| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Community Health Sciences, School of Public Health, University of California, Los Angeles.
2 Institute for Health, Health Care Policy, and Aging Research, Rutgers University, New Brunswick, New Jersey.
3 Division of Geriatrics, David Geffen School of Medicine, University of California, Los Angeles.
Address correspondence to Dr. Carol S. Aneshensel, Department of Community Health Sciences, School of Public Health, University of California, Los Angeles, Box 951772, 650 Young Drive South, Los Angeles, CA 90095-1772. E-mail: anshnsl{at}ucla.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Survey data are from the Study of Assets and Health Dynamics Among the Oldest Old (AHEAD), a 1993 U.S. national probability sample of noninstitutionalized persons born in 1923 or earlier (i.e. people aged 70 or older). Neighborhood data are from the 1990 Census at the tract level. Hierarchical linear regression is used to estimate multilevel models.
Results. The average number of depressive symptoms varies across Census tracts independent of individual-level characteristics. Symptoms are not significantly associated with neighborhood SES, ethnic composition, or age structure when individual-level characteristics are controlled statistically. However, net of individual-level characteristics, symptoms are positively associated with neighborhood residential stability, pointing to a complex meaning of residential stability for the older population.
Discussion. This study shows that apparent neighborhood-level socioeconomic effects on depressive symptoms among urban-dwelling older adults are largely if not entirely compositional in nature. Further, residential stability in the urban neighborhood may not be emotionally beneficial to its aged residents.
AN explosion of recent research links neighborhood to health, generally showing modest contextual effects of the socioeconomic environment over and above larger compositional effects on health outcomes (Diez Roux, 2002
; Pickett & Pearl, 2001). A growing body of literature demonstrates this connection specifically for the older population as it relates to self-rated health (Cagney, Browning, & Wen, 2005), cognitive function (Wight et al., 2006), and physical function (Balfour & Kaplan, 2002). The current study extends this line of research by asking whether depressive symptoms among older adults also vary according to place, and, if so, whether this variation is due to contextual effects of neighborhood characteristics that are beyond the compositional effects of individual characteristics.
The recognition that neighborhood socioeconomic status (SES) is relevant to mental health dates to Faris and Dunham (1939/1960), who documented high rates of mental disorder in Chicago areas characterized by social disorganization and isolation. Recent interest in neighborhood and mental health flows largely from the work of Wilson (1987, 1991) and Massey and Denton (1993), especially their descriptions of the emergence of hypersegregated underclass urban communities, where they view the combination of disadvantage and segregation as especially pernicious. These structural models posit a main effect of neighborhood disadvantage on individual well-being, which Jencks and Mayer (1990, p. 113) labeled the "advantages of advantaged neighbors."
For example, Massey and Denton (1993) contended that the physical deterioration of the environment influences the social behavior of residents in the community in ways that are consequential to well-being. Specifically, the presence of ambient hazards such as abandoned housing leads to visible deviance such as public drunkenness, prompting residents to withdraw and spend more time indoors, thus focusing social interactions on family and close friends and decreasing involvement in the larger community. This line of reasoning is of direct relevance to older persons, some of whom find themselves in deteriorating neighborhoods as they and their neighborhoods age in unison. Support networks that are already diminished through death may be further jeopardized if older persons turn inward, away from neighbors. Krause (1996) contended that dilapidated and crime-ridden neighborhoods project an atmosphere that inhibits social interaction by making older adults suspicious and distrustful of others, leading to social isolation, especially with regard to friendships.
Recent work in the structural tradition emphasizes collective efficacy theory (e.g., Cagney et al., 2005). Especially influential is the work of Sampson and colleagues (Sampson, Morenoff, & Earls, 1999; Sampson, Raudenbush, & Earls, 1997), which has identified informal social control and social cohesion as mechanisms that link well-being to neighborhood structural factors, such as socioeconomic disadvantage. As Cagney and associates (2005) observed, this orientation points to two key attributes of neighborhoods—poverty and residential stability. These investigators also noted that affluence may be additionally important in generating a cohesive and trusting environment, a proposition substantiated by their empirical findings.
Structural models implicitly assume that attributes of the environment influence similarly all persons within that environment, but ecological models explicitly treat the impact of environmental factors as conditional upon attributes of the person. For example, Jencks and Mayer (1990, p. 116) referred to the "disadvantages of advantaged neighbors," wherein affluent environments benefit affluent residents more than disadvantaged residents through processes such as relative deprivation. Lawton's seminal ecological model of aging (e.g., Lawton, 1982) is especially relevant because it treats health outcomes as a function of the interaction of the person and the environment. Specifically, the model considers personal characteristics as competencies (e.g., monetary resources) and environmental characteristics as press (e.g., poverty), or recently, as having buoying effects (e.g., social services; Glass & Balfour, 2003). If these elements are out of balance (e.g., when competence is low and press is high), negative affect is likely. From the ecological perspective, health is not the simple sum of exposure to noxious or beneficial stimuli, but rather an interactive function of person–environment fit.
Glass and Balfour (2003) presented a compelling rationale for studying neighborhood specifically among older persons. They ascribed greater vulnerability to neighborhood factors to four mechanisms: longer duration of exposure; increased biological and psychological vulnerability along with changes in cognitive capacity; changing patterns of spatial use so that the older person experiences the neighborhood as the most salient environment; and a reliance on access to community sources of integration, such as senior centers, because other social networks often contract. In addition, Cagney and colleagues (2005) argued that neighborhood is especially consequential to older adults because many age in place. We do not specifically test these mechanisms in the current study, because this would require comparisons to younger populations; however, these considerations led us to focus specifically on the older population.
Few studies have focused on neighborhood SES and emotional well-being specifically among aged individuals, and results are mixed. For example, Kubzansky and associates (2005) found that neighborhood poverty is associated with depressive symptoms net of the individual characteristics for a community sample of residents of New Haven aged 65 and older in the Established Populations for Epidemiologic Studies of the Elderly study. In contrast, La Gory and Fitzpatrick (1992) found no independent effect of neighborhood poverty on depressive symptoms after controlling for personal characteristics for a sample of Alabama residents aged 55 and older. Similarly, for a sample of North Carolina residents aged 65 and older, Hybels and associates (2006) found no neighborhood-level effects on depressive symptoms after controlling for individual-level characteristics for neighborhood socioeconomic disadvantage, racial/ethnic heterogeneity, residential stability, and age structure. Researchers have reported similar results for persons aged 75 and older in Britain (Walters et al., 2004). For the population as a whole, findings are similarly mixed, with some studies showing an association between neighborhood disadvantage and depressive symptoms (e.g., Ross, 2000) and major depression net of individual-level characteristics (Silver, Mulvey, & Swanson, 2002), and other studies finding that apparent neighborhood effects are explained by statistical controls for personal characteristics (Henderson et al., 2005; Yen & Kaplan, 1999).
Studies focusing on the emotional impact of both socioeconomic disadvantage and the racial/ethnic composition of neighborhoods in late life are rare. The Hispanic Established Populations for Epidemiologic Studies of the Elderly study examined Mexican Americans aged 65 and older in five states (Ostir, Eschbach, Markides, & Goodwin, 2003). After adjusting for demographic and other individual level factors, neighborhood poverty was associated with depressive symptoms but only in conjunction with neighborhood proportion Mexican American. The adverse impact of socioeconomic deprivation was offset by the apparent sociocultural benefit of living in an ethnic enclave. These benefits may include shared language and customs, the provision of mutual aid, and identification with others who are of the same background—factors that promote social cohesion. Research on adolescents supports the ideas that ethnic isolation may be a risk for depressive symptoms (Wight, Aneshensel, Botticello, & Sepúlveda, 2005) and that a high concentration of minority residents may offset the risk posed by socioeconomic disadvantage (Aneshensel & Sucoff, 1996). Thus, the concentration of ethnic minority members may reflect a supportive ethnic enclave that is protective of the emotional well-being of residents (Ostir et al., 2003; Shaw & McKay, 1969).
However, the structural perspective posits a less benign interpretation because it sees residential segregation as leading to high densities of minorities, especially African Americans, in underclass neighborhoods (Massey & Denton, 1993; Wilson, 1987, 1991). From this perspective, poverty and segregation each exert negative influences on well-being, influences that may be synergistic.
The current study addresses these issues in evaluating potential neighborhood-level effects on depressive symptoms among older persons living in urban areas. Our conceptual model posited that three neighborhood-level factors influence the level of depressive symptoms among older adults net of the individual-level characteristics of these adults. The first dimension is socioeconomic disadvantage. Following Massey and Denton (1993), we posited that neighborhood poverty erodes the quality of life of residents by making the neighborhood threatening and leading residents to live lives that are disconnected from their neighbors. In contrast, affluence benefits the emotional well-being of residents by generating a cohesive and trusting environment. The second element is the racial/ethnic composition of the neighborhood, specifically the relative presence of members of minority groups. Here we tested two conflicting perspectives: (a) that residential segregation is associated with the types of neighborhood conditions that are also associated with poverty (Massey & Denton, 1993), or (b) that high concentrations of persons of similar backgrounds create ethnic enclaves that promote well-being because people live with people like themselves (e.g., Ostir et al., 2003). Third, we posited that two structural aspects of neighborhoods are beneficial to older residents—residential stability and proportion of adults older than age 65—because these attributes reflect a stable environment of persons like one's self.
In our analysis, we tested both structural and ecological models of these three factors. Our basic structural model contained main effects for each of these factors as described here. An important variant tested neighborhood-level interactions, specifically the idea that it is the combination of neighborhood disadvantage and residential segregation that is especially harmful. Our basic ecological model posited interactions between neighborhood- and individual-level characteristics. The most important interaction here concerned the person's own ethnicity and the racial/ethnic composition of his or her neighborhood, testing the ethnic enclave hypothesis.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
, for a detailed description of the sample derivation). We focused on urban areas because the theory linking concentrated poverty to adverse living conditions was developed with specific reference to urban underclass areas, especially those that are hypersegregated (Massey & Denton, 1993; Wilson, 1987, 1991), and because neighborhood has meaning primarily as applied to urban areas, whereas rural neighbors are typically geographically dispersed. Sample weights adjusted for probabilities of selection. Consequently, the analytic sample was nationally representative of noninstitutionalized persons 70 years or older living in urban areas in 1993.
Measures
The dependent variable was depressive symptoms measured with eight items (e.g., felt depressed) from the longer Center for Epidemiologic Studies–Depression scale (Radloff, 1977), which was modified by changing response codes to yes (1) or no (0) for experiencing the symptom during "much of the time in the past week" (Soldo et al., 1997). We reverse-coded positively worded items and summed items. Reliability (
=.77) and construct validity for the eight-item version of the Center for Epidemiologic Studies–Depression scale have been documented (Steffick, 2000; Turvey, Wallace, & Herzog, 1999).
We analyzed two sets of individual-level independent variables. First were demographic characteristics typically associated with depression. We operationalized SES with three variables: educational attainment (coded as the highest grade of school completed); household wealth (in tens of thousands of dollars, logged), and household income (in thousands of dollars, logged). For income and wealth, the logged transformation was preferable to other functions tested; specifically there was no indication of a threshold effect. We included religion in this set because religious coping may influence the impact of neighborhood deterioration during late life (Krause, 1998). We measured other demographic variables with standard survey items and operationalized them as shown in Table 1.
|
), which has established reliability and validity (e.g., Herzog, & Wallace, 1997). AHEAD researchers interviewed cognitively impaired persons by proxy, and consequently we omitted them from this analysis. We operationalized contextual-level constructs with 1990 U.S. census data for 1,217 tracts; the number of participants per tract ranged from 1 to 31. We assessed neighborhood-level SES with two measures. First, we operationalized socioeconomic disadvantage with a principal component comprising the proportion of (a) residents aged 25 or older without a high school degree, (b) households receiving public assistance income, (c) residents living below the poverty level, and (d) residents aged 16 or older who were unemployed. We repeated analyses for each indicator separately, but we do not report them here because we found no major differences. Second, we assessed affluence as the proportion of households with incomes of $50,000 or more. We used two indicators of racial/ethnic composition: proportion of residents who were African American, and proportion of residents who were Hispanic. Residential stability was the proportion of people aged 5 or older who had lived in the same house for the past 5 years. The proportion aged was the proportion of persons who were older than age 65. We tested census-tract variables for non-linear associations, especially threshold effects, but we found no significant departures from linearity.
Analysis
We calculated descriptive statistics with the SVY procedure in the Stata software package (StataCorp, 2001) and estimated hierarchical linear models with HLM software, version 6.01 (Raudenbush, Bryk, & Cheong, 2000) using full maximum likelihood. We estimated models assuming both a Poisson distribution (appropriate for the count of symptoms) and a normal distribution (because we assumed associations to be linear, not proportional). We found no notable differences, so we present the normal distribution findings for ease of interpretation.
Power calculations took into consideration the design effect, which reduced the effective sample size to 2,790. There was excellent power (99%) to detect partial correlations as small as.10, at an alpha of.05 (Hsieh, Lavori, Cohen, & Feussner, 2003). Assuming a test efficiency of 25%, there was 80% power at an alpha of.05 to detect interaction effect sizes (squared partial correlation) of 1.1% (Aiken & West, 1991).
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
There was considerable variation in census tract characteristics, as shown in Table 2. For each element of the socioeconomic disadvantage principal component, the minimum approached 0 (not shown). However, other areas were characterized by concentrated disadvantage, as evidenced by the maximum values: without a high school degree, 86.3%; receiving public assistance, 73.5%; below poverty level, 86.0%; and unemployment, 48.7%. Socioeconomic disadvantage was most strongly correlated with the absence of affluence, as one would expect, and both measures were moderately correlated with the density of African Americans and Hispanics.
|
=.239; p <.001). The intraclass correlation, which was the ratio of between tract variation to total variation, however, was small (
=.064), indicating that most of the variation in symptoms was at the individual rather than the tract level. The second step asked whether there was any overall association between depressive symptoms and each of the tract-level variables. These multilevel regression coefficients and standard errors appear in the final columns of Table 2. Most of the Level 2 variables had large regression coefficients that were highly statistically significant. Symptoms tended to be high in tracts that were socioeconomically disadvantaged, had few affluent residents, and had large communities of color. Symptoms were also high in neighborhoods with a high proportion of long-term residents, which was in the opposite direction as expected. The relative presence of older residents was not significantly associated with symptoms.
The pivotal issue, however, was whether these associations persisted when we considered individual-level characteristics in the third step. Model 1 in Table 3 presents the regression of depressive symptoms on individual-level demographic and health characteristics. Symptoms were higher on average among women than men and declined slightly with advancing age. Symptoms tended to be higher among persons with limited education. Persons who were widowed and those who were divorced or separated tended to have higher levels of symptoms than married persons. Older adults who were Catholic or Jewish tended to have higher levels of symptoms than those who were Protestant, controlling for other factors in the model. All of the (poor) health variables had statistically significant positive coefficients with depressive symptoms with the exception of having had a stroke (which was not significant) and cognitive function (which was negative in sign because a high score indicates better functioning).
|
2(20, N = 3,442) = 732.844, p 
.001. In combination, the health and demographic variables accounted for 82.4% of the original between-group variance and 15.5% of the original within-group variance in depressive symptoms. However, the small amount of between-tract variation remaining net of these individual-level characteristics was statistically significant. The small amount of residual variation between tracts made it unlikely that any neighborhood variables would be systematically associated with depressive symptoms. Nevertheless, we tested each of the Level 2 variables (see Table 2) individually as additions to Model 1. Only one variable attained statistical significance as shown in Model 2, Table 3. As the proportion of long-term residents increased, the average level of depressive symptoms increased.
Like previous studies of neighborhood SES and depressive symptoms among aged individuals, we included health status as a statistical control. However, health status may be the mechanism linking neighborhood conditions to depressive symptoms, as poor health is predictive of depressive symptoms (Aneshensel, Frerichs, & Huba, 1984). In other words, if health status is a mediator, then controlling it is an instance of overcontrolling. To test this possibility, we removed the measures of health status, as shown in Model 3, Table 3, and then tested each of the neighborhood-level variables individually as additions to the model. As shown, the only variable to attain statistical significance was the proportion of residents who were long-term residents, which was positively associated with the occurrence of depressive symptoms net of individual-level demographic characteristics.
Interaction Models
Existing theories provide competing hypotheses for the interaction of two neighborhood characteristics: socioeconomic disadvantage and the racial/ethnic composition of the neighborhood. From the structural perspective, one would expect socioeconomic disadvantage to have an especially adverse effect when coupled with the segregation of minority residents (cf., Massey & Denton, 1993). From a cultural perspective, however, one would think the presence of ethnic enclaves would offset the impact of disadvantage (e.g., Ostir et al., 2003). Both theories posit an interaction between these characteristics, but of opposite sign. However, the interaction term was not statically significant for either proportion African American or proportion Hispanic.
The ecological model posited cross-level interactions between neighborhood-level and individual-level characteristics. Thus, we also tested whether the impact of socioeconomic disadvantage differed for select subpopulations by testing cross-level interactions. To ascertain whether disadvantage was more depressing among poor persons, we interacted neighborhood-level socioeconomic disadvantage with each of the individual-level SES variables; none of these terms were statistically significant. We tested similar cross-level interactions for the individual's ethnicity, specifically being African American or being Hispanic; these terms did not attain statistical significance. In addition, we tested whether living in a similar ethnic community was associated with depression by testing the cross-level interaction of proportion Hispanic with being Hispanic, and proportion African American with being African American; these terms were not significant.
Because several studies have suggested that neighborhood effects may be especially pronounced among older persons who are in poor health (La Gory & Fitzpatrick, 1992; Muramatsu, 2003), we tested interactions between socioeconomic disadvantage and each of the health indicators. None of these tests were statistically significant. In addition, Ross, Reynolds, and Geis (2000) suggested that the impact of residential stability is contingent upon the socioeconomic circumstances of the neighborhood. We tested this interaction, and it was not statistically significant. We also tested whether the residential stability effect was contingent upon age structure, but this term also was not significant. Finally, Krause (1998) reported a stress-buffering function of religious coping, but interactions of socioeconomic disadvantage and religious affiliation, a weak indicator of religious coping, were not significant.
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; La Gory & Fitzpatrick, 1992; Walters et al., 2004) but differ from two other studies that reported an effect of neighborhood SES on depressive symptoms while controlling for individual-level SES (Kubzansky et al., 2005; Ostir et al., 2003). To what can we attribute the discrepancy?
Researchers often criticize contextual effects as representing misspecification of the individual-level model, specifically the omission of relevant control variables (e.g., Diez Roux, 2002). In anticipation of this critique, we included three individual-level measures of SES: education, income, and wealth. Including wealth is important because it is a better indicator of SES than income for retired persons; discrepant previous studies did not include wealth. However, our findings did not change when we re-estimated equations without the wealth variable (data not shown). We also used more precise measures of SES than did previous discrepant studies, but when we rescored our measures to match these studies, the neighborhood-level SES variables remained nonsignificant. Thus, differences in the measurement of individual-level SES do not seem to account for the discrepant findings across studies. Nevertheless, we can attribute the poor performance of the neighborhood-level SES variables to the strength of our individual-level model insofar as it accounted for almost all of the variation in symptoms between tracts; that is, there was very little between-tract variation remaining to be explained by any neighborhood-level factor.
Differences in findings may be due to variation in the populations studied. AHEAD is a nationally representative sample of older adults, which makes it considerably more diverse than the samples of the other studies, which typically focused on one locale. For example, the Established Populations for Epidemiologic Studies of the Elderly study represents only one metropolitan area: New Haven (Kubzansky et al., 2005). It is possible that neighborhood-level SES effects are confined to select subpopulations.
We examined this issue by testing cross-level interactions between neighborhood-level socioeconomic disadvantage and (a) the individual-level SES variables, and (b) being African American or being Hispanic. We also tested neighborhood-level interactions between socioeconomic disadvantage and the racial/ethnic composition of the tract. As reported previously, these tests were not statistically significant. These findings do not support the presence of a systematic SES-related process that renders some older subpopulations (e.g., poor persons, ethnic minorities) especially vulnerable to neighborhood socioeconomic disadvantage, at least with regard to their emotional well-being.
The one other study that examined a national sample used the same AHEAD data as the current study but analyzed the entire sample (not just urban residents) and analyzed county-rather than tract-level data. Although Muramatsu (2003) found a contextual effect for income inequality, the coefficient for county-level income was not statistically significant, similar to our findings for tract-level socioeconomic disadvantage. County-level analyses are useful for administrative purposes and provide insight into how large social contexts may affect health. However, tract-level analyses are appropriate for operationalizing the concept of neighborhood because tracts are designed to be relatively homogeneous units with respect to population characteristics, economic status, and living conditions. The dense concentration of persons makes tracts particularly suitable representations of neighborhoods in urban areas, especially because they encompass a relatively small geographical area. Thus, our approach to the AHEAD data applied a meaningful operationalization of neighborhood to the urban setting. Nevertheless, our findings apply only to the tract operationalization of neighborhood; other aggregations, both larger and smaller, may reveal different neighborhood SES effects.
Residential stability was the sole neighborhood characteristic to significantly contribute to the explanation of depressive symptoms once we had taken individual-level characteristics into consideration. We conceptualized residential stability as beneficial to emotional well-being, whereas the observed association, in both bivariate and multivariate analysis, was in the opposite direction than expected. We note that residential stability was positively associated with affluence and a high concentration of African Americans, even though affluence was negatively associated with a high concentration of African Americans (see Table 2). Residential stability was also positively associated with the relative concentration of older residents and inversely associated with a high concentration of Hispanics. These correlations point to a complex meaning of residential stability for the older population. Our findings point to aspects of stable neighborhoods inhabited by older residents that may be emotionally deleterious, such as the possibility that stable neighborhoods are a sign that older residents generally are not able to move from the neighborhood, although other interpretations also are plausible. However, the amount of variation between neighborhoods that this association accounted for was very small, indicating that readers should use caution in its interpretation.
A comment is in order about the association between age and depression. The zero-order correlation coefficient for these variables was significantly positive (r =.100; p .001). This association was not significant, however, when we controlled for individual-level demographic characteristics (data not shown). Moreover, the association became significantly negative when we controlled for health status (Model 1, Table 3). Our zero-order findings are consistent with previous research reporting that symptoms of depression increase appreciably at older ages (Frerichs, Aneshensel, & Clark, 1981; Kessler, Foster, Webster, & House, 1992; Mirowsky & Ross, 1992; Newmann, 1989). The change in sign observed here suggests that one can attribute increasing symptoms among the older population to the occurrence of poor health and that, were it not for these health problems, symptoms would decline slightly with age.
Our analysis assumes that tract- and individual-level SES are associated with each other because the tract is the sum total of individual-level SES; we do not attribute a causal connection to this association. Other interpretations of this association, however, are possible. For instance, it may be the result of selection of individuals into and out of neighborhoods on the basis of SES. Alternatively, processes operating at the neighborhood level might influence how people accumulate wealth. Probing the nature of this association is an important research topic in its own right.
There are several limitations to the current study that one should take into consideration. First, in contrast to the just-described merits of using census tracts, this operationalization is problematic because tracts are official boundaries that create artificial neighborhoods rather than subjective definitions of neighborhood used by residents. However, the availability of official data concerning tracts—data that are not readily available for other definitions of neighborhood—justifies this practice. Second, an inherent limitation of the AHEAD study, which organizers did not design with multilevel analysis as a scientific objective, is that a substantial proportion of sampled census tracts contain only one respondent (44%), meaning there is no within-tract variation. Our ability to find an effect of a neighborhood characteristic, however, was determined not by the number of participants per tract, but by the distribution of the characteristic in the study sample. Thus, the singleton tracts did not limit our ability to make inferences about measured neighborhood characteristics insofar as there were sufficient tracts with different values for all of the Level 2 variables used in these analyses. The singleton tract limitation also was offset by the diversity of the tracts included in the analysis, given that this was a national sample. Another analysis of the AHEAD data found a neighborhood-level SES effect for the outcome of cognitive functioning, including a cross-level interaction (Wight et al., 2006). Thus, any limitations of the tract-level data do not appear to be sufficient to fully account for our negative findings for depressive symptoms for the neighborhood SES variables.
Third, although the measurement of SES is a study strong point, the measurement of depressive symptoms is abbreviated, which restricts the amount of variation available for analysis. This may compromise power to detect neighborhood variation in symptoms. In this regard, we should note that the two studies reporting a neighborhood SES effect among older persons employed the full 20-item version of the Center for Epidemiologic Studies–Depression scale (Kubzansky et al., 2005; Ostir et al., 2003), but this is also the case for studies that failed to find this effect (Hybels et al., 2006; La Gory & Fitzpatrick, 1992). Thus, it is unlikely that our negative findings for neighborhood SES are due solely to the abbreviated measurement of depressive symptoms.
In sum, the discrepant findings across types of neighborhood studies point to the need to specify theoretically the circumstances under which specific neighborhood effects should be manifest. Although we found that the stability of urban neighborhood residents may not be emotionally beneficial in late life, this finding merits additional future research. In consideration of the strengths and limitations of the current study, we conclude that the apparent neighborhood effects of socioeconomic disadvantage and ethnic minority concentrations on depressive symptoms among the urban population of older persons are largely, if not entirely, compositional in nature.
|
Acknowledgments |
|---|
We wish to thank Abdelmonem Afifi for his advice on analysis and three anonymous reviewers for their helpful suggestions. Any errors are the sole responsibility of the authors.
|
Footnotes |
|---|
Received for publication December 12, 2005. Accepted for publication August 29, 2006.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C Mair, A V D. Roux, and S Galea Are neighbourhood characteristics associated with depressive symptoms? A review of evidence J Epidemiol Community Health, November 1, 2008; 62(11): 940 - 946. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Kim Blues from the Neighborhood? Neighborhood Characteristics and Depression Epidemiol. Rev., November 1, 2008; 30(1): 101 - 117. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. A. Lang, D. J. Llewellyn, K. M. Langa, R. B. Wallace, and D. Melzer Neighbourhood deprivation and incident mobility disability in older adults Age Ageing, July 1, 2008; 37(4): 403 - 410. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
