Journal oj Gerontology: SOCIAL SCIENCES
1996. Vol. 5IB, No. 2. S82-S9O
Copyright 1996 by The Gerontological Society of America
Opportunity Wages and Workforce Adjustments
Understanding the Cost of In-Home Elder Care
Shelley White-Means' and Deborah Chollet2
'Department of Economics, The University of Memphis, Memphis, Tennessee.
2
The Alpha Center, Washington, DC.
Time spent providing informal care represents a real cost to caregivers and potentially affects their decisions about
alternative activities, including paid work. This study offers estimates of opportunity wages incurred by informal
caregivers and explores the impact of these costs on their decisions to reduce or abandon workforce participation. We
consider caregivers at two points in time: 1982 and 1989. Our estimates of caregivers' imputed wages vary more widely
and range higher than those reported elsewhere, varying from $0.78 to $27.18 per hour, compared to values in the
literature ranging from $4.64 to $10.30. In both years, caregivers with higher imputed wages were significantly more
likely to work, with this effect stronger in 1989 as all noneconomic factors became less important. However,
accommodation (working fewer hours or forgoing job opportunities) decreased with higher imputed wages in 1982
only. In 1989, elders' frailty became a significant determinant of accommodation.
I
NFORMAL caregivers provide essential services that enable frail elderly persons to maintain their independence
and live at home in the community. As informal caregiving
efforts may delay institutionalization of frail older persons,
these efforts are also perceived to save public resources.
Given the significant role of informal caregivers in the
long-term care system for older persons, it is important to
recognize that their efforts are not costless; they require
sacrifices of time and energy (Headen, 1992). Informal caregivers contribute significant hours to support the in-home care
of elderly patients (Stephens and Christianson, 1986). Many
contribute hours comparable to part-time jobs and similar to
full-time jobs, if the caregivers are not employed in the labor
market. Past research has begun to document both caregivers'
hours of assistance (Brody, 1981; Finley, 1989; Matthews
and Rosner, 1988; Stephens and Christianson, 1986; WhiteMeans and Thornton, 1990; White-Means, 1992) and the
stresses that they encounter (Abel, 1990; Cox and Monk,
1990; Miller and McFall, 1991).
At least two studies (Ernst and Hay, 1994; and Hu,
Huang, and Cartwright, 1986) go beyond the documentation
of time efforts to assign a monetary value to the time
provided by caregivers for older persons disabled by dementia. Based on the researchers' calculation that care for senile/
demented elderly persons requires an average effort of 6.28
hours per day and valuing caregivers' time at the median
wage of nurse's aide, Hu, Huang, and Cartwright (1986)
estimated the value of caregiving at $29.15 per day. On the
same basis, Ernst and Hay (1994) assign a much higher
hourly rate of $9.40 to $10.30 per hour to informal caregivers, producing an estimate of gross value of unpaid
caregiving at $20,900 per year per older person afflicted with
dementia.
Estimates like those described above are one measure of
the opportunity cost of informal caregiving — the value of
resources that would be devoted to caregiving were that
activity paid at the current market wage of formal caregivers.
S82
However, such estimates may not accurately reflect informal
caregivers' actual opportunity cost. The argument for anticipating that an informal caregiver's time may not relate in any
particular way to the value of a formal caregiver's time
pertains to researchers' inability to observe some important
features of that market. In particular, informal caregivers
(whom, we assume, manage the decision to substitute formal caregiving services for their own time) may find the
qualitative aspects of formal caregiving services very difficult to evaluate and subjectively mark up the price of
formal caregivers to reflect their anticipated search and
failure costs. In addition, disabled older persons may place a
high value on informal caregiving relative to formal caregiving; as an agent of the disabled older person, informal
caregivers in this situation may exhibit a reduced demand for
substitutive formal caregiving at any price. In such a market,
it is not reasonable to assume that informal caregivers'
opportunity wages are related in any predictable way to the
actual market value of formal caregiver services. Instead,
the direction of the relationship is an empirical question.
Caregivers' opportunity costs include forgone wages and
current non-wage benefits, including health insurance, pension accruals, and Social Security accruals. They also include forgone future income, if absence from the paid
workforce reduces future earnings and benefits. Most of
these costs remain undocumented in monetary terms. An
exception is a recent study by Kingson and O'GradyLeShane (1993), who estimated that early retirement due to
leaving the labor force to engage in caregiving results in a
loss of $127 in monthly Social Security benefits to the
caregiver. However, this study was unable to differentiate
between caregiving responsibilities associated with an older
person versus a child, and failed to address other differences
in retirement income due to differences in the probability of
pension receipt.
White-Means (1994) reports that over 12 percent of caregivers for frail elders perceive that their physical health is
OPPORTUNITY WAGES AND WORK ADJUSTMENTS
worse due to elder care responsibilities. Moreover, studies
by George and Gwyther (1986) and Haley et al. (1987)
established that informal caregivers to older persons with
dementia also incur higher health care costs, including significantly higher use of physician services, prescription
drugs, and hospital days. Thus, estimates of informal caregiver opportunity cost that rely on the value of formal
caregivers' time probably also do not capture these informal
caregiver costs.
Caregivers' opportunity costs may affect decisions to
reduce or abandon participation in the paid workforce, with
higher opportunity costs associated with less workforce
adjustments. We have limited information about this supposition because national caregiver data bases do not include
explicit measures of short- and long-term caregiver costs, as
described above. Past research suggests that a significant
minority of informal caregivers do make labor force adjustments to accommodate the personal care needs of disabled
older persons. Using data from the 1982-84 National Long
Term Care Survey, Stone, Cafferata, and Sangl (1987)
reported that 8.9 percent of the nation's caregivers quit their
jobs because of caregiving responsibilities. This group of
caregivers represented about 22 percent of employed caregivers. Among caregivers who continued their labor force
activities, 21 percent worked fewer hours, and 19 percent
took time off without pay. Daughters were more likely than
sons to undertake either of these adjustments in labor force
activity. Using the same survey data, White-Means and
Thornton (1990) reported that African American caregivers
were less likely than White ethnic caregivers to adjust their
labor force activity.
Available research also suggests that informal caregivers
make greater adjustments in their labor force activity when
the health of frail older persons declines to the point that they
would qualify for hospice care. Muurinen (1986) reported
that 33 percent of employed caregivers for hospice patients
had quit their jobs, and 60 percent had lost income due to
caregiving responsibilities.
This study offers estimates of the current opportunity
wages incurred by informal caregivers engaged in providing
personal care services to disabled older persons. To understand the nature of informal caregivers' workforce behavior,
we estimate a behavioral model of workforce participation
among informal caregivers. The model considers a number
of variables related to caregivers' production of informal
long-term care services, preferences, and economic situation
— including the imputed value of the caregiver's time. We
estimate the relationship of these variables to the caregiver's
workforce adjustment (accommodation) decision and the
decision to work in paid employment, respectively and
separately for two years: 1982 and 1989.
The rationale for estimating the relationship for these
years is twofold. First, we anticipate that caregivers' decisions about accommodation and employment may differ as a
result of changing labor market conditions. The year 1982
was one of high unemployment; by 1989, unemployment
had declined and labor force opportunities for caregivers
potentially had improved. Second, between 1982 and 1989,
Medicare had phased in prospective payment for hospital
care. This change in Medicare payment encouraged reduced
S83
rates of hospital admission and shorter lengths of stay among
Medicare beneficiaries. As a result, Medicare beneficiaries
released into formal long-term care were reported to be
frailer at the point of hospital discharge in 1989 than in 1982.
We presume that this change in Medicare payment also may
have affected the health status at discharge of Medicare
patients released to home care — either formal or informal.
The first section reviews some essential findings from the
1989 National Long Term Care Survey (NLTCS), describing the workforce adjustments of informal caregivers in the
survey. Next we describe the authors' estimation of informal
caregivers' opportunity cost, calculated as imputed wages.
The third section explores the impact of the informal caregivers' opportunity wage on their decisions to participate in
the paid workforce. This analysis considers caregivers' decisions in both 1982 and 1989 in order to identify changes in
decision-making that may have occurred over the decade.
Finally, a summary and preliminary conclusions are provided. Analysis of important long-term aspects of the opportunity cost of informal caregiving — forgone nonwage
benefits, future wages, and health status — is deferred to the
next phase of research to be conducted under this project.
Reported Workforce Adjustments by Caregivers
Although about one-half of informal caregivers are of
conventional working age (younger than age 65), only 27
percent of informal caregivers who were included in the
1989 NLTCS report paid employment. Of these, only 19
percent report full-time paid employment (Table 1).
A significant minority of informal caregivers adjust their
paid work effort to accommodate their caregiving responsibilities. Of caregivers who were employed at the time of the
NLTCS, 22 percent reported that they worked fewer hours
than they would have preferred because of caregiving.
Among all caregivers, 13 percent reported having quit or
turned down a job at some time because of caregiving
responsibilities.
The likelihood that caregivers had ever adjusted their
workforce participation (that is, reduced hours worked, or
quit or turned down a job because of their caregiving responsibilities) varies systematically with selected measures of the
burden of caregiving. Simple descriptive statistics of these
relationships are presented in Table 2; these relationships are
explored further in a multivariate analysis in the section,
Caregiver Workforce Adjustments.
In general, informal caregivers are more likely to have
made some workforce accommodation to their caregiving
responsibilities as the older person becomes more frail. In
1989, caregivers who either were not working at the time of
the survey or had made some workforce accommodation to
caregiving were providing care for much frailer older persons
than those who were working or had made no accommodation. Similarly, caregivers who had adjusted their paid work
effort were on average caring for older persons with more
ADL limitations than caregivers who had not (although the
cognitive impairment of these older persons was less severe).
Finally, caregivers who had adjusted their workforce effort
also were more likely to be caring for an older person who
was receiving Medicaid. This relationship suggests that Medicaid recipiency may have a real income effect on care-
WHITE-MEANS AND CHOLLET
S84
Table 1. Workforce Participation of Informal Caregivers to Disabled Older Persons, 1989 (Percentages)
All Caregivers
Total {N = 741)
Caregiver has ever worked for pay
Caregiver currently works full-time
Caregiver has ever worked fewer hours than would like because of helping disabled
older person
Caregiver has ever quit or turned down job or not looked for job to care for disabled
older person
Caregiver Does
Not Work
Caregiver Works
100.0%
88.1
19.0
27.1%
100.0
70.0
72.9%
79.6
NA
8.4
22.0
3.4
13.2
15.0
12.5
Source: 1989 National Long Term Care Survey.
Table 2. Means for Selected Characteristics of Disabled Older Persons
by Informal Caregivers' Reported Workforce Adjustments, 1989 (Percentages)
Extent of frailty*
Extent of cognitive impairment'
Number of ADL limitations
Number of IADL limitations
Caregiver reporting that:
Older person requires more care now than when caregiving was initiated
Caregiving by respondent commenced in 1982 or before
Older person has someone else who could help them if caregiver (respondent) could not
Older person receives Medicaid
Older person has private insurance
Caregiver Works
Caregiver Has
Ever Adjusted
Workforce Effort
Yes
Yes
19.4%
21.3
2.6
3.6"
30.6
30.6
68.6
23.4
60.1
No
b
24.4%
23.6
2.8
3.6
36.6%
15.8'
3.4"
4.4
32.7
32.7
54.6
18.7
63.8
62.4
33.3
54.0
24. lb
58.2
No
c
17.8%
25.5
2.5
3.3
47.4
31.8
61.2
17.8
64.5
Source: 1989 National Long Term Care Survey.
"A continuous measure of the probability of a particular disability profile. Each measure varies from 0 to 100, as developed for the NLTCS based on the
Grade of Membership (GOM) model by the Center for Demographic Studies at Duke University.
•"Significantly different from comparison group at 0.95 confidence level.
'Significantly different from comparison group at 0.99 confidence level.
givers' decisions to work in paid employment, but it also is
consistent with Medicaid's meager financing for in-home
services. For a discussion of Medicaid financing of in-home
care, see Rowland and Lyons (1991). Current workforce
participation was less sensitive than other types of accommodations to most measures of the burden of caregiving.
Imputation of the Opportunity Wage of Informal Caregivers
Economic theory suggests that market wage rates are
reliable measures of the current opportunity value of an hour
of a caregiver's time. To understand this conclusion, we
begin with the supposition that caregivers allocate their time
among labor market employment, caregiving, other household responsibilities, and leisure in a way that maximizes
utility, constrained by financial resources and a 24-hour day.
This occurs when the caregiver's productivity from the last
hour spent in each activity is equal. The market wage rate
reflects the productive value of the last hour spent in the
labor market. Among employed caregivers who have allocated their time optimally, this value equals the value of the
last hour spent in other uses of time. Thus, when time use is
optimized, the market wage equals the productive value of
all other time uses. When the wage rate that an individual
could earn in the labor market is lower than the value of their
time in non-labor market activities, the individual will
choose to allocate time to non-market activities only. For
such individuals, the market wage rate is a lower bound
estimate of the value of an hour of their time (Zick and
Bryant, 1983).
One measure of the opportunity value of an hour of a
caregiver's time is the caregiver's replacement cost; that is,
the market price of an hour of a home health worker's time.
However, since the wage of home health workers reflects the
productive value of the last hour of their time, it may not be
reasonable to assume that the informal caregiver's time
value is related in any predictable way to that of the formal
caregiver. Indeed, research that compares market replacement cost estimates of domestic and home service workers
with measures of opportunity wages of housewives indicates
that replacement cost measures significantly underestimate
the value of housewives' time (Zick and Bryant, 1983).
The NLTCS does not include market wage data for caregivers and, in any case, would include wages only for
caregivers who worked in paid employment. We imputed
the market value of informal caregivers' time using information about the characteristics of workers who are employed,
OPPORTUNITY WAGES AND WORK ADJUSTMENTS
as reported in a separate survey, the March 1992 Current
Population Survey (CPS). A household survey conducted
monthly by the Bureau of the Census, the CPS supports
many of the federal government's population data series,
including national and state population estimates, and estimates of unemployment. Each year, the March CPS includes
extensive questioning about income, family composition,
and workers' labor force activity and wages.
To impute wages, we stratified the 1992 CPS into four
groups reflecting differences in the workers' race (Black and
White) and gender (male and female). Our sample included
only workers aged 18 to 64 who reported themselves to be
White or Black, worked full- or part-time, were employed in
private sector nonagricultural jobs, and earned $1 to $50 per
hour. For each group, we estimated parameters for a standard human-capital model, regressing the logarithm of hourly
wages (annual personal earnings divided by hours of work
per year) on experience (measured as age minus education
minus 6), the square of experience, years of education, and a
series of dummy variables for full-time work, metropolitan
location, geographic location, occupation, and industry.
S85
We correct this wage equation for sample selection bias
(Greene, 1993). This is because wage data were reported
only for those who were employed in the labor market.
Thus, the determinants of these wages are systematically
correlated with the probability of employment. Estimation
of the wage equation for workers, without accounting for
this correlation, may lead to biased wage equation estimates.
Our wage equation estimation is performed using Limdep
version 5.1 (Greene, 1993). We first estimate the probability
of employment, using a binary probit criterion function.
Second, the lambda correction function is computed. Third,
we estimate selection corrected wage equations for the
sample of employed persons only. These wage estimates are
generalizable to a sample of persons who are not employed
in the labor market.
The parameter estimates for these selection-corrected
equations are reported in Table 3. The results are similar to
those found elsewhere in the literature, with slight variations
by race and gender. Earnings increase with experience at a
decreasing rate. Workers who have more education, are
employed full-time, and live in metropolitan areas earn
Table 3. Parameter Estimates for Wage Equations for All Workers (Selection Corrected)'
Experience
Experience squared
Education
Full-time
Metropolitan
South
Manage
Sales
Laborer
Agriculture
Transportation/utilities
Trade
Personal/business services
Professional services
Intercept
Lambda
Adjusted R2
Sample size
White
Males
Black
Males
White
Females
Black
Females
.041
(39.61)**
-.006
(-25.94)**
.065
(38.95)**
.222
(18.55)**
.115
(13.21)**
-.084
(-11.31)**
.243
(22.92)**
.082
(7.96)**
-.084
(-8.78)**
.006
(.45)
.017
(1.41)
-.192
(-20.65)**
-.222
(-18.11)**
-.193
(-15.17)**
.997
(34.53)**
-.261
(-17.11)**
.383
22,746
.031
(7.96)**
-.0004
(-4.28)**
.059
(7.9)**
.188
(5.2)**
.1091
(2.3)*
-.166
(-6.11)**
.231
(5.04)**
.145
(3.78)**
-.008
(-.24)
-.094
(-1.80) +
.017
(.39)
-.209
(-5.83)**
-.193
(-4.56)**
-.123
(-2.48)**
1.05
(8.03)**
-.145
(-2.94)**
.276
1,741
.287
(27.44)**
-.0005
(-20.58)**
.064
(30.09)**
.166
(19.15)**
.188
(21.01)**
-.087
(-11.35)**
.411
(33.34)**
.202
(19.57)**
.006
(.38)
-.074
(-2.47)**
.063
(3.30)**
-.193
(-15.98)**
-.220
(-14.61)**
-.111
(_8.74)**
.703
(19.44)**
.004
(.20)
.312
19,339
.026
(7.73)**
-.0003
(-4.49)**
.062
(8.31)**
.211
(7.65)**
.170
(5.11)**
-.106
(-4.72)**
.439
(11.92)**
.208"
(7.15)**
.009
(.24)
.022
(.16)
.072
(1.39)
-.233
(-5.96)**
-.277
(-6.53)**
-.197
(_5.04)**
.661
(5.17)**
.008
(.20)
.337
1,953
Source: OLS regressions based on March 1992 Current Population Survey,
'/-statistics are in parentheses.
+ p< A0;*p< .05;**p< .01.
S86
WHITE-MEANS AND CHOLLET
higher wages; workers in the South have lower wages than
workers elsewhere.
We used these regression estimates to impute wages for
each of the sample caregivers who participated in the 1989
NLTCS. For caregivers who ever had worked in paid employment, the NLTCS asks information about each of the
variables that we used to estimate the wage equations. To
produce an imputed wage for each of the NLTCS caregivers,
we applied the CPS-estimated parameters to their responses
for each variable.
Among all caregivers responding to the NLTCS, 741
responded to all the workforce questions asked in the survey.
Based on these responses, the average imputed wage for
these caregivers is $6.23. Caregivers' imputed wages vary
from $0.78 per hour to $27.18. The mean values of the
imputed wages for this caregiver sample, by selected characteristics of the caregiver, are reported in Table 4 and are
consistent with general wage patterns. For example, men,
full-time workers, those who live in metropolitan areas, and
those who live outside the South have higher imputed wages.
Somewhat counterintuitive is the finding that imputed wages
are higher for Blacks than for Whites. In part, this is
explained by the relatively small representation of men
among the caregiver sample and the greater likelihood that
Black women work full-time, live in metropolitan areas, and
work in professional and related services industries. Also
note that the average imputed wage of caregivers who work
is higher than that for caregivers who do not work, consistent
with our general expectation that informal caregiving may
decline as the caregivers' opportunity costs are higher.
Caregiver Workforce Adjustments
and Employment Decisions
Underlying behavioral model. — The caregiver utility
maximization model presented in White-Means and Chang
(1991) is the underlying behavioral model for our analysis.
The model describes decision making among caregivers who
provide in-home services to older persons who have recon-
Table 4. Means of Imputed Wages by Sample
Caregiver Characteristics
Caregiver Characteristics
All caregivers
Caregiver works
Caregiver does not work
Female
Male
Black
White
Full-time worker
Part-time worker or not working
Metropolitan residence
Non-metropolitan residence
Southern residence
Non-Southern residence
Mean Imputed
Wage*
6.23
8.41
5.26
6.01
11.84
7.01
6.23
9.26
5.42
7.70
5.96
5.85
6.69
•The differences in the reported means within groups are significant at .01
or better.
ciled to stay out of an institution and to receive care in the
home. In this model, caregivers gain utility from marketpurchased and household-produced home health care, as well
as from others' goods. Caregivers can allocate time among
four types of activities: production of home health care,
production of other household goods, employment, and leisure. The caregiver's productivity in home health services
depends on the home health production technology, where
EHC = E(N,Z,H)
(1)
and
EHC = production technology of informal home health
services,
N = number of home health workers,
Z = degree of household worker substitution among
various home health tasks,
H = health needs of the elderly patient.
Thus, the caregiver's derived demand for use of time in
labor market activities is specified as follows:
M = M(W, V, P x , P c , SM, EHC, EHP)
(2)
where M = time allocated to labor market employment,
W = the caregiver's wage (opportunity cost of
time),
V = other family income,
Px = price of market goods,
Pc = price of market-purchased home health
services,
SM = preferences shifters for labor market
employment,
EHC = production technology of informal home
health services,
EHP = production technology of other household
goods.
Analysis plan. — We examine the caregiver's labor force
decision (equation 2 of the behavioral model) within the
framework of a nested decision process. Similar to Stone and
Short (1990), we model the three choices of the caregiver: to
work in the labor market or not, to work with accommodation, and to work without accommodation. Within a nested
logit framework, whether the caregiver works is influenced
by the probability of accommodation. Thus, we first estimate the probability of accommodation and, given these
results, predict the probability of employment in the labor
market. For this analysis, we restrict the sample to include
only those caregivers who, at some time, have worked in the
labor market.
Data. — The principal data source for this analysis is the
National Long Term Care Survey (NLTCS), 1982/84 and
1989. The NLTCS is a nationally representative and longitudinal data base, sponsored by the Office of the Assistant
Secretary for Planning and Evaluation (U.S. Department of
Health and Human Services), the Health Care Financing
Administration, and the Bureau of Census (U.S. Department
of Commerce). In 1982, the NLTCS surveyed approximately 6,400 persons who were aged 65 or older, were
Medicare recipients, and affirmed a chronic limitation in
activities of daily living (ADLs) or instrumental activities of
OPPORTUNITY WAGES AND WORK ADJUSTMENTS
daily living (IADLs). Approximately 1,900 caregivers for
the primary survey population were interviewed. These
caregivers were questioned about the types of informal
support services that they provided, their monetary and time
contributions, how their efforts affected their labor market
employment and family life, and their general attitudes
about institutional nursing home services.
In 1989, the NLTCS questioned 1,007 caregivers for
older persons who (a) had been interviewed in 1982; or (b)
were caregivers for older persons that NLTCS had included
in the 1984 follow-up NLTCS; or (c) were identified for the
1989 NLTCS as meeting the survey's age and disability
criteria. The survey instrument used to interview caregivers
in 1989 was similar to the 1982 NLTCS instrument and was
supplemented with more detailed questions about the labor
market activities of the caregivers. This reinterview pattern
allows observation of the characteristics of known longtime
caregivers, compared to other caregivers in the survey.
While the 1989 NLTCS asks some labor market questions
of caregivers, the primary aim of the longitudinal NLTCS
was to examine changes in the health status of older persons
and the ways in which older persons met their needs in the
community. As a result, the NLTCS labor market questions
of the caregivers were not sufficiently comprehensive, used
alone, to support estimates of caregiver's cost. For example,
the NLTCS data do not include information on wages or
pension participation among caregivers who work in paid
employment. Consequently, we use the CPS estimates to
impute key information to the NLTCS.
Variable definitions. — We use two dichotomous dependent variables in our regression analysis of labor market
decisions: (1) the caregiver works (or not); and (2) the
caregiver accommodates (or not). Among employed caregivers, accommodation is indicated by one or more positive
responses to a number of questions, indicating that the
caregiver works part-time rather than full-time because of
caregiving responsibilities or that the caregiver had ever quit
a job, taken time off without pay, not looked for market
employment, or turned down a job due to caregiving responsibilities. A summary of other variable measures is included
in Table 5.
The estimation incorporates variable measures (in Table
5) that reflect the seven theoretical variables of Equation 2.
Some of these measures require further discussion.
The caregiver's opportunity cost (W) is measured as the
caregiver's imputed wage. Imputed wages for 1989 and
1982 were adjusted by the CPI.
The price of market-purchased home health (Pc) is represented by the disabled older person's health insurance status
— specifically, coverage by Medicaid or private health
insurance. Since the sample for the NLTCS was drawn from
Medicare records, all disabled older persons in the survey
were enrolled in Medicare.
The caregivers' preference shifters (SM) are assumed to
correlate with demographic variables such as marital status,
age of caregiver, whether the caregiver and the older disabled
person live together, race, gender, and family relationship.
Factors that are assumed to influence overall household
productivity (EHP) were the caregiver's education, health
S87
Table 5. Production Variable Definitions
Name
Production Variables:
Backup
Education
Fair/poor health
Other ADL caregivers
Other IADL caregivers
Children
More care
Cognitive health of
older person
Frail
Description
= 1 If an alternate caregiver is available
Years of formal education
= 1 If caregiver's health is fair or poor
= 1 If other persons provide ADL
assistance
= 1 If other persons provide IADL
assistance
Number of children who live at home with
the caregiver
= 1 If caregiver provides more care than
former years
GOM measure of the degree of cognitive
impairment
GOM measure of the elder's degree of
frailty
Income and Prices of Home Care:
Family Income
In hundreds of dollars
Wage rate
Imputed wages as a measure of the
opportunity cost of time
Medicaid
= 1 If Older Person has Medicaid coverage
Private
= 1 If Older Person has private health
insurance
Carein82
= 1 If the caregiver has provided assistance
for 7 or more years
Preference Shifters:
Married
Age of caregiver
Livewith
Black
Gender
Immediate
= 1 If caregiver is married
In years
= 1 If caregiver lives with older person
= 1 If caregiver is Black
= 1 If caregiver is female
= 1 If caregiver is an immediate family
member of the older person
status, and the number of children living in the caregiver's
household. The technology of home care (EHC) is measured
by a series of dummy variables indicating (1) whether there
are other ADL or IADL caregivers, (2) whether a substitute
caregiver is available, (3) the older person's degree of
cognitive and physical disability, (4) whether the older
person requires more care than when caregiving began, and
(5) whether the caregiver had provided assistance for seven
or more years.
Measurement of disability status among older persons
(reflecting one component of the theoretical construct, EHC,
in Equation 2) is complicated by the existence of comorbid
medical conditions and multiple functional limitations. Yet,
it is this complex mix of conditions and limitations that
determines the caregiver's perception of the demand of
caregiving and whether or not it is possible to balance labor
force activities and caregiving responsibilities. In this study,
we use new measures of disability status that are based on a
Grade of Membership (GOM) Model and developed for
NLTCS by the Center for Demographic Studies at Duke
University. A detailed description of the methods used to
derive these disability measures is provided in a recent
article by Manton, Corder, and Stallard (1993). The GOM
model provides six profiles of the disability status of older
S88
WHITE-MEANS AND CHOLLET
persons surveyed in the NLTCS: healthy, moderate cognitive impairment, mild instrumental and physical impairment, serious physical impairment, moderate ADL and
serious physical impairment, and frail. The GOM measures
are continuous and reflect the probability that an older person
has a particular disability state. We consider two disability
states: moderate cognitive impairment and frailty.
Persons with a higher probability of mild cognitive impairment are more likely to be limited in their ability to
perform IADLs (including heavy work, laundry, cooking,
grocery shopping, traveling, managing money, and taking
medicine) and difficulties with functional tasks (such as
putting on socks, reaching overhead, combing/washing their
hair, grasping objects and seeing well enough to read a
newspaper).
Persons with a higher probability of frailty are more likely
to be limited in ADLs (especially getting in or out of bed,
getting around the house, dressing, bathing, and using the
toilet), to be limited in instrumental activities of daily living,
and to have greater difficulty with functional tasks (such as
climbing stairs, holding 10 pounds or bending).
Regression results. — In 1989, caregivers were slightly
more likely to work with accommodation (30%, vs 29% in
1982), and less likely to be employed at all (29%, vs 33% in
1982). The difference in average behavior between 1982 and
1989 reflects differences in the nature of caregivers' accommodation and employment decisions; see Table 6.
In 1982, production factors, selected income and prices of
home care factors, and selected caregiver preferences influenced the accommodation decision. Accommodation was
more likely if the caregiver's health was either fair or poor, or
if the caregiver was older. If the imputed wage was high, or if
other caregivers were available to give support in IADLs,
employed caregivers were less likely to accommodate.
Among the production factors that were significant in
caregivers' accommodation decisions in 1982, only the
caregiver's health status remained significant in 1989. In
1989, the older person's frailty became significant, raising
the caregiver's likelihood of accommodation. The caregiver's immediate family relationship to the older person
also became important, raising the likelihood of accommodation. And while the caregiver's opportunity wage was no
longer important in accommodation decisions, the older
person's Medicaid coverage (consistent with the effect of a
real income transfer) increased the caregiver's probability of
accommodation.
In both years, we found that caregivers' decision to work
was unrelated to their probability of accommodation, a
finding that contrasts with that reported by Stone and Short
(1990). However, like the decision to accommodate, the
decision to work was independently related to a number of
production factors, selected income and price factors, and
selected caregiver preferences. Also, like the decision to
accommodate, the absolute and relative importance of these
factors changed markedly between 1982 and 1989.
In 1982, caregivers were more likely to work if they had
an available backup, a high opportunity wage, or were
female. They were less likely to work if they were older,
Black or married, if their health was fair or poor, or if the
older person in their care was more frail or received Medicaid. Additionally, caregivers with more years of formal
education were less likely to work, potentially reflecting
greater productivity in caregiving and also better retirement
planning during the caregiver's working career.
In 1989, the caregiver's education, opportunity wage, age,
and gender remained factors in the caregiver's work decisions. Of these, the caregiver's imputed wage was most significant and had the greatest impact on the caregiver's work
decisions. Family income, Medicaid, marital status, and race
were no longer significant. Furthermore, while having an
available backup (or temporary substitute caregiver) no
longer affected the caregiver's work decision, the availability
of an ADL caregiver (a time-intensive, ongoing secondary
caregiver) became significant in the decision to work.
Discussion. —Structural changes between 1982 and 1989
in Medicare reimbursements to hospitals have led to reductions in hospital admissions among older persons, reduced
lengths of hospital stay, and increased use of skilled nursing
and home health care. It is also reasonable to expect that
these changes have increased demands on family members
who serve as informal caregivers. The differences that we
observe between 1982 and 1989 in caregivers' decisions to
accommodate or to work at all are consistent with these
changes.
Between 1982 and 1989, the frailty of the disabled older
person became a more important factor in caregivers' decisions to accommodate. Caregivers in situations with greater
real income in the form of the older person's Medicaid
eligibility also became more likely to accommodate, but the
importance of the imputed value of their time in paid employment declined.
The nature of caregivers' decisions to work at all also
changed between 1982 and 1989. Specifically, caregivers
became more sensitive to the value of their time in paid
employment, so that caregivers with a higher opportunity
wage were still more likely to work in paid employment than
they had been in 1982. This finding is consistent with the
lower unemployment rates and enhanced labor market opportunities that characterized 1989. That is, greater earnings
opportunities in 1989 may have made leaving paid employment a more difficult choice, sharpening the caregiver's
focus on the opportunity wage. This explanation (coupled
with the declining real wages among low- and middleincome workers that characterized the market at large) also is
consistent with the insignificance of income variables —
both actual income and Medicaid eligibility — as determinants of caregivers' decisions to work in 1989.
Between 1982 and 1989, caregivers' decisions to work
became less sensitive to caregivers' ability to substitute
against the role of caregiver. That is, the importance of
having an ongoing secondary ADL caregiver increased, as
the importance of more remote temporary backups declined.
Summary
This study offers new estimates of the value of the informal caregiver's time and explores the importance of the
imputed value* of caregiver time to caregivers' decisions to
alter their workforce behavior, either by accommodating
OPPORTUNITY WAGES AND WORK ADJUSTMENTS
S89
Table 6. Logit Coefficients: The Probability of Work Accommodation
and the Probability of Employment Among Caregivers"
Probability of Work
Accommodation Among
Employed Caregivers
1982
Inclusive value
Production variables:
Backup
Education
Fair/poor health
Other ADL caregivers
Other 1ADL caregivers
Children
More care
Cognitive health of elder
Frail
Income and prices of home care:
Family income (in hundreds)
Wage rate
Medicaid
Private
Care in '82
Preference shifters:
Married
Age of caregiver
Live with
Black
Gender
Immediate
Probability of Employment
Among Caregivers
1989
1982
1989
—
-0.825
(-1.08)
-0.333
(-0.37)
-0.323
(-1.19)
0.049
(1.09)
1.020
(3.15)**
0.181
(0.54)
-0.921
(-2.60)**
0.027
(0.11)
0.434
(1.53)
-0.015
(-1.58)
0.006
(1.26)
-0.406
(-1.09)
-0.025
(-0.26)
0.862
(1.98)*
-0.148
(-0.27)
0.535
(1.25)
-0.087
(-0.29)
0.414
(1.06)
0.005
(0.72)
0.018
(2.34)*
0.478
(2.01)*
-0.148
(-3.60)**
-0.922
(-1.87) +
0.006
(0.03)
0.784
(159)
-0.161
(-0.80)
0.129
(0.47)
0.011
(1.11)
-0.011
(-2.63)**
0.229
(0.67)
-0.180
(-2.92)**
-0.806
(-1.41)
0.651
(1.73) +
-0.588
(-1.33)
0.239
(0.89)
0.062
(0.17)
0.002
(0.34)
-0.015
(-0.01)
0.002
(1.67) +
-0.124
(-2.91)**
0.176
(0.54)
0.362
(1.25)
0.099
(0.35)
0.001
(0.67)
-0.055
(-0-71)
0.840
(1.75) +
0.073
(1.18)
-0.103
(-0.25)
0.003
(2.31)*
0.355
(4.65)**
-0.571
(-2.68)**
-0.393
(-1.52)
0.017
(0.09)
0.002
(0.14)
0.541
(6.24)**
-0.118
(-0.21)
0.275
(0.99)
0.127
(0.47)
-0.236
(-0.79)
0.024
(1.77) +
-0.218
(-0.72)
0.614
(1.34)
-0.403
(-1.20)
0.306
(0.93)
0.165
(0.37)
-0.033
(-1.87) +
0.046
(0.10)
-5.63
(-1.06)
-0.379
(-0.50)
1.09
(2.10)*
-0.376
(-1.68) +
-0.042
(-2.86)**
-0.061
(-0.27)
-1.23
(-3.01)**
0.839
(2.69)**
0.073
(0.27)
-0.202
(-0.67)
-0.053
(-2.34)*
-0.036
(-0.12)
0.387
(0.79)
1.610
(2.52)*
-0.190
(-0.25)
N=
359
193
818
591
-2 Log Likelihood =
378.13
205.99
886.20
480.83
235.53**
265.86**
Chi2 =
"/-statistics are in parentheses.
+ p< .10; *p< .05;**p< .01.
52.21**
29.98 +
S90
WHITE-MEANS AND CHOLLET
(reducing the number of hours worked or declining or not
seeking employment) or by not working at all.
Consideration of the literature suggests that the market
value of formal caregiving may be a poor measure of informal
caregivers' opportunity wages. Indeed, our estimates of informal caregivers' imputed wages suggest that the variance of
the value of informal caregivers' time is quite high, varying
from $0.78 per hour to $27.18 per hour, compared to assumed values ranging from $4.64 to $10.30 that appear in the
literature. Our estimates also suggest that caregivers' time
value is an important element in their decision, despite
sometimes substantial caregiving responsibilities.
On average, the probability that caregivers worked at all
in 1989 was slightly less than in 1982 and independent of the
probability of accommodation. The differences between
1982 and 1989 in the significance patterns of factors that
determine caregiver's accommodation and work decisions
are striking. In making decisions to reduce work or forgo job
opportunities (that is, to accommodate), employed caregivers have focused increasingly on the frailty of the disabled older person, the immediate relationship to the older
person, and the income constraints they may feel in making
these decisions. We speculate that between 1982 and 1989,
Medicare's practice of paying hospitals — and as a result,
the greater caregiving burdens thrust on informal caregivers
— may explain some of the changes in caregiver decisions
about accommodation.
In deciding to work at all, caregivers seem to have focused
increasingly on their opportunity wage in paid employment.
Furthermore, this focus may be stronger in periods of enhanced earnings opportunities (with lower general unemployment). While other factors remained significant (education, the availability of ongoing secondary ADL caregiver,
age, and gender), the magnitude of their effects declined.
Taken as a whole, these results suggest both a complex
decision process and a strong economic focus among caregivers in making workforce decisions.
ACKNOWLEDGMENTS
The research on which this study is based was financed in part by a grant
from the Retirement Research Foundation. The authors thank David Kass,
Mahmud Hassan, Bernard Friedman, and participants at the annual meeting
of the American Public Health Association held in October, 1994, for many
helpful comments and insights. We also thank David Burchfield, Vanina
Farber, and Stacy Halbert for their research assistance.
Address correspondence to Dr. Shelley I. White-Means, Department of
Economics, Fogelman College of Business and Economics, University of
Memphis, Memphis, TN 38152; Internet: White.meanss@msuvx2.
memphis.edu.
REFERENCES
Abel, E. 1990. Who Cares for Elderly? Public Policy and Experiences of
Adult Daughters. Philadelphia: Temple University Press.
Brody, E. 1981. "Women in the Middle and Family Help to Older
People.'' The Gerontologist 21:471 -480.
Cox, C. and A. Monk. 1990. "Minority Caregivers of Dementia Victims:
A Comparison of Black and Hispanic Families." Journal of Applied
Gerontology 9:340-354.
Ernst, R.L. and J.W. Hay. 1994. "The US Economic and Social Costs of
Alzheimer's Disease Revisited." American Journal of Public Health
84:1261-1264.
Finley, N. 1989. "Theories of Family Labor as Applied to Gender Differences in Caregiving for Elderly Parents." Journal of Marriage and the
Family 51:79-86.
George, L.K. and L.P. Gwyther. 1986. "Caregiver Well-being: A Multidimensional Examination of Family Caregivers of Demented Adults."
The Gerontologist 26:253-259.
Greene. W. 1993. Econometric Analysis. New York: McMillan.
Haley, W.E., E.G. Levine, S.L. Brown, J.W. Berry, and G.H. Hughes.
1987. " Psychological, Social, and Health Consequences of Caring for a
Relative with Senile Dementia." Journal of the American Geriatrics
Society 35:405-411.
Headen, A. 1992. "Time Costs and Informal Social Support as Determinants of Differences Between Black and White Families in the Provision
of Long Term Care." Inquiry 29:440-450.
Hu, T., L. Huang, and W. Cartwright. 1986. "Evaluation of the Costs of
Caring for the Senile Demented Elderly: A Pilot Study." The Gerontologist 26:158-163.
Kingson, E. and R. O'Grady-LeShane. 1993. "The Effects of Caregiving on
Women's Social Security Benefits." The Gerontologist 33:230-239.
Manton, K., L. Corder, and E. Stallard. 1993. "Estimates of Change in
Chronic Disability and Institutional Incidence and Prevalence Rates in
the U.S. Elderly Population From the 1982, 1984, and 1989 National
Long Term Care Survey." Journal of Gerontology: Social Sciences
48:S153-S166.
Matthews, S. and T. Rosner. 1988. "Shared Filial Responsibility: The
Family as the Primary Caregiver." Journal of Marriage and the Family
50:185-195.
Miller, B. and S. McFall. 1991. "The Effect of Caregiver's Burden on
Change in Frail Older Persons' Use of Formal Helpers." Journal of
Health and Social Behavior 32:165-179.
Muurinen, J.M. 1986. "The Economics of Informal Care: Labor Market
Effects in the National Hospice Study." Medical Care 24:1007-1017.
Rowland, D. and B. Lyons. 1991. Financing Home Care: Improving
Protection for Disabled Elderly People. Baltimore: Johns Hopkins
University Press.
Stephens, S. and J. Christianson. 1986. Informal Care of the Elderly.
Lexington, MA: D.C. Heath.
Stone, R., G. Cafferata, and J. Sangl. 1987. "Caregivers of the Frail
Elderly: A National Profile." The Gerontologist 27:616-626.
Stone, R. and P. Short. 1990. "The Competing Demands of Employment
and Caregiving to Disabled Elders." Medical Care 28:513-526.
White-Means, S. 1992. "Allocation of Labor to Informal Home Health
Production: Health Care for Frail Elderly, If Time Permits." Journal of
Consumer Affairs 26:69-89.
White-Means, S. 1994. "Expanded Home Health Financing and Economic
Consequences for Informal Caregivers." Paper presented at the Fifth
Annual Health Economics Conference, June 1994, Philadelphia,
Pennsylvania.
White-Means, S. and C. Chang. 1991. "Family Choices with Managed Care
for the Home Bound Elderly." Southern Economic Journal 58:203-224.
White-Means, S. and M. Thornton. 1990. "Labor Market Choices and
Home Health Care Provision Among Employed Ethnic Caregivers."
The Gerontologist 30:769-775.
Zick, C. and W.K. Bryant. 1983. "Alternative Strategies for Pricing Home
Work Time." Home Economic Research Journal 12:134-144.
Received November 21, 1994
Accepted August 22, 1995