ARTICLE IN PRESS
Social Science & Medicine 63 (2006) 2604–2616
www.elsevier.com/locate/socscimed
Urban neighborhoods, chronic stress, gender and depression$
Flora I. Mathesona,b,, Rahim Moineddinc, James R. Dunna,d,
Maria Isabella Creatorea, Piotr Gozdyraa, Richard H. Glaziera,c,e
a
Centre for Research on Inner City Health, St. Michael’s Hospital, Toronto, Ontario, Canada
b
Department of Public Health Sciences, University of Toronto, Toronto, Ontario, Canada
c
Department of Family and Community Medicine, University of Toronto, Toronto, Ontario, Canada
d
Department of Geography, University of Toronto, Toronto, Ontario, Canada
e
Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada
Available online 22 August 2006
Abstract
Using multilevel analysis we find that residents of ‘‘stressed’’ neighborhoods have higher levels of depression than
residents of less ‘‘stressed’’ neighborhoods. Data for individuals are from two cycles of the Canadian Community Health
Survey, a national probability sample of 56,428 adults living in 25 Census Metropolitan Areas in Canada, with linked
information about the respondents’ census tracts. Depression is measured with the Center for Epidemiologic StudiesDepression Scale Short Form and is based on a cutoff of 4+ symptoms. Factor analysis of census tract characteristics
identified two measures of neighborhood chronic stress—residential mobility and material deprivation—and two measures
of population structure—ethnic diversity and dependency. After adjustment for individual-level gender, age, education,
marital and visible minority status and neighborhood-level ethnic diversity and dependency, a significant contextual effect
of neighborhood chronic stress survives. As such, the daily stress of living in a neighborhood where residential mobility
and material deprivation prevail is associated with depression. Since gender frames access to personal and social resources,
we explored the possibility that women might be more reactive to chronic stressors manifested in higher risk of depression.
However, we did not find random variation in depression by gender across neighborhoods.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Canada; Depression; Gender; Neighborhood; Chronic stress
Introduction
$
This research was made possible through the assistance of
the Social Sciences and Humanities Research Council of Canada,
Standard Research Grant 410-2005-2306.
Corresponding author. Tel.: +1 416 864 6060;
fax: +1 416 864 5485.
E-mail addresses: [email protected]
(F.I. Matheson), [email protected]
(R. Moineddin), [email protected] (J.R. Dunn),
[email protected] (M.I. Creatore),
[email protected] (P. Gozdyra),
[email protected] (R.H. Glazier).
0277-9536/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2006.07.001
The idea that features of the social environment
in which we live are important determinants of
healthy lives is not new. In one of the earliest studies
on environment and mental health, Faris and
Dunham (1939) showed that psychiatric admissions
in Chicago varied by location within the city with
higher rates for those living in the inner-city core
than in outlying areas. The ecological approach to
the study of health was largely supplanted by
ARTICLE IN PRESS
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
individualistic approaches, but more recently the
hybrid approach of multilevel modeling has become
more popular. Historical and modern writings
advocate for population-level studies of health,
reasoning that while risk factors originate in
individuals (compostional effects), many of these
risks propagate and become reinforced within social
contexts (contextual effects), especially risks associated with health and health-related behavior.
Durkheim (1951) studied the social causes of suicide
and found regularities in suicide rates among
specific population groups. He argued that social
integration and regulation, characteristics of the
social environment rather than the individual,
contributed to the suicide rate in a particular group
thus producing differences from the rate in other
groups. According to Dunn, Frohlich, Ross, Curtis,
and Sanmartin (2005) social facts may provide
details on the uniqueness of specific places and
particularly those stressors that influence, enhance
and undermine the health of the population. They
argue that it is imperative to examine these underlying phenomena to inform a global discussion of
population health. Geoffrey Rose (1985) argued
that groups of individuals function as a collectivity
and these groups are affected by the average
functioning of the group. Multilevel modeling
allows researchers to simultaneously study individual and contextual effects and to ask questions
about the relationship between neighborhood and
personal health.
The nature of chronic stress in context
So what is it about the social environment that
might lead to deterioration in the health of
individuals? One issue, often identified, is chronic
stressors in the local residential environment. In the
social science literature contextual features such as
crowding (Gillis, 1979a; Gillis, Richard, & Hagan,
1986), density, (Gillis, 1974, 1979b; Gillis & Hagan,
1982; Regoeczi, 2002; Sampson, 1983) and housing
(Dunn, 2002; Dunn & Hayes, 2000; Gillis, 1977)
have been examined as sources of chronic stress and
have been found to be associated with health and
well-being and health-related behaviors (Glazier
et al., 2004; Hyndman, Holman, & Jamrozik,
1997; Katz, Zemencuk, & Hofer, 2000).
Wheaton (1999) defines stressors (chronic, life
event and daily hassles) as ‘‘conditions of threat,
demands, or structural constraints that, by their
very occurrence or existence, call into question the
2605
operating integrity of the organism’’ (Wheaton,
1999, p. 177). Chronic stress is insidious, with a slow
and imperceptible onset and an open-ended recurring character; this form of stress is especially
important since it can arise from contextual
features. Chronic stressors in the local residential
environment (ecological) are difficult to avoid (e.g.,
uncontrollable) and represent ongoing stress that is
a part of daily living. Ecological chronic stressors
are different from role-related stressors since they
originate at a level above the individual and his or
her interaction with role partners and peers (Serido,
Almeida, & Wethington, 2004; Wheaton, 1999).
This form of stress can include reduced or lack of
access to opportunity or to the necessary means to
achieve ends as well as structural reduction in
available alternatives or choices (Wheaton, 1999).
Research suggests that these chronic sources of
stress or ‘‘quotidian’’ stressors (Pearlin & Skalkidou, 1995; Wheaton, 1999) more strongly affect
well-being than the less frequent, but more major
class of stressors known as life events which tend to
be more transient in nature. Recent research on
material deprivation, socioeconomic disadvantage,
neighborhood disorder and instability points to the
negative impact of chronic stressors on health
(Aneshensel & Sucoff, 1996; Boardman, 2004;
Boardman, Finch, Ellison, Williams, & Jackson,
2001; Dunn & Hayes, 2000; Feldman & Steptoe,
2004; Linsky, 1969; Ross & Mirowsky, 2001). The
ecology of chronic stress is an important area of
sociological inquiry (Hill, Ross, & Angel, 2005). The
strain created by the chronic nature of neighborhood stressors will not only have an impact on
individual health, but can lead to the deterioration
of the capacity of the population within a neighborhood to resist the pathological effects of ambient
stress.
A growing body of research examines the
relationship between urban stressors and health
outcomes with only a handful of these studies
focusing on the life stress paradigm. Boardman and
colleagues (2001) found that socio-economically
disadvantaged neighborhoods had a significant
impact on the likelihood of adult drug use net of
individual factors like socioeconomic status (SES)
and socio-demographic characteristics (Boardman
et al., 2001). Their results show that the effect of
neighborhood disadvantage is most pronounced
among poorer residents. Elliott (2000), using a
composite measure of neighborhood socioeconomic
status found that social support was protective of
ARTICLE IN PRESS
2606
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
mental and physical health among residents of
higher-SES neighborhoods. He was unable to
directly examine neighborhood stressors, but rather
assumed that lower-SES neighborhoods are more
stressful with less resources than higher-SES neighborhoods. In a more recent paper, Boardman (2004)
found that neighborhood residential instability was
an important factor in explaining physical health.
Residing in relatively unstable neighborhoods had a
greater negative impact on health.
Silver, Mulvey, and Swanson (2002) examined the
impact of neighborhood disadvantage and residential mobility on several mental disorders after
controlling for individual-level risk factors. They
found that major depression was more prevalent in
disadvantaged and residentially mobile neighborhoods. Their findings mirror that of Ross, Reynolds, and Geis (2000) who found that the effect of
residential stability on depression depended on the
level of poverty in the neighborhood. Under
conditions of high poverty, residents of neighborhoods with little residential turnover had higher
levels of depression and anxiety than residents of
more mobile neighborhoods.
Recently, Latkin and Curry (2003) examined the
impact of neighborhood stress on depression using a
prospective multilevel approach. They found that
neighborhood perceptions of social disorder were
positively associated with depressive symptoms and
argue that social disorganization is an important
chronic stressor among inner-city populations. They
emphasize that depression should not only be
viewed as an individual level phenomenon, but that
ecological factors are important for identifying and
understanding neighborhoods that are at risk for
depression.
Using Rose’s (1985) logic we ask, ‘‘Why do some
neighborhood populations have more depression.’’
We seek an explanation in sources of chronic
stressors that exist at the neighborhood level using
the framework of the stress process paradigm. The
original stress process paradigm (Lin & Ensel, 1989;
Pearlin, 1989) suggests that the pathway to wellbeing is a process that includes individual risk factors
(e.g., gender), individual stressors (e.g., life events
and chronic strains) and social resources (e.g.,
supportive friendship networks), but Pearlin (1989)
also suggested that stressful experiences do not exist
in a vacuum, but can be traced to social structures
and the location of individuals within these structures. Stressful experiences are interpreted within a
person-environment transactional framework where
transactions depend on the impact of the external
stressor. This is mediated by the person’s appraisal of
the stressor and the social and cultural resources at
his or her disposal (Antonovsky & Kats, 1967;
Cohen, 1984; Lazarus, 1966) and these appraisals are
thought to be influenced by gender roles and gender
socialization processes.
We use the general framework of the stress
process paradigm to examine neighborhood sources
of stress and their association with depression.
Instead of individual-level stress we construct
aggregate measures of neighborhood-level chronic
stressors. The evidence that chronic stressors in the
local residential environment are associated with
physical and mental health is growing (Hill et al.,
2005) and is evident in the literature on material
deprivation from the UK (Carstairs & Morris,
1989a, b; Jarman, 1983, 1984; Townsend, 1987).
More recently, in two separate studies in Canada,
Pampalon and Raymond (2000) developed a deprivation index for urban and rural areas in Quebec
and Frohlich and Mustard (1996) developed a
composite index of neighborhood disadvantage for
the province of Manitoba. Like these previous
studies we wanted to explore the notions of
deprivation and residential instability as potential
chronic stressors within the local residential environment. Using similar census measures to those
from the UK and Canada on material deprivation
and on residential instability (primarily from the
US) we sought to define chronic stressors related to
deprivation and instability for 25 urban areas in
Canada. This paper therefore provides an exploration of the association of neighborhood chronic
stressors with individual reports of depression
within a national sample of 3619 socially and
ethnically diverse neighborhoods.
Gender and neighborhood chronic stress
The stress process model posits that social status
determines one’s position in the social structure, and
thus the types and intensity of stressors to which one
is exposed. Gender is recognized as a status position
that organizes our lives and frames our access to
personal and social resources primarily because
gender encompasses both biological/genetic and
social learning differences. Explanations of gender
differences in depression, with women at greater risk,
reflect both sex-related biological and social aspects
as well as the interplay between these two forces
(Verbrugge, 1989). Social explanations of gender
ARTICLE IN PRESS
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
differences in health posit that women report higher
levels of depression and other health problems
because of reduced access to material and social
conditions of life that foster health (Arber & Cooper,
1999; Ross & Bird, 1994). In comparison to men,
women are less likely to be employed, more likely to
work in lower status positions when employed, to
have lower incomes and be single parents (Denton &
Walters, 1999; Ross & Bird, 1994).
While we know from research that living in
economically disadvantaged and residentially mobile neighborhoods leads to poor physical and
mental health outcomes, we know little of the
potential impact of neighborhood chronic stressors
on depressive outcomes in men and women. There
are grounds to believe that neighborhood stressors
may be more detrimental for women’s health.
Suicide literature suggests that gender roles veritably weave women more wholly into the social fabric
(Danigelis & Pope, 1979; Steffensmeier, 1984)
Women know more of their neighbors by name,
and talk or visit with them more frequently than do
men (Campbell & Lee, 1992; Kessler & McLeod,
1984) suggesting that women have larger social
networks than men and are more embedded in their
communities. If women are more integrated into
their neighborhoods, maintaining more intense and
multiplex relations (Campbell & Lee, 1992) we
might envision that neighborhood stressors will
have greater impact on women’s mental health
rather than men’s health; this primarily because
residential mobility and material deprivation interfere with maintenance of social bonds and social
support both of which are often considered protective for women’s health (Denton & Walters, 1999;
Magdol, 2002; Propper et al., 2005).
Research questions
This paper addresses the four following research
questions:
(1) Is neighborhood stress correlated with higher
risk of depression?
(2) To what extent does chronic stress account for
neighborhood variation in depression?
(3) Does the association of neighborhood chronic
stress remain after controlling for individual
characteristics, neighborhood ethnic diversity
and dependency?
(4) Is neighborhood chronic stress associated with
gender differences in depression?
2607
Data and methods
Data sources
Two data sources were used to investigate the
association between neighborhood context and
depression—the Canadian Community Health Survey (CCHS) and the 2001 census of Canada, both
collected by Statistics Canada. The CCHS is a crosssectional nationally representative survey that provides detailed information on health determinants
and health outcomes (see Beland, 2002 for detailed
methodology). Two cycles of the CCHS (cycle 1.1
and cycle 2.1 collected in 2000–2001 and 2003–2004,
respectively) were combined to create a large crossnational dataset of 267,108 respondents. The
combined CCHS represents two surveys with the
same subject matter and identical populations.
There is considerable interest in combining datasets
when possible because the combined data provide
more efficient estimates by increasing sample size
within neighborhoods and by increasing the number
of neighborhoods. Statistics Canada’s data publication guidelines were followed throughout the
analysis. All analyses were weighted using the
combined sampling weight provided by Statistics
Canada. Bootstrapping with 500 replications was
used to obtain p values for logistic analysis shown in
Table 2. All p values were two-sided. SAS version
9.1 (SAS Institute Inc., Cary, NC, USA) was used
for data manipulation and statistical analysis.
Ethics approval was obtained from the St. Michael’s
Hospital Research Ethics Review Board.
Respondents were assigned to census tracts (CTs)
using the Postal Code Conversion File Plus
(PCCF4F+), an automated system that uses postal
codes to assign census geography (Wilkins, 2005).
PCCF4F+ is based on the latest Postal Code
Conversion File and Postal Code Population
Weight File produced by the Geography Division
of Statistics Canada. (CMAs) uses weights to
allocate postal codes linked to multiple dissemination areas according to the distribution of population using a given postal code.
Sample and unit of analysis
For practical reasons our focus here is on urban
neighborhoods since 80% of Canadians are urban
dwellers. The concentration of people in cities
makes them the primary locus of neighborhood
effects. We included respondents aged 18–74 living
ARTICLE IN PRESS
2608
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
in the 25 Census Metropolitan Areas PCCF4F+
included in the survey (CCHS) representing approximately 40% of the full sample. Respondents
who were less than 18 and greater than 74 were
excluded due to high rates of proxy interviews. After
exclusion of observations with missing data the final
sample was 56,428 respondents in 3619 CTs. CTs
are small, relatively stable geographic units (populations between 2500 and 8000) constructed similarly with respect to economic status and social
conditions. Fig. 1 visually depicts the 25 CMAs
included in the CCHS that are the focus of this
study.
Neighborhood measures
The selection of neighborhood characteristics for
this analysis was guided by previous neighborhoodbased research and theory linking neighborhood
stressors to poor health (Hill et al., 2005; Kornhauser, 1978; Pampalon & Raymond, 2000; Sampson &
Groves, 1989; Sampson, Raudenbush, & Earls,
1997; Shaw & McKay, 1969). The literature on
deprivation and residential instability was pivotal as
a starting point and provided the input variables for
factor analysis. In total, 18 CT measures were
extracted from the census data, describing the socioeconomic and demographic character of the CT. To
avoid the problem of multicollinearity (often
associated with census data) that would result from
analyzing these measures simultaneously in a
regression equation, we followed the recommendation of Land, McCall, and Cohen (1990) and factor
analyzed them. Factors were constructed using
oblique rotation which allows the factors to co-vary
(Kim, 1978a, b). We also estimated an orthogonal
factor matrix that yielded substantively identical
results. Measures of neighborhood context were
constructed from census data available through the
Canadian University Data Liberation Initiative.
With an eigenvalue of 7.8, the first factor is
dominated by moderate to high loadings (460) for
percentage living alone, percentage youth 5–15
years (reverse coded), persons per dwelling (reverse
coded), percentage living in apartment buildings,
percentage married (reverse coded), percentage
home ownership (reverse coded), and percentage
moving within the last 5 years. This first factor is
essentially a measure of residential instability.
The second dimension (eigenvalue 2.8) is a
measure of material deprivation. The six variables
that define this dimension are percentage 20+
without high school graduation, percentage lone
parent families, percentage of families receiving
government transfer payments, percentage 15+
unemployed, percentage living below the low
income cut off (a Statistics Canada defined measure
that is adjusted for community size, family size and
inflation), and percentage of homes needing major
repair.
The third dimension, dependency (eigenvalue 2.5)
includes high factor loadings (480) for percentage
Fig. 1. Map of the 25 Canadian census metropolitan areas for which data from the 2001 CCHS was used. Source: Statistics Canada and
ESRI Canada.
ARTICLE IN PRESS
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
seniors (age 65+), the ratio of the population age
0–14 and 65+ divided by the population aged
15–64, and labor force participation (reverse coded).
The final dimension is ethnic diversity (eigenvalue
1.7), which is dominated by high loadings (490) for
percentage recent immigrants and percentage visible
minorities.
The loadings for each factor were used to
compute four separate indices for residential instability, material deprivation, dependency, and
ethnic diversity from the factor scores generated
for each subject (after reverse coding the variables
with negative weights). All four indices are standardized scores.
In assessing these four factors, residential instability and material deprivation capture the
concept of chronic stress to a much larger degree
than ethnic diversity and dependency. The latter
two measures better reflect demographic characteristics related to ethnicity and population age
structure and thus were used as neighborhood-level
control variables in the analysis.
Depression
The CCHS interview incorporated a brief predictive instrument to identify major depression
occurring during the preceding year. The Composite
Diagnostic Interview Schedule Short Form for
major depression (CIDI-SF MD) is a brief version
of a fully structured diagnostic interview for major
depression (Kessler, Andrew, Mroczek, Üstün, &
Wittchen, 1998) that can be administered by trained
lay interviewers in large-scale national health
surveys and includes questions covering the symptoms of major depression that are critical to the
diagnostic definition within the DSM-IV (American
Psychiatric Association, 2000). To construct the
individual-level measure of depression we selected a
probability of caseness of 0.8125 corresponding to a
short form major depression score of 4 or greater. In
creating the outcome measure, each respondent was
categorized as depressed or not based on a symptom
score of 4 or greater.
2609
we include several categorical variables: gender, age
of respondent, (age 18–29, age 30–39, age 40–49, age
50–59 with age 60–74 used as the reference
category), married, education and visible minority
status which includes persons, other than Aboriginal persons, who are not white in race or color.
There is considerable debate in the literature on
the treatment of individual-level covariates in multilevel modeling. Some investigators include an
extensive list of covariates while others are highly
selective (see Bingenheimer & Raudenbush, 2004).
We began with a relatively extensive list, but found it
necessary to cull the list due to a high degree of
collinearity among individual-level covariates. In
addition to the measures described above, the initial
list included low income status, single parenthood
status, unemployment, aboriginal status, housing
ownership, immigrant status, living alone, and
sources of income. We found strong associations
between these variables and the five variables that
have been retained for multilevel analysis. Other
variables often examined within the stress-process
model, specifically those related to individual-level
personal and social resources, life event and chronic
stress, were not collected in all regions of the country
and as such could not be included in the analysis.
Statistical procedures
The analysis begins with a description of the
sample followed by standard logistic regression to
assess the association of each predictor and control
variable with depression. We then conduct multilevel logistic modeling to allow for the hierarchical
nature of the data since individuals are nested
within neighborhoods. We first determine how
much of the variance in the outcome measure of
depression can be attributed to the neighborhood
level (null model). We follow this with a series of
models first introducing individual-level variables
then neighborhood-level variables in separate models and finally including the individual and neighborhood level variables together in a final model.
Hierarchical logistic regression was performed using
the SAS NLMIXED procedure.
Individual level measures
Results
Individual-level variables that reflect socio-demographic characteristics related to depression are
included to adjust for their potential impact on
individual-level depression when neighborhood
stressors are included in the model. From the CCHS
Descriptive statistics
Correlations among the factors and their components are presented in the third column of Table 1.
ARTICLE IN PRESS
2610
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
Table 1
Factor pattern and correlations for 3619 Canadian neighborhoods
Variable
Factor
loading
Correlation
with factor
96
95
94
0.95***
0.91***
0.87***
Residential instability
Living alone
Youth (reverse coded)
Persons per dwelling (reverse
coded)
Apartments
Married (reverse coded)
Owner-occupied house (reverse
coded)
Residential mobility (5 year)
79
77
71
0.87***
0.87***
0.87***
64
0.67***
Material deprivation
No high school graduation
Lone parent families
Government transfers
Unemployment
Low income
Homes needing major repair
91
84
81
70
67
51
0.88***
0.87***
0.86***
0.79***
0.77***
0.52***
90
88
0.91***
0.86***
79
0.82***
96
95
0.94***
0.92***
Dependency
Seniors
Ratio population (age
0–14+65+)/(age 15–64)
Labour force participation (reverse
coded)
Ethnic diversity
Recent immigrants
Visible minorities
Notes: *pp0.05; **pp0.01; ***pp0.001.
Correlations among residential instability and its
component measures range from 0.67 to 0.95; those
for the components of material deprivation range
from 0.52 to 0.88; those for the components of
dependency range from 0.82 to 0.91; and those
between ethnic diversity and its components range
from 0.92 to 0.94.
As shown in Table 2 the prevalence of depression
in the sample was 9% (N ¼ 5497). Approximately
51% of the sample were female, 56% were married,
84% reported having graduated from high school
and 22% reported being of visible minority status.
Age is shown as a categorical variable: 23% age
18–29, 23% age 30–39, 23% age 40–49, and 16%
age 50–59 (age 60–74 was the reference category).
The odds of depression by each predictor are
shown in the last column of Table 2 (standard logit
model). The risk of depression is almost 2 times
greater for women than men (OR ¼ 1.79, pp0:001).
Those in younger age groups are at greater risk of
depression relative to those aged 60–74 (pp0:001).
Being married is associated with a 51% decreased
risk (OR ¼ 0.49, pp0:001), graduation from high
school is associated with a 20% decreased risk
(OR ¼ 0.80, pp0:001), and visible minority status is
associated with a 37% decreased risk (OR ¼ 0.63,
pp0:001) of depression.
Table 2 also provides the odds of depression by
each of the neighborhood characteristics. Both
measures of neighborhood chronic stress are associated with increased odds of depression, with
residential instability associated with an 11% and
deprivation with a 6% increased risk (pp0:001 and
p0:01, respectively). Dependency (OR ¼ 0.94,
pp0:01) and ethnic diversity (OR ¼ 0.92,
pp0:001) are associated with decreased risk of
depression. These results provide an affirmative
response to the first research question: Is neighborhood stress correlated with the risk of depression?
Multilevel logistic regression
We begin by considering a fully unconditional
ordered logit model of the dichotomous outcome of
depression (not shown) in order to decompose
variance in the dependent variable across levels of
analysis. The intraclass correlation (ICC) for
logistic regression can be calculated as r ¼
ðs2u =s2u þ s2e Þ, where s2e ¼ p2 =3 (Guo & Zhao,
2000).The estimate of the intercept variance
(0.089, OR ¼ 1.09, pp0:001) yields an intra-class
correlation (ICC) of 0.026, suggesting that, on
average, 2.6% of the variance in the dependent
variable can be attributed to the neighborhood
level. In interpreting the value of the ICC it is
important to keep the following points in mind:
while most of the variability in depression is at the
individual level, the magnitude of the ICC is within
the typical range since ICCs rarely exceed 0.20
(Snijders & Bosker, 1999); the size of the ICC does
not rule out relatively large effects of neighborhood
level measures (Duncan & Raudenbush, 1999); and,
to the extent that the final model includes significant
cross-level interactions, ICCs will vary across level
two units.
Table 3 reports the results of a series of
hierarchical ordered logit models predicting depression. Model 1 reports the results of a regression
predicting depression based on demographic background variables. The unadjusted odds ratios for
women remained relatively unchanged from results
ARTICLE IN PRESS
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
2611
Table 2
Descriptive statistics (N ¼ 56; 428)
Variables
Percent
N
Bivariate odds of depression
Individual characteristics
Depression (4+ symptoms)
Female
Age 18–29*
Age 30–39
Age 40–49
Age 50–59
Married
High school graduation
Visible minority
9%
51%
23%
23%
23%
16%
56%
84%
22%
5497
30,414
11,688
12,768
12,433
9499
28,126
47,038
8578
—
1.79***
2.41***
2.01***
1.97***
1.59***
0.49***
0.80***
0.63***
Neighborhood characteristicsa
Residential instability
Material deprivation
Dependency
Ethnic diversity
—
—
—
—
—
—
—
—
1.11***
1.06**
0.94**
0.92***
Notes: Weighted means and odds ratios and non-weighted N’s are presented.*pp0.05; **pp0.01; ***pp0.001.
a
Factor scores are standardized with a mean of 0 and standard deviation of 1.
Table 3
Multilevel logistic regression odds ratios of depression by
individual and neighborhood characteristics (N ¼ 56; 428)
Depression (4+ symptoms)
Variables
Model 1
Individual level
Female
Age 18–29
Age 30–39
Age 40–49
Age 50–59
Married
High school graduation
Visible minority
1.87***
2.46***
2.84***
2.78***
2.20***
0.49***
0.73***
0.65***
Model 2
Model 3
1.87***
2.44***
2.81***
2.76***
2.19***
0.51***
0.74***
0.66***
Neighborhood level
Residential instability
Material deprivation
Dependency
Ethnic diversity
1.12***
1.12***
0.92***
0.90***
1.04*
1.05**
0.97
0.97
Intercept
0.06***
Intercept Variance Component 1.07***
0.11***
1.05***
0.06***
1.06***
Notes: *pp0.05; **pp0.01; ***pp0.001.
reported in Table 2 after adjusting for age, marital
status, high school graduation and visible minority
status. The odds ratios for all demographic background variables were similar in size and effect to
those presented in Table 2. Despite the inclusion of
individual-level characteristics, the variance of the
random intercept remains significant indicating that
there is still unexplained variance at the neighborhood-level (pp0:001) net of individual-level variables.
Model 2 reports results for depression regressed
on neighborhood characteristics including the
two measures of chronic stress—residential instability and material deprivation. The two neighborhood stressors—residential instability (OR ¼ 1.12,
pp0:001) and material deprivation (OR ¼ 1.12,
pp0:001)—are each significantly and independently
associated with a 12% increased risk of depression
in Model 2. The positive effect of both measures of
neighborhood stress is consistent with our conceptualization of the stress process model suggesting
that neighborhood stressors account for some of the
variation in depression after controlling for neighborhood population structure (dependency) and
ethnic diversity. Both dependency (OR ¼ 0.92,
pp0:001) and ethnic diversity (OR ¼ 0.90, pp
0:001) are negatively associated with depression.
Model 3 includes neighborhood characteristics as
well as demographic background characteristics.
Tracking changes in neighborhood stressors (our
primary interest) across models 2 and 3 indicate that
neighborhood stressors contribute independently to
depression, so although the social composition
(individual-level) of urban Canadian neighborhood
matters, context matters as well. While there is a
reduction in the size of the odds ratios for material
deprivation (OR ¼ 1.05, pp0:01) and residential
instability (OR ¼ 1.04, pp0:05) (see model 3 as
ARTICLE IN PRESS
2612
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
compared to model 2), each continues to contribute
independently to depression. When individual characteristics are controlled, neighborhood ethnic
diversity and dependency exert no independent
effects on depression. The variance estimate of the
random intercept remains statistically significant
indicating that neighborhood chronic stress continues to explain some of the variation in depression
when individual and other contextual factors are
included in the model.
In separate analyses, we specified a model with
random slopes for gender in order to determine
whether these effects vary across neighborhoods.
The fixed effect of gender was significant in the
models reported in Table 3; however, allowing the
slope of gender effect to vary revealed no significant
variance component suggesting that the gender
difference in depression is constant across CTs of
25 of Canada’s CMAs. This finding means that we
have a negative response to our research question: Is
neighborhood chronic stress associated with gender
differences in depression? Neighborhood chronic
stress does not seem to differentially affect risk of
depression for men and women.
Taken together, the analyses indicate that neighborhood stressors—residential instability and material deprivation—exert significant effects on
depression, even after controlling for individual
demographic variables and relevant neighborhood
level processes (ethnic diversity and dependency).
Discussion
Rose (1985) argues that we should shift some of
our focus from treatment and prevention at the
individual level to prevention for ‘‘sick populations’’. In this study we asked: ‘‘Why do some
neighborhoods have more depression than others’’?
We argued that at least part of the answer might lie
in the existence of chronic stressors in the local
residential environment. Identification of chronic
stressors that may be uncontrollable at the individual-level, but which are embedded in the social and
political structure of the community, may lead to
population-based initiatives and policies that reduce
the impact of these chronic stressors on the
population.
To investigate this possibility, we selected neighborhood characteristics from the census that were
identified in the literature on material deprivation
and residential instability. Our focus on neighborhood material deprivation and residential instability
was based on the fact that these concepts seem to
possess inherent qualities of chronic stressors. When
present, deprivation and residential instability are
uncontrollable, especially at the individual level,
and tend to be persistent features of a neighborhood. Uncontrollable stressors are documented to
undermine health more so than controllable ones
(Friis, Wittchen, Pfister, & Lieb, 2002; Perkins,
1982; Zautra & Reich, 1983). Factor analysis of the
selected census characteristics produced two chronic
stress indices, neighborhood material deprivation
and residential instability. Two other dimensions
emerged from the data which reflect ethnic diversity
and what we called dependency, which essentially
reflects the age structure of the population.
Several features of our study are particularly
noteworthy: (1) we develop a model that embodies
research on neighborhood context using knowledge
from literature on the stress process which has
guided the majority of the micro-level research on
social antecedents of health; (2) we define and
operationalize two sources of chronic stress at the
neighborhood level as multi-item composite indicators of residential instability and material deprivation; (3) we examine the impact of neighborhood
chronic stress at a national level using a large
general population sample of over 50,000 respondents located in 25 large urban centers which
comprise over 3000 neighborhoods.
Our results indicate that neighborhood residential
instability and material deprivation contribute
significantly to the likelihood of depression above
and beyond the influence of individual-level sociodemographic factors. Moreover, although the estimated net effects of residential instability and
material deprivation are relatively small in magnitude, such effects may be especially important
because even modest shifts in neighborhood instability and deprivation have the ability to impact
thousands of individuals.
Gender has long been recognized as a status
position that frames our access to personal and
social resources. In this study we examined whether
gender framed one’s response to neighborhood
chronic stressors thinking that women might be
more reactive to such stress, however we did not
find random variation in depression by gender
across neighborhoods. So, while women were at
greater risk of depression the gender difference in
depression was relatively constant across our
neighborhood units of CTs. The null findings for
depression do not mean that this will be the case for
ARTICLE IN PRESS
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
other health outcomes. We believe that it is
important to seek those contextual features that
might differentially impact the genders in an effort
to discover the mechanisms that differentially affect
the health of men and women.
Limitations
Census variables are commonly used to characterize neighborhoods, but they tend to be relatively
non-specific. Better measures are required to elucidate pathways and provide opportunities for intervention. A shortcoming of this research is that our
exploration of neighborhood chronic stressors and
their impact on depression was restricted to urban
centers. Much of rural Canada is not tracted so
census characteristics cannot be ascribed to these
areas at the same geographic level. Census information is available for smaller geographic areas (i.e.,
dissemination and enumeration areas), but using
such areas would lead to sparse data at the
individual level and this would affect estimation of
logistic models for low-prevalent outcomes. In a
recent study Weich, Twigg, and Lewis (2006) found
that people living in rural areas experience better
mental health than those living in urban areas.
Rural areas differ from urban areas in availability of
health services, access to transport, social support
and social networks and potentially in exposure to
life event and chronic stressors. Development of
models to investigate the impact of context on
mental health in rural areas is relatively understudied and needs attention.
A further limitation of the data is the lack of
individual information on mediating variables like
personal (e.g., mastery and self-esteem) and social
resources (e.g., social support and social networks),
life events (e.g., divorce) and chronic stress (e.g.,
long-term illness). Questions that would capture this
information were part of the optional content of the
CCHS and administered only to a few health
regions across Canada. This is disappointing
because while knowing that neighborhood chronic
stressors affect the health of individuals is important, it is also important to understand how microlevel resources and stressors either exacerbate or
attenuate the impact of neighborhood chronic
stressors on mental health. In addition, data that
capture neighborhood resources that may attenuate
the negative impact of neighborhood chronic
stressors on health should also be included in future
research.
2613
The Achilles heel in cross-sectional studies of
neighborhood and individual-level mental health is
that it is not possible to get a clear picture of the
extent to which neighborhood accounts for depression. This problem applies not only to research
conducted at the individual level, but inevitably to
approaches that combine the individual and the
ecological. Without longitudinal data it is impossible to state unequivocally that chronic stress at the
level of the neighborhood ‘‘causes’’ depression;
neither can we say that onset and perseverance of
depression leads to downward drift and migration
of those with poor mental health to less economically viable areas; obviously, not everyone in these
neighborhoods migrates in because of downward
drift. The present state of any neighborhood reflects
broader socioeconomic transformations which in
turn influence migration patterns: healthy, economically advantaged individuals may migrate to more
affluent areas of the city from neighborhoods laden
with chronic stress; healthy and economically
disadvantaged individuals may be unable to move
despite desire to escape the neighborhood; and,
unhealthy and economically disadvantaged individuals may migrate in. The point to emphasize here
is that both neighborhoods and individuals are
dynamic with a constant interplay between the
neighborhood and its residents.
Conclusions
Rose (1985) and Weich (1997) emphasize that
approaches to prevention and treatment need to go
beyond the individual to the level of the population
and have suggested that prevention and treatment
within the community or neighborhood should be
considered, not as an alternative to clinical practice,
but as an approach to improve the general health of
a large network of people. Studies that employ
multilevel methods will assist in this endeavor
through identification of communities and neighborhoods that have higher burden of chronic
stressors and a concomitant high number of people
reporting depression or other common mental
disorders. Rose (1993) asks us to reflect on what
might be the ‘‘psychiatric counterpart’’ to the
reduction of cholera through the control of water
pollution. He calls for research that can bridge
clinical skills, epidemiology and the social sciences
and prevention techniques that bridge medicine,
social policy and politics. This approach suggests
several layers, including identifying communities or
ARTICLE IN PRESS
2614
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
neighborhoods in need, public and political partners
that can come to the table with concrete strategies
to address the problem, find solutions and to build
sustainability.
The stigma associated with the common mental
disorders such as depression means that these
damaging illnesses often go largely undetected and
untreated. Environmental and contextual factors
like chronic stress in the neighborhood are modifiable. The findings linking social context to mental
health that are derived from the foregoing analysis
are important because they may help to provide
information that will better help guide policies
specifically aimed at ameliorating persistent health
inequalities.
The findings presented above are important for
health-related policy makers because they suggest
where mental health resources might be better
located to provide interventions that reduce neighborhood chronic stress which may improve the
health of at-risk populations. Such information
could improve the effectiveness of existing community-based mental health care agencies. If we begin
to identify neighborhoods at greater risk then we
can uncover modifiable structural mechanisms
aimed at reducing depression for those living in
those areas; this can happen through strategic
geographic allocation of mental health services
and resources or through community re-development to foster supportive social networks and build
community efficacy. We have known for some time
that people matter in the sense that individual-level
intervention and prevention can benefit health.
Emerging research using multilevel modeling is
pointing to the conclusion that place also matters.
We need to be cognizant of this discovery and work
towards defining those cross-level mechanisms that
explain why neighborhood context is important for
individual-level health.
References
American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (DSM-IV-TR). Washington,
DC: American Psychiatric Press.
Aneshensel, C. S., & Sucoff, C. A. (1996). The neighborhood
context of adolescent mental health. Journal of Health &
Social Behavior, 37(4), 293–310.
Antonovsky, A., & Kats, R. (1967). The life crisis history of a
tool in epidemiological research. Journal of Health and Social
Behavior, 8(1), 15–21.
Arber, S., & Cooper, H. (1999). Gender differences in health in
later life: The new paradox? Social Science & Medicine, 48(1),
61–76.
Beland, Y. (2002). Canadian community health survey—Methodological overview. Health Reports, 13(3), 9–14.
Bingenheimer, J. B., & Raudenbush, S. W. (2004). Statistical and
substantive inferences in public health: Issues in the application of multilevel models. Annual Review of Public Health,
25(1), 53–77.
Boardman, J. D. (2004). Stress and physical health: The role of
neighborhoods as mediating and moderating mechanisms.
Social Science & Medicine, 58(12), 2473–2483.
Boardman, J. D., Finch, B. K., Ellison, C. G., Williams, D. R., &
Jackson, J. S. (2001). Neighborhood disadvantage, stress, and
drug use among adults. Journal of Health & Social Behavior,
42(2), 151–165.
Campbell, K. E., & Lee, B. A. (1992). Sources of personal
neighbor networks: Social integration. Social Forces, 70(4),
1077.
Carstairs, V., & Morris, R. (1989a). Deprivation: Explaining
differences in mortality between Scotland and England and
Wales. British Medical Journal, 299(6704), 886–889.
Carstairs, V., & Morris, R. (1989b). Deprivation and health.
British Medical Journal, 299(6713), 1462.
Cohen, F. (1984). Coping. In J. D. Matarazzo, S. M. Weiss, J. A.
Herd, N. E. Miller, & S. M. Weiss (Eds.), Behavioral health: A
handbook of health enhancement and disease prevention. New
York: Wiley.
Danigelis, N., & Pope, W. (1979). Durkheim’s theory of suicide
as applied to the family: An empirical test. Social Forces,
57(4), 1081–1106.
Denton, M., & Walters, V. (1999). Gender differences in
structural and behavioral determinants of health: An analysis
of the social production of health. Social Science & Medicine,
48(9), 1221–1235.
Duncan, G. J., & Raudenbush, S. W. (1999). Assessing the
effects of context in studies of child and youth development. Educational Psychologist [H.W. Wilson—EDUC],
34(1), 29.
Dunn, J. R. (2002). Housing and inequalities in health: A study of
socioeconomic dimensions of housing and self reported health
from a survey of Vancouver residents. Journal of Epidemiology and Community Health, 56(9), 671–681.
Dunn, J. R., Frohlich, K. L., Ross, N. A., Curtis, L. J., &
Sanmartin, C. (2005). Role of geography in inequalities in
health and human development. In J. Heyman, C. Hertzman,
M. L. Barer, & R. G. Evans (Eds.), Healthier societies: From
analysis to action (pp. 554–598). New York: Oxford University Press.
Dunn, J. R., & Hayes, M. V. (2000). Social inequality, population
health, and housing: A study of two Vancouver neighborhoods. Social Science & Medicine, 51(4), 563–587.
Durkheim, E. (1951). Suicide. New York, NY, USA: Free Press.
Elliott, M. (2000). The stress process in neighborhood context.
Health & Place, 6(4), 287–299.
Faris, R. E. L., & Dunham, H. W. (1939). Mental disorders in
urban areas: An ecological study of schizophrenia and other
psychoses. Chicago: University of Chicago Press.
Feldman, P. J., & Steptoe, A. (2004). How neighborhoods and
physical functioning are related: The roles of neighborhood
socioeconomic status, perceived neighborhood strain, and
individual health risk factors. Annals of Behavioral Medicine,
27(2), 91–99.
Friis, R. H., Wittchen, H. U., Pfister, H., & Lieb, R. (2002). Life
events and changes in the course of depression in young
ARTICLE IN PRESS
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
adults. European Psychiatry: The Journal of the Association of
European Psychiatrists, 17(5), 241–253.
Frohlich, N., & Mustard, C. (1996). A regional comparison of
socioeconomic and health indices in a Canadian province.
Social Science and Medicine, 42(9), 1273–1281.
Gillis, A. R. (1974). Population density and social pathology: The
case of building type, social allowance and juvenile delinquency. Social Forces, 53(2), 306–314.
Gillis, A. R. (1977). High-rise housing and psychological strain.
Journal of Health & Social Behavior, 18(4), 418–431.
Gillis, A. R. (1979a). Coping with crowding: Television, patterns
of activity, and adaptation to high density environments. The
Sociological Quarterly, 20(2), 267–277.
Gillis, A. R. (1979b). Household density and human crowding:
Unravelling a non-linear relationship. Journal of Population,
2(2), 104–117.
Gillis, A. R., & Hagan, J. (1982). Density, delinquency, and
design: Formal and informal control and the built environment. Criminology, 19(4), 514–529.
Gillis, A. R., Richard, M. A., & Hagan, J. (1986). Ethnic
susceptibility to crowding: An empirical analysis. Environment
& Behavior, 18(6), 683–706.
Glazier, R. H., Creatore, M. I., Gozdyra, P., Matheson, F. I.,
Steele, L. S., Boyle, E., et al. (2004). Geographic methods for
understanding and responding to disparities in mammography use in Toronto, Canada. Journal of General Internal
Medicine, 19(9), 952–961.
Guo, G., & Zhao, H. (2000). Multilevel modeling for binary data.
Annual Review of Sociology, 26, 441–462.
Hill, T. D., Ross, C. E., & Angel, R. J. (2005). Neighborhood
disorder, psychophysiological distress, and health. Journal of
Health and Social Behavior, 46(2), 170.
Hyndman, J., Holman, C. D., & Jamrozik, K. (1997). The effect
of spatial definition on the allocation of clients to screening
clinics. Social Science & Medicine, 45(2), 331–340.
Jarman, B. (1983). Identification of underprivileged areas. British
Medical Journal, 286(6379), 1705–1709.
Jarman, B. (1984). Underprivileged areas: Validation and distribution of scores. British Medical Journal, 289(6458), 1587–1592.
Katz, S. J., Zemencuk, J. K., & Hofer, T. P. (2000). Breast cancer
screening in the United States and Canada, 1994: Socioeconomic gradients persist. American Journal of Public
Health, 90(5), 799–803.
Kessler, R. C., Andrew, G., Mroczek, D., Üstün, T. B., &
Wittchen, H. U. (1998). The world health organization
composite international diagnostic interview short form
(CIDI-SF). International Journal of Methods in Psychiatric
Research, 7(4), 171–185.
Kessler, R. C., & McLeod, J. D. (1984). Sex differences in
vulnerability to undesirable life events. American Sociological
Review, 49(5), 620–631.
Kim, J. O. (1978a). Factor analysis: Statistical methods and
practical issues. Beverly Hills, CA: Sage Publications.
Kim, J. O. (1978b). Introduction to factor analysis: What it is and
how to do it. Beverly Hills, CA: Sage Publications.
Kornhauser, R. R. (1978). Social sources of delinquency: An
appraisal of analytic models. Chicago: University of Chicago
Press.
Land, K. C., McCall, P. L., & Cohen, L. E. (1990). Structural
covariates of homicide rates: Are there any invariances across
time and social space? American Journal of Sociology, 95(4),
922–963.
2615
Latkin, C. A., & Curry, A. D. (2003). Stressful neighborhoods
and depression: A prospective study of the impact of
neighborhood disorder. Journal of Health & Social Behavior,
44(1), 34–44.
Lazarus, R. S. (1966). Psychological stress and the coping process.
New York: McGraw-Hill.
Lin, N., & Ensel, W. M. (1989). Life stress and health: Stressors
and resources. American Sociological Review, 54(3), 382–399.
Linsky, A. S. (1969). Community structure and depressive
disorders. Social Problems, 17(1), 120–131.
Magdol, L. (2002). Is moving gendered? The effects of residential
mobility on the psychological well-being of men and women.
Sex Roles, 47(11–12), 553–560.
Pampalon, R., & Raymond, G. (2000). A deprivation index for
health and welfare planning in Quebec. Chronic Diseases in
Canada, 21(3), 104–113.
Pearlin, L. I. (1989). The sociological study of stress. Journal of
Health and Social Behavior, 30(3), 241–256.
Pearlin, L.I., & Skalkidou, A. (1995). Stressors and adaptation in
late life. In M. Gatz (Ed.) American Psychological Association (Eds.), Emerging issues in mental health and aging.
Washington, DC: American Psychological Association.
Perkins, D. V. (1982). The assessment of stress using life event
scales. In L. Goldberger, & S. Breznitz (Eds.), Handbook of
stress: Theoretical and clinical aspects (pp. 320–331). New
York: Free Press.
Propper, C., Jones, K., Bolster, A., Burgess, S., Johnston, R., &
Sarker, R. (2005). Local neighbourhood and mental health:
Evidence from the UK. Social Science & Medicine, 61(10),
2065–2083.
Regoeczi, W. C. (2002). The impact of density: The importance of
nonlinearity and selection on flight and fight responses. Social
Forces, 81(2), 505–530.
Rose, G. (1985). Sick individuals and sick populations. International Journal of Epidemiology, 14(1), 32–38.
Rose, G. (1993). Mental disorder and the strategies of prevention.
Psychological Medicine, 23(3), 553–555.
Ross, C. E., & Bird, C. E. (1994). Sex stratification and
health lifestyle: Consequences for men’s and women’s
perceived health. Journal of Health & Social Behavior, 35(2),
161–178.
Ross, C. E., & Mirowsky, J. (2001). Neighborhood disadvantage,
disorder, and health. Journal of Health and Social Behavior,
42(3), 258–276.
Ross, C. E., Reynolds, J. R., & Geis, K. J. (2000). The contingent meaning of neighborhood stability for residents’
psychological well-being. American Sociological Review,
65(4), 581–597.
Sampson, R. J. (1983). Structural density and criminal victimization. Criminology, 21(2), 276–293.
Sampson, R. J., & Groves, W. B. (1989). Community structure
and crime: Testing social-disorganization theory. American
Journal of Sociology, 94(4), 774–802.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997).
Neighborhoods and violent crime: A multilevel study of
collective efficacy. Science, 277(5328), 918–924.
Serido, J., Almeida, D. M., & Wethington, E. (2004). Chronic
stressors and daily hassles: Unique and interactive relationships with psychological distress. Journal of Health and Social
Behavior, 45(1), 17–33.
Shaw, C. R., & McKay, H. D. (1969). Juvenile delinquency and
urban areas; a study of rates of delinquency in relation to
ARTICLE IN PRESS
2616
F.I. Matheson et al. / Social Science & Medicine 63 (2006) 2604–2616
differential characteristics of local communities in American
cities. USA: Chicago University Press.
Silver, E., Mulvey, E. P., & Swanson, J. W. (2002). Neighborhood
structural characteristics and mental disorder: Faris and
Dunham revisited. Social Science & Medicine, 55(8), 1457–1470.
Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An
introduction to basic and advanced multilevel modeling.
Thousand Oaks, CA: SAGE.
Steffensmeier, R. H. (1984). Suicide and the contemporary
woman: Are male and female suicide rates converging? Sex
Roles, 10(7–8), 613–631.
Townsend, P. (1987). Deprivation. Journal of Social Policy, 16(2),
125–146.
Verbrugge, L. M. (1989). The Twain meet: Empirical explanations of sex differences in health and mortality. Journal of
Health and Social Behavior, 30(3), 282–304.
Weich, S. (1997). Prevention of the common mental disorders:
A public health perspective. Psychological Medicine, 27(4),
757–764.
Weich, S., Twigg, L., & Lewis, G. (2006). Rural/non-rural
differences in rates of common mental disorders in Britain:
Prospective multilevel cohort study. The British Journal of
Psychiatry: The Journal of Mental Science, 188, 51–57.
Wheaton, B. (1999). The nature of stressors. In A. V. Horwitz, &
T. L. Scheid (Eds.), A handbook for the study of mental health:
Social contexts, theories, and systems (pp. 176–197).
New York: Cambridge University Press.
Wilkins, R. (2005). PCCF+ Version 4F User’s Guide. Ottawa:
Health Analysis and Measurement Group, Statistics Canada.
Zautra, A. J., & Reich, J. W. (1983). LIfe events and perceptions
of life quality: Developments in a two-factor approach.
Journal of Community Psychology, 1, 121–132.