Annals of Behavioral Medicine

, Volume 43, Issue 1, pp 50–61 | Cite as

Neighborhood Contexts and the Mediating Role of Neighborhood Social Cohesion on Health and Psychological Distress Among Hispanic and Non-Hispanic Residents

Original Article



Neighborhood social cohesion (NSC) may contribute to understanding how neighborhood contexts influence the physical and mental health of residents.


We examined the relation of NSC to self-rated mental and physical health and evaluated the mediating role of NSC on relations between neighborhood socioeconomic status, ethnic composition, and health.


A sample of 3,098 Hispanic and non-Hispanic residents within 597 census tracts in metropolitan Phoenix, Arizona rated their health, psychological distress, and their perceptions of NSC. Census tract estimates provided neighborhood contextual measures.


Neighborhood social cohesion was significantly related to better physical and mental health. Both individually rated NSC and neighborhood-level NSC mediated relations between neighborhood contexts and health outcomes. Substantive findings were consistent across Hispanic and non-Hispanic residents.


The findings have implications for improving ethnic and socioeconomic disparities in physical and mental health through attention to social cohesion among neighborhood residents.


Ethnic composition Hispanics Multilevel mediation Neighborhood social cohesion Self-rated health Psychological distress 

Life in the USA is replete with social and economic inequalities. Health disparities have been increasingly well documented for ethnic minorities and the economically disadvantaged. Relative to higher socioeconomic status (SES), lower SES is associated with poorer health [1, 2], lower life satisfaction [3, 4], and greater psychological distress [5, 6]. Racial and ethnic minorities in the USA achieve lower income and educational levels compared with non-Hispanic Whites, and Hispanics consistently report poorer self-rated physical health compared to Whites [3, 7]. However, despite relative economic and health disadvantage, rates of mental health problems among Hispanics and other ethnoracial minorities do not tend to reflect the expected mental health disadvantage based on socioeconomic group differences [8, 9], suggesting that factors other than SES influence health-related outcomes.

Neighborhood social environments shape and perpetuate health disparities, beyond the influence of individual sociodemographic characteristics. Neighborhood economic disadvantage is associated with poorer self-rated health [10] and poorer mental health [5, 11], including more depressive symptoms [12]. Research on neighborhood contextual factors has tended to focus on neighborhood economic and social disadvantage and has not adequately addressed the influence of protective aspects of collective neighborhood social, cultural, and ethnic neighborhood contexts [13]. One such collective social resource, social cohesion, encompasses strength of social relations, sense of belonging, shared values, common identity, trust, and the existence of equal opportunities versus social exclusion within a community [14]. Social cohesion within neighborhoods has been suggested as a key factor for understanding how neighborhood influences affect mental and physical health [9].

The construct of social cohesion, first characterized by sociologists and now broadly employed, describes social bonds among people in groups that contribute to group formation and ongoing participation [14, 15]. Closely related to social cohesion are the constructs of social capital, the “stock” of social networks and resources [16], and collective efficacy [17], the ability to take collective action in service of the common good. This study employs a broad conceptualization of social cohesion that encompasses norms and sentiments among neighbors, mutual trust among residents, and sense of connection to one’s neighborhood.

Social cohesion and related constructs have been linked to physical and mental well-being. A composite of social cohesion, social control, and respondents’ neighbor-based friendship networks was positively associated with individual self-rated health, and social cohesion partially mediated the relationship between neighborhood concentrated affluence and self-rated health, after controlling for individual social, demographic, and behavioral factors [10]. Collective efficacy was positively associated with self-rated health beyond individual characteristics [18] and protected against mortality among older individuals with serious medical diagnoses [19]. Adolescents who perceive their neighborhoods as socially cohesive report better global self-rated health [20].

A relatively small number of studies have examined aspects of social cohesion and mental health [21]. Social cohesion reduced the deleterious effect of neighborhood income deprivation on mental health [22], while low social capital led to feelings of pessimism and low self-esteem through truncated social networks that provided little support [23]. Yet, social ties are not uniformly protective. Community level stressors may moderate the impact of informal ties among neighbors. In disorganized or disadvantaged communities, such ties are not necessarily protective and may even promote delinquency [24]. Thus, protective effects of social cohesion on health may depend on local neighborhood contexts.

Neighborhood contexts also shape cohesion among neighbors. For example, higher neighborhood SES was associated with higher social capital [25], and concentrated neighborhood disadvantage was associated with lower collective efficacy [17]. The ethnic makeup of neighborhoods is posited to influence health of residents through the impact on processes of social cohesion versus social exclusion. Key aspects of social cohesion—shared cultural identify, sense of belonging, and norms of mutual aid and reciprocity—are often protective characteristics of ethnic minority communities [7] that enable residents to cope with economic challenges and promote health. Homogenous ethnic neighborhoods are linked to positive physical and mental health among Hispanics. Residents of high-density Mexican American and Cuban American neighborhoods reported better health than those in low-density neighborhoods of similar SES [26]. Within similar socioeconomic geographic areas, older Mexican Americans in ethnically homogeneous neighborhoods reported higher self-rated health than those in more diverse neighborhoods [27], and teen birth rates among Latinas were lower in communities high in social capital [28].

However, cohesion among ethnic enclaves may be coupled with social exclusion in the wider community [14, 29]. Health inequalities may be perpetuated through isolation of low income and minority residents from key social resources that support social cohesion [17]. Recent findings linked greater ethnic diversity to depleted social capital across city, state, and national data [30]. Ethnic heterogeneity was associated with lower social capital among neighborhoods in Britain [29]. Yet ethnic heterogeneity was not associated with collective efficacy across small neighborhood designations [31]. The question of how ethnic composition relates to aspects of social cohesion warrants further clarification, and neighborhood processes relating social cohesion to physical and mental health may lead to a greater understanding of how neighborhoods—the contexts in which we live and work, thrive, and endure—favor the health of some above others.

The purpose of this study is to examine relations between perceived neighborhood social cohesion (NSC), residents’ physical and mental health, and sociodemographic neighborhood contextual variables—ethnic composition and average SES. The first aim was to examine the role of neighborhood social cohesion as a promising social collective resource associated with residents’ health that partially accounts for the effects of sociodemographic neighborhood contexts on health. Our second aim was to determine whether neighborhood social cohesion was differentially related to neighborhood contexts and health outcomes for ethnic minority Hispanic and non-Hispanic residents. We tested the following hypotheses:
  1. Hypothesis 1:

    Residents’ perceptions of neighborhood social cohesion would be associated with greater physical and mental health.

  2. Hypothesis 2:

    Residents’ perceptions of neighborhood social cohesion would partially mediate the effects of neighborhood SES and ethnic composition on residents’ self-rated physical and mental health.

To address the second aim, we probed individual ethnicity as a moderator variable in each substantive model we tested in order to examine whether Hispanic/non-Hispanic ethnicity affected our findings.


Participants and Neighborhoods

The sample consisted of 3,098 adult residents of Maricopa County, Arizona who participated in the 2008 Arizona Health Survey, funded by St. Luke’s Health Initiatives. Table 1 provides a side-by-side comparison of demographic characteristics of the survey sample with county demographic characteristics. The use of a landline telephone-based sampling method resulted in underrepresentation of sociodemographic groups that parallels differences in documented rates of landline telephone use collected nationally over a similar time period [32]. Compared to county census data, Hispanic respondents in the sample were underrepresented, and female respondents, respondents with a bachelor’s degree or higher, and older respondents were overrepresented. In order to increase the accuracy of population parameters estimates, we employed sampling weights, an approach customarily employed in population health surveys (e.g., [33]).
Table 1

Demographic characteristics of the 2008 AHS respondents and county characteristics















 Hispanic or Latino




 Non-Hispanic or Latino













 Native American or Alaska Native




 Native Hawaiian or other Pacific Islander








 Some other race





 18–24 years old




 25–34 years old




 35–44 years old




 45–59 years old




 60–74 years old




 75–84 years old




 85+ years old





 Bachelor’s degree or higher




Household income



 Median income




 Mean income




aRace categories add up to greater than the sample size because some respondents chose multiple racial categories

bThe Arizona Health Survey (AHS) sample included adults aged 18 and over

cCounty percentages refer to percentage of total adult population aged 18 and over, unless otherwise specified

dStatistic refers to percentage of population age 25 years and over

In Arizona, Mexican Americans predominantly make up the Hispanic population.1 Among the 405 Hispanics in the sample, 64% (n = 260) considered Spanish to be their first language, and 48% (n = 195) of participants chose to respond to the survey in Spanish. Forty-four percent of Hispanics (n = 179) in the sample were born in the USA and 48% were born in Mexico (n = 195). Of the remaining (n = 31) Hispanic participants, approximately equal proportions were born in Central America (n = 10), South America (n = 10), and Puerto Rico (n = 7).

Neighborhoods were defined using boundaries designated by the US Census Bureau as census tracts: subdivisions that generally follow visible geographic features and are relatively homogeneous units with respect to population characteristics, economic status, and living conditions. Census tracts have been found to yield estimates of contextual effects similar to estimates made using smaller census geographic designations such as census block groups [35]. Of the 663 tracts spanning the county, survey respondents resided in 597 (90%) of tracts. Between 1 and 42 respondents (mean = 5.2, SD = 4.5) resided in each tract.

Descriptive statistics for the 597 census tracts illustrate a great diversity among neighborhoods in socioeconomic characteristics, ethnic composition, and total population (see Table 2). The ranges for neighborhood percentage of the population aged 25 years and older with a bachelor’s degree, median household income, and percent of Hispanic residents varied widely across neighborhoods. The mean census tract population was 4,802 (SD = 2,052) and ranged between 37 and 15,675. The 66 census tracts that were not represented in the survey (i.e., 663 tracts in all minus 597 represented tracts) were smaller, poorer, and less well educated than the 597 represented tracts. They had on average 1,700 fewer residents, a lower median household income by $11,000, and a 7% lower percentage of the population with at least a bachelor’s degree. Unrepresented neighborhood tracts also had a higher mean percentage of Hispanic residents (38.9%) compared to tracts in the sample (23.5%).
Table 2

Descriptive statistics for demographic and key study variables



M (SD)




AHS variable



56.0 (17.6)




Annual household income


8.1 (3.7)




Educational attainment level


5.0 (2.0)




Neighborhood social cohesiona


3.08 (0.48)




Kessler psychological distress


1.57 (0.59)




Self-rated health


3.48 (1.10)




Neighborhood variable

Ethnic composition: Hispanic percentb


23.5 (22.4)




Neighborhood education: percent with BAc


25.6 (15.8)




Neighborhood median household income


50,200 (21,200)




Neighborhood Mean NSC (NSCMean)


3.02 (0.33)




All neighborhood variables except NSC were taken from the 2000 Decennial Census; neighborhood mean NSC scores were first computed separately for each census tract and then averaged across tracts. N differs across variables due to missing data.

aN for NSC excludes respondents who selected “don’t know” as well as missing responses

bEthnic composition was defined as the percentage of Hispanic residents in each census tract

cMeasured as percentage of population 25 years and over with a bachelor’s degree or higher

AHS Arizona Health Survey; NSC Neighborhood social cohesion


The sample was drawn using a list-assisted, random-digit-dialing (RDD) approach to select a representative sample of the residential population by selecting households and residents within households. List-assisted RDD is the current standard method of choice for telephone surveys [36]. Group households with more than nine unrelated persons, institutional facilities, the homeless, and those living in military barracks were excluded from the sample. People living in residences without landline telephones were also excluded.

Of the 96,435 telephone numbers that were selected, 59,424 numbers were determined to be out of scope (e.g., nonresidential or nonworking) and were removed prior to recruitment and interviewing. Interviewers had attempted to contact 37,011 numbers at the time that recruitment goals were reached. Of these, a total of 5,760 (16%) numbers were successfully screened. An additional 7,315 (20%) of these numbers were out of scope. Call attempts yielded 20 ineligible, 6,793 (18%) noncontact, 10,217 (28%) refusals, and 783 (2%) other nonresponse contacts.

Of the 5,760 screened, 3,139 (55%) completed interviews. In each household, screening interviewers identified one adult resident 18 years or older using a selection algorithm [37]. Sixty-five percent of screened numbers were pre-selected for refusal conversion, where further attempts to contact cases were made upon initial refusal. The survey questionnaire was administered by phone interview in English or Spanish and included items assessing demographics, physical, and mental health and perceptions of one’s neighborhood. The adult questionnaire took approximately 29 min to administer in English and 42 min in Spanish. Ninety-seven trained interviewers administered surveys between March 24 and June 9, 2008. The refusal rate for completion of the interview among those who were successfully screened was 31%. Other reasons for nonresponse, accounting for 14% of those screened, were language or hearing problems, illness, and exceeding the designated maximum number of 23 call attempts. An additional 41 cases were removed because they were performed by proxy respondent or were assigned to census tracts outside Maricopa County; the final sample size was 3,098.

To balance for underrepresentation of sociodemographic characteristics and to reduce variance of estimates, person-level sampling weights were developed using a classical design-based approach and a raking procedure. Raking adjusts a dataset so that marginal totals match known population totals on a specified set of variables [38]; raking variables included socioeconomic variables, age, race, and ethnicity. Item imputation was employed for missing raking variables. A detailed description of the sampling and weighting procedures is available online (

Individual Level Measures

Sociodemographic Measures

Participants reported their gender and age in years; age ranged from 18 to 96. Educational attainment was measured by asking “What is the highest grade of education you have completed and received credit for?” Responses were coded into 11 categories of increasing educational attainment, scaled 0 to 10. Ethnicity was assessed by asking respondents to report whether or not they were “Latino or Hispanic.” Non-Hispanics were assigned a code of 0, and Hispanics received a code of 1.

Self-Rated Health

A single-item assessed self-rated global health from the SF-36 [39]. The question “In general, would you say that your health is…” was rated on a five-point scale from poor to excellent. The validity of using this single-item measure of self-rated health has been well established among all major ethnic groups in the USA [40], including Hispanics [41].

Psychological Distress

Six items from the Kessler Psychological Distress Scale measuring depressive and anxious symptoms [42] were used to measure psychological distress. This measure has established reliability and validity for use in community and national samples [42, 43]. Using a time frame of 30 days, respondents were asked how often they felt “nervous,” “hopeless,” “restless or fidgety,” “so depressed so that nothing could cheer you up,” “that everything was an effort,” and “worthless.” Respondents rated the items on a five-point scale (1 = all of the time, 5 = none of the time). Scale scores were constructed using mean scores across reverse-coded items, so that the higher scale scores represent higher psychological distress. Cronbach’s alpha was 0.80.

Neighborhood Social Cohesion

Five items measured the Social Cohesion dimension of the Collective Efficacy Scale [17]. The items ask respondents to rate on a four-point scale how much they agreed or disagreed with statements regarding their neighborhood and neighbors, such as “people in this neighborhood can be trusted,” “people in this neighborhood do NOT share the same values,” and “there are people I can count on in this neighborhood.” A mean scale score across items was created for respondents who gave responses to at least three items in the scale, yielding scores for 97% of all respondents. Cronbach’s alpha was 0.78.

Neighborhood-Level Measures

Neighborhood-level measures were drawn from the 2000 Decennial Census for all census tracts making up Maricopa County from the American FactFinder website (

Neighborhood Socioeconomic Status

Two census tract indicators were combined to create a composite measure of neighborhood SES. Neighborhood educational attainment level was represented using population estimates for the percent of the population age 25 or over who completed at least a bachelor’s degree. Median household income of each census tract measured neighborhood income level. Sums of z scores for the two neighborhood SES indicators provided a composite measure of neighborhood SES.

Ethnic Composition

Ethnic composition was defined as the percentage of Hispanic residents per census tract (percent Hispanic), computed from estimates of population counts in each census tract. Percentages were divided by 10 to rescale the variable on a scale from 0.00 to 10.00. Given that all other minorities (African-American, Native American, and Asian American) constituted only 7% of the population, more complex calculations of ethnic composition were highly correlated with percent Hispanic in each tract.

Statistical Analysis

Path analysis using Mplus in a multilevel framework was used to test hypotheses. Multilevel modeling is appropriate for the handling of hierarchical data because of its ability to appropriately accommodate for non-independence of observations and to simultaneously model individual and community influences contributing to health differences. Mplus computational software additionally has the ability to incorporate sampling weights into inferential analysis and was well suited to meet computational needs. In the current study, level 1 represents data at the individual level, and level 2 represents the neighborhood level.

Mediation terminology is employed, where a paths refer to the relation of independent to mediator variables, b paths refer to the relation of the mediator to dependent variables, and c paths refer to the total relation of independent to dependent variables. A final path, c′, refers to the remaining relationship of independent to dependent variables once intervening mediated relations have been accounted for [44]. The hypothesized mediation model was in the form of level 2→1→1, where mediation went from the independent variable measured at level 2, through the mediator at level 1, to outcomes at level 1.

Centering variables into deviation form leads to interpretable path coefficients in multilevel modeling [45, 46]. Two forms of centering are (a) centering around the grand mean (CGM), the deviation of each score from the grand mean of a variable, and (b) centering within context (CWC), the deviation of each score from its cluster mean. Choice of centering depends on the specific research question [45, 46]. Only CGM is employed at level 2. CWC removes all between-cluster mean differences and is employed at level 1 to estimate the relation of an individual’s standing, relative to his/her cluster, to an individual outcome. CGM for level 1 covariates, appropriate for models with both level 1 and level 2 predictors, provides statistical control at both levels.


Descriptive Analyses

Unweighted descriptive statistics for demographic and key survey and neighborhood variables are presented in Table 2. Neighborhood mean NSC was calculated as the arithmetic mean NSC score of participants in each census tract.2 All individual level and neighborhood variables have skew and kurtosis within the acceptable range for use of maximum likelihood estimation [47]. Very few data points were missing on outcome measures—one respondent on psychological distress and five on self-rated health. On the neighborhood social cohesion scale, 101 (3.3%) scale scores were missing. Intraclass correlations were 0.11 for NSC, 0.03 for self-rated health, and 0.07 for psychological distress.

Table 3 presents bivariate correlations among demographic and key variables measured at the individual level. The two dependent variables were negatively correlated; individually rated NSC was positively correlated with self-rated health and negatively correlated with psychological distress. A series of one-way ANOVAs indicated that Hispanic individuals reported lower NSC than did non-Hispanics (Hispanics, M = 2.86, SD = 0.46; non-Hispanics, M = 3.11, SD = 0.47). Hispanic respondents reported significantly poorer self-rated health (Hispanics, M = 3.2, SD = 1.1; non-Hispanics, M = 3.5, SD = 1.1) and greater psychological distress (Hispanics, M = 1.65, SD = 0.62; non-Hispanics M = 1.55, SD = 0.59) compared to non-Hispanics. Within the Hispanic subsample, a comparison of differences in key variables according to nativity and language of interview revealed similar patterns. Hispanics who were interviewed in Spanish reported lower mean scores on NSC (M = 2.7, SD = 0.40, Spanish vs. M = 3.0, SD = 0.47, English) and self-rated health (M = 2.9, SD = 0.94, Spanish; M = 3.5, SD = 1.1, English) than Hispanics who were interviewed in English. Similarly, foreign-born Hispanics reported lower NSC and self-rated health than US-born Hispanics. There were no differences in psychological distress according to nativity or language use.
Table 3

Correlations among individual level variables









1. Age


2. Gendera



3. Ethnicityb




4. Educational attainment





5. Income level






6. Neighborhood social cohesion







7. Self-rated health







8. Psychological distress








These bivariate correlations are total correlations (i.e., do not take into account clustering within neighborhoods)

*p < 0.05; **p < 0.01; ***p < 0.001

a1 = female, 0 = male

b1 = Hispanic, 0 = non-Hispanic

Hispanic survey respondents resided in 246 of the 597 census tracts. At the neighborhood level, percentage of Hispanic residents was inversely associated with neighborhood educational attainment level (r = −0.70; p < 0.001) and median household income (r = −0.59; p < 0.001). The percentage of Hispanic residents was inversely correlated with mean neighborhood NSC scores (r = −0.48, p < 0.001).

Inferential Analyses

A series of multilevel models was employed to test hypotheses. Individual level gender, education, ethnicity, and age were controlled in all models. All regression coefficients reported are standardized (β) unless noted as unstandardized (B). Census tract number was used as the clustering variable to identify neighborhoods. Sampling weights, described above, were included in all analyses.

Is Neighborhood Social Cohesion Associated with Better Physical and Mental Health?

Hypothesis 1 predicted that individual ratings of NSC would be positively associated with self-rated health and negatively associated with psychological distress. CWC was applied to the predictor and all control variables. In initial model specification, both random intercepts to reflect differences in mean outcomes across neighborhood tracts and random slopes to represent varying relationships of predictors to outcomes across tracts were permitted.

Neither variance component representing slope variation across neighborhood tracts was statistically significant (for self-rated health, τ11 = 0.00, SE = 0.05, p = 0.99; for psychological distress, τ11 = 0.02, SE = 0.02, p = 0.38). Moreover, models with random versus fixed slopes resulted in the same estimates of path coefficients and their statistical significance. We therefore retained the more parsimonious model with fixed slopes. Within neighborhoods, individual ratings of NSC, relative to averages of social cohesion within the neighborhood, were associated with higher self-rated health and lower psychological distress. Standardized coefficients for the effects of NSC were as follows: on self-rated health, β = 0.11, SE = 0.02, p < 0.001, and on psychological distress, β = −0.10, SE = 0.02, p < 0.00. Significant variance components for random intercepts for both self-rated health (τ00 = 3.48, SE = 0.02, p < 0.001) and psychological distress (τ00 = 1.61, SE = 0.01, p < 0.001) indicated that mean scores on outcomes significantly varied across neighborhoods.

Moderation of B Paths by Ethnicity

The interaction between Hispanic ethnicity and NSC in predicting outcomes was included in a subsequent model to test whether NSC functioned equivalently across ethnic groups. A within-level interaction term was computed by multiplying individual Hispanic ethnicity (1 = Hispanic, 0 = non-Hispanic) with NSC. Hispanic ethnicity moderated the effects of NSC on self-rated health (B = −0.17, SE = 0.03, p < 0.001, unstandardized), suggesting that the relation varied across ethnic groups. However, upon estimating the relation within each ethnic subgroup, we found that NSC was positively associated with self-rated health for both groups, and the magnitude of coefficient was only slightly larger among Hispanics (B = 0.42, SE = 0.17, p = 0.01, unstandardized) compared to non-Hispanics (B = 0.28, SE = 0.06, p = 0.001, unstandardized). The interaction term for the prediction of psychological distress approached significance (B = 0.03, SE = 0.02, p = 0.09, unstandardized). The path coefficient for the effect of NSC on psychological distress was in the same direction and slightly larger among Hispanics (B = −0.37, SE = 0.09, p < 0.001, unstandardized) compared to non-Hispanics (B = −0.16, SE = 0.04, p < 0.001, unstandardized).

Does Neighborhood Social Cohesion Mediate the Effects of Neighborhood Contexts on Health Outcomes?

Hypothesis 2, the test of NSC as a mediator carrying the influence of neighborhood SES and ethnic composition on residents’ self-rated physical and mental health, was tested in three steps. First, we estimated the c paths, or total effects, of neighborhood independent variables on dependent variables. Second, a paths, or the cross-level effects of neighborhood independent variables (level 2 neighborhood SES and ethnicity) on level 1 NSC were tested. In a third step, level 1 health outcomes were regressed on the mediator (level 1 NSC) and all level 2 neighborhood variables (SES, ethnic composition, neighborhood mean NSC) to yield b and c′ path estimates. Mediated effects (ab) were calculated and asymmetric confidence limits for the mediated effects were obtained using the PRODCLIN program [48].

Step 1: Total Effects of Neighborhood Independent Variables on Outcomes (c Paths)

The c paths of the mediation analysis characterize the relationship of neighborhood-level independent variables (neighborhood SES, percent Hispanic) to individual level health outcomes. We used CGM centering of level 1 and level 2 predictor and control variables. Due to high collinearity between neighborhood SES and percent Hispanic (r = −0.68, p < 0.001), with both variables entered simultaneously into a single contextual model, only the effects of neighborhood SES were statistically significant (on self-rated health, c = 0.62, SE = 0.19, p = 0.001; on psychological distress, c = −0.43, SE = 0.17, p = 0.01), over and above the influence of individual characteristics. Percent Hispanic did not contribute to the prediction of psychological distress (c = 0.04, SE = 0.18, p = 0.81), but the negative association with self-rated health approached significance (c = −0.31, SE = 0.17, p = 0.07). In neighborhoods with a greater percentage of Hispanic residents, ratings of self-rated health were marginally poorer than those in more homogenously non-Hispanic neighborhoods.

Moderation of c Paths by Ethnicity

To determine whether the total effects of neighborhood contextual variables on outcomes differed between Hispanic and non-Hispanic residents, cross-level interaction terms were created by multiplying level 1 Hispanic ethnicity with CGM centered level 2 neighborhood SES and percent Hispanic, respectively. These interaction terms were added to the previous model. No interaction was significant; Hispanic ethnicity did not moderate the relation between neighborhood SES (on self-rated health, B = −0.03, SE = 0.06, p = 0.60; on psychological distress, B = −0.02, SE = 0.03, p = 0.59) or percent Hispanic (on self-rated health, B = 0.005, SE = 0.05, p = 0.91; on psychological distress, B = −0.02, SE = 0.03, p = 0.62).

Step 2: Effects of Neighborhood Independent Variables on Mediator (a Paths)

In tests of the association between neighborhood independent variables and the mediator, NSC, a path coefficients for both neighborhood SES and percent Hispanic were statistically significant in predicted directions. Neighborhood SES was associated with greater NSC (a = 0.26, SE = 0.09, p < 0.01), and greater percent Hispanic was associated with lower NSC (a = −0.61, SE = 0.10, p < 0.01), with the influence of level 1 covariates controlled. Neighborhood SES and percent Hispanic accounted for unique portions of variance in NSC.

Moderation of a Paths

Tests of interaction terms between each neighborhood context and individual ethnicity suggested that the relation between neighborhood SES and NSC was similar across ethnic groups (B = 0.08, SE = 0.06, p = 0.18, unstandardized). The relation of percent Hispanic to social cohesion was significantly different between Hispanics and non-Hispanics (B = 0.08, SE = 0.03, p = 0.007, unstandardized). The association was negative in both groups but differed slightly in magnitude between non-Hispanics (B = −0.05, SE = 0.01, p < 0.001, unstandardized), compared with Hispanics (B = −0.03, SE = 0.02, p = 0.04, unstandardized).

Step 3: Tests of the Mediated Effects (b, c′, and ab Paths)

A final set of multilevel equations predicting outcomes from the mediator, NSC, and both neighborhood independent variables was specified to obtain estimates for b and c′ paths. In the form of a contextual model [46], we entered aggregated mean NSC, estimated as the arithmetic mean of social cohesion within a neighborhood (NSCMEAN) at level 2. This contextual model yielded two distinct, non-overlapping sources of mediation. First was the hypothesized mediation through perceived NSC at the individual level (i.e., within level 1). We additionally tested mediation through level 2 NSCMEAN, which yielded an estimate of the extent to which the ambient level of social cohesion within each neighborhood mediated the relationship of neighborhood characteristics (SES and percent Hispanic) to health outcomes. Level 1 and level 2 control variables were centered using CGM. Results of the mediation model are displayed graphically in Fig. 1. Each arrowed line designates a statistically significant path, and all path coefficients are standardized. Note that in addition to individually rated NSC, NSCMEAN related to self-rated health (γ01 = 0.69, SE = 0.14, p < 0.001) and psychological distress (γ01 = −0.73, SE = 0.13, p < 0.001), indicating that both within and between neighborhood differences in NSC predicted outcomes.
Fig. 1

Mediation model. Figures are regression path coefficients. Solid lines depict paths that were statistically significant at the p < 0.05 level; dashed lines represent paths that were not statistically significant. The estimates for individual level 1 control variables are not displayed. NSCMean neighborhood social cohesion, neighborhood tract mean level; NSCIndividual neighborhood social cohesion, individual level

As Fig. 1 shows, all four of the mediated effects in the mediation model through level 1 NSC were statistically significant. For the mediated effects between neighborhood SES and outcomes, the mediator NSC accounted for a significant portion of the relation of neighborhood SES to self-rated health (ab = 0.03, SE = 0.01, p < 0.05) and psychological distress (ab = −0.03, SE = 0.01, p < 0.05). Because mediated paths included a combination of positive and negative effects, proportions of the total effects mediated, ab/(c′ + ab), were computed with absolute values [49]. NSC accounted for 7% of the relation with self-rated health and 10% with psychological distress. Level 1 NSC significantly mediated the relation between ethnic composition and self-rated health (ab = −0.07, SE = 0.02, p < 0.01); mediation was complete (i.e., accounted for 96% of the relationship). For the mediated effect between ethnic composition and psychological distress, the mediated effect was significant (ab = 0.06, SE = 0.02, p < 0.01) and accounted for 24% of the total effect.

We performed a second analysis to estimate the mediation through NSCMEAN, that is, level 2→2→1. Level 2 NSCMEAN partially mediated the effects of percent Hispanic on self-rated health (ab = −0.27, SE = 0.07, p < 0.01) and psychological distress (ab = 0.28, SE = 0.06, p < 0.05) and partially mediated the effects of neighborhood SES on self-rated health (ab = 0.08, SE = 0.04, p < 0.01) and psychological distress (ab = −0.09, SE = 0.04, p < 0.05).3 The full model, as shown in Fig. 1, with between and within class effects of NSC, accounted for 11% of the variance in self-rated health and 6% in psychological distress.

Further Exploration of Ethnic Composition and NSC

In a final supplemental analysis, we explored the possibility of a curvilinear component to the relation between ethnic composition and NSC, reflecting a potential increase in NSC among more densely Hispanic communities. A squared (or quadratic) term was created from centered (CGM) percent Hispanic; the quadratic model predicted NSCMEAN. Both the overall negative first-order effect between percent Hispanic and NSC (β = −0.37, SE = 0.15, p < 0.01) and a positive second-order effect of the squared term (β = −1.1, SE = 0.13, p < 0.001) were statistically significant. The quadratic term accounted for 3.8% of variance in NSCMEAN beyond the linear effect. The negative trend in mean neighborhood NSC with increasing percent Hispanic decreased in slope above approximately 40% Hispanic and appeared relatively flat above 70% Hispanic. There was no evidence of an increase in NSC at high percent Hispanic; however, sparseness of neighborhoods with high percent Hispanic may have limited the potential to observe this effect.


The results provide evidence in support our hypotheses; NSC appears to serve as a protective factor for health and mental health among residents and that factor accounts, in part, for differences in health of residents associated with neighborhood SES and ethnic diversity. The present investigation is one of a handful of studies [19, 22, 50] that found support for NSC or related constructs as a mediator of neighborhood effects on individual outcomes and the only study, of which we are aware, to examine both self-rated physical and mental health. Consistent with research associating SES of geographic area with physical [10, 50] and mental health [5, 12], we found neighborhood SES to be associated with higher self-rated health and lower psychological distress of residents, over and above individual SES. The mediated effects through NSC explain a dimension of neighborhood risk and resilience. Educational and income resources within neighborhoods bolster levels of NSC, which in turn are associated with better physical and mental health. Conversely, lower income, less educated, and more ethnically heterogeneous neighborhoods house residents with poorer mental and physical health. Lower levels of social cohesion in these neighborhoods are an important part of this association and highlight the key role of social processes in the link between neighborhoods and health.

The neighborhood context of ethnic makeup yielded perhaps the most striking findings. Within a major Southwestern metropolitan area, where the presence of Hispanic cultures is central to the cultural and sociopolitical landscape, increasing percentage of Hispanics in neighborhoods was associated with poorer self-rated health among individuals through its negative relation to perceived NSC. This negative relation was clearly evidenced across Hispanics and non-Hispanics alike. The findings are consistent with a large study of social capital across US communities of varying sizes from very large (states or nations) to very small rural communities. People in more ethnically diverse areas report lower trust in their neighbors, regardless of the ethnicity of the respondent [30], but they also report less trust in people with ethnic backgrounds like their own. Putnam observes that ethnic diversity is often associated with a withdrawal from community life, a “hunkering down,” rather than inter-ethnic conflict. That defensive posture appears to be present most often in Arizona neighborhoods of mixed ethnicity.

A subtle curvilinear effect suggested that the significant negative relationship between percent Hispanic and NSC did not hold at the high end of the range of percent Hispanic (above about 70%), where there was an absence of relationship between NSC and percent Hispanic. Considering that neighborhoods with the greatest concentration of Hispanic residents were underrepresented, the question remains whether we would have observed an increase in NSC in neighborhoods with the highest proportion of Hispanic residents given sufficient representation of homogeneous Hispanic neighborhoods. Although social withdrawal in the presence of increasing diversity was observed among Hispanics in our study, we may not have been able to detect the positive social effects of highly homogeneous groups residing in small Hispanic enclaves. Protective social processes for mental health occur within very small geographic designations [51].

In addition, patterns of migration, acculturation, and social mobility are important dimensions of Hispanic heterogeneity that further define the complex nature of ethnic diversity in Maricopa County. Mexicans from lower socioeconomic strata in the community of origin tend to immigrate to the USA, and subsequent generations are highly socially mobile [52]. Across metropolitan Phoenix, lower neighborhood SES is associated with greater percentage of Hispanic residents. This may be an indication that with acculturation and movement up the socioeconomic ladder, Hispanics move into higher SES neighborhoods with lower concentrations of Hispanic residents and report an increased sense of connection with their neighborhoods. Indeed, social cohesion ratings in this sample were lower for Mexican-born respondents and those who were interviewed in Spanish compared to Hispanics that were born in the USA and completed the survey in English.

The present findings provide strong evidence that NSC is associated with residents’ physical and mental health. Consistent with prior studies, individually rated NSC was associated with higher self-rated health [10, 20] and lower psychological distress [22], with all between neighborhood differences partialed out. Aggregated mean NSC garnered effects at the neighborhood level as well, independent from the effects of individual level NSC. This finding suggests that not only may individuals’ ratings of NSC be protective but perceptions of NSC among one’s neighbors may protect health as well. NSC was similarly protective within and across neighborhoods, whether disadvantaged, affluent, ethnically diverse, or homogenous.

Several potential mechanisms have been posited linking social cohesion to health and well-being. One potential mechanism is the enhancement of collective efficacy in neighborhoods with greater social cohesion. Through the neighborhood context of mutual trust and shared values, neighborhood residents increase their expectations that together they can achieve common goals [13]. As such, healthy lifestyle behaviors are promoted through collective efforts to protect safe public spaces for activity, clean and safe housing, and availability of nutritional foods. Cohesive neighborhoods also foster a sense of community, which can be considered an affective component of social cohesion that positively impacts quality of life and may benefit self-rated physical and mental health directly.

Despite lower SES and a trend for reporting worse self-rated health and higher psychological distress, Mexicans and Mexican Americans show a mortality [52] and psychiatric illness advantage among Mexican-origin Hispanics [53]. This has been attributed to protective cultural factors such as prosocial norms fostering cohesion and the migration of disproportionately healthy individuals [52]. In this study, NSC ratings among Hispanics did not demonstrate a culturally specific protective effect. We found that within neighborhoods, social cohesion was protective of physical and mental health for both Hispanics and non-Hispanics. Ratings of social cohesion among one’s neighbors may be distinct from social cohesion within one’s cultural group.

Limitations and Future Directions

Certain limitations to representativeness of the survey sample are inherent in the telephone-based sampling design utilized in the study. Underrepresentation of Hispanic, male, young adult, and less educated residents among the sample places limits on the generalizability of the findings, but those effects may be relatively minor [32]. Due to the study’s cross-sectional design, the association may reflect bidirectional causality, where higher social cohesion may lead to more positive self-reported health, but healthier people also may tend to rate their neighborhoods as more socially cohesiveness. A strength of the data analysis was the use of sampling weights, a methodological strategy designed to reduce bias in population parameter estimates.

The link between social cohesion and health is not confined to a single pathway, and more detailed inquiry is needed in future studies. Comparative analyses of the role of social cohesion in specific neighborhoods, including ethnographic or mixed qualitative and quantitative investigations and longitudinal studies to observe the development and maintenance of NSC and to establish its causal effects on health, are warranted. Such research should seek to specify how processes of social cohesion and social exclusion, in the context of ethnic diversity, influence health and well-being, and shape health inequalities. This study was limited by the use of a measure of physical health that was self-reported and measured by a single item. Further inquiry should also seek to determine whether the protective benefits of social cohesion extend to health indicators such as body mass index, health behaviors, or other objective outcomes of health and disease.

Undoubtedly, there are other characteristics of urban settings that are relevant to the way that ethnic composition shapes NSC that were not included in the study. Our measure of ethnic composition, the percentage of Hispanic residents in each neighborhood, served as a proxy measure for ethnic diversity. Percent Hispanic provides a straightforward measure of ethnic composition in the Hispanic southwest but is limited in its ability to take into consideration the extent of integration versus segregation between people of different backgrounds within neighborhoods, an aspect of ethnic diversity that some posit is the root of health inequalities [54]. Neighborhood contexts such as availability of parks and public spaces, the extent of illegal or violent activities, the extent of civic involvement and volunteerism, and the social and political history of the area are highly relevant to consider in understanding how neighborhood contexts shape perceptions of NSC, health, and well-being.

Taken together, we consider these findings to be strong evidence for the health protective benefits of NSC, whether conceived as individuals’ perceptions or a community’s aggregate perceptions, within the context of an increasingly diverse society. We found evidence linking ethnic diversity to poorer self-rated health and well-being through lower social cohesion, yet over time the community benefits of ethnic diversity are positive [30] and these conditions are not likely immutable. Recent advances in resilience research offer guidance on ways to enhance trust and mutually beneficial social connection in community settings [55, 56].

A sense of the societal benefit of socially cohesive neighborhoods is intuitive, yet the direct health benefits may not be obvious to policy makers or health practitioners. This study has implications for public policy and health promotion initiatives to foster the development of cohesive, resilient communities [56]. For example, identifying pathways to enhance NSC through intervention or policy or bolstering existing population health initiatives by adding a component to facilitate social cohesiveness may be worthwhile future directions. In the Southwestern United States, the current sociopolitical context includes heated conflict over immigration during a troubling economic downturn. This situation, with its deep historical roots, poses a great challenge to social cohesiveness among area residents. Perhaps along with current challenges comes a great opportunity for community and cultural leaders, as well as academic institutions, to develop ways to foster social cohesion as a resource for promoting resilience in these trying times.


  1. 1.

    Similar to national census data on racial categories endorsed by Hispanics [34], the majority of Hispanics did not endorse a racial category provided; 57% of Hispanics selected “other race” (N = 164), “don’t know” (N = 60), or refused to respond (N = 7). Thirty-seven percent identified themselves as White, 4% selected Native American, 1% selected Asian, 0.5% selected African-American, and 0.5% selected Pacific Islander categories.

  2. 2.

    For the 93 (15.6%) neighborhood tracts for which there was one AHS respondent, neighborhood mean NSC was represented by the individual’s score.

  3. 3.

    Mediation models were re-estimated excluding the 93 tracts containing only one respondent per neighborhood, and all results were replicated. This subsample consisted of 503 neighborhood tracts and 2,878 residents.



The authors acknowledge St. Luke’s Health Initiatives for providing access to the 2008 Arizona Health Survey data. We also thank Manuel Barrera, Jr., Mary C. Davis, and John Hall for their valuable comments on an earlier version of this manuscript.

Conflict of Interest Statement

The authors have no conflict of interest to disclose.


  1. 1.
    Adler NE, Ostrove JM. Socioeconomic status and health: What we know and what we don't. Ann N Y Acad Sci. 1999;896:3–15.PubMedCrossRefGoogle Scholar
  2. 2.
    Marmot MG. The status syndrome: How social standing affects our health and longevity. New York: Henry Holt & Co.; 2004.Google Scholar
  3. 3.
    Markides KS, Eschbach K. Aging, migration, and mortality: Current status of research on the Hispanic paradox. J Gerontol B Psychol Sci Soc Sci. 2005;60 Spec No 2:68–75PubMedCrossRefGoogle Scholar
  4. 4.
    Markus HR, Ryff CD, Curhan KB, Palmersheim KA. In their own words: Well-being at midlife among high school-educated and college-educated adults. In: Brim OG, Ryff CD, Kessler RC, eds. How healthy are we? A national study of well-being at midlife. Chicago: The University of Chicago Press; 2004:373–397.Google Scholar
  5. 5.
    Silver E, Mulvey EP, Swanson JW. Neighborhood structural characteristics and mental disorder: Faris and Dunham revisited. Soc Sci Med. 2002;55(8):1457–1470.PubMedCrossRefGoogle Scholar
  6. 6.
    Turner RJ, Lloyd DA. The stress process and the social distribution of depression. J Health Soc Behav. 1999;40(4):374–404.PubMedCrossRefGoogle Scholar
  7. 7.
    Angel JL, Angel RJ. Minority group status and healthful aging: Social structure still matters. Am J Public Health. 2006;96(7):1152–1159.PubMedCrossRefGoogle Scholar
  8. 8.
    Ryff CD, Keyes CLM, Hughes DL. Psychological well-being in MIDUS: Profiles of ethnic/racial diversity and life-course uniformity. In: Brim OG, Ryff CD, Kessler RC, eds. How healthy are we? A national study of well-being at midlife. Chicago: The University of Chicago Press; 2004:398–424.Google Scholar
  9. 9.
    Snowden LR. Racial, cultural and ethnic disparities in health and mental health: Toward theory and research at community levels. Am J Community Psychol. 2005;35(1–2):1–8.PubMedCrossRefGoogle Scholar
  10. 10.
    Wen M, Browning CR, Cagney KA. Poverty, affluence, and income inequality: Neighborhood economic structure and its implications for health. Soc Sci Med. 2003;57(5):843–860.PubMedCrossRefGoogle Scholar
  11. 11.
    Muntaner C, Eaton WW, Miech R, O’Campo P. Socioeconomic position and major mental disorders. Epidemiol Rev. 2004;26:53–62.PubMedCrossRefGoogle Scholar
  12. 12.
    Ross CE. Neighborhood disadvantage and adult depression. Health and Social Behavior. 2000;41:177–187.CrossRefGoogle Scholar
  13. 13.
    Sampson RJ. The neighborhood context of well-being. Perspectives in Biology and Medicine. 2003;46(3 suppl):S53–S64.PubMedGoogle Scholar
  14. 14.
    Berger-Schmitt R. Considering social cohesion in quality of life assessments: Concept and measurement. Social Indicators Research. 2002;58:403–428.CrossRefGoogle Scholar
  15. 15.
    Friedkin NE. Social cohesion. Annual Review of Sociology. 2004;30:409–425.CrossRefGoogle Scholar
  16. 16.
    Lynch J, Kaplan G. Understanding how inequality in the distribution of income affects health. Journal of Health Psychology. 1997;2:297–314.PubMedCrossRefGoogle Scholar
  17. 17.
    Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: A multilevel study of collective efficacy. Science. 1997;277(5328):918–924.PubMedCrossRefGoogle Scholar
  18. 18.
    Browning C, Cagney K. Collective efficacy and health: Neighborhood social capital and self-rated physical functioning in an urban setting. Journal of Health and Social Behaviors. 2002;43(4):383–399.CrossRefGoogle Scholar
  19. 19.
    Wen M, Cagney KA, Christakis NA. Effect of specific aspects of community social environment on the mortality of individuals diagnosed with serious illness. Soc Sci Med. 2005;61(6):1119–1134.PubMedCrossRefGoogle Scholar
  20. 20.
    Abada T, Hou F, Ram B. Racially mixed neighborhoods, perceived neighborhood social cohesion, and adolescent health in Canada. Soc Sci Med. 2007;65(10):2004–2017.PubMedCrossRefGoogle Scholar
  21. 21.
    Whitley R, McKenzie K. Social capital and psychiatry: Review of the literature. Harv Rev Psychiatry. 2005;13(2):71–84.PubMedCrossRefGoogle Scholar
  22. 22.
    Fone D, Dunstan F, Lloyd K, Williams G, Watkins J, Palmer S. Does social cohesion modify the association between area income deprivation and mental health? A multilevel analysis. Int J Epidemiol. 2007;36(2):338–345.PubMedCrossRefGoogle Scholar
  23. 23.
    Cattell V. Poor people, poor places, and poor health: The mediating role of social networks and social capital. Soc Sci Med. 2001;52(10):1501–1516.PubMedCrossRefGoogle Scholar
  24. 24.
    Dupere V, Perkins DD. Community types and mental health: A multilevel study of local environmental stress and coping. Am J Community Psychol. 2007;39(1–2):107–119.PubMedCrossRefGoogle Scholar
  25. 25.
    Steptoe A, Feldman PJ. Neighborhood problems as sources of chronic stress: Development of a measure of neighborhood problems, and associations with socioeconomic status and health. Ann Behav Med. 2001;23(3):177–185.PubMedCrossRefGoogle Scholar
  26. 26.
    Eschbach K, Ostir GV, Patel KV, Markides KS, Goodwin JS. Neighborhood context and mortality among older Mexican Americans: Is there a barrio advantage? Am J Public Health. 2004;94(10):1807–1812.PubMedCrossRefGoogle Scholar
  27. 27.
    Patel KV, Eschbach K, Rudkin LL, Peek MK, Markides KS. Neighborhood context and self-rated health in older Mexican Americans. Ann Epidemiol. 2003;13(9):620–628.PubMedCrossRefGoogle Scholar
  28. 28.
    Denner J, Kirby K, Coyle K, Brindis C. The protective role of social capital and cultural norms in Latino communities: A study of adolescent births. Hispanic Journal of Behavioral Sciences. 2001;23(1):3–21.CrossRefGoogle Scholar
  29. 29.
    McCulloch A. An examination of social capital and social disorganisation in neighbourhoods in the British household panel study. Soc Sci Med. 2003;56(7):1425–1438.PubMedCrossRefGoogle Scholar
  30. 30.
    Putnam R. E pluribus unum: Diversity and community in the twenty-first century. The 2006 Johan Skytte Prize Lecture. Scandinavian Political Studies. 2007;30(2):137–174CrossRefGoogle Scholar
  31. 31.
    Duncan TE, Duncan SC, Okut H, Strycker LA, Hix-Small H. A multilevel contextual model of neighborhood collective efficacy. Am J Community Psychol. 2003;32(3–4):245–252.PubMedCrossRefGoogle Scholar
  32. 32.
    Blumberg SJ, Luke JV. Reevaluating the need for concern regarding noncoverage bias in landline surveys. Am J Public Health. 2009;99(10):1806–1810.PubMedCrossRefGoogle Scholar
  33. 33.
    Botman SL, Moore TF, Moriarty CL, Parsons VL, eds. Design and estimation for the National Health Interview Survey, 1995–2004. Hyattsville: National Center for Health Statistics; 2000 Vital Health Statistics.Google Scholar
  34. 34.
    Tafoya S. Shades of belonging. Retrieved from Pew Hispanic Center website:; 2004.
  35. 35.
    Borrell LN, Diez Roux AV, Rose K, Catellier D, Clark BL, Atherosclerosis Risk in Communities Study. Neighbourhood characteristics and mortality in the Atherosclerosis Risk in Communities Study. Int J Epidemiol. 2004;33(2):398–407.PubMedCrossRefGoogle Scholar
  36. 36.
    Tucker C, Lepkowski J, Piekarski L. The current efficiency of list-assisted telephone sampling designs. Public Opinion Quarterly. 2002;66:321–338.CrossRefGoogle Scholar
  37. 37.
    Rizzo J, Brick JM, Park I. A minimally intrusive method for sampling persons in random digit dial surveys. Public Opinion Quarterly. 2004;68(2):267–274.CrossRefGoogle Scholar
  38. 38.
    Battaglia MP, Izrael D, Hoaglin DC, Frankel MR. Practical considerations in raking survey data. Accessed June 23, 2010.
  39. 39.
    Ware JE, Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care. 1992;30(6):473–483.PubMedCrossRefGoogle Scholar
  40. 40.
    McGee DL, Liao Y, Cao G, Cooper RS. Self-reported health status and mortality in a multiethnic US cohort. Am J Epidemiol. 1999;149(1):41–46.PubMedGoogle Scholar
  41. 41.
    Finch BK, Hummer RA, Reindl M, Vega WA. Validity of self-rated health among Latino(a)s. Am J Epidemiol. 2002;155(8):755–759.PubMedCrossRefGoogle Scholar
  42. 42.
    Kessler RC, Andrews G, Colpe LJ, et al. Short screening scales to monitor population prevalences and trends in non-specific psychological distress. Psychol Med. 2002;32(6):959–976.PubMedCrossRefGoogle Scholar
  43. 43.
    Furukawa TA, Kessler RC, Slade T, Andrews G. The performance of the K6 and K10 screening scales for psychological distress in the Australian National Survey of Mental Health and Well-Being. Psychol Med. 2003;33(2):357–362.PubMedCrossRefGoogle Scholar
  44. 44.
    MacKinnon DP. Introduction to statistical mediation analysis. New York: Lawrence Erlbaum Associates; 2008:477.Google Scholar
  45. 45.
    Enders CK, Tofighi D. Centering predictor variables in cross-sectional multilevel models: A new look at an old issue. Psychol Methods. 2007;12(2):121–138.PubMedCrossRefGoogle Scholar
  46. 46.
    Kreft IGG, de Leeuw J, Aiken LS. The effect of different forms of centering in hierarchical linear models. Multivariate Behavioral Research. 1995;30(1):1–21.CrossRefGoogle Scholar
  47. 47.
    West SG, Finch JF, Curran PJ. Structural equation models with non-normal variables: Problems and remedies. In: Hoyle R, ed. Structural equation modeling: Concepts, issues and applications. Newbury Park: Sage; 1995:56–75.Google Scholar
  48. 48.
    MacKinnon DP, Fritz MS, Williams J, Lockwood CM. Distribution of the product confidence limits for the indirect effect: Program PRODCLIN. Behav Res Methods. 2007;39(3):384–389.PubMedCrossRefGoogle Scholar
  49. 49.
    MacKinnon DP, Krull JL, Lockwood CM. Equivalence of the mediation, confounding and suppression effect. Prev Sci. 2000;1(4):173–181.PubMedCrossRefGoogle Scholar
  50. 50.
    Franzini L, Caughy M, Spears W, Fernandez Esquer ME. Neighborhood economic conditions, social processes, and self-rated health in low-income neighborhoods in Texas: A multilevel latent variables model. Soc Sci Med. 2005;61(6):1135–1150.PubMedCrossRefGoogle Scholar
  51. 51.
    Whitley R, Prince M, Cargo M. Thinking inside the bubble: Evidence for a new contextual unit in urban mental health. J Epidemiol Community Health. 2005;59(10):893–897.PubMedCrossRefGoogle Scholar
  52. 52.
    Zsembik BA, Fennell D. Ethnic variation in health and the determinants of health among Latinos. Soc Sci Med. 2005;61(1):53–63.PubMedCrossRefGoogle Scholar
  53. 53.
    Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005;62(6):593–602.PubMedCrossRefGoogle Scholar
  54. 54.
    Gee GC, Payne-Sturges DC. Environmental health disparities: A framework integrating psychosocial and environmental concepts. Environ Health Perspect. 2004;112(17):1645–1653.PubMedCrossRefGoogle Scholar
  55. 55.
    Cacioppo JT, Reis HT, Zautra AJ. Social resilience: The value of social fitness with an application to the military. Am Psychol. 2011;66(1):43–51.PubMedCrossRefGoogle Scholar
  56. 56.
    Zautra AJ, Hall JS, Murray KE. Resilience: A new definition of health for people and communities. In: Reich JW, Zautra AJ, Hall JS, eds. Handbook of adult resilience. New York: Guilford; 2010:3–29.Google Scholar

Copyright information

© The Society of Behavioral Medicine 2011

Authors and Affiliations

  • Rebeca Rios
    • 1
    • 2
  • Leona S. Aiken
    • 1
  • Alex J. Zautra
    • 1
  1. 1.Department of PsychologyArizona State UniversityTempeUSA
  2. 2.Johns Hopkins Bayview Medical CenterJohns Hopkins Burn CenterBaltimoreUSA

Personalised recommendations