ARTICLE IN PRESS

Social Science & Medicine 65 (2007) 245–261

www.elsevier.com/locate/socscimed

Meta-analysis of stigma and mental health

Winnie W.S. Maka,, Cecilia Y.M. Poonb, Loraine Y.K. Puna, Shu Fai Cheungc
a
Department of Psychology, The Chinese University of Hong Kong, China
b
Department of Psychology, University of Southern California, USA
c
Department of Psychology, University of Macau, China
Available online 25 April 2007

Abstract

Recent research has emphasized the adverse effects of stigma on minority groups’ mental health. Governments and
service agencies have put much effort into combating stigma against a variety of conditions. Nevertheless, previous
empirical research on the stigma–mental health relationship has yielded inconclusive findings, varying from strong negative
to zero correlations. Thus, whether stigma is related significantly to mental health is yet to be confirmed. Using meta-
analysis, the associations between stigma and mental health from 49 empirical studies were examined across various
stigmatized conditions and mental health indices. Possible moderators were also explored. The mean correlation between
stigma and average mental health scores corrected for sampling error, unreliability, and other artifacts was .28
(N ¼ 10,567, k ¼ 52). No strong moderators were found, yet meaningful patterns were observed. Implications of the
results are discussed.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Meta-analysis; Stigma; Mental health; Review

Introduction (Crocker, Major, & Steele, 1998, p. 505). From the

general public’s view, stigma represents the endorse-
Social stigma has been identified and recognized ment of a set of prejudicial attitudes, negative
as a major concern in health care and human emotional responses, discriminatory behaviors, and
services across societies worldwide. It was first biased social structures towards members of a
defined as a spoiled identity that discredits a person subgroup (Corrigan, 2000). It involves labeling,
in society (Goffman, 1963). Individuals being stereotyping, separation, status loss, and discrimi-
relegated to this stigmatized status are believed ‘‘to nation of the stigmatized individuals in a power
possess some attribute that conveys a social identity situation (Link & Phelan, 2001).
that is devalued in a particular social context’’ Stigma can also be perceived by minority
members and their associates. Termed as self-
Corresponding author. Tel.: +852 2609 6577;
stigma, it is the internalization of the stigma
responses from the general public by the target
fax: +852 2603 5019.
E-mail addresses: [email protected] (W.W.S. Mak),
individuals. It occurs when members of a subgroup
[email protected] (C.Y.M. Poon), [email protected] internalize the prejudicial attitudes and apply these
(L.Y.K. Pun), [email protected] (S.F. Cheung). attitudes toward themselves, leading to negative

0277-9536/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2007.03.015
 ARTICLE IN PRESS
246 W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

emotional responses and behaviors (Corrigan & about US$2.5 million was committed to the
Watson, 2002). Similar internalization of stigma can program supporting up to 12 new competitive
occur among the associates. Affiliate stigma is the grants. From the advancements in public policies,
psychological responses and identification of indi- it could be shown that the influence of stigma on
viduals who are closely associated with the target mental health has been widely recognized and
individuals (e.g., caregivers, family, and friends). regarded as important in the government policy-
Through association, these affiliates may internalize making process.
the stigma attached to the target, affecting their own In addition to the growing research showing the
well-being. It should be noted that although we have potential impact of stigma on a variety of health
made a fine distinction among different concepts of conditions, health agencies have begun taking
stigma, the term stigma is used indiscriminately in initiatives to combat stigma. The World Health
the literature. For the purpose of this study, stigma Report (WHO, 2001) discussed possible ways to
is operationalized as internalized stigma construed reduce stigma, including organizing anti-stigma
by the minority members (a.k.a. self-stigma) and activities and campaigns through partnerships with
their associates (a.k.a. affiliate stigma) across non-government organizations. The International
different stigmatized conditions. In this meta- HIV/AIDS Alliance identified stigma as a barrier to
analysis we examine the relations between stigma HIV/AIDS prevention and began stigma reduction
and various positive and negative indicators of programs in some developing countries. The Global
mental health. Health Council also conducted research and pro-
Although the concept of stigma has been around vides resources to AIDS stigma prevention pro-
for decades, only in the past decade has it gained grams across the world. Various virtual campaigns
increasing attention in research and in policy- against stigma of mental illness were also developed.
making. A PsycINFO and PubMed search with These campaigns attempted to use the Internet and
the keyword stigma found more than 4278 related media to overcome stigma and monitor acts of
articles, with more than half of the articles being discrimination at both the local and the global level.
published between the years 2000 and 2005. This It is evident from these efforts that stigma has
indicates a growing interest in the concept of stigma been generally recognized as a global concern that
and its influence on psychological and medical must be addressed and overcome at multiple levels
research. Research on stigma encompassed a wide (e.g., individual, institutional systems, community,
range of conditions, including mental illness, and regional). Governments, agencies, organiza-
intellectual disability, HIV/AIDS, ethnicity, and tions, and research put resources into combating
sexual orientation, to name only a few. stigma. However, the empirical research evidence
In the Surgeon General’s Report on Mental shows an inconsistent relation between stigma and
Health (US Department of Health and Human mental health, ranging from non-significance to
Services, 1999), strong emphasis was put on under- strong correlations. Some researchers argued that
standing the roots of stigma, its effects on mental stigma does not necessary affect mental health
health, and ways to overcome stigma. Following the (Corrigan & Kleinlein, 2005; Crocker & Major,
issuance of this landmark report, in 2001, the 1989). Moreover, research suggests that stigma
National Institute of Mental Health (NIH) orga- varies in degree, depending on the specific diagnosis
nized a major international conference, ‘‘Stigma and and sociocultural group to which the target group
Global Health: Developing a Research Agenda,’’ to belongs (e.g., Lau & Cheung, 1999; Lee, Lee, Chiu,
arouse attention and research efforts in examining & Kleinman, 2005). Therefore, before concluding
the causes and consequences of stigma. Following that stigma and its effect on mental health are as
the NIH’s organizational restructuring in 2002, a important as we have been ardently suggesting, we
new Stigma and Health Disparities Program under need to empirically integrate and consolidate these
the Health and Behavior Research Branch in the findings to examine their association and to seek out
Division of AIDS and Health and Behavior possible moderators that could help us understand
Research was formally established. The program the relationship.
focused on the issues of stigma; such as supporting One means of aggregating research findings is
research to understand its mechanisms, developing through meta-analysis, a statistical analysis of
strategies to reduce stigma and examining media previous studies quantitatively to integrate findings
influences on mental illness perception. In 2003, (Glass, 1976). It helps researchers to integrate and
 ARTICLE IN PRESS
W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261 247

interpret primary research findings, which may coverage of mental health-related disciplines includ-
diverge both in strength and direction, and make a ing psychology, social work, psychiatry, nursing,
more stringent conclusion about the subject of health administration, pharmacology, rehabilita-
interest. Meta-analysis outweighs traditional narra- tion, epidemiology, and other human services
tive review methods in several ways. First, it is based professions. PubMed covers an array of life science
on shared subjectivity of primary research rather and biomedical journals that span fields of medi-
than the reviewer’s own stance. Second, statistical cine, nursing, dentistry, and health care system.
artifacts affecting the reported results of primary Both bibliographic databases have worldwide cov-
research can be corrected before aggregating and erage; nevertheless, they include mostly records
making judgments. Moreover, moderators can be written in English.
examined through statistical methods in meta- To maximize the search in the databases, both
analysis. keywords were exploded using the thesaurus func-
The purpose of our study is to synthesize previous tion. Keywords for stigma included stigma, atti-
findings on stigma and mental health in under- tudes, discrimination, labeling, prejudice, social
standing the overall association between these two acceptance, social approval, social discrimination,
constructs. In addition to examining the integrated social perception, stereotypes, and stereotypes-
effects between stigma and mental health, we attitudes. Keywords for mental health included
conducted a quality assessment of the studies. mental health, emotional adjustment, well-being,
According to Khan, ter Riet, Popay, Nixon, and life satisfaction, quality of life, mental disorders,
Kleijnen (2001) and Moher, Cook, Eastwood, distress, and depression. Based on the above
Olkin, Rennie, and Stroup (1999), assessment of keywords, a total of 808 studies were identified
study quality is an important aspect in synthesized from the two databases. Four research assistants
reviews and its rigor is an indication of the quality with background in psychology independently
of the meta-analysis itself. Based on Wortman reviewed the abstracts of these studies based on
(1994), the present study assessed study quality the following inclusion criteria: (1) empirical and
along four dimensions: theory, publication bias, quantitative studies reported in English, (2) the
design, and sources of heterogeneity. Theory relationship between stigma and mental health was
evaluates construct validity, or whether the study mentioned and, (3) at least one measure of stigma
focuses on testing the relationship between stigma and mental health was available, respectively.
and mental health and whether an operational Eighty-two studies met the above criteria. An
definition is provided for the theoretical constructs. additional 14 studies were included through the
Publication bias assesses external validity or the reference lists of the identified articles. Key authors
extent to which the results can be generalized to the in the field were contacted for additional unpub-
population. Design evaluates internal validity, or lished work. However, no authors provided us with
the extent to which certain threats may system- additional studies.
atically bias the results. Sources of heterogeneity The 96 studies which have met the inclusion
refer to statistical conclusion validity or attention to criteria were further examined to determine whether
types of conditions, settings, and respondents that necessary information was available. Five articles
may produce random heterogeneity. were duplicates (e.g., a dissertation that was
subsequently published or two articles based on
Method the same sample) and so were excluded. Nineteen
articles were excluded because access to the full-text
Rules for inclusion in the meta-analysis was not available. An additional 23 studies were
excluded because either the studies did not have
In order to identify studies containing informa- quantitative measures for stigma or mental health
tion about the relation between stigma and mental or statistical information was inadequate for coding
health, studies of all types spanning the years 1985 even after authors were contacted for clarification.
to January 2005, including journal articles, book Forty-two articles and 7 dissertations were retained
chapters, and dissertations, were searched and through the process. Dissertations were obtained
identified in PsycINFO and PubMed. These two from the ProQuest Electronic Dissertation Data-
databases were chosen because they cover journals base that contained full-text dissertations completed
across relevant disciplines. Specifically, PsycInfo has only after 1997. In the 49 studies, two studies
 ARTICLE IN PRESS
248 W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

contained several independent samples reporting negative mental health indicators (34 correlations),
results separately for different ethnic groups. Those were also generated for analyses.
reported correlations were separately recorded as
independent samples. In sum, 49 studies containing Meta-analytic procedures
52 independent samples were included in the
analyses. This sample of studies was representative The meta-analytic procedures of Hunter and
of various stigmatized conditions (see Table 1). Schmidt (2004) were adopted using the Hunter–
References and details of the included articles are Schmidt Meta-Analysis Programsr (Schmidt &
available upon request to the first author. Each of Huy, 2004) to correct the observed mean correla-
the independent samples was carefully coded tions for artifacts, due to sampling error and
separately by two coders and their inter-rater unreliability in stigma and mental health measures.
reliability was satisfactory. Because most of the selected studies did not provide
Stigma–mental health correlations were obtained all the information required to correct for all
either directly from the articles, or through trans- artifacts, a two-step meta-analysis using artifact
forming relevant statistics (e.g., standardized regres- distributions was conducted. A meta-analysis was
sion coefficients) reported in the articles. Authors of first performed, generating a mean correlation
the studies for which data were missing were corrected for the individually known artifacts and
contacted for additional information. Many of the sampling error. Results from the first meta-analysis
studies reported several measures of stigma or were then corrected for the average attenuation
mental health and thus reported several correlations factor, which was derived from the means of
between those measures. Counting those correla- separate attenuation factors. This final outcome
tions from one single study as separate entries was the mean and standard deviation of true score
would violate statistical independence; instead, a correlation.
composite correlation was calculated for each study. Besides obtaining the mean true score correlation,
Using a conservative but widely accepted method it is also important to describe the variability
(DeCoster, 2004; Rosenthal & Rubin, 1986), one in the correlations before concluding about the
score of predictor (stigma) reliability, criterion strength of relationship. The 80% credibility inter-
(mental health) reliability, and correlation was vals (80% CV), an estimate of the variability of the
obtained by averaging across the scores of the individual correlations across studies, were re-
several measures reported for each study. A total of ported. The confidence intervals (95% CI) repre-
52 composite correlations were generated for the senting the variability of the estimated mean true
analyses. Another two sets of composite correla- correlations in the population were also reported.
tions, between stigma and positive mental health The 80% CV was generated from the Hunter–
indicators (31 correlations) and between stigma and Schmidt Meta-Analysis Programsr (Schmidt &
Huy, 2004), and the 95% CI was calculated based
on a formula suggested by Hunter and Schmidt
Table 1 (2004).
Types of stigmatized conditions included in the meta-analysis
In terms of assessment study quality, Table 2 lists
Type of stigma Number of studies the aspects coded in each dimension and their
descriptions. With regard to the theory dimension,
Mental disorder 19 the focus of the study and explanation of the
Sexual orientation 6
HIV/AIDS 6
theoretical constructs of stigma and mental health
Intellectual/learning disability 4 were coded. To examine the possibility of publica-
Medical disorder 3 tion bias (i.e., the file-drawer problem), wherein
Physical disability 3 statistically significant results are more likely to be
Gender 2 published than non-significant results (Scargle,
Race/ethnicity 2
Child sexual abuse 1
2000), the following aspects were coded: review
Obesity 1 process, publication type, publication date, and
Unemployed 1 impact factor. Initially, retrieval method (e.g., based
Others (e.g. mixed stigmatized conditions) 1 on literature database, contacts of field experts, and
Total 49 references from other studies) was included. How-
ever, because studies identified by the field experts
 ARTICLE IN PRESS
W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261 249

Table 2
Variables in assessment of study quality

Dimensions of quality assessment Components within quality dimensions and their operational definitions

Theory
Focus of the study Whether the relationship between stigma and mental health is the focus of the study

 Main focus
 Auxiliary
Explanation of the theoretical constructs of Whether the constructs are well-defined in the study
stigma and mental health
 Well-defined
 Unclear or not defined

Publication bias
Type of review process Whether the article has undergone peer review

 Peer-reviewed
 Not peer-reviewed

Publication type What type of publication is the article?

 Journal article
 Book or book chapter
 Doctoral dissertation
Publication date When is the article published?

 Before 2000
 2000 and later
Journal impact factor What is the impact factor of the journal in which the article is published?

 0–1.0
 1.1–2.0
 2.1 or above

Design
Sampling What is method by which the participants are obtained for the study?

 Randomized sample of target groups

 Combination of random sampling and convenience sampling
 Convenience sampling of target groups
Causality Whether causation between stigma and mental health can be determined or inferred in
the study

 Experimental study
 Correlational—longitudinal study
 Correlational—cross-sectional study

Measures used in the study What is the nature of measures that are used in the study?

 Established measures
 Combination of established and self-developed measures
 Self-developed measures
 ARTICLE IN PRESS
250 W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

Table 2 (continued )

Dimensions of quality assessment Components within quality dimensions and their operational definitions

Reporting of statistics Whether bivariate correlation is explicitly reported or the statistic has to be converted
from other statistics

 Reported
 Converted

Sources of heterogeneity
Region of study Region from which the study was conducted

 North America
 Europe and Australia
 Asia
Type of stigmatized condition Category of stigmatized conditions to which the target sample of the study belong

 Physical—related to medical diseases, physical disabilities, and other physical

conditions
 Mental—related to psychological disorders and intellectual disabilities
 Social—related to social categories such as gender, ethnicity, sexual orientation
Major types of stigma Major categories of stigmatized conditions found in the literature

 Stigma towards mental illness

 Stigma towards sexual orientation
 Stigma towards HIV/AIDS

Concealability of stigma Whether the stigmatized condition is concealable or not

 Concealable—the condition is not directly apparent from view

 Conspicuous—the condition can be directly observed

Table 3 tized conditions were included. Rather than

Measures of stigma included in the meta-analysis combining quality assessment components into
Stigma measure Frequency
global indices of quality, moderator analyses were
conducted separately for each quality assessment
Devaluation-Discrimination scale 9 component. This method was preferred so the
Szivos-Bach Stigma scale 3 effects of individual components of quality assess-
Perceived Social Stigma scale 2
ment could be directly observed and no arbitrary
Demi Stigma scales 2
Explanatory Model Interview Catalogue 2 weighting of individual components was necessary
Other stigma scales 31 to compute the summary scores (Khan, ter Riet,
Popay, Nixon, & Kleijnen, 2001; Wortman, 1994).

and reference lists were already identified in the Results

database, this category was dropped from analysis.
As to design, we assessed sampling, causality The instruments measuring stigma and mental
between stigma and mental health, quality of health were presented in Tables 3 and 4. For stigma
measures used, and reporting of statistical findings. measures, one-fourth of the studies adopted or
Finally, with reference to sources of heterogeneity, modified Link’s (1987) scale of Devaluation-Dis-
the region where the study was conducted, stigma- crimination; other studies employed a variety of
tized conditions, and concealability of the stigma- scales such as Szivos-Bach (1993) Stigma scale for
 ARTICLE IN PRESS
W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261 251

individuals with intellectual disability and the Overall analyses

Perceived Social Stigma Scale (Rybarczyk et al.,
1992). For mental health measures, scales such as Three sets of meta-analysis were performed
the Center for Epidemiologic Studies-Depression between stigma and: (1) a mental health composite
Scale (Radloff, 1977), General Health Question- averaging all positive and negative indicators of
naire (Goldberg, 1987), Rosenberg self-esteem scale mental health, (2) a composite averaging all positive
(Rosenberg, 1965) and Quality of Life Interview indicators of mental health, and (3) a composite
(Lehman, 1983) were used. averaging all negative indicators of mental health
(see Table 5). For ease of comparison, all correla-
tions were reported in one direction with higher
Table 4 correlations representing stronger relationships be-
Measures of mental health included in the meta-analysis tween stigma and (good) mental health. For the
Scale Frequency
overall average analysis, the sample size weighted
mean correlation was .22. The estimated popula-
Positive Negative tion correlation corrected for sampling error,
indicators indicators unreliability and other artifacts was .28. Accord-
Self-esteem measures ing to Cohen (1992), it was a medium correlational
Szivos-Bach Self-esteem scale 2 effect size, which could be recognized in daily life.
Rosenberg Self-esteem scale 10 The 80% CV excluded zero (ranged .57 to .01),
Others 5 indicating a substantial variability in the individual
Psychotic symptoms and depression measures correlations across studies. The 95% CI (ranged
Center for Epidemiologic Studies 9 .29 to .26) suggested that the average relation
Depression between stigma and mental health was non-zero.
Beck Depression Inventory 3
Results with positive and negative indicators were
Hopkins Symptom Checklist 2
Brief Symptom Inventory 1 similar to the overall analysis. For the positive
Others 7 indicator analysis, population correlation was .34.
The 80% CV had a large variability (ranged .56 to
Mental health measures
General Health Questionnaire 2 .11) and the 95% CI excluded zero (ranged .35
Others 2 to .33). For the negative indicator analysis, the
population correlation was .28. The 80% CV had
Quality of life measures 7
Well-being measures 4 a large variability (ranged .58 to .02) and the 95%
Life satisfaction measures 3 CI (ranged .30 to .26) excluded zero.
Psychological distress 2
Happiness measures 2 Moderator analyses
Mastery measures 2
Loneliness measures 2
Burden measures 2 As observed in the overall analyses, the credibility
Other measures 4 6 intervals were relatively wide, indicating substantial
variation in the individual correlations among the

Table 5
Summary of stigma and mental health correlationsa

k N rw SDw p SDp 80% CV 95% CI % variance

Lower Upper Lower Upper

Averaged mental health score 52 10567 .22 .18 .28 .23 .57 .01 .29 .26 12.97
Averaged positive indicators 31 5995 .26 .14 .34 .17 .56 .11 .35 .33 21.50
Averaged negative indicators 34 7562 .23 .19 .28 .23 .58 .02 .30 .26 10.76

Note: k ¼ number of correlations; N ¼ total sample size; r ¼ uncorrected mean observed correlation; SDr ¼ standard deviation of
uncorrected mean correlation; rw ¼ sample size weighted mean observed correlation; SDw ¼ standard deviation of sample size weighted
mean observed correlation; p ¼ estimated true score correlation; SDp ¼ standard deviation of estimated true score correlation; 80%
CV ¼ 80% credibility interval; 90% CI ¼ 90% confidence interval; % variance ¼ percentage of the variance in correlations attributable to
all artifacts.
 ARTICLE IN PRESS
252 W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

meta-analyzed studies. Sampling error and measure- correlation between stigma and mental health than
ment error accounted for only 13%, 22%, and 11% did those that used random sampling (p ¼ .02)
of the variation in average mental health, positive and longitudinal design (p ¼ .12). Similar patterns
indicators, and negative indicators, respectively, were observed for negative and positive indicators.
leaving a great proportion of the variation unex- Nevertheless, the number of studies that used
plained. To examine sources of this variation, random sampling and longitudinal design was very
moderator analyses based on various components small.
in quality assessment were conducted. The overall
results and the results of positive and negative Sources of heterogeneity
indicator analyses are presented in Tables 6, 7 and With regard to regions, the true score correlation
8, respectively. was the lowest in North America (p ¼ .29).
Comparatively, studies in Europe and Australia
Theory (p ¼ .48) and Asia (p ¼ .37) reported higher true
Findings showed that the stigma–mental health score correlations between stigma and mental
relation across three indicators was slightly stronger health, with a smaller within population variability.
in studies that did not specifically focus on stigma This pattern was also observed using negative
and mental health or have these two constructs well indicators but more attenuated with positive in-
defined. dicators. Although there were far fewer European,
Australian and Asian studies than North American
Publication bias ones, the former included studies of different
The stigma–mental health relation was much stigmatized conditions.
stronger in peer-reviewed articles (p ¼ .35) than To further analyze how the stigmatized condi-
in non-peer-reviewed articles in which the relation tions may affect the stigma–mental health relation,
was reduced to near zero (p ¼ .08). Similar we grouped them in different ways. First, we divided
findings were observed for publication type. Upon the studies into three groups of stigma, namely,
closer examination, this difference was observed physical, social, and mental. Results showed that
only in negative indicators but not in positive across the three indicators, the mental stigma group
indicators. reported the weakest stigma–mental health relation,
Findings on publication date showed that studies followed by the social stigma group, and with the
published since the year 2000 had a slightly higher strongest in the physical stigma group.
stigma–mental health correlation than the earlier We then extracted for finer analysis the most
studies. Again, this difference was more apparent commonly reported types of stigma from each
among negative indicators than among positive category, namely, stigma towards HIV+/AIDS
indicators. Finally, we examined whether studies patients, stigma towards mental illness, and stigma
reporting a stronger relation between stigma and towards one’s sexual orientation. Results for
mental health would concentrate in journals with a average mental health and negative indicators
higher impact factor. We used impact factors to suggested a strong effect on HIV+/AIDS patients,
classify the studies into three groups at arbitrary a moderate effect on sexual orientation, and a small
cut-offs: impact factor 0.1–1.1; impact factor effect on mental illness. In regards to positive
1.1–2.0 and impact factor 2.1 or above. The results indicators, the relationship was stronger on mental
showed that true score correlations of the three illness (p ¼ .22) than on sexual orientation
groups were similar, suggesting that a study’s (p ¼ .06) but the number of studies was very
reported strength of stigma–mental health relation small.
was not crucial to the journals’ decision of Lastly, we grouped the studies by concealability,
acceptance. or the extent to which the condition is apparent
from view. We classified such conditions as HIV+/
Design AIDS and sexual orientation as concealable, and
Whereas the choice of measures and reporting obesity and physical disability as conspicuous.
practices of statistics did not make a difference to HIV+/AIDS were classified as concealable because
the relationships, studies that used convenience most of the samples in this category were asympto-
sampling (p ¼ .34) and a cross-sectional design matic. Stigmatized conditions of ambiguous con-
(p ¼ .38) tended to show a stronger true score cealability (e.g. intellectual disability which depends
Table 6
Moderator analyses of study quality components on average mental health scorea

k N rw SDw p SDp 80% CV 95% CI % variance

W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

Lower Upper Lower Upper

Focus of the study

Main focus 37 7804 .20 .19 .25 .23 .54 .04 .26 .23 11.76
Auxiliary 15 2763 .29 .15 .38 .19 .62 .13 .39 .36 23.12

ARTICLE IN PRESS
Explanation of constructs
Well-defined 21 5727 .18 .16 .22 .20 .48 .04 .24 .21 11.90
Unclear or not defined 30 4576 .27 .19 .33 .23 .63 .04 .35 .32 16.26
Type of review process
Peer-reviewed 44 7720 .28 .16 .35 .19 .60 .10 .36 .34 18.48
Not peer-reviewed 8 2847 .06 .14 .08 .18 .30 .15 .10 .06 12.97
Publication type
Journal article 44 7720 .28 .16 .35 .19 .60 .10 .36 .34 18.48
Book or book chapter 1 741 — — — — — — — — —
Doctoral dissertation 7 2106 .03 .15 .04 .20 .29 .21 .07 .02 12.86
Publication date
Before 2000 16 4028 .15 .19 .19 .24 .50 .11 .22 .17 10.10
2000 and later 31 5469 .26 .17 .33 .21 .59 .06 .34 .31 16.6
Journal impact factor
0–1.0 7 846 .28 .11 .34 .13 .50 .17 .35 .32 37.34
1.1–2.0 17 3319 .24 .15 .30 .18 .53 .07 .32 .29 22.35
2.1 or above 7 2130 .31 .18 .39 .22 .67 .10 .42 .35 8.32
Sampling
Random sampling 4 1830 .016 .17 .02 .22 .31 .27 .06 .018 7.30
Random and convenience sampling 4 2026 .22 .10 .28 .13 .45 .11 .30 .27 14.92
Convenience sampling 44 6711 .28 .16 .34 .20 .59 .09 .35 .33 19.89
Causality
Experimental 0 0 — — — — — — — — —
Correlational—longitudinal 9 4313 .09 .15 .12 .19 .36 .13 .14 .10 8.19
Correlational—cross-sectional 42 6174 .14 .17 25.35

253
.31 .38 .60 .16 .40 .36
 254
Table 6 (continued )

k N rw SDw p SDp 80% CV 95% CI % variance

Lower Upper Lower Upper

Measures
Established measures 18 2130 .23 .18 .28 .21 .55 .01 .30 .26 20.11
Established & self-developed measures 34 8507 .22 .20 .27 .25 .59 .04 .29 .26 9.49
Self-developed measures 0 0 — — — — — — — — —
Reporting of statistics

W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

Reported 32 5672 .21 .20 .26 .25 .58 .06 .28 .24 12.19
Converted 20 4895 .24 .16 .30 .19 .54 .05 .31 .28 14.61
Region of study
North America 40 8466 .23 .17 .29 .21 .56 .03 .30 .28 15.24
Europe and Australia 7 456 .39 .13 .48 .16 .69 .28 .51 .46 41.27
Asia 2 425 .30 .13 .37 .16 .57 .16 .40 .33 18.29

ARTICLE IN PRESS
Type of stigmatized condition
Physical 14 1818 .37 .17 .44 .20 .71 .18 .48 .45 17.90
Mental 18 3982 .09 .19 .12 .23 .42 .18 .14 .09 11.70
Social 19 4704 .26 .11 .33 .13 .50 .16 .34 .32 28.18
Major types of stigma
Stigma towards mental illness 15 3817 .08 .17 .10 .22 .38 .18 .12 .08 11.53
Stigma towards sexual orientation 6 2430 .23 .06 .28 .07 .37 .19 .29 .27 42.39
Stigma towards HIV/AIDS 5 791 .43 .09 .53 .10 .66 .39 .54 .51 37.64
Concealability of stigma
Concealable 12 3356 .28 .11 .34 .13 .50 .18 .35 .33 22.34
Conspicuous 6 1356 .32 .11 .41 .14 .59 .23 .43 .39 26.09

Note: k ¼ number of correlations; N ¼ total sample size; r ¼ uncorrected mean observed correlation; SDr ¼ standard deviation of uncorrected mean correlation; rw ¼ sample size
weighted mean observed correlation; SDw ¼ standard deviation of sample size weighted mean observed correlation; p ¼ estimated true score correlation; SDp ¼ standard deviation of
estimated true score correlation; 80% CV ¼ 80% credibility interval; 90% CI ¼ 90% confidence interval; % variance ¼ percentage of the variance in correlations attributable to all
artifacts.
 ARTICLE IN PRESS
W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261 255

Table 7
Moderator analyses of study quality components on positive indicatorsa

k N rw SDw p SDp 80% CV 95% CI % variance

Lower Upper Lower Upper

Focus of the study

Main focus 19 3483 .24 .12 .30 .14 .48 .11 .31 .29 26.62
Auxiliary 11 1600 .29 .21 .43 .27 .78 .82 .47 .39 31.42
Explanation of constructs
Well defined 10 2485 .23 .11 .28 .14 .46 .11 .30 .27 22.20
Unclear or not defined 20 2598 .28 .18 .35 .22 .63 .07 .37 .33 23.15
Type of review process
Peer reviewed 26 4584 .25 .16 .32 .20 .57 .07 .33 .30 20.11
Not peer reviewed 2 317 .30 .06 .35 .07 .44 .26 .37 .34 59.19

Publication type
Journal article 26 4584 .25 .16 .32 .19 .56 .07 .33 .30 20.14
Book or book chapter — — — — — — — — — — —
Doctoral dissertation 2 317 .30 .06 .35 .07 .44 .26 .37 .34 59.19
Publication date
Before 2000 10 2432 .22 .11 .27 .13 .44 .10 .28 .26 25.09
2000 and later 21 3476 .30 .16 .39 .19 .63 .14 .40 .37 22.05

Journal impact factor

0–1.0 8 846 .37 .17 .43 .20 .69 .18 .46 .41 21.10
1.1–2.0 10 2343 .21 .15 .27 .18 .50 .03 .29 .25 19.24
2.1 or above 6 1834 .28 .12 .36 .15 .55 .17 .38 .34 18.40
Sampling
Random sampling — — — — — — — — — — —
Random and convenience sampling 2 373 .19 .07 .23 .08 .33 .12 .24 .21 50.79
Convenience sampling 27 4384 .28 .15 .35 .18 .58 .11 .36 .33 24.00
Causality
Experimental — — — — — — — — — — —
Correlational—longitudinal 4 1678 .21 .06 .26 .08 .36 .16 .27 .25 36.33
Correlational—cross-sectional 26 3405 .28 .18 .35 .21 .63 .08 .37 .33 22.54
Measures
Established measures 11 1500 .20 .12 .24 .14 .42 .05 .25 .22 31.40
Established and self-developed measures 19 3583 .28 .16 .36 .19 .31 .12 .38 .34 22.00
Self-developed measures — — — — — — — — — — —

Reporting of statistics
Reported 18 2835 .27 .14 .33 .16 .54 .12 .35 .32 23.42
Converted 12 2248 .24 .17 .29 .21 .56 .03 .32 .27 17.08
Region of study
North America 24 5030 .26 .14 .33 .17 .56 .11 .35 .32 20.89
Europe and Australia 6 609 .33 .17 .39 .20 .65 .14 .43 .36 22.57
Asia 1 — — — — — — — — — —

Type of stigmatized condition

Physical 8 940 .38 .08 .46 .09 .57 .34 .47 .45 56.58
Mental 12 1993 .20 .15 .24 .18 .47 .01 .26 .22 19.89
Social 11 3062 .27 .13 .36 .16 .57 .16 .38 .35 24.44
Major types of stigma
Stigma towards mental illness 10 1912 .19 .14 .22 .17 .44 .01 .24 .20 19.90
Stigma towards sexual orientation 3 1385 .05 .02 .06 .03 .09 .02 .06 .05 81.93
Stigma towards HIV/AIDS 0 — — — — — — — — — —
 ARTICLE IN PRESS
256 W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

Table 7 (continued )

k N rw SDw p SDp 80% CV 95% CI % variance

Lower Upper Lower Upper

Concealability of stigma
Concealable 7 1993 .33 .10 .40 .12 .56 .24 .41 .39 21.45
Conspicuous 5 1244 .31 .02 .41 .01 .41 .40 .41 .40 99.56

Note. k ¼ Number of correlations; N ¼ total sample size; r ¼ uncorrected mean observed correlation; SDr ¼ standard deviation of
uncorrected mean correlation; rw ¼ sample size weighted mean observed correlation; SDw ¼ standard deviation of sample size weighted
mean observed correlation; p ¼ estimated true score correlation; SDp ¼ standard deviation of estimated true score correlation; 80%
CV ¼ 80% credibility interval; 90% CI ¼ 90% confidence interval; % variance ¼ percentage of the variance in correlations attributable to
all artifacts.

on its severity) were not included in this analysis. growth than an exacerbating effect on psychological
Small difference in true score correlations bet- distress.
ween conspicuous and concealable stigma was The present study did not identify any significant
found on mental health and negative indicators, moderators that can explain the relationship be-
whereas similar relationships were found on positive tween stigma and mental health. Different types of
indicators. stigma on different types of people could create very
It should be noted that although differences in different mental health outcomes. Studies have
true score correlations could be observed through shown the relationship between stigma and self-
within-moderator analyses, none of the moderators esteem is contingent upon situational contexts and
met the 75% rule as suggested by Hunter and collective representations of the label to the
Schmidt (2004). The 75% rule indicates that a individuals (Crocker, 1999). Corrigan and Watson
successful moderator should reduce the heterogene- (2002) also argued that not all stigmatized indivi-
ity of population, as evidenced by an increase in the duals go through the process of internalization.
percentage of variance due to all artifacts to about Some individuals may not be aware of stigma,
75%. As observed in our results, no moderators whereas others may be energized by the oppression
satisfied this rule, although an increase in the and fight against the odds. Only when stigmatized
percentage of variance due to all artifacts was individuals perceive the negative acts committed by
observed. Therefore, the comparisons made be- others towards them as legitimate would their
tween different groups in the moderator analyses mental health be adversely affected. If stigmatized
should be interpreted with caution. individuals view the negative acts as not legitimate,
they would either show righteous anger or feel
Discussion indifferent, depending on how much they identify
with the group. Therefore, it is not surprising that a
The present study integrated findings from large unexplained variance is left in the stigma–
empirical studies to establish the overall significance mental health relationship. Future research should
of stigma in relation to mental health. Based on our explore whether the variations come from psycho-
research synthesis, the relation between stigma and metric limitations or different mechanisms by which
mental health had a medium correlational effect stigma operates in different conditions.
size, which indicated that it is strong enough to be Although the moderators in the present study did
observed in everyday life. In other words, stigma not meet the statistical 75% rule, some interesting
does have an observable association with stigma- patterns in the relations between stigma and mental
tized groups’ mental health. Across stigmatized health are worthy of attention. The present findings
conditions, stigma was found to have a stronger showed a large difference in mean correlation
relationship with positive mental health indicators between peer-reviewed and non-peer-reviewed stu-
than with negative ones. Given mental health is dies and between journal articles and dissertations
not merely the absence of mental illness or distress, for mental health and negative indicators. In other
this pattern of relationships suggested that stigma words, the file-drawer problem may suggest a bias
has a stronger negative effect on adjustment and towards positive findings in the literature. Such a
Table 8
Moderator analyses of study quality components on negative indicatorsa

k N rw SDw p SDp 80% CV 95% CI % variance

W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

Lower Upper Lower Upper

Focus of the study

Main focus 28 5952 .20 .21 .24 .25 .56 .07 .27 .22 9.96
Auxiliary 6 1610 .32 .05 .41 .07 .49 .32 .41 .40 53.14

ARTICLE IN PRESS
Explanation of constructs
Well-defined 19 4818 .17 .18 .21 .22 .49 .07 .23 .19 10.66
Unclear or not defined 15 2636 .31 .17 .39 .21 .65 .12 .41 .37 15.55
Type of review process
Peer-reviewed 28 5050 .33 .11 .40 .13 .57 .23 .41 .39 27.86
Not peer-reviewed 6 2585 .02 .14 .03 .16 .24 .18 .01 .05 10.39
Publication type
Journal article 28 4977 .33 .11 .40 .14 .58 .23 .41 .40 28.25
Book or book chapter 1 741 — — — — — — — — —
Doctoral dissertation 5 1844 .03 .14 .03 .19 .21 .27 .01 .06 11.78
Publication date
Before 2000 14 3308 .14 .20 .17 .26 .50 .15 .21 .14 9.40
2000 and later 20 4254 .30 .15 .36 .17 .58 .13 .37 .34 16.27
Journal impact factor
0-1.0 10 1091 .31 .11 .36 .07 .45 .27 .37 .35 69.64
1.1–2.0 9 1513 .28 .14 .33 .16 .54 .12 .35 .31 21.20
2.1 or above 5 1636 .35 .08 .46 .11 .59 .32 .47 .44 29.41
Sampling
Random sampling 4 1830 .01 .17 .02 .22 .31 .27 .06 .02 7.32
Random and convenience sampling 3 1746 .22 .17 .27 .20 .53 .01 .31 .23 5.77
Convenience sampling 27 3986 .32 .12 .39 .14 .56 .21 .39 .38 30.18
Causality
Experimental — — — — — — — — — — —
Correlational—longitudinal 7 3325 .07 .14 .09 .17 .31 .13 .11 .07 9.80
Correlational—cross-sectional 27 4236 .13 .16 24.38

257
.34 .42 .62 .22 .43 .41
 258
Table 8 (continued )

k N rw SDw p SDp 80% CV 95% CI % variance

Lower Upper Lower Upper

Measures
Established measures 10 754 .25 .07 .31 .32 .72 .10 .37 .24 14.96
Established and self-developed measures 24 6808 .22 .18 .27 .22 .56 .01 .29 .26 9.73
Self-developed measures — — — — — — — — — — —
Reporting of statistics

W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

Reported 21 3709 .18 .24 .23 .30 .62 .15 .27 .20 8.80
Converted 13 3853 .26 .11 .32 .13 .48 .15 .33 .31 21.19
Region of study
North America 27 6836 .21 .19 .26 .23 .56 .03 .28 .24 9.80
Europe and Australia 4 221 .28 .16 .36 .21 .63 .09 .42 .31 37.63
Asia 2 425 .38 .14 .46 .17 .67 .25 .50 .42 15.67

ARTICLE IN PRESS
Type of stigmatized condition
Physical 8 1002 .38 .25 .48 .30 .87 .09 .54 .43 10.89
Mental 10 2827 .08 .17 .11 .22 .38 .18 .13 .08 10.54
Social 16 3733 .29 .09 .35 .11 .49 .22 .36 .34 32.59
Major types of stigma
Stigma towards mental illness 8 2712 .10 .14 .13 .18 .36 .10 .15 .11 12.72
Stigma towards sexual orientation 5 1605 .20 .08 .23 .09 .34 .12 .24 .22 33.19
Stigma towards HIV/AIDS 5 791 .43 .09 .53 .10 .66 .39 .54 .51 37.64
Concealability of stigma
Concealable 10 2395 .30 .12 .34 .13 .51 .17 .35 .33 21.30
Conspicuous 4 1168 .31 .21 .56 .35 -1.02 .11 .64 .48 12.74

Note: k ¼ number of correlations; N ¼ total sample size; r ¼ uncorrected mean observed correlation; SDr ¼ standard deviation of uncorrected mean correlation; rw ¼ sample size
weighted mean observed correlation; SDw ¼ standard deviation of sample size weighted mean observed correlation; p ¼ estimated true score correlation; SDp ¼ standard deviation of
estimated true score correlation; 80% CV ¼ 80% credibility interval; 90% CI ¼ 90% confidence interval; % variance ¼ percentage of the variance in correlations attributable to all
artifacts.
 ARTICLE IN PRESS
W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261 259

bias might lead us to overestimate the association sufficient for research synthesis, the number re-
between stigma and psychological distress. Another mained small for moderator analyses, which may
interesting finding is the stronger stigma-mental increase sampling error in subgroup analyses.
health relationship observed among studies that do Moreover, diverse measures were used to assess
not explicitly focus on stigma and mental health, stigma and mental health across the meta-analyzed
that do not have the constructs clearly defined, as studies. Except for the Devaluation-Discrimination
well as those that used convenience sampling and Scale that has been used in nine studies to measure
cross-sectional design. Unfortunately, these studies stigma, many studies employed self-developed scales
made up the majority of literature reviewed; because of limited validation evidence (Link, Yang, Phelan,
of their relatively weak internal validity, caution & Collins, 2004). Although moderator analyses did
should be used in interpreting the findings. not show a difference in the strength of the
Findings also showed a trend for the stigma- stigma–mental health relationship between estab-
mental health relation to be stronger in Europe and lished and self-developed measures, for the con-
Australia than in North America. Future studies solidation and accumulation of research findings on
may examine closely the situational effects of stigma stigma, unified stigma scales should be developed
on mental health. Perhaps the difference is due to and validated, so that future studies can utilize
the varying institutional and societal efforts that measures that are consistent in conceptualization
different regions have put into stigma reduction. and operationalization. With a greater number of
Because an overwhelming majority of the studies studies accumulated, researchers can also perform
were conducted in North America, a fine-grained separate analysis on measurements in future meta-
assessment of this regional effect was not possible in analysis.
meta-analysis. Similarly, different types of mental health mea-
Differential effect sizes between stigmatized condi- sures were included in our meta-analysis. Although
tions were observed in the present analysis. Com- we have already divided the measures into positive
pared to physical and social conditions, stigma and negative indicators, analyses with these compo-
related to mental conditions has a smaller mean site scores can only provide us with a general idea
correlation with mental health. This may seem a bit about how stigma is related to mental health. Given
surprising; however, upon closer examination, the the relation between stigma and mental health states
difference may be due to the large effect size observed may vary, analysis of specific measures is important.
in studies with HIV/AIDS condition, which make up However, the sample size of mental health sub-
one-third of the studies in the physical condition. groups is too small to allow for separate analyses.
Studies have found AIDS stigma to be the greatest We suggest that future research in stigma and
among infectious diseases and medical conditions mental health use more comprehensive positive and
such as genital herpes, hepatitis, tuberculosis, drug negative mental health indicators, so that results can
abuse, and cancer (Crawford, 1996; Mak et al., 2006). provide within-study validation as well as between-
It may be conjectured that the higher the level of study validation in future meta-analytic summaries.
stigma, the stronger the impact it may have on one’s The present meta-analytic study focused on
mental health. Another explanation of this trend is bivariate correlational data. Due to the cross-
that the condition of mental illness itself may have sectional nature of data, causal inference should
already taken up a great deal of variance in mental not be drawn between stigma and mental health.
health outcomes. Thus, stigma is only secondary to Meta-analysis on stigma studies that utilized experi-
mental illness in affecting well-being. Furthermore, mental design is necessary to establish the causal
due to the impairment inherent in severe psychiatric link between stigma and the various indices of
disorders, individuals who are subjected to strong mental health. None of the meta-analyzed studies
stigmatization may lack the cognitive insight to be used an experimental design. Finally, the present
aware of stigmatization targeted to them (Corrigan & meta-analysis focused on studies written in English.
Watson, 2002; Mak & Wu, 2006). Although the two databases used have a worldwide
coverage and study region was included as a
Limitations of study moderator, future meta-analytic attempts should
include non-English articles so that the relationship
Several limitations must be born in mind. between stigma and mental health can be examined
Although the number of studies identified was cross-culturally.
 ARTICLE IN PRESS
260 W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261

Notwithstanding the limitations, the present Goffman, E. (1963). Stigma: Notes on the management of spoiled
study provided support for the importance of identity. New York: Simon & Schuster Inc.
stigma in relation to stigmatized groups’ mental Goldberg, D. P. (1987). Manual for General Health Questionnaire.
London: NFER-Nelson.
health. Efforts to combat stigma are worthy, as the Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis:
mitigation of stigma may enhance people’s well- Correcting error and bias in research findings. Sage: Thousand
being. We believe that summarizing research find- Oaks.
ings at this point is important for researchers and Khan, K. S., ter Riet, G., Popay, J., Nixon, J., & Kleijnen (2001).
policy makers to take stock of our knowledge and Study quality assessment. In CRD Report #4 (2nd Ed.),
Undertaking systematic reviews of research on effectiveness:
understanding of the stigma–mental health relation- CRD’s guidance for those carrying out or commissioning
ship and to chart our directions for future research reviews. York, UK: Centre for Reviews and Dissemination,
and policy making. University of York.
Lau, T. F., & Cheung, C. K. (1999). Discriminatory attitudes to
people with intellectual disability or mental health difficulty.
Acknowledgments International Social Work, 42, 431–444.
Lee, S., Lee, M. T. Y., Chiu, M., & Kleinman, A. (2005).
Experience of social stigma by people with schizophrenia in
We would like to thank Anna Ho, Gladys Ho, Hong Kong. British Journal of Psychiatry, 186, 153–157.
Venus Yiu, and Bauhinia Yong for their involve- Lehman, A. F. (1983). The well-being of chronic mental patients:
ment in the initial screening of the research studies, Assessing their quality of life. Archives of General Psychiatry,
Ivy Ng for her involvement in coding of the 40, 369–373.
qualified research studies, and Rebecca Cheung for Link, B. G. (1987). Understanding labeling effects in the area of
mental disorders: An assessment of the effects of expectations
her assistance in calculations. Special thanks go to of rejection. American Sociological Review, 52, 96–112.
Dianne van Hemert for her advice on statistical Link, B. G., & Phelan, J. C. (2001). Conceptualizing stigma.
analysis, and to all the authors who have provided Annual Review of Sociology, 27, 363–385.
additional statistical information to us. Link, B. G., Yang, L. H., Phelan, J. C., & Collins, P. Y. (2004).
Measuring mental illness stigma. Schizophrenia Bulletin,
30(3), 511–541.
Mak, W. W. S., Mo, P. K. H., Cheung, R. Y. M., Woo, J.,
References Cheung, F., & Lee, D. (2006). Comparative stigma of HIV/
AIDS, SARS, and TB in Hong Kong. Social Science &
Cohen, J. (1992). A power primer. Psychological Bulletin, 112, Medicine, 63(7), 1912–1922.
155–159. Mak, W. W. S., & Wu, C. F. M. (2006). Cognitive insight and
Corrigan, P. W. (2000). Mental health stigma as social attribu- casual attribution in the development of self-stigma among
tion: Implications for research methods and attitude change. individuals with schizophrenia. Psychiatric Services, 57(12),
Clinical Psychology: Science and Practice, 7(1), 48–67. 1800–1802.
Corrigan, P. W., & Kleinlein, P. (2005). The impact of mental Moher, D., Cook, D. J., Eastwood, S., Olkin, I., Rennie, D., &
illness stigma. In P. W. Corrigan (Ed.), On the stigma of Stroup, D. F. (1999). Improving the quality of reports of
mental illness: Practical strategies for research and social meta-analyses of randomized controlled trials: The QUOR-
change (pp. 11–44). Washington, DC: American Psychologi- OM statement. The Lancet, 354, 1896–1900.
cal Association. Radloff, L. S. (1977). The CES-D scale: A self report depression
Corrigan, P. W., & Watson, A. C. (2002). The paradox of self- scale for research in the general population. Applied
stigma and mental illness. Clinical Psychology: Science and Psychological Measurement, 1, 385–401.
Practice, 9, 35–53. Rosenberg, M. (1965). Society and the adolescent self-image.
Crawford, A. M. (1996). Stigma associated with AIDS: A meta- Princeton, NJ: Princeton University Press.
analysis. Journal of Applied Social Psychology, 26(5), 398–416. Rosenthal, R., & Rubin, D. B. (1986). Meta-analytic procedures
Crocker, J. (1999). Social stigma and self-esteem: Situational for combining studies with multiple effect sizes. Psychological
construction of self-worth. Journal of Experimental Social Bulletin, 99, 400–406.
Psychology, 35, 89–107. Rybarczyk, B., Gallagher-Thompson, D., Rodman, J., Zeiss, A.,
Crocker, J., & Major, B. (1989). Social stigma and self-esteem: et al. (1992). Applying cognitive–behavioral psychotherapy to
The self-protective properties of stigma. Psychological Review, the chronically ill elderly: Treatment issues and case illustra-
96, 608–630. tion. International Psychogeriatrics, 4, 127–140.
Crocker, J., Major, B., & Steele, C. (1998). Social stigma. In D. T. Scargle, J. D. (2000). Publication bias: The ‘‘file-drawer’’ problem
Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of in scientific inference. Journal of Scientific Exploration, 14(1),
social psychology, Vol. 2 (pp. 504–553). New York: McGraw- 91–106.
Hill. Schmidt F. L., & Huy, A. L. (2004). Hunter– Schmidt meta-
DeCoster, J. (19 September 2004). Meta-analysis notes. Retrieved analysis programs [Computer software].
8 August 2005, from /http://www.stat-help.com/meta.pdfS. Szivos-Bach, S. E. (1993). Social comparisons, stigma and
Glass, G. V. (1976). Primary, secondary and meta-analysis of mainstreaming: The self esteem of young adults with a mild
research. Educational Researcher, 5, 3–8. mental handicap. Mental Handicap Research, 6, 217–238.
 ARTICLE IN PRESS
W.W.S. Mak et al. / Social Science & Medicine 65 (2007) 245–261 261

US Department of Health and Human Services. (1999). World Health Organization. (2001). The World Health Report
Mental health: A report of the Surgeon General. Rockville, 2001—Mental health: New understanding, new hope. Retrieved
MD: US Department of Health and Human 1 December 2005, from /http://www.who.int/whr/2001/en/
Services, Substance Abuse and Mental Health Services index.htmlS.
Administration, Center for Mental Health Services, Wortman, P. M. (1994). Judging research quality. In H. Cooper,
National Institutes of Health, National Institute of Mental & L. V. Hedges (Eds.), The handbook of research synthesis
Health. (pp. 97–109). New York: Russell Sage Foundation.