Family and Neighborhood Fit or Misfit and the Adaptation of Mexican Americans
Introduction
Numerous studies have shown that neighborhood characteristics are related to adult and child physical, social, and psychological problems (e.g., Dupere and Perkins 2007; Kupersmidt, Giesler, DeRosier, Patterson, & Davis, 1995; Roosa et al., 2005). Most studies have characterized neighborhoods as sources of stress with which residents contend. Despite the utility of stress process models for understanding variations in adjustment due to neighborhood characteristics, these usually are applied mechanistically such that neighborhood characteristics influence individuals who are characterized as relatively passive participants in the process (Roosa, Jones, Tein, & Cree, 2003). In contrast, human development and adaptation clearly unfold because of the constant interaction between individuals and the contexts in which they are embedded (Bronfenbrenner & Ceci, 1994; Lerner, 1983; 1985; Rutter et al., 1997). Thus, adaptation depends upon characteristics of the context, characteristics of the individual, and relationships between characteristics of the context and those of the individual. Missing from most studies of the relations between neighborhood characteristics and individual well-being is the simultaneous consideration of the characteristics of individuals and their neighborhoods.
In the current study, the person-environment fit model was tested using data from a study of 750 Mexican American families with children in 5th grade. Rather than using a variable-centered approach to exploring relationships between neighborhood characteristics, family characteristics, and adjustment, this study used a combination of a family-centered (i.e., using multiple family characteristics simultaneously; Weaver and Kim 2008) and a neighborhood-centered (using multiple characteristics of neighborhoods simultaneously) approach to better understand the complex influences on adjustment. Similar dimensions were used to characterize families and neighborhoods. A socioeconomic dimension included family variables of parent education, income, and number of parents and the neighborhood variables of percent of families living below the poverty line and percent of individuals with a high school or higher level of education. A cultural dimension was represented by mothers' nativity (U.S. versus Mexico) at the family level and, at the neighborhood level, the percent of Hispanics within a neighborhood. In contrast to other studies that tested a person-environment fit model to understand how neighborhood characteristics relate to adjustment, this study used an ethnic homogenous design; in more integrated samples, race/ethnicity and social class often are confounded making it difficult to interpret results (Mertens, 1998). In addition, the sample was very diverse in terms of social class, family structure, cultural orientation, and neighborhood characteristics, whereas most studies of Mexican Americans have focused on low-income, English speaking, inner city residents (Gonzales, Knight, Morgan-Lopez, Saenz, & Sirolli, 2002). We used a variation of the person-centered approach (Bergman & Magnusson, 1997) via latent class cluster analysis to identify clusters of relatively homogenous types of families and of neighborhoods, respectively. Then we integrated the family- and neighborhood-centered methods by using these clusters as variables in hierarchical linear models to determine how well, individually and jointly, they helped explain diversity in adjustment of Mexican American adults and children.
Methods
Sample
Data for this study came from an investigation of the roles of culture and context in the lives of Mexican Americans in a large southwestern metropolitan area (Roosa et al., 2008). Participants were 750 Mexican American students in 5th grade and their families who were selected from schools in ethnically, economically, and linguistically diverse communities. Eligible families met the following criteria: (a) the child and mother agreed to participate; (b) they had a fifth grader attending a sampled school; (c) the participating mother was the child's biological mother, lived with the child, and self-identified as Mexican or Mexican American; (d) the child's biological father was of Mexican origin; (e) the child was not severely learning disabled; and (f) no step-father was living with the child. Of the 570 two-parent families (76.6% of the sample) in the study, 460 (80.7%) fathers were interviewed. On average, mothers were about 36 years old, fathers about 38 years old, and children about 10 years old. Most mothers (69.9%) and fathers (76.7%) were interviewed in Spanish while most children (82.5%) were interviewed in English. Table 1 presents demographic characteristics of families included in the current analyses (n = 738). Twelve families were excluded due to missing data for at least one of the variables used in the creation of the family and neighborhood profiles.
Analytic Procedures
To investigate whether family-neighborhood fit or misfit was related to parent and child psychological outcomes, three-level hierarchical linear models were tested. Multilevel modeling was used to account for the possible non-independency of the variables due to the fact that families (level-1) were nested within neighborhoods (level-2), which, in turn, were nested within school catchment areas (level-3). However, the intraclass correlations (ICC) for psychological problems were generally low for this sample (i.e., mean = .03; range = 0 to .08 for neighborhood-level variances; mean = .01; range = 0 to .03 for school-level variances) across the studied variables. In the multilevel model, family type was the level-1 predictor and neighborhood type was the level-2 predictor. The cross-level interaction of family type and neighborhood type signified fit/misfit effects. Specifically, the interaction evaluated whether neighborhood types had similar or different effects on psychological problems across family types and vice versa. SAS PROC MIXED was used to conduct the analyses.
Results
Family and Neighborhood Latent Classes
Parental education, household income, family structure (1- vs. 2-parent household), and maternal nativity (born in Mexico vs. born in U.S.) were specified as indicators in the family latent class models. Household income and parental education were allowed to be correlated within each class. Models consisting of one- through seven-classes were estimated. As shown in Table 2, the six-class solution was supported by both information criterion indices (AIC and SABIC) as being the best fitting model among the admissible solutions, although this solution was not a statistically better fit to the data than the five-class solution (see LMR LRT tests). The primary distinction between the five- and six-class solutions was the emergence of a class comprised of families characterized by the highest parental education and household income levels, a subpopulation rarely examined in research with Mexican Americans. Overall, we viewed the six-class solution as being the most interpretable and conceptually meaningful.
Testing the Relations of Family and Neighborhood Fit/Misfit to Psychological Adjustment
Using the family type (with five levels after combining the two higher income groups) and the neighborhood type (with three levels) variables that resulted from the latent class analyses as the level-1 and level-2 predictors, respectively, and the interaction of the two types as the cross-level predictor, multilevel analyses were conducted on each of the psychological outcomes. Table 5 summarizes the Type III F statistics for the overall fixed effects. Family type appeared to have stronger effects on parent and child psychological problems than neighborhood type. There were six significant main effects for family type but only one significant main effect for neighborhood type. There was one significant cross-level interaction. For significant main effects not conditioned by an interaction, we conducted post-hoc pairwise comparisons of the least-square means using Tukey-Kramer tests (West, Welch, & Galecki, 2006). For the cross-level interaction, with three levels of neighborhood type and five levels of family type, post-hoc comparisons could be done in many different ways. We focused on the simple effects of family type differences for each neighborhood type. Scheffe tests, which are more appropriate for post-hoc comparisons with interaction effects, were conducted to test for significant differences of the simple effects.
Discussion
This study applied a person-environment fit model to examine how family and neighborhood characteristics, separately and jointly, were related to variations in adjustment in a sample of Mexican Americans. Latent class analysis was used to identify six distinct family types based on combinations of parent education, income, family structure, and mother's nativity. Similarly, three distinct neighborhood types were identified based on distributions of families living in poverty, individuals with a high school or higher education, and Hispanics. Multilevel modeling was used to determine the main and interactive effects of family and neighborhood types on parent and child adjustment. Not surprisingly, family type was a better direct predictor of adjustment. The value of including neighborhood quality in studies of adjustment is the possibility of adding to the variance explained by models using information about more proximal environments (Leventhal & Brooks-Gunn, 2000). Neighborhoods are distal influences while the family is both more proximal to the individual and the primary developmental context for children (Roosa et al., 2003). Moreover, for ADHD, neighborhood effects were dependent on the characteristics of resident families, or vice versa.
Neighborhood type had a direct effect only on mother's depression. Interestingly, Mexican American mothers in Middle SES neighborhoods reported greater depression than those in Upper SES neighborhoods; mothers from Lower SES neighborhoods had scores between these groups. Mothers in Upper SES neighborhoods likely had resources that made meeting family needs easy while experiencing few contextual stressors. Lower SES neighborhood mothers likely experienced many contextual stressors but may have received sufficient support from neighbors coping with similar stressors (Suarez-Orozco & Suarez-Orozco, 2001). Social bonds with neighbors similar to themselves may have provided some protection for these mothers. On the other hand, mothers in Middle SES neighborhoods may have struggled to meet the social and economic demands of a middle class lifestyle while being socially isolated because of the norm for independence in such neighborhoods. Furthermore, Middle SES neighborhoods were the most ethnically integrated, perhaps adding to social isolation or reduced social support among neighbors. These are hypotheses to be tested in future studies.
Results of multilevel analyses demonstrated the value of using a family-neighborhood fit model to understand variations in ADHD in Mexican Americans; children's levels of ADHD symptoms depended on the combination of family and neighborhood types for several groups. For instance, Middle Class -- Later Generation children were at greater risk for ADHD problems when their families lived in Middle SES neighborhoods than when they lived in Upper SES neighborhoods (there were too few of these families in Lower SES neighborhoods to reliable comparisons). The privileged status of these children (i.e., high parents' income and education) may have made it difficult for them to fit in with children in neighborhoods where relatively few parents had a high school level education. In addition, because of their later-generation status, few of these children spoke Spanish making it difficult for them to communicate with many of the adults and at least some of their peers in Middle SES neighborhoods.
On the other hand, children in Economically Distressed families reported fewer adaptation problems if they lived in Lower SES neighborhoods. Despite the high number of single parents in this group, Lower SES neighborhoods probably provided these families in which most mothers were born in Mexico with more opportunities for social interaction and access to social support than they would find in other neighborhood types. Relatively high levels of similarity with their neighbors may have made Lower SES neighborhoods relatively comfortable for these families (cf.: Georgiades et al., 2007). Further, the experiences of most mothers in this group in Mexico may have prepared them for survival in Lower SES contexts (Roosa et al., 2002). Although one might expect a protective effect from living in a less risky neighborhood, these families had very low incomes and may have struggled economically and socially in Middle SES neighborhoods. In addition, many of the mothers in these families were Spanish speakers, which would make them more isolated in Middle SES neighborhoods.
Overall, the use of family and neighborhood typologies and a family-neighborhood fit model provided useful insights into the diversity in mental health among Mexican Americans. The family-environment fit model may be particularly helpful in studying adjustment among populations with large portions of immigrants because of the challenges they face adapting to what are often dramatically different living circumstances than they experienced in their home countries. Obviously, more study is needed to help explain the specifics of fit/misfit found in this study. In addition, the family-environment fit model should be tested longitudinally to help determine causality. Still, the results of this study provide a beginning point for developing interventions for different Mexican American subgroups based on both family and neighborhood characteristics.