Journal of Gerontology: SOCIAL SCIENCES
1999, Vol. 54B, No. 6, S329-S338
Copyright 1999 by The Geruntological Society ofAmerica
Multiple Roles and Well-Being Among Midlife Women:
Testing Role Strain and Role Enhancement Theories
Jennifer Reid and Melissa Hardy
Department of Sociology and Pepper Institute on Aging and Public Policy, Florida State University, Tallahassee.
Objective. Research on women's multiple roles frequently adopts one of two perspectives: role strain, which argues that
assuming multiple roles is detrimental to mental well-being, or role enhancement, which argues that engaging in multiple roles
enhances mental well-being. We argue that the relationship between role occupancy and well-being is manifested through multiple dimensions of role experiences. We investigate the association between depressive symptomatology and various dimensions of the roles of wife, mother, paid worker, and informal caregiver to aging parents.
Methods. Data are from the 1992 wave of the Health and Retirement Study. Depressive symptomatology, measured by a
subset of the CES-D scale, is the dependent variable. To assess the robustness of findings relative to different functional forms
of the dependent variable, we estimate multiple regression, log-linear regression, and multinomial logit models. Independent
variables include demographic characteristics, measures of role occupancy, role demands, and role satisfaction.
Results. Although the number of roles women assume affects their reports of depressive symptoms, once the demand and
satisfaction associated with these roles is controlled, number has no effect; that is, the effect of the number of roles is indirect.
Discussion. Our results highlight the importance of women's perceptions of the quality of their roles in relation to their
overall well-being. Future investigations of women's multiple roles should examine how roles may provide rewards, impose
constraints, or generate conflict, as well as the extent to which the willingness to assume multiple roles and the reported levels
of role satisfaction and mental well-being may be jointly endogenous.
T
HAT women spend much of their adult lives juggling the
demands of multiple roles is not news. Because the roles
that women occupy depend, in part, on age and family structure, roles common to women of one age may be much less
common and, perhaps, much less enjoyable for women of different ages. As more people survive to old age and live longer
as elders, more women find themselves assuming the role of
caregiver to multiple generations of family members while they
continue in their roles as wives and paid workers. Managing
multiple roles can be both stressful and satisfying. Even very
demanding roles can be associated with enhanced well-being if
they reinforce a positive sense of identity. Studies that examine
the impact of women's multiple roles on their well-being generally explore the roles of wife, mother, and paid worker, but
exclude the role of informal caregiver to aging adults. In addition, although role occupancy and role demands are generally
included, role quality often is not.
How are women's experiences in multiple roles related to
their well-being in midlife? Recent cohorts of women—particularly married women and women with children—exhibit
higher rates of labor force participation than previous generations (Moen, 1992). Women also constitute the majority of informal caregivers to elderly people (Brody, Kleban, Johnsen,
Hoffman, & Schoonover, 1987; Guberman, Maheu, & Maille,
1992; Kramer & Kipnis, 1995; Levinson, 1996; Mui, 1995).
Thus, midlife women are increasingly likely to be engaged in
various types of paid and unpaid labor. The obligations and expectations that accompany these roles may be incompatible,
negatively affecting women's well-being and resulting in symptoms of depression. On the other hand, roles may be reinforcing
and rewarding, leading to enhanced well-being.
Two perspectives central to role theory—role strain and role
enhancement—predict different outcomes for women occupying multiple roles. Whereas a role strain perspective suggests
that multiple roles can make women feel overburdened, thereby
having a detrimental effect on mental and physical well-being
(Goode, 1960; Mui, 1995; Pearlin, 1989; Young & Kahana,
1989), a role enhancement perspective argues that multiple
roles improve mental well-being (Brody, Litvin, Hoffman, &
Kleban, 1992; Moen, Robison, & Dempster-McClain, 1995;
Parris Stephens & Franks, 1995; Stoller & Pugliesi, 1989).
Women's roles outside the home may put them in touch with
more emotional and economic resources and temper their conflicting responsibilities. Having more outlets for the release of
tension and more sources of support may be beneficial for
women. Furthermore, a difficulty in one role can be offset by
satisfactions in another (Baruch & Barnett, 1986; Crosby, 1991;
Parris Stephens & Franks, 1995).
Empirical studies often address the number of roles a woman
occupies, but multiple roles are not uniformly beneficial or
detrimental. The quality of those roles can also affect wellbeing (Baruch & Barnett, 1986; Parris Stephens, Franks, &
Townsend, 1994). In fact, satisfaction with a role may have a
greater effect on well-being than actual role demands (Baruch
& Barnett, 1986). Here we examine how women's mental wellbeing is linked to the number, demands, and satisfaction with
the multiple roles they occupy.
Role strain or role enhancement.—The role strain perspective emphasizes the potential negative consequences that the
stress and competing demands of multiple roles may have for
individuals (Goode, 1960; Mui, 1992). Given a limited amount
S329
S33O
REID AND HARDY
of time and energy, adapting to multiple roles may adversely
affect mental well-being (Brody, 1990) and compromise friendships and other emotional resources (Stoller & Pugliesi, 1989).
Because relationships formed through sets of roles tend to be
stable and important to individuals, stress can increase when
these relationships become problematic (Pearlin, 1983, 1989).
Chronic role strains applicable to women in this study include
role overload (e.g., when demands exceed capacities, common
to informal caregivers), interrole conflict (e.g., when demands
of multiple roles are incompatible), role captivity (e.g., when an
individual takes on a role unwillingly), and role restructuring
(e.g., when adult children assume increasing responsibility for
their parents). These changes in relationships can result in feelings of betrayal, status loss, and the violation of expectations.
According to the role enhancement perspective, the accumulation of multiple roles can increase social integration leading
to an increase in "power, prestige, resources, and emotional
gratification, including social recognition and a heightened
sense of identity" (Moen et al., 1995, p. 260). A corollary of
this perspective is that too few roles may be detrimental to mental well-being. Women who are not employed outside the home
may experience stress from isolation, have fewer outlets for the
release of tension, and be less satisfied with their lives (Oakley,
1974). In general, the demands of multiple roles tend to be offset by the support received (Brody et al., 1992; Parris Stephens
& Franks, 1995; Scharlach, 1994).
Experiences in one role may affect experiences in another
through a "role spillover" effect (Parris Stephens & Franks,
1995). For example, for an employed woman with a fulfilling
job but a difficult informal caregiving relationship, the positive
employment experience may make the caregiving easier, or her
negative caregiving experience may interfere with her paid
work. Parris Stephens and colleagues (1994) examine role quality in terms of the stresses and rewards and argue that although
the accumulation of stress across roles may be detrimental to
women's well-being, the accumulation of rewards across roles
may be beneficial. Role history—an individual's accumulated
experience in a role—also suggests that some life events may
be nonproblematic or even beneficial to individuals by offering
an escape from a chronically stressful role situation (Wheaton,
1990).
Mental well-being, central to both perspectives, has been
linked to life satisfaction, on the one hand, and to depressive
symptoms, on the other. Well-being and distress (which may be
manifested as depression) may be conceptualized as opposite
ends of a continuum (Mirowsky & Ross, 1989); therefore, tests
of role theories can examine the correlates of either well-being
or distress to establish the connections between role occupancy
and mental health outcomes.
Women's roles at midlife.—Some studies of women's roles
at midlife emphasize changes within a role that occur in
midlife; others examine the interaction between roles. One
midlife transition that affects the mother role is the departure of
children from the home, often referred to as the "empty nest"
syndrome (Levinson, 1996). Some women may become depressed when their children leave home, while others experience a sense of liberation from the maternal role (Apter, 1995;
Levinson, 1996). Although most midlife women no longer have
young children at home, an increasing number of older children
have been returning to live at home (Aquilino & Supple, 1991).
For women who enjoyed freedom from the maternal role, the
return of adult children to the parental household may cause
tension. The quality of the relationship with the adult child is
central to mothers' satisfaction with the living arrangements
(Aquilino & Supple, 1991).
The effects of being an employee, a wife, and a caregiver on
well-being depends, in part, on the conditions under which the
roles are enacted. Employed women are less distressed than
women who are exclusively homemakers, but more distressed
than employed men (Mirowsky & Ross, 1989). However, little
difference persists between homemakers and employed women
once the working conditions of the roles (e.g., security, complexity, autonomy, discretion) are controlled (Lennon, 1994).
Married people score higher than unmarried people on measures of well-being, although the difference is greater for men
than women (Baruch & Barnett, 1986; Lee, Seccombe, &
Shehan, 1991; Mirowsky & Ross, 1989; Nolen-Hoeksema,
1990). However, marriages characterized by inequality, conflict, and a lack of consideration may decrease well-being
(Mirowsky & Ross, 1989). The subjective experience of the
wife role or an unequal division of labor within the household
may also affect marital satisfaction (Glass & Fujimoto, 1994).
Finally, caregiving can complement or interfere with other role
obligations. Role strain may result from trying to balance caregiving for elderly parents with career and family responsibilities (Lerner, Somers, Reid, Chiriboga, & Tierney, 1991; Mui,
1995; Young & Kahana, 1989) and may be exacerbated when
high levels of assistance are necessary (Stoller & Pugliesi,
1989). Caregiving may also negatively affect performance at
work, increasing absenteeism, tardiness, work interruptions,
missed advancement opportunities, and thus increasing job
stress (Pearlin, 1989; Scharlach, 1994).
Positive aspects associated with performing multiple roles
include enhanced self-esteem, identity strengthening, informal
support networks, and improved access to material and social
resources (Brody, 1985; Parris Stephens & Franks, 1995;
Stoller & Pugliesi, 1989). In fact, many married women caregivers find that the esteem and effectiveness they experience in
this role benefits their relationships with their husbands (Parris
Stephens & Franks, 1995). Married women caregivers also
have more socioeconomic and instrumental support, and financial resources; they report less depression than unmarried caregivers (Brody et al., 1992). Employment provides increased income to informal caregivers and can lead to an improved
relationship between the caregiver and care recipient
(Scharlach, 1994). In sum, a sense of accomplishment and extended interpersonal relationships can overshadow the negative
aspects of combining caregiving with employment.
Both the role strain and role enhancement perspectives lack
dimensions that allow for the complexity of women's lives as
they balance multiple roles. Studies of women's multiple roles
display a mixture of methodologies; their samples are often
small and unrepresentative, and gender is not always controlled.
In addition, the range of roles in which women engage is often
limited to two or three, with informal caregiver and employee
being the most frequently excluded. For these reasons, the relationship between multiple roles and well-being is an open question. Although the number of roles coupled with the demands
and quality of each role are all implicated in well-being, most
MULTIPLE ROLES AND WELL-BEING
research focuses primarily on the number of roles women assume and the compatibility of those roles, but excludes enjoyment of those roles and their implied relationship. We build on
previous research [e.g., Baruch & Barnett (1986), Parris
Stephens et al. (1994), Parris Stephens & Franks (1995)], as we
examine the interrelationship between women's roles using
measures of role occupancy, role demand, and role satisfaction.
In this article, we assess whether the effects of multiple roles
are additive or interactive; that is, are certain combinations of
roles more strongly linked to well-being or distress? Second,
we test whether the relationships between well-being and various role dimensions are direct or indirect.
METHODS
Data.—Our data are from the 1992 wave of the Health and
Retirement Study (HRS), which provides a nationally representative sample of men and women aged 51-61 and their spouses
or partners. Cases were drawn from a sample of 70,000 households that were screened for age-eligible respondents. The sample includes 12,654 respondents from 7,703 households. The
survey was designed to study labor force participation and pensions, health conditions and health status, family structure and
transfers, and economic status (Burkhauser & Gertler, 1995).
From these data, we analyze age-eligible women (5,116), as the
out-of-age-range spouses are not representative of their age
groups. The survey oversamples African Americans, Hispanics,
and Florida residents. Our analyses are based on the personweighted data to readjust for the differential sampling weights.
Given that the average age of female adult-child caregivers is
52.4 (Stone, Cafferata, & Sangl, 1987), HRS respondents are
an appropriate sample in which to include informal caregiver as
an optional role. Women in this age group have parents who are
likely to be experiencing chronic and acute illness that may require informal caregiving or who may need help with activities
of daily living (ADLs). Therefore, they occupy the midlife age
range in which the multiple roles of wife, worker, parent, and
caregiver are likely to overlap.
However, we also face data limitations. HRS includes measures of satisfaction for the paid worker and wife roles; it does
not offer measures of satisfaction specific to the parental or informal caregiving roles. However, a measure of satisfaction with
"family life" is available. Clearly, the roles of wife, mother, and
parental caregiver all involve family life and relationships. Given
the separate measure of marital satisfaction (which should partial out this dimension of family life for all wives), we are left
with a measure that is relevant to both the role of parent and the
role of caregiver, as well as to other potential family connections. Also, because this study is cross-sectional, it is not possible to establish direct causal relationships between the independent and dependent variables, nor can we control for preexisting
levels of depressive symptoms. We can, however, investigate the
correlative linkage between women's experiences of their multiple roles and their reported symptoms of depression.
Independent variables.—The analyses include four categories of independent variables that are described in Table 1. A
few points should be noted. First, the informal caregiver role
distinguishes women who assisted their own or their spouse's
parents with ADLs such as bathing, eating, and dressing in the
S331
past year from those who did not. This definition of informal
caregiving does not include activities such as transportation,
shopping, or cooking. Thus, the informal caregivers in this analysis are involved in the most intensive hands-on caregiving activities, though the time they spend engaged in these activities
varies widely. We include measures of the demand associated
with the roles of paid worker, mother, and informal caregiver,
but not wife. For both paid work and caregiving, we created
dummy variables indicative of levels of demand consistent with
natural cutpoints in the distribution; the omitted (reference) categories are the highest levels of demand. By treating the continuous variables for hours at work and hours of informal care as
sets of dummy variables, we are better able to capture nonuniform effects across levels of demand. Measures of role demands and role quality apply only to women who occupy those
roles. The distributions on these independent variables are
therefore a combination of zeros (for women not in the role)
and observed values (for women occupying the role). The interpretation of these "embedded variables" is discussed in the
Analysis Plan section.
Dependent variable.—The dependent variable for this analysis is the respondent's score on a 14-item scale that is a subset
of the Center for Epidemiological Studies-Depression (CESD) 20-item scale; it serves as a measure of depressive symptomatology. Depressive symptomatology has been found to be
a fit indicator of distress for women. It is important to distinguish here between distress and depression and their different
manifestations for men and women. Measures of alcohol consumption and other outward displays of depressive symptoms
may be more suitable for men than women, because alcoholism, for example, has been suggested as a masculine form
of depressive behavior (Nolen-Hoeksema, 1990; Pearlin, 1989).
The CES-D was designed to measure symptoms of depression
in the general population. Its purpose is not to diagnose depression but rather to assess the level of symptoms of depression
with an emphasis on depressed mood (Radloff, 1977). Thus,
we employ the 14-item scale not as a diagnostic tool, but as an
indicator of distress. Many researchers use this shorter version
of the CES-D scale and have found it to be highly internally
consistent (Wallace & Herzog, 1995). The scale has an alpha of
0.88 for the present sample and is therefore considered a reliable measure of depressive symptomatology.
Analysis plan.—We estimated ordinary least squares (OLS)
and multinomial logit regression models with different specifications of the dependent variable (raw score, logged score, binary, multinomial) to test the strength of our findings under
varying sets of assumptions. In each case, we examined the net
effects of the independent variables in stages, moving from less
to more complex specifications, that is, from basic demographic
characteristics, to role occupancy (wife, mother, paid worker,
and informal caregiver to parents), role demands (for mother,
paid worker, and informal caregiver), and finally role quality.
The combination of the first and second models can be written
as:
= a + pd Demogj + 7,Rolesr + ei;
(1)
where Y; is the raw score for the 14-item CES-D scale, a is the
constant, (3d are the coefficients for the "d" demographic char-
REID AND HARDY
S332
Table 1. Independent Variables, Variable Names, and Coding
Variable Names
Independent Variables
Coding
Demographics
Socioeconomic status
Education
Age
Race
Physical health
Logged household income
Education (years of school completed)
Age
Race
Fair or poor health
0-17
51-61
1 = White, 0 = non-White (including Hispanics)
1 = fair/poor health, 0 = good, very good, or excellent
Role Occupancy
Wife
Paid worker
Informal caregiver to parents
Mother
Married
Employed
Caregiver
Children >18 in household
1 = married, 0 = unmarried
1 = working for pay, 0= nonemployed
1 = provided care, 0 = did not
1 = children >18 in household, 0 = no children >18 in household
Paid Work Dummy 1
Paid Work Dummy 2
Paid Work Dummy 3
(reference category)
Caregiving Dummy 1
Caregiving Dummy 2
Caregiving Dummy 3
Caregiving Dummy 4
Caregiving Dummy 5
(reference category)
No. children >18 living at home
1-19 hours
20-39 hours
40 hours
>40 hours
6-96 hours
97-168 hours
169-365 hours
366-700 hours
701-1277 hours
>1277 hours
1-6
Role Demand
Paid worker
Informal caregiver
Mother
Role Satisfaction
Marital satisfaction
Job satisfaction
Family satisfaction: Caregiver
Family satisfaction: Child > 18
1-5
1-5
1-5
1-5
acteristics, and 7 r are the coefficients for the "r" role variables.
Taking each category in turn, the coefficients in £d estimate the
net difference of an additional year of age or an additional year
of schooling on the expected CES-D score; the coefficient for
race estimates the net difference in average CES-D scores for
Whites and non-Whites, and the coefficient for logged household income estimates the net difference of a percentage change
in income in the expected CES-D score. Because the role occupancy variables are dummy-coded, the coefficients in -yr can be
thought of as increments or decrements to the intercept, showing how role occupants differ in their average CES-D scores
from those not in the role when other variables are controlled.
The full model adds embedded measures of demand and role
quality. It can be written as:
=a +
+ 7rRolesr
j + 8kQualityk + s,, (2)
where a, f3d, and 7r are as before, 8j are the coefficients for the j
role demand variables, and 8k are the coefficients for the k role
quality measures. These remaining independent variables are
examples of embedded variables in that they are constructed so
that the coefficients refer only to the women engaged in those
roles. For example, women who are not employed are coded
zero for the employee role and, therefore, also coded zero on all
the dummy variables for hours worked, and zero for job satisfaction. The same situation holds for women who are not caregivers. For married women, the distribution on marital satisfaction reflects this same logic; however, we have no measure of
marital demand. By controlling for role occupancy, we limit the
(5 = highest
(5 = highest
(5 = highest
(5 = highest
satisfaction)
satisfaction)
satisfaction)
satisfaction)
coefficients for role demand and role quality to those women
who occupy the role. For example, the coefficients for the
dummy variables indicating hours of work estimate how different work regimens increase or decrease the expected CES-D
score among women who work. Similarly, the coefficient for
job satisfaction describes the net effect of role quality on wellbeing for working women.
RESULTS
Descriptive statistics.—Thefirstcolumn of Table 2 displays
the mean values for all variables. The typical respondent was a
White, 56-year-old woman with a high school diploma. The
majority of women were married and employed. Six percent of
the sample were informal caregivers to aging parents who averaged between 2 and 3 hours caregiving per day. Women who
were employed worked on average almost 40 hours per week.
One third had children over 18 years old at home, with one
child over 18 as the average. Four out of five women reported
being in good to excellent health. Married women were generally satisfied with their marriages (only 10% scored 3 or
below), and employed women expressed somewhat less satisfaction with their jobs (15% scored 3 or below).
Given the prominence of employment and marital status in
the literature, we report within-category means by both employment status and marital status. The second and third
columns of Table 2 display the group means for employed and
nonemployed women. Although employed and nonemployed
S333
MULTIPLE ROLES AND WELL-BEING
Table 2. Descriptive Statistics for Employed vs Nonemployed and Married vs Unmarried Women (A^= 5116)
Overall M
(SD)
Employed
Age 51-61 years
Child >18 in household
No. children >18 in household8;
1-6 children (n = 1842)
Caregiver
No. hours spent caregivingb (n = 314)
Fair or poor health
White
Logged household income
Years of education; 0-17 years
Married
(n = 3426)
Unmarried
(n =1690)
t Value
.58
(.49)
.68
(.47)
55.92
(3.16)
.34
(.48)
56.12
(3.23)
2.47*
.34
(.47)
.118
.06
(.24)
.05
(.23)
.17
(.38)
.86
(.34)
.29
(.45)
.71
(.46)
18.08***
10.69
(.78)
34.74***
18.48***
12.35
(2.60)
9.86
(.98)
12.04
(3.16)
7.96
(5.81)
10.10
(7.12)
11.45***
.38
.23
.15
.24
.26
.20
.16
.37
.67
(.47)
.63
(•48)
55.98
(3.19)
.34
(.48)
55.69
(3.14)
.73
(-45)
56.44
(3.21)
.36
(.48)
.33
(.47)
2.57**
7.79***
.06
(.24)
886.99
(1320.01)
.06
(.23)
.07
(.25)
1.80
.21
(.41)
.12
(.32)
.83
(.38)
.35
(.48)
.79
(.41)
10.14
(1.08)
11.44
(2.98)
8.43***
o, 33***
1.35
(.64)
.81
(•39)
10.41
(.94)
12.25
(2.80)
10.58
(.79)
12.76
(2.56)
21.81***
3.66***
1.50
10.84***
14.48***
3 91***
4.63
(.97)
3.95
(1.43)
Family satisfaction: caregiver1' 1-5:
5 = highest satisfaction (n = 314)
1.46
(.84)
Family satisfaction: child >18a 1-5:
5 = highest satisfaction (n =1842)
1.43
(.79)
CES-D (groups)
0-5
6-8
9-11
12+
t Value
36.91
(11.37)
Marital satisfaction 1-5:
5 = highest satisfaction
Job satisfaction 1-5:
5 = highest satisfaction
CES-D scale
Not Employed
(n = 2037)
.61
(.49)
Paid work hours per week, 1^40 hours
Married
Employed
(n = 3079)
8.68
(6.36)
7.81
(5.14)
10.05
(7.42)
.34
.22
.15
.28
.38
.24
.15
.23
.29
.19
.14
.37
12.19***
"Among those with children >18 in the household.
b
Among those who assume the informal caregiving role.
*p < .05; **p < .01; ***p < .001.
women are equally likely to be caregivers, nonemployed caregivers spend significantly more hours providing care than employed women. Nonemployed women are also more likely to
be married, non-White, somewhat older, in poorer health, have
less schooling, and lower household incomes. Unmarried
women, compared to married women, are significantly more
likely to be employed, older, in poorer health, non-White, have
lower socioeconomic status, and higher CES-D scores (see
columns 5,6,7 of Table 2).
For all women in the sample, the mean score on the CES-D
subset is 8.68, which is a low level. Nonemployed women average somewhat higher levels of symptoms than employed
women, as do unmarried women compared to married women.
At the bivariate level, these findings support a relationship between marriage and employment, on the one hand, and better
physical and mental health, on the other. However, these findings cannot distinguish between two competing hypotheses:
Does marriage or employment enhance women's health, or do
healthier women tend to be married and/or employed?
Regression results.—Thefirstregression model estimates the
net coefficients for basic demographic characteristics. The coefficients and their standard errors are found in thefirstcolumn of
Table 3. This initial model explains approximately one fifth of
S334
REID AND HARDY
the variance in depressive symptoms. Years of school and age
have somewhat modest net effects on depressive symptoms—a
one- to two-tenth decrement for each additional year. Household income is also negatively associated with depressive
symptoms and registers a relatively strong effect. CES-D scores
for women in fair or poor health are almost 6 points higher than
scores for healthy women. Race, however, is not a significant
net predictor of CES-D scores.
The second regression model explores how assuming different roles relates to expected CES-D scores. These coefficients
are found in the second column of Table 3. Once we add measures for the various types of roles women can assume, the explanatory power of the model increases only marginally, explaining 22% of the variance in CES-D scores. Nevertheless,
when we control for demographic differences, these roles are
significantly related to the level of depressive symptoms.
Women who have children over 18 at home tend to have
marginally higher CES-D scores than women who do not, and
women who are caregivers tend to have higher CES-D scores
than noncaregivers. In contrast, married women have lower
CES-D scores than nonmarried women, and employed women
have lower CES-D scores on average than nonemployed
women.
To further explore the relationship between marital status
and employment status, we include an interaction term to determine whether the positive effect of employment status is the
same regardless of marital status. In fact, it is not. The negative
effect of employment on CES-D scores is larger for nonmarried
women than married women. Similarly, marital status reduces
CES-D scores for nonemployed women more than for employed women. Overall, women who are either married or employed, or both married and employed, have similar scores, indicating significantly lower CES-D scores than for women who
occupy neither role. Hypothesized interactions between other
combinations of roles were not supported, indicating that, with
the exception of wife and paid worker, the effects of other roles
on predicted CES-D scores were uniform.
Including the role indicators in this model had little effect on
the estimates for education, age, and health, and somewhat reduced the effect for household income (which was probably a
function of adding employment status and marital status, because they are also related to household income). The largest
change occurred in estimated racial differences. When we compare White and non-White women with the same demographic
characteristics who occupy the same roles, White women have
somewhat higher CES-D scores than non-White women. The
effect is small but nonetheless significant.
The third regression model explores the demand associated
with three of the four roles in Model 2. Rather than entering
hours spent caregiving and hours of paid work as interval measures, these distributions were classified into categories of demand so we could test for nonlinearities in their effects. The
third column of Table 3 reports these results. In this model, the
demand variables act as embedded measures. Because they
only apply to those who occupy the role, the demand variables
are essentially interaction terms that measure the effects of
higher levels of demand on CES-D scores among those who
perform the role (Cohen & Cohen, 1983).
As in Model 2, we are interested in comparing those who are
not caregivers with those who provide care. However, by
adding the role demand dimension, we elaborate this comparison. We now compare those who are not caregivers to several
different groups of caregivers providing different amounts of
care. When we hold other variables constant, 7cg = -.70 estimates the net difference in CES-D scores, comparing those not
providing care to women providing care at the highest level. In
Model 3, therefore, the net group differences depend on the
level of demand. For remaining levels of caregiving, the net
group difference is the sum of the coefficients for the dummy
variable caregiving (-.6944) and the coefficient for the specified level of hours spent caregiving (8j). For example, between
noncaregivers and caregivers providing between 400 and 700
hours of care, the net difference on their expected symptom
scores is -.70 + 1.77 = 1.07. Ironically, the strongest relationship between caregiving and CES-D scores occurs among
women who spend the fewest hours in the role. Although all of
the coefficients are positive, only the lowest demand levels have
significant coefficients.
The effects for employment are similarly constructed. Again,
the reference category represents the highest level of demand
(more than 40 hours per week). The significant coefficient for
the dummy variable employed (-2.22) suggests that women
working the most demanding schedules have the lowest scores
on the CES-D subset scale.
Models 4 and 5 add in role satisfaction measures ranging in
value from 1 to 5, with 5 indicating the most satisfaction. As
HRS provides measures of satisfaction specific to marriage and
job, Model 4 adds in only those variables. In Model 5 we added
proxy measures of satisfaction for the roles of caregiver and
parent that were based on the respondent's satisfaction with
family life. Although slightly more than one third of the caregivers had children over age 18 in the home, only 6% of parents with children over 18 in the home were also providing care
to older parents. For these women (caregivers to both generations), the measure of satisfaction for these two roles is identical. Clearly, we would prefer separate measures for each role,
but they are not available.
Results from Model 5 show that the average CES-D score is
lower among employed women, married women, parents with
children over age 18 in the home, and parental caregivers (in a
one-tailed test only) as their levels of satisfaction with these
roles increase. Reviewing the coefficients for previously included variables, we find that, for the most part, patterns of effects are unchanged. Examining the dummy variables for role
occupancy, we find that (a) having older children in the home,
and (b) the interaction between married and paid worker, are
the only statistically significant occupancy effects.
As an illustration of how these multiple-role dimensions can
be translated into various levels of depressive symptoms, consider three fictional women. All three are 55-year-old White
women in good health, with 12 years of school, no children in
the home, and a household income of $36,000. The first
woman is married, employed, and working 30 hours per week,
a caregiver providing 150 hours of care each year, and has a
19-year-old living in the household; she is very satisfied in all
her roles. The second woman differs from the first only in that
she is not satisfied with any of her roles. The third woman is
neither married nor employed, provides the lowest level of
care, and is not satisfied. Respectively, the predicted CES-D
subset scores for these three women are 7.58, 14.94, and 14.96.
MULTIPLE ROLES AND WELL-BEING
S335
Table 3. Effects of Independent Variables on Depressive Symptomatology (A^ =5116)
Independent Variables
Logged household income
Education, 0-17 years
Race, 1 = White, 0 = non-White
Age, 51-61 years
Fair or poor health, 1 = fair/poor, 0 = good/excellent
Model 1
Model 2
Model 3
Model 4
Model 5
b
(SE)
b
(SE)
b
(SE)
b
(SE)
b
(SE)
-.61***
(.11)
-.61***
(.11)
_ J7#**
(.03)
— 84***
(.09)
-.61***
(.11)
_]§***
(.03)
-.18***
(.03)
.29
(.21)
-.11***
(.03)
5.88***
(.21)
Married, 1 = married, 0 = unmarried
_ j7##*
(.03)
.61**
(-22)
-.11***
(.03)
-.11***
(.03)
5 ^###
-.17***
(.03)
.68**
(•21)
-.11***
(.02)
.64**
(.21)
_ j j###
(.02)
(.21)
-2.28***
(.31)
-.05
(.33)
5.30***
(•21)
1.37*
(-56)
5.30***
(.21)
.41
(.58)
-.15
(.32)
3.48***
(.71)
-1.90***
(.31)
.79*
(.33)
-2.22***
(.36)
-.58
(.44)
-.69
(.44)
1.81***
(.36)
1.89***
(.36)
-.65
(-75)
j 97***
(.35)
2.23
(1.67)
j 9]***
(.35)
.45**
(.17)
Caregiver, 1 = provided care, 0 = did not
(.11)
.56**
(.20)
5.56***
(.21)
-2.27***
(.31)
Children >18 in household, 1 = yes, 0 = no
Employed, 1 = working for pay, 0 = not
_63***
Married X Employed
No. of children >18 in household, 1-6 children
-.69
(.76)
.37
(.21)
3.69**
(1.25)
Caregiving Dummy 1 (6-96 hours)a
.40
(.21)
.38
(.21)
3.74**
(1.24)
3.73**
(1.23)
1.89
(1.01)
.84
(1.06)
Caregiving Dummy 2 (100-168 hours)
2.07*
(1.03)
Caregiving Dummy 3 (170-365 hours)
.70
(1.08)
1.89
(1.01)
.77
(1.07)
Caregiving Dummy 4 (400-700 hours)
1.77
(1.10)
1.56
(1.09)
1.10
(1.10)
Caregiving Dummy 5 (720-1277 hours)
1.66
(1.08)
1.60
(1.06)
1.46
(1.06)
Paid Work Dummy 1 (l-19hours) b
-.61
(.43)
-.62
(.43)
-.52
(.43)
Paid Work Dummy 2 (20-39 hours)
.42
(-27)
.61*
(-27)
.61*
(.27)
Paid Work Dummy 3 (40 hours)
.48
(.25)
.58*
(.25)
.57*
(.25)
-.81***
(.10)
-.47***
(•07)
-.58***
(.11)
-.43***
(.07)
_ g]***
Marital satisfaction 1-5 (5 = highest satisfaction)
Job satisfaction 1-5 (5 = highest satisfaction)
Family satisfaction: child >18
(.14)
Family satisfaction: caregiver
Constant
R2
-.63
(.33)
24.45***
(1.74)
23.69***
(1.75)
.21
FforAR2
a
.22
15 24***
The reference category for all Caregiving Dummy variables is >1277 hours of care.
The reference category for all Paid Work Dummy variables is >40 hours of paid work.
*p < .05, **p < .01, ***p < .001.
23.58***
(1.75)
23.47***
(1.73)
23.44***
(1.72)
.22
.25
.25
2.56**
71.40***
19.43***
S336
REID AND HARDY
In other words, a greater difference exists between two women
in the same set of roles, at the same level of demand, who differ only in role quality, than between two women who share
only one role, neither of which is wife or worker. From these
results, it appears that role satisfaction measures capture dimensions of the multiple role experience that would otherwise
be confounded or excluded. Women's sense of well-being is
linked not only to what roles they occupy and the demands
these roles place on them, but also to quality of their experiences in these roles.
Alternative models.—To test the robustness of these findings
under different sets of assumptions, we estimated models with
the same set of explanatory variables for alternative specifications of the dependent variable. Results based on the natural log
of the CES-D subset scores were consistent with the findings
reported above. We also estimated binary logistic and multinomial logistic models for categorizations of the dependent variable. We adjusted Radloff's (1977) cutpoint of 16 for the 20point scale to this 14-point subscale by straightforward
extrapolation (0.7 * 16 = 11.2), coding a binary dependent variable "0" if equal to or below the threshold of 11 (and therefore
nonproblematic) and " 1 " if above the threshold of 11 (depicting
a potentially problematic level). Once again, the binary logit
models essentially reproduced the findings from earlier models.
However, results from the multinomial logistic regressions provide some additional insight into the nature of these relationships. From the CES-D, we constructed a 4-category dependent
variable coded as follows: 0 = 0-5, 1 = 6-8, 2 = 9-11,3=12
and higher. Multinomial logistic regression normalizes results
relative to the "0" category; therefore, estimation of the final
model produces three sets of coefficients and standard errors.
Each set of coefficients reports the net effects of the independent variable on the log-odds of being in categories 1, 2, or 3
(respectively) versus category 0. In other words, the log-odds of
reporting symptoms consistent with a score of 6 to 8 rather than
0 to 5 (the lowest level of symptoms) are reported in the first
column of Table 4; effects on the log-odds of a score from 9 to
11 versus 0 to 5 are reported in the second column; and effects
on the log-odds of a score 12 or higher versus 0 to 5 are reported in the third column.
We can summarize the detailed findings reported in Table 4
by categorizing independent variables into those that affect the
log-odds across the entire range of scores, those that affect the
log-odds toward the upper end of scores, and those that affect
the log-odds only at the extremes. Health, race, and job satisfaction have consistent effects across the range of CES-D
scores. Not only do these measures discriminate those with
problematic from nonproblematic levels of symptoms, but they
discriminate among those with virtually no symptoms and
those with a low level of depressive symptomatology. Age, having children over age 18 in the household, and family satisfaction (with children over 18) distinguish those with the lowest
symptoms from those with scores of 9 or higher (the upper two
categories). Finally, household income, education, working a
traditional full-time or part-time schedule (40 hours per week
or 20-39 hours per week), supplying the lowest levels of care,
and marital satisfaction are added to the previous list of variables, all of which discriminate those with the highest levels of
symptoms from those with the lowest levels of symptoms.
Table 4. Multinomial Logistic Regression Coefficients
and Standard Errors for CES-D Scale (A^ = 5116)a
Categories
6-8/0-5
b
(SE)
9-11/0-5
b
(SE)
12+/O-5
b
(SE)
-.02
-.07
(.06)
(.06)
Education, 0-17 years
.01
-.01
(.02)
(.02)
Race, 1 = White, 0 = non-White
.34**
.26*
(.11)
(.12)
Age, 51-61 years
-.01
-.05***
(.01)
(.01)
Fair or poor health,
.85*** 1.31***
1 = fair/poor, 0 = good-excellent
(.14)
(.13)
Married, 1 = married, 0 = unmarried
.20
.13
(.36)
(-34)
Children >18 in household,
1.24**
-.19
1 = yes, 0 = no
(.48)
(.46)
Employed, 1 = working for pay, 0 = not
.73**
.11
(.24)
(.26)
Caregiver, 1 = provided care, 0 = did not
-.49
-.59
(1.19)
(1.1)
Married X employed
.15
-.01
(.20)
(.22)
.19
No. children >18 in household,
.21
(.12)
1-6 children
(.11)
Caregiving Dummy lb,
-.01
-.98
(.96)
(6-96 hours)
(.62)
Caregiving Dummy 2,
-.23
-1.01
(.67)
(100-168 hours)
(.49)
-.20
Caregiving Dummy 3,
-.01
(170-365 hours)
(.50)
(.60)
Caregiving Dummy 4,
-.26
-.03
(.66)
(400-700 hours)
(.53)
Caregiving Dummy 5,
.78
-.18
(.55)
(•54)
(720-1277 hours)
c
Paid Work Dummy l ,
-.16
-.33
(1-19 hours)
(.20)
(.25)
Paid Work Dummy 2,
.17
.26
(20-39 hours)
(.15)
(•13)
Paid Work Dummy 3,
.09
.26*
(40 hours)
(.14)
(.12)
Marital satisfaction 1-5
-.09
-.10
(.07)
(5 = highest satisfaction)
(.06)
_2j***
Job satisfaction 1-5
-.10*
(.04)
(5 = highest satisfaction)
(.04)
.04
-.25**
Family satisfaction: child >18
(.09)
(.10)
Family satisfaction: caregiver
.14
.17
(.23)
(.24)
2.31*
Constant
.07
(1.00)
(•88)
Log likelihood
-6285.16
Log likelihood (restricted)
-6869.39
R2 count
.47
R2 adjusted count
.18
_27***
(.05)
_07***
(.02)
.28*
(.11)
-.05***
(.01)
2.12***
(.12)
.46
(.31)
1.48***
(.42)
.27
(.24)
.10
(.93)
.33
(.19)
.27*
(.11)
1.40*
(.65)
1.06*
(.53)
.70
(.57)
1.15*
(.57)
.68
(.59)
-.14
(.24)
.56***
(.15)
52***
(.13)
-.23***
(.06)
-.26***
(.04)
-.35***
(.08)
-.12
(.19)
6.09***
(.90)
Logged household income
a
From the CES-D we constructed a 4-category dependent variable coded as
follows: 0 = 1-5, 1 = 6-8, 2 = 9-11, 3 = 12 and higher. Thus, each set of coefficients reports the net effects of the independent variable on the log-odds of
being in categories 1, 2, or 3 (respectively) versus category 0.
b
The reference category for all Caregiving Dummy variables is >1277 hours
of care.
The reference category for all Paid Work Dummy variables is >40 hours of
paid work.
*p < .05; **p < .01; ***/? < .001.
MULTIPLE ROLES AND WELL-BEING
DISCUSSION
Although role enhancement and role strain perspectives provide useful points of departure for studying women's multiple
role experiences, neither perspective completely captures the
relationship between multiple roles and mental well-being.
First, we hypothesized that the effects of multiple roles may not
be additive, but that certain combinations of roles may ameliorate the negative aspects of multiple roles or enhance the positive aspects. Our findings show that only the combination of
marriage with employment interacts in this way; other combinations were not significant. Overall, these findings support the
importance of social integration for well-being, but suggest that
there may be a threshold effect: Being either married or employed is associated with lower levels of depressive symptoms,
but being both married and employed does not provide any additional increment to well-being. Second, we hypothesized that
the effects of role occupancy measures would be translated
through different indicators of demand and role satisfaction.
Our findings support this view. Thus, although women's reports
of depressive symptoms are linked to the kinds of roles they occupy, role occupancy alone does not fully reflect the complexity of the relationship of both demand levels and satisfaction to
well-being.
This analysis explores in more detail the nature of role experiences and how those experiences may be correlated with mental health. Roles that involve women outside of the home, such
as employment, are positively linked to mental health. Furthermore, roles that integrate women into family settings, such as
marital status, may also have a positive relationship with mental
health. However, in both cases the effects are indirect. Greater
role satisfaction is linked to lower depressive symptomatology
and, in the case of job satisfaction, its effect is consistent across
the range of symptom levels.
The relationship between employment and well-being is of
particular interest. Women who work 20-40 hours per week expressed the highest level of symptoms. Based on this analysis,
we can only speculate as to why this may be the case. Some
possibilities may be the nature of their work or their need for
income, and the possibility that they would prefer either to work
more or fewer hours. Indeed, the specific qualities of their jobs
may be important predictors of their well-being. Job characteristics such as autonomy,flexibilitywith scheduling, and the extent to which their employment promotes positive self-identity
formation may affect how women experience their paid work
roles (Lennon, 1994; Lennon & Rosenfield, 1992). Occupational sex segregation may also influence women's paid work
experiences because predominantly female jobs are often inflexible and unaccommodating to family caregiving responsibilities (Glass, 1990).
It is unfortunate that this survey does not include specific
measures of satisfaction with the informal caregiver role, nor
with the role of parent. Although results are consistent with expectations, the overlap among quality measures for family roles
complicates the estimation. The pattern of the relationship between well-being and caregiving demand remained U-shaped.
Again, we can only speculate that this pattern may involve
some necessary renegotiations of relationships between the
caregiver and the care recipient. Specifically, at the lowest levels of care, the care recipient may be adjusting to the transition
from independence to partial dependency, while the caregiver is
S337
struggling to define her responsibility and renegotiate a relationship that takes this increased dependency into account. At
higher levels of care, the issue may be one of disability. Because
the caregiving measure was specifically limited to ADL assistance, the interventions are likely linked to health-related declines. Thus, the relationship between caregiver and care recipient may be strained as the family copes with anticipatory
bereavement or faces questions of the need for professional
care or institutionalization (Abel, 1986).
The chronic role strains identified by Pearlin are useful for
describing many informal caregivers' experiences. Pearlin
views informal caregiving as an extreme form of role restructuring, as changes in the parent-child relationship create stress
for the caregiver. The stress associated with role captivity may
be a reality for women who reluctantly assume the informal
caregiving role—and this may be especially true for women
who spend the least amount of time in the caregiving role.
Pearlin (1989) and Greene and Monahan (1989) also highlight
the importance of coping strategies and social support as mediators of informal caregivers' stress. While these dimensions of
caregivers' experiences are not included in this analysis, future
investigations should explore these factors and their effects on
caregivers' mental health.
Limits of the study.—Several limitations of this analysis
merit discussion. First, well-being is indicated only by the respondent's level of depressive symptomatology. There are other
dimensions of well-being, such as self-esteem and the respondent's perception of the quality of her life in general, that would
also be useful indicators of well-being. Second, we are also
limited as to numbers of roles and an uneven coverage of levels
of demand and role quality. Future data collection efforts that
carefully assess a more extensive repertoire of women's roles
along with associated levels of demand and quality of experience would be welcome. Third, this research is cross-sectional,
and therefore cannot take into account preexisting emotional
problems or the conditions under which women undertake multiple roles. For example, it is not possible to know whether
women who have low levels of depressive symptoms take on
roles because they are able to, or if having multiple roles lowers
women's depressive symptomatology. Indeed, Moen and colleagues (1995) found that the effects of caregiving on women's
emotional health are moderated by their previous mental wellbeing. Further studies of these sorting processes are needed.
Finally, although our analysis treated role quality dimensions as
exogenous, future work will attempt to assess inherent endogeneities in these processes. Testing whether the complexity of
role sets, role demands, role satisfaction, and well-being are
jointly determined would be an appropriate next step in this
research.
Contributions to the literature.—On the positive side, this research contributes to the literature on women's multiple roles in
four important ways. First, our sample is more readily generalizable to the population at large than many previous studies; the
sample is representative of the population, includes a sufficiently large number of minority women to allow an assessment of racial/ethnic differences, and identifies a substantial
number of caregivers to elderly parents. Second, this research
adds to the role enhancement literature, which argues that en-
S338
REID AND HARDY
gaging in multiple roles is not uniformly detrimental for
women's well-being (Stoller & Pugliesi, 1989). Instead, as
Moen and coworkers (1995) conclude, the answer to the role
strain versus role enhancement debate is that "it depends" (p.
270). Third, our findings add support for and supplement the
"role spillover" literature, which emphasizes the interrelations
between roles (Parris Stephens & Franks, 1995; Parris Stephens
et al., 1994). Finally, we extend previous research by Baruch
and Barnett (1986) by including role quality in our conceptualization of women's multiple roles.
The present inquiry highlights the importance of respondents' perceptions of the quality of their roles. A realistic conceptualization of women's lives must include the possibility of
women assuming many combinations of these roles. Moreover,
we must examine not only how the roles may constrain or conflict but also how women themselves manage their obligations.
As we attempt to understand how women juggle competing demands, how they are trapped in unsatisfactory situations, or
how they successfully weave together multiple threads of responsibility, we must find better ways to assess not only the opportunities and the constraints that women face, but also the resources that promote their well-being.
ACKNOWLEDGMENTS
We thank Jill Quadagno, Irene Padavic, and Jim Orcutt for their wise counsel
and guidance and for their valuable editorial comments on numerous drafts of
this article. We also thank Kim Shuey for her help with the data set and for her
support throughout this research process. John Reynolds also deserves our gratitude for his advice. Finally, we thank Fredric D. Wolinsky and anonymous reviewers for their patience and guidance.
Address correspondence to Jennifer Reid, Florida State University, Claude
Pepper Center, Room 203, Tallahassee, FL 32306-1121. E-mail: jlr5335@
garnet.acns.fsu.edu
REFERENCES
Abel, E. (1986). Adult daughters and care for the elderly. Feminist Studies, 12,
479-497.
Apter, T. (1995). Secret paths: Women in the new midlife. New York: W. W.
Norton.
Aquilino, W. S., & Supple, K. R. (1991). Parent-child relations and parent's satisfaction with living arrangements when adult children live at home.
Journal of Marriage and the Family, 53, 13—27.
Baruch, G. K., & Barnett, R. (1986). Role quality, multiple role involvement,
and psychological well-being in midlife women. Journal of Personality and
Social Psychology, 51, 578-585.
Brody, E. M. (1985). Parent care as a normative family stress. The Gerontologist, 25, 19-29.
Brody, E. M. (1990). Women in the middle: Their parent care years. New York:
Springer.
Brody, E. M, Kleban, M. H., Johnsen, P. T., Hoffman, C, & Schoonover, C. B.
(1987). Work status and parent care: A comparison of four groups of
women. The Gerontologist, 27, 201-208.
Brody, E. M, Litvin, S. J., Hoffman, C , & Kleban, M. H. (1992). Differential
effects of daughters' marital status on their parent care experiences. The
Gerontologist, 32, 58-67.
Burkhauser, R. V., & Gertler, P. J. (1995). Introduction to special issue in the
Health and Retirement Survey/data quality and early results. Journal of
Human Resources, 30, S1-S6.
Cohen, J., & Cohen, P. (1983). Applied multiple regression (2nd ed.). Hillsdale,
NJ: Lawrence Erlbaum.
Crosby, F. (1991). Juggling. New York: Free Press.
Glass, J. (1990). The impact of occupational segregation on working conditions. Social Forces, 68, 779-796.
Glass, J., & Fujimoto, T. (1994). Housework, paid work, and depression among
husbands and wives. Journal of Health and Social Behavior, 35, 179-191.
Goode, W. J. (1960). A theory of role strain. American Sociological Review, 25,
483^96.
Greene, V. L., & Monahan, D. J. (1989). The effect of a support and education
program on stress and burden among family caregivers to frail elderly persons. The Gerontologist, 29, All-All.
Guberman, N., Maheu, P., & Maille, C. (1992). Women as family caregivers:
Why do they care? The Gerontologist, 32, 607-617.
Kramer, B. J., & Kipnis, S. (1995). Eldercare and work-role conflict: Towards
an understanding of gender differences in caregiver burden. The Gerontologist, 35, 340-348.
Lee, G. R., Seccombe, K., & Shehan, C. L. (1991). Marital status and personal
happiness: An analysis of trend data. Journal of Marriage and the Family,
53, 839-44.
Lennon, M. C. (1994). Women, work, and well-being: The importance of work
conditions. Journal of Health and Social Behavior, 35, 235—247.
Lennon, M. C , & Rosenfield, S. (1992). Women and mental health: The interaction of job and family conditions. Journal of Health and Social Behavior,
33, 316-327.
Lerner, M. J., Somers, D. G., Reid, D., Chiriboga, D., & Tierney, M. (1991).
Adult children as caregivers: Egocentric biases in judgements of sibling
contributions. The Gerontologist, 31, 746-755.
Levinson, D. J. (1996). The seasons of a woman's life. New York: Alfred J.
Knopf.
Mirowsky, J., & Ross, C. E. (1989). Social causes ofpsychological distress.
New York: Aldine de Gruyter.
Moen, P. (1992). Women's two roles: A contemporary dilemma. Westport, CT:
Auburn House.
Moen, P., Robison, J., & Dempster-McClain, D. (1995). Caregiving and
women's well-being: A life course approach. Journal of Health and Social
Behavior, 36, 259-273.
Mui, A. C. (1992). Caregiver strain among black and white daughter caregivers:
A role theory perspective. The Gerontologist, 32, 203-212.
Mui, A. C. (1995). Caring for frail elderly parents: A comparison of adult sons
and daughters. The Gerontologist, 35, 86-93.
Nolen-Hoeksema, S. (1990). Sex differences in depression. Stanford, CA:
Stanford University Press.
Oakley, A. (1974). The sociology of housework. Oxford, UK: Martin Robertson.
Parris Stephens, M. A., & Franks, M. M. (1995). Spillover between daughters'
roles as caregiver and wife: Interference or enhancement? Journal of
Gerontology: Psychological Sciences, 50B, P9-P17.
Parris Stevens, M. A., Franks, M. M., & Townsend, A. L. (1994). Stress and rewards in women's multiple roles. Psychology and Aging, 9, 45-52.
Pearlin, L. I. (1983). Role strains and personal stress. In H. B. Kaplan (Ed.),
Psychosocial stress: Trends in theory and research. New York: Academic
Press.
Pearlin, L. I. (1989). The sociological study of stress. Journal of Health and
Social Behavior, 30, 241-256.
Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1,
385-401.
Scharlach, A. E. (1994). Caregiving and employment: Competing or complementary roles? The Gerontologist, 34, 378-385.
Stoller, E. P., & Pugliesi, K. L. (1989). Other roles of caregivers: Competing responsibilities or supportive resources? Journal of Gerontology: Social
Sciences, 44, S231-S238.
Stone, R., Cafferata, G. L., & Sangl, J. (1987). Caregivers of the frail elderly.
The Gerontologist, 27, 616-626.
Wallace, R. B., & Herzog, R. (1995). Overview of health measures in the health
and retirement survey. Journal of Human Resources, 30, S84-S107.
Wheaton, B. (1990). Life transitions, role histories, and mental health.
American Sociological Review, 55, 209—223.
Young, R. F, & Kahana, E. (1989). Specifying caregiver outcomes: Gender and
relationship aspects of caregiver strain. The Gerontologist, 29, 660-666.
Received February 19, 1998
Accepted May 4, 1999