ArticlePDF AvailableLiterature Review

Understanding the association between socioeconomic status and physical health: Do negative emotions play a role?


Abstract and Figures

In this article, the authors evaluate the possible roles of negative emotions and cognitions in the association between socioeconomic status (SES) and physical health, focusing on the outcomes of cardiovascular diseases and all-cause mortality. After reviewing the limited direct evidence, the authors examine indirect evidence showing that (a) SES relates to the targeted health outcomes, (b) SES relates to negative emotions and cognitions, and (c) negative emotions and cognitions relate to the targeted health outcomes. The authors present a general framework for understanding the roles of cognitive-emotional factors, suggesting that low-SES environments are stressful and reduce individuals' reserve capacity to manage stress, thereby increasing vulnerability to negative emotions and cognitions. The article concludes with suggestions for future research to better evaluate the proposed model.
No caption available
No caption available
Content may be subject to copyright.
Understanding the Association Between Socioeconomic Status
and Physical Health: Do Negative Emotions Play a Role?
Linda C. Gallo
San Diego State University Karen A. Matthews
University of Pittsburgh School of Medicine
In this article, the authors evaluate the possible roles of negative emotions and cognitions in the
association between socioeconomic status (SES) and physical health, focusing on the outcomes of
cardiovascular diseases and all-cause mortality. After reviewing the limited direct evidence, the authors
examine indirect evidence showing that (a) SES relates to the targeted health outcomes, (b) SES relates
to negative emotions and cognitions, and (c) negative emotions and cognitions relate to the targeted
health outcomes. The authors present a general framework for understanding the roles of cognitive–
emotional factors, suggesting that low-SES environments are stressful and reduce individuals’ reserve
capacity to manage stress, thereby increasing vulnerability to negative emotions and cognitions. The
article concludes with suggestions for future research to better evaluate the proposed model.
Health disparities associated with socioeconomic status (SES)
have existed for centuries (G. D. Smith, Carroll, Rankin, &
Rowan, 1992) and have been recognized by researchers for many
decades (Chapin, 1924; Warren & Sydenstricker, 1916). Recent
research within the United States and other industrialized countries
demonstrates that SES is associated with diverse health outcomes
(Adler, Marmot, McEwen, & Stewart, 1999), and some evidence
suggests that SES inequalities in mortality may even be widening
(e.g., Drever, Whitehead, & Roden, 1996; Pappas, Queen, Hadden,
& Fisher, 1993; Phillimore, Beattie, & Townsend, 1994). Despite
the consistent pattern of these findings, the mechanisms that un-
derlie the graded relationship between SES and health have not
been clearly elucidated. In part, SES disparities in health are
clearly due to differences in the distribution of basic resources
such as health care, nutrition, and sanitary living environments
(e.g., Antonovsky, 1967; see also Lynch, Smith, Kaplan, & House,
2000). This focus may be particularly important to explaining poor
health in groups characterized by poverty, but the impact of SES
on health is not only at the poverty line. Rather, health discrepan-
cies have a monotonic relationship with SES, so that even rela-
tively affluent groups exhibit worse health than their higher SES
counterparts (e.g., Kitagawa & Hauser, 1973; Kraus, Borhani, &
Franti, 1980). Thus, numerous interconnected factors appear to
contribute to SES disparities in health, and researchers have there-
fore cast a wider net in attempting to explain the SES gradient.
One prominent explanation is that cognitive–emotional factors
and disorders play a role in understanding how low SES results in
risk for early death and disability (Adler et al., 1994; Kaplan &
Keil, 1993; Matthews, 1989; Taylor, Repetti, & Seeman, 1997).
Low-SES environments may kindle disproportionate levels of neg-
ative emotions and attitudes, and likewise, these variables may
have deleterious effects on health. However, the literature has not
been reviewed systematically to support or refute this hypothesis.
In the current article, we evaluate the evidence for the tenet that
cognitive–emotional factors may, in part, mediate the relationship
between SES and health. First, we provide a brief overview of the
conceptual issues important in examining socioeconomic and emo-
tional factors. We then review the evidence that bears on the
mediational hypothesis. For mediation by cognitive–emotional
factors to be tenable, the research must show (a) that SES relates
to health; (b) that SES relates to negative emotions and cognitions;
(c) that negative emotions and cognitions relate to health; and
finally (d) that when all factors are examined within a single
methodological frame, the relationship between SES and health is
attenuated if effects for negative cognitive–emotional factors are
statistically controlled (Baron & Kenny, 1986). Unfortunately, few
studies have evaluated these four criteria. The indirect evidence for
mediation is more compelling, and we therefore discuss this re-
search at length. Many detailed reviews concerning the associa-
tions between SES and health and between cognitive–emotional
factors and health have been published recently. Rather than du-
plicate this work, we summarize and update it. We focus on
cardiovascular diseases—the leading cause of mortality in the
United States (American Heart Association, 2000)—and all-cause
mortality because these outcomes have been examined with suf-
ficient frequency and rigor to provide an estimate of the proposed
Linda C. Gallo, Department of Psychology, San Diego State University;
Karen A. Matthews, Department of Psychiatry, University of Pittsburgh
School of Medicine.
Portions of this review were presented at the New York Academy of
Sciences Conference on Social Class and Health, Washington, DC, May
1999, and the conference proceedings are published in the Annals of the
New York Academy of Sciences (Gallo & Matthews, 1999). This research
was supported in part by National Institutes of Health Grants HL25767,
HL07560, HL65111, and HL65112 and by the John D. and Catherine T.
MacArthur Foundation Research Network on Socioeconomic Status and
Health. We thank Edith Chen, Timothy W. Smith, and Shelley E. Taylor
for their helpful comments on earlier versions of this article.
Correspondence concerning this article should be addressed to Linda C.
Gallo, Department of Psychology, San Diego State University, 6363 Al-
varado Court, Suite 103, San Diego, California 92120. E-mail:
Psychological Bulletin Copyright 2003 by the American Psychological Association, Inc.
2003, Vol. 129, No. 1, 10–51 0033-2909/03/$12.00 DOI: 10.1037/0033-2909.129.1.10
links and because SES gradients are particularly strong for these
health outcomes (e.g., M. A. Gonzalez, Rodriguez, & Calero,
1998; Kaplan & Keil, 1993; Lynch et al., 2000). The associations
between SES and cognitiveemotional factors have not been pre-
sented in any recent, enumerative reviews (but see the review of
SES and psychiatric disorders by Kohn, Dohrenwend, & Mi-
rotznik, 1998), and we therefore analyze this research in more
detail. Following our review and critical analysis, we present a
framework for understanding the pathways that may dynamically
link SES, cognitiveemotional factors, and health. Finally, we
conclude with recommendations for future research to better ad-
dress the proposed mediation hypothesis.
Conceptualizing SES
SES is an aggregate concept defined according to ones level of
resources or prestige in relation to others (Krieger, Williams, &
Moss, 1997; Lynch & Kaplan, 2000). Resource-based measures
assess access to material and social assets, including income,
wealth, and educational attainment. Prestige-based measures refer
to an individuals rank or status in a social hierarchy, typically
evaluated by access to and consumption of goods, services, and
knowledge as linked to occupational prestige and education. Social
class refers to groups defined by interdependent economic and
legal relationships, based on an individuals structural location
within the economy (e.g., employer vs. employee).
SES can be assessed at the level of the individual, household
unit, and neighborhood or community. Most commonly, SES is
assessed through single individual-level indicators, but these may
not accurately characterize the status of the family or household.
Moreover, research shows that neighborhood-level indicators (e.g.,
Eaton & Muntaner, 1999; Haan, Kaplan, & Camacho, 1987) and
contextual income distribution (Daly, Duncan, Kaplan, & Lynch,
1998; Soobader & LeClere, 1999) predict health above the effects
of individual-level indicators. Studies that evaluate SES at multiple
levels can therefore provide a more accurate assessment of the
association between SES and health outcomes (e.g., Diez-Roux et
al., 1997; B. P. Kennedy, Kawachi, Glass, & Prothrow-Stith,
1998). Perhaps more important, they provide clues to the material,
social, and psychological mechanisms that may account for the
association between SES and health.
In research concerning health, the most commonly used indica-
tors of SES include educational attainment, occupational prestige,
and income. These indicators are related, but not fully overlapping,
and they may impact health through disparate pathways. Further-
more, each is associated with distinct advantages and disadvan-
tages for research (Krieger et al., 1997). For example, questions
about income are prone to missing and distorted responses.
Occupation-based measures cannot be used to indicate SES for
individuals who are not working for reasons of retirement, unem-
ployment, homemaking, or caretaking. Income and occupation
may also be influenced by reverse causationthat is, the effects of
psychiatric or medical illness on SES. In these respects, education
affords some advantage because ones education is often com-
pleted before the onset of chronic illness, and individuals are
typically willing and able to accurately represent their education.
Overall, education may be the most appropriate assessment of SES
for women because of the proportion of women who do not work
outside the home (approximately 40%; Bureau of Labor Statistics,
2000). Their own income or occupation may therefore underrep-
resent the SES of the household. On the other hand, the implica-
tions of education for economic and health standing may differ by
age, ethnicity, and gender (Krieger et al., 1997; Oliver & Shapiro,
1995). Thus, SES indicators should be chosen according to the
specific research questions and populations of interest.
Another factor important in measuring SES is the temporal
nature of the assessment. SES tends to be stable across the life span
and across generations of family members (e.g., Lynch, Kaplan, &
Salonen, 1997), but some indicators of SES are quite dynamic. For
example, more than half of U.S. households sampled experienced
an increase of 50% or a decrease of more than 33% in monthly
income in 1984 (U.S. Bureau of the Census, 1996). Experiencing
a single marked income drop during a 5-year period resulted in a
30% increase in mortality risk, whereas two or more decreases
predicted a 70% increase in mortality risk (Duncan, 1996). An-
other recent study found a doseresponse association between the
frequency of exposure to low income and physical, psychological
(i.e., depression, hostility, low optimism), and cognitive function-
ing (Lynch, Kaplan, & Shema, 1997). Similarly, accumulated
exposure to low occupational prestige of parents and low SES at
several points in young adulthood predicted self-rated health in
midlife (Power, Manor, & Matthews, 1999). Thus, the common
approach of measuring SES at a single time point may be inade-
quate to capture the full impact of exposure to low-SES
An incorporation of time in assessing SES would also help
elucidate directional effects in the association between SES and
health, as proposed in the social causation and social drift hypoth-
eses. According to the social causation perspective, socioeconomic
standing has a causal role in determining health or emotional
problems. Social drift interpretations assert that individuals with
worse physical or emotional health may drift downthe socio-
economic hierarchy or fail to rise in socioeconomic standing as
would be expected on the basis of familial origins or changes in
societal affluence. That is, the social drift model views health
problems as exerting a causal influence on social status. These
directional hypotheses have been the subject of considerable re-
search and discussion in the literature (e.g., B. P. Dohrenwend et
al., 1998; Eaton & Muntaner, 1999; Kessler, 1982; Lichtenstein,
Harris, Pedersen, & McClearn, 1993; Marmot, Kogevinas, & El-
ston, 1987; J. P. Smith, 1999). In actuality, the associations are
probably dynamic and reciprocal. We further discuss these issues
below, but it is important to note that we view social drift as a
phenomenon that, in part, maintains the adverse effects of SES.
To summarize, in evaluating research concerning the association
between SES and mental health and the association between SES
and physical health, the adequacy of the SES assessment is an
important consideration. The fact that studies of SES and health
incorporating a single indicator measured at one level and point in
time have revealed statistically significant effects attests to the
robust nature of the relationship. Nevertheless, this common mea-
surement technique may underestimate the nature and scope of
SEShealth disparities and could impede efforts to delineate the
numerous paths that create them. Future research should attempt to
examine how the multiple levels of SES interrelate and affect
health and should also consider the temporal and dynamic nature
of socioeconomic position (for further discussion and review, see
Anderson, 1999; Krieger et al., 1997; Robert & House, 2000).
Conceptualizing Emotion
The conceptual and theoretical description of emotion has been
the source of considerable attention and disagreement in the liter-
ature (e.g., Ekman & Davidson, 1994). However, most researchers
concur that emotions comprise affective, cognitive, and behavioral
components, along with concomitant physiological changes that
occur to ready the body for action. Moreover, many definitions and
theories assert that emotional experiences reflect two underlying
dimensions that characterize the valence, or pleasantness, and
arousal, or attention, of the emotion (Plutchik, 1980; Russell,
1997; Watson & Tellegen, 1985). Emotions that are similar in
terms of valence and/or arousal tend to cluster (e.g., L. A. Feld-
man, 1995). For example, individuals who report depressed affect
also frequently report feeling anxious (Clark & Watson, 1991).
Negative emotions and cognitions are also closely linked (Clore,
1994). For example, hostile cognition frequently occurs in con-
junction with angry affect (T. W. Smith, 1994), and hopeless
cognition often occurs with depressed affect (G. W. Brown &
Harris, 1978).
Emotion researchers continue to disagree about whether positive
affect and negative affect form two poles of a single dimension or
distinct dimensions. Substantial evidence suggests that these con-
structs are orthogonal (Watson & Clark, 1997), yet a recent com-
prehensive review concluded that bipolarity represents the best fit
to existing data (Russell & Carroll, 1999). Still other research has
found support for both models of affectivity, showing that the
structure of emotion is importantly influenced by individual-level
variables (Reich, Zautra, & Potter, 2001). Resolving this issue is
beyond the scope of this review. However, for our purposes, we
consider the possibility that positive and negative affect represent
discrete dimensions and, as such, that they may provide distinct
information about the factors connecting SES with health.
Regardless of valence and arousal, all emotions are fundamen-
tally adaptive inasmuch as they serve a communicative function
(Lazarus, 1991). Maladaptive emotions are distinguished from
adaptive emotions by their inappropriateness, frequency, intensity,
and duration (Frijda, 1994). Although maladaptive emotional ex-
periences form an integral part of psychiatric disorders (e.g.,
depressed mood is fundamental to clinical depression), they should
be not be considered interchangeable (Santor & Coyne, 2001).
Emotional disorders are broader syndromes that comprise a cluster
of symptoms, behaviors, and cognitiveaffective processes. Here,
we summarize research that has analyzed cognitiveemotional
states or symptoms as well as clinical disorders. This distinction is
important because psychiatric disorders occur relatively infre-
quently in the population, whereas symptoms are fairly common.
In specific, our review addresses depression, anxiety, anger and
hostility, and hopelessness. These constructs clearly overlap sub-
stantially with more general constructs such as psychological
distress and with trait measures of negative affect (e.g., Clark &
Watson, 1991; Clark, Watson, & Mineka, 1994). However,
broader conceptualizations have not received rigorous attention in
physical health research, perhaps in part because they may be too
vague to contribute information regarding mechanisms or targets
for intervention. In addition, broader conceptualizations of nega-
tive affectivity (e.g., Neuroticism) often relate to somatic com-
plaints but not necessarily to objective health outcomes (Watson &
Pennebaker, 1989). Thus, we focus explicitly on the cognitive
emotional constructs that, at this point, have the greatest potential
to help explain the health disadvantage experienced by people with
low SES.
Definition and Measurement of Depression
Depression is an unpleasant emotion that is generally associated
with low arousal. Clinical depression represents a constellation of
emotional and behavioral symptoms forming several Diagnostic
and Statistical Manual of Mental Disorders (4th ed.; DSM–IV;
American Psychiatric Association, 1994) diagnoses. A diagnosis
of major depression entails depressed mood and/or a lack of
interest or pleasure in most activities for at least 2 weeks. Four
additional symptoms must be present, such as a change in appetite
or sleeping habits, fatigue, psychomotor retardation or agitation,
thoughts of guilt or worthlessness, problems thinking or concen-
trating, and suicidal ideation or intent. Over a 1-month period,
approximately 5% of the U.S. population experiences a major
depressive episode (Blazer, Kessler, McGonagle, & Swartz, 1994).
Lifetime prevalence is considerably higher, at about 13% (Kessler
et al., 1994). Prevalence is likely to be elevated in certain sub-
groups, including medically ill populations (e.g., Stevens, Meri-
kangas, & Merikangas, 1995). For example, estimates of major
depression prevalence rates among heart disease patients range
from 15% to 23% (Carney, Freedland, Sheline, & Weiss, 1997;
Frasure-Smith, Lespe´rance, & Talajic, 1993; M. B. Gonzalez et al.,
1996). Other clinical depressive diagnoses include minor depres-
sion, depressive symptoms that are subthreshold in severity to
major depression, and dysthymia, subthreshold depressive symp-
toms that endure for at least 2 years. Lifetime prevalence of
dysthymia has been estimated at 5% (Kessler et al., 1994), and
lifetime prevalence of subthreshold depressive symptoms is much
higher, at approximately 23% (Horwath, Johnson, Klerman, &
Weissman, 1992).
The research concerning depression and physical health has
typically examined major depression or depressive symptoms.
Research pertaining to socioeconomic predictors of depression has
analyzed depressive symptoms as well as clinical depressive di-
agnoses. Studies that use measures of depressive symptoms can
provide information concerning the doseresponse relationship
between depression and health outcomes, which occurs if the
severity of health problems increases with the severity of the
symptoms. This pattern of association suggests a gradient, or linear
relationship, between depression and health. In contrast, a thresh-
old interpretation of the depression and health relationship would
be tenable if the research suggested that only those individuals
with major depression suffered a health disadvantage. On the other
hand, these patterns are difficult to discriminate because high
levels of depressive symptoms portend risk for full-syndrome
depression (Judd & Akiskal, 2000), which in turn may have a
deleterious effect on health.
With some exceptions, most studies concerning clinical depres-
sion have used well-validated structured interview assessments,
such as the Diagnostic Interview Schedule (Robins, Helzer,
Croughan, & Ratcliff, 1981) or the Schedule for Affective Disor-
ders and Schizophrenia (SADS; Endicott & Spitzer, 1979). Some
of the research concerning depressive symptoms has used well-
validated measures such as the Beck Depression Inventory (BDI;
Beck & Beamesderfer, 1974) or the Center for Epidemiological
Studies Depression Scale (CESD; Radloff, 1977), whereas other
research has relied on study-specific measures, sometimes com-
posed of one or only a few questions. Less confidence can be
placed in findings derived from very few items, given the lower
reliability typically associated with such assessment tools. Symp-
tom measures can be examined on a continuum, with higher scores
indicating higher levels of symptoms, or they can be dichotomized
using a standard cutoff to form depressed and nondepressed
Definition and Measurement of Hopelessness
Hopelessness describes negative cognition about the self and the
future, which is likely to accompany severe negative emotions
such as those associated with clinical depression. Along these
lines, hopelessness is often conceptualized as a symptom of de-
pression (G. W. Brown & Harris, 1978). However, some research
suggests that this construct affects health beyond its association
with depression (Everson et al., 1996). Furthermore, hopelessness
and depression are uniquely associated with other psychological
phenomena, including suicidal intent (Greene, 1989). Studies that
have examined hopelessness have generally assessed this construct
using a few items or even a single one. Despite the limitations
associated with this type of assessment, the research has been
fairly consistent in showing that hopelessness has important health
Definition and Measurement of Anxiety
Anxiety is an emotion of negative valence and positive arousal.
Fear or apprehension about the future represents the cardinal
feature of all anxiety disorders, but corollary symptoms vary
widely, consistent with the number of DSMIV (American Psy-
chiatric Association, 1994) anxiety diagnoses. Diagnostic criteria
for generalized anxiety disorderthe least situation- and stimulus-
specific diagnosisinclude excessive anxiety and worry for at
least 6 months; difficulty controlling the anxiety; and three addi-
tional symptoms, such as restlessness, fatigue, difficulty thinking
or concentrating, irritability, muscle tension, and sleep distur-
bance. Panic disorder has frequently been examined in relationship
to physical health. Panic disorder involves recurrent attacks of
sudden intense fear that occur without an identifiable cause and
that are accompanied by somatic (e.g., hot flashes, shortness of
breath, sweating, gastrointestinal distress) and cognitive symptoms
(e.g., fear of losing control or dying). Agoraphobia, or the enact-
ment of behaviors designed to avoid triggers of panic attacks such
as refusal to drive or leave ones house, may occur in association
with panic disorder. Prevalence rates of anxiety disorders vary
considerably. In a 1-year period, approximately 0.9% of the pop-
ulation meet criteria for panic disorder, whereas about 9.7% meet
criteria for any phobic diagnosis (Eaton, Dryman, & Weissman,
1991). Most prior research concerning SES has focused on anxiety
disorders, whereas many studies related to physical health have
examined anxiety symptoms (e.g., phobic symptoms or general-
ized worry). A number of studies of physical health outcomes have
applied the CrownCrisp Experiential Index, a reliable and valid
measure of somatic anxiety, phobic anxiety, and obsessionality
(see, e.g., M. W. Ross & Hafner, 1990). Other studies have used
less well-known measures of anxiety, impeding the evaluation of
its health correlates.
Definition and Measurement of Hostility
Research concerning hostility involves analyzation of three in-
terrelated emotional, behavioral, and cognitive constructs (T. W.
Smith, 1994), consistent with the definition of emotion outlined
above. Anger represents the emotional component and is charac-
terized by negative valence and moderate to high arousal. Resent-
ment, scorn, and derision are closely related emotional constructs.
The behavioral component consists of verbal and physical aggres-
sive acts, involving harmful intent. Hostility represents negative
attitudes and beliefs about others, such as cynicism and mistrust.
Studies of hostility have variously used interview ratings or ques-
tionnaire assessment methods, and the two strategies have pro-
duced only moderately related findings (Dembroski, MacDougall,
Williams, Haney, & Blumenthal, 1985). Interview ratings typically
assess the behavioral aspects of this construct. Perhaps the most
frequently used hostility self-report measure is the CookMedley
hostility inventory (Ho; Cook & Medley, 1954), which is best
described as a measure of hostile cognition (e.g., Barefoot, Dodge,
Peterson, Dahlstrom, & Williams, 1989; T. W. Smith & Frohm,
1985). Despite widespread usage, the Ho has been criticized for
questionable construct validity, heterogeneity of item content, and
close association with Neuroticism (e.g., Barefoot, 1992; Barefoot
et al., 1989; Carmody, Crossen, & Wiens, 1989).
SES and Health
The evidence for SES disparities in health is long-standing and
compelling. More than 3 decades ago, Antonovsky (1967) found
consistent support for an inverse relationship between SES and
mortality in a review of more than 30 studies. The coherence of the
association was notable given the diverse populations and meth-
odologies used. More recently, many well-designed studies have
identified a socioeconomic differential in morbidity and mortality
in the United States (e.g., Backlund, Sorlie, & Johnson, 1996; J. J.
Feldman, Makuc, Kleinman, & Cornoni-Huntley, 1989; Kitagawa
& Hauser, 1973; Lantz et al., 1998; Pappas et al., 1993) and in
other industrialized countries (e.g., Doornbos & Kromhout, 1990;
Holme, Helgeland, Hjermann, & Leren, 1982; Marmot, Shipley, &
Rose, 1984; Marmot et al., 1991; Salonen, 1982). A number of
reviews have provided detailed and comprehensive analyses of this
research and have presented the conclusion that SES has a pro-
found influence on health (Adler et al., 1994; Adler, Boyce,
Chesney, Folkman, & Syme, 1993; Carroll, Bennett, & Davey
Smith, 1993; Lynch et al., 2000; Marmot et al., 1987; Williams,
1990; Williams & Collins, 1995).
The association between SES and health can be summarized as
monotonic, so that as individuals or groups move up in the SES
continuum, mortality and morbidity rates decrease, with the gra-
dient steepest at the lowest levels (Backlund et al., 1996; Ecob &
Davey Smith, 1999). This pattern emerges across socioeconomic
indicators (e.g., occupation, income, education), suggesting that a
general underlying social ordering is important (e.g., Adler et al.,
1994). In addition, both community (e.g., census tract) and indi-
vidual indicators of SES predict health outcomes (e.g., Adler et al.,
1993; Robert & House, 2000). The gradient extends to mortality
from all causes (e.g., Adler et al., 1994; Anderson & Armstead,
1995) and to diverse specific health outcomes, including cardio-
vascular disease, renal disease, diabetes, cancer, arthritis, and
infant mortality (e.g., Illsey & Baker, 1991; Pincus, Callahan, &
Burkhauser, 1987). Two reviews have concluded that SES affects
the development and progression of cardiovascular diseases (M. A.
Gonzalez, Rodriguez, & Calero, 1998; Kaplan & Keil, 1993).
Researchers have examined a number of possible explanations
for SEShealth disparities, including access to health care, resi-
dential characteristics, environmental exposure, physiological pro-
cesses, health behaviors, and psychosocial factors (e.g., Anderson
& Armstead, 1995; Kaplan & Keil, 1993; Macintyre, 1997; Wil-
liams, 1990). What is clear is that none of these factors provides a
complete explanation for the gradient. For example, SES relates to
health behaviors and established biological health risk factors, but
these variables do not account statistically for the gradient (e.g.,
Lantz et al., 1998; see also Adler et al., 1994; Evans, Barer, &
Marmor, 1994; Marmot, Bobak, & Smith, 1995, for discussion and
review). Inequalities in access to health care do not provide an
exhaustive explanation, as evidenced by the fact that countries
with nationally funded health care programs show a linear, albeit
less steep, association between SES and health (Adler et al., 1993;
Williams, 1990). SES gradients also exist for causes of death that
are not amenable to medical treatment (Marmot et al., 1987). In
addition, although health problems do affect social mobility, social
drift does not appear to explain a substantial portion of the asso-
ciation between SES and health (e.g., Haan et al., 1987; Macintyre,
1997; Wilkinson, 1986). Thus, the paths linking SES to health are
multifarious and not amenable to simple solutions.
Integrative Studies of SES, CognitiveEmotional Factors,
and Health: Evidence for Mediation?
Few studies meet the criteria outlined above as necessary to
examine whether cognitive and emotional factors partially mediate
the association between SES and health. Consequently, we do not
limit this portion of our review to studies focusing on mortality and
cardiovascular diseases, including any study that simultaneously
examined SES, cognitive or emotional factors, and a physical
health outcome (regardless of whether they tested all statistical
criteria). Seven studies included all three types of variables.
Cohen, Kaplan, and Salonen (1999) examined the extent to
which psychosocial characteristics (e.g., stress, personal control,
social support, anger and hostility, depression and hopelessness)
and health behaviors contributed to the association between SES
and perceived health in samples from the United States (the Harris
Poll Study) and Finland (the Kuopio Ischemic Heart Disease
Study). Both studies revealed a graded, inverse association be-
tween SES and the odds of poor self-rated health. In the U.S.
study, income and education were also significantly associated
with personal control, social support, stress, life events, and anger,
and these variables predicted the likelihood of perceived poor
health. Statistical control for psychosocial variables attenuated the
odds ratio (OR) for poor health for all lower education groups
compared with the beyond college group as follows: from 5.0
to 3.6 for the 8th grade or less group, from 2.6 to 2.2 for the some
high school or degree group, and from 1.6 to 1.4 for the some
college or degree group. Excess risk for the two lower income
groups compared with the highest income group (i.e., $45,000)
was also attenuated (ORs reduced from 4.6 to 2.6, $10,000 or less
group; from 3.1 to 2.2, $10,000$20,000 group). In the Finnish
study, income and education were strongly inversely associated
with hostility, depression, and hopelessness. Social support was
positively related to income but only marginally to education, and
life events scores were unrelated to income and positively related
to education. Increases in life events, depression, and hopelessness
were positively associated with the risk of perceived poor health,
and social support was inversely associated with the risk of per-
ceived poor health. Hostility did not predict perceived health.
Statistical control for psychosocial variables attenuated the OR for
perceived poor health associated with less than elementary educa-
tion from 4.4 to 2.5, for elementary or part junior high from 3.4
to 2.3, and for less than high school from 2.1 to 1.7 (with high
school education serving as the comparison). Similarly, ORs for
the lowest income group were attenuated from 5.3 to 3.4, the next
lowest from 2.9 to 1.7, the middle income group from 2.2 to 1.5,
and the next to highest income group from 1.2 to 0.9, when
compared with the highest income group.
Thus, in both studies statistical control for the psychosocial
variables reduced the excess risk for poor health associated with
the lowest income and education groups by the greatest proportion
but attenuated risk for all lower education and income groups to
varying degrees. In commenting on this pattern of findings, Cohen
et al. (1999) suggested that psychological characteristics might be
more likely to covary with other sources of risk in lower SES
groups (i.e., environmental factors, poor nutrition) and, therefore,
that psychosocial factors capture multiple influences on health at
low levels of SES. Limitations of this study include the use of a
cross-sectional design, a self-reported health outcome, and some
nonvalidated measures of psychosocial constructs. In addition,
interactive associations were not tested statistically, nor were types
of mediators examined individually (impeding identification of the
relative impact of the cognitive and emotional variables). None-
theless, the results support the view that psychosocial functioning
in part mediates the association between SES and perceived health.
Levenstein and Kaplan (1998) investigated the degree to which
psychological characteristics explained the association between
SES and the occurrence of ulcer. For women, educational attain-
ment was a significant predictor of new ulcers during a 9-year
follow-up period. Compared with participants who attended at
least some college, those with less than a high school education
were more than three times as likely (OR 3.3), and those with
a high school degree were twice as likely (OR 2.0), to develop
an ulcer. For men, the association between educational attainment
and incident ulcer was only marginally significant (OR 1.9, less
than high school; OR 1.8, high school graduates). For women,
adjustment for a composite psychological index comprising de-
pression, hostility, anomy, and personal uncertainty reduced the
OR for new ulcer associated with the less than high school group
from 3.3 to 2.4 and from 2.0 to 1.5 for high school graduates. For
men, adjustment for psychological factors had very little impact on
the association between education and ulcer incidence (with ORs
reduced from 1.9 to 1.8, less than high school; from 1.8 to 1.7, high
school graduates), but given that the initial effects were nonsig-
nificant, this lack of mediation effect is not particularly informa-
tive. The extent to which psychosocial factors related to SES, or
were independently related to the occurrence of ulcer, was not
reported. Levenstein and Kaplan concluded that overall, the results
suggest that psychosocial factors (in addition to other factors)
might contribute to the association between SES and ulcer, at least
in women.
Lynch, Kaplan, Cohen, Tuomilehto, and Salonen (1996) exam-
ined the effects of SES on incident myocardial infarction (MI),
cardiovascular mortality, and all-cause mortality across a 410-
year follow-up period and evaluated the degree to which biologi-
cal, behavioral, and psychosocial risk factors explained the effects
of SES. The associations between the psychosocial factors and the
health outcomes and between the psychosocial factors and SES
were not reported. Nonetheless, simultaneous adjustment for de-
pression, hopelessness, marital status, participation in organiza-
tions, and social support reduced the excess risk associated with
the lowest income group by 52% (from 3.14 to 2.03) for all-cause
mortality and by 57% (from 2.66 to 1.71) for cardiovascular
mortality. In contrast, very little of the relationship between SES
and incident MI was explained by psychosocial factors (OR for the
lowest income group was reduced from 4.34 to 4.25). Because
social (i.e., resource) and cognitiveemotional factors were con-
sidered in a single block, the independent effects of depression and
hopelessness cannot be determined. Furthermore, interactions
among variables were not tested. The results of this study suggest
that psychosocial factors contribute to the association between
SES and all-cause or cardiovascular mortality but not to the
association between SES and MI.
Another study (Gump, Matthews, & Ra¨ikko¨nen, 1999) used
structural equation monitoring to examine the interrelationships
among family and neighborhood SES, hostility, cardiovascular
reactivity to stressful tasks, and left ventricular mass, in a group of
children and adolescents. Low family and neighborhood SES
predicted higher levels of hostility in Black but not White partic-
ipants. Low SES was also associated with augmented cardiovas-
cular reactivity, which, in turn, was related to left ventricular mass.
For Black children, the association between SES and reactivity
was mediated by individual differences in hostility. This study is
limited by use of a cross-sectional design, with a relatively small
and nonrandomly selected sample. In addition, children and ado-
lescents were examined, and the results might not generalize to
adults. Finally, clinical outcomes were not examined, although
cardiovascular reactivity represents a possible physiological path-
way through which psychosocial factors may affect cardiovascular
health outcomes (Manuck, 1994), and left ventricular mass pre-
dicts cardiovascular morbidity and mortality (Casale, Devereux, &
Milner, 1986; Levy, Garrison, Savage, Kannel, & Castelli, 1990).
In summary, this study provides preliminary evidence that in
Blacks, SES might influence the development of cardiovascular
disease, in part, through its association with hostility.
Two articles based on the Beta Blocker Heart Attack Trial
(BHAT) examined the extent to which psychosocial factors ex-
plained the association between SES and cardiovascular outcomes.
In the first, Ruberman, Weinblatt, Goldberg, and Chaudhary
(1984) examined psychosocial factors including social isolation,
stress, depression, Type A behavior, and educational attainment on
survival in men recovering from MI. Individuals with lower edu-
cation had higher mortality rates following MI compared with
participants with more education. In addition, less-educated indi-
viduals had higher levels of stress and social isolation. Type A
behavior and depression were not related to educational attain-
ment, and overall, the proportion of men with high depression
scores was quite low. Life stress and social isolation predicted
mortality rates. When education, life stress, and social isolation
were included simultaneously in a regression equation, education
no longer predicted survival. It is notable that occupational status
formed a component of the life stress assessment because this
creates interpretive ambiguities. That is, social isolation and stress
may have mediated the association between education and sur-
vival, or life stress and education could simply have represented
confounded indicators of SES. In addition, as noted, depression
rates were very low in this study; thus, range restriction may have
limited power.
A later study (Ickovics, Viscoli, & Horwitz, 1997) from the
BHAT examined the associations among SES, depression, and
functional status in individuals 12 months after MI. The initial
criteria for mediation were fulfilledSES was related to depres-
sion, life stress, and social support, and SES was inversely asso-
ciated with improvement in functional status. Likewise, life stress,
social isolation, and depression predicted changes in functional
status. However, after controlling for clinical features (i.e., medical
history, health behaviors, treatment group, severity of MI) and
demographic variables (i.e., age, ethnicity), the psychosocial fac-
tors explained only a small amount of the risk for no improvement
associated with lower social class; the OR for showing no im-
provement for low versus high social class decreased from 1.62
to 1.51. One limitation of this study is that it did not allow for
possible interactions among stress, social isolation, and depression.
In addition, it would have been informative if the authors had
presented information about potential discrepancies between anal-
yses examining the mediational effects of psychosocial factors
both before and after controlling for clinical features, thereby
allowing for the possible indirect effects of psychosocial factors
(e.g., through comorbid health problems, severity of MI, health
Finally, Fiscella and Franks (1997) examined the extent to
which psychological factors explained the association between
SES and mortality over a 1216-year follow-up period. Individuals
with low income had significantly higher levels of depression,
hopelessness, and life dissatisfaction at baseline compared with
their higher SES counterparts, and these psychosocial constructs
significantly predicted mortality risk. In multivariate analyses,
statistical control for depression and hopelessness accounted for a
small amount of the association between SES and mortality (3%
11%). Limitations of this research include the fact that psycholog-
ical factors were measured over a brief period and that they were
not reassessed during follow-up. In addition, the measures of
depression and hopelessness consisted of only a few items.
In summary, few studies have integrated socioeconomic, cog-
nitive or emotional, and health variables within the same method-
ological framework. The studies that have evaluated these factors
concurrently provide inconclusive evidence for the model, given
their conflicting findings and methodological limitations. None of
the studies statistically tested interactions among psychosocial
factors, or between SES and psychosocial factors, and they may
therefore underestimate the true effects of negative emotions and
cognitions (and other psychosocial paths). Therefore, we now turn
to the more comprehensive indirect evidence from studies that
have examined the associations between SES and negative emo-
tions or cognitions and between negative emotions or cognitions
and health. Cognitive and emotional factors have received consid-
erable attention as health risk factors in recent years, and a number
of excellent reviews are available through which to evaluate this
portion of the mediation hypothesis. Rather than duplicate this
work, we provide a synopsis and update to these reviews. We
address the associations between SES and cognitiveemotional
factors in greater detail.
Negative CognitiveEmotional Factors and SES
In the following sections, we review the research that has
examined SES in relation to cognitive and emotional symptoms
and to psychiatric disorders that have negative emotion as a
primary component, which some evidence indicates could relate to
health (i.e., major and minor depression, dysthymia, phobic and
panic disorders, generalized anxiety disorder). We review works
that have been published since 1990, as well as older studies that
are frequently cited in the literature. Reviewed studies were iden-
tified through MEDLINE and PsycINFO searches, crossing the
key words hostility,anger,depression,anxiety,hopelessness,af-
fective disorder, and anxiety disorder with socioeconomic status
and social status (searches were limited to studies that were
published in the English language between 1990 and 2001 and that
used an adult population sample). We identified additional studies
through the ancestry method. An effort was made to avoid pre-
senting redundant findings. Thus, when several studies from the
same research were published, we attempted to present the most
recent findings. Cross-sectional studies are reviewed summarily,
and because of their ability to provide information concerning
directionality, prospective studies are reviewed in detail. To ex-
amine the association between SES and emotional disorders, we
rely primarily on evidence from the Epidemiologic Catchment
Area (ECA; Robins & Regier, 1991) and National Comorbidity
Studies (NCS; Kessler et al., 1994). These studies both involved
the administration of structured psychiatric interviews to large
probability samples of U.S. residents. The authors of the ECA
administered the Diagnostic Interview Schedule for DSMIII (Di-
agnostic and Statistical Manual of Mental Disorders, 3rd ed.;
American Psychiatric Association, 1980) to approximately 20,000
U.S. residents over age 18. The authors of the NCS used the
Composite International Diagnostic Interview for DSMIIIR(Di-
agnostic and Statistical Manual of Mental Disorders, 3rd ed., rev.;
American Psychiatric Association, 1987) to assess more
than 8,000 U.S. residents between the ages of 15 and 54.
We rated all studies for methodological rigor on a scale of 0 to 6,
according to the following: design (1 point for including a pro-
spective component, thereby facilitating the determination of di-
rectionality), sample selection (1 point for random or population-
based sampling, which would improve the generalizability of the
findings), sample size (1 point for N300 and 0 points for smaller
samples, which would have lower power), measure of SES (1 point
for at least one continuous assessment of SES, 0 points if the study
dichotomized SES, thereby attenuating power and preventing ex-
amination of a gradient effect), measure of emotion and/or cogni-
tion (1 point for use of a structured diagnostic interview or well-
validated measure of symptoms and 0 points for study-specific
and hence potentially less reliable or validmeasures), and
control or restriction of potential confounds (1 point for control of
factors such as age, sex, and ethnicitya procedure that would
facilitate examining the uniqueeffect of SES). We then grouped
studies according to whether they provided evidence of an inverse
relationship between SES and the cognitiveemotional construct
(i.e., a statistically significant association, at p.05), mixed
evidence (i.e., an inverse association in some groups but not
others), or null evidence (i.e., no significant association between
SES and the cognitiveemotional factor). To examine the possi-
bility that associations depend on the type of socioeconomic indi-
cator, we examined the proportion of positive, mixed, and null
evidence separately for education, income, occupation, and other
measures of SES.
Depression and Hopelessness and SES
Table 1 summarizes the research concerning the association
between depression or hopelessness and SES. Nine of the studies
reviewed examined the cross-sectional association between de-
pressive or hopeless symptoms and indicators of SES. Of the five
that examined education, two showed evidence of an inverse,
linear association (Lynch, Kaplan, & Salonen, 1997; Salokangas &
Putanen, 1998), and the remaining three identified mixed evidence
(Comstock & Helsing, 1976; Craig & Van Natta, 1979; West,
Reed, & Gildengorin, 1998). Three out of four studies identified an
inverse association between income and depressive symptoms
(Fiscella & Franks, 1997; Salokangas & Putanen, 1998; West et
al., 1998), whereas one revealed mixed evidence (Comstock &
Helsing, 1976). Four studies used composite measures of SES; of
these, three identified an inverse association (Ickovics et al., 1997;
Steele, 1978; Warheit, Holzer, & Arey, 1975), and one showed
mixed evidence (Lynch, Kaplan, & Salonen, 1997). One study
found an inverse association between occupational prestige and
depressive and hopeless symptoms (Lynch, Kaplan, & Salonen,
1997). Thus, overall, 64% of the examined associations suggested
an inverse relationship between SES and depressive symptoms,
whereas 36% showed an inverse association for some groups or
measures and nonsignificant associations for others. None of the
studies identified completely null findings. In studies that identi-
fied an inverse relationship, the association tended to be linear, so
that at each successive decrease in SES, depressive and hopeless
symptoms increased. The obvious limitation of all these studies is
their cross-sectional design, which impedes determination of
Seven of the reviewed studies examined SES and the prevalence
of depressive disorders, as shown in Table 1. One of three that
included assessments of educationKessler et al. (1994), using
the NCSidentified an inverse association, and two found no
association: Bebbington, Hurry, Tennant, Sturt, and Wing (1981)
and, using the ECA, Weissman, Bruce, Leaf, Florio, and Holzer
(1991). The NCS also found an inverse association between in-
come and prevalence of affective disorder (Kessler et al., 1994),
whereas the ECA findings regarding income were again null
(Weissman et al., 1991). Likewise, the ECA identified no associ-
ation between occupation and depression prevalence (Weissman et
al., 1991). Four out of six studies identified an inverse association
between other indicators of SES and the prevalence of depressive
disorders (Bebbington et al., 1981; Murphy et al., 1991; Weissman
et al., 1991; Wilson, Chen, Taylor, McCracken, & Copeland,
1999), and two showed mixed evidence (Regier et al., 1993;
Weissman & Myers, 1978). Thus, 50% of the studies identified an
(text continues on page 22)
Table 1
Research Addressing the Relationship Between Socioeconomic Status (SES) and Depression and/or Hopelessness
Study Population/source Design SES measure Depression/hopelessness
measure Controls Evidence for inverse
association Strength of
Studies examining the cross-sectional association between SES and depressive or hopeless symptoms
Comstock & Helsing
(1976) 3,845 residents of
Kansas City,
MO, and
County, MD,
18 yrs old
Cross-sectional Income, education Depressive symptoms,
CESD16 Sex, age, marital status,
employment status,
income, or education
Income: strength and
pattern of association
varies by ethnicity and
Education: strength
and pattern of association
varies by ethnicity and
Craig & Van Natta
(1979) 1,614 community
and 45
psychiatric in-
Cross-sectional Education CESD, prevalence
(symptoms in prior
week), persistence
(symptoms 57
days in prior week)
Sex, age, marital status,
clinical status
For prevalence, inverse
association in 2 of 16
symptoms; for
persistence, inverse
association in 10 of 16
4/Clinical sample not
selected randomly,
symptoms examined
Fiscella & Franks
(1997) 6,582 U.S.
residents, 25
74 yrs old/
1216-yr f/u
for health
General Well Being
Scale, high versus
low depressive
symptoms, hopeless
affect in past
month; high
hopeless outlook in
past week
Adjustment for
oversampling and
Rates of all
outcomes 1.62.0 times
higher for low income
2/Distress measured over
brief time period and
with minimal items; no
adjustment for sex, age,
Ickovics et al.
(1997) 2,145 male post-
MI patients,
2969 yrs old
Insurance Plan
substudy of
12-mo f/u for
Education and
Three-item measure
of depression, low
versus high
None Inverse, linear
relationship 2/Brief depression measure;
no controls for possible
confounds; sample not
selected randomly (post-
MI patients)
Lynch, Krause, et al.
(1997) 2,674 Finnish
Ischemic Heart
Disease Study
Cross-sectional Aggregate index
of recalled
childhood SES
early adulthood
Occupation (adult
Depression: shortened
Hopelessness: two
items, study
Age Childhood SES: poor-
and middle-SES groups
more likely to have high
hopelessness, not
Education: gradient,
inverse for hopelessness,
Occupation: blue collar
workers (compared with
white collar) with more
high hopelessness,
5/Specific population
(Finnish men); data
cross-sectional, but
capitalized on unique
approach to capturing
SES at different life
(table continues)
Study Population/source Design SES measure Depression/hopelessness
measure Controls Evidence for inverse
association Strength of
Studies examining the cross-sectional association between SES and depressive or hopeless symptoms (continued)
Salokangas &
Poutanen (1998) 1,643 patients of
health centers,
Finland, 1864
yrs old
Cross-sectional Education, income Depressive symptoms,
Depression Project
Multivariate analysis
with age, self and
spouse health,
alcohol, housing
problems, other
psychosocial factors
Education: inverse,
gradient association;
nonsignificant in
multivariate analysis,
most likely because of
strong correlation with
physical health
4/Specific population
(primary care patients);
study-specific depression
Income: low earnings
predicted higher
depressive symptoms
Steele (1978) 134 upper- and
class residents,
New Haven,
Cross-sectional Education and
Zung Depression
Scale and several
lesser known
None Lowest class had higher
scores compared with
higher classes
2/Participants not selected
randomly and small
sample; low social
classes excluded;
ethnicity confounded
with social class; no
Warheit et al. (1975) 1,645 Florida
residents Cross-sectional Income,
18-item measure of
Multivariate analysis
with ethnicity, age,
sex, and SES
Inverse, linear
association 4/Study-specific measure
used, but good
psychometrics reported
West et al. (1998) 2,025 affluent
Marin County,
CA, residents,
55 yrs old
Cross-sectional Household
Depressive symptoms,
CESD16 Age; in multivariate
analysis: health
behaviors, disability,
social support
Income: inverse
association up to income
$74,999; association
nonsignificant in
multivariate analysis, due
to relationship between
income and other
psychosocial factors
Education: inverse
association in women
only; marginal in
multivariate analysis
Studies examining the cross-sectional association between SES and depressive disorders
Bebbington et al.
(1981) 800 (Stage 1),
310 (Stage 2)
Cross-sectional Education, social
1-mo prevalence of
affective disorders,
Present State
Examination (no
symptoms, minimal
symptoms, below
threshold, threshold
disorder, definite
Sex-specific analyses,
and no overall age
differences were
Social class: significant
association between total
symptoms and social
class, when both
measured continuously
Table 1 (continued)
Study Population/source Design SES measure Depression/hopelessness
measure Controls Evidence for inverse
association Strength of
Studies examining the cross-sectional association between SES and depressive disorders (continued)
Kessler et al. (1994) 8,098 U.S.
residents, 15
54 yrs old/NCS
Cross-sectional Income, education 12-mo and lifetime
prevalence of
affective disorders
(major depression,
mania, dysthymia),
accounted for 91%
of 12-mo and 89%
of lifetime preva-
Rates weighted for
deviation from
Income: for lifetime and
12-mo prevalence, lowest
differed from highest
income group (ORs
1.56 and 1.73,
Education: for lifetime
and 12-mo prevalence,
all lower education
groups differed from
highest group (ORs
1.86, 1.76, and 1.44,
and 1.79, 1.38, and 1.37,
5/Lifetime diagnoses rely
on memory and
accuracy; affective
disorders aggregated
(includes mania, bipolar);
rates for specific
disorders not reported
Regier et al. (1993) 18,571 U.S.
residents/ECA Cross-sectional Nam Index,
composite of
income, own
education, and
divided into
Major depression,
dysthymia, DIS for
Age, sex, ethnicity,
marital status
Linear inverse
association between SES
and OR for major
depression, until highest
SES when trend began to
reverse; lowest differed
significantly from next to
highest (OR 2.62);
nonsignificant trend
toward higher ORs for
dysthymia in lower SES
Weissman et al.
(1991) 19,182 U.S.
residents/ECA Cross-sectional Occupation,
past 5 yrs),
dependence on
1-yr prevalence of
major depression,
via DIS
Sex, age, ethnicity Occupation
Unemployment (OR
3.25) and financial
dependence (OR 2.56)
4/Crude distinctions in
socioeconomic groupings
(SES measures
Weissman & Myers
(1978) 511 treated and
(report on
89-yr f/u of
Social class Point prevalence,
lifetime prevalence
of major and minor
depression, SADS
None For point prevalence,
minor depression, not
major depression, was
more common in lower
class groups. No
consistent relationship for
lifetime prevalence, and
rates tended to be higher
in upper classes
4/Many participants were
lost from original cohort;
differential attrition (e.g.,
mortality) could have
affected results; no
controls for possible
(table continues)
Table 1 (continued)
Study Population/source Design SES measure Depression/hopelessness
measure Controls Evidence for inverse
association Strength of
Studies examining the prospective association between SES and depressive disorders or symptoms
Anthony & Petronis
(1991) 5,969 U.S.
residents, 1844
yrs old/ECA
1-yr f/u Employment, job
Major depression, DIS
for DSMIII Analyzed in multiple
regression with
ethnicity, marital
status, education,
employment, job
prestige; also
stratified by age,
census tract
Job prestige (inverse
Employment: working
for pay associated with
decreased incidence (RR
6/Brief f/u period
Bruce et al. (1991) 3,497 New
Haven, CT,
residents, 18
yrs old/ECA
6-mo f/u Poverty Incident major
depression, DIS for
Sex, age, ethnicity,
psychiatric history
Inverse association, with
OR 2.29 for poverty
group; about 10% of
cases attributable to
5/Possible under reporting
of past episodes (may
overrepresent true
incidence); brief f/u
Coryell et al. (1992) 965 relatives
(727), controls
(150), and
spouses (88) of
persons with
disorders who
had no history
of mental dis-
orative Study
6-yr f/u Education,
occupation SADS Multiple logistic
regression analysis
with age, sex, marital
status, place of
residence and
education or
Education: inverse linear
association for men,
except the least educated
had the lowest rates;
positive association for
womenthe most
educated had the highest
Occupation: No
consistent relationship
4/Sample nonrandomly
selected, and 75% were
first-degree relatives of
individuals with affective
disorders; completers
were younger, better
educated, and more often
female and single than
J. J. Gallo et al.
(1993) 7,737 ECA
40 yrs old
1-yr f/u Education,
unemployment DIS for DSMIII
occurrence of major
Sex, marital status,
minority status,
neighborhood factors,
employment or
Education: adults with
12 yrs of education had
higher risk (RR 2.20)
compared with those with
12 yrs of education. F/u
analyses showed an
interaction with sex
association significant for
women only
5/3,049 individuals lost to
f/u (death, refusal,
inability to attend the
interview, or
whereabouts unknown);
differential attrition may
have impacted the
results; brief f/u period
incidence not greater in
Kaplan et al. (1987) 6,928 Alameda
County, CA,
9-yr f/u Education, income Validated 18-item
measure of
symptoms, incident
Multivariate analyses
with age, sex,
education, income,
ethnicity, baseline
physical health,
perceived health
Income: linear, inverse
association; inadequate
income group differed
from very/adequate
income group (RR
1.46); nonsignificant
in multivariate analysis
Education: linear, inverse
association (RRs 1.86
and 1.53 for low and
medium vs. high
6/No interim depression
Table 1 (continued)
Study Population/source Design SES measure Depression/hopelessness
measure Controls Evidence for inverse
association Strength of
Studies examining the prospective association between SES and depressive disorders or symptoms (continued)
Murphy et al. (1991) 593 rural
16-yr f/u Material
(low, average,
or high)
Presence of a
depressive disorder
(inferred through
responses to a
computer program)
Rates age and sex
For prevalence at
baseline and f/u, inverse
association for men and
For incidence across f/u,
gradient inverse
association for men and
5/Study-specific assessment
approach; no interim
assessment of depression
Wilson et al. (1999) 5,222, 65 yrs
old, Liverpool,
cohort, 2-yr
Townsend Index
(higher scores
mean more
according to
census by
postal codes)
Geriatric Mental State
(prevalent and
incident depression
Age, sex At baseline, depressive
cases (OR 5.57) and
subcases (OR 5.74)
had higher Townsend
Index scores compared
with well participants
(OR 5.05)
Significant trend for
higher odds of incident
depression with
increasing Townsend
Index; however, most
affluent group had
highest odds of incident
6/District SES does not
equate to individual
SEScannot directly
link Townsend Index
score and psychiatric
Note. MO Missouri; MD Maryland; yr year; CESDCenter for Epidemiologic Studies Depression Scale; ⫾⫽mixed findings (i.e., significant inverse association for some groups but
not all); U.S. United States; NHANES National Health and Nutrition Examination Study; f/u follow-up; ⫹⫽evidence for an inverse association at p.05; MI myocardial infarction;
BHAT Beta Blocker Heart Attack Trial; mo month; MMPI Minnesota Multiphasic Personality Inventory; CT Connecticut; CA California; ⫺⫽no significant inverse association; NCS
National Comorbidity Study; CIDI Composite International Diagnostic Interview; DSMIIIRDiagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric
Association, 1987); OR odds ratio; ECA Epidemiologic Catchment Area Study; DIS Diagnostic Interview Schedule; SADS Schedule for Affective Disorders and Schizophrenia; DSMIII
Diagnostic and Statistical Manual of Mental Disorders (3rd ed.; American Psychiatric Association, 1980); RR relative risk.
Table 1 (continued)
inverse association between SES and prevalence of depressive
disorders, whereas 17% showed mixed evidence and 33% showed
null evidence. It is notable that three of the four null associations
derived from the ECA (see Weissman et al., 1991) and involved
dichotomized assessments of education, income, and occupation.
As shown in Table 1, seven studies examined the association
between SES and depression using prospective methods. Inasmuch
as these studies demonstrate that low SES precedes the develop-
ment of depressive symptoms or disorders, they provide support
for the social causation hypothesis. Kaplan, Roberts, Camacho,
and Coyne (1987) found a prospective relationship between SES
and depressive symptoms in nearly 7,000 Alameda County, Cali-
fornia, residents. For individuals who were not depressed at base-
line, low and medium levels of education were associated with a
greater risk of high depression after a 9-year follow-up period
compared with rates associated with high levels of education.
Inadequate income also predicted risk of high depression symp-
toms at follow-up. In a prospective epidemiologic study based in
Stirling County (a pseudonym for the research site), Atlantic
Canada, Murphy et al. (1991) used a computer algorithm to diag-
nose depressive disorders in 593 men and women who were then
followed over 16 years. A composite assessment of material pos-
sessions served as the indicator of SES. Persons with low SES had
a higher incidence of depression across follow-up. A trend for
depression to predict downward social drift also emerged, but
given the close association between depression and SES at base-
line, the power to evaluate this effect was quite low. The level of
social deprivation in participantsresidential district also predicted
incident major depression in a study of residents in Camberwell,
England; however, the most affluent level had the highest rate of
new depression cases (Wilson et al., 1999). Finally, Coryell, En-
dicott, and Keller (1992) reported findings from the Collaborative
Study, in which the SADS was used to evaluate incident major
depression in spouses, relatives, and controls (age- and sex-
matched to relatives) of individuals with affective disorders. Ed-
ucation was inversely associated with incident depression in men
but positively associated with depression in women. Occupation
did not predict depression. It is important to note that the sampling
strategy used in this study, as well as the differential attrition
across gender and age, creates interpretive ambiguities.
The ECA also included a prospective component, which exam-
ined sociodemographic predictors of psychiatric disorder. In a
1-year follow-up of younger individuals (1844 years old) en-
rolled in the ECA (Anthony & Petronis, 1991), unemployment and
job prestige at the first assessment predicted new cases of major
depression at follow-up, whereas education did not. Conversely,
J. J. Gallo, Royall, and Anthony (1993) examined the older par-
ticipants (40 years old) from the ECA across the same follow-up
period and found that lower education predicted higher risk of
first-time major depression after controlling for other factors (al-
though this was statistically significant for women only), whereas
unemployment at baseline did not. These contrasting findings
suggest that distinct socioeconomic indicators could be of impor-
tance at different developmental stages or for different cohorts. In
a third study from the ECA, a prospective analysis of the New
Haven, Connecticut, sample showed that individuals reporting
poverty-level income and no history of depression at baseline had
higher rates of incident major depression across a 6-month
follow-up period (Bruce, Takeuchi, & Leaf, 1991). In summary,
four of seven (57%) prospective studies found evidence of an
inverse association between various indicators of SES and incident
depression. The remaining three studies (43%)two from the
ECAidentified mixed findings.
In the studies that examined SES and depression using cross-
sectional methods, 6 of 14 received a rating of 5 (out of a possible
6) according to our scheme for evaluating the strength of the
evidence. Among the three methodologically rigorous studies that
examined depressive symptoms, the evidence was mixed, with one
identifying an inverse association for some subsamples but not
others (Comstock & Helsing, 1976) and the remainder identifying
mixed evidence across indicators (Lynch, Kaplan, & Salonen,
1997; West et al., 1998). The three rigorous studies that examined
depressive disorders showed positive (Kessler et al., 1994) or
mixed evidence (Bebbington et al., 1981; Regier et al., 1993).
Thus, although few studies were rated as methodologically rigor-
ous, the distribution of positive and mixed findings was roughly
similar to that observed across all studies, and none of the rigorous
studies identified null findings. All of the prospective studies were
rated as methodologically rigorous.
In summary, the majority of the evidence suggests that individ-
uals with low SES have higher levels of depressive symptoms and
depressive disorders. The evidence is strongest for a cross-
sectional association between depressive symptoms and SES and
between incident depressive disorders and SES, although half of
the reviewed studies also suggest that SES is associated with
prevalent depressive disorders. The evidence seems most consis-
tent for comparisons involving income or composite measures of
SES, as opposed to education measures, although this pattern is by
no means conclusive. Very few studies examined occupational
prestige. Several studies suggest that higher SES is associated with
decreasing rates of depressive disorders and symptoms only up to
a high affluence level, at which point the effect appears to reverse
(e.g., Regier et al., 1993; West et al., 1998; Wilson et al., 1999).
Studies demonstrating that low SES precedes the development of
depression suggest that social causation explains at least part of the
association between the two.
Anxiety and SES
As shown in Table 2, our review revealed only two studies that
examined the association between SES and anxiety symptoms.
These studies suggest an inverse, linear association for education,
income (Himmelfarb & Murrell, 1984), and composite social class
(Warheit et al., 1975) with anxiety symptoms.
Of the eight studies reviewed that examined the cross-sectional
association between SES and anxiety disorders, four derived from
the NCS and three derived from the ECA. Two of four studies that
included an assessment of educationboth from the NCSiden-
tified an inverse association with prevalent panic (Eaton, Kessler,
Wittchen, & Magee, 1994) and phobic disorders (Magee, Eaton,
Wittchen, McGonagle, & Kessler, 1996). The ECA identified
mixed evidence for an association between education and preva-
lent panic and phobic disorders (Eaton et al., 1991) and null
evidence for education and generalized anxiety disorder (Blazer,
Hughes, George, Swartz, & Boyer, 1991). Three of five studies,
each from the NCS, identified an inverse, linear association be-
tween income and prevalence of various anxiety disorders (Kessler
(text continues on page 26)
Table 2
Research Addressing the Relationship Between Socioeconomic Status (SES) and Anxiety
Study Population/source Design SES measure Anxiety measure Controls Evidence for inverse
association Strength of
Studies addressing the cross-sectional association between SES and anxiety symptoms
Himmelfarb &
Murrell (1984) 2,051 Kentucky
residents, 55
yrs old
Cross-sectional Education income Spielberger Trait
Anxiety Scale Sex-specific analyses
performed; age range
Education: inverse, linear
association for men and
Income: inverse
correlation between
income and anxiety
Warheit et al. (1975) 1,645 Florida
residents Cross-sectional Income,
occupation, and
Phobic symptoms (10
items), general
anxiety, symptoms
(12 items), anxiety
function (11 items)
Multivariate analysis
with ethnicity, age,
sex, and SES
For phobic symptoms,
inverse linear association
leveled off at highest SES
group; inverse linear
association for general
anxiety symptoms; for
anxiety function, inverse
association, but second
lowest SES group had the
highest score
4/Study-specific measures
used, but good
psychometrics reported
Studies addressing the cross-sectional association between SES and anxiety disorders
Blazer et al. (1991) 8,205 U.S.
residents/ECA Cross-sectional Income,
dependence on
1-yr prevalence of
GAD, DIS Rates weighted to site
Income: overall inverse
association, but variable
Occupation: inverse
association, but
Financial dependence:
3.8% versus 1% in
Durham, NC, 4.6% versus
1% in Los Angeles
5/Data collected in only
three ECA sites and
during second study
wave only; assessment of
GAD varied across sites
Eaton et al. (1991) 19,498 U.S.
(panic); 14,263
U.S. residents
Cross-sectional Education,
dependence on
1-yr prevalence of
panic and phobic
disorders, DIS for
Sex-specific analyses
performed; report
that age adjustment
did not affect
education results
Education: trend toward
inverse association for
men only, phobic
disorders only; ethnicity-
specific analyses show
significant trend for
Whites only
5/Relatively low reliability
assessments for panic
diagnoses from DIS
Occupation: no association
for phobic disorder; for
panic, trend toward inverse
association in men, positive
association in women
Financial dependence:
inverse association for
phobia and panic disorders,
both men and women (table continues)
Study Population/source Design SES measure Anxiety measure Controls Evidence for inverse
association Strength of
Studies addressing the cross-sectional association between SES and anxiety disorders (continued)
Eaton et al. (1994) 8,098 U.S.
residents, 15
54 yrs old/NCS
Cross-sectional Education, income Prevalence of panic
attacks, disorder,
and panic with
agoraphobia in past
month, CIDI
Rates weighted for
selection, population
Report that
adjustment had
negligible effect
Education: significant
inverse association; least
educated group 4 times
likely to have a panic
attack, 10 times likely
to have panic disorder,
7 times likely to have
panic with agoraphobia
compared with college
Income: no association
Kessler et al. (1994) 8,098 U.S.
residents, 15
54 yrs old/NCS
Cross-sectional Income, education Lifetime and 12-mo
prevalence of any
anxiety disorder,
Rates weighted for
deviation from
Income: for lifetime
prevalence, all lower
differed from highest
income group ($70,000;
ORs 2.00, 1.52, and 1.48
for lowest to second highest
income groups); similar for
12-mo prevalence (ORs
2.12, 1.56, and 1.50)
5/Lifetime diagnoses rely
on memory and accuracy
Education: inverse, linear
association for lifetime
prevalence (ORs 1.86,
1.76, and 1.44 for lowest
to second highest educated
groups); similar for 12-mo
prevalence (ORs 2.82,
2.10, and 1.60)
Magee et al. (1996) 8,098 U.S.
residents, 1554
yrs old/NCS
Cross-sectional Education, income 30-day prevalence of
agorophobia, social
phobia, simple
phobia, CIDI
Rates weighted for
deviation from
Education: inverse linear
association for
agoraphobia, simple
phobia, and social phobia
Income: inverse linear
association for
agoraphobia, simple
phobia, and social phobia
Regier et al. (1993) 18,571 U.S.
residents/ECA Cross-sectional Nam Index,
composite of
income, own
education, and
divided into
1-mo prevalence of
phobia, panic, DIS
Age, sex, ethnicity,
marital status
For phobia, inverse linear
relationship between SES
and OR of 1-mo phobia
(lowest 2.43; second
lowest 1.84); SES
quartiles had higher OR
compared with highest
quartile; linear inverse
relationship for panic but
only lowest quartile
differed significantly from
highest (OR 11.58)
5/Relatively low reliability
assessments for panic
diagnoses from DIS
Table 2 (continued)
Study Population/source Design SES measure Anxiety measure Controls Evidence for inverse
association Strength of
Studies addressing the cross-sectional association between SES and anxiety disorders (continued)
Wittchen et al.
(1994) 8,098 U.S.
residents, 1554
yrs old/NCS
Cross-sectional Income Lifetime prevalence
of GAD, CIDI Multivariate analysis
controlling for sex,
age, ethnicity, marital
status, employment,
and region
Inverse linear association;
lowest differs from highest
income group; effect
nonsignificant in
multivariate analysis
5/Lifetime diagnoses relied
on memory, accuracy
Studies examining the prospective association between SES and anxiety disorders
Bruce et al. (1991) 3,497 New
Haven, CT
residents, 18
yrs old/ECA
6-mo f/u Poverty Incident panic, phobia
DIS Sex, age, ethnicity,
psychiatric history
Nonsignificant inverse
association for both,
with 1.7% and 6.7% of
cases attributable to
poverty for panic and
phobia, respectively
5/Relatively low reliability
assessments for panic
diagnoses from DIS
Eaton & Keyl (1990) 11,756 U.S.
residents/ECA Prospective,
1-yr f/u Education,
occupation 1-yr incidence of
agoraphobia, DIS
for DSMIII, split
among situational
cases, classic cases
(fear of going out
alone), and cases
with panic
Age, gender, ethnicity,
marital status, living
situation, education,
or occupation
Education: tendency for
fewer education (19 yrs)
to predict increased
incidence of agoraphobic
cases, but this was signif-
icant for classicagora-
phobia only (OR 1.77)
Occupation: tendency for
higher occupational rank
to be protective, but again,
this was significant for
classicagoraphobia only
(OR 0.86)
6/Education was
dichotomized; brief f/u
period; possible
underreporting of past
episodes could have
inflated incidence rates
Keyl & Eaton (1990) 12,823 U.S.
residents/ECA Prospective,
1-yr f/u Occupation Incident panic
disorder, severe
unexplained panic
attacks, other panic
attacks, DIS for
Age, ethnicity, marital
Increasing occupational
prestige was associated
with lower odds of
incident panic disorder
(OR 0.80), but the
effects for other types of
panic were nonsignificant
6/Brief f/u period;
relatively low reliability
assessments for panic
diagnoses from DIS
Murphy et al. (1991) 593 rural
16-yr f/u Material
(low, average,
Presence of an
anxiety disorder
Rates were
standardized by age
and sex
No association for
prevalence at baseline; a
marginal association for
prevalence at f/u; no
association for incidence
at f/u
assessment; few incident
J. C. Wells et al.
(1994) 9,437 U.S.
residents/ECA Case-control,
1-yr f/u Education 1-yr incidence of
social phobia, DIS
Gender, marital status,
age (living alone,
rural vs. nonrural
residence did not
predict social phobia)
Inverse relationship. RR
for 08 yrs of education
(vs. 13)2.14. RR for
912 yrs of education (vs.
6/Brief f/u period; site
differences were
observed, the meaning of
which are unclear
Note. yr year; ⫹⫽evidence for an inverse association at p.05; U.S. United States; ECA Epidemiologic Catchment Area Study; GAD generalized anxiety disorder; DIS Diagnostic
Interview Schedule; ⫾⫽mixed findings (i.e., significant inverse association for some groups but not all); ⫺⫽no significant inverse association; NC North Carolina; DSMIII Diagnostic and
Statistical Manual of Mental Disorders (3rd ed.; American Psychiatric Association, 1980); NCS National Comorbidity Study; CIDI Composite International Diagnostic Interview; mo month;
OR odds ratio; CT Connecticut; f/u follow-up; RR relative risk.
Table 2 (continued)
et al., 1994; Magee et al., 1996; Wittchen, Zhao, Kessler, & Eaton,
1994); studies from the ECA identified mixed findings (Blazer et
al., 1991) and null findings for the income and anxiety disorder
association (Eaton et al., 1994). The two studies from the ECA also
identified null associations between occupation and prevalent anx-
iety disorders (Blazer et al., 1991; Eaton et al., 1994). The ECA did
find a positive association between financial dependence on the
government and 1-year prevalence rates of generalized anxiety
disorders (Blazer et al., 1991) and phobic and panic disorders
(Eaton et al., 1991) and an association between a composite
assessment of SES (i.e., the Nam index) and 1-month prevalence
of panic and phobic disorders (Regier et al., 1993). Murphy and
colleagues (1991) found no association between material posses-
sions and prevalent anxiety disorders at a baseline assessment, and
only a marginal inverse association at a follow-up assessment.
Thus, the majority of the comparisons (53.3%) identified inverse
associations between indicators of SES and prevalent anxiety
disorders; 13.3% identified mixed findings, and 33.3% identified
null findings.
As shown in Table 2, five studies examined whether SES
predicts incident anxiety disorders, four of which represent find-
ings from the ECA. In a 6-month longitudinal study of the New
Haven ECA sample, poverty (i.e., very low income) did not
significantly predict incident panic or phobic disorders (Bruce et
al., 1991). However, few new cases were observed in this brief
follow-up period, suggesting that low power might have contrib-
uted to the null results (Bruce et al., 1991). Eaton and Keyl (1990)
examined the associations between occupation and education and
1-year incidence of agoraphobia in the ECA. Lower SES tended to
predict a higher incidence of agoraphobia, but this was statistically
significant only for classicagoraphobia (i.e., the fear of going
out alone), which is the most debilitating type (Eaton & Keyl,
1990). Keyl and Eaton (1990) also found a prospective inverse
association between occupational prestige and 1-year incidence of
panic disorder but not other types of panic attacks. J. C. Wells,
Tien, Garrison, and Eaton (1994) found that lower education was
associated with higher incidence rates of social phobia over a
1-year follow-up of the ECA participants. Finally, Murphy and
colleagues (1991) incorporated an assessment of anxiety disorders
in their prospective study of SES and psychiatric status and found
that SES did not show a clearly patterned association with incident
anxiety disorders. In this longitudinal study, neither the social drift
nor the social causation hypothesis received substantial support.
Thus, of the six comparisons within, 17% (i.e., one) supported an
inverse association between SES and incident anxiety disorders,
50% showed mixed findings, and 33% showed null findings. All
but one of the mixed or null findings derived from studies asso-
ciated with the ECA.
In summary, although based on only two studies, our review of
this topic does suggest that lower levels of SES are associated with
higher levels of anxiety symptoms. More than half of the reviewed
comparisons also suggested an inverse association between SES
and prevalent anxiety disorders. In contrast, only one of six com-
parisons suggested that lower SES leads to a higher incidence of
anxiety disorders, with half of the comparisons providing mixed
evidence. Thus, the extent to which a social causation versus a
social drift interpretation explains the association between anxiety
and SES is unclear. Overall, the mixed findings concerning SES
and anxiety disorder are not easily explained in terms of type of
socioeconomic indicator or type of anxiety disorder. All but one of
the studies concerning SES and anxiety received a strength of
evidence rating of 5 or 6.
Hostility and SES
As shown in Table 3, all of the studies that examined the
association between SES and hostility used cross-sectional ap-
proaches. Four studies examined the association between SES and
scores on the Ho. The studies were consistent in identifying a
linear, inverse relationship between education (Barefoot et al.,
1991; Lynch, Kaplan, & Salonen, 1997; Scherwitz, Perkins,
Chesney, & Hughes, 1991), income (Barefoot et al., 1991; Lynch,
Kaplan, & Salonen, 1997), and occupation (Barefoot et al., 1991)
and scores on the Ho full and individual subscales. In addition,
Carmelli, Rosenman, and Swan (1988) found an inverse associa-
tion between scores on the Hollingshead (1975) measure of SES
and on the Ho, and Lynch, Kaplan, and Salonen (1997) showed
that recalled childhood SES related to adult Ho scores. As noted
above, the Ho is primarily a measure of the cognitive aspects of the
hostility construct. Thus, this research suggests that individuals
with lower SES tend to maintain attitudes of cynical distrust about
others. In addition, the studies by Barefoot et al. (1991) and
Scherwitz et al. (1991) showed that these inverse associations
extended to subscales from the Ho that assess behavioral and
affective components of hostility.
Four studies using measures of hostility besides the Ho also
showed an inverse relationship with SES. Ranchor, Bouma, and
Sanderman (1996) found inverse associations between education
and occupation and scores on a multidimensional measure of anger
and hostility. Matthews, Kelsey, Meilahn, Kuller, and Wing (1989)
showed that women with lower levels of education had higher
scores on measures of trait anger and the tendency to suppress
but not expressanger. Finally, Mittleman, Maclure, Nachnani,
Sherwood, and Muller (1997) found that less-educated individuals
were more likely to report experiencing an episode of anger prior
to MI compared with their more highly educated counterparts.
In summary, 11 of 12, or 92%, of the comparisons concerning
SES and hostility identified inverse and generally linear relation-
ships, and the remaining study identified an inverse relationship
for two of three anger scales. The strength of the evidence was
mixed, with only three studies receiving a methodological rating
of 5. However, given the consistency of the findings, the associ-
ation is likely to be reliable. Thus, SES appears to be associated
with the cognitive, affective, and behavioral correlates of hostility.
Of note, two studies suggested that these associations might be
stronger in ethnic minorities (Barefoot et al., 1991; Scherwitz et
al., 1991). Because all of the reviewed studies concerning SES and
hostility used cross-sectional methods, the direction of this asso-
ciation cannot be determined. However, some research suggests
that the association between SES and hostility might begin in
childhood (Gump et al., 1999; Lynch, Kaplan, & Salonen, 1997).
Longitudinal research is needed to further explore the temporal
association between SES and hostility.
The reviewed literature suggests that an association between
SES and negative emotions and attitudes is likely. The evidence is
particularly strong for an association between low SES and de-
Table 3
Research Addressing the Relationship Between Socioeconomic Status (SES) and Hostility
Study Population/source Design SES measure Hostility measure Controls Evidence for inverse
association Strength of
Studies examining the cross-sectional association between SES and hostility as assessed by the CookMedley Hostility Inventory (Ho)
Barefoot et al.
(1991) 2,536 U.S.
residents, 18
90 yrs old
Cross-sectional Income,
Ho full scale and
subscales Multivariate analysis
with age, sex,
ethnicity, income,
occupation, marital
status, and education
Income: inverse linear
association for full scale,
Cynicism, Hostile
Attribution, Aggressive
Responding subscales,
but weaker for Social
Avoidance, Other;
interaction between
ethnicity and income for
full scale: association for
non-Whites only
Occupation: inverse
linear association, but
significant for full scale
Education: inverse, linear
association for full scale,
Cynicism, Hostile
Attribution, Aggressive
Responding, but weaker
for Social Avoidance,
Carmelli et al.
(1988) 37 mono- and 60
White male
twin pairs, CA
Cross-sectional Hollingshead Ho full scale,
Cynicism and
Paranoid Alienation
Age Significant inverse
association between
Hollingshead ranking and
full scale as well as
subscale scores
3/Small sample, not
selected randomly
Lynch, Krause, et al.
(1997) 2,674 Finnish
Ischemic Heart
Disease Study
Cross-sectional Aggregate index
of recalled
childhood SES,
early adulthood
(adult SES)
Abbreviated Ho Age Childhood SES: poor-
and middle-SES groups
have more high hostility
Education: Gradient,
inverse association
between education and
hostility; least educated
(compared with most
educated) more likely to
have high hostility
Occupation: blue collar
workers (compared with
white collar) more likely
to have high hostility
5/Specific population
(Finnish men); data
cross-sectional, but
capitalized on unique
approach to capturing
SES at different life
(table continues)
Table 3 (continued)
Study Population/source Design SES measure Hostility measure Controls Evidence for inverse
association Strength of
Studies examining the cross-sectional association between SES and hostility as assessed by the CookMedley Hostility Inventory (Ho; continued)
Scherwitz et al.
(1991) 5,115 residents of
AL, Chicago,
MN, Oakland,
CA, 1830 yrs
Cross-sectional Education (high
school or less,
more than high
Ho full scale and sub-
scales Multivariate analysis
with ethnicity,
gender, age,
education, all
interaction effects
Inverse association for
full scale and all
subscales except Social
Avoidance (i.e.,
Cynicism, Hostile
Attributions, Hostile
Affect, Aggressive
Responding, Other);
associations stronger for
non-Whites than for
4/Education dichotomized
Studies examining the cross-sectional association between SES and scores on other measures of hostility
Matthews et al.
(1989) 541 women,
Pittsburgh, PA,
4250 yrs old/
Women Study
Cross-sectional Education Spielberger Trait
Anger Scale,
Anger-Out and
Anger-In subscales
None Inverse association with
linear trend for
Spielberger Trait Anger
and Anger-In; no
association for Anger-Out
4/Sample was healthier and
better educated than
nonparticipants; no
controls for possible
Mittleman et al.
(1997) 1,623 post-MI
patients Case crossover Education Onset Anger Scale
(one item), Anger-
In 2 hr preceding
Separate analyses of
groups differing in
aspirin use,
diabetes, overweight
Inverse, gradient
association between
education and anger
triggering in all analyses
2/Retrospective report of
anger; single-item
assessment; specific
population, not selected
randomly (post-MI
Ranchor et al.
(1996) 6,989 men, the
3070 yrs old
Cross-sectional Education,
occupation BussDurkee
Resentment, and
Suspicion Scales
from factor analysis
Age, social desirability Education: inverse, linear
association for all three
Occupation: inverse,
linear association for all
three subscales
5/Specific population
(Dutch men)
Note. U.S. United States; yr year; ⫹⫽evidence for an inverse association at p.05; CA California; AL Alabama; IL Illinois; MN Minnesota; CARDIA coronary artery risk
development in young adults; PA Pennsylvania; ⫾⫽mixed finding (i.e., significant inverse association for some groups but not all); MI myocardial infarction.
pressive, hopeless, anxious, and hostile symptoms. Some evidence
also suggests that SES exhibits a cross-sectional association with
depressive and anxiety disorders, but these findings are less con-
sistent. Likewise, some research suggests that low SES precedes
the development of depressive symptoms and disorders and, to a
lesser extent, anxiety disorders. These studies suggest the role of
social causation in connecting SES with negative cognitive
emotional factors. Our review points to a number of areas for
further research, including (a) additional studies examining the
association between SES and anxiety symptoms, (b) further re-
search concerning the prospective association between SES and
cognitive and emotional symptoms, and (c) continued research into
the socioeconomic correlates and consequences of emotional
It is interesting that the NCS provided substantially stronger
evidence of an association between SES and anxiety disorders
compared with the ECA. A similar difference, though perhaps not
as marked, emerged in relation to depressive disorders. In brief,
differences between the studies include the fact that the NCS
examined a younger age distribution, presented all screening items
at the outset of the interview to avoid participantstendency to
minimize pathology to shorten the interview, and included adjust-
ments for nonparticipantshigher rates of disorder (Eaton et al.,
1994). In addition, the NCS was based on DSMIIIR(American
Psychiatric Association, 1987) criteria, whereas the ECA was
based on DSMIII (American Psychiatric Association, 1980) cri-
teria. The NCS also included more thorough probing, for example,
in relation to phobias and in respect to lifetime recall, and more
stem questions, for example, in relation to major depression
(Kessler et al., 1994). Overall, the NCS identified higher preva-
lence rates for many disorders than the ECA (e.g., Kessler et al.,
1994). Hence, increased power is probably one factor explaining
the stronger associations observed in the NCS. Further research is
needed to determine which study has provided the more accurate
assessment of links between SES and prevalent and incident anx-
iety and depressive disorders.
Negative Emotions and Cognitions and Health
If negative emotions and cognitions are important in under-
standing the SEShealth link, research should show that these
factors predict verifiable health outcomes. In the following sec-
tions, we discuss this evidence, focusing on cardiovascular dis-
eases and all-cause mortality. To demonstrate that these factors
predict these outcomes, we draw on the work of other researchers
who have provided recent, detailed reviews concerning the health
correlates of depression (e.g., Glassman & Shapiro, 1998; Wulsin,
Vaillant, & Wells, 1999), anxiety (e.g., Fleet & Beitman, 1998;
Hayward, 1995; Kubzansky, Kawachi, Weiss, & Sparrow, 1998;
Rozanski, Blumenthal, & Kaplan, 1999), and anger and hostility
(e.g., Miller, Smith, Turner, Guijarro, & Hallet, 1996; Rozanski et
al., 1999). In addition, we discuss prospective studies that have
been performed since the publication of these reviews.
Depression and Hopelessness and Health
A number of reviews have summarized the evidence linking
depression and hopelessness with all-cause mortality and with
cardiovascular morbidity and mortality. It is not surprising that
these reviews overlap considerably. Rozanski et al. (1999) evalu-
ated the prospective evidence assessing the effects of depression
and hopelessness on cardiovascular outcomes. They limited their
focus to studies that included a hardendpoint, such as MI or
cardiac death. The studies varied according to whether they as-
sessed depressive or hopeless symptoms or clinical depression
(i.e., major depression, dysthymia). Depression or hopelessness
predicted increased risk of future coronary events in six of eight
studies of community populations (Anda et al., 1993; Aromaa et
al., 1994; Barefoot & Schroll, 1996; Everson et al., 1996; Ford et
al., 1998; Pratt et al., 1996). Relative risks for depressed versus
nondepressed individuals ranged from 1.5 (for depressive symp-
toms; Anda et al., 1993) to 4.5 (for major depression; Pratt et al.,
1996). Mixed findings were reported for a study by Wasserthal-
Smoller and colleagues (1996), who found that depressive symp-
toms did not relate to cardiovascular outcomes during the rela-
tively brief 4.5-year follow-up period but that increases in
depressive symptoms predicted cardiovascular events. Negative
findings were identified by Vogt, Pope, Mullooly, and Hollis
(1994), who examined a study-specific measure of depression in
relation to all-cause mortality. All reviewed studies that examined
effects of depression on outcomes in coronary heart disease (CHD)
patient populations found positive evidence for an association
(Ahern et al., 1990; Barefoot et al., 1996; Carney, Rich, Freedland,
& Sanai, 1988; Denollet & Brutsaert, 1998; Frasure-Smith, Les-
pe´rance, Juneau, Talajic, & Bourassa, 1999; Frasure-Smith, Les-
pe´rance, & Talajic, 1995a; Hermann et al., 1998; G. J. Kennedy et
al., 1987). Rozanski and colleagues concluded that overall, there is
support for an association between major depression as well as
depressive symptoms and future coronary events. They noted that
a number of studies suggest a doseresponse relationship between
hopelessness or depression and level of risk (e.g., Anda et al.,
1993; Barefoot & Schroll, 1996; Everson et al., 1996; Pratt et al.,
Musselman, Evans, and Nemeroff (1998) provided a detailed
review of studies published between 1966 and 1997, which eval-
uated the prospective association of depressive symptoms, includ-
ing hopeless cognition, or major depression, with cardiovascular
outcomes. These studies generally concerned community popula-
tions, although one study examined hypertensive patients (Simon-
sick, Wallace, Blazer, & Berkman, 1995). Major depression or
depressive symptoms predicted incident cardiovascular disease or
cardiac death in 10 of 13 studies. The consistency of these findings
is notable given the rigorous methodological restrictions imposed
by Musselman et al. Specifically, the review included only those
studies that controlled for other cardiovascular risk factors (e.g.,
hypertension, hypercholesterolemia, smoking history) and socio-
demographic characteristics (e.g., age, sex, SES). The authors
concluded that depression is an independent risk factor in the
pathophysiologic progression of CVD [cardiovascular disease],
rather than merely a secondary emotional response to the illness
(Musselman et al., 1998, p. 581).
In another review, Glassman and Shapiro (1998) reported that 9
of 10 studies comparing patients with major depression with
general population samples found increased risk of cardiovascular
mortality in depressed patients. Two community studies that
lacked controls for smoking also found increased mortality in
depressed individuals, and six of seven community studies that
controlled for smoking and other cardiovascular risk factors found
increased cardiovascular morbidity and/or mortality in depressed
individuals. Finally, six studies of cardiac patients found excess
risk of death for depressed individuals, after controlling for other
cardiovascular risk factors and medical and social variables. Glass-
man and Shapiro noted that the risk of cardiovascular mortality
was about 1.52.0 times higher for depressed than for nonde-
pressed healthyindividuals and about double that for depressed
versus nondepressed patients recovering from MI.
Finally, a detailed and comprehensive review by Wulsin et al.
(1999) examined the association between depression and all-cause
mortality in 57 studies published between 1966 and 1996. The
authors rated each study in terms of methodological rigor, accord-
ing to the following criteria: (a) sample size, (b) measure of
depression (with a structured interview given the highest weight
and a self-report measure the lowest weight), (c) chosen compar-
ison group (with matched control given the highest weight, cohort
the next highest, and a general population control group the lowest
weight), and (d) factors controlled for, with the highest rating
given to studies that accounted for age, sex, and at least two of the
following: physical illness, smoking, alcohol, and suicide. Twenty-
nine of the studies (51%) provided consistent positive evidence, 13
(23%) provided consistent negative evidence, and 15 (26%) pro-
vided mixed evidence for an association between depression and
mortality. It is important to note, however, that only 37% of the
examined studies met the authorscriteria for bettermethodol-
ogy. The relative percentage of positive (48%), negative (29%),
and mixed findings (23%) was similar among these studies, though
slightly less supportive of an independent effect of depression.
Wulsin et al. also examined specific causes of death, including
cardiovascular causes. They reported that 60% (15 of 25) of
studies examining depression and cardiovascular mortality identi-
fied positive evidence for an association between them, with the
remainder split between negative and mixed findings. Wulsin et al.
concluded that suicide seemed to explain only a small amount of
the association between depression and mortality and that mixed
findings were not well explained, either in terms of an identifiable
pattern or in terms of hypotheses put forth by the reviewed studies
authors. Their overall conclusions were that depression has a
substantial effect on mortality in certain populations and that it
appears to be linked most closely with cardiovascular deaths. They
recommended future research that follows a large sample using
longitudinal design while controlling for important confounds.
Since the publication of these reviews, several studies have
examined the association between depression and mortality or
cardiovascular disease using a prospective design. Jonas and Mus-
solino (2000) examined stroke incidence in a probability sample
of 6,095 Black and White U.S. residents. After adjustment for
diverse potential confounds, individuals with high depressive
symptoms had a higher risk (between 50% and 160%) of stroke
over the next 22 years compared with individuals with low de-
pressive symptoms. This association was present in all gender and
ethnicity groups and was linear in nature. A study of 573 older
hospitalized patients (Covinsky et al., 1999) found that patients
who had more depressive symptoms when admitted were more
likely to die during a 3-year follow-up than were patients with
fewer symptoms. This excess risk was attenuated but remained
statistically significant after controlling for potential confounding
factors, including sociodemographics, indicators of illness sever-
ity, and functional status. Again, the association between depres-
sion and risk of death was linear in nature. Another study (Les-
pe´rance, Frasure-Smith, Juneau, & The´roux, 2000) found that
patients hospitalized with unstable angina and who had high BDI
scores were more than four times as likely to experience a subse-
quent morbid or mortal coronary event compared with nonde-
pressed patients, even after controlling for other identified predic-
tors (e.g., left ventricular ejection fraction, number of diseased
vessels, electrocardiographic evidence of ischemia). Finally, Pen-
ninx et al. (2001) found an increased risk of cardiac mortality in
initially healthy, community-dwelling, depressed older persons
and in those with diagnosed cardiac disease in a 50-month pro-
spective study. Individuals with minor depression (as diagnosed
according to CESD score cutoffs) showed moderate levels of risk,
and the associations persisted after control for diverse potential
confounds. Contradictory data were presented in a recently pub-
lished 12-month prospective study of a small sample of MI pa-
tients that found that depressive symptoms (and anxiety symp-
toms) failed to predict mortality (Lane, Carroll, Ring, Beevers, &
Lip, 2001).
In conclusion, although mixed findings have appeared in the
literature, the weight of the evidence supports the conclusion that
depression and hopelessness predict negative health outcomes. The
support for an association between depression and hopelessness
and cardiovascular morbidity and mortality seems especially com-
pelling, with more conflicting findings emerging in the literature
assessing mortality from all causes. A number of more recent
studies that used rigorous methodologies to examine depression or
hopelessness and outcomes in relatively large community or pa-
tient populations provide especially convincing evidence that these
constructs are important in the etiology and course of cardiovas-
cular disease (e.g., Anda et al., 1993; Barefoot & Schroll, 1996;
Everson et al., 1996; Everson, Roberts, Goldberg, & Kaplan, 1998;
Frasure-Smith, et al., 1993, 1995a, 1999; Jonas & Mussolino,
2000; Penninx et al., 2001; Pratt et al., 1996). Nevertheless, as
noted by Wulsin and colleagues (1999), mixed findings and a
plethora of less methodologically sound research suggest the need
for additional longitudinal research.
Anxiety and Health
Several recent reviews summarize the evidence concerning the
health implications of anxiety. Rozanski and colleagues (1999)
examined nine studies investigating the association between anx-
iety and cardiovascular outcomes. Three studies of community
populations showed a positive association between phobic anxiety
symptoms at baseline and risk of cardiac death during follow-up
(Haines, Imeson, & Meade, 1987; Kawachi, Sparrow, Vokonas, &
Weiss, 1994; Kawachi, Colditz, et al., 1994), even after control for
a variety of potential confounds. Level of anxiety appears to exert
a graded, or doseresponse effect on mortality risk (Kawachi,
Colditz, et al., 1994; Kawachi, Sparrow, et al., 1994). It is notable
that these studies support an association between anxiety and
sudden coronary death rather than nonfatal coronary heart disease,
suggesting the possible role of fatal ventricular arrhythmia trig-
gered by acute anxiety (e.g., Amsterdam, 1990; Lown, 1982).
Another community study (Kubzansky et al., 1997) found that
level of worry, a symptom of generalized anxiety disorder, pre-
dicted incident MI and marginally predicted incident CHD. Again,
results suggest a doseresponse association. Finally, Rozanksi et
al. reviewed the findings of a community study (Weissman,
Markowitz, Ouellette, Greenwald, & Kahn, 1990) that found in-
creased risk of MI for people with panic disorder compared with
those without panic disorder. Rozanski et al.s review also showed
that in three recent studies of patients recovering from MI, anxiety
symptoms predicted physical outcomes (Denollet & Brutsaert,
1998; Frasure-Smith, Lespe´rance, & Talajic, 1995b; Moser &
Dracup, 1996). In these studies, patients with high levels of anxiety
had between 2.5 and 4.9 times the likelihood of negative outcomes
including unstable angina, re-infarction, and cardiac death after
controlling for other risk factors as compared with patients with
lower anxiety. Finally, a study of medical in-patients (Hermann et
al., 1998) showed that those with high levels of anxiety symptoms
who were admitted with cardiopulmonary diagnoses were signif-
icantly more likely to die during follow-up than their less anxious
counterparts. Rozanski et al. suggested that large-scale epidemio-
logical studies of coronary artery disease patients should be at-
tempted and that studies of women in particular are needed.
Fleet and Beitman (1998) provided a critical review and analysis
of the six studies that, before 1997, examined the association
between panic disorder or panic-like anxiety symptoms and death
from cardiovascular causes. Three of these studies used retrospec-
tive methodologies. In the first, male but not female in-patients
with probable panic disorder (according to chart review) had
increased risk of death from circulatory diseaseacross follow-up
compared with age- and gender-matched population mortality
rates (Coryell, Noyes, & Clancy, 1982). Similar results were
identified by the same group of researchers in a retrospective study
of out-patients with probable panic disorder (Coryell, Noyes, &
House, 1986). As discussed by Fleet and Beitman, the method-
ological limitations of these studies include the following: (a)
current diagnostic criteria for panic may not have been met; (b)
circulatory disease was not defined, and the number of deaths by
circulatory causes was quite small; (c) the contribution of medical
and psychiatric disorders that developed across follow-up was not
assessed; and (d) the contribution of other coronary risk factors,
such as smoking, was not assessed. In the third retrospective study
of a random sample of more than 18,000 adults (Weissman et al.,
1990), panic disorder was assessed by structured interview, assur-
ing adherence to DSMIII criteria. Patients with panic were more
likely to have experienced hypertension, heart attack, or stroke,
according to their own self-reports, after adjustment for a number
of demographic variables. The use of self-reported outcomes is an
important limitation of this study. In addition, the interview as-
sessed lifetime diagnoses, making it impossible to identify the
direction of any association between psychiatric problems and
health outcomes.
Fleet and Beitman (1998) also reviewed three prospective stud-
ies (Haines et al., 1987; Kawachi, Colditz, et al., 1994; Kawachi,
Sparrow, et al., 1994) that were likewise described by Rozanski
and colleagues (1999). Consistent with this review, Fleet and
Beitman concluded that the research suggests a doseresponse
relationship between symptoms of anxiety and future risk of sud-
den coronary death. Furthermore, they pointed out that associa-
tions persisted even after controlling for a wide range of potential
confounds (e.g., smoking, blood pressure, family history of CHD).
However, they also noted the possible limited clinical implications
of the findings in light of the fact that very few sudden cardiac
deaths occurred in any of the studies. In addition, the studies
assessed symptoms of anxiety, which do not necessarily constitute
a panic disorder or any other anxiety disorder.
Finally, Kubzansky et al. (1998) provided a comprehensive
review of studies published between 1980 and 1996 as well as
especially relevant earlier reports of the association between anx-
iety and CHD. Many of the studies reviewed, including the three
major prospective community studies, have already been discussed
in this article. The authors also described the findings of Eaker,
Pinsky, and Castellis (1992) follow-up from the Framingham
Study of factors predicting incident CHD (MI, cardiac death) in
women across 20 years; anxiety and tension symptoms were
associated with excess risk of incident CHD in homemakers, but
not in employed women, after controlling for a wide range of
additional risk factors. Kubzansky et al. (1998) also summarized
three earlier prospective studies that found no relationship between
anxiety and CHD (Allgulander & Lavori, 1991; Martin, Cloninger,
Guze, & Clayton, 1985; Wheeler, White, Reed, & Cohen, 1950).
All three of these examined psychiatric patient populations. How-
ever, the review authors delineated numerous methodological lim-
itations of these studies. For example, standardized mortality rates
were examined rather than deaths in appropriate control samples.
One of the studies had a relatively high number of deaths due to
suicide (Allgulander & Lavori, 1991), and another study had no
CHD cases in patients with panic disorder (Martin et al., 1985). In
concluding their review, Kubzansky et al. (1998) noted that anx-
iety is a potentially important risk factor for CHD and that further,
interdisciplinary research is needed. In particular, the authors
called for research including women and studies examining the
effects of nonpathological levels of anxiety. In addition, they
suggested research that allows the effects of interrelated negative
emotions (i.e., depression, anxiety) to be distinguished.
To our knowledge, no prior reviews have aggregated the find-
ings assessing the association between anxiety and all-cause mor-
tality. Our examination of the literature suggested only limited
available evidence. Murphy, Monson, Olivier, Sobol, and Leighton
(1987) found no association between presence of an anxiety dis-
order at baseline and later mortality in either men or women. Two
studies (Kubzansky et al., 1997; Vogt et al., 1994) reported no
association between baseline levels of worry and mortality during
follow-up. In contrast, in their analysis of the association between
psychiatric status and 9-year mortality rates from the New Haven
ECA, Bruce, Leaf, Rozal, Florio, and Hoff (1994) found that
individuals with a history of panic disorder were approximately
three times more likely to die during follow-up, after adjustment
for age and sex, than individuals without a history of panic
disorder. A history of phobia did not predict mortality risk. De-
nollet and Brutsaert (1998) reported that Type D personality (i.e.,
high trait anxiety and social inhibition) predicted an increased risk
of death from noncardiac as well as cardiac mortality in patients
with established CHD, even after control for severity of disease.
In summary, several well-designed, prospective studies of com-
munity populations suggest that higher levels of anxiety symptoms
confer excess risk for negative cardiac outcomes, and in particular,
sudden coronary death. The evidence for an association between
anxiety and mortality from all causes is less compelling, as is the
research concerning anxiety disorders and health outcomes. Addi-
tional research is needed to confirm whether the health effects of
anxiety extend to women.
Hostility and Health
Like the research concerning depression and anxiety, the most
recent review of hostility and health was performed by Rozanski
and colleagues (1999). These authors discussed 11 studies that
examined the association between hostility and cardiovascular
outcomes or all-cause mortality in healthy samples and noted that
the findings from this research are quite mixed. For example, of
eight studies that administered the Ho to healthy samples, four
(Hearn, Murray, & Luepker, 1989; Leon, Finn, Murray, & Bailey,
1988; Maruta et al., 1993; McCranie, Watkins, Brandsma, &
Sisson, 1986) found no evidence for an association with cardio-
vascular health outcomes or all-cause mortality, whereas the other
four found that Ho scores predicted cardiovascular outcomes
(Barefoot, Larsen, von der Lieth, & Schroll, 1995), all-cause
mortality (Barefoot et al., 1989), or both (Barefoot et al., 1983;
Shekelle, Gale, Ostfeld, & Paul, 1983). Among the remaining three
studies that used measures other than the Ho scale, one identified
a marginal association between hostility and all-cause mortality
(Koskenvuo et al., 1988), one found that cynical distrust predicted
MI and cardiac death, but only before other risk factors were
statistically controlled (Everson et al., 1997), and the final study
found a positive and graded association between anger and likeli-
hood of experiencing any coronary event during follow-up, even
after control for potential confounds (Kawachi et al., 1996). The
authors summarized four additional studies showing that higher
levels of hostility predicted more negative outcomes in patients
with known coronary artery disease (Dembroski et al., 1985;
Hecker, Chesney, Black, & Frautchi, 1988; Koskenvuo et al.,
1998; Mendes de Leon, Kop, de Swart, Bar, & Appels, 1996) and
two showing that hostility (Julkunen, Salonen, Kaplan, Chesney,
& Salonen, 1994) and anger (Matsumoto et al., 1993) predicted
atherosclerotic progression. In discussing the varied findings,
Rozanski and colleagues pointed out that the negative studies are
of disparate quality. For example, in some cases high rates of
attrition (Maruta et al., 1993) and brief follow-up periods (Kos-
kenvuo et al., 1988) could have obscured underlying effects.
Perhaps the most comprehensive review of the literature ad-
dressing the association between hostility and physical health was
performed by Miller et al. (1996). Forty-five studies were exam-
ined by these authors and subjected to both qualitative and quan-
titative (i.e., meta-analytic) analysis. Unique features of this re-
view include efforts to disentangle the relative predictive utility of
different facets of hostility and to examine methodological limi-
tations contributing to contradictory findings. Twelve reviewed
studies used interview-based ratings, which are most closely re-
lated with expressive aspects of the hostility construct (i.e., verbal
or physical aggressive behaviors; Miller et al., 1996). Seven found
positive evidence for a relationship with CHD outcomes, 2 showed
null findings, and 3 identified mixed findings, such as an associ-
ation for some measures and not others (n2) or in older
participants only (n1). In a meta-analysis of studies that used
interview-based assessments of hostility to predict CHD, the
weighted mean effect size was .18. In contrast, scores on the Ho
and on other self-report measures of the experiential aspects of
hostility (i.e., hostile cognition, angry affect) were less consistently
associated with CHD outcomes. Three out of 12 studies identified
an association between Ho scores and CHD outcomes, and 9
identified null results. Fourteen out of 31 studies identified an
association between scores on other self-report measures of hos-
tility CHD outcomes; 13 identified null results, 3 identified mixed
results (e.g., positive findings for some measures only), and 1
identified contradictory results (i.e., an effect in the direction
opposite to that predicted). The weighted mean effect size for the
association across all studies including the Ho and CHD outcomes
was .07, as was the weighted mean effect size representing the
association between other self-report assessments of cognitive
experiential aspects of hostility and CHD outcomes. This repre-
sents a small but statistically significant effect. Additional analyses
showed that self-report measures of emotional or expressive hos-
tility did not consistently predict CHD. Miller and colleagues
noted that overall, adjustment for CHD risk factors had little
impact on the associations between hostility and CHD.
Fifty percent of studies reviewed by Miller et al. (1996) that
administered the Ho (three of six) found positive evidence for an
association with all-cause mortality. The studies that administered
other self-report assessments of experiential aspects of hostility
found either positive (three of five) or mixed (two of five) evi-
dence for an association with all-cause mortality. The weighted
mean effect size for all studies that examined hostility and all-
cause mortality was .16. Statistical adjustment for other risk fac-
tors had little effect on the association between hostility and
all-cause mortality. In summarizing their findings, the authors
noted that research is most clearly indicative of an association
between behavioral ratings of hostility and CHD and between
cognitiveexperiential self-report measures of hostility and all-
cause mortality. Null findings appear to have been influenced by
variability in measurements and samples and by the use of high-
risk populations. That is, if hostility is a predictor of CHD, studies
that include only persons with CHD or those at high risk for CHD
will sample, on average, a higher and more limited range of
hostility scores.
Recent studies provide additional evidence for the deleterious
health effects of anger and hostility. For example, Everson et al.
(1999) showed that high scores on a measure of expressed anger
predicted incident stroke in Finnish men. Men with a history of
cardiovascular disease who had high anger-expression scores were
more than six times as likely to experience a stroke, when com-
pared with their less angry counterparts, after adjustment for
numerous other risk factors. The tendency to suppress and control
anger was not significantly associated with stroke risk. Williams,
Nieto, Sanford, Couper, and Tyroler (2002) showed that trait anger
was associated with increased stroke incidence among younger
participants and among those with less atherogenic lipid profiles
(i.e., among individuals who were otherwise at lower risk for
stroke). Similarly, trait anger was associated with an increased risk
of CHD in the same cohort, but only among individuals who were
not hypertensive (Williams et al., 2000). Two recent case-
crossover studies (Mittleman et al., 1995; Mo¨ller et al., 1999)
showed that episodes of intense angry affect increased the risk of
suffering an acute MI in the following 12 hr. Finally, recent
studies have shown that facets of hostility prospectively predict the
severity and progression of atherosclerosis in healthy populations
(Iribarren et al., 2000; Matthews, Owens, Kuller, Sutton-Tyrrell, &
Jansen-McWilliams, 1998; Whiteman, Dreary, & Fowkes, 2000).
In summary, the weight of the evidence suggests that hostility
and related constructs pose significant health risks. Research sug-
gests an effect of hostility on health outcomes such as incident
CHD, atherosclerotic progression, stroke, and all-cause mortality.
Detailed quantitative analyses indicate that distinct facets of the
hostility construct could be important for different health outcomes
(Miller et al., 1996)an assertion that may help explain contra-
dictory findings. In addition, studies concerning high-risk popula-
tions could impede the identification of hostilitys health effects
(cf. Miller et al., 1996), as demonstrated by studies showing that
anger increases risk for cardiac events only in lower risk groups
(e.g., Williams et al., 2000, 2002). Well-designed studies that have
controlled for diverse potential confounds (e.g., Everson et al.,
1999; Kawachi et al., 1996; see also Miller et al., 1996, for a
review) provide especially compelling evidence that hostility
poses health risks.
The reviewed research provides evidence that negative emotions
and attitudes predict health outcomes. The evidence is most com-
pelling for the effects of depression, hopelessness, and hostility on
cardiovascular morbidity and mortality and for anxiety on sudden
cardiac death. These associations have been observed in initially
healthy populations and in individuals with diagnosed cardiovas-
cular diseases. Some research also suggests that negative
cognitiveemotional constructs may be important predictors of
all-cause mortality. The evidence is particularly strong for aspects
of hostility and all-cause mortality, although hopelessness and
depression have also been shown to predict mortality in some
research. The cited reviews suggest a number of methodological
issues that need to be confronted before any definitive conclusions
can be drawn, and they offer some possible explanations for
inconsistencies observed in previous research.
It is important to note that a number of studies suggest a
doseresponse relationship between cognitiveemotional symp-
toms and health risks. Given that an individual diagnosed with a
psychiatric disorder would tend to have more severe symptoms,
one might infer that psychiatric disorders would be particularly
strong predictors of morbidity and mortality. Moreover, although
the existence of a doseresponse relationship implies that even
relatively small elevations of cognitiveemotional symptoms may
be associated with worse health outcomes, methodological limita-
tions of previous research limit definitive conclusions. For exam-
ple, as noted in a recent editorial commentary (Carney, Freedland,
& Jaffe, 2001), many studies include only a single, baseline
assessment of depression. Because subclinical depressive symp-
toms are strongly predictive of future major depression (Judd &
Akiskal, 2000), it is unclear whether negative health outcomes
develop only in those individuals who eventually suffer a full-
blown major depressive episode. Thus, further research that in-
cludes assessments of cognitiveemotional symptoms and disor-
ders (e.g., Frasure-Smith et al., 1993, 1995a; Rovner et al., 1991),
preferably at multiple time points, is clearly warranted. What we
can safely conclude from the reviewed patterns of findings is that
SES is consistently related to cognitive and emotional symptoms,
elevations of which do predict worse health outcomes.
How Do Low-SES Environments Influence Negative
Emotions and Cognitions?
If cognitiveemotional factors play a role in connecting SES
with health, a key question becomes how the environments that
people with low SES inhabit lead them to experience negative
emotions and cognitions, which, in turn, engender early morbidity
and mortality. To address this question, we offer a general model
outlining plausible psychosocial pathways among SES, cognitive
emotional factors, and health. The model and associated literature
review are intended to serve two functions. First, if available
evidence supports the posited pathways, the model would reinforce
the primary mediational hypothesis, providing a potential frame-
work for understanding the roles of cognitiveemotional factors in
the SES and health association. Second, the model is intended to
guide future research by more clearly identifying the circum-
stances under which low-SES environments may impact health
through cognitiveemotional factors. The latter focus is developed
in the final portion of the article.
In the current section, we begin our presentation of the model
with consideration of the leading psychosocial candidate for con-
necting SES with negative emotions and cognitions: exposure and
psychological reactivity to stress. Other pathways, such as envi-
ronmental toxins or a lack of mental health services, are also likely
to be important, but they are not considered here. We then artic-
ulate how low levels of tangible, interpersonal, and intrapersonal
resources, that is, a deficient reserve capacity, may exacerbate the
impact of SES-associated stress on negative emotions and atti-
tudes, recognizing that low SES can both deplete resources and
thwart the development and restoration of the resource bank. A
challenge of any model of SES and cognitiveemotional factors is
to consider the potential for bidirectional relationships; our model
explicitly posits that emotions and cognitions can influence the
availability of resources and that they can alter interpretations of
stressful circumstances. However, the model does not specifically
address the connection from health outcomes, or intermediate
paths, back to emotions and cognitions. Clearly, health endpoints
such as CHD can affect the intermediate psychosocial factors
included in the model, as well as socioeconomic standing (e.g.,
Dew, 1998). However, CHD and premature mortality represent
distal health outcomes that (typically) occur late in the natural
history of disease, well after the usual establishment of socioeco-
nomic position. In the final step of the model, we review the
behavioral and physiological pathways connecting emotions and
cognitions with health.
It is important to note that we do not intend to assert that
psychosocial aspects of low-SES environments affect health ex-
clusively through cognitiveemotional paths. Indeed, there is rea-
sonable evidence that low-SES environments directly impact be-
havioral and biological mechanisms that in turn affect health (see
Taylor et al., 1997, for further discussion). Similarly, Link and
Phelan (1995, 1996) have pointed out that across time and chang-
ing patterns of disease, individuals with low SES are less able to
avoid risks for health problems because they maintain less money,
power, and information with which to adopt newly identified
health-protective lifestyles and prevention strategies. However, in
line with the primary focus of this review, our model centers on
cognitiveemotional pathways.
SES, Exposure, and Reactivity to Stress
From a psychosocial perspective, the frequency and intensity of
exposure to harmful or potentially threatening situations and to
rewarding or potentially beneficial situations is the primary step
linking SES with negative emotions and attitudes (see Arrow A of
Figure 1). Indeed, a number of studies indicate that individuals
with low SES encounter more frequent negative life events and
chronic stressors (B. S. Dohrenwend, 1973; Langer & Michael,
1963; McLeod & Kessler, 1990; Murrell & Norris, 1991; Myers,
Lindenthal, & Pepper, 1974; Stansfeld, Head, & Marmot, 1998)
and interpret ambiguous social events more negatively (Chen &
Matthews, 2001; Flory, Matthews, & Owens, 1998). For example,
Matthews, Ra¨ikko¨nen, et al. (2000) documented that, throughout 3
days of experiential sampling, men and women employed in po-
sitions with low occupational prestige experienced more interper-
sonal conflict relative to those with higher prestige positions.
Exposure to chronic and acute stressors, in turn, has a direct
negative impact on emotional experiences (Figure 1, Arrow B;
e.g., Alec, 1996; Ensel & Lin, 1991; Paykel, 1994; Stansfeld,
North, White, & Marmot, 1995) and a direct association with
pathways affecting health outcomes (Figure 1, Arrow C; e.g.,
McEwan & Stellar, 1993). However, studies that account for initial
differences in stress exposure suggest that at every level of stress,
individuals with low SES report more emotional distress than
those with higher SES (G. W. Brown & Harris, 1978; Kessler &
Cleary, 1980; McLeod & Kessler, 1990; Turner & Noh, 1983).
Why might individuals residing in low-SES environments be
more reactive to stress? Our framework suggests that low-SES
individuals maintain a smaller bank of resourcestangible, inter-
personal, and intrapersonalto deal with stressful events com-
pared with their higher SES counterparts (Figure 1, Arrow D).
Resources tend to occur in aggregate or to be absent in aggregate
(e.g., Rini, Dunkel-Schetter, Wadhwa, & Sandman, 1999; Turner,
Lloyd, & Roszell, 1999), suggesting the existence of a general
protective influence, or resource bank (Hobfoll, 1998, 2001). Bor-
rowing a concept from the aging literature, we label this reserve
capacity. Low-SES personsreserve capacity to deal with stressful
environments may be low for two reasons: (a) Low-SES individ-
uals are exposed to more situations in which they must use their
resources and (b) their environments prevent the development and
replenishment of resources to be kept in reserve.
Consistent with our model, having few tangible, interpersonal,
or intrapersonal resources exacerbates the effects of stressful
events on outcomes such as depression (e.g., G. W. Brown &
Bifulco, 1990; G. W. Brown & Harris, 1978; G. W. Brown &
Moran, 1997; Cohen & Wills, 1985; Hobfoll, 1988, 1989; Holahan
& Moos, 1987, 1991). Furthermore, once an individual has been
exposed to stress, resources tend to deteriorate, leaving the indi-
vidual even more vulnerable to future strains (e.g., Bolger, Foster,
Vinokur, & Ng, 1996; Ensel & Lin, 1991; path not shown in Figure
1). In fact, degradations in resources may help explain the effects
of negative events on subsequent depressive symptoms (Holahan,
Moos, Holahan, & Cronkite, 1999, 2000). Hence, having fewer
stress-dampening resources, which are further reduced by stress
Figure 1. The reserve capacity model for the dynamic associations among environments of low socioeconomic
status (SES), stressful experiences, psychosocial resources, emotion and cognition, and biological and behavioral
pathways predicting cardiovascular disease and all-cause mortality over time. Dashed lines depict possible
influences. Arrow A depicts the direct influence of SES on exposure to stressful experiences. Arrow B indicates
the direct impact of stressful experiences on emotion and cognition. Arrow C shows the effects of stress on
intermediate pathways hypothesized to affect health outcomes. Arrow D shows that socioeconomic environ-
ments condition and shape the bank of resources (i.e., the reserve capacity) available to manage stress. Arrow
E (dashed line) shows that the reserve capacity represents a potential moderator of the association between stress
and cognitiveemotional factors. Arrows F and G indicate the direct impact of cognitiveemotional factors and
reserve capacity resources, respectively, on intermediate pathways. Arrows H, I, and J (dashed lines) depict the
possible reverse influence of cognitiveemotional factors on reserve capacity resources, the experience of stress,
and SES, respectively. Arrows K and L show that emotion and cognition are hypothesized to affect health
outcomes through a variety of interrelated behavioral and physiological pathways. HPA hypothalamic
pituitaryadrenocortical axis; SAM sympathetic adrenalmedullary axis.
exposures, individuals with low SES are likely to show increased
responsiveness when faced with stress (Figure 1, Arrow E).
SES and Reserve Capacity
What is the evidence that SES is associated with low reserve
capacity (i.e., Figure 1, Arrow D)? Consistent with resource-based
definitions, individuals with low SES have access to fewer finan-
cial and material goods, which might otherwise offset tangible
stressors such as job loss, illness, or disability (Thoits, 1995).
Low-SES environments also contain deficient community re-
sources, such as safe neighborhoods, parks, transportation, and
child care (Macintyre, Maciver, & Solomon, 1993; Sooman &
Macintyre, 1995; Troutt, 1993). If present, these resources might
reduce chronic or daily stress.
SES is also associated with diverse aspects of social functioning,
including contact with others, network size, reciprocity in relation-
ships, satisfaction with support, the tendency to seek social sup-
port, work support, and generalized support perceptions (e.g.,
Belle, 1982, 1990; Bosma, Van de Mheen, & Mackenbach, 1999;
Cohen et al., 1999; House et al., 1994; Krause, 1991; Krause &
Borawski-Clark, 1995; Matthews et al., 1989; Murrell & Norris,
1991; Oakley & Rajan, 1991; Ranchor, Bouma, & Sanderman,
1996; Stansfeld et al., 1998; Turner & Marino, 1994). Further-
more, neighborhood as well as individual SES appears to influence
social experiences (e.g., Gracia, Garcia, & Musitu, 1995; Tigges,
Browne, & Green, 1998). Social stressors typical of low-SES
environments (e.g., crowding, violence, high crime rates) may
interfere with the development of supportive contacts by discour-
aging interpersonal trust (Krause, 1991; Roschelle, 1997; C. E.
Ross & Jang, 2000). In addition, individuals with low SES may be
vulnerable to factors that could degrade social support and increase
social stress, such as marital instability (Tzeng & Mare, 1995),
domestic violence (Aldarondo & Sugarman, 1996; Christmas,
Wodarski, & Smokowski, 1996; Lockhart, 1987; Straus, 1980),
substance abuse (Kessler et al., 1994), and single parenting (Bi-
anchi, 1995). Contradictory findings have been reported (e.g.,
Sokolovsky & Cohen, 1981; Stack, 1974), possibly reflecting the
multidimensional nature of social support (e.g., Krause &
Borawski-Clark, 1995), ethnic and cultural differences in the as-
sociation between SES and social functioning (e.g., Stack, 1974),
or methodological limitations such as inadequate sample sizes and
time frames. However, overall, the weight of the evidence suggests
that low-SES environments shape interpersonal experiences in a
negative manner, although the effect sizes may be small.
Further research indicates an inverse relationship between SES
and resilient intrapersonal characteristics such as self-efficacy,
mastery, or a sense of perceived control (Cohen et al., 1999;
Lachman & Weaver, 1998; Marmot, Ryff, Bumpass, Shipley, &
Marks, 1997; Matthews et al., 1989; Mirowsky, Ross, & Willigen,
1996; Ranchor et al., 1996). A recent meta-analysis suggests that
individuals with low SES have lower levels of self-esteem, with
the effects stronger for occupation and education than for income
(Twenge & Campbell, 2002). The association between SES and
intrapersonal resources might originate through asymmetrical re-
lationships associated with status hierarchies, which bring unpleas-
ant feelings of inferiority for those low in the hierarchy and
pleasant feelings of superiority for those high in the hierarchy (R.
Brown, 1965). These interpersonal experiences are likely to begin
in early development (Kessler & Cleary, 1980; Repetti, Taylor, &
Seeman, 2002) and may be maintained through socialization ex-
periences related to low-SES work environments later in life (e.g.,
Gecas & Seff, 1989). For example, downward social mobility in
employment status is related to lower levels of personal control
and mastery (Pearlin, Lieberman, Menaghan, & Mullan, 1981),
and individuals with low SES tend to report low control, decision-
making latitude, skill discretion, and work support and high de-
mands in their jobs (e.g., Marmot & Theorell, 1988; Stansfeld et
al., 1998). Like the research regarding social support, that con-
cerning SES and intrapersonal resources has been somewhat in-
consistent, perhaps in part because of gender, cultural, and ethnic
differences in the nature of the association (e.g., Richman, Clark,
& Brown, 1985). For example, the association between SES and
self-esteem has become stronger in women over time, is higher
among Asian Americans than other groups, and increases with age,
peaking in midlife (Twenge & Campbell, 2002).
Evidence Supporting the Moderating Roles of Reserve
Capacity Resources
Consistent with the reserve capacity model (Figure 1, Arrow E),
studies showing that social support (e.g., Turner & Noh, 1983;
Stansfeld et al., 1998; C. E. Ross & Mirowsky, 1989), negative
work characteristics (Stansfeld et al., 1998), and intrapersonal
resources (e.g., Lachman & Weaver, 1998; Mirowsky et al., 1996;
C. E. Ross & Mirowsky, 1989; Turner, Lloyd, & Roszell, 1999)
help explain low-SES individualsexcess vulnerability to distress
provide direct support for the moderating roles of reserve capacity
resources (although see contradictory evidence in Murrell & Nor-
ris, 1991; Thoits, 1982, 1984). For example, Turner and Noh
(1983) found that individuals with low SES but high social support
and perceived control did not display elevated emotional distress
relative to those with high social status. Similarly, Lachman and
Weaver (1998) found that the positive effects of control on health
outcomes (i.e., perceived health, physical symptoms, depression,
life satisfaction) were greater at lower levels of education or
income; individuals with low SES and strong control beliefs had
health outcomes similar to their higher SES counterparts. Thus,
negative emotions and cognitions may be more likely to serve as
mediators of the SEShealth association in the context of low
reserve capacity resources.
Evidence Supporting the Mediating Roles of Reserve
Capacity Resources
Although less germane to our primary focus, it is important to
note that reserve capacity resources may also serve mediating roles
in the associations between SES and emotional and physical
health. For example, prior research indicates an association be-
tween reserve capacity resources and cognitiveemotional states,
as shown in Arrow F of Figure 1. Specifically, research has shown
that low self-esteem predisposes individuals to depression (Hokan-
son, Rubert, Welker, Hollander, & Hedeen, 1989), although con-
tradictory findings have been reported (Lewinsohn, Steinmetz,
Larson, & Franklin, 1981; Robertson & Simons, 1989). Social
supportand in particular social integrationalso has a direct,
positive association with mental health (Cohen & Wills, 1985;
Thoits, 1995). Evidence for the mediating roles of resources was
provided in a study by Link, Lennon, and Dohrenwend (1993), in
which the lack of control and planning associated with low-SES
occupations related to lower levels of personal control, which in
turn predicted depression and distress. Similarly, Ennis, Hobfoll,
and Schro¨der (2000) showed that changes in material resources led
to increased depression in inner-city women; this association was
mediated by psychosocial resources (i.e., mastery, social support),
which were inversely associated with SES.
Reserve capacity resources could also contribute directly to the
association between low SES and physical health outcomes
through intermediate pathways, as shown in Arrow G of Figure 1
(Taylor & Seeman, 1999). Substantial research suggests that the
same types of interpersonal resources deficient in low-SES envi-
ronments protect against negative health outcomes (e.g., Adler &
Matthews, 1994; Krantz & McCeney, 2002; Rozanksi et al., 1999).
Some studies have also shown that intrapersonal resources predict
physical health outcomes (Rodin, 1990; Seeman & Lewis, 1995).
Arrow G is consistent with the reviewed mediation studies, which
included resource variables such as social support and control (in
addition to cognitiveemotional factors) to examine mediating
roles of psychosocial factors in studies of SES and physical health
(e.g., Cohen et al., 1999; Levenstein & Kaplan, 1998; Lynch et al.,
1996). Further evidence is provided by a recent study (Marmot,
1998) showing that work characteristics (e.g., skill discretion,
decision authority) and perceived efficacy and control partially
mediated the association between SES and perceived health and
the association between SES and waist-to-hip ratio. Thus, both
mediating roles and moderating roles of psychosocial resources
should be considered in future research.
Bidirectionality in the Associations Between Emotions and
Cognitions and SES
Our framework considers that distress may negatively impact
available resources, which would result in the perpetuation and
intensification of negative cognitiveemotional experiences (Ar-
row H, Figure 1). For example, research has shown that hostile
individuals experience less social support and more social conflict
compared with their nonhostile counterparts (Barefoot, Dahlstrom,
& Williams, 1983; Houston & Kelly, 1989; Scherwitz et al., 1991).
Similarly, depressed persons experience greater interpersonal dif-
ficulties and lesser social support (e.g., Brugha, 1995), and this
association appears to be reciprocal in nature (e.g., T. P. Johnson,
1991; see also discussions in Hammen, 1991; Joiner & Coyne,
1999). Likewise, depression (Beach & OLeary, 1993; S. L. John-
son & Jacob, 1997) and hostility (Miller, Marksides, Chiriboga, &
Ray, 1995; Newton & Kiecolt-Glaser, 1995; T. W. Smith, Sanders,
& Alexander, 1990) predict poorer marital functioning.
As shown by Arrow I (Figure 1), cognitiveemotional states
may also influence appraisals of external stimulia tenet that
underlies cognitive theories of depression (Beck, 1971). Specifi-
cally, depressed individuals may distort processing of information
in a manner that serves to reinforce negative mood (Haaga, Dyck,
& Ernst, 1991). Similarly, hostile individuals may be more likely
to interpret ambiguous stimuli in a negative or threatening light
(Chen & Matthews, 2001; Flory, Matthews, & Owens, 1998).
Thus, negative emotions and cognitions may increase the likeli-
hood that ambiguous stimuli are viewed as threatening or harmful,
thereby resulting in further degradations of the reserve capacity.
Negative emotional and cognitive experiences could also feed
back to SES, in accord with the social drift, or social selection,
hypotheses (Figure 1, Arrow J; e.g., Kessler, 1979). According to
this view, psychological disorder impairs ones ability to attain a
higher social class and/or causes one to drift down the socioeco-
nomic hierarchy. For example, a study by Kessler, Foster, Saun-
ders, and Stang (1995) showed that an early history of conduct
disorder, substance use, anxiety, and mood disorders predicted the
likelihood that one would fail to complete high school. However,
this study was not longitudinal, thereby creating interpretive am-
biguities (J. G. Johnson, Cohen, Dohrenwend, Link, & Brook,
1999). In a longitudinal study by Caspi, Elder, and Bem (1987),
the presence of temper tantrums in late childhood predicted later
socioeconomic attainment. Specifically, men with a history of
temper tantrums experienced downward occupational mobility,
and women with a history of tantrums tended to marry men with
lower status occupations. Another study found that individuals
with panic disorder had an increased likelihood of initiating dis-
ability payments across a 1-year follow-up compared with those
without a panic diagnosis (Kouzis & Eaton, 2000).
However, some research suggests that the social drift effect
might be more relevant to very debilitating psychiatric disorders,
such as schizophrenia (e.g., B. P. Dohrenwend, 1990; Wyatt &
Clark, 1987) and substance problems (e.g., Kessler et al., 1995),
and less so to depression and anxiety. For example, a recent study
showed that SES in childhood was a strong prospective predictor
of future depressive and anxiety disorders, but depression and
anxiety did not predict further decrements in SES (J. G. Johnson et
al., 1999). In contrast, alcohol- and substance-related diagnoses
predicted downward socioeconomic drift. A recent prospective
study of adolescents found no evidence of downward social selec-
tion, in respect to educational attainment, for participants with
depression or anxiety at baseline (Miech, Caspi, Moffitt, Wright,
& Silva, 1999). Overall, studies have supported the causation
hypothesis over the drift, or selection hypothesis, in respect to
depression (see also B. P. Dohrenwend, 2000; Moos, Cronkite, &
Moos, 1998; Ritsher, Warner, Johnson, & Dohrenwend, 2001).
Ultimately, identifying directionality in the associations be-
tween SES and emotions and cognitions represents an exceedingly
difficult goal, as evidenced by the fact that after many years of
investigation, the primary questions remain unanswered (Fox,
1990). Considerable methodological obstacles exist. For example,
in longitudinal studies, previous subclinical episodes or periodic
increases in symptoms might have predated the initial study period
and, therefore, impacted the socioeconomic standing of the partic-
ipant as measured at baseline. Even past threshold episodes may be
underestimated because of inaccurate memory or intentional mis-
representation in reporting lifetime occurrences. Furthermore, lon-
gitudinal studies may underemphasize the impact of psychological
disorder on social standing by disregarding the influence of psy-
chological distress on the failure to rise(i.e., to achieve a higher
socioeconomic standing than ones origins) or to keep up with
societal increases in affluence (e.g., B. P. Dohrenwend & Dohren-
wend, 1969). Studies that measure psychological dysfunction in
relation to SES at a single point in time also cannot integrate the
influence of previous exposures to low SES. Finally, research
concerning reciprocal influences of SES and cognitiveemotional
symptoms involves complexity in unraveling state and trait influ-
ences and in inferring datable onsets. The association between SES
and psychological functioning is sure to be complex and, most
likely, dynamic. In aggregate, we interpret the literature as show-
ing that to some extent, low SES can be considered causally
antecedent to psychological distress and that psychological dys-
function and low SES may be mutually reinforcing.
Pathways From CognitiveEmotional Factors to Health
As shown in Figure 1, Arrows K and L, negative emotions and
cognitions appear to affect health through several interrelated
behavioral and physiological pathways. These constructs influence
health behavior choices and adherence to intervention and preven-
tion regimens. For example, depressed individuals are more likely
to display negative health behaviors, such as smoking (Anda et al.,
1990; Hughes et al., 1986) and leading a sedentary lifestyle (Farm-
er et al., 1988). Other research suggests that depression leads to
poor adherence to coronary treatment recommendations (Blumen-
thal, Williams, Wallace, Williams, & Needles, 1982; Carney,
Freedland, Eisen, Rich, & Jaffe, 1995; Guiry, Conroy, Hickey, &
Mulcahy, 1987). High levels of anxiety (Breslau, Kilbey, & An-
dreski, 1991; Pohl, Yeragani, Balon, Lycaki, & McBride, 1992)
and hostility (Leiker & Hailey, 1988; Siegler, 1994) also increase
the likelihood that young people will engage in negative health
behaviors later in life. Negative cognitive and emotional processes
could affect health behavior choices through their association with
factors such as locus of control, health beliefs, or self-efficacy
(Bandura, 1989; Lau, 1988; Strickland, 1978). In addition, nega-
tive health practices and poor adherence may represent maladap-
tive coping strategies associated with emotional distress. Consis-
tent with this assertion, some research suggests that individuals
experiencing stress are more likely to engage in negative health
behaviors (Horowitz et al., 1979; Schachter, Silverstein, Kozlow-
ski, Herman, & Liebling, 1977). However, studies that have ex-
amined health practices suggest that they do not account com-
pletely for associations of negative emotions and cognitions with
health. Investigators have therefore begun to explore a number of
plausible biological pathways.
Negative emotions and attitudes appear to produce diverse phys-
iological alterations that are linked to increased risk of cardiovas-
cular morbidity and mortality, including hyperactivity of the sym-
pathetic adrenalmedullary (SAM) system, dysregulation of the
hypothalamicpituitaryadrenocortical axis, exaggerated platelet
reactivity, reduced heart rate variability, heightened inflammatory
processes, and ventricular instability and ischemia in response to
stress (see reviews by Carney, Rich, & Jaffe, 1995; Ehlert &
Straub, 1998; Glassman & Shapiro, 1998; Kop, 1999; Krantz &
McCeney, 2002; Kubzansky et al., 1998; Markovitz & Matthews,
1991; Miller et al., 1996; T. W. Smith, & Gallo, 2001; Van Kanel,
Mills, Fainman, & Dimsdale, 2001). More speculative is the pos-
sibility that negative emotions and cognitions, particularly hostility
and depression, may be associated with altered central serotonergic
function (Manuck et al., 1998), which, in turn, is associated with
SES and several cardiovascular risk factors (Matthews, Flory,
Muldoon, & Manuck, 2000; Muldoon et al., 1998). Some research
suggests that negative emotions and attitudes might also influence
cardiovascular morbidity and mortality through their association
with established cardiovascular risk factors such as hypertension
(Everson et al., 1998; Jonas, Franks, & Ingram, 1997; Markovitz,
Matthews, Kannel, Cobb, & DAgostino, 1993; Markovitz, Mat-
thews, Wing, Kuller, & Meilahn, 1991; Paterniti et al., 1999; K. B.
Wells, Golding, & Burnam, 1989) and hyperlipidemia (Bajwa,
Asnis, Sanderson, Irfan, & van Praag, 1992; Oxenkrug, Brancon-
nier, Harto-Traux, & Cole, 1983; Weidner, Sexton, McLellarn,
Connor, & Matarazzo, 1987; see also Hayward, 1995, for a re-
view) and metabolic factors such as diabetes (Eaton, Armenian,
Gallo, Pratt, & Ford, 1996), obesity, and central adiposity
(Ra¨ikko¨nen, Matthews, Kuller, Reiber, & Bunker, 1999). In addi-
tion, research showing that negative emotions and attitudes affect
immune functioningpossibly through SAM activation (Cohen &
Herbert, 1996)provides evidence that they alter vulnerability to
infectious diseases, cancer, and coronary artery disease (Appels,
Ba¨r, Ba¨r, Bruggeman, & de Baets, 2000; Christensen, Edwards,
Wiebe, Benotsch, & McKelvey, 1996; Herbert & Cohen, 1993;
Weisse, 1992). Thus, negative emotions and attitudes could shape
health and disease through a number of behavioral and biological
Summary and Implications
The presented model and supporting evidence provide a frame-
work for integrating disparate literatures with the unique goal of
evaluating a frequently posited psychosocial pathway connecting
SES and health. Understanding whether and how cognitive
emotional factors play a role in SEShealth links is of vast
importance, and the model suggests specific ways to evaluate that
role. More specifically, the model has heuristic value, leading to a
number of testable predictions.
For example, the association between SES and negative emo-
tions may begin in childhood as children develop attitudes about
others and process information about potentially threatening or
harmful environments (Chen, Matthews, & Boyce, 2002). The role
of reserve capacity should increase with age because low-SES
environments are proposed to cause increased use of available
resources and to interfere with the development of a resource bank
as one ages. Similarly, the association between SES and negative
emotions should increase in size longitudinally, given the proposed
effects of negative emotion feedback on social position. Thus,
studies that examine associations among SES, cognitive
emotional factors, and intermediate pathways or health in multiple
age groups or, ideally, with longitudinal methodologies may help
elucidate when and how these relationships emerge. Indices of
social position that are more dynamic in nature, such as income,
should be more sensitive to the bidirectional relationships between
SES and negative cognitiveemotional factors and, presumably,
health. Finally, the influence of SES on negative emotions and
cognitions may be enhanced with consideration of resources or
reserve capacity. Cognitiveemotional factors and resources may
exhibit systematic and potentially synergistic relationships. Hence,
any study that considers either in isolation of the other may
underestimate their impact on health. We further develop these
assertions and their implications for future research below.
Future Research Directions
SES is inversely associated with anxiety and depression symp-
toms and negative cognitive styles, such as hopeless and hostile
attitudes. Low SES may also be associated with higher rates of
anxiety and depressive disorders, although the evidence for these
associations is less consistent. As suggested by a number of
prospective studies (e.g., Bruce et al., 1991; Kaplan et al., 1987;
J. C. Wells et al., 1994; Wilson et al., 1999), and as depicted in
Figure 1, at least part of the discrepant vulnerability associated
with low SES likely stems from social causation. Low-SES envi-
ronments are associated with events and situations that have a
direct impact on negative emotional and cognitive states. In addi-
tion, lower SES individuals do not appear to benefit from having
as large a reserve capacity, developed over time from the aggre-
gation of positive interpersonal experiences and intrapersonal
characteristics, which might otherwise attenuate the effects of their
stressful environments. Thus, low SES is associated with greater
exposure and reactivity to stress (cf. Bolger & Zuckerman, 1995).
Previous research also suggests that negative emotions and
attitudes have deleterious effects on health, and a number of
behavioral and biological mechanisms could underlie these asso-
ciations. In aggregate, the findings suggest that experiencing
higher levels of negative emotions and cognitions represents one
possible mediator of the association between SES and health
particularly cardiovascular health and all-cause mortality. Studies
that have addressed mediation more directly provide limited evi-
dence for the roles of cognitiveemotional factors. However, these
studies are associated with important methodological limitations
that create interpretive ambiguities. Specifically, prior studies of
mediation have not considered the types of reciprocal associations
between resources and cognitiveemotional factors that are de-
picted in Figure 1 and supported by previous research. Further-
more, previous mediation studies have not typically considered
SES from a multilevel, dynamic perspective, nor have they ad-
dressed the aggregate influence of negative emotions and cogni-
tions or the possible protective effects of positive emotions and
attitudes. Prior research also has been limited by including homog-
enous samples and by underutilizing developmental and lifestyle
trajectory approaches. We now present what we consider to be the
most comprehensive methods for future studies addressing the
mediation framework. We also present suggestions regarding ini-
tial steps for preliminary tests of the proposed associations.
Comprehensive Tests of Mediation: What Types of
Research Are Needed?
To further evaluate the tenet that negative emotions and cogni-
tions contribute to the association between SES and health, addi-
tional integrative research is needed. Our review has uncovered
very few studies that could be used to directly evaluate mediation,
and studies that did include all pieces of the framework provide
very limited evidence for the dynamic links suggested in our
model. What characteristics of studies would allow a more com-
prehensive analysis of mediation in future research?
Consideration of aggregate effects among risk factors. First,
consistent with the reserve capacity model presented in Figure 1,
we recommend that future mediation studies consider joint effects
among variables. Studies that control for all psychosocial factors
simultaneously and then examine the degree to which the excess
risk of low-SES individuals is attenuated may underestimate the
true roles of psychosocial factors by neglecting additive, interac-
tive, or synergistic effects. Risk factors tend to aggregate within
lower SES individuals (e.g., Lynch, Kaplan, & Salonen, 1997), and
attention to these aggregations might contribute to psychologists
understanding of how SES affects health. For example, Kaplan
(1995) described data from the Alameda County Study (Berkman
& Breslow, 1983) in which low income, social isolation, and
depression each independently predicted mortality from all causes
over 9 years. Individuals who were isolated, depressed, and poor
had a dramatically increased risk of death, which was nearly four
times higher than the referent group (i.e., not poor, isolated, or
depressed). In the Kuopio Ischemic Heart Disease Study (Kaplan
et al., 1994; see also Kaplan & Salonen, 1990), low income, social
isolation, and cynical distrust each increased risk for premature
mortality. Again, individuals who experienced these risk factors in
combination evidenced a markedly elevated risk of death (nearly
four times as high) relative to the comparison group (Kaplan,
1995). Another recent study (Frasure-Smith et al., 2000) showed
that depression and social support interacted to predict outcomes
following MI. Individuals who had high depression but low social
support were at higher risk of a recurrent event, whereas those who
had high depression and high social support were not at increased
risk. Thus, joint effects among psychosocial risk factors may be
important to predicting health outcomes and should be considered
in tests of mediation.
It is important to note that covariation in high-risk psychological
(i.e., depression, anger, low perceived control) and social (i.e.,
social support, work environments) characteristics do not occur
randomly. Rather, these factors are likely to be reciprocally deter-
mined through recurring transactions between individuals and their
social contexts (e.g., Revenson, 1990; T. W. Smith, 1995). Well-
articulated theories based on this transactional view have been
advanced to explain the close association between depression
(Joiner & Coyne, 1999) and hostility (T. W. Smith, 1995) and
personal control (Krause, 1997) with features of the social envi-
ronment. The socioeconomic context is likely to play an important
role in shaping these recurring psychosocial patterns, as shown in
Figure 1. Thus, through the common epidemiological practice of
statistically controlling for other risk factors in an effort to identify
the independent predictive value of a single variable, information
about dynamic risk processes may be lost (see also L. C. Gallo &
Smith, 1999; Revenson, 1990; T. W. Smith & Gallo, 2001). We
suggest that a more comprehensive understanding of risk may be
gained by examining psychosocial risk factors in their naturally
occurring configurations. For example, cluster analytic techniques
could be applied to existing data sets to describe specific patterns
of person- and social-level risk factors (L. C. Gallo & Smith,
1999), and resulting groupings could then be included as predictors
of subsequent health.
Previous studies that have examined mediation have also tended
to control for other types of risk factors prior to examining the
effects of psychosocial variables. This is logical in cases in which
the goal is to determine whether psychosocial factors add to
prediction over and above traditional risk factors. However, this
approach may underestimate the true impact of psychosocial fac-
tors on the outcome. For example, in part, psychosocial factors are
believed to affect health by influencing health behaviors. Statisti-
cally controlling for health behaviors to examine the independent
effect of psychosocial factors eliminates consideration of these
indirect effects (e.g., Siegler, 1994). This assertion is supported by
a recent study (Whiteman et al., 2000) in which hostility directly
predicted atherosclerotic progression and also indirectly predicted
progression through smoking. Moreover, psychosocial factors are
believed to be importantly related to a variety of health outcomes;
therefore, covarying out prior health history could have a similar
effect. Likewise, psychosocial factors could contribute to disease
severity. For example, individuals with high levels of negative
emotions and cognitions and low levels of resources could have
more severe disease because of the cumulative exposure to these
psychosocial risk factors. Thus, additional studies should focus on
modeling both the direct and indirect effects of psychosocial risk
factors using path analysis or structural equation modeling. Alter-
natively, both psychosocial and behavioral or biological risk fac-
tors could be included in attempts to identify naturally occurring
patterns of risk through cluster analytic techniques.
Dynamic conceptualization of SES and cognitiveemotional
factors. Second, ideally future studies concerning the hypothe-
sized associations would include multilevel assessments of socio-
economic variables, as contextual SES variables appear to have
important health implications over and above the effects of indi-
vidual factors (Diez-Roux et al., 2001; Lynch et al., 2000; Robert,
1998; Yen & Kaplan, 1999; see also Diez-Roux, 1998; Macintyre
& Ellaway, 2000, for further discussion). In addition, SES should
be assessed at multiple time points because, as noted above,
previous research suggests that repeated exposure to low SES
could generate cumulative effects (e.g., Lynch, Kaplan, & Shema,
1997). Multiple assessments would also allow an evaluation of the
effects of changes in SES on psychological and physical health
(e.g., Duncan, 1996; Matthews et al., 2001; McDonough, Duncan,
Williams, & House, 1997). Although this type of dynamic evalu-
ation is best suited to a longitudinal framework, exposures and
changes in SES could also be assessed in cross-sectional research.
For example, future studies could evaluate periods of unemploy-
ment in ones lifetime or the number of times ones income has
fallen below the poverty line.
Likewise, having more than one psychological disorder or hav-
ing high levels of many types of emotional and cognitive symp-
toms may pose even greater risk than having only one disorder or
one type of symptoms. Thus, we recommend that future research
also focus on the aggregate health impact of negative emotions,
emotional disorders, and other psychiatric disorders. Further-
more, future research should consider the cumulative impact of
exposure to negative emotions, cognitions, or emotional disor-
ders across time, consistent with the recommendations for con-
ceptualizing SES. In line with our discussion of conceptualizing
emotion, we also note the importance of using well-validated
and reliable self-report and interview measures of emotion and
Positive emotions and attitudes. Future studies of mediation
should also focus explicitly on the health-protective effects of high
levels of positive emotions and cognitions, which may be more
likely to occur in high-SES environments. Some previous studies
concerning SES and negative emotions or cognitions have used
continuous assessments and identified inverse, approximately lin-
ear relationships between the variables (e.g., Barefoot et al., 1991;
Himmelfarb & Murrell, 1984; Ickovics et al., 1997; Kessler et al.,
1994; Scherwitz et al., 1991; Wilson et al., 1999; Wittchen et al.,
1994). To the extent that individuals with high SES report lesser
negative emotions and cognitions, they may be protected from
deleterious health effects. However, if affect consists of two va-
lence dimensions, as some research and theories proffer (e.g.,
Watson & Clark, 1997; Zautra, Potter, & Reich, 1997), then future
studies should also include distinct measure of positive cognition
and emotion.
Reflecting a rather ubiquitous negative research bias (cf. Selig-
man & Csikszentmihalyi, 2000), researchers have not typically
conceptualized the SES and health gradient from the perspective of
positive emotions and cognitions. However, they may protect
against health through mechanisms similar to those proposed to
link negative emotions and cognitions with health. Resilient inter-
personal and intrapersonal characteristics could increase the
chances that one will adaptively cope, thereby reducing reliance on
health-detrimental coping behaviors. In addition, at least prelimi-
nary evidence suggests that positive emotions produce beneficial
physiological effects, including short-circuiting cardiovascular re-
activity associated with negative emotions (Fredrickson & Leven-
son, 1998) and augmenting immune functioning (Stone, Cox,
Valdimarsdottir, & Jandorf, 1987; Stone, Neale, Cox, & Napoli,
1994). Additional research suggests that positive cognition about
interpersonal partners may dampen reactivity to stress (T. W.
Smith, Ruiz, & Uchino, 2001). Thus, future research should in-
clude separate measures of positive and negative emotional and
cognitive experiences (and resources) to examine whether these
constructs make independent or aggregate contributions to SES
health discrepancies.
Inclusion of diverse samples. Also important for future studies
of associations among SES, psychosocial resources, emotions and
cognitions, and physical health is the inclusion of diverse samples.
For example, many prior studies evaluating physical endpoints
have included samples of men only, and additional research is
therefore needed to examine the proposed mediational framework
in women. This seems particularly important because women show
consistently higher rates of depressive (Comstock & Helsing,
1976; Fiscella & Franks, 1997; Weissman & Myers, 1978) and
anxiety symptoms (Himmelfarb & Murrell, 1984) than men.
Women also experience major depression approximately twice as
often as men (Kessler et al., 1994; Weissman et al., 1991), and they
may also have higher rates of most anxiety disorders (Kessler et
al., 1994). In contrast, men display higher rates of hostility (Bare-
foot et al., 1991; Scherwitz et al., 1991). Thus, we recommend that
future research consider whether gender affects the nature of
relationships among SES, cognitiveemotional factors, and health.
Future research should also include ethnically diverse samples
to evaluate whether the proposed paths are influenced by ethnic or
cultural differences. Several studies have shown that individuals of
ethnic minority status may have higher rates of cognitive
emotional symptoms (e.g., Bruce et al., 1991; Comstock & Hels-
ing, 1976; Warheit et al., 1975), although some have suggested
that these differences could be attributed largely to the effects of
social class (e.g., Comstock & Helsing, 1976; Warheit et al.,
1975). On the other hand, some research indicates that members of
ethnic minorities obtain higher scores on measures of hostile,
mistrustful attitudes compared with non-Hispanic White respon-
dents (Barefoot et al., 1991; Gump et al., 1999; Scherwitz et al.,
1991). Hostile, mistrustful attitudes could therefore play a more
important mediating role for members of ethnic minority groups
than for Whites. Indeed, the study by Gump and colleagues (1999)
found that hostility represented a mediating factor connecting low
SES with cardiovascular reactivity and left ventricular hypertrophy
for Black but not White adolescents.
In addition, psychosocial resources could have disparate mean-
ings, structures, or roles for individuals of different ethnic or
cultural backgrounds. For example, some research suggests that
kin support may function differentially for Blacks and Hispanics
when compared with non-Hispanic Whites, although the pattern of
differences has varied (e.g., Hogan, Hao, & Parish, 1990; Jung,
1989; Roschelle, 1997). Furthermore, the association between
racial discrimination and SES appears to be complex and varied.
For example, Blacks with higher SES are sometimes more likely to
report discrimination relative to Blacks with lower SES (e.g.,
Gary, 1995; Landrine & Klonoff, 1996). Nonetheless, discrimina-
tory experiences deserve further attention as factors that could
increase the stressful experiences associated with low-SES envi-
ronments for non-White individuals (Krieger, 2000; Williams,
1999). Thus, we recommend that future research concerning the
mediation hypothesis incorporate diverse samples to evaluate the
roles of discrimination and to examine consistency in the hypoth-
esized pathways.
Developmental research and life-course trajectory approaches.
Given our focus on cardiovascular endpoints and all-cause mor-
tality, the reserve capacity model has been discussed in the context
of understanding the roles of psychosocial factors in adult health
gradients. However, the gradient between SES and other health
problems can be observed in samples of children aged 1 to 18
years (Goodman, 1999; Johnson et al., 1999; for reviews, see also
Bradley & Corwyn, 2002; Chen et al., 2002). This finding suggests
the importance of future longitudinal research that incorporates a
consideration of developmental factors in the processes outlined in
our model. Chen et al. (2002) pointed out that because of normal
developmental changes in cognition and affect, the transition from
childhood to adolescence is the period in which negative emotions
and cognitions are likely to begin influencing the links between
SES and health outcomes. Adolescence is an intriguing period
because for some health outcomes (e.g., blood pressure), SES links
are strong in early childhood, disappear by adolescence, and re-
emerge by adulthood (Chen et al., 2002). For other health out-
comes, such as physical activity and smoking, links with SES
develop during adolescence (Chen et al., 2002). Our model sug-
gests that increased exposure across the life span to the processes
outlined in the reserve capacity model contribute to SEShealth
links considerably prior to midlife and perhaps in adolescence and
young adulthood.
Developmentally focused approaches are also of value because
they may facilitate an understanding of why childhood SES (i.e.,
SES of family of origin) is an important predictor of adult psy-
chosocial characteristics (Bosma et al., 1999; Lynch, Kaplan, &
Salonen, 1997), behavioral and biological health risk factors
(Blane et al., 1996; Lynch, Kaplan, & Salonen, 1997; Van de
Mheen, Stronks, Looman, & Mackenbach, 1998), and morbidity
and mortality (Ben-Shlomo & Davey Smith, 1991; Davey Smith,
Hart, Blane, Gillis, & Hawthorne, 1997; Davey Smith, Hart, Blane,
& Hole, 1998; Kaplan & Salonen, 1990; Rahkonen, Lahelma, &
Huuhka, 1997) independently of adult SES. Furthermore, they may
provide another perspective about why biological factors, such as
in-uterine stress and low infant weight and height, that are corre-
lated with low SES may have a direct association with later health
(i.e., one that is not mediated by psychosocial factors). Biological
factors may represent markers for stressful psychosocial environ-
ments associated with being born into a lower SES environment,
and they may relate to health by indicating future adult socioeco-
nomic circumstances (as discussed in Kaplan & Keil, 1993; Mar-
mot, Shipley, Brunner, & Hemingway, 2001; Power & Hertzman,
In aggregate, these findings suggest that a comprehensive anal-
ysis of the paths creating socioeconomic disparities in health
would ultimately require longitudinal research that examines re-
sources, emotions, attitudes, behavioral and biological health risk
factors, and psychological and physical health outcomes as they
unfold across the life span. Life-course trajectory approaches have
become more realistic with the accessibility of multilevel model-
ing statistical procedures, which facilitate analysis of clustered
data and intrapersonal change. Even with careful longitudinal
methodologies, identifying temporal precedence and causal rela-
tionships is a difficult task (e.g., Bradley & Corwyn, 2002; Kuh &
Ben-Shlomo, 1997). Nevertheless, by comparing life-course tra-
jectories in individuals with low versus high SES, future research
can at least begin to elucidate the intricate pathways we have
presented. Thus, we suggest that additional longitudinal studies
that incorporate developmental-stage analyses be conducted.
What Intermediate Research Steps May Be Taken to
Evaluate the Mediation Hypothesis?
The proposed framework for an ideal test to the mediation
hypothesis is daunting given the complexity of addressing recip-
rocal relationships, the requirement of detailed measures taken
repeatedly over an appropriate period of time, the need to assess
symptoms, disorders, and multiple levels of SES, the analytic
challenges, and the need to take into account the general changes
in the population. Thus, we propose a number of initial research
steps that may provide a foundation for future, more comprehen-
sive research.
Analysis of existing data sets. Our review has uncovered very
few studies that examined the degree to which negative emotions
and cognitions account for SES and cardiovascular morbidity and
mortality or all-cause mortality. Yet, many large-scale databases
exist in which measures of each of these variables are included.
Thus, we recommend that formal mediation tests that include
analysis of each criterion presented by Baron and Kenny (1986) be
applied to existing data sets. We also recommend that these anal-
yses incorporate a consideration of interactive effects, especially
when measures of stress or interpersonal and intrapersonal re-
sources are included. Similarly, the procedures outlined above for
examining naturally occurring aggregations of socioeconomic,
psychosocial, and/or other types of risk factors could be applied to
existing data sets to perform a more comprehensive analysis of the
roles of emotions, cognitions, and other psychosocial factors.
Tests of mediation with intermediate health outcomes. As a
further step toward testing the proposed model, we recommend
that future research focus on the intermediate paths depicted in
Figure 1. For example, mediational tests could be applied in
studies that have examined the association between SES and
cardiovascular reactivity, an approach adopted by Gump et al.
(1999). Similarly, future research could examine whether cogni-
tive and emotional factors help explain the association between
SES and metabolic factors, health behaviors, and other risk factors
(e.g., Kubzansky, Kawachi, & Sparrow, 1999; Marmot, 1998).
Finally, not depicted in Figure 1, subclinical cardiovascular dis-
ease measures could be used as an intermediate outcome. Mea-
sures such as carotid artery disease, coronary calcification, or
endothelial dysfunction that occur without clinical signs can be
measured reliably and noninvasively, can predict subsequent clin-
ical cardiovascular events, and recently have been used to examine
psychosocial hypotheses (e.g., L. C. Gallo, Matthews, Kuller,
Sutton-Tyrrell, & Edmundowicz, 2001; Matthews et al., 1998).
These endpoints would allow a less time- and resource-intense
approach to evaluating the model, which could provide a basis for
future research with more distal health endpoints. They also have
the advantage of not being as subject to an interpretation of reverse
causality, because intermediate disease outcomes occur without
symptoms. Consistent with our recommendations for comprehen-
sive mediation approaches, we suggest that such tests consider the
roles of additive and synergistic effects among cognitive
emotional and resource variables.
Experimental research approaches. Experimental research
approaches may also facilitate tests of associations represented in
the reserve capacity model. For example, using social
psychological procedures, one could experimentally manipulate
key components of SESsuch as access to tangible resources or
prestigeand then measure the impact on mood states and inter-
mediate health outcomes (e.g., cardiovascular responses to acute
stress). The moderating effect of provision of interpersonal re-
sources could also be evaluated through experimental manipula-
tion. There may also be opportunities to piggy-back onto random-
ized interventions designed to reduce negative affect to examine
whether therapeutic techniques that improve symptoms also lead
to improvement in intermediate health outcomes when compared
with control groups. Finally, natural experiments, such as plant
closings or planned layoffs, may provide the context for examining
the impact of changing social status on negative emotions and
attitudes and health outcomes.
Available evidence supports the plausibility of the hypothesis
that the association between SES and health is mediatedat least
in partby cognitiveemotional factors. Cognitiveemotional
factors may play a particularly salient role in the context of a low
reserve capacity, and we therefore recommend further studies that
adopt a more integrative approach to examining the roles of
psychosocial factors. Cognitiveemotional factors are only one
potential influence to consider in unraveling the links between low
SES and poor health. The challenge is to identify those factors
susceptible to intervention to promote better health in the popula-
tion and reduce the substantial health variability that exists accord-
ing to SES. This review points to the importance of cognitive and
emotional factors in the quest to reduce health inequalities.
Adler, N. E., Boyce, T., Chesney, M. A., Cohen, S., Folkman, S., Kahn,
R. L., & Syme, S. L. (1994). Socioeconomic status and health: The
challenge of the gradient. American Psychologist, 49, 1524.
Adler, N. E., Boyce, T., Chesney, M. A., Folkman, S., & Syme, L. (1993).
Socioeconomic inequalities in health: No easy solution. Journal of the
American Medical Association, 269, 31403145.
Adler, N. E., Marmot, M., McEwen, B. S., & Stewart, J. (Eds.). (1999).
Annals of the New York Academy of Sciences: Vol. 896. Socioeconomic
status and health in industrialized nations. New York: New York
Academy of Sciences.
Adler, N. E., & Matthews, K. (1994). Health psychology: Why do some
people get sick and some stay well? Annual Review of Psychology, 45,
Ahern, D. K., Gorkin, L., Anderson, J. L., Tierney, C., Hallstrom, A.,
Ewart, C., et al. (1990). Biobehavioral variables and mortality or cardiac
arrest in the Cardiac Arrhythmia Pilot Study (CAPS). American Journal
of Cardiology, 66, 5962.
Aldarondo, E., & Sugarman, D. B. (1996). Risk marker analysis of the
cessation and persistence of wife assault. Journal of Consulting and
Clinical Psychology, 64, 10101019.
Alec, R. (1996). Psychosocial factors and depression. Journal of Nervous
and Mental Disease, 184, 509510.
Allgulander, C., & Lavori, P. W. (1991). Excess mortality among 3302
patients with pureanxiety neurosis. Archives of General Psychia-
try, 48, 599602.
American Heart Association. (2000). Heart and stroke statistical update.
Dallas, TX: Author.
American Psychiatric Association. (1980). Diagnostic and statistical man-
ual of mental disorders (3rd ed.). Washington, DC: Author.
American Psychiatric Association. (1987). Diagnostic and statistical man-
ual of mental disorders (3rd ed., rev.). Washington, DC: Author.
American Psychiatric Association. (1994). Diagnostic and statistical man-
ual of mental disorders (4th ed.). Washington, DC: Author.
Amsterdam, E. A. (1990). Emotions, cardiac arrhythmias, and sudden
death. In D. G. Byrne & R. H. Rosenmann (Eds.), Anxiety and the heart:
The series in health psychology and behavioral medicine (pp. 251258).
New York: Hemisphere.
Anda, R. F., Williamson, D. F., Escobedo, L. G., Mast, E. E., Giovino,
G. A., & Remington, P. L. (1990). Depression and the dynamics of
smoking: A national perspective. Journal of the American Medical
Association, 264, 15411545.
Anda, R., Williamson, D., Jones, D., Macera, C., Eaker, E., Glassman, A.,
& Marks, J. (1993). Depressed affect, hopelessness, and the risk of
ischemic heart disease in a cohort of U.S. adults. Epidemiology, 4,
Anderson, N. B. (1999). Solving the puzzle of socioeconomic status and
health: The need for integrated, multilevel, interdisciplinary research. In
N. E. Adler, M. Marmot, B. S. McEwen, & J. Stewart (Eds.), Annals of
the New York Academy of Sciences: Vol. 896. Socioeconomic status and
health in industrialized nations (pp. 302312). New York: New York
Academy of Sciences.
Anderson, N. B., & Armstead, C. A. (1995). Toward understanding the
association of socioeconomic status and health: A new challenge for the
biopsychosocial approach. Psychosomatic Medicine, 57, 213225.
Anthony, J. C., & Petronis, K. R. (1991). Suspected risk factors for
depression among adults 1844 years old. Epidemiology, 2, 123132.
Antonovsky, A. (1967). Social class, life expectancy and overall mortality.
Milbank Quarterly, 45, 3173.
Appels, A., Ba¨r, F. W., Ba¨r, J., Bruggeman, C., & de Baets, M. (2000).
Inflammation, depressive symptomatology, and coronary artery disease.
Psychosomatic Medicine, 62, 601605.
Aromaa, A., Raitasalo, R., Reunanen, A., Impivaara, O., Heliovaara, M.,
Knekt, P., et al. (1994). Depression and cardiovascular diseases. Acta
Psychiatrica Scandinavica, 89(Suppl. 377),7782.
Backlund, E., Sorlie, P. D., & Johnson, N. J. (1996). The shape of the
relationship between income and mortality in the United States: Evi-
dence from the National Longitudinal Mortality Study. Annals of Epi-
demiology, 6,1220.
Bajwa, W. K., Asnis, G. M., Sanderson, W. C., Irfan, A., & van Praag,
H. M. (1992). High cholesterol levels in patients with panic disorder.
American Journal of Psychiatry, 149, 376378.
Bandura, A. (1989). Human agency in social cognitive theory. American
Psychologist, 44, 11751184.
Barefoot, J. C. (1992). Developments in the measurement of hostility. In
H. S. Friedman (Ed.), Hostility, coping, and health (pp. 1331). Wash-
ington, DC: American Psychological Association.
Barefoot, J. C., Dahlstrom, W. G., & Williams, R. B. (1983). Hostility,
CHD incidence, and total mortality: A 25-year follow-up study of 255
physicians. Psychosomatic Medicine, 45, 5963.
Barefoot, J. C., Dodge, K. A., Peterson, B. L., Dahlstrom, W. G., &
Williams, R. B., Jr. (1989). The CookMedley Hostility Scale: Item
content and ability to predict survival. Psychosomatic Medicine, 51,
Barefoot, J. C., Helms, M. S., Mark, D. B., Blumenthal, J. A., Califf, R. M.,
Haney, T. L., et al. (1996). Depression and long-term mortality risk in
patients with coronary artery disease. American Journal of Cardiol-
ogy, 78, 613617.
Barefoot, J. C., Larsen, S., von der Lieth, L., & Schroll, M. (1995).
Hostility, incidence of acute myocardial infarction, and mortality in a
sample of older Danish men and women. American Journal of Epide-
miology, 142, 477484.
Barefoot, J. C., Peterson, B. L., Dahlstrom, W. G., Siegler, I. C., Anderson,
N. B., & Williams, R. B., Jr. (1991). Hostility patterns and health
implications: Correlates of CookMedley Hostility Scale scores in a
national survey. Health Psychology, 10, 1824.
Barefoot, J. C., & Schroll, M. (1996). Symptoms of depression, acute
myocardial infarction, and total mortality in a community sample. Cir-
culation, 93, 19761980.
Baron, R. M., & Kenny, D. A. (1986). The moderatormediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychol-
ogy, 51, 11731182.
Beach, S. R., & OLeary, K. D. (1993). Dysphoria and marital discord: Are
dysphoric individuals at risk for marital maladjustment? Journal of
Marital & Family Therapy, 19, 355368.
Bebbington, P., Hurry, J., Tennant, J. C., Sturt, E., & Wing, J. K. (1981).
The epidemiology of mental disorders in Camberwell. Psychological
Medicine, 11, 561580.
Beck, A. T. (1971). Cognition, affect, and psychopathology. Archives of
General Psychiatry, 24, 495500.
Beck, A. T., & Beamesderfer, A. (1974). Assessment of depression: The
Depression Inventory. In P. Pinchot (Ed.), Psychological measurements
in psychopharmacology (pp. 151169). Basel, Switzerland: Karger
Belle, D. E. (1982). The impact of poverty on social networks and supports.
Marriage & Family Review, 5, 89103.
Belle, D. (1990). Poverty and womens mental health. American Psychol-
ogist, 45, 385389.
Ben-Shlomo, Y., & Davey Smith, G. (1991). Deprivation in infancy or
adult life: Which is more important for mortality risk? Lancet, 337,
Berkman, L. F., & Breslow, L. (1983). Health and ways of living: The
Alameda County Study. New York: Oxford University Press.
Bianchi, S. (1995). The changing demographic and socioeconomic char-
acteristics of single parent families. Marriage & Family Review, 20,
Blane, D., Hart, C. L., Davey Smith, G., Gillis, C. R., Hole, D. J., &
Hawthorne, V. M. (1996). Association of cardiovascular disease risk
factors with socioeconomic position during childhood and during adult-
hood. British Medical Journal, 313, 14341438.
Blazer, D. G., Hughes, D., George, L. K., Swartz, M., & Boyer, R. (1991).
Generalized anxiety disorder. In L. N. Robins & D. A. Regier (Eds.),
Psychiatric disorders in America (pp. 180203). New York: Free Press.
Blazer, D. G., Kessler, R. C., McGonagle, K. A., & Swartz, M. S. (1994).
The prevalence and distribution of major depression in a national com-
munity sample: The National Comorbidity Survey. American Journal of
Psychiatry, 151, 979986.
Blumenthal, J. A., Williams, R. S., Wallace, A. G., Williams, R. B., &
Needles, T. L. (1982). Physiological and psychological variables predict
compliance to prescribed exercise therapy in patients recovering from
myocardial infarction. Psychosomatic Medicine, 44, 519527.
Bolger, N., Foster, M., Vinokur, A. D., & Ng, R. (1996). Close relation-
ships and adjustment to a life crisis: The case of breast cancer. Journal
of Personality and Social Psychology, 70, 283294.
Bolger, N., & Zuckerman, A. (1995). A framework for studying person-
ality in the stress process. Journal of Personality and Social Psychol-
ogy, 69, 890902.
Bosma, H., Van de Mheen, H. D., & Mackenbach, J. P. (1999). Social class
in childhood and general health in adulthood: Questionnaire study of
contribution of psychological attributes. British Medical Journal, 318,
Bradley, R. H., & Corwyn, R. F. (2002). Socioeconomic status and child
development. Annual Review of Psychology, 53, 371399.
Breslau, N. M., Kilbey, M., & Andreski, P. (1991). Nicotine dependence,
major depression, and anxiety in young adults. Archives of General
Psychiatry, 48, 10691074.
Brown, G. W., & Bifulco, A. (1990). Motherhood, employment, and the
development of depression: A replication of a finding. British Journal of
Psychiatry, 156, 169179.
Brown, G. W., & Harris, T. (1978). Social origins of depression. London:
Brown, G. W., & Moran, P. M. (1997). Single mothers, poverty, and
depression. Psychological Medicine, 27, 2133.
Brown, R. (1965). Social psychology. New York: Free Press.
Bruce, M. L., Leaf, P. J., Rozal, G. P., Florio, L., & Hoff, R. A. (1994).
Psychiatric status and 9-year mortality data in the New Haven Epide-
miologic Catchment Area Study. American Journal of Psychiatry, 151,
Bruce, M. L., Takeuchi, D. T., & Leaf, P. J. (1991). Poverty and psychiatric
status. Archives of General Psychiatry, 48, 470474.
Brugha, T. (1995). Social support and psychiatric disorder. Cambridge,
England: Cambridge University Press.
Bureau of Labor Statistics. (2000). Current population survey: Civilian
employmentpopulation ratio (Series LFS1601702). Washington, DC:
Carmelli, D., Rosenman, R. H., & Swan, G. E. (1988). The Cook and
Medley Ho Scale: A heritability analysis in adult male twins. Psycho-
somatic Medicine, 50, 165170.
Carmody, T. P., Crossen, J. R., & Wiens, A. N. (1989). Hostility as a health
risk factor: Relationships with neuroticism, Type A behavior, attentional
focus, and interpersonal style. Journal of Clinical Psychology, 45, 754
Carney, R. M., Freedland, K. E., Eisen, S. A., Rich, M. W., & Jaffe, A. S.
(1995). Major depression and medication adherence in elderly patients
with coronary artery disease. Health Psychology, 14, 8890.
Carney, R. M., Freedland, K. E., & Jaffe, A. S. (2001). Depression as a risk
factor for coronary heart disease mortality. Archives of General Psychi-
atry, 58, 229230.
Carney, R. M., Freedland, K. E., Sheline, Y., & Weiss, E. (1997). Depres-
sion and coronary heart disease: A review for cardiologists. Clinical
Cardiology, 20, 196200.
Carney, R. M., Rich, M. W., Freedland, K. E., & Sanai, J. (1988). Major
depressive disorder predicts cardiac events in patients with coronary
artery disease. Psychosomatic Medicine, 50, 627633.
Carney, R. M., Rich, M., & Jaffe, A. S. (1995). Depression as a risk factor
for cardiac events in established coronary heart disease: A review of
possible mechanisms. Annals of Behavioral Medicine, 17, 142149.
Carroll, D., Bennett, P., & Davey Smith, G. (1993). Socio-economic health
inequalities: Their origins and implications. Psychological Health, 8,
Casale, P. N., Devereux, R. B., & Milner, M. (1986). Value of echocar-
diographic measurement of left ventricular mass in predicting cardio-
vascular morbid events in hypertensive men. Annals of Internal Medi-
cine, 105, 173178.
Caspi, A., Elder, G. H., & Bem, D. J. (1987). Moving against the world:
Life-course patterns of explosive children. Developmental Psychol-
ogy, 23, 308313.
Chapin, C. F. (1924). Deaths among taxpayers and non-taxpayers of
income tax, Providence, 1865. American Journal of Public Health, 14,
Chen, E., & Matthews, K. A. (2001). Cognitive appraisal biases: An
approach to understanding the relationship between socioeconomic sta-
tus and cardiovascular reactivity in children. Annals of Behavioral
Medicine, 23, 101111.
Chen, E., Matthews, K. A., & Boyce, W. T. (2002). Socioeconomic
differences in childrens health: How and why do these relationships
change with age? Psychological Bulletin, 128, 295329.
Christensen, A. J., Edwards, D. L., Wiebe, J. S., Benotsch, E. G., &
McKelvey, L. (1996). Effect of verbal self-disclosure on NK cell activ-
ity: Moderating influence of cynical hostility. Psychosomatic Medi-
cine, 58, 150155.
Christmas, A. L., Wodarski, J. S., & Smokowski, P. R. (1996). Risk factors
for physical child abuse: A practice theoretical paradigm. Family Ther-
apy, 23, 233248.
Clark, L. A., & Watson, D. (1991). Theoretical and empirical issues in
differentiating depression from anxiety. In J. Becker & A. Kleinman
(Eds.), Psychosocial aspects of depression (pp. 3965). Hillsdale, NJ:
Clark, L. A., Watson, D., & Mineka, S. (1994). Temperament, personality,
and the mood and anxiety disorders. Journal of Abnormal Psychology,
103, 103116.
Clore, G. L. (1994). Why emotions require cognition. In P. Ekman & R. J.
Davidson (Eds.), The nature of emotion: Fundamental questions (pp.
181191). New York: Oxford University Press.
Cohen, S., & Herbert, T. B. (1996). Health psychology: Psychological
factors and disease from the perspective of human psychoneuro-
immunology. Annual Review of Psychology, 47, 113142.
Cohen, S., Kaplan, G. A., & Salonen, J. T. (1999). The role of psycholog-
ical characteristics in the relation between socioeconomic status and
perceived health. Journal of Applied Social Psychology, 29, 445468.
Cohen, S., & Wills, T. A. (1985). Stress, social support, and the buffering
hypothesis. Psychological Bulletin, 98, 310357.
Comstock, G. W., & Helsing, K. J. (1976). Symptoms of depression in two
communities. Psychological Medicine, 6, 551563.
Cook, W. W., & Medley, D. M. (1954). Proposed hostility and pharisaic-
virtue scales for the MMPI. Journal of Applied Psychology, 6, 414418.
Coryell, W., Endicott, J., & Keller, M. (1992). Major depression in a
nonclinical sample: Demographic and clinical risk factors for first onset.
Archives of General Psychiatry, 49, 117125.
Coryell, W., Noyes, R., & Clancy, J. (1982). Excess mortality in panic
disorder. Archives of General Psychiatry, 39, 701703.
Coryell, W., Noyes, R., & House, D. (1986). Mortality among outpatients
with anxiety disorders. American Journal of Psychiatry, 143, 608510.
Covinsky, K. A., Kahana, E., Chin, M., Palmer, R., Fortinsky, R., &
Landefeld, S. (1999). Depressive symptoms and 3-year mortality in
older hospitalized medical patients. Annals of Internal Medicine, 130,
Craig, T. J., & Van Natta, P. A. (1979). Influence of demographic char-
acteristics on two measures of depressive symptoms. Archives of Gen-
eral Psychiatry, 36, 149154.
Daly, M., Duncan, G., Kaplan, G. A., & Lynch, J. W. (1998). Macro-to-
micro linkages in the inequality-mortality relationship. Milbank Quar-
terly, 76, 315339.
Davey Smith, G., Hart, C., Blane, D., Gillis, C., & Hawthorne, V. (1997).
Lifetime socioeconomic position and mortality: Prospective observa-
tional study. British Medical Journal, 314, 547552.
Davey Smith, G., Hart, C., Blane, D., & Hole, D. (1998). Adverse socio-
economic conditions in childhood and cause specific adult mortality:
Prospective observational study. British Medical Journal, 316, 1631
Dembroski, T. M., MacDougall, J. M., Williams, R. B., Haney, T. L., &
Blumenthal, J. A. (1985). Components of Type A, hostility and anger-in:
Relationship to angiographic findings. Psychosomatic Medicine, 47,
Denollet, J., & Brutsaert, D. L. (1998). Personality, disease severity, and
the risk of long-term cardiac events in patients with a decreased ejection
fraction after myocardial infarction. Circulation, 97, 16773.
Dew, M. A. (1998). Psychiatric disorder in the context of physical illness.
In B. P. Dohrenwend (Ed.), Adversity, stress, and psychopathology (pp.
177218). New York: Oxford University Press.
Diez-Roux, A. V. (1998). Bringing context back into epidemiology: Vari-
ables and fallacies in multilevel analysis. American Journal of Public
Health, 88, 216222.
Diez-Roux, A. V., Merkin, S. S., Arnett, D., Chambless, L., Massing, M.,
Nieto, F. J., et al. (2001). Neighborhood of residence and incidence of
coronary heart disease. New England Journal of Medicine, 345, 99106.
Diez-Roux, A. V., Nieto, F. J., Muntaner, C., Tyroler, H. A., Comstock,
G. W., Shahr, E., et al. (1997). Neighborhood environments and coro-
nary heart disease: A multilevel analysis. American Journal of Epide-
miology, 146, 4863.
Dohrenwend, B. P. (1990). Socioeconomic status (SES) and psychiatric
disorders: Are the issues still compelling? Social Psychiatry and Psy-
chiatric Epidemiology, 25, 4147.
Dohrenwend, B. P. (2000). The role of adversity and stress in psychopa-
thology: Some evidence and its implications for theory and research.
Journal of Health and Social Behavior, 41, 119.
Dohrenwend, B. P., & Dohrenwend, B. S. (1969). Social status and
psychological disorder. New York: Wiley.
Dohrenwend, B. P., Levav, I., Shrout, P., Schwartz, S., Naveh, G., Link,
B. G., et al. (1998). Ethnicity, socioeconomic status, and psychiatric
disorders: A test of the social causationsocial selection issue. In B. P.
Dohrenwend (Ed.), Adversity, stress, and psychopathology (pp. 285
318). New York: Oxford University Press.
Dohrenwend, B. S. (1973). Social status and stressful life events. Journal
of Personality and Social Psychology, 28, 225235.
Doornbos, G., & Kromhout, D. (1990). Educational level and mortality in
a 32-year follow-up study of 18-year-old men in the Netherlands. Inter-
national Journal of Epidemiology, 19, 374379.
Drever, F., Whitehead, M., & Roden, M. (1996). Current patterns and
trends in male mortality by social class (based on occupation). Popula-
tion Trends, 86, 1520.
Duncan, G. J. (1996). Income dynamics and health. International Journal
of Health Services, 26, 419444.
Eaker, E. D., Pinsky, J., & Castelli, W. P. (1992). Myocardial infarction
and coronary death among women: Psychosocial predictors from a
20-year follow-up of women in the Framingham Study. American Jour-
nal of Epidemiology, 135, 854864.
Eaton, W. W., Armenian, H., Gallo, J., Pratt, L., & Ford, D. E. (1996).
Depression and risk for onset of Type II diabetes: A prospective
population-based study. Diabetes Care, 19, 10971102.
Eaton, W. W., Dryman, A., & Weissman, M. M. (1991). Panic and phobia.
In L. N. Robins & D. A. Regier (Eds.), Psychiatric disorders in America
(pp. 155179). New York: Free Press.
Eaton, W. W., Kessler, R. C., Wittchen, H. U., & Magee, W. J. (1994).
Panic and panic disorder in the United States. American Journal of
Psychiatry, 151, 413420.
Eaton, W. W., & Keyl, P. M. (1990). Risk factors for the onset of
diagnostic interview schedule/DSMIII agoraphobia in a prospective
population-based study. Archives of General Psychiatry, 47, 819824.
Eaton, W. W., & Muntaner, C. (1999). Socioeconomic stratification and
mental disorder. In A. V. Horwitz & T. L. Scheid (Eds.), A handbook for
the study of mental health: Social contexts, theories, and systems (pp.
259283). New York: Cambridge University Press.
Ecob, R., & Davey Smith, G. (1999). Income and health: What is the nature
of the relationships? Social Science & Medicine, 48, 693705.
Ehlert, U., & Straub, R. (1998). Physiological and emotional response to
psychological stressors in psychiatric and psychosomatic disorders. In P.
Csermely (Ed.), Annals of the New York Academy of Sciences: Vol. 851.
Stress of life: From molecules to man (pp. 477486). New York: New
York Academy of Sciences.
Ekman, P., & Davidson, R. (1994). The nature of emotion: Fundamental
questions. New York: Oxford University Press.
Endicott, J., & Spitzer, R. L. (1979). Use of the Research Diagnostic
Criteria and the Schedule for Affective Disorders and Schizophrenia to
study affective disorders. American Journal of Psychiatry, 136, 5256.
Ennis, N. E., Hobfoll, S. E., & Schro¨der, K. E. E. (2000). Money doesnt
talk, it swears: How economic stress and resistance resources impact
inner-city womens depressive mood. American Journal of Community
Psychology, 28, 149173.
Ensel, W. M., & Lin, N. (1991). The life stress paradigm and psychological
distress. Journal of Health and Social Behavior, 32, 321341.
Evans, R. G., Barer, M. L., & Marmor, T. R. (Eds.). (1994). Why are some
people healthy and others not? The determinants of health of popula-
tions. New York: de Gruyter.
Everson, S. A., George, K. A., Goldberg, D. E., Lakka, T. A., Sivenius, J.,
& Salonen, J. T. (1999). Anger expression and incident stroke: Prospec-
tive evidence from the Kuopio Ischemic Heart Disease Study. Stroke, 30,
Everson, S. A., Goldberg, D. E., Kaplan, G. A., Cohen, R. D., Pukkala, E.,
Tuomilehto, J., & Salonen, J. T. (1996). Hopelessness and risk of
mortality and incidence of myocardial infarction and cancer. Psychoso-
matic Medicine, 58, 113121.
Everson, S. A., Kauhanen, J., Kaplan, G. A., Goldberg, D. E., Julkunen, J.,
Tuomilehto, J., & Salonen, J. T. (1997). Hostility and increased risk of
mortality and acute myocardial infarction: The mediating role of behav-
ioral risk factors. American Journal of Epidemiology, 146, 142152.
Everson, S. A., Roberts, R. E., Goldberg, D. E., & Kaplan, G. A. (1998).
Depressive symptoms and increased risk of stroke mortality over a
29-year period. Archives of Internal Medicine, 158, 11331138.
Farmer, M. E., Locke, B. Z., Moscicki, E. K., Dannenberg, A. L., Larson,
D. B., & Radloff, L. S. (1988). Physical activity and depressive symp-
toms: The NHANES I epidemiologic follow-up study. American Journal
of Epidemiology, 128, 13401351.
Feldman, J. J., Makuc, D. M., Kleinman, J. C., & Cornoni-Huntley, J.
(1989). National trends in educational differentials in mortality. Amer-
ican Journal of Epidemiology, 129, 919933.
Feldman, L. A. (1995). Variations in the circumplex structure of mood.
Personality and Social Psychology Bulletin, 21, 806817.
Fiscella, K., & Franks, P. (1997). Does psychological distress contribute to
racial and socioeconomic disparities in mortality? Social Science &
Medicine, 45, 18051809.
Fleet, R. P., & Beitman, B. D. (1998). Cardiovascular death from panic
disorder and panic-like anxiety: A critical review of the literature.
Journal of Psychosomatic Research, 44, 7180.
Flory, J. D., Matthews, K. A., & Owens, J. F. (1998). A social information
processing approach to dispositional hostility: Relationships with nega-
tive mood and blood pressure elevations at work. Journal of Social and
Clinical Psychology, 17, 491504.
Ford, D. E., Mead, L. A., Chang, P. P., Cooper-Patrick, L., Wang, N., &
Klag, M. J. (1998). Depression is a risk factor for coronary artery disease
in men. Archives of Internal Medicine, 158, 14221426.
Fox, J. W. (1990). Social class, mental illness, and social mobility: The
social selection-drift hypothesis for serious mental illness. Journal of
Health and Social Behavior, 31, 344353.
Frasure-Smith, N., Lespe´rance, F., Gravel, G., Masson, A., Juneau, M.,
Talajic, M., & Bourassa, M. G. (2000). Social support, depression, and
mortality during the first year after myocardial infarction. Circulation,
101, 19191924.
Frasure-Smith, N., Lespe´rance, F., Juneau, M., Talajic, M., & Bourassa,
M. G. (1999). Gender, depression, and one-year prognosis after myo-
cardial infarction. Psychosomatic Medicine, 61, 2637.
Frasure-Smith, N., Lespe´rance, F., & Talajic, M. (1993). Depression fol-
lowing myocardial infarction: Impact on 6-month survival. Journal of
the American Medical Association, 270, 18191825.
Frasure-Smith, N., Lespe´rance, F., & Talajic, M. (1995a). Depression and
18-month prognosis after myocardial infarction. Circulation, 91, 9991005.
Frasure-Smith, N., Lespe´rance, F., & Talajic, M. (1995b). The impact of
negative emotions on prognosis following myocardial infarction: Is it
more than depression? Health Psychology, 14, 388398.
Fredrickson, B. L., & Levenson, R. W. (1998). Positive emotions speed
recovery from the cardiovascular sequelae of negative emotions. Cog-
nition & Emotion, 12, 191220.
Frijda, N. H. (1994). Emotions are functional most of the time. In P. Ekman
& R. J. Davidson (Eds.), The nature of emotion: Fundamental questions
(pp. 112122). New York: Oxford University Press.
Gallo, J. J., Royall, D. R., & Anthony, J. C. (1993). Risk factors for the
onset of depression in middle age and later life. Social Psychiatry and
Psychiatric Epidemiology, 28, 101108.
Gallo, L. C., & Matthews, K. A. (1999). Do negative emotions mediate the
association between socioeconomic status and health? In N. E. Adler, M.
Marmot, B. S. McEwen, & J. Stewart (Eds.), Annals of the New York
Academy of Sciences: Vol. 896. Socioeconomic status and health in
industrialized nations (pp. 226245). New York: New York Academy
of Sciences.
Gallo, L. C., Matthews, K. A., Kuller, L., Sutton-Tyrrell, K., & Edmun-
dowicz, D. (2001). Educational attainment and coronary and aortic
calcification in post-menopausal women. Psychosomatic Medicine, 63,
Gallo, L. C., & Smith, T. W. (1999). Patterns of hostility and social
support: Conceptualizing psychosocial risk factors as characteristics of
the person and the environment. Journal of Research in Personality, 33,
Gary, L. (1995). African American mens perceptions of racial discrimi-
nation. Social Work Research, 19, 207217.
Gecas, V., & Seff, M. (1989). Social class, occupational conditions, and
self-esteem. Sociological Perspectives, 32, 353364.
Glassman, A. H., & Shapiro, P. A. (1998). Depression and the course of
coronary artery disease. American Journal of Psychiatry, 155, 411.
Gonzalez, M. A., Rodriguez, A. F., & Calero, J. R. (1998). Relationship
between socioeconomic status and ischemic heart disease in cohort and
case-control studies: 19601993. International Journal of Epidemiol-
ogy, 27, 350358.
Gonzalez, M. B., Snyderman, T. B., Colket, J. T., Arias, R. M., Jiang,
J. W., OConnor, C. M., & Krishnan, K. R. (1996). Depression in
patients with coronary artery disease. Depression, 4, 5762.
Goodman, E. (1999). The role of socioeconomic status gradients in ex-
plaining differences in U.S. adolescentshealth. American Journal of
Public Health, 89, 15221528.
Gracia, E., Garcia, F., & Musitu, G. (1995). Macrosocial determinants of
social integration: Social class and area effect. Journal of Community &
Applied Social Psychology, 5, 105119.
Greene, S. M. (1989). The relationship between depression and hopeless-
ness: Implications for current theories of depression. British Journal of
Psychiatry, 154, 650659.
Guiry, E., Conroy, R. M., Hickey, N., & Mulcahy, R. (1987). Psycholog-
ical response to an acute coronary event and its effect on subsequent
rehabilitation and lifestyle change. Clinical Cardiology, 10, 256260.
Gump, B. B., Matthews, K. A., & Ra¨ikko¨nen, K. (1999). Modeling
relationships among socioeconomic status, hostility, cardiovascular re-
activity, and left ventricular mass in African American and White
children. Health Psychology, 18, 140150.
Haaga, D. A. F., Dyck, M. J., & Ernst, D. (1991). Empirical status of
cognitive theory of depression. Psychological Bulletin, 110, 215236.
Haan, M., Kaplan, G. A., & Camacho, T. (1987). Poverty and health:
Prospective evidence from the Alameda County Study. American Jour-
nal of Epidemiology, 125, 989998.
Haines, A. P., Imeson, J. D., & Meade, T. W. (1987). Phobic anxiety and
ischemic heart disease. British Medical Journal, 295, 297299.
Hammen, C. (1991). Generation of stress in the course of unipolar depres-
sion. Journal of Abnormal Psychology, 100, 555561.
Hayward, C. (1995). Psychiatric illness and cardiovascular disease risk.
Epidemiologic Reviews, 17, 129138.
Hearn, M. D., Murray, D. M., & Luepker, R. V. (1989). Hostility coronary
heart disease, and total mortality: A 33-year follow-up study of univer-
sity students. Journal of Behavioral Medicine, 12, 105121.
Hecker, M. H. L., Chesney, M. A., Black, G. W., & Frautchi, N. (1988).
Coronary-prone behaviors in the Western Collaborative Group Study.
Psychosomatic Medicine, 50, 153164.
Herbert, T. B., & Cohen, S. (1993). Depression and immunity: A meta-
analytic review. Psychological Bulletin, 113, 472486.
Hermann, C., Brand-Driehorst, S., Kaminsky, B., Leibing, E., Staats, H., &
Ruger, U. (1998). Diagnostic groups and depressed mood as predictors
of 22-month mortality in medical inpatients. Psychosomatic Medi-
cine, 60, 570577.
Himmelfarb, S., & Murrell, S. A. (1984). The prevalence and correlates of
anxiety symptoms in older adults. Journal of Psychology, 116, 159167.
Hobfoll, S. E. (1988). The ecology of stress. New York: Hemisphere.
Hobfoll, S. E. (1989). Conservation of resources: A new attempt at con-
ceptualizing stress. American Psychologist, 44, 513524.
Hobfoll, S. E. (1998). Stress, culture, and community: The psychology and
philosophy of stress. New York: Plenum Press.
Hobfoll, S. E. (2001). The influence of culture, community, and the
nested-self in the stress process: Advancing conservation of resources
theory. Applied Psychology, 50, 337370.
Hogan, D. P., Hao, L. X., & Parish, W. L. (1990). Race, kin networks, and
assistance to mother-headed families. Social Forces, 68, 797812.
Hokanson, J. E., Rubert, M. P., Welker, R. A., Hollander, G. R., & Hedeen,
C. (1989). Interpersonal concomitants and antecedents of depression
among college students. Journal of Abnormal Psychology, 98, 209217.
Holahan, C. J., & Moos, R. H. (1987). Risk, resistance, and psychological
distress: A longitudinal analysis with adults and children. Journal of
Abnormal Psychology, 96, 313.
Holahan, C. J., & Moos, R. H. (1991). Life stressors, personal and social
resources, and depression: A 4-year structural model. Journal of Abnor-
mal Psychology, 100, 3138.
Holahan, C. J., Moos, R. H., Holahan, C. K., & Cronkite, R. C. (1999).
Resource loss, resource gain, and depressive symptoms: A 10-year
model. Journal of Personality and Social Psychology, 77, 620629.
Holahan, C. J., Moos, R. H., Holahan, C. K., & Cronkite, R. C. (2000).
Long-term posttreatment functioning among patients with unipolar de-
pression: An integrative model. Journal of Consulting and Clinical
Psychology, 68, 226232.
Hollingshead, A. B. (1975). Four Factor Index of Social Status. New
Haven, CT: Hollingshead.
Holme, I., Helgeland, A., Hjermann, I., & Leren, P. (1982). Socio-
economic status as a coronary risk factor: The Oslo study. Acta Medica
Scandinavica, 660(Suppl.), 147151.
Horowitz, M. J., Berfari, R., Hulley, S., Blair, S., Alvarez, W., Borhani, N.,
et al. (1979). Life events, risk factors, and coronary disease. Psychoso-
matics, 20, 586592.
Horwath, E., Johnson, J., Klerman, G. L., & Weissman, M. M. (1992).
Depressive symptoms as relative and attributable risk factors for first-
onset major depression. Archives of General Psychiatry, 49, 817823.
House, J. S., Lepkowski, J. M., Kinney, A. M., Mero, R. P., Kessler, R. C.,
& Herzog, A. R. (1994). The social stratification of aging and health.
Journal of Health and Social Behavior, 35, 213234.
Houston, B. K., & Kelly, K. E. (1989). Hostility in employed women:
Relation to work and marital experiences, social support, stress, and
anger expression. Personality and Social Psychology Bulletin, 15, 175
Hughes, J. R., Hatsukami, D. K., Mitchell, J. E., & Dahlgren, L. A. (1986).
Prevalence of smoking among psychiatric outpatients. American Journal
of Psychiatry, 143, 993997.
Ickovics, J. R., Viscoli, C. M., & Horwitz, R. I. (1997). Functional
recovery after myocardial infarction in men: The independent effects of
social class. Annals of Internal Medicine, 127, 518525.
Illsey, R., & Baker, D. (1991). Contextual variations in the meaning of
health inequality. Social Science & Medicine, 32, 359365.
Iribarren, C., Sidney, S., Bild, D. E., Liu, K., Markovitz, J. H., Roseman,
J. M., & Matthews, K. (2000). Association of hostility with coronary
artery calcification in young adults: The CARDIA study. Coronary
artery risk development in young adults. Journal of the American Med-
ical Association, 283, 25462551.
Johnson, J. G., Cohen, P., Dohrenwend, B. P., Link, B. G., & Brook, J. S.
(1999). A longitudinal investigation of social causation and social se-
lection processes involved in the association between socioeconomic
status and psychiatric disorders. Journal of Abnormal Psychology, 108,
Johnson, S. L., & Jacob, T. (1997). Marital interactions of depressed men
and women. Journal of Consulting and Clinical Psychology, 65, 1523.
Johnson, T. P. (1991). Mental health, social relations, and social selection:
A longitudinal analysis. Journal of Health and Social Behavior, 32,
Joiner, T., & Coyne, J. (1999). The interactional nature of depression:
Advances in interpersonal approaches. Washington, DC: American Psy-
chological Association.
Jonas, B. S., Franks, P., & Ingram, D. D. (1997). Are symptoms of anxiety
and depression risk factors for hypertension? Longitudinal evidence
from the National Health and Nutrition Examination Survey I epidemi-
ologic follow-up study. Archives of Family Medicine, 6, 4349.
Jonas, B. S., & Mussolino, M. E. (2000). Symptoms of depression as a
prospective risk factor for stroke. Psychosomatic Medicine, 62, 463
Judd, L. L., & Akiskal, H. S. (2000). Delineating the longitudinal structure
of depressive illness: Beyond clinical subtypes and duration thresholds.
Pharmacopsychiatry, 33, 37.
Julkunen, J., Salonen, R., Kaplan, G. A., Chesney, M. A., & Salonen, J. T.
(1994). Hostility and the progression of carotid atherosclerosis. Psycho-
somatic Medicine, 56, 519525.
Jung, J. (1989). The relationship of daily hassles, social support, and
coping to depression in Black and White students. Journal of General
Psychology, 116, 407417.
Kaplan, G. A. (1995). Where do shared pathways lead? Some reflections
on a research agenda. Psychosomatic Medicine, 57, 208212.
Kaplan, G. A., & Keil, J. E. (1993). Socioeconomic factors and cardiovas-
cular disease: A review of the literature. Circulation, 88, 19731998.
Kaplan, G. A., Roberts, R. E., Camacho, T. C., & Coyne, J. (1987).
Psychosocial predictors of depression. American Journal of Epidemiol-
ogy, 125, 206220.
Kaplan, G. A., & Salonen, J. T. (1990). Socioeconomic conditions in
childhood and ischemic heart disease during middle age. British Medical
Journal, 301, 11211123.
Kaplan, G. A., Wilson, T. W., Cohen, R. D., Kauhanen, J., Wu, M., &
Salonen, J. T. (1994). Social functioning and overall mortality: Prospec-
tive evidence from the Kuopio Ischemic Heart Disease Risk Factor
Study. Epidemiology, 5, 495500.
Kawachi, I., Colditz, G. A., Ascherio, A., Rimm, E. B., Giovannucci, E.,
Stampfer, M. J., & Willett, W. C. (1994). Prospective study of phobic
anxiety and risk of coronary heart disease in men. Circulation, 89,
Kawachi, I., Sparrow, D., Spiro, A., Vokonas, P., & Weiss, S. T. (1996).
A prospective study of anger and coronary heart disease: The normative
aging study. Circulation, 94, 20902095.
Kawachi, I., Sparrow, D., Vokonas, P. S., & Weiss, S. T. (1994). Symp-
toms of anxiety and risk of coronary heart disease: The normative aging
study. Circulation, 90, 22252229.
Kennedy, B. P., Kawachi, I., Glass, R., & Prothrow-Stith, D. (1998).
Income distribution, socioeconomic status, and self-rated health in the
United States: Multilevel analysis. British Medical Journal, 317, 917
Kennedy, G. J., Hofer, M. A., Cohen, D., Shindledecker, R., & Fisher, J. D.
(1987). Significance of depression and cognitive impairment in patients
undergoing programmed stimulation of cardiac arrhythmias. Psychoso-
matic Medicine, 49, 410421.
Kessler, R. C. (1979). Stress, social status, and psychological distress.
Journal of Health and Social Behavior, 20, 259272.
Kessler, R. C. (1982). A disaggregation of the relationship between socio-
economic status and psychological distress. American Sociological Re-
view, 47, 752764.
Kessler, R. C., & Cleary, P. D. (1980). Social class and psychological
distress. American Sociological Review, 45, 463478.
Kessler, R. C., Foster, C. L., Saunders, W. B., & Stang, P. E. (1995). Social
consequences of psychiatric disorders, I: Educational attainment. Amer-
ican Journal of Psychiatry, 152, 10261032.
Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C. B., Hughes, M.,
Eshleman, S., et al. (1994). Lifetime and 12-month prevalence of DSM
IIIRpsychiatric disorders in the United States. Archives of General
Psychiatry, 51, 819.
Keyl, P. M., & Eaton, W. W. (1990). Risk factors for the onset of panic
disorder and other panic attacks in a prospective, population-based
study. American Journal of Epidemiology, 131, 301311.
Kitagawa, E. M., & Hauser, P. M. (1973). Differential mortality in the
United States: A study in socioeconomic epidemiology. Cambridge, MA:
Harvard University Press.
Kohn, R., Dohrenwend, B. P., & Mirotznik, J. (1998). Epidemiological
findings on selected psychiatric disorders in the general population. In
B. P. Dohrenwend (Ed.), Adversity, stress, and psychopathology (pp.
235284). New York: Oxford University Press.
Kop, W. J. (1999). Chronic and acute psychological risk factors for clinical
manifestations of coronary artery disease. Psychosomatic Medicine, 61,
Koskenvuo, M., Kaprio, J., Rose, R. J., Kesaniemi, A., Sarna, S., Heikkila,
K., & Langinvainio, H. (1988). Hostility as a risk factor for mortality and
ischemic heart disease in men. Psychosomatic Medicine, 50, 330340.
Kouzis, A. C., & Eaton, W. W. (2000). Psychopathology and the initiation
of disability payments. Psychiatric Services, 51, 908913.
Krantz, D. S., & McCeney, M. K. (2002). Effects of psychological and
social factors on organic disease: A critical assessment of research on
coronary heart disease. Annual Review of Psychology, 53, 341369.
Kraus, J. F., Borhani, N. O., & Franti, C. E. (1980). Socioeconomic status,
ethnicity, and risk of coronary heart disease. American Journal of Public
Health, 82, 703710.
Krause, N. (1991). Stress and isolation from close ties in later life. Journal
of Gerontology: Social Sciences, 46, 183195.
Krause, N. (1997). Social support and feelings of personal control in later
life. In G. R. Pierce & B. Lakey (Eds.), Sourcebook of social support and
personality (pp. 335355). New York: Plenum Press.
Krause, N., & Borawski-Clark, E. (1995). Social class differences in social
support among older adults. Gerontologist, 35, 498508.
Krieger, N. (2000). Discrimination and health. In L. F. Berkman & I.
Kawachi (Eds.), Social epidemiology (pp. 3675). New York: Oxford
University Press.
Krieger, N., Williams, D. R., & Moss, N. E. (1997). Measuring social class
in U.S. public health research: Concepts, methodologies, and guidelines.
Annual Review of Public Health, 18, 341378.
Kubzansky, L. D., Kawachi, I., & Sparrow, D. (1999). Socioeconomic
status, hostility, and risk factor clustering in the Normative Aging Study:
Any help from the concept of allostatic load? Annals of Behavioral
Medicine, 21, 330338.
Kubzansky, L. D., Kawachi, I., Spiro, A., Weiss, S. T., Vokonas, P. S., &
Sparrow, D. (1997). Is worrying bad for your heart? A prospective study
of worry and coronary heart disease in the Normative Aging Study.
Circulation, 95, 818824.
Kubzansky, L. D., Kawachi, I., Weiss, S. T., & Sparrow, D. (1998).
Anxiety and coronary heart disease: A synthesis of epidemiological
psychological, and experimental evidence. Annals of Behavioral Medi-
cine, 20, 4758.
Kuh, D., & Ben-Shlomo, Y. (1997). A life course approach to chronic
disease epidemiology. Oxford, England: Oxford University Press.
Lachman, M. E., & Weaver, S. L. (1998). The sense of control as a
moderator of social class differences in health and well-being. Journal of
Personality and Social Psychology, 74, 763773.
Landrine, H., & Klonoff, E. A. (1996). The schedule of racist events: A
measure of racial discrimination and a study of its negative physical and
mental health consequences. Journal of Black Psychology, 22, 144168.
Lane, D., Carroll, D., Ring, C., Beevers, D. G., & Lip, G. Y. (2001).
Mortality and quality of life 12 months after myocardial infarction:
Effects of depression and anxiety. Psychosomatic Medicine, 63, 221
Langer, T. S., & Michael, S. T. (1963). Life stress and mental health. New
York: Free Press.
Lantz, P. M., House, J. S., Lepowski, J. M., Williams, D. R., Mero, R. P.,
& Chen, J. (1998). Socioeconomic factors, health behaviors, and mor-
tality: Results from a nationally representative prospective study of U.S.
adults. Journal of the American Medical Association, 279, 17031708.
Lau, R. R. (1988). Beliefs about control and health behavior. In D. S.
Gochman (Ed.), Health behavior: Emerging research perspectives (pp.
4363). New York: Plenum Press.
Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford Uni-
versity Press.
Leiker, M., & Hailey, B. J. (1988). A link between hostility and disease:
Poor health habits? Behavioral Medicine, 3, 129133.
Leon, G. R., Finn, S. E., Murray, D., & Bailey, J. M. (1988). The inability
to predict cardiovascular disease from hostility scores or MMPI items
related to Type A behavior. Journal of Consulting and Clinical Psychol-
ogy, 56, 597600.
Lespe´rance, F., Frasure-Smith, N., Juneau, M., & The´roux, P. (2000).
Depression and 1-year prognosis in unstable angina. Archives of Internal
Medicine, 160, 13541360.
Levenstein, S., & Kaplan, G. A. (1998). Socioeconomic status and ulcer. A
prospective study of contributory risk factors. Journal of Clinical Gas-
troenterology, 26, 1427.
Levy, D., Garrison, R. J., Savage, D. D., Kannel, W. B., & Castelli, W. P.
(1990). Prognostic implications of echocardiographically determined
left ventricular mass in the Framingham Heart Study. New England
Journal of Medicine, 322, 15611566.
Lewinsohn, P. M., Steinmetz, J. L., Larson, D. W., & Franklin, J. (1981).
Depression-related cognitions: Antecedent or consequence? Journal of
Abnormal Psychology, 90, 213219.
Lichtenstein, P., Harris, J. R., Pedersen, N. L., & McClearn, G. E. (1993).
Socioeconomic status and physical health, how are they related? An
empirical study based on twins reared apart and twins reared together.
Social Science & Medicine, 36, 441450.
Link, B. G., Lennon, M. C., & Dohrenwend, B. P. (1993). Socioeconomic
status and depression: The role of occupations involving direction,
control, and planning. American Journal of Sociology, 98, 13511387.
Link, B. G., & Phelan, J. C. (1995). Social conditions as fundamental
causes of disease. Journal of Health and Social Behavior, 92, 8094.
Link, B. G., & Phelan, J. C. (1996). Understanding sociodemographic
differences in healththe role of fundamental social causes. American
Journal of Public Health, 86, 471473.
Lockhart, L. L. (1987). A reexamination of the effects of race and social
class on the incidence of marital violence: A search for reliable differ-
ence. Journal of Marriage and the Family, 49, 603610.
Lown, B. (1982). Mental stress, arrhythmias and sudden death. American
Journal of Medicine, 72, 177180.
Lynch, J., & Kaplan, G. (2000). Socioeconomic position. In L. F. Berkman
& I. Kawachi (Eds.), Social epidemiology (pp. 1335). New York:
Oxford University Press.
Lynch, J. W., Kaplan, G. A., Cohen, R. D., Tuomilehto, J., & Salonen, J. T.
(1996). Do cardiovascular risk factors explain the relation between
socioeconomic status, risk of all-cause mortality, cardiovascular mortal-
ity, and acute myocardial infarction? American Journal of Epidemiol-
ogy, 144, 934942.
Lynch, J. W., Kaplan, G. A., & Salonen, J. T. (1997). Why do poor people
behave poorly? Variation in adult health behaviours and psychosocial
characteristics by stages of the socioeconomic lifecourse. Social Science
& Medicine, 44, 809819.
Lynch, J. W., Kaplan, G. A., & Shema, S. J. (1997). Cumulative impact of
sustained economic hardship on physical, cognitive, psychological, and
social functioning. New England Journal of Medicine, 337, 18891895.
Lynch, J. W., Krause, N., Kaplan, G. A., Tuomilehto, J., & Salonen, J. T.
(1997). Workplace conditions, socioeconomic status, and the risk of
mortality and acute myocardial infarction: The Kuopio Ischemic Heart
Disease Risk Factor Study. American Journal of Public Health, 87,
Lynch, J. W., Smith, G. D., Kaplan, G. A., & House, J. S. (2000). Income
inequality and mortality: Importance to health of individual income,
psychosocial environment, or material conditions. British Medical Jour-
nal, 320, 12001204.
Macintyre, S. (1997). The Black Report and beyond: What are the issues?
Social Science & Medicine, 44, 723745.
Macintyre, S., & Ellaway, A. (2000). Ecological approaches: Rediscover-
ing the role of the physical and social environment. In L. F. Berkman &
I. Kawachi (Eds.), Social epidemiology (pp. 332348). New York:
Oxford University Press.
Macintyre, S., Maciver, S., & Solomon, A. (1993). Area, class and health:
Should we be focusing on places or people? Social Policy, 22, 213234.
Magee, W. J., Eaton, W. W., Wittchen, H. U., McGonagle, K. A., &
Kessler, R. C. (1996). Agoraphobia, simple phobia, and social phobia in
the National Comorbidity Survey. Archives of General Psychiatry, 53,
Manuck, S. B. (1994). Cardiovascular reactivity in cardiovascular disease:
Once more unto the breach.International Journal of Behavioral
Medicine, 1, 431.
Manuck, S. B., Flory, J. D., McCaffery, J. M., Matthews, K. A., Mann,
J. J., & Muldoon, M. F. (1998). Aggression, impulsivity, and central
nervous system serotonergic responsivity in a nonpatient sample. Neu-
ropsychopharmacology, 19, 287299.
Markovitz, J. H., & Matthews, K. A. (1991). Platelets and coronary heart
disease: Potential psychophysiologic mechanisms. Psychosomatic Med-
icine, 53, 643668.
Markovitz, J. H., Matthews, K. A., Kannel, W. B., Cobb, J. L., &
DAgostino, R. B. (1993). Psychological predictors of hypertension in
the Framingham Study: Is there tension in hypertension? Journal of the
American Medical Association, 6, 24392443.
Markovitz, J. H., Matthews, K. A., Wing, R. R., Kuller, L. H., & Meilahn,
E. N. (1991). Psychological, biological and health behavior predictors of
blood pressure changes in middle-aged women. Journal of Hyperten-
sion, 9, 399406.
Marmot, M. G. (1998). Contribution of psychosocial factors to socioeco-
nomic differences in health. Milbank Quarterly, 76, 403448.
Marmot, M., Bobak, M., & Smith, D. G. (1995). Explanations for social
inequalities in health. In B. Amick III, S. Levine, A. R. Tarov, & D.
Walsh (Eds.), Society and health (pp. 172210). New York: Oxford
University Press.
Marmot, M. G., Kogevinas, M., & Elston, M. A. (1987). Social/economic
status and disease. Annual Review of Public Health, 8, 111135.
Marmot, M., Ryff, C. D., Bumpass, L. L., Shipley, M., & Marks, N. F.
(1997). Social inequalities in health: Next questions and converging
evidence. Social Science & Medicine, 44, 901910.
Marmot, M., Shipley, M., Brunner, E., & Hemingway, H. (2001). Relative
contribution of early life and adult socioeconomic factors to adult
morbidity in the Whitehall II Study. Journal of Epidemiology and
Community Health, 55, 301307.
Marmot, M. G., Shipley, M. J., & Rose, G. (1984). Inequalities in death:
Specific explanations of a general pattern? Lancet, 1, 10031006.
Marmot, M. G., Smith, G. D., Stansfeld, S., Patel, C., North, F., Head, J.,
et al. (1991). Health inequalities among British civil servants: The
Whitehall II Study. Lancet, 337, 13871393.
Marmot, M., & Theorell, T. (1988). Social class and cardiovascular dis-
ease: The contribution of work. International Journal of Health Ser-
vices, 18, 659674.
Martin, R. L., Cloninger, R., Guze, S. B., & Clayton, P. J. (1985). Mortality
in a follow-up of 500 psychiatric outpatients. Archives of General
Psychiatry, 42, 5866.
Maruta, T., Hamburgen, M. E., Jennings, C. A., Offord, K. P., Colligan,
R. C., Frye, R. L., & Malinchoc, M. (1993). Keeping hostility in
perspective: Coronary heart disease and the Hostility Scale on the
Minnesota Multiphasic Personality Inventory. Mayo Clinic Proceed-
ings, 68, 192193.
Matsumoto, Y., Uyama, O., Shimizu, S., Michishita, H., Mori, R., Owada,
T., & Sugita, M. (1993). Do anger and aggression affect carotid athero-
sclerosis? Stroke, 24, 983986.
Matthews, K. A. (1989). Are sociodemographic variables markers for
psychological determinants of heath? Health Psychology, 8, 641648.
Matthews, K. A., Flory, J. D., Muldoon, M. F., & Manuck, S. B. (2000).
Does socioeconomic status relate to central serotonergic responsivity in
healthy adults? Psychosomatic Medicine, 62, 231237.
Matthews, K. A., Kelsey, S. F., Meilahn, E. N., Kuller, L. H., & Wing,
R. R. (1989). Educational attainment and behavioral and biologic risk
factors for coronary heart disease in middle-aged women. American
Journal of Epidemiology, 129, 11321144.
Matthews, K. A., Kiefe, C. I., Lewis, C. E., Liu, K., Sidney, S., & Yunis,
C. (2001). Socioeconomic trajectories in a biracial cohort of young
adults: Risk for incident hypertension. Psychosomatic Medicine, 63,
Matthews, K. A., Owens, J. F., Kuller, L. H., Sutton-Tyrrell, K., Jansen-
McWilliams, L. (1998). Are hostility and anxiety associated with carotid
atherosclerosis in healthy postmenopausal women? Psychosomatic Med-
icine, 60, 633638.
Matthews, K. A., Ra¨ikko¨nen, K., Everson, S. A., Flory, J. D., Marco, C. A.,
Owens J. F., & Lloyd, C. E. (2000). Do the daily experiences of healthy
men and women vary according to occupational prestige and work
strain? Psychosomatic Medicine, 62, 346353.
McCranie, E. W., Watkins, L. O., Brandsma, J. M., & Sisson, B. D. (1986).
Hostility, coronary heart disease (CHD) incidence, and total mortality:
Lack of association in a 25-year follow-up study of 478 physicians.
Journal of Behavioral Medicine, 9, 119125.
McDonough, P., Duncan, G. J., Williams, D., & House, J. (1997). Income
dynamics and adult mortality in the United States, 19721989. American
Journal of Public Health, 87, 14761483.
McEwan, B. S., & Stellar, E. (1993). Stress and the individual: Mecha-
nisms leading to disease. Archives of Internal Medicine, 1534, 1093
McLeod, J. D., & Kessler, R. C. (1990). Socioeconomic status differences
in vulnerability to undesirable life events. Journal of Health and Social
Behavior, 31, 162172.
Mendes de Leon, C. F., Kop, W. J., de Swart, H. B., Bar, F. W., & Appels,
A. P. (1996). Psychosocial characteristics and recurrent events after
percutaneous transluminal coronary angioplasty. American Journal of
Cardiology, 77, 252255.
Miech, R. A., Caspi, A., Moffitt, T. E., Wright, B. R. E., & Silva, P. A.
(1999). Low socioeconomic status and mental disorders: A longitudinal
study of selection and causation during young adulthood. American
Journal of Sociology, 104, 10961131.
Miller, T. Q., Marksides, K. S., Chiriboga, D. A., & Ray, L. A. (1995). A
test of the psychosocial vulnerability and health behavior models of
hostility: Results from an 11-year follow-up study of Mexican Ameri-
cans. Psychosomatic Medicine, 57, 572581.
Miller, T. Q., Smith, T. W., Turner, C. W., Guijarro, M. L., & Hallet, A. J.
(1996). A meta-analytic review of research on hostility and physical
health. Psychological Bulletin, 119, 322348.
Mirowsky, J., Ross, C. E., & Willigen, M. V. (1996). Instrumentalism in
the land of opportunity: Socioeconomic causes and emotional conse-
quences. Social Psychology Quarterly, 59, 322337.
Mittleman, M. A., Maclure, M., Nachnani, M., Sherwood, J. B., & Muller,
J. E. (1997). Educational attainment, anger, and the risk of triggering
myocardial infarction onset. The determinants of myocardial infarction
onset study investigators. Archives of Internal Medicine, 157, 6975.
Mittleman, M. A., Maclure, M., Sherwood, J. B., Mulry, R. P., Tofler,
G. H., Jacobs, S. C., et al. (1995). Triggering of acute myocardial
infarction onset by episodes of anger. Circulation, 92, 17201725.
Mo¨ller, J., Hallqvist, J., Diderichsen, F., Theorell, T., Reuterwall, C., &
Ahlbom, A. (1999). Do episodes of anger trigger myocardial infarction?
A case-crossover analysis in the Stockholm Heart Epidemiology Pro-
gram (SHEEP). Psychosomatic Medicine, 61, 842849.
Moos, R. H., Cronkite, R. C., & Moos, B. S. (1998). The long-term
interplay between family and extrafamily resources and depression.
Journal of Family Psychology, 12, 326434.
Moser, D. K., & Dracup, K. (1996). Is anxiety early after myocardial
infarction associated with subsequent ischemic and arrhythmic events?
Psychosomatic Medicine, 58, 395401.
Muldoon, M. F., Sved, A. F., Flory, J. D., Perel, J. M., Matthews, K. A.,
& Manuck, S. B. (1998). Inverse relationship between fenfluarmine-
induced prolactin release and blood pressure in humans. Hyperten-
sion, 32, 972975.
Murphy, J. M., Monson, R. R., Olivier, D. C., Sobol, A. M., & Leighton,
A. H. (1987). Affective disorders and mortality: A general population
study. Archives of General Psychiatry, 44, 473480.
Murphy, J. M., Olivier, D. C., Monson, R. R., Sobol, A. M., Federman,
E. B., & Leighton, A. H. (1991). Depression and anxiety in relation to
social status. Archives of General Psychiatry, 48, 223229.
Murrell, S. A., & Norris, F. H. (1991). Differential social support and life
change as contributors to the social class-distress relationship in older
adults. Psychology and Aging, 6, 223231.
Musselman, D. L., Evans, D. L., & Nemeroff, C. B. (1998). The relation-
ship of depression to cardiovascular disease. Archives of General Psy-
chiatry, 55, 580592.
Myers, J. K., Lindenthal., J. L., & Pepper, M. P. (1974). Social class, life
events, and psychiatric symptoms: A longitudinal study. In B. S.
Dohrenwend & B. P. Dohrenwend (Eds.), Stressful life events: Their
nature and effects (pp. 191206). New York: Wiley.
Newton, T. L., & Kiecolt-Glaser, J. K. (1995). Hostility and marital quality
during early marriage. Journal of Behavioral Medicine, 18, 601619.
Oakley, A., & Rajan, L. (1991). Social class and social support: The same
or different? Sociology, 25, 3159.
Oliver, M. L., & Shapiro, T. M. (1995). Black wealth/White wealth. New
York: Routledge.
Oxenkrug, G. F., Branconnier, R. J., Harto-Truax, N., & Cole, J. O. (1983).
Is serum cholesterol a biological marker for major depressive disorder?
American Journal of Psychiatry, 140, 920921.
Pappas, G., Queen, S., Hadden, W., & Fisher, G. (1993). The increasing
disparity in mortality between socioeconomic groups in the U.S., 1960
1986. New England Journal of Medicine, 329, 103109.
Paterniti, S., Alpe´rovitch, A., Ducimetie`re, P., Dealberto, M. J., Le´pine,
J. P., & Bisserbe, J. C. (1999). Anxiety but not depression is associated
with elevated blood pressure in a community group of French elderly.
Psychosomatic Medicine, 61, 7783.
Paykel, E. S. (1994). Life events, social support, and depression. Acta
Psychiatrica Scandinavica, 879(Suppl. 377), 5058.
Pearlin, L. I., Lieberman, M. A., Menaghan, E. G., & Mullan, J. T. (1981).
The stress process. Journal of Health and Social Behavior, 18, 221.
Penninx, B. W. J. H., Beekman, A. T. F., Honig, A., Deeg, D. J. H.,
Schoevers, R. A., van Eijk, T. M., & van Tilburg, W. (2001). Depression
and cardiac mortality: Results from a community-based longitudinal
study. Archives of General Psychiatry, 58, 221227.
Phillimore, P., Beattie, A., & Townsend, P. (1994). Widening inequality of
health in Northern England, 19811991. British Medical Journal, 308,
Pincus, T., Callahan, L. F., & Burkhasuer, R. V. (1987). Most chronic
diseases are reported more frequently by individuals with fewer than 12
years of education in the 1864 United States population. Journal of
Chronic Diseases, 40, 865874.
Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New York:
Harper & Row.
Pohl, R., Yeragani, R., Balon, R., Lycaki, H., & McBride, R. (1992).
Smoking in patients with panic disorder. Psychiatry Research, 43, 253