Full text of DOI:10.1207/S15327558IJBM0903_07

INTERNATIONAL JOURNAL OF BEHAVIORAL MEDICINE, 9(3), 263–285
Copyright © 2002, Lawrence Erlbaum Associates, Inc.
Which Aspects of Socio-Economic
Status are Related to Health in
Mid-Aged and Older Women?
Gita D. Mishra, Kylie Ball, Annette J. Dobson,
Julie E. Byles, and Penny Warner-Smith
A population-based study was conducted to validate gender- and age-specific indexes
of socio-economic status (SES) and to investigate the associations between these in-
dexes and a range of health outcomes in 2 age cohorts of women. Data from 11,637
women aged 45 to 50 and 9,510 women aged 70 to 75 were analyzed. Confirmatory
factor analysis produced four domains of SES among the mid-aged cohort (employ-
ment, family unit, education, andmigration) andfour domains among the oldercohort
(family unit, income, education, and migration). Overall, the results supported the
factor structures derived from another population-based study (Australian Bureau of
Statistics, 1995), reinforcing the argument that SES domains differ across age groups.
In general, the findings also supported the hypotheses that women with low SES
would have poorer health outcomes than higher SES women, and that the magnitude
of these effects would differ according to the specific SES domain and by age group,
with fewer and smaller differences observed among older women. The main excep-
Kylie Ball, School of Health Sciences, Deakin University, Victoria, Australia; Annette J. Dobson,
School of Population Health, University of Queensland, Queensland, Australia; Julie E. Byles, Centre
for Clinical Epidemiology & Biostatistics, University of Newcastle, Australia; Penny Warner-Smith,
Department of Leisure and Tourism Studies, University of Newcastle, Australia.
The research on which this article is based was conducted as part of the Australian Longitudinal Study
on Women’s Health (Women’s Health Australia).
Kylie Ball is supported by a Public Health Postdoctoral Research Fellowship from the Australian Na-
tional Health and Medical Research Council, ID 136925.
WearegratefultotheAustralianCommonwealthDepartmentofHealthandAgeingforfundingandto
thestaff oftheResearch CentreforGenderand Health at theUniversityofNewcastle, Australia, fortheir
assistance. Wearealsogratefultotheeditorandrefereesfortheirhelpfulcommentsandsuggestions.
Correspondence concerning this article should be addressed to Kylie Ball, School of Health Sciences,
Deakin University, 221 Burwood Highway, Burwood VIC 3125, Australia. E-mail:
kball@deakin.edu.au

264
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
tion was that in the older cohort, the education domain was significantly associated
with specific health conditions. Results suggest that relations between SES and health
are highly complex and vary by age, SES domain, and the health outcome under
study.
Key words: socio-economic status (SES), measurement, physical and mental health, health
care utilization, health behavior, women’s health
A vast body of research demonstrates associations between socio-economic sta-
tus (SES) and morbidity and mortality from a range of physical and mental
health conditions. Low SES has been linked with the prevalence of cardiovascu-
lar disease (Brezinka & Kittel, 1996; Hallqvist, Lundbert, Diderichsen, &
Ahlbom, 1998; McElduff & Dobson, 2000; Osler et al., 2000; Tyroler, 1999),
obesity (Sarlio-Lahteenkorva & Lahelma, 1999; Sobal & Stunkard, 1989), Type
2 diabetes (Evans, Newton, Ruta, MacDonald, & Morris, 2000), hypertension
(Vargas, Ingram, & Gillum, 2000), complaints and symptoms (Der, MacIntyre,
Ford, Hunt, & West, 1999; Mackenbach, 1992), including constipation
(Johanson & Sonnenberg, 1990), perceived general health (Hemingway, Nichol-
son, Stafford, Roberts, & Marmot, 1997; Mackenbach, 1992), and psychosocial
stress (Mackenbach, 1992). Those with low SES are also reportedly at increased
risk of engaging in unhealthy behaviors including smoking (Graham & Hunt,
1994; Power & Matthews, 1997), poor diet (Baghurst et al., 1990), physical in-
activity (Crespo, Ainsworth, Keteyian, Heath, & Smit, 1999), and decreased use
of health care services (Mackenbach, 1992).
SES inequalities in health may vary by gender and age. There is some incon-
sistency in results of studies that have compared SES differentials among men
and women. Although some findings demonstrate stronger associations of SES
with health outcomes among men (e.g., Arber, 1997; Elo & Preston, 1996), oth-
ers show no sex differences (e.g., Hemingway et al., 1997). Some evidence sug-
gests that inequalities increase over the adult years until later in life (House et
al., 1994) and that SES differences in relative rates of mortality and morbidity
are strongest among young to mid-aged adults (Mustard, Derksen, Bethelot,
Wolfson, & Roos, 1997). It has been suggested that SES inequalities in health
and mortality are small in the elderly, perhaps due to a narrowing of social and
economic differentials once individuals leave employment (Anderson, Sorlie,
Backlund, Johnson, & Kaplan, 1997). However, these findings may depend on
country- or time-specific findings, such as social welfare support for the elderly.
In addition, contradictory findings show SES gradients in health do exist among
the elderly (Berkman & Gurland, 1998; Broom, 1984; Hart, Davey Smith, &
Blane, 1998). Studies showing a narrowing of SES differentials among older
adults have tended to rely on income as an indicator of SES, and the inclusion of

SOCIO-ECOMOMIC STATUS AND WOMEN’S HEALTH
265
education in addition to income in the study by Berkman and Gurland (1998)
may have contributed to its contradictory findings.
Such contradictory findings and use of varying indicators is likely to be attrib-
utable in part to disagreement over the conceptualization of SES. SES is the term
commonly used to refer to the expression and distribution of such attributes as oc-
cupation, income, and status (Liberatos, Link, & Kelsey, 1988), although Kreiger
et al. (1997) preferred the term socio-economic position, arguing that SES includes
both access to and possession of material resources as well as prestige and position
in a hierarchical ranking. According to seminal researcher Max Weber, differen-
tial social position is based on three dimensions: class (an economic concept, indi-
cated, e.g., by income), power (related to political context), and status (“access to
life chances” based on social and cultural factors; Liberatos et al., 1988; Weber,
1946). According to this definition, factors such as family background and life-
style could be considered sociocultural factors reflective of the “status” dimension
(Liberatos et al., 1988). Although not typically considered a component of SES, it
is also now recognized that “ethnicity” is a social rather than biological construct
(Kreiger, 2000) and ethnicity and migration history may also reflect “access to life
chances.” Migrants in Australia, for example, experience higher rates of unem-
ployment and lower labor force participation rates than the Australian-born popu-
lation (Australian Bureau of Statistics [ABS], 1999). This may be attributable to
noneconomic factors, including migrants’ experiences of discrimination in their
new country. Although associations between health and SES are commonly ex-
plained in terms of the material reality of modern life, it is important to acknowl-
edge the complex contribution of social, cultural, and behavioral factors
(Scambler & Higgs, 1999), particularly when considering mental, as well as physi-
cal, health. Structural disadvantage and poorer life chances shape psychosocial
orientations and lifestyle behaviors that are prejudicial to health. Hence the inclu-
sion of migrant history as a social indicator along with conventional measures of
SES provides a broader picture of socio-economic position and its association with
a range of physical and mental health outcomes. Although we accept the “blur-
ring” of material and social power that occurs with the use of SES, we adopt SES
rather than socio-economic position in this article as the more widely understood
term.
The measurement of SES is particularly problematic for women, because previ-
ous typically used methods of assigning SES, such as allocating a woman the oc-
cupational class of her husband, are often inappropriate (Koskinen & Martelin,
1994; McDonough, Williams, House, & Duncan, 1999). The increasing labor
force participation rate of women, which is seeing dual earning as the most com-
mon form of family life (Bradley, 1998), now affords the potential to analyze pat-
terns of inequalities in health using women’s own occupational status, and in fact
demands that this be done (Arber, 1997; Baxter, 1991). Similarly, measures such
as income or occupation are often not appropriate for assigning SES to older

266
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
adults, who may be retired and/or on pensions (Daniel, 1984). The applicability of
the SES measures to the specific populations being studied is critical (Liberatos et
al., 1988). With older adults in particular, most research has used only limited indi-
cators (notably education or income). In one of the few studies to investigate the
use of multiple indexes of SES among older adults (Robert & House, 1996), it was
found that the association between financial assets and health remained until quite
late in life, and financial assets become a more important predictor of some mea-
sures of health than either education or income. This finding lends strength to
claims that relying on a single indicator of SES is problematic, because different
indicators of SES have different associations with specific health outcomes (e.g.,
Chandola, 2000). The use of several indicators, or of multidimensional methods of
assigning SES, has been recommended (Liberatos et al., 1988; Martikainen,
1995).
In a recent study that attempted to address these problems, a set of individ-
ual-based, age-, and gender-specific indexes for assigning SES was developed
(Mishra, Ball, Dobson, Byles, & Warner-Smith, 2001). Factor analysis of data
from the 1995 Australian National Health Survey (NHS; ABS, 1995) produced
consistent results that were interpreted in terms of five conceptually meaningful
domains or factors of SES: employment, income, migration, family unit, and ed-
ucation. A factor can be interpreted as a dimension or construct that is a con-
densed statement of the relations between a set of variables (Kline, 1994).
Results showed that age- and gender-specific SES scores based on these factors
had stronger associations with physical and mental health than either an
area-based index or the SES indicators developed in this study for middle-aged
(40- to 44-year-old) men, and applied to the other age and gender groups. These
results were interpreted as evidence that SES measures composed of social and
demographic items demonstrate important age- and gender-specific differences
that are relevant for health. However, further validation of the SES indexes, and
investigation of their predictive ability for a range of health outcomes, is re-
quired.
The baseline surveys for the Australian Longitudinal Study on Women’s
Health (also known as the Women’s Health Australia [WHA] project) presented
an opportunity to examine these issues as they relate to women. Four main hypoth-
eses examined in this article are:
1. The age-specific SES domains (employment, family unit, income, migra-
tion, and education) for women obtained for the NHS data will be replicated in the
WHA data.
2. Self-reported measures of health will differ among SES groups defined for
each of the SES domains, with women in low SES reporting poorer health out-
comes than higher SES women.

SOCIO-ECOMOMIC STATUS AND WOMEN’S HEALTH
267
3. The magnitude of SES differentials will vary among SES domains for differ-
ent aspects of health; specifically, it is hypothesized that physical health and health
care utilization will show stronger associations with economic domains and mental
health will show stronger associations with sociocultural domains.
4. The SES differentials in health measures will be smaller for the older cohort
than for the mid-aged cohort.
METHOD
The WHA project is a longitudinal study of factors affecting the health and well-be-
ing of three national cohorts of women who were aged 18 to 23 years (“young”), 45
to 50 years (“mid-aged”), and 70 to 75 years (“older”) at the time of Survey 1 in
1996. This study, which is designed to track the health of women over a period of up
to 20 years, will provide longitudinal data on health, health service use,
sociodemographics, and personal information from 41,500 women. Since Survey
1, the three age cohorts have been surveyed annually on a rolling basis. It was not
feasible to include the young cohort in this study, because education is known to be
an important indicator of SES, and many of the young cohort of women were still in
the process of acquiring educational qualifications.
Study Sample
The original WHA study sample was selected randomly from the national
Medicare health insurance database (which incorporates all residents of Australia
regardless of age, including immigrants and refugees). Women from rural and re-
mote areas of Australia are overrepresented in the sample. Further details of the re-
cruitment methods have been described elsewhere (Brown et al., 1998).
At Survey 1, in 1996, a total of 14,065 mid-aged women and 12,624 older
women responded to the mailed surveys. The mid age cohort was surveyed for the
second time in 1998, and the older cohort in 1999. There were two versions of Sur-
vey 2—a long version administered via mail (mid- aged: N = 11,637; older: N =
9,510) and a short version consisting of only selected questions, which was admin-
istered via telephone interview (mid-aged: N = 691; older: N = 920). The response
rates for Survey 2 were 92% (for the mid-aged) and 91% (for older women) of
those women who had consented at Survey 1 to further contact and had not subse-
quently died. The nonrespondents consisted of those who did not return Survey 2
(mid-aged: 6.5%; older: 7.6%) and those who declined to participate (mid-aged:
1.5%; and older: 1.8%). Women who responded to the short version of Survey 2
were excluded from the analyses because some of the variables relevant to this ar-

268
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
ticle were not collected in the short version of the survey. The sample for this study
consisted of the 11,637 mid-aged women and 9,510 older women who responded
to both surveys.
Measures
All of the demographic and socio-economic items in Surveys 1 and 2 were selected
for analysis (25 items for the mid-aged women and 18 items for older women). For
a few items, some categories have been collapsed due to small numbers of women.
Most of the demographic and socio-economic items used in the analyses were col-
lected in Survey 1. The items selected, and their response options, are outlined in
Table 1.
A range of health conditions and behaviors that have been established in previ-
ous studies to be associated with indicators of SES were selected from Survey 1 as
indicators of health.
1. Health conditions: Three specific health conditions that are known to be as-
sociated with SES were selected for inclusion in analyses. These conditions were:
having ever been told by a doctor they had hypertension (two response options), be-
ing told by a doctor they had diabetes (two response options), and experiencing
constipation in the last 12 months (four response options from never to often; re-
sponses of sometimes and often were used to estimate the prevalence).
2. Medical Outcomes Study Health Survey Short-Form (SF–36): Participants
completed the SF–36 (Ware, Kosinski, & Keller, 1994), a widely used and vali-
dated measure of health-related quality of life, separated into physical and mental
health component summary scores denoted as physical health (PCS) and mental
health (MCS), respectively.
3. Health care utilization: The use of general practitioner services has been
shown to be inversely related to SES levels (Young, Dobson & Byles, 2001). To
measure the number of times health services were utilized in the last 12 months,
respondents were asked, How many times have you consulted the following for
your own health in the last 12 months? … Family doctor or another general prac-
titioner; hospital doctor; specialist doctor, allied health professional; “alterna-
tive” health practitioner. The response options were: none, once or twice, three
or four times, five or six times, seven or more times. Responses were scored
based on approximate annual frequencies of health care utilization (none = 0;
once or twice = 1.5; three or four times = 3.5; five or six times = 5.5; seven or
more times = 8). These were summed over the five different types of providers
of health services to give an overall measure of health care utilization ranging
from 0 to 40.

SOCIO-ECOMOMIC STATUS AND WOMEN’S HEALTH
269
TABLE 1
Items and Response Options for the Items Included in Factor Analyses
Item
Order of Response Options (Number of Options)
Marital status
Country of birth
Never married/separated/divorced/widowed, married/defacto (2)
Other, Asia, Europe (Non-English speaking), Other English
speaking, Australia (5)
Year of arrival in Australia
Mid cohort: 1966 or later,1965 or earlier, Australian born (3)
Older cohort: 1956 or later,1955 or earlier, Australian born (3)
Non-English, English (2)
Usual language spoken at
home
Area of residence
Other rural/remote areas, large/small rural centres, capital
city/other metropolitan centres (3)
Employment status
Mid cohort only: No paid work (unemployed , studying, unpaid
voluntary work, sick, other), home duties, work without pay (eg
family business), Employedûpart time, employed ûfull time (5)
Usual hours worked each week Not applicable, 1û24, 25û40, 41 or more hours (4)
Whether in paid shift work Not applicable, yes, no (3)
Whether in paid work at night Not applicable, yes, no (3)
Occupation
Never had a paid job/other, manual workers/machine operators or
drivers, sales and personal service workers/clerks,
tradespersons/para-professional; professionals/managers or
administrators (5)
Partner’s occupation
Not applicable/never had a paid job/other, manual
workers/machine operators or drivers, Sales and personal
service workers/clerks, tradespersons/para-professional,
professionals/managers or administrators (5)
Source of income for
self/partner
Wage or salary
Mid cohort: not applicable, no, yes (3)
Business/farm/partnership Mid cohort: not applicable, no, yes (3)
Government pension or
allowance
Older cohort: not applicable, no, yes
Superannuation or other
private income
Older cohort: not applicable, no, yes
Gross personal income
per annum
Mid cohort: Not applicable/don’t know/not stated, $15,999 or less,
$16,000–36,999, $37,000 or more (4)
Gross personal income of Mid cohort: not applicable/don’t know/not stated, $15,999 or less,
partner per annum
$16,000–36,999, $37,000 or more (4)
Highest qualification
Mid cohort: No formal qualification, School certificate, Higher
school certificate, trade/apprenticeship/certificate/diploma,
bachelor degree/higher degree (5)
Older cohort: No formal qualification/School certificate, higher
school certificate,
Trade/apprenticeship/certificate/diploma/bachelor degree/higher
degree (3)
(CONTINUED)

270
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
TABLE 1
(CONTINUED)
Item
Order of Response Options (Numger of Options)
Age first left school
Mid cohort: not applicable/16 years or younger, 17 years or older
(2)
Older cohort: Not applicable/14 years or younger, 15–16 years, 17
years or older (3)
Self reported class
Whether has private hospital
insurance coverage
Whether has private health
insurance for ancillary
services
Mid cohort: don’t know/missing, working, middle/upper (3)
No, Yes (2)
No, Yes (2)
Type of dwelling
Name in the ownership/
purchasing/tenancy
agreement
Caravan/tent/cabin/houseboat/other, Flat/unit/apartment, house (3)
Not applicable/other family member/other, partner or spouse, self
with partner or spouse, or self with others, or just self (3)
Whether lived alone
Whether lived with partner or
spouse
Yes, No (2)
No, Yes (2)
Whether lived with own
children
No, Yes (2)
4. Height and weights: Self-reported height and weight were used to compute
body mass index (BMI = weight in kilograms/square of height in meters). BMI was
classified according to the Australian National Health and Medical Research Coun-
cil (1997) guidelines: underweight (<20.00 kg/m2), acceptable weight
(20.00-25.00 kg/m2), overweight (25.01–30.00kg/m2), and obese (>30.00kg/m2.
5. Health behaviors: Physical activity scores were derived from self-reported
frequency and intensity of leisure time physical activity. Scores were classified as:
none or low, moderate, or high level of physical activity (Brown, Mishra, Lee, &
Bauman, 2000). Cigarette smoking status was defined as nonsmoker or smoker.
Statistical Analysis
All demographic and socio-economic items were considered as ordinal variables.
For each item, response options were arranged in ascending order with respect to
SES.Aseparatecategorywascreatedfortheresponse“notapplicable.”Theorderof
the categories in each item are given in Table 1. With the sample stratified by age,
confirmatoryfactoranalysisusingthemethodofprincipalcomponentsandvarimax
rotation was performed on the demographic and socio-economic items. Items that

SOCIO-ECOMOMIC STATUS AND WOMEN’S HEALTH
271
cross-loaded on several factors or had loadings of 0.5 or less on all the factors were
subsequently eliminated. Inter-item reliability for each factor was assessed by
Cronbach’s coefficients for standardized variables. Kaiser’s measure of sampling
adequacywasusedtoquantifythedegreeofintercorrelationamongtheitemsandthe
appropriateness of factor analysis was also reported (Hair, Anderson, Tatham, &
Black, 1997). In addition, the factor structures were compared with the results from
thesamplesaftertheyhadbeenrandomlysplitintotwosubsamplesandtheanalyses
repeated on each half.
Factor scores for each of the resultant SES domains were grouped into tertiles
2
(or in some cases dichotomized). Univariate analyses (χ test) and the
Cochran–Armitage test for trend were used to compare percentages of women
reporting medical history, symptoms, and health behaviors by SES tertiles (low,
middle, and high SES). Multiple linear regression models were used to analyze
the relation between physical health (the outcome variable) and the four SES do-
mains simultaneously (four explanatory variables). The analyses were repeated
for the outcome variables of mental health and health care utilization. Means
and 95% confidence intervals (CI) were estimated for physical health, mental
health, and health care utilization variables using the least square means option
of the general linear models procedure of SAS (SAS Institute Inc., 1989).
Bonferroni corrections were used to reduce the effects of inflated Type 1 errors
due to multiple comparisons (Neter, Kutner, Nachtsheim, & Wasserman, 1996).
To estimate the magnitude of SES differentials, mean difference and 95% confi-
dence interval between the high and low tertiles were also calculated using
means option of the general linear models procedure.
RESULTS
Factor analysis confirmed four independent factors for the mid-aged cohort and
four independent factors for the older cohort. Table 2 sets out the factors, which to-
gether explain 66% and 69% of the variation in the data for the mid-aged and older
women, respectively. The main difference was that the employment domain was
the primary factor for the mid group; in the older group, this was replaced by the
family unit domain followed by the income domain, which consisted of items relat-
ing to health insurance and source of income. The education and migration domains
were the next most important factors for both cohorts.
Tables 3 and 4 show the relation between the SES domains and various
health outcomes for the mid-aged cohort. Because the distribution of factor
scores for the employment domain was bimodal, factor scores were
dichotomized to low and high groups. For other factors, tertiles of scores were
used, with the lowest tertile representing the most disadvantaged group. Physical
and mental health scores generally increased across increasing SES tertiles,

272

p
273

274

275

276
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
whereas health care utilization decreased for all domains. The most marked
change occurred between the lowest tertile and the other two. For specific health
measures shown in Table 4, there were significant linear trends across SES
tertiles for the employment, family unit, and education domains except for ciga-
rette smoking status (statistically significant for family unit and education only)
and physical activity level (significant for employment and education only).
Physical health differentials and health service utilization were largest for the
employment domain and mental health differentials were largest for the family
unit domain. Health behaviors and conditions differed most for the education
domain.
There were substantially different relations between SES domains and health
measures for the older cohort shown in Tables 5, and 6, with fewer statistically sig-
nificant differences and smaller absolute differences. The family unit domain did
not exhibit any significant associations with physical or mental health. The income
domain had significant associations with the tertiles for mental health and health
care utilization. As for the mid-aged cohort, physical and mental health generally
increased across the education domain tertiles. With few exceptions, the only evi-
dence of linear trends with specific health measures was found in the education do-
main. These tended to decline across education domain tertiles (with exceptions
being physical activity, which markedly increased, and cigarette smoking, which
was not significantly associated). The income domain showed significant differen-
tials for cigarette smoking and obesity.
DISCUSSION
Overall, the results from the confirmatory factor analysis supported the factor
structures derived from the NHS analysis with the exception of the housing do-
main, which did not appear in the WHA cohorts. This could be due to the different
questions on housing included in each study. Five SES domains were identified
overall, with four age-specific domains overlapping but differing slightly between
mid-aged and older women. Age differences were consistent with those found in
NHS analyses (Mishra et al., 2001). As expected, employment was not the primary
factor among the older cohort, as only a small proportion of that age group reported
that they were still in paid employment. Other income items such as gross personal
annual income rarely appeared in the factors. This may be due to the high level (ap-
proximately20%)of “don’tknow,”“don’twanttoanswer,”ormissingresponses.
The results also supported the second and third hypotheses; namely that the
SES groups would differ on health outcomes and that the magnitude of these SES
differentials would vary according to the SES domain and health measure exam-
ined. Consistent with a vast body of previous research, results provide evidence of
an inverse relation between SES and self-reported health. This study advanced

277

278

279

280
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
previous findings by showing that the magnitude of the SES–health associations
differed depending on SES domain and the specific health outcome, with mental
health differentials larger for economic aspects of SES (e.g., employment), but
physical health and health care utilization differentials larger for sociocultural as-
pects (e.g., family unit). The results also partly supported the final hypothesis that
SES differentials in health measures would be smaller for the older cohort than for
the mid-aged cohort. The main exception was that in the older cohort, education
was significantly associated with all the specific health measures, except health
care utilization. This may be attributable to the likelihood that education level is
stable in later adult life, unlike family unit or income, which are affected by wid-
owhood or ceasing employment. Hence older women may become a more homog-
enous group in terms of their family situation and income, whereas differentials in
education and associations with health outcomes remain. However, certain aspects
(e.g., mental health and health care utilization) were significantly associated with
either income or family unit. The implications of these findings for future research
into health inequalities are that different domains of SES may be useful for identi-
fying aspects of SES that are important for different health outcomes, but that at a
minimum, education should be included as a measure of the SES of older adults.
The migration domain was found to be associated with symptoms (diabetes, hy-
pertension, constipation) and health behaviors but not with self-rated physical and
mental health or health care utilization. This may reflect cultural differences in
perceptions of health. Further research into discrepancies between perceived and
objective health in migrant groups is warranted, because the findings have impli-
cations for the responsiveness of the health care system to the needs of migrant
women.
The fact that the relations between most of the SES domains and health out-
comes were significant and meaningful (in terms of size of effects) when included
in the regression models simultaneously (particularly for mid-aged women) dem-
onstrates that these SES domains are independent predictors of health outcomes.
In conjunction with findings that the magnitude of SES–health associations dif-
fered according to SES domain, these results suggest that different domains of
SES are independently associated with different health outcomes. Acknowledging
this study’s cross-sectional design, these findings may reflect different pathways
through which aspects of SES impact on health.
The findings support the validity of the measures developed in an earlier
study by Mishra et al. (2001), reinforcing the argument that SES domains differ
across age and gender groups and supporting arguments for the use of age- and
gender-specific indexes in studies of health inequalities. Such a focus on specific
population groups is particularly relevant for women. Women’s increasing labor
force participation, accompanied by growing numbers of female-headed families
and the decline of fertility rates, challenges the traditional assumptions that the
male breadwinner is the determinant of a household’s socio-economic position

SOCIO-ECOMOMIC STATUS AND WOMEN’S HEALTH
281
(Hayes & Jones, 1992). In the late 1980s, fewer women in their 50s than in their
40s were in paid work, and they tended to work from financial necessity (Arber
& Ginn, 1995). Research suggests that this may be changing and that many
women now in this age group are more ambitious than men of similar age
(Bradley, 1998). Many young Australian women perceive a combination of
motherhood and paid work to be the norm and aspire to such a lifestyle by the
age of 35 (Wicks & Mishra, 1997). Despite their increasing participation in the
paid workforce, women today still perform the majority of domestic and caring
work (Bittman, 1995). Women are likely to be engaged in multiple social roles
related to paid employment, family, child-care, and caring for others (Pugh &
Moser, 1990); their lives thus remain markedly different from those of men. The
need for gender-specific measures of SES, particularly those that take
sociocultural factors into account, is therefore critical. In addition, factors such
as delayed retirement and self-funded retirement will increasingly impinge on
the material well being of older women. The “complex mosaic” (Mowl &
Turner, 1995) of women’s lives as they move through different life stages calls
for more sensitive measures of SES to take account of this diversity.
Gender differences in health vary according to stage of the life cycle (Arber
& Cooper, 1999) and increasing evidence suggests that socio-economic factors
acting over the lifetime may have cumulative effects on health (Hart et al., 1998;
Smith, Hart, Blane, Gillis & Hawthorne, 1997). The longitudinal nature of the
WHA study affords the opportunity to further explore, over time, the marked
difference between the age groups in regard to the influence of paid employ-
ment. Will continued participation in paid work through their young and middle
years affect the SES of women after retirement? Will employment remain less
relevant to older women? Or is the association demonstrated in this study a gen-
erational effect specific to these cohorts, derived from the fact that the mid-aged
women have typically engaged in some form of paid work while the other older
group has not?
Associations between the type of family unit and women’s health require further
investigation. Evidence from British studies suggests that differences between
never-married women and divorced and separated women may vary by age and/or
birthcohort.AmongolderBritishwomen,divorcedandseparatedwomenmayhave
experienced more harmful health effects than never-married women; however,
among younger women, this difference may be absent or possibly reversed
(Waldron, Weiss, & Hughes, 1997). The longitudinal WHA study affords opportu-
nities to examine these associations over time for Australian women.
Strengths of this study include the large, representative population samples of
women surveyed and the comprehensive range of health indicators assessed. Limi-
tations include the fact that only those socio-economic and demographic items in-
cluded in the WHA surveys could be included in analyses. Other potentially
important proxies for SES, such as inherited wealth or material possessions, may

282
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
havealteredthefactorstructureiftheyhadbeenincluded.Inaddition,earlylifeSES,
whichemergingevidencesuggestsplaysakeyroleincontributingtohealthinequal-
ities throughout life (e.g., Power & Matthews, 1997; Wamala, Lynch, & Kaplan,
2001), was not assessed comprehensively in this study. The data presented are
cross-sectional, and hence results of the investigation described in this article dem-
onstrate associations only. However, again the longitudinal WHA study provides
great potential for future exploration of causal connections and the cumulative ef-
fects of SES on health over time, and in the context of significant life events. For ex-
ample, what is the effect of children leaving home? Does the removal of such a
demandonparentalresourcesmeanthathousingunitbecomeslessrelevantasacon-
tribution to SES?
These findings add to the growing body of evidence demonstrating links be-
tween low SES and poor health and indicate that these associations differ de-
pending on the domain of SES and the age group under study. The finding that
different SES domains were differently associated with health outcomes across
age groups suggests that underlying relations between SES and health are likely
to be complex and governed by different mechanisms depending partly on age
and aspect of health. Further research is required to untangle these complex pat-
terns and investigate the causal mechanisms between SES and health outcomes.
A better understanding of these causal pathways is critical to begin to address
SES-related health inequalities.
REFERENCES
Anderson, R. T., Sorlie, P., Backlund, E., Johnson, N., & Kaplan, G. A. (1997). Mortality effects of com-
munity socioeconomic status. Epidemiology, 8, 42–47.
Arber, S. (1997). Comparing inequalities in women’s and men’s health: Britain in the 1990s. Social Sci-
ence and Medicine, 44, 773–787.
Arber, S., & Ginn, J. (1995). The mirage of gender equality: Occupational success in the labour market
and within marriage. British Journal of Sociology, 46, 21–43.
Arber S., & Cooper, H. (1999). Gender differences in health in later life: The new paradox? Social Sci-
ence and Medicine, 48, 61–76.S
Australian Bureau of Statistics. (1995). National Health Survey: User’s guide (Catalogue No. 4363.0).
Canberra: Australian Bureau of Statistics.
Australian Bureau of Statistics. (1999). Labour force status and other characteristics of migrants, Aus-
tralia (Catalogue No. 6250.0). Canberra: Australian Government Publishing Service.
Austrialian National Health and Medical Research Council. (1997). Acting on Australia’s weight: A
strategic plan for the prevention of overweight and obesity. Canberra: Australian Government Pub-
lishing Service.
Baghurst, K. I., Record, S. J., Baghurst, P. A., Syrette, J. A., Crawford, D., & Worsley, A. (1990).
Sociodemographic determinants in Australia of the intake of food and nutrients implicated in cancer
aetiology. The Medical Journal of Australia, 153, 444–452.
Baxter, J. (1991). The class location of women: Direct or derived? In J. Baxter & M. Western (Eds.),
Class analysis and contemporary Australia (pp. 202–222). South Melbourne: MacMillan.

SOCIO-ECOMOMIC STATUS AND WOMEN’S HEALTH
283
Berkman, C. S., & Gurland, B. J. (1998). The relationship among income, other socioeconomic indica-
tors, and functional level in older persons. Journal of Aging and Health, 10, 81–98.
Bittman, M. (1995). Recent changes in unpaid work (Occasional Paper, ABS Catalogue No. 4154.0).
Canberra: Australian Bureau of Statistics.
Bradley, H. (1998). A new gendered order? Researching and rethinking women’s work. Sociology, 33,
869–873.
Brezinka, V., & Kittel, F. (1996). Psychosocial factors of coronary heart disease in women: A review.
Social Science & Medicine, 42, 1351–1365.
Broom, D. H. (1984). The social distribution of illness: Is Australia more equal? Social Science and
Medicine, 18, 909–917.
Brown, W. J., Bryson, L. B., Byles, J. E., Dobson, A. J., Lee, C., Mishra, G., & Schofield, M. (1998).
Women’s Health Australia: Recruitment for a national longitudinal cohort study. Women and
Health, 28, 23–40.
Brown, W. J., Mishra, G., Lee, C., & Bauman, A. (2000). Leisure time physical activity in Australian
women: Relationship with well-being and symptoms. Research Quarterly for Exercise and Sport,
71, 206–216.
Chandola, T. (2000). Social class differences in mortality using the new UK national statistics
socio-economic classification. Social Science and Medicine, 50, 641–649.
Crespo, C. J., Ainsworth, B. E., Keteyian, S. J., Heath, G. W., & Smit, E. (1999). Prevalence of physical
inactivity and its relation to social class in U.S. adults: Results from the Third National Health and
Nutrition Examination Survey, 1988–1994. Medicine and Science in Sports and Exercise, 31,
1821–1827.
Daniel, A. (1984). The measurement of social class. Community Health Studies, 8, 218–222.
Der, G., MacIntyre, S., Ford, G., Hunt, K., & West, P. (1999). The relationship of household income to a
range of health measures in three age cohorts from the West of Scotland. European Journal of Public
Health, 9, 271–277.
Elo, I. T., & Preston, S. H. (1996). Educational differentials in mortality: United States, 1979–1985. So-
cial Science & Medicine, 42, 47–57.
Evans, J. M. M., Newton, R. W., Ruta, D. A., MacDonald, T. M., & Morris, A. D. (2000). Socio-eco-
nomic status, obesity, and prevalenceofType1 andType2 diabetes mellitus. Diabetic Medicine, 17,
478–480.
Graham, H., & Hunt, S. (1994). Women’s smoking and measures of women’s socioeconomic status in
the United Kingdom. Health Promotion International, 9, 81–88.
Hair, J. F., Jr., Anderson, R. E., Tatham R. L., & Black, W. C. (1997). Multivariate data analysis with
readings (4th ed.). Upper Saddle River, NJ: Prentice Hall International Editions.
Hallqvist, J., Lundbert, M., Diderichsen, F., & Ahlbom, A. (1998). Socioeconomic differences in risk of
myocardial infarction 1971–1994 in Sweden: Time trends, relative risks, and population attributable
risks. International Journal of Epidemiology, 27, 410–415.
Hart, C. L., Davey Smith G., & Blane D. (1998). Inequalities in mortality by social class measured at
three stages of the life course. American Journal of Public Health, 88, 471–474.
Hayes, B., & Jones, F. (1992). Class identification among Australian couples: Are wives’ characteristics
relevant? British Journal of Sociology, 43, 463–484.
Hemingway, H., Nicholson, A., Stafford, M., Roberts, R., & Marmot, M. (1997). The impact of socio-
economic status on health functioning as assessed by the SF–36 questionnaire: The Whitehall 2
study. American Journal of Public Health, 87, 1484–1490.
House, J., Lepkowski, J., Kinney, A., Mero, R. Kessler, R., & Herzog, A. (1994). The social stratifica-
tion of aging and health. Journal of Health and Social Behaviour, 35, 213–34.
Johanson, J. F., & Sonnenberg, A. (1990). The prevalence of hemorrhoids and chronic constipation: An
epidemiologic study. Gastroenterology, 98, 380–386.
Kline, P. (1994). An easy guide to factor analysis. London: Routledge.

284
MISHRA, BALL, DOBSON, BYLES, WARNER-SMITH
Koskinen, S., & Martelin, T. (1994). Why are socioeconomic mortality differences smaller among
women than among men? Social Science and Medicine, 38, 1385–1396.
Kreiger, N. (2000). Refiguring “race”: Epidemiology, racialized biology, and biological expressions of
race relations. International Journal of Health Services, 30, 211–216.
Kreiger, N., Williams, D., & Moss, N. (1997). Measuring social class in U.S. public health research:
Concepts, methodologies, and guidelines. Annual Review of Public Health, 18, 341–378.
Liberatos, P., Link, B. G., & Kelsey, J. L. (1988). The measurement of social class in epidemiology. Epi-
demiology Review, 10, 87–122.
Mackenbach, J. P. (1992). Socio-economic health differences in the Netherlands: A review of recent
empirical findings. Social Science & Medicine, 34, 213–226.
Martikainen, P. (1995). Mortality and socio-economic status among Finnish women. Population
Studies, 49, 71–90.
McDonough, P., Williams, D. R., House, J. S., & Duncan, G. J. (1999). Gender and the socioeconomic
gradient in mortality. Journal of Health and Social Behaviour, 40, 17–31.
McElduff, P., & Dobson, A. J. (2000). Trends in coronary heart disease: Has the socio-economic differ-
ential changed? Australian and New Zealand Journal of Public Health, 24, 465–473.
Mishra, G. D., Ball, K., Dobson, A. J., Byles, J. E., & Warner-Smith, P. (2001). The measurement of so-
cioeconomic status: Investigation of gender- and age-specific indicators in Australia: National
Health Survey ‘95. Social Indicators Research, 56, 73–89.
Mowl, G., & Towner, J. (1995). Women, gender, leisure, and place: Towards a more humanistic geogra-
phy of women’s leisure. Leisure Studies, 14, 102–116.
Mustard, C. A., Derksen, S., Bethelot, J. M., Wolfson, M., & Roos, L. L. (1997). Age-specific education
and income gradients in morbidity and mortality in a Canadian province. Social Science and Medi-
cine, 45, 383–397.
Neter, J., Kutner, M. H., Nachtsheim, C. J., & Wasserman, W. (1996). Applied linear statistical model.
Chicago: Irwin.
Osler, M., Gerdes, L. U., Davidsen, M., Bronnum-Hansen, H., Madsen, M., Jorgensen, T., & Schroll, M.
(2000). Socioeconomic status and trends in risk factors for cardiovascular diseases in the Danish
MONICA population, 1982–1992. Journal of Epidemiology & Community Health, 54, 108–113.
Power, C., & Matthews, S. (1997). Origins of health inequalities in a national population sample. Lan-
cet, 350, 1584–1589.
Pugh, H., & Moser, K. (1990). Measuring women’s mortality differences. In H. Roberts (Ed.), Women’s
health counts (pp. 93–112). London: Routledge.
Robert, S., & House, J. (1996). SES differentials in health by age and alternative indicators of SES. Jour-
nal of Aging and Health, 8, 359–388.
Sarlio-Lahteenkorva, S., & Lahelma, E. (1999). The association of body mass index with social and eco-
nomic disadvantage in women and men. International Journal of Epidemiology, 28, 445–449.
SAS Institute Inc. (1989). SAS/STAT user’s guide (Version 6, 4th ed., Vol. 2). Cary, NC: Author.
Scambler, G., & Higgs, P. (1999). Stratification, class, and health: Class relations and health inequalities
in high modernity. Sociology, 33, 275–288.
Smith, G. D., Hart, C., Blane, D., Gillis, C., & Hawthorne, V. (1997). Lifetime socioeconomic position
and mortality: Prospective observational study. British Medical Journal, 314, 547–552.
Sobal, J., & Stunkard, A. J. (1989). Socioeconomic status and obesity: A review of the literature. Psy-
chological Bulletin, 105, 260–275.
Tyroler, H. A. (1999). The influence of socioeconomic factors on cardiovascular disease risk factor de-
velopment. Preventive Medicine, 29(Suppl. S), S36–S40.
Vargas, C. M., Ingram, D. D., & Gillum, R. F. (2000). Incidence of hypertension and educational attain-
ment: The NHANES I epidemiologic follow-up study. First National Health and Nutrition Exami-
nation Survey. American Journal of Epidemiology, 152, 272–278.

SOCIO-ECOMOMIC STATUS AND WOMEN’S HEALTH
285
Waldron, I., Weiss, C. C., & Hughes, M. E. (1997). Marital status effects on health: Are there differences
between never married women and divorced and separated women? Social Science & Medicine, 45,
1387–1397.
Wamala, S. P., Lynch, J., & Kaplan, G. A. (2001). Women’s exposure to early and later life socioeco-
nomic disadvantage and coronary heart disease risk: The Stockholm Female Coronary Risk Study.
International Journal of Epidemiology, 30, 275–284.
Ware, J. E., Kosinski, M., & Keller, S. D. (1994). SF–36 physical and mental health summary scales: A
user’s manual. Boston: The Health Institute, New England Medical Center.
Weber, M. (1946). Class, status, and party. In H. Gerth & W. W. Mills (Eds.), From Max Weber: Essays
in sociology (pp. 180–195). New York: Oxford University Press.
Wicks, D., & Mishra, G. (1997). Young Australian women and their aspirations for work, education,
and relationships. In E. Carson, A. Jamrozik, & T. Winefield (Eds.), Unemployment: Economic
promise and political will (pp. 89–100). London: Academic.
Young, A. F., Dobson, A. J., & Byles, J. E. (2001). Determinants of general practitioner use among
women in Australia. Social Science and Medicine, 53, 1641–1651.