investigating socio-economic explanations for gender and ethnic inequalities in health
TRANSCRIPT
Social Science a Medicine 54 (2002) 693–706
Investigating socio-economic explanations for gender andethnic inequalities in health
Helen Cooper*
Methodology Group, Office for National Statistics, Titchfield, Fareham PO15 5RR, UK
Abstract
This paper examines inequalities in the self-reported health of men and women from white and minority ethnic
groups in the UK using representative data from the Health Survey for England, 1993–1996. The results show
substantially poorer health among all minority ethnic groups compared to whites of working-age. The absence of
gender inequality in health among white adults contrasts with higher morbidity for many minority ethnic women
compared to men in the same ethnic group. The analysis addresses whether socio-economic inequality is a potential
explanation for this pattern of health inequality using measures of educational level, employment status, occupational
social class and material deprivation. There are marked socio-economic differences according to gender and ethnic
group; high morbidity is concentrated among adults who are most socio-economically disadvantaged, notably
Pakistanis and Bangladeshis. Logistic regression analyses show that socio-economic inequality can account for a
sizeable proportion of the health disadvantage experienced by minority ethnic men and women, but gender inequality in
minority ethnic health remains after adjusting for socio-economic characteristics. Crown Copyright r 2002 Published
by Elsevier Science Ltd. All rights reserved.
Keywords: Gender; Ethnicity; Health inequality; Socio-economic
Introduction
The established finding of higher morbidity among
women than men (Nathanson, 1975, 1977; Verbrugge,
1979) was drawn into question in the 1990s by research
showing much smaller gender inequality in self-reported
health than hitherto reported. Macintyre, Hunt, a
Sweeting (1996) found no consistent disadvantage in
self-reported health for women relative to men and
argued that the typical characterisation of a female
‘excess’ in morbidity may over-simplify the relationship
between gender and health.
This paper argues that ethnicity is a neglected
dimension in comparative studies of gender and health
and that it is timely to assess the health of men and
women in terms of ethnic group, not least because
minority ethnic groups comprise a growing proportion
of the UK population. Census records show that
minority ethnic groups comprised 6.2 percent of the
population of England in 1991, of which 1 percent were
classified as Black Caribbean and 3 percent as ‘South
Asian’; a diverse ethnic category that includes Indian,
Pakistani and Bangladeshi adults (Owen, 1992).
British surveys have found considerable variation in
reported health among these ethnic groups (Nazroo,
1997; Fenton, Hughes, a Hine, 1995; Rudat, 1994).
Indians typically have a more ‘advantaged’ health
profile, not too dissimilar to white adults, but reported
morbidity is substantially higher for other minority
ethnic groupsFparticularly for Pakistanis and Bangla-
deshis. Nazroo (1997) found gender differences in health
among adults within the same ethnic group, reporting
that white, Black Caribbean, Indian, Pakistani and
Bangladeshi women interviewed in the Fourth National
Survey of Ethnic Minorities (FNS) were more likely
than men in these ethnic groups to rate their health as
‘fair’ or ‘poor’.*Tel.: +44-1392-733886; fax: +44-1329-841760.
E-mail address: [email protected] (H. Cooper).
0277-9536/02/$ - see front matter Crown Copyright r 2002 Published by Elsevier Science Ltd. All rights reserved.
PII: S 0 2 7 7 - 9 5 3 6 ( 0 1 ) 0 0 1 1 8 - 6
Whilst the socio-economic basis of health inequality
has been compared for men and women (Arber, 1997a;
Denton a Walters, 1999), there has been neglect of
potential ethnic differences in the nature of these
relationships. Although this neglect partly reflects the
lack of survey data with sufficiently large samples of
minority ethnic men and women, a general criticism
levelled at ethnic health research is an undue emphasis
on cultural or behavioural factors specific to minority
ethnic groups, rather than general explanations
grounded in the social and economic living conditions
of white and minority ethnic groups (Nazroo, 1998).
However, where British research has shown a strong
socio-economic basis to ethnic inequality in health (e.g.
Nazroo, 1997; Fenton et al., 1995), standardising for the
effects of gender in these analyses obscures differences
between men and women in the same ethnic group.
Connecting gender and ethnicity draws attention to
ethnic differences in the health and socio-economic
position of women and men, as well as gender inequality
within ethnic groups.
The analysis presented in this paper focuses on the
interplay between gender and ethnicity by first examin-
ing the extent and nature of health inequality among
men and women classified as; white, Black Caribbean,
Indian, Pakistani and Bangladeshi, and second, examin-
ing to what extent key measures of individual socio-
economic position can explain differences in the self-
reported health of these ‘gender and ethnic groups’.
Results are based on a pooled sample of working-age
adults (age 20–60) interviewed in 4 years of the Health
Survey for England (1993–1996). The next section
reviews research on gender, ethnicity and health and
considers the extent to which socio-economic inequality
may underlie these relationships.
Gender and ethnic inequality in health
Both gender and ethnicity are significant factors for
health. Higher reported morbidity among women than
men was a common finding in the 1970s (Nathanson,
1975; Verbrugge, 1979) and British minority ethnic
groups report substantially poorer health compared to
the UK average (Rudat, 1994). The Fourth National
Survey (FNS) represents the most comprehensive data
on British minority ethnic groups to date, showing
inequality in self-assessed health across and within
ethnic groups according to gender (Nazroo, 1997).
There is considerable debate about the interplay
between biological, genetic and social factors for the
health of men and women and members of minority
ethnic groups. Unequal social relations, characterised by
discrimination, exclusion and exploitation, are thought
to have profound consequences for the economic and
social well-being of gender and ethnic groups that may
ultimately be expressed as inequalities in health (Krie-
ger, 2000). Minority ethnic women in particular may be
exposed to adverse health consequences associated with
‘multiple discrimination’ on the grounds of gender and
ethnicity. Thus, it is important to consider the extent to
which the health of men and women from different
ethnic groups is mediated by inequalities in their socio-
economic position.
An extensive literature testifies to the fact that men
and women differ markedly in social roles within the
family and that gender differences in type of occupation,
pay, work-hours and experience in the workplace often
place women at a disadvantage to men (Dakin a
Doyal, 1999; Annandale a Hunt, 2000). Occupational
gender segregation is a key characteristic of the labour
market; women are disproportionately concentrated in a
narrow range of low paid and low status jobs; women
who are employed in traditionally male-dominated jobs
tend to remain in the least senior positions whereas men
working in ‘female’ occupations are over-represented at
a senior level (Jacobs, 1993). However, structural
changes in the nature of employment over recent
decades have begun to change these traditional gender
patterns, with evidence that men suffered disproportio-
nately from the decline in manufacturing, for example.
Annandale a Hunt (2000) argue that because these
changes have not been uniform for different subgroups
of men and women, new forms of inequality between the
sexes have emerged, and ethnicity may be a key
parameter of difference among men and women.
Brah (1993) emphasises that labour market experi-
ences are mediated not only through relations of gender
but also of ‘race’ or ethnicity. Racial discrimination may
function to confine minority ethnic workers to certain
types of low paid and low status occupations on the
periphery of the labour market associated with poor pay
and working conditions. Studies confirm that minority
ethnic workers are under-represented at a managerial
level and are more likely than white adults to be
employed in temporary or shift-work in labour-intensive
occupations (Modood et al., 1997; Office for National
Statistics, 1996; Jones, 1993). Minority ethnic adults are
more likely to be outside the formal labour market than
whites; their unemployment rate is considerably higher
and of longer duration (Amin a Oppenheim, 1996) and
recruitment procedures may discriminate against min-
ority ethnic applicants.
There is, however, considerable diversity in the socio-
economic profiles of minority ethnic groups. Although
the socio-economic position of Black Caribbean and
Indian adults is less advantaged than for whites,
Pakistanis and Bangladeshis of both sexes emerge as
the poorest and most deprived groups across a range of
social indicators (Modood et al., 1997). However, the
economic activity and employment of ethnic groups is
inextricably linked with gender and leads to differences
H. Cooper / Social Science a Medicine (2002) 693–706694
in employment status, class position and material
resources for men and women. The following section
considers in more detail the inter-relationships between
gender, ethnicity and socio-economic position and their
significance for assessing health inequalities among these
groups.
Socio-economic inequality and health
The finding that health is influenced by socio-
economic position is well established; reported poor
health is greatest for individuals at the bottom of the
social hierarchy, but a step-wise gradient in morbidity
extends across the whole social spectrum. It is argued
that the social and economic standing of individuals’ in
society shapes exposure to health-damaging agents, as
well as determining individual resources to promote
health (Lynch a Kaplan, 2000).
Social class and employment
Of the number of different measures used to assess the
extent and magnitude of socio-economic inequality in
health, the most common in UK health research is
occupational social class, based on current or last main
occupation. Class gradients in self-assessed health have
been found for men and women (Matthews, Orly, a
Power, 1999; Arber, 1996) although some report that
individual class differences are less marked for women’s
health (Yuen, Machin, a Balarajan, 1990; Stronks, van
de Mheen, van den Bos, a Mackenbach, 1995).
However, these studies have not considered how class
position may be differentially related to the reported
health of men and women from different ethnic groups.
Nazroo (1997) reports a similar relationship between
class and self-assessed health for white and minority
ethnic groups, based on the social class of the household
rather than the individual. However, this analysis
standardised for sex and age, thus precluding compar-
ison of gender and ethnic differences in the relationship
between social class and health. Investigation of social
class and ethnic health inequality rarely use individual
occupational class for men and women, although many
studies of gender inequality and health advocate the use
of women’s own occupational class to assess socio-
economic inequalities in health (e.g. Arber, 1996),
particularly at a time of increased employment partici-
pation among white and minority ethnic women
(Bhopal, 1998).
In part, this approach reflects conceptual and
measurement problems associated with the use of
occupational class particularly for women and members
of minority ethnic groups. The internal heterogeneity of
occupational groups and evidence that, within any given
class, women and adults from minority ethnic groups
are disproportionately disadvantaged in terms of pay,
working conditions and job status are two main reasons
why occupational class may present a misleading picture
of socio-economic inequality in health (Macran, Clarke,
Sloggett, a Bethune, 1994; Emslie, Hunt, a Macintyre,
1999).
The reliance of occupational class on information
about the current or previous occupation of individuals
means that it is not inclusive of those who have never
had a paid job. Although the never employed constitute
a very small proportion of working-age adults (Arber,
1997b), this proportion varies markedly by gender and
ethnicity; a sizeable proportion of Pakistani and
Bangladeshi women report never having had a paid
job (West a Pilgrim, 1995; Modood et al., 1997)
although some of these women may be engaged in
home-working (Phizacklea a Wolkowitz, 1995).
Social class is also likely to be a less sensitive measure
of current social and economic conditions for the
growing number of ‘non-employed’ adults outside the
labour market. Reasons for non-employment are
gendered; approximately one-quarter of non-employed
women are housewives, whereas a greater percentage of
men are unemployed (Arber, 1996).
Studies show that the employment characteristics of
men and women are also related in different ways to
ethnic group. Minority ethnic men have a greater risk of
unemployment and non-employment than white men.
With the notable exception of Indian men, minority
ethnic men are more likely to work part-time and are
disproportionately concentrated in occupations asso-
ciated with low pay and job insecurity (ONS, 1996).
A low level of formal employment among Pakistani
and Bangladeshi women can largely be understood from
their domestic and child-care responsibilities (West a
Pilgrim, 1995). White, Black Caribbean and Indian
women of working-age are more likely to be employed
by comparison, but only white women have a high level
of part-time employment and a low risk of unemploy-
ment (ONS, 1996).
Within white and minority ethnic groups, women are
more likely than men to be non-employed, but this
gender difference is much greater for Pakistanis and
Bangladeshis than for other ethnic groups because of the
low economic activity of Pakistani and Bangladeshi
women (Modood et al., 1997; ONS, 1996). Unlike other
ethnic groups, where women tend to be disadvantaged
relative to men in terms of occupational level, pay and
working conditions, studies suggest greater non-manual
employment, lower unemployment and higher income
for Black Caribbean women than for Black Caribbean
men (Modood et al., 1997; ONS, 1996). Thus, patterns
of employment vary within and across ethnic groups
according to gender.
The likelihood of non-employment is greater for lower
social class groups and is associated with high levels of
H. Cooper / Social Science a Medicine (2002) 693–706 695
reported morbidity. An analysis of the General House-
hold Survey by Arber (1996) found stronger class
gradients in health for non-employed men compared
with those in paid employment, but less class variation
for non-employed women. Given the variation in
employment status between gender and ethnic groups
described above, it is important to include employment
status as a structural variable in analyses of health and
to distinguish between those in full-time and part-time
employment.
Education
Level of educational qualification may be important
in the creation and maintenance of social inequalities in
health, through shaping cognitive skills and learning
that are important for maintaining good health (Lynch
a Kaplan, 2000) or determining future labour market
success and material resources (Wadsworth, 1991).
A socio-economic measure based on educational
qualifications is more inclusive than occupational class
because it can represent adults who have never had a
paid job. Arber (1997a) found that educational qualifi-
cations were strongly associated with the general health
of working-age adults and unlike class, education could
differentiate the health of women who were non-
employed.
As well as being of particular value for assessing
socio-economic inequality in women’s health, education
may be better suited to the investigation of ethnic
inequality in health than social class because it can
overcome difficulties associated with the classification of
never employed and non-employed groups. Surveys
show marked differences in the educational profiles of
white and minority ethnic groups with lower educational
qualifications among Pakistanis and Bangladeshis than
for other ethnic groups (Modood et al., 1997). Within
ethnic groups, women tend to have a poorer educational
profile than men, with the notable exception of Black
Caribbean women (Blackburn, Dale, a Jarman, 1996).
To date, educational level has rarely been used in
British research to examine the socio-economic basis of
ethnic health inequality. It is, however, important to
examine the relationship between education and health
along with other socio-economic measures because for
any given level of qualification, the ‘economic return’ in
terms of future employment status and occupational
class may be less for minority ethnic men and women
because of inequality in job opportunities, pay and
working conditions (Lynch a Kaplan, 2000; Krieger,
Rowley, Herman, Avery a Phillips, 1993).
Material deprivation
Socio-economic measures grounded in everyday
material conditions or ‘standard of living’ are indepen-
dently associated with health inequalities among men
and women (Yuen et al., 1990; Arber, 1996). Indices of
deprivation, as well as single-item measures of car
ownership and housing tenure, often show a linear
relationship with health (Arber, 1997a). Material
resources may themselves have immediate benefits for
health in terms of improved living conditions, or may
primarily reflect labour market position or income.
Nazroo (1997) found that standard of living among
different ethnic groups was better able to explain the
poor self-assessed health of minority ethnic groups than
either household class or housing tenure. Thus, material
conditions may more directly reflect health-related
exposures and resources for members of minority
ethnic groups, although it is impossible to establish the
causal direction between health and material resources
from cross-sectional data (Berkman a Macintyre,
1997).
Methodology
The aim of this paper is to investigate patterns of
reported health among gender and ethnic groups and to
assess both the relative and overall contribution of
education, occupational class, employment status and
material circumstances to these gender and ethnic health
inequalities.
The paper analyses data from the Health Survey for
England (HSE) combined over 4 years from 1993 to
1996. The HSE provides representative data for white
and minority ethnic men and women living in private
households in England. Approximately 16,000 inter-
views were gained with adults aged 16 and above in 1993
and 1994 (Bennett et al., 1995; Colhoun a Prescott-
Clarke, 1996). The 1995 and 1996 surveys each include
interviews with more than 19,000 adults (Prescott-
Clarke a Primatesta, 1998, 1997).
The overall response rate to the HSE was approxi-
mately 77 percent in each year of the survey used here
(Colhoun a Prescott-Clarke, 1996). The paper analyses
approximately 43,500 adults in the combined HSE data-
set who were of working-age, defined here as between 20
and 60 years. Combining 4 years of survey data in this
way increased the number of minority ethnic men and
women in the sample. Ethnic group is based on the self-
identification of respondents with the following ethnic
groups; white (N=41,500), Black Caribbean (N=519),
Indian (N=900), Pakistani (N=430) and Bangladeshi
(N=116). Although the measurement of ethnicity in this
way is considered preferable to interviewer identification
of ethnic group, the meaning of fixed-choice categories
used in this type of survey question has been contested
(Senior a Bhopal, 1994).
H. Cooper / Social Science a Medicine (2002) 693–706696
Socio-economic measures
Four measures are used to represent individual socio-
economic position in this analysis. As each may have a
different meaning and significance for health, the use of
multiple socio-economic indicators provides a fuller and
more adequate adjustment for socio-economic differ-
ences between gender and ethnic groups than any single
measure (Smaje, 1995).
1. Educational level is based on the highest educational
qualification of each respondent. These are divided into
the following five categories; higher qualifications
(degree, professional or nursing qualifications), A’ Level
or equivalent; GCSE/ O’Level or equivalent, other
qualifications (e.g. vocational) and no qualifications.
2. Employment status is based on the individuals’
current labour market position. Distinctions are made
between adults who are currently employed, who are
unemployed (defined as actively seeking work) or have
never had a paid job. Due to the numbers in the HSE
sample, non-employed groups are grouped together in a
separate category that includes any of the following;
full-time students, retired adults, those looking after
home or family, the sick and/or disabled. It is, however,
recognised that this composite measure will conceal
diversity among different non-employed groups of men
and women. In the multivariate analysis, the
employment status variable further distinguishes
workers who are employed full-time (more than 30 h/
week) or part-time (30 h/week or less), as well as
adults who are ‘looking after the home’ because the
proportion of adults in these groups varies by gender
and ethnicity.
3. Occupational social class (socio-economic group,
SEG) is based on the individuals’ current or last main
job. Socio-economic groups include; professional and
managerial, routine non-manual, skilled manual and
semi-/unskilled manual workers. A separate category
includes the ‘never employed’ who constitute a large
proportion of Pakistani and Bangladeshi women.
3. An index of material deprivation was constructed
from five questions about individuals’ access to material
resources within their household. A score of +1 was
added for any of the following five items that applied; no
central heating in household; no telephone in household;
no car in household; home not owned; Income Support
received by anyone in the household. This gave a
maximum material deprivation score of 5 and a
minimum score of 0. Scores of 3 or more represent a
high level of material deprivation.
Analyses
The analysis first examines the nature of gender
inequality in health among different ethnic groups. The
differential socio-economic position of gender and
ethnic groups is then focused upon, before using
multivariate logistic regression analysis to show the
relative and overall contribution of socio-economic
position to inequalities in self-assessed health. All
analyses control for age, important because minority
ethnic groups have a much younger age structure than
the white population (ONS, 1996) and age is likely to
influence both socio-economic position and reported
health. In tables, age standardised percentages are
calculated by direct age-standardisation in 10-year age
groups, using the combined HSE sample as the standard
population. All logistic regression tables include age as
an independent variable in each model.
Results
Gender and ethnic inequality in health
A commonly used global indicator of morbidity is
self-assessed health, which is measured by the HSE
question ‘How is your health in general? Would you say it
was; very good, good, fair, bad or very bad?’ Poor self-
assessed health has been shown to be a good predictor of
mortality in other studies (Idler a Benyamini, 1997)
and has good test re-test reliability (Lundberg a
Manderbacka, 1996).
Table 1 shows that the proportion of all adults who
rated their health as ‘less than good’ (responses of ‘fair’,
‘bad’ or ‘very bad’ combined) was very similar for men
and women aged between 20 and 60 years. There was
very little gender difference using this measure of
reported morbidity for white adults, although the
slightly higher morbidity of white women can be seen
from the sex ratio of 1.06 in Table 1.
In contrast to whites, there were larger gender
differences in health for each minority ethnic group.
Black Caribbean, Indian and Pakistani women were
more likely than men in these ethnic groups to report
poor health. It is notable that gender differences in
health became more pronounced for minority ethnic
groups after standardising for age (see sex ratios in
Table 1). For Bangladeshis, higher morbidity among
men than women of working-age was reversed after
standardising for age-related differences. However, the
generalisability of this finding is limited by the small
number of Bangladeshis in the HSE sample.
The higher morbidity of minority ethnic women than
men after age standardisation occurs because gender
differences in the timing of migration mean that
minority ethnic women tend, on average, to be of a
younger age than minority ethnic men and white adults
(Blakemore a Boneham, 1994) and might therefore be
expected to report better, not worse, health than these
groups.
Table 2 compares the health of white women and
minority ethnic groups with that of white men, because
white men have the lowest level of reported morbidity in
H. Cooper / Social Science a Medicine (2002) 693–706 697
Table 1. Results are presented in this way to show the
extent and nature of health inequality among gender and
ethnic groups, and it is not intended that the health of
white men be interpreted as the ‘norm’ from which other
gender and ethnic groups deviate.
Ratios are presented in Table 2 to show the relative
health ‘advantage’ of white men compared to men and
women from other ethnic groups after standardising for
age. The health ratios show poorer health for minority
ethnic groups than for white men, especially for
Pakistanis and Bangladeshis. Compared to white men,
Black Caribbean, Indian, Pakistani and Bangladeshi
women were at a greater health disadvantage than men
in these minority ethnic groups.
Socio-economic inequality among gender and ethnic
groups
This part of the analysis examines the nature and
magnitude of socio-economic inequality associated with
gender and ethnicity as a precursor to exploring the
health implications of socio-economic disadvantage.
Pakistani and Bangladeshi groups, who had in common
a high level of reported morbidity, are combined where
age-standardised percentages are shown. Whilst this was
necessary due to the small numbers of Pakistani and
Bangladeshi men and women in the HSE sample, Fig. 1
also reports unstandardised percentages for both of
these ethnic groups to show any socio-economic
differences between them.
Fig. 1 shows that the likelihood of being in paid
employment varied according to gender and ethnic
group. For white, Indian and Pakistani/Bangladeshi
groups, a greater proportion of men than women were in
paid employment. The exception is for Black Caribbean
adults where there was little gender difference in overall
employment, but greater unemployment among men
than women.
For white and minority ethnic groups, a higher
percentage of women than men were non-employed; a
gender difference that will primarily reflect the number
of working-age women who report looking after family
and home. However, it is notable that nearly 20 percent
of Pakistani and Bangladeshi men aged 20–60 were non-
employed; this is much higher than for other men and
may reflect withdrawal from paid work on the grounds
Table 1
Gender differences in reported ‘less than good’ health by ethnic groupa
Men Women Sex ratio
Men/women
Age std
sex ratio
% Age std % N % Age std % N
All adults aged 20–60 18 18 20,793 19 19 23,923 1.06*** 1.06
White 17 17 19,330 18 18 22,233 1.06** 1.06
Black Caribbean 23 23 206 32 36 312 1.39* 1.56
Indian 22 24 431 28 30 469 1.27* 1.25
Pakistani 31 34 214 35 42 216 1.13 (ns) 1.23
Bangladeshi 36 36 66 28 48 50 0.77 (ns) 1.33
a (1)*Statistical significance of gender difference in health; *Po0:05; **Po0:01; ***Po0:001: (2) Age standardisation in 10-year age
groups. Source: Health Survey for England (1993–96).
Table 2
Gender and ethnic differences in reported ‘less than good health’: a comparison with white men
White Black Indian Pakistani Bangladeshi
Caribbean
Age standardised percentages
Men 17 23 24 34 36
Ratioa 1.00 1.35 1.41 2.00 2.12
Women 18 36 30 42 48
Ratio 1.06 2.12 1.76 2.47 2.82
Base numbers
Men 19,330 206 431 214 66
Women 22,233 312 469 216 50
aRatio=white women or minority ethnic groups/ white men. Source: Health Survey for England (1993–96).
H. Cooper / Social Science a Medicine (2002) 693–706698
of ill-health. Nearly 60 percent of Pakistani and
Bangladeshi women of working-age reported never
having had a paid job, with only 15 percent currently
in paid employment. This differs markedly from the
employment profile of other women, where the majority
were employed. Greater unemployment and economic
inactivity among working-age Pakistani and Banglade-
shi women suggests that they are likely to be most
disadvantaged in terms of income, material resources
and work-related benefits.
Table 3 compares the percentage of men and women
in each ethnic group who are located in low educational
and social class groups, and who score highly on an
index of material deprivation, because these positions of
‘socio-economic disadvantage’ are thought to be most
potentially damaging to health.
For all adults of working-age, women were more
likely than men to have no educational qualifications or
to be in semi or unskilled manual jobs and approxi-
mately 10 percent of men and women were in the most
materially deprived group with a score of 3 or more on
the material deprivation index. These results suggest
that working-age women are disproportionately located
in low socio-economic positions relative to men.
White women were more likely than white men to be
without formal qualifications, but there was no greater
educational disadvantage for Black Caribbean and
Indian women than for men in these respective ethnic
groups after adjusting for age-related differences (Table
3a). The results suggest that a greater proportion of
Pakistani women were without educational qualifica-
tions than Pakistani men, but the opposite was found for
Bangladeshis. More than half of Bangladeshi men and
women of working-age were without formal qualifica-
tions, highlighting that educational disadvantage in this
ethnic group extends to both sexes.
Fig. 1. Employment status of men and women aged 20–60 by ethnic group.
H. Cooper / Social Science a Medicine (2002) 693–706 699
Based on current or last main occupation, the
disadvantaged class position of white, Black Caribbean
and Indian women relative to men in their respective
ethnic groups is evident from Table 3b. Within each of
these ethnic groups, a greater proportion of women than
men were classified in semi-skilled or unskilled manual
occupations, and these gender differences remained after
adjusting for age. A small proportion of all Pakistani
and Bangladeshi women were in the semi-skilled or
unskilled manual class compared to men in this ethnic
group. This gender difference results from the high
percentage of never employed Pakistani and Banglade-
shi women (see Fig. 1), who are excluded from occupa-
tional class measures of socio-economic position.
In all ethnic groups, a greater proportion of women
than men were in the most materially deprived group
(Table 3c). Adjusting for age, approximately one-
quarter of Black Caribbean women had poor material
living conditions according to this measure, compared
with 21 percent of Black Caribbean men. The magnitude
of gender inequality in material deprivation was less
marked for other ethnic groups by comparison, and
Indian men and women were least likely to be classed as
materially deprived.
These results show that women and minority ethnic
groups were often over-represented in positions asso-
ciated with socio-economic disadvantage, whereas white
men had a high level of paid employment and were least
likely to be in low educational and class groups. What is
striking is that patterns of socio-economic disadvantage
among gender and ethnic groups appear to follow
patterns of health inequality reported in Table 1; the
poorest health and greatest socio-economic disadvan-
tage was concentrated among Pakistanis and Banglade-
shis, although there were marked gender differences in
employment participation within this ethnic group.
There was also considerable variation among gender
and ethnic groups according to the socio-economic
measure used; the finding of low material deprivation
among Indian men and women highlights that minority
groups cannot be characterised as uniformly disadvan-
taged. These socio-economic differences suggest, firstly,
Table 3
Socio-economic disadvantage among men and women aged 20–60 by ethnic group
Men Women
% Age std % N % Age std % N
(a) % with no educational
qualifications
White 23 23 19,323 29 29 22,233
Black Caribbean 35 35 207 23 35 311
Indian 24 25 431 35 25 468
PakistaniBangladeshi
3862
�47 214
664654
�54 216
50
All 24 23 20,782 29 31 23,915
(b) % in semi or unskilled manual
socio-economic groupsa
White 16 14 19,161 27 27 22,211
Black Caribbean 22 21 206 28 31 311
Indian 20 20 430 29 30 469
PakistaniBangladeshi
2951
�35 214
65186
�32 216
50
All 16 16 20,609 27 27 23,894
(c) % in most materially disadvantaged
group (score of 3+)
White 9 9 18,835 11 11 21,682
Black Caribbean 21 21 202 28 26 304
Indian 6 6 418 8 7 460
PakistaniBangladeshi
1327
�17 208
661732
�19 210
50
All 10 10 20,801 11 11 23,569
aClass based on current or last main job. Source: Health Survey for England (1993–96).
H. Cooper / Social Science a Medicine (2002) 693–706700
that socio-economic correlates of health may differ for
gender and ethnic groups, and secondly, that the sole use
of occupational class measures of socio-economic
inequality will be inadequate for some groups of
minority ethnic women because they exclude the never
employed.
Multivariate analyses
This section uses logistic regression analyses to
examine how gender and ethnic differences in self-
assessed health change when the socio-economic char-
acteristics of individuals are included in the same
analysis. The aim of this section is to assess the extent
to which inequality in health across gender and ethnic
groups can be accounted for by their differential socio-
economic characteristics. The interaction between gen-
der and ethnicity is included as a single independent
variable in the logistic models presented in Table 4.
Model 1 gives the odds ratios for reported ‘less than
good’ health for gender and ethnic groups with white
men defined as the reference category (1.00), and age in
5-year groups included in the model. Being white and
male was associated with the best health, with the odds
of ‘less than good’ health significantly higher for
minority ethnic men, white and minority ethnic women
in comparison. The odds ratios of poor health were
more than two times higher for Black Caribbean and
Indian women and over three times greater for Pakistani
women. The odds ratios of poor health for Black
Caribbean, Indian and Pakistani men were somewhat
lower than for women in each of these ethnic groups, but
minority ethnic men were clearly disadvantaged in their
health compared to white men. Bangladeshi men had a
higher odds ratio of poor health than Bangladeshi
women, but for both sexes the odds were markedly
higher than the reference category of white men. These
substantial gender and ethnic differences in health were
all highly statistically significant in the model.
Subsequent models in Table 4 show how these odds
ratios of poor health were modified by controlling for
different socio-economic characteristics. Socio-economic
measures are entered sequentially into the logistic
regression models in order to assess their relative
contribution to the self-assessed health of gender and
ethnic groups. Model 2 includes educational level and
this shows a consistent linear relationship with health.
Adults with the highest level of education were least
likely to report ‘less than good’ health and the odds of
poor health became consistently greater with a lower
level of educational qualification and were over 3.5 for
those with no qualifications compared to those with a
degree or above.
Black Caribbean men did not have a significantly
higher odds ratio of morbidity when education was
added to the model, but Black Caribbean women
continued to have poorer health than white men
(OR=2.46). Educational level made little difference to
the odds ratio of poor health for Indian men, and
although the odds were reduced for Indian women, their
health remained significantly poorer than that of white
men. For Pakistanis, the odds of poor health were
substantially reduced for both sexes once education was
added to the model. A similarly large reduction in the
odds ratio of poor health was found for Bangladeshi
men, but the greatest change was for Bangladeshi
women where the odds of poor health were no longer
significantly different from white men.
After controlling for education, there was no gender
difference in health for white adults, but women who
were Black Caribbean, Indian or Pakistani continued to
have higher odds ratios of poor health than men from
the same ethnic group. The exception was a higher odds
ratio for Bangladeshi men than for Bangladeshi women.
The differential impact of education on the health of
gender and ethnic groups suggests that educational
disadvantage is a major factor in accounting for the
higher morbidity of Black Caribbean men, white and
Bangladeshi women relative to white men, and to some
extent contributes to the poor health of Pakistanis,
Indian women and Bangladeshi men. However, adjust-
ing for education does little to alter gender differences in
health found within minority ethnic groups.
One way in which educational qualifications may
influence health is through labour market position.
Model 3 shows that both education and employment
status have strong and independent relationships with
health. Adults of working-age who were employed full-
time had the best health, with the odds significantly
higher for part-time workers in comparison. Being
unemployed or looking after the home were both
associated with high reported morbidity, with odds
approximately twice as high odds ratio as for the full-
time employed. It is, however, impossible here to assess
the extent to which poor health precedes job loss or
economic inactivity. The highest odds ratio of poor
health was for other non-employed groups, which is
expected as this category includes the long-term sick and
disabled. Adults of working-age who have never been
employed approx 3 times higher odds of poor health
than the ref. cat.
As shown in Fig. 1, the never employed group
included a large proportion of Pakistani and Banglade-
shi women; controlling for employment status substan-
tially reduced their odds of poor health relative to white
men (to under 1.00 for Bangladeshi women). The odds
of poor health were lowered by over one-third for Black
Caribbean and Indian women along with Pakistani men,
all of whom had a lower level of employment than white
men in Fig. 1. With the exception of Indian men, whose
employment profile was comparable to white men in
Fig. 1, these results show that the poor position of
H. Cooper / Social Science a Medicine (2002) 693–706 701
minority ethnic men and women in the labour market
may serve to disadvantage their health.
The odds of poor health for white women, a
disproportionate number of whom are employed
part-time, is significantly lower than for white men
once employment status is included in the model.
Thus, controlling for both education and employment
status reverses the gender difference in health for white
adults of working-age found in Model 1. Employment
status could account for a greater proportion of the
poor health reported by Pakistani and Indian women
than for men in these respective ethnic groups, thus the
gender gap is narrowed for Indians and reversed for
Pakistanis. By contrast, marked gender differences in
health remain for Black Caribbean and Bangladeshi
adults.
Model 4 considers how the occupational class of all
those currently or previously employed is related to
Table 4
Logistic regression models of ‘less than good’ healtha
Model 1 Model 2 Model 3 Model 4 Model 5
Age (in 5 year groups) +++ +++ +++ +++ +++
Gender and ethnic group +++ +++ +++ +++ +++
White men 1.00 1.00 1.00 1.00 1.00
White women 1.08** 0.99 0.89** 0.97 0.99
Black Caribbean men 1.61** 1.39 1.25 1.22 1.15
Black Caribbean women 2.55*** 2.46*** 2.09*** 2.26*** 1.98***
Indian men 1.46** 1.47** 1.42** 1.43** 1.50**
Indian women 2.07*** 1.81*** 1.47** 1.53*** 1.70***
Pakistani men 2.31*** 1.97*** 1.61** 1.57** 1.52*
Pakistani women 3.24*** 2.38*** 1.43* 1.53** 1.68**
Bangladeshi men 2.75*** 1.94* 1.62 1.66 1.56
Bangaldeshi women 2.31** 1.57 0.93 1.03 1.05
Educational qualifications +++ +++ +++ +++
Higher 1.00 1.00 1.00 1.00
A level or equivalent 1.45*** 1.35*** 1.23*** 1.23***
GCSE/O’Level or equivalent 1.55*** 1.53*** 1.34*** 1.31***
Other 2.01*** 1.90*** 1.59*** 1.50***
None 3.57*** 3.04*** 2.34*** 2.03***
Employment status +++ +++ +++
Employed full-time (30+ h/week) 1.00 1.00 1.00
Employed part-time (o30 h/week) 1.22*** 1.16** 1.13**
Unemployed 2.12*** 2.01*** 1.48***
Looking after home 1.92*** 1.83*** 1.51***
Other non-employed 7.24*** 7.00*** 5.77***
Never been employed 2.97*** 3.78*** 2.74***
Socio-economic Group (SEG) +++ +++
Professional /managerial 1.00 1.00
Routine non-manual 1.03 1.03
Skilled manual 1.49*** 1.42***
Semi or unskilled manual 1.60*** 1.42***
Material deprivation score +++
0 (none) 1.00
1 1.34***
2 1.77***
3+ 2.26***
DLLR (Ddf) 1023 (16) 1219 (4) 2404 (5) 192 (3) 363 (3)
N= 42202
a (1)+++ Statistical significance of variable in the model: +Po0:05; ++Po0:01; +++Po0:001. (2)**Statistical significance of
difference from the reference category; *Po0:05; ***Po0:001 Source: Health Survey for England (1993-96).
H. Cooper / Social Science a Medicine (2002) 693–706702
health (the never employed are not excluded from this
model). The odds of poor health were increased by 49
percent for the skilled manual class and 60 percent for
those classified in semi-skilled or unskilled manual
occupations compared to the professional/managerial
class, but the health of the routine non-manual class was
not significantly different to this group.
Controlling for occupational class made little differ-
ence to the pattern of health inequality across gender
and ethnic groups; the odds of poor reported health
remained significantly higher for Indians, Pakistanis and
Black Caribbean women compared to white men. Prior
to adding occupational class, white women reported
significantly better health than men, but the odds of
occupational class in model 4 removed this gender
difference for whites but did not alter the gender
difference for minority ethnic groups.
The index of material deprivation is the final socio-
economic measure added in Model 5. There was a highly
significant material deprivation gradient in health; the
best health was found for materially advantaged adults
on this measure (score 0) rising to an odds ratio of 2.26
for those with a score of 3+ on the material deprivation
index. It is notable that material living conditions
appear to reduce the reported health disadvantage for
Black Caribbean women to a greater extent than either
education or occupational class, although Black Car-
ibbean women continue to have a significantly high odds
ratio of poor health. This could suggest that education
and, particularly social class, poorly represent the socio-
economic position of these women, if as some authors
have suggested, there is a disparity between educational
qualifications, occupation and material living conditions
(Bruegel, 1994; Krieger et al., 1993).
The results did not suggest that the material living
conditions of Indian men and women contributed to
their high morbidity, and this is likely to reflect the
smaller proportion of Indians than white men and
women living in the most materially disadvantaged
conditions (see Table 3c). There was higher morbidity
among Indian and Pakistani women than men after
controlling for the measure of material deprivation, but
the odds of poor health were significantly greater for
both sexes relative to white men.
The overall contribution of socio-economic position
to gender and ethnic inequality in health is shown in
Fig. 2 where unadjusted odds ratios can be compared
with odds ratios adjusted for socio-economic position. A
key finding is that socio-economic characteristics sub-
stantially reduce the magnitude of ethnic inequality in
health, especially for Black Caribbean, Pakistani and
Bangladeshi adults, but taking into account socio-
Fig. 2. Odds ratios of ‘less than good’ health for gender and ethnic groups: figures adjusted for age and socio-economic position.
H. Cooper / Social Science a Medicine (2002) 693–706 703
economic inequality does little to alter gender differences
in health within minority ethnic groups.
Socio-economic disadvantage made most contribu-
tion to the poor health reported by Bangladeshis, and
this was more marked for women than for men.
Morbidity reported by Pakistanis can also be largely
attributed to poor socio-economic circumstances,
particularly for women, although the odds of poor
health for Pakistanis remained significantly higher
relative to white men after adjusting for socio-economic
position.
Measures of socio-economic position made less over-
all contribution to the poor self-assessed health of
Indian adults, most notably men, whose socio-economic
position was most comparable to that of white men. A
sizeable gender difference in morbidity remains for
Black Caribbean adults; only women in this ethnic
group have a significantly higher odds ratio of poor
health.
Conclusions
The finding of little overall gender difference in self-
assessed health for white adults of working-age contrasts
with substantial inequality in health among men and
women from different ethnic groups. Consistent
with other studies, reported morbidity was greater
for minority ethnic groups than for whites of
both sexes, with the greatest health disadvantage
found for Pakistanis and Bangladeshis. An additional
finding of this study was of marked gender differences in
health within minority ethnic groups. These gender
inequalities were accentuated by standardising for age,
suggesting that minority ethnic women report particu-
larly poor health despite their younger average age
profile.
The key finding of this paper is that, despite the
problems associated with the use of socio-economic
measures for gender and ethnic groups, these
accounted for a substantial proportion of ethnic
inequality in health. Adjusting for educational qualifica-
tions substantially reduced the likelihood of poor health
for Black Caribbean men, Pakistanis and Bangladeshi
women. Being in paid employment was positively
associated with good health, and controlling for employ-
ment status reduced the odds of poor health for
working-age Pakistani and Bangladeshi womenFa
substantial proportion of whom were non-employed.
As expected, occupational class made less contribution
to patterns of gender and ethnic inequality in health
than education or employment status, and is a
particularly poor marker of socio-economic inequality
in health for minority ethnic women because of its
reliance on previous occupation. Material deprivation
was independently associated with health and the results
suggest that this measure better accounts for high
morbidity among Black Caribbean women than other
socio-economic measures.
Whilst socio-economic disadvantage can explain in
large part why many minority ethnic adults report
poorer health than white men, significant ethnic inequal-
ity in health remained after adjusting for socio-economic
position, particularly for minority ethnic women. This
suggests that socio-economic measures are important,
but cannot fully ‘explain’ gender and ethnic inequality in
health for the following reasons;
Firstly, it is recognised that ethnicity is not simply
‘reducible’ to socio-economic position (Nazroo, 1998).
The findings from this study show considerable diversity
among ethnic groups who cannot be characterised as
uniformly disadvantaged relative to whites. The poor
health reported by Indian men, for example, was not due
to their socio-economic disadvantage relative to white
men, since there were more similarities than differences
in socio-economic position for men in these ethnic
groups. However, for other minority ethnic groups, it is
important to show that poor socio-economic conditions
have a sizeable impact on the health because this
detracts from an undue emphasis on individual or
cultural explanations that risk stereotyping assumed
differences from the white population.
Secondly, although this analysis examines how key
socio-economic measures are related to the pattern of
health inequality across gender and ethnic groups, this
does not represent a ‘complete’ adjustment for social
and economic living conditions, or the economic and
emotional health consequences of discrimination. After
adjusting for socio-economic position, many minority
ethnic women had a higher odds ratio of morbidity than
men in the same ethnic group. This gender difference
was most marked for Black Caribbean women who,
unlike Black Caribbean men, continued to have
significantly higher odds of poor health relative to white
men. Part of the explanation may concern a disparity
between educational qualifications and class positionFwhere studies suggest Black Caribbean women are more
‘advantaged’ than Black Caribbean menFand actual
living conditions that are relevant to health (Blackburn
et al., 1996).
Socio-economic position is of course only one of
many possible determinants of health. Further research
should address how socio-economic circumstances
intersect with family responsibilities or the health-
related behaviours of gender and ethnic groups. It is
important to investigate, for example, whether the high
level of smoking reported for Bangladeshi men (Rudat,
1994) makes an independent contribution to their poor
health and how socio-economic disadvantage experi-
enced by gender and ethnic groups may be compounded
by large family sizes, caring for dependent children or
domestic work.
H. Cooper / Social Science a Medicine (2002) 693–706704
Acknowledgements
This doctoral research was funded by the ESRC and
data from the Health Survey for England was made
available from the Data Archive on-line at MIMAS.
The author would like to thank Sara Arber and Chris
Smaje at the University of Surrey for their helpful
comments and discussion relating to this paper.
References
Amin, K., a Oppenheim, C. (1996). Poverty in black and white:
Deprivation and ethnic minorities. London: CPAG Runny-
mede Trust.
Annandale, E., a Hunt, K. (2000). Gender inequalities in
health: research at the crossroads. In E. Annandale, a K.
Hunt (Eds.), Gender inequalities in health. Buckingham:
Open University Press.
Arber, S. (1996). Integrating non-employment into research
on health inequalities. Health and Social Policy, 26(3),
445–481.
Arber, S. (1997a). Comparing inequalities in women’s and
men’s health: Britain in the 1990s. Social Science a
Medicine, 44, 773–787.
Arber, S. (1997b). Insights about the non-employed, class and
health: Evidence from the General Household Survey. In D.
Rose, a K. O’Reilly (Eds.), Constructing classes (pp. 78–
92). Swindon: ESRC/ONS.
Bennett, N., et al. (Eds.) (1995). Health survey for England:
1993. London: H.M.S.O.
Berkman, L.F., a Macintyre, S. (1997). The measurement of
social class in health studies: Old measures and new
formulations. In: Kogevinas, M., et al. (Eds.), Social
inequalities and cancer. IARC Scientific publications, No.
138.
Bhopal, K. (1998) How gender and ethnicity intersect: The
significance of education, employment and marital status.
Sociological Research Online, 3, 3. URL: http://www.socre-
sonline.org.uk/socresonline/3/3/6.html
Blackburn, R. M., Dale, A., a Jarman, J. (1996). Ethnic
differences in attainment in education, occupation and
lifestyle. In Kahn, V. (Ed.), Ethnicity in the 1991 Census. Vol
4. London: The Stationary Office.
Blakemore, K., a Boneham, M. (1994). Age, race and
ethnicity: A comparative approach. Buckingham: Open
University Press.
Brah, A. (1993). ‘Race’ and ‘culture’ in the gendering of labour
markets: South Asian young Muslim women and the labour
market. New Community, 45819(3), 441.
Bruegel, I. (1994). Labour market prospects for women from
ethnic minorities. In R. Lindley (Ed.), Labour market
structures and prospects for women. Manchester EOC.
Colhoun, H., Prescott-Clarke, P. (Eds.) (1996). Health Survey
for England, 1994. London: H.M.S.O.
Emslie, C., Hunt, K., a Macintyre, S. (1999). Problematising
gender, work and health: The relationship between gender,
occupational grade, working conditions and minor morbid-
ity in full-time bank employees. Social ScienceaMedicine,
48(1), 33–48.
Fenton, S., Hughes, A. O., a Hine, C. E. (1995). Self-assessed
health, economic status and ethnic origin. New Community,
21(1), 55–69.
Idler, E. L., a Benyamini, Y. (1997). Self-rated health
and mortality: A review of twenty-seven community
studies. Journal of Health and Social Behavior, 38(1),
21–37.
Jacobs, J. A. (1993). Men in female-dominated fields trends and
turnover. In C. L. Williams (Ed.), Doing womens work. Men
in non-traditional occupations (pp. 49–63). Newbury Park,
CA: Sage.
Jones, T. (1993). Britains ethnic minorities. London: Policy
Studies Institute.
Krieger, N. (2000). Discrimination and health. In L. F.
Berkman, a I. Kawachi (Eds.), Social epidemiology.
Oxford: Oxford University Press.
Krieger, N., Rowley, P. L., Herman A. A., Avery, B., a
Phillips, M. T. (1993). Racism, sexism and social class:
Implications for studies of health, disease and well-being.
American Journal of Preventative Medicine, 9(6 Suppl), 82–
122.
Lundberg, O., a Manderbacka, K. (1996). Assessing the
reliability of a measure of self-rated health. Scandinavian
Journal of Social Medicine, 24(3), 218–224.
Lynch, J., a Kaplan, G. (2000). Socioeconomic position. In L.
F. Berkman, a I. Kawachi (Eds.), Social epidemiology.
Oxford: Oxford University Press.
Macintyre, S., Hunt, K., a Sweeting, H. (1996). Gender
differences in health: Are things really as simple as they
seem? Social Science a Medicine, 42(4), 614–624.
Macran, S., Clarke, L., Sloggett, A., a Bethune, A. (1994).
Womens socio-economic status and self-assessed health:
Identifying some disadvantaged groups. Sociology of Health
and Illness, 16, 182–208.
Matthews, S., Orly, M., a Power, C. (1999). Social inequalities
in health: are there gender differences? Social Science a
Medicine, 48(1), 49–60.
Modood, T., Berthoud, R., Lakey, J., Nazroo, J., Smith, P.,
Virdee, S., a Beishon, S. (1997). Ethnic minorities in
Britain: Diversity and disadvantage. London: Policy Studies
Institute.
Nathanson, C. (1975). Illness and the feminine role: A
theoretical review. Social Science a Medicine, 9, 57–62.
Nathanson, C. (1977). Sex, illness and medical care: A review of
data, theory and method. Social Science a Medicine, 11,
13–25.
Nazroo, J. Y. (1997). The health of Britains ethnic minorities.
London: Policy Studies Institute.
Nazroo, J. Y. (1998). Genetic, cultural or socio-economic
vulnerability? Explaining ethnic inequalities in health. In M.
Bartley, D. Blane, a G. Davey Smith (Eds.), The sociology
of health inequalities. Oxford: Blackwell Publishers.
Office for National Statistics (1996). Social focus on ethnic
minorities. London: H.M.S.O.
Owen, D. (1992). Ethnic minorities in Britain: Settlement
patterns. Coventry: University of Warwick, Centre for
Research.
Phizacklea, A., a Wolkowitz, C. (1995). Homeworking women:
gender, racism and class at work. London: Sage.
Prescott-Clarke, P., a Primatesta, P. (1997). In Health Survey
for England, 1995. London: H.M.S.O.
H. Cooper / Social Science a Medicine (2002) 693–706 705
Prescott-Clarke, P., a Primatesta, P. (1998). In Health Survey
for England, 1996. London: H.M.S.O.
Rudat, K. (1994). Black and minority ethnic groups in England.
London: Health Education Authority.
Senior, P. A., a Bhopal, R. (1994). Ethnicity as a variable in
epidemiological research. British Medical Journal, 309, 327–
330.
Smaje, C. (1995). Health, ‘race’ and ethnicity: Making sense of
the evidence. London: Kings Fund Institute.
Stronks, K., van de Mheen, H., van den Bos, J., a Mackenbach,
J. P. (1995). Smaller socio-economic inequalities in health
among women: The role of employment status. International
Journal of Epidemiology, 24(3), 559–568.
Verbrugge, L. (1979). Females and illness: Recent trends in sex
differences in the United States. Journal of Health and Social
Behavior, 17, 387–403.
Wadsworth, M. (1991). The imprint of time. Oxford: Clarendon
Press.
West, J., a Pilgrim, S. (1995). South Asian women in
employment: The impact of migration, ethnic origin and
the local economy. New Community, 21(3), 357–378.
Yuen, P., Machin, D., a Balarajan, R. (1990). Inequalities in
health: Socio-economic differences in self-reported morbid-
ity. Public Health, 104(1), 65–71.
H. Cooper / Social Science a Medicine (2002) 693–706706