The Role of Families in Enhancing or Tempering Neighborhood and School Impacts

Just as interactions between parents and children are bi-directional (Bell & Chapman, 1986; Collins, Maccoby, Steinberg, Hetherington, & Bornstein, 2000), the interactions among families and their neighborhoods and schools are expected to be bidirectional as well (Leventhal & Brooks-Gunn,

2000). Children, and adolescents in particular, can choose their peers, the neighborhoods in which they hang out, and, in cities and locales such as New York, have a role in choosing (or qualifying for) their high schools. More importantly, the decisions to live in certain neighborhoods and to attend certain schools depend on parents' background characteristics and current circumstances. For example, among a sample of incarcerated adolescents, those who reported high levels of family conflict and family socio-economic risks (e.g., family criminality, parent unemployment) were more likely to report perpetration of, witnessing of, and victimization by violence (Halliday-Boykins & Graham,

2001), likely because families with socio-economic disadvantage and forced to live in disadvantaged and dangerous neighborhoods.

These important individual- and family-level determinants of how families come to live in or remain in certain neighborhoods and of which schools children attend result in a logical and statistical problem of endogeneity (Duncan & Raudenbush, 2001), such that the process of identifying neighborhood or school impacts on individuals is clouded by the fact that individuals can choose their neighborhoods and schools. The best non-experimental solution to this problem of endogeneity is to "measure the unmeasured" (Duncan & Raudenbush, 2001) parent characteristics to reduce the endogenous membership bias. Typical family-level controls are income, education, race/ethnicity, maternal age at birth, family structure, and family size (Leventhal & Brooks-Gunn, 2000).

Beyond family socio-demographic factors that may determine the neighborhoods in which children live and the schools they attend, parents play an active role in helping children negotiate certain neighborhood challenges. Following from McLoyd's (1990) assertions that parents in dangerous neighborhoods will rely on harsh punishment of their adolescents' misbehaviors because the consequences of misbehavior are potentially grave, studies linking neighborhood characteristics to parent behavior and parenting practices abound. Parents who live in disadvantaged neighborhoods tend to show less warmth and to be more controlling of their children (Earls, McGuire, & Shay, 1994; Furstenberg, 1993; Furstenberg, Cook, Eccles, Elder, & Sameroff, 1999; Klebanov, Brooks-Gunn, & Duncan 1994; Pinderhughes, Nix, Foster, Jones, the Conduct Problems Prevention Research Group., 2001). Other neighborhood characteristics, such as dissatisfaction with public services and perceived danger, have also been linked with more harsh, inappropriate, and inconsistent discipline (Hill & Herman-Stahl, 2002; Pinderhughes et al., 2001). Neighborhood influences have been associated with parenting practices over and above family income and education (Klebanov et al., 1994).

Families formulate different strategies for raising children in high risk neighborhoods (Fursten-berg, 1993; Jones, 2001) and these strategies, in turn, can either buffer or exacerbate the effects of disadvantaged neighborhoods on adolescents. For example, the success of different parent strategies in promoting adolescent autonomy has been found to depend on the level of environmental risk (Boykin McElhaney & Allen, 2001). As evidence of buffering, parents' reliance on practices high in control has been found to be beneficial for adolescents living in high-risk neighborhoods whereas low control is beneficial for adolescents in low-risk neighborhoods (Eamon, 2001; Gershoff et al., 2004; Gonzales et al., 1996; Lamborn, Dornbusch, & Steinberg, 1996). Similarly, high parental monitoring of children's activities is associated with decreased externalizing behavior problems among families who reside in neighborhoods characterized by high residential instability (Beyers, et al., 2003) or low collective efficacy (Rankin & Quane, 2002). High support from parents has been found to moderate the negative impact of disadvantaged and high risk neighborhoods on children's mental health, social competence, and intrusive thinking (Kliewer et al., 1998; Krenichyn, Saegert, & Evans, 2001; Stiffman, Hadley-Ives, Elze, Johnson, & Doré 1999). Children who reside in disadvantaged neighborhoods but whose parents are nurturant and involved are less likely to associate with deviant peers (Brody et al., 2001). In general, neighborhood social capital has been found to increase positive parenting and in turn decrease children's adjustment (Dorsey & Forehand, 2003). Consistent with Bronfenbrenner and Ceci's (1994) hypothesis that the ability of parent behavior to buffer children against negative outcomes will be greater in impoverished and unstable environments, there is indeed some evidence that parenting may play a more important role than temperament in determining children's behavior in disadvantaged neighborhoods. Specifically, violent behaviors among adolescents in advantaged neighborhoods appear to be determined more from temperament-based factors, such as hyperactivity and aggression, whereas violence among adolescents in disadvantaged neighborhoods is influenced more by quality of communication with parents (Beyers, Loeber, Wilkstrom, & Stouthamer-Loeber, 2001).

The extent to which parents are aware of the risks their children are exposed to can impact the quality of the parent-child relationship as well as children's mental health. When parents are more aware of and thus in agreement with children's reports of children's exposure to violence, children experience fewer PTSD symptoms (Ceballo et al., 2001). When faced with daily dangers, children who are able to discuss them with their parents, and perhaps collaboratively devise strategies for avoiding them, are likely better able to cope with such dangers as exposure to violence.

Transmission of negative neighborhood influences is also possible through increased parent stress and harsh behavior. Mothers' ratings of neighborhood danger are associated with increased maternal depression (Hill & Herman-Stahl, 2002), and levels of parents' stress have been found to mediate the effects of family socio-economic characteristics and community violence on children's behavior problems (Linares et al., 2001). Positive links between social support and parenting behavior are attenuated when parents live in primarily low-income and high-crime neighborhoods (Ceballo & McLoyd, 2002). The negative effects of community disadvantage on adolescent boys' mental health have been found to be mediated through parent behaviors such as monitoring, discipline, and communication (Simons et al., 1996).

Parents also can play a role in moderating the effects schools have on their children. Family-level resources can compensate for poor resources at school (Parcel & Dufur, 2001). For example, family social capital is more strongly associated with children's internalizing and externalizing problems than school social capital: parental monitoring, parent knowledge of child's friends, and family attendance at church have dampening effects on behavior problems, whereas teacher-student ratios, school social problems, and perceptions of teachers as caring are not significantly associated with behavior problems (Parcel & Dufur, 2001).

However, the extent to which parents, whose domain is clearly the home and neighborhood, can influence the behaviors of their children in a separate domain, that of school, is not fully understood. There is some evidence that parents' reach does not extend to the school context: Parent involvement in adolescents' school-related activities was not associated with students' levels of misbehavior at school (Otto & Atkinson, 1997). Parents' influence on their children's mental health and behavior in the context of school requires further study.


Without appropriate datasets that include appropriate sampling strategies as well as valid and reliable measures of conceptually key constructs at both individual and context levels, it was impossible until recently to follow Bronfenbrenner's (1977, 1986) exhortation to examine development ecologically. With increasingly sophisticated analytic techniques, academic and social-emotional success have been simultaneously linked with the quality of family, peer, school, and neighborhood contexts (Cook et al., 2002). In order to model context effects, it is necessary to have data sets that sample and assess children over time and nested in complex environments. Yet such data pose challenges both in precisely conceptualizing the constructs of interest and in statistically decomposing the impacts of contextual variables on individual level outcomes. Other recent reviews have detailed the theoretical, methodological, and analytic shortcomings in the literatures on neighborhood and school impacts on child development (Duncan, & Raudenbush, 1999; Sampson, Morenoff, & Gannon-Rowley, 2002). We wish in this section to highlight two of the most important of these issues.

Operationalization of Neighborhoods

Deciding and defining what constitutes a neighborhood (or neighborhoods) of interest are key initial steps in the study of neighborhood impacts. We discuss several issues related to operationalizing neighborhoods here.

Objective Neighborhood Characteristics

Many studies of neighborhood effects rely exclusively on census-based, tract-level demographic characteristics of neighborhoods (e.g., poverty level, racial heterogeneity, residential instability) and use such characteristics as proxies for neighborhood disorganization and danger (Aber, 1994). Although the census data are important because they are comprehensive, publicly available, and include important demographic characteristics at the tract level, these characteristics are largely used by researchers as proxies for particular social characteristics of interest, including level of violence and neighborhood cohesion (Aber, 1994). Census definitions of neighborhoods as "tracts" have been criticized as not representing the ways in which the people who live in these neighborhoods define them, thus implying that meaningful measures of neighborhood boundaries must be gleaned from the perceptions of those individuals who live there (Small & Supple, 1999).

Studies using census-based sources of neighborhood characteristics typically have two main problems (Duncan & Raudenbush, 1999, 2001): (1) they use only one neighborhood characteristic (e.g., tract poverty rate) as an index of neighborhood influence, without taking into account other neighborhood dimensions that might be correlated with that one characteristic (such as crime, drug activity, collective efficacy, and school quality); and (2) they propose to test process models that are not readily testable by Census-based data. These drawbacks can be addressed in analytic steps recommended by Duncan and Aber (1997). First, descriptive associations between neighborhood characteristics and developmental outcomes should be presented. Second, these associations should be adjusted for effects of correlated family-level influences. Third, the potential for family processes to mediate neighborhood effects should be examined. Finally, developmental models should be examined within both resource-rich and resource-poor neighborhoods.

We wish to illustrate from our own work an alternative approach to using individual Census-based indicators. We have taken a multi-stage approach to characterizing the neighborhoods of the New York City youth in our study (Gershoff, Pedersen, Ware, & Aber, 2004). The 893 youth for whom we had neighborhood data resided in 369 census tracts across the city; with an average of only 2.4 participants per tract, we were more interested in the type of neighborhoods in which youth lived, rather than their physical address. Based on recommendations that factors and/or clusters of neighborhood quality will be more reliable than individual indicators (Duncan & Aber, 1997) and on previously successful efforts at using cluster analysis to characterize types of neighborhoods (Anehensel & Sucoff, 2001; Sampson et al., 1997), we thus decided to create clusters of neighborhood types within which youth would be nested. This approach entailed coding youths' home addresses into census tracts, linking tracts to socio-economic, demographic, public health and public safety data, factor analyzing the neighborhood variables, and finally clustering the factors to obtain neighborhood types.

We used socio-economic and demographic variables from the 2000 Decennial U.S. Census Neighborhood Change Database (NCDB) developed by the Urban Institute for Geolytics, Inc., and public health, public services, and public safety variables from data maintained by Community Studies of New York, Inc. ( and by the New York City government ( gov). We submitted 17 variables to a factor analysis and obtained four factors: (1) socio-economic status, race-ethnicity, and family structure; (2) housing; (3) recreation and public safety services; and (4) danger (see Table 25.1 for specific items). We first conducted a hierarchical clustering procedure with these four factors and obtained a six cluster solution. We then conducted an iterative clustering procedure in which cases were reassigned to clusters if such reassignment would increase within-cluster homogeneity along the four factors. Through this iterative procedure we obtained a final six cluster solution (see Figure 25.3).

The six clusters that emerged from this analysis varied in size from 29 tracts to 104 tracts. Three of the clusters were comprised of low SES neighborhoods. Neighborhoods in the first cluster (Low SES/ High Danger: n = 29) had, on average, low SES and very high danger. Neighborhoods in the second cluster (Low SES/High Danger/High Services: n = 38) also had, on average, low SES residents and dangerous neighborhoods but also had very high access to services. The third cluster (Low SES/Low Services: n = 89) also contained neighborhoods of low SES but had marginally low access to services and average levels of danger and housing density/stability. The largest cluster (Low Risk/High Density: n = 104) contained neighborhoods that averaged close to mean levels across most dimensions except for high housing stability/ density. Two middle income SES clusters also emerged. The first middle income SES cluster (Low Risk/High Services: n = 43) was comprised of neighborhoods that exhibited very high access to services and low levels of danger. Neighborhoods in the second middle income SES cluster (Low Risk/Low Services: n = 67) had, on average, low housing stability/density, low danger, and marginally low access to services. We thus obtained six distinct types of neighborhoods that include tracts from across the city. The next stages of our research will involve (1) replicating the factor and

TABLE 25.1

Factors and Individual Indicators of Neighborhood Quality

Factor 1: Socio-Economic Status, Race-Ethnicity, and Family Structure

• Median family income (reversed)

• Percent of adults unemployed

• Percent of people in poverty

• Percent of households receiving public assistance

• Percent of adults over 25 without high school diploma or GED

• Percent non-White

• Percent Hispanic

• Percent children under 18 years of age

• Percent female-headed households or subfamilies

Factor 2: Housing

• Residential instability

• Density (number of people per housing unit) Factor 3: Recreation and Public Safety Services

• Recreational facilities (parks, playgrounds, libraries)

• Protective services (police station, firehouse, hospital)

Factor 4: Danger

• Rate of substantiated child maltreatment cases per number of children

• Rate of violent crime per number of residents

• Arrest rate of youth less than 20 years old

• Death rate of youth less than 20 years old

Source: Gershoff et al., 2004.

cluster analyses with all of the Census tracts in New York City so that we might place our sample of tracts within the population of tracts (Gershoff, Pedersen, & Aber, 2005); and (2) nesting youth and their families within the neighborhood clusters to see if the effects of violence exposure and received parenting on youth vary by neighborhood type.

The approach we have described here is innovative in three main ways. First, it incorporates data from multiple sources, supplementing commonly used census demographic and socio-economic


td "g


td "g

Low SES/ High Danger

Low SES/

High Danger/

High Services

Low SES/

Low Services

Low Danger/ High Density

Low SES Services Housing Danger

Low Risk/ Low Risk/

High Low Services Services

Figure 25.3 Clusters of select New York City Neighborhoods (2002-2003). From Gershoff, Pedersen, Ware, and Aber (2004).

indicators with key aspects of neighborhood service availability and danger. We were fortunate that New York City makes a supreme effort to both collect and disseminate up-to-date neighborhood quality statistics at the tract level. By including data on service availability and danger, we were able to make valuable distinctions between neighborhoods with similar income profiles but differing on factors which may prove important to youth development. Second, by combining data reduction through factor analysis with exploratory cluster analyses, we had the analytic power to include a greater number of neighborhood-level indicators than is typically included in studies that use such indicators as separate independent variables. Third, our approach compensates for not sampling at the neighborhood level by clustering participants' home tracts into neighborhood types. This method is a promising way to conduct neighborhood impact analyses that both maximize the ability of researchers to characterize neighborhoods accurately (without sampling at the neighborhood level) and to retain the ability to conduct multilevel analyses of participants nested within types of neighborhoods.

Perceived Neighborhood Characteristics

With increased interest in neighborhood constructs such as collective efficacy, researchers are interested in individuals' perceptions of their neighborhoods, rather than, or in addition to, objective characterizations of neighborhoods such as those from census data. Although neighborhood perception measures can be crucial indices of how individuals experience their neighborhoods, neighborhood data directly from adolescents or their parents is problematic because of shared error variance across reports of parenting and adolescent outcomes (Duncan & Raudenbush, 2001). This problem can be reduced if outside data sources of youth outcomes are used (e.g., test scores, attendance, arrest records). Other solutions can be achieved with the neighborhood-level data itself. One would be to get independent ratings of neighborhoods and pool them to create context-level measures (Duncan & Raudenbush, 2001); Sampson et al. (1997) did so successfully with 15-30 informants per context. A second would be to have multiple informants of perceived neighborhood characteristics, such as collective efficacy or violence. Colder et al. (2000) successfully used multiple raters of perceived neighborhood danger (child, parent) and child aggression (child, teacher, parent) in a structural equation model in order to minimize shared error variance. Unfortunately, parents tend to underreport the extent to which their children are exposed to violence (Ceballo et al., 2001; Kuo et al., 2000), which leads to an attenuation of associations between exposure and child outcomes when parent-reports or averaged parent and child reports are used (Kuo et al., 2000). Kuo et al. (2000) suggest using multiple raters in a multilevel analytic context to preserve the unique contributions of each rater.

Multiple "Neighborhoods"

Although ethnographic studies of youth have revealed that they spend their time in multiple neighborhoods and that studying just their residential neighborhood will not give a complete picture of the contexts in which they live (Burton, 2001), the empirical literature has neglected to consider neighborhood contexts other than that of the youths' home neighborhoods. In our current research, we are examining the impacts of both home and school neighborhoods on youth outcomes through cross-classified multi-level analyses (Gershoff & Aber, 2003). The vast majority of studies of which we are aware use "neighborhood" to mean the neighborhood in which youth reside; thus, we will be among the first to examine the neighborhoods surrounding youths' schools for their effects on youth development. For many children who attend neighborhood public schools, their home and school neighborhoods are the same. Yet in our New York City youth sample, the overlap of home and school neighborhoods is quite small, as high school placement in New York depends not on residence but rather on a combination of test score requirements, choice, and lottery. As a result, we will be able to identify whether the negative impacts of living in a disadvantaged neighborhood are offset by the benefits of attending school in a more advantaged neighborhood.

Single Parenting Becoming the Best Parent For Your Child

Single Parenting Becoming the Best Parent For Your Child

Parenting is a challenging task. As a single parent, how can you juggle work, parenting, and possibly college studies single handedly and still manage to be an ideal parent for your child? Read the 65-page eBook Single Parenting Becoming The Best Parent For Your Child to find out how. Loaded with tips, it can inspire, empower, and instruct you to successfully face the challenges of parenthood.

Get My Free Ebook

Post a comment