage

Anderson

(1999). Code of the street: Decency, violence and the moral life of the inner city. New York, NY: W.W. Norton.

Bámaca

M. Y.

Umaña-Taylor

A. J.

Shin

Alfaro

E. C.

(2005). Latino adolescents’ perceptions of parenting behaviors and self-esteem: Examining the role of neighborhood risk. Family Relations, 54, 621–632.

Booth

Crouter

A. C.

(2001). Does it take a village? Community effects on children, adolescents, and families. New York, NY: Lawrence Erlbaum.

Bronfenbrenner

(1979). The ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard University Press.

Buss

Plomin

(1984). Temperament: Early developing personality traits. Hillsdale, NJ: Lawrence Erlbaum.

Butcher

K. E.

Piehl

A. M.

(1999). Cross-city evidence on the relationship between immigration and crime. Journal of Policy Analysis and Management, 17, 457–493.

Caldwell

Bradley

(1984). Home observation for measurement of the environment (HOME)—Revised edition. Little Rock: University of Arkansas, Little Rock.

10.

Cohen

Felson

(1979). Social change and crime rate trends: A routine activities approach. American Sociological Review, 44, 588–607.

11.

Cook

T. D.

(2003). The case for studying multiple contexts simultaneously. Addiction, 98, 151–155.

12.

Cuéllar

Arnold

González

(1995). Cognitive referents of acculturation: Assessment of cultural constructs. Journal of Community Psychology, 23, 339–356.

13.

Desmond

S. A.

Kubrin

C. E.

(2009). The power of place: Immigrant communities and adolescent violence. The Sociological Quarterly, 50, 581–607.

14.

Earls

F. J.

Raudenbush

S. W.

Reiss

A. J.

Jr Sampson

R. J.

(1995). Project on human development in Chicago neighborhoods (PHDCN): Systematic social observation, 1995 [Computer file]. ICPSR13578-v1. Boston, MA: Harvard Medical School [producer], 2002. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], July 18, 2005.

15.

Elliott

D. S.

Menard

Rankin

Elliott

Huizinga

Wilson

W. J.

(2006). Good kids from bad neighborhoods: Successful development in social context. Cambridge, England: Cambridge University Press.

16.

Estrada-Martínez

L. M.

Padilla

M. B.

Caldwell

C. H.

Schulz

A. J.

(2011). Examining the influence of family environments on youth violence: A comparison of Mexican, Puerto Rican, Cuban, non-Latino Black, and non-Latino White adolescents. Journal of Youth and Adolescence, 40, 1039–1051.

17.

Fagan

A. A.

Wright

E. M.

(2012). The effects of neighborhood context on youth violence: Does gender matter? Youth Violence and Juvenile Justice, 10, 64–82.

18.

Farrington

D. P.

(2002). Families and crime. In Wilson

J. Q.

Petersilia

(Eds.), Crime: Public policies for crime control (2nd ed., pp. 129–148). Oakland, CA: Institute for Contemporary Studies Press.

19.

Farrington

D. P.

Loeber

(1999). Transatlantic replicability of risk factors in the development of delinquency. In Cohen

Slomkowski

Robins

L. N.

(Eds.), Historical and geographic influences on psychopathology (pp. 299–329). Mahwah, NJ: Lawrence Erlbaum.

20.

Furstenberg

F. F.

Jr Cook

T. D.

Eccles

Elder

G. H.

Jr Sameroff

(1999). Managing to make it: Urban families and adolescent success. Chicago, IL: The University of Chicago Press.

21.

Germán

Gonzales

N. A.

Dumka

(2009). Familism values as a protective factor for Mexican-origin adolescents exposed to deviant peers. Journal of Early Adolescence, 29, 16–42.

22.

Gorman-Smith

Henry

D. B.

Tolan

P. H.

(2004). Exposure to community violence and violence perpetration: The protective effects of family functioning. Journal of Clinical Child & Adolescent Psychology, 33, 439–449.

23.

Gottfredson

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

24.

Harker

(2001). Immigrant generation, assimilation, and adolescent well-being. Social Forces, 79, 969–1004.

25.

Harris

K. M.

(1999). The health status and risk behaviors of adolescents in immigrant families. In Hernandez

D. J.

(Ed.), Children of immigrants: Health, adjustment, and public assistance (pp. 286–347). Washington, DC: National Academies Press.

26.

Haynie

D. L.

Osgood

D. W.

(2005). Reconsidering peers and delinquency: How do peers matter? Social Forces, 84, 1109–1130.

27.

Hirschi

(1969). Causes of delinquency. Berkeley, CA: University of California Press.

28.

Horowitz

(1983). Honor and the American dream: Culture and identity in a Chicano community. New Brunswick, NJ: Rutgers University Press.

29.

Jarrett

R. L.

(1997). African American family and parenting strategies in impoverished neighborhoods. Qualitative Sociology, 20, 275–288.

30.

T. N.

Stockdale

(2008). Acculturative dissonance, ethnic identity, and youth violence. Cultural Diversity and Ethnic Minority Psychology, 14, 1–9.

31.

Leventhal

Brooks-Gunn

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

32.

Levine-Coley

Hoffman

L. W.

(1996). Relations of parental supervision and monitoring to children’s functioning in various contexts: Moderating effects of families and neighborhoods. Journal of Applied Developmental Psychology, 17, 51–68.

33.

Loeber

Stouthamer-Loeber

(1986). Family factors as correlates and predictors of juvenile conduct problems and delinquency. In Morris

(Ed.), Crime and justice: An annual review of research (Vol. 7., pp. 29–159). Chicago, IL: University of Chicago Press.

34.

MacLeod

(2008). Ain’t no makin’ it: Aspirations and attainment in a low-income neighborhood. Boulder, CO: Westview Press.

35.

Maimon

Browning

C. R.

(2010). Unstructured socializing, collective efficacy, and violent behavior among urban youth. Criminology, 48, 443–474.

36.

Markides

K. S.

Coreil

(1986). The health of Hispanics in the southwestern United States: An epidemiologic paradox. Public Health Reports, 101, 253–265.

37.

Marsiglia

F. F.

Parsai

Kulis

(2009). Effects of familism and family cohesion on problem behaviors among adolescents in Mexican immigrant families in the Southwest U.S. Journal of Ethnic and Cultural Diversity in Social Work, 18, 203–220.

38.

Martinez

Jr . (2002). Latino homicide: Immigration, violence, and community. New York, NY: Routledge.

39.

McCord

(1979). Some child-rearing antecedents of criminal behavior in adult men. Journal of Personality and Social Psychology, 37, 1477–1486.

40.

Merton

R. K.

(1938). Social structure and anomie. American Sociological Review, 3, 672–682.

41.

Morenoff

J. D.

Astor

(2006). Immigrant assimilation and crime: Generational differences in youth violence in Chicago. In Martinez

Jr Valenzuela

Jr . (Eds.), Immigration and crime: Race, ethnicity, and violence (pp. 36–63). New York: New York University Press.

42.

Nye

F. I.

(1958). Family relationships and delinquent behavior. New York, NY: Wiley.

43.

Osgood

D. W.

Anderson

A. L.

(2004). Unstructured socializing and rates of delinquency. Criminology, 42, 519–549.

44.

Portes

Fernández-Kelly

(2008). No margin for error: Educational and occupational achievement among disadvantaged children of immigrants. The ANNALS of the American Academy of Political and Social Science, 620, 12–36.

45.

Portes

Rumbaut

R. G.

(2001). Legacies: The story of the immigrant second generation. Berkeley: University of California Press.

46.

Portes

Zhou

(1993). The new second generation: Segmented assimilation and its variants. Annals of the American Academy of Political and Social Sciences, 530, 74–96.

47.

Rankin

B. H.

Quane

J. M.

(2002). Social contexts and urban adolescent outcomes: The interrelated effects of neighborhoods, families, and peers on African-American youth. Social Problems, 49, 79–100.

48.

Raudenbush

S. W.

Bryk

A. S.

Cheong

A. S.

Fai

Y. F.

Congdon

R. T.

du Toit

(2011). HLM 7: Hierarchical linear and nonlinear modeling, Lincolnwood, IL: Scientific Software International.

49.

Raudenbush

S. W.

Johnson

Sampson

R. J.

(2003). A multivariate, multilevel Rasch model for self-reported criminal behavior. Sociological Methodology, 33, 169–211.

50.

Reiss

A. J.

Jr . (1951). Delinquency as the failure of personal and social controls. American Sociological Review, 16, 196–207.

51.

Rodriguez

L. J.

(1993). Always running: La vida loca, gang days in L.A. New York, NY: Touchstone.

52.

Royston

(2005). Multiple imputation of missing values: Update. The Stata Journal, 5, 1–14.

53.

Sampson

R. J.

(1992). Family management and child development: Insights from social disorganization theory. In McCord

(Ed.), Facts, frameworks, and forecasts: Advances in criminological theory (Vol. 3, pp. 63–93). New Brunswick, NJ: Transaction.

54.

Sampson

R. J.

(2008). Rethinking crime and immigration. Contexts, Winter, 7, 28–33.

55.

Sampson

R. J.

Laub

J. H.

(1993). Crime in the making: Pathways and turning points through life. Cambridge, MA: Harvard University Press.

56.

Sampson

R. J.

Morenoff

Gannon-Rowley

(2002). Assessing “neighborhood effects”: Social processes and new directions in research. Annual Review of Sociology, 28, 443–478.

57.

Sampson

R. J.

Morenoff

Raudenbush

S. W.

(2005). Social anatomy of racial and ethnic disparities in violence. American Journal of Public Health, 95, 224–232.

58.

Sampson

R. J.

Raudenbush

S. W.

Earls

(1997). Neighborhoods and violent drime: A multilevel study of collective efficacy. Science, 227, 918–923.

59.

Shaw

C. R.

McKay

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

60.

Smith

C. A.

Stern

S. B.

(1997). Delinquency and antisocial behavior: A review of family processes and intervention research. Social Service Review, 71, 382–420.

61.

Smith

R. C.

(2008). Horatio Alger lives in Brooklyn: Extrafamily support, intrafamily dynamics, and socially neutral operating identities in exceptional mobility among children of Mexican immigrants. The Annals of the American Academy of Political and Social Science, 620, 270–290.

62.

StataCorp. (2005). Stata statistical software: Release 9, College Station, TX: StataCorp LP.

63.

Steinberg

Mounts

N. S.

Lamborn

S. D.

Dornbusch

S. M.

(1991). Authoritative parenting and adolescent adjustment across varied ecological niches. Journal of Research on Adolescence, 1, 19–36.

64.

Thomas

W. I.

Znaniecki

(1984 [1918–1920]). The Polish peasant in Europe and America: Monograph of an immigrant group, edited and abridged by Zaretsky

Eli

. Chicago: University of Illinois Press.

65.

Turney

Kao

(2009). Barriers to school involvement: Are immigrant parents disadvantaged? The Journal of Educational Research, 102, 257–271.

66.

Updegraff

K. A.

McHale

S. M.

Whiteman

S. D.

Thayer

S. M.

Delgado

M. Y.

(2005). Adolescent sibling relationships in Mexican American families: Exploring the role of familism. Journal of Family Psychology, 19, 512–522.

67.

Waters

(1999). Crime and immigrant youth. Thousand Oaks, CA: Sage.

68.

Wilkinson

(1993). WRAT-3: Wide-range achievement test, administration manual. Wilmington, DE: Wide Range.

69.

Xie

Greenman

(2005). Segmented assimilation theory: A reformulation and empirical test. Research Report 05-581, Population Studies Center, University of Michigan Institute for Social Research, Ann Arbor, MI.

70.

Zimmerman

G. M.

Messner

S. F.

(2010). Neighborhood context and the gender gap in adolescent violent crime. American Sociological Review, 75, 958–980.