Journal of Health Economics 28 (2009) 143154

Contents lists available at ScienceDirect

Journal of Health Economics

journal homepage: www.elsevier.com/locate/econbase

Does informal care from children to their elderly parents substitute for
formal care in Europe?
Eric Bonsang
CREPP, HEC-ULg Management School, Bd du Rectorat (B31), 7, 4000 Lige, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: This paper analyzes the impact of informal care by adult children on the use of long-term
Received 22 December 2007 care among the elderly in Europe and the effect of the level of the parents disability on this
Received in revised form 22 August 2008
relationship. We focus on two types of formal home care that are the most likely to inter-
Accepted 5 September 2008
act with informal care: paid domestic help and nursing care. Using recent European data
Available online 13 September 2008
emerging from the Survey on Health, Ageing and Retirement in Europe (SHARE), we build a
two-part utilization model analyzing both the decision to use each type of formal care or not
JEL classication:
I11 and the amount of formal care received by the elderly. Instrumental variables estimations
I12 are used to control for the potential endogeneity existing between formal and informal
J14 care. We nd endogeneity of informal care in the decision to receive paid domestic help.
Estimation results indicate that informal care substitutes for this type of formal home care.
Keywords:
However, we nd that this substitution effect tends to disappear as the level of disability of
Long-term care
the elderly person increases. Finally, informal care is a weak complement to nursing care,
Informal care
Elderly
independently of the level of disability. These results highlight the heterogeneous effects
SHARE of informal care on formal care use and suggest that informal care is an effective substitute
for long-term care as long as the needs of the elderly are low and require unskilled type of
care. Any policy encouraging informal care to decrease long-term care expenditures should
take it into account to assess its effectiveness.
2008 Elsevier B.V. All rights reserved.

1. Introduction

Ageing of the population in most developed countries will undoubtedly have important effects on the demand for long-
term care.1 The growing proportion of the elderly in the population is likely to increase substantially the demand for long-term
care (Yang et al., 2003; Pezzin et al., 1996). As a percentage of GDP, long-term care expenditure is projected to increase by
168% in Germany, by 149% in Spain, and by 138% in Italy between 2000 and 2050 (Comas-Herrera et al., 2003). The evolution
of long-term care expenditures depends on many factors, such as demographic trends or health, but also on the change in

This paper uses data from release 2 of SHARE 2004. The SHARE data collection was primarily funded by the European Commission through the 5th
framework program (project QLK6-CT-2001-00360 in the thematic program Quality of Life). Additional funding came from the US National Institute on
Ageing (U01 AG09740-13S2, P01 AG005842, P01 AG08291, P30 AG12815, Y1-AG-4553-01 and OGHA 04-064). Data collection in Austria (through the Austrian
Science Foundation, FWF), Belgium (through the Belgian Science Policy Ofce) and Switzerland (through BBW/OFES/UFES) was nationally funded. The SHARE
data collection in Israel was funded by the US National Institute on Aging (R21 AG025169), by the German-Israeli Foundation for Scientic Research and
Development (G.I.F.), and by the National Insurance Institute of Israel. Further support by the European Commission through the 6th framework program
(projects SHARE-I3, RII-CT-2006-062193, and COMPARE, CIT5-CT-2005-028857) is gratefully acknowledged. For methodological details see Brsch-Supan
and Jrges (2005).
Tel.: +32 4 366 31 10; fax: +32 4 366 31 06.
E-mail address: Eric.Bonsang@ulg.ac.be.
1
Long-term care is dened as a range of services for persons who are dependent on help with basic activities of daily living.

0167-6296/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhealeco.2008.09.002
144 E. Bonsang / Journal of Health Economics 28 (2009) 143154

informal care over time. One suggested solution for slowing down the increase in long-term care expenditure is to encourage
the development of informal care provided by the family to their frail, elderly family members. Indeed, family has always
been a major source of care for frail older individuals and it is thought that informal care is less costly than more formal care
arrangements. However, this solution is likely to lessen long-term care expenditure only if the informal care provided is an
effective substitute for formal care.
In this paper, we assess the effect of a change in the provision of informal care by extra-resident children on the use of
formal home care among the elderly using data from the Survey of Health, Ageing and Retirement in Europe (SHARE). We
focus on two types of formal home care that are the most likely to interact with informal care: paid domestic help and nursing
care. Paid domestic help consists of professional or paid home help in carrying out tasks such as doing work around the house
or the garden or shopping for groceries. Nursing care is dened as nursing or personal care provided by professionals.
Many past studies analyzing the relationship between formal and informal care were concerned about of the effect of pub-
lic support on informal care. Literature about this issue provides mixed results. Christianson (1988) shows that the increase
in formal care that occurred due to the Channeling (a National long-term care demonstration that took place during the
1980s in US) had virtually no effect on the supply of informal care. Moreover, Langa et al. (2001) nd a rather complemen-
tary relationship: the increase in home health care that took place during the 1990s in the United States mostly beneted
individuals with a relatively high level of social support. Motel-Klingebiel et al. (2005) nd no evidence of substantial crowd-
ing out of family help using data from Norway, England, Germany, Spain, and Israel. By contrast, Ettner (1994) shows that
Medicaid home care subsidies have increased the use of formal care and have reduced the amount of informal care among
no-institutionalized persons regarding no medical care in the United States. Pezzin et al. (1996) also nd a limited substitu-
tion between publicly provided home care and informal care. Instead of relieving informal caregivers, this public provision
of care can result in more help being given to the elderly. Stabile et al. (2006) show that increased availability of publicly
nanced home care is associated with an increase in its utilization and a decline in informal care giving in Canada. Viitanen
(2007) uses longitudinal data from 12 European countries (from the European Community Household Panel 19942001) and
nd that an increase in long-term care expenditure decreases informal caregiving undertaken outside the household.
By contrast, recent studies (Pezzin and Schone, 1999; Van Houtven and Norton, 2004, 2008; Bolin et al., 2008) analyzing
the relationship from the other point of view, i.e. the effect of informal care on the use of formal care, usually nd that
informal care substitutes for formal care, once controlling for endogeneity. Pezzin and Schone (1999) nd a substitution
relationship between paid home care when analyzing the informal care given by adult daughters to their elderly parents.
Van Houtven and Norton (2004) obtain different results depending on the type of care granted. They nd a net substitution
for all types of care except for outpatient surgery. Bolin et al. (2008) have also examined the effect of informal care on the use
of different types of formal and medical care among single-living elderly in Europe. They nd that informal care substitutes
for formal home care, while also being a complement to doctor and hospital visits.
In the present study, we extend the previous ndings in different directions: rst, we compare the effect of informal care
on skilled (nursing care) and unskilled (paid domestic help) formal home care. This is in contrast with Bolin et al. (2008) and
Van Houtven and Norton (2004), who consider the utilization of any formal home care without making such a distinction. We
show that informal care has a different effect with respect to the home care being considered. Second, we analyze whether
the relationship between formal and informal care differs according to the level of disability of the elderly person concerned.
Our results suggest that the relationship between formal and informal care is sensitive to the needs of the elderly, represented
by a constructed disability index.
Due to the sampling procedure of the rst wave of SHARE, the analysis does not include elderly living in nursing home,
which may represent a non-negligible proportion of formal care in several European countries, especially in Northern coun-
tries. This may potentially affect our results if the decision to institutionalize an older individual represents a substantial
substitution of formal care for informal care (Pezzin et al., 1996). Moreover, elderly co-residing with their children are
excluded from the analysis due to the difculties in evaluating the existence, the direction, and the importance of the intra-
household transfers taking place between co-residents (McGarry, 1999; Norton and Van Houtven, 2006). However, Section
6.2 analyzes the use of formal home care taking into account the effect of co-residence on the use of formal home care and
show consistent results with the main analysis.
The remainder of this paper is organized as follows. The next section briey discusses the conceptual framework and
highlights the main issues raised in analyzing the relationship between formal and informal care. Section 3 explains the
empirical model and discusses the econometric issues. Section 4 provides a description of the SHARE data and the variables
used in the empirical model. Section 5 presents the estimation results and Section 6 checks those results by performing
some sensitivity analyses regarding the set of instruments considered, and the inclusion of co-residence as another source
of informal care by the adult children. Finally, Section 7 concludes.

2. Formal versus informal care

Theoretical models related to the utilization of formal and informal care among the elderly are mainly based on the
family-decision making process and a health production function (Grossman, 1972; Van Houtven and Norton, 2004) or the
ability by the elderly person to perform activities of daily living function (Stabile et al., 2006), using as inputs formal and
informal care. According to these models, the relationship between informal and formal care depends therefore on the sign
of the derivative of the marginal product of formal care (in the production of health) with respect to informal care, which
 E. Bonsang / Journal of Health Economics 28 (2009) 143154 145

means that complementarity or substitution between formal and informal care is essentially an empirical issue (Bolin et
al., 2008). In this paper, we test whether the relationship between formal and informal care differs according to the level
of disability of the elderly person. Conceptually, it consists of analyzing whether the level of disability has an effect on the
derivative of the marginal product of formal care with respect to informal care in the production of health or the ability to
perform activities of daily living.
The empirical relationship between formal and informal care is not straightforward. As mentioned previously, prior
studies on this topic provide mixed results regarding the nature and the sign of the link between these two types of care.
This inconsistency highlights the complexity of the mechanisms lying behind the relationship between formal and informal
care. Several aspects are important to take into account in order to analyze the provision of formal and informal care.
First, endogeneity of informal care in a model explaining the use of formal care is likely to be present for at least two
reasons: the decision to provide informal care to parents and the decision to use formal care are simultaneously determined
(Van Houtven and Norton, 2004). Moreover, a spurious positive correlation may exist between formal and informal care.
This may be due to unobserved negative health characteristics, which are likely to increase the demand for both formal and
informal care (Charles and Sevak, 2005), or to unobserved preferences for care affecting both formal and informal care use.
Second, the relationship between formal and informal care is likely to differ according to the type of formal care used (Van
Houtven and Norton, 2004; Bolin et al., 2008). In practice, care encompasses a great number of services ranging from personal
care to gardening or shopping. While it is difcult to distinguish the type of informal care due to its multidimensional nature,
formal care providers can specialize in certain aspects of care and the type of care provided can be better identied. Informal
care is likely to be a substitute for formal care that requires low level skills such as grocery shopping or cleaning the house.
However, this substitution effect may not apply to formal care demanding higher level skills such as personal or nursing
care: children are less likely to be able to perform such tasks or the parent may be reluctant for their children to help them
with dressing, bathing or going to the toilet. Note however that there is evidence that children do perform personal care to
their parents (Romoren, 2003). As a result, we expect that the substitution effect between informal care and formal care will
be larger for paid domestic help (a low skilled care) than for nursing care.
Finally, the level of disability of the elderly person may change the relationship between these two types of care: informal
care is likely to be a substitute for an elderly person with a low level of disability, while it may become complementary for
highly handicapped individuals. When the dependency level is low, informal care is likely to substitute for formal care as the
type of help required demands low level skills and few commitments. However, as the level of disability increases, the burden
imposed by care becomes so heavy that it requires both formal and informal care. At this stage, the relationship between
formal and informal care is likely to become complementary, the informal carer acting as the agent of the dependent elderly
person in order to improve the formal care services. The informal carer is likely to be better informed about the needs of the
dependent and the formal care services than the dependent himself/herself, especially in case of mental health disease.

3. Empirical model

The empirical model consists of analyzing the causal effect of informal care from adult children to their elderly parents
on the utilization of formal home care. The analysis addresses the following issues: endogeneity, differential effects by types
of formal home care and by level of disability of the elderly person.
Two types of formal home care are considered: nursing care and paid domestic help. We use a two-part model introduced
by Duan et al. (1983), which allows the separation of behavior into two stages: rst a decision regarding receiving the
treatment and second, a decision regarding the level of this treatment, conditional on receiving any. The parents utilization of
formal care (gij ) is a function of informal care (hi ), the level of disability (Di ) and a vector of socio-demographic characteristics
(Xi ). The subscript i represents the individual and the j = 1, 2 the two types of formal care analyzed in this paper. The two-part
model assumes that part one, P(gij > 0), is described by a binary probit model such that:

P(gij > 0|Xi , hi , Di ) = (1j + Xi Xj + hj ln(1 + hi ) + f (Di , Dj )), (1)

where () represents the cumulative density function of the standard normal, f() is a polynomial in disability (Di ) allowing
non-linear disability effect on the utilization of formal care (our model uses the quadratic function) and  ij ,  Xj ,  hj , and
 Dj are the parameters to be estimated. Part two corresponds to the following equation assuming that the logarithm of the
positive values of gij is linear in Xi , ln(1 + hi ), and f(Di , ):

E[ln(gij )|gij > 0, Xi , hi , Di ] = 1j + Xi Xj + hj ln(1 + hi ) + f (Di , Dj ), (2)

where 1j , Xj , hj , and Dj are the parameters to be estimated by Ordinary Least Squares.

As mentioned earlier, we suspect that the main variable of interest, informal care (hi ), is endogenous. Regarding the
part one of the two-part model, the standard approach of instrumental variables estimation used in linear model provides
inconsistent estimates when applied to nonlinear model (Terza et al., 2008). Instead, Rivers and Vuong (1988) propose
a two-step approach in order to obtain consistent estimates for the probit model in the presence of endogeneity. This
method consists of including the residuals of the rst stage equation in the second stage equation. It is worth noting that
the resulting coefcients are estimated only up to scale. However, their sign and signicance are of interest. Furthermore,
we can consistently estimate the average partial effect and thus the elasticity of formal care with respect to informal care
146 E. Bonsang / Journal of Health Economics 28 (2009) 143154

by adjusting the scaled coefcients appropriately (Wooldridge, 2002, p. 474). Part one of the two-part model is estimated by
using Amemiyas Generalized Least Squares (Newey, 1987).2 Part two uses the standard two-stage least squares estimation.
The choice of the instrumental variables is driven by the potential effect that they can have on the receipt of informal
care by the children. Geographical distance from children is a possible candidate as it might represent a substantial cost
of caring on the part of the informal caregiver. Several studies have shown the importance of this variable for the provi-
sion of informal care (Charles and Sevak, 2005; Stern, 1995). Children living further away from their parents are less likely
to provide informal care than closer children. However, the condition requiring no correlation between the error terms of
the formal care equation and the instrumental variable is not necessarily satised. Parents may choose to move nearer to
their children when their own health deteriorates, or children may choose to live near their disabled parents. If the model
does not fully take into account the effects of health on the use of formal care, the error terms are likely to be negatively
correlated to the distance to the nearest child. However, Charles and Sevak (2005) nd no evidence that children are more
likely to live close by when their parent is in bad health. Bolin et al. (2008) use a dummy indicating whether the elderly
parent has children living less than 100 km away as an instrument using SHARE data and their instruments pass the
overidentication restriction test. Moreover, Stern (1995) shows that, although it is endogenous to the supply of informal
care by the children, distance is a strong predictor of informal care supply and the related endogeneity bias is very lim-
ited. A sensitivity analysis in Section 6.1 will test the effect of excluding distance from the set of instruments. Another
component with potential impact on informal care is the gender of the children, represented by the proportion of daugh-
ters. Numerous studies show that daughters provide more care to their parents than do sons (see, for example, Horowitz,
1985), although the potential of sons to provide informal care may be underestimated in modern societies (Carmichael and
Charles, 2003; Romoren, 2003). A test of overidentifying restriction will assess the validity of our set of instruments for
each of the model considered in this paper. It is worth noting that we tried other instruments for informal care receipt:
the number of grandchildren, the age of the children, and their level of education. All these instruments have only a poor
predictive power regarding the informal care receipt and are therefore not used as instruments for informal care in our main
analysis.
The model is rst estimated ignoring the possibility that the coefcient on informal care may differ according to the
disability level of the elderly. In a second step, this assumption is relaxed by including the interactions of informal care and
the disability level.

4. Data

SHARE is a European multi-disciplinary survey including more than 30,000 persons aged 50 and over, and who come
from 11 European countries ranging from Scandinavia to the Mediterranean, and Israel.3 We use in this paper release 2 of the
rst wave of the survey, which was conducted in 2004. The survey brings together many disciplines, including demography,
economics, epidemiology, psychology, and sociology. The data were collected using a computer assisted personal interview-
ing (CAPI) program, supplemented by a self-completion paper and pencil questionnaire. For more details on the sampling
procedure, questionnaire contents and eldwork methodology, readers should refer to Brsch-Supan and Jrges (2005).

4.1. Sample selection criteria

In order to be selected in our analysis, respondents had to be at least 65 years old. They also had to have between one
and four children4 and to live neither with one of their children nor in an institution. We excluded individuals living with
their children because it is difcult to distinguish the way, the type, and the importance of the transfers that take place
within a household. However, Section 6.2 is devoted to the analysis of the effect of co-resident children on the utilization of
formal home care by the parents. Moreover, the analysis focuses on the formal care in the community because the sampling
design of the rst wave of SHARE does not include institutionalized individuals. This might potentially affect our results if the
decision to institutionalize an older individual represents a substantial substitution of formal care for informal care (Pezzin
et al., 1996). Moreover, we discard observations with missing or unreliable values for the variables of interest and the other
explanatory variables. Finally, we do not use data from Greece, Israel, or Switzerland because we did not have information
regarding formal home care use in these countries. Our nal sample includes 7329 observations for nine European countries.
Table 1 presents the summary statistics of the main variables used in the model.

4.2. Dependent variables: the utilization of formal care

This paper analyzes the utilization of two types of formal home care: paid domestic help and nursing care. In SHARE,
respondents were asked to report the number of weeks and the average number of hours per week they received either paid

2
The statistical package used is Stata 9 and the estimation command is ivprobit.
3
The rst wave of SHARE data includes 12 countries: Austria (AU), Belgium (BE), Germany (DE), Denmark (DK), France (FR), Greece (GR), Italy (IT), The
Netherlands (NL), Spain (ES), Sweden (SE), Switzerland (CH) and Israel (IL).
4
Parents with more than four children account only for a small fraction of the population.
 E. Bonsang / Journal of Health Economics 28 (2009) 143154 147

Table 1
Summary statistics.

All countries Austria Belgium Denmark France Germany Italy Netherlands Spain Sweden

Number of observations 7,329 605 1,166 521 921 1,004 618 802 578 1,114

Receive paid domestic help (%) 8.7 5.0 14.7 17.5 10.3 2.2 2.6 12.7 6.2 6.5
Receive nursing care (%) 6.1 4.0 11.7 5.6 16.7 1.9 1.9 3.7 4.2 1.7

Hours of informal care from 5.6 7.7 8.1 4.6 7.0 11.6 11.6 1.6 12.4 3.4
the children per month
Woman (%) 53.5 58.8 54.2 55.7 58.1 51.6 53.1 48.5 54.8 50.2
Age 73.3 73.1 73.3 74.3 74.0 72.3 72.7 73.1 73.8 73.7
Single household (%) 27.7 42.5 29.2 39.7 34.1 21.6 20.9 21.8 18.0 25.9
Years of education 9.4 10.9 9.4 11.7 7.5 13.2 5.8 10.5 4.2 9.5

Limited in activities because of a health problem

Severely limited (%) 17.2 15.9 18.2 15.2 20.4 21.6 16.0 19.6 6.4 15.4
Limited, but not severely (%) 33.9 40.3 28.9 37.6 27.8 41.0 35.0 26.3 42.6 32.6
Not limited (%) 49.0 43.8 52.9 47.2 51.8 37.4 49.0 54.1 51.0 52.0

Yearly household gross income 40,816 36,612 41,715 40,614 47,429 40,157 27,332 52,937 18,231 47,855
in D (median)
Net worth in D (median) 342,175 176,239 412,665 399,135 447,532 351,774 336,401 348,514 288,654 262,531

Number of sons 1.1 1.0 1.1 1.2 1.1 1.0 1.1 1.2 1.1 1.2
Number of daughters 1.1 1.1 1.1 1.2 1.1 1.1 1.1 1.2 1.1 1.1
Total number of children 2.2 2.1 2.3 2.4 2.2 2.1 2.2 2.4 2.3 2.3

Distance from the nearest child 42.9 39.9 21.7 48.1 71.3 51.0 27.9 25.8 32.4 59.7
(in kilometers)

Source: SHARE 2004. Sample includes all individuals being 65 year old or over having at least one child and not living with them.

or professional home help, which they could not perform due to health problem, or professional or paid nursing or personal
care during the 12 months preceding the survey. The dependent variables are expressed as the average number of hours
received per month.

4.3. Informal care receipt and the instrumental variables

The informal care measure used in this paper is the average total number of hours of informal care received from the chil-
dren of the respondent per month.5 We also include informal care provided by a childs spouse and children (the respondents
grandchildren). This variable is based on three questions that the respondent was asked in order to describe the relationship
with the caregiver (if any), the frequency (daily, weekly, monthly or less often) of informal care received and the average
number of hours per day/week/month/year, respectively. We transform these variables into a measure of the average total
number of hours of informal care received from the children per month.
Informal care given by the children is assumed to depend on several characteristics of the children that are independent
of the use of formal care. As discussed in the previous section we assume that geographical proximity and the gender of the
children mainly explain informal care receipt of older individuals. Gender composition of the children is represented by the
proportion of daughters. Geographical proximity corresponds to the distance of the nearest child from the elderly parent. In
SHARE, this variable is allowed to take the following categories: the children can live either in the same building (but not the
same household), less than 1 km away, between 1 and 5 km away, between 5 and 25 km away, between 25 and 100 km away,
between 100 and 500 km away or more than 500 km away. From this variable, we compute a new variable, the distance from
the nearest child, by assigning the number of kilometers corresponding to the middle of the bandwidth of each possible
categorical answer.

4.4. The parents explanatory variables

The utilization of formal care is assumed to depend on various characteristics of the parent. First, our empirical model
includes several socio-demographic variables: gender, age, number of years of education,6 and household composition
(whether the parent lives alone or not). Moreover, the model also includes dummies related to the country-specic

5
In SHARE, information regarding the amount of informal care received is collected for only up to three potential informal caregivers. So, if the respondent
has more than three caregivers, it is possible that our variable of informal care underestimates the amount of informal care provided by all the children
(and the childrens spouse and their children).
6
The 1997 International Standard Classication of Education (OECD, 1999) is used for the computation of the years of education variable for the different
SHARE-participating countries.
148 E. Bonsang / Journal of Health Economics 28 (2009) 143154

net-worth quartile (including housing wealth) and gross household income quartile of the respondent in order to take into
account the ability of the elderly parent to pay for care.7

4.5. The disability index

A major issue when dealing with disability or health status is how to measure it. In survey data, they are mainly measured
using self-reported health or limitations with daily activities. However, these subjective variables raise a number of potential
problems (Bound, 1991; Bound et al., 1999; Dwyer and Mitchell, 1999; Baker et al., 2004). More recent household surveys, like
SHARE, also include several more objective and detailed health indicators that are less susceptible to measurement error
and endogeneity, since questions are narrower and more concrete (Bound et al., 1999). However, these variables also raise
several problems. First, even if we have numerous measures available, they may only provide a partial measure of overall
health (Bound, 1991; Dwyer and Mitchell, 1999). Second, the objective health variables are also subject to measurement
errors (Baker et al., 2004). Third, including all the objective health characteristics directly into the model is likely to induce
multi-collinearity because of co-morbidity and may make the interpretation of these results difcult. In order to create
an index of disability, we follow the strategy of Bound et al. (1999) by using objective health variables available in SHARE
to predict self-reported measure of functional limitation. This disability index attenuates the reporting bias in the self-
reported variable, and does not suffer from multi-collinearity (Bound et al., 1999). This index is constructed on the basis of a
subjective categorical variable from the SHARE questionnaire that asks to which extent, if any, the respondent is limited in his
daily activities because of health-related problem, and numerous objective variables on the health status of the individual.
These objective variables include a set of dummies related to chronic diseases, reported symptoms, limitations with the
Activities of Daily Living (Katz et al., 1970) and Instrumental Activities of Daily Living (Lawton and Brody, 1969), and the
mobility limitations of the respondent. The disability index of the respondent is obtained by performing an ordered probit
model with the limitation question as dependent variable and the set of dummies as explanatory variables.8 Most of the
variables have an expected sign, and are signicantly different from zero. From these results, we compute the predicted
latent variable and use its standardized value as disability index.

5. Estimation results

In this section, we present the estimation results for our model. We show both the estimation of the two-part model with
regard to the provision of informal care as exogenous and the same model taking into account the possible endogeneity of
this variable on the utilization of formal care. Moreover, we distinguish two types of formal home care: nursing care and
paid domestic help. All equations include country-dummies in order to take into account the cross-country heterogeneity
regarding the provision of formal and informal care.

5.1. Informal care receipt equation

The rst column of Table 2 presents the estimation results for the informal care equation. The model has an adjusted-R2 of
17.8%. The variable corresponding to the proportion of daughters has a signicant positive impact on the provision of informal
care to older parents. The distance to the nearest child and its square are highly signicant regarding informal receipt. The
further away the children are, the lower the provision of informal care. The F-test of the excluded instruments suggested by
Bound et al. (1995) conrms that these are signicant predictors of informal care receipt (F(3, 7305) = 40.07).
It is worth noting that informal care depends on several other variables included in the model. First, informal care provision
is higher among older individuals and those with low level of education. Moreover, individuals living alone are more likely
to receive informal care. Finally, results indicate that the disability level of the elderly signicantly increases informal care
receipt.

5.2. Paid domestic help utilization equation

The last four columns of Table 2 present the estimation results of the two-part model for the use of paid domestic help.
Looking at the model assuming exogeneity of informal care by adult children, the estimated coefcients on the logarithm
of total hours of informal care are positive and signicant for both the choice and the intensity equation. However, the null
hypothesis of the Wald test of exogeneity of informal care is signicantly rejected in the instrumental variable (IV) probit
equation (21 = 13.48) indicating that informal care is endogenous to the decision to use paid domestic help. As a result, the
coefcient estimates from the simple probit equation are inconsistent. The model passes the overidentication restriction
test (22 = 1.38) suggesting that the selected instruments are independent of the error terms of the structural equation. The IV
probit equation reveals a negative and highly signicant relationship between the decision to obtain paid domestic help and
informal care. This conrms our expectations that informal care is a substitute for paid domestic help. Regarding the intensity

7
Household income is only available in gross amount in the wave 1 of SHARE.
8
The results from the ordered probit model are available upon request.
 E. Bonsang / Journal of Health Economics 28 (2009) 143154 149

Table 2
Two-part model of paid domestic help utilization and the rst-stage equation of informal care receipt.

Dependent variable Informal care Paid domestic help

OLS Probit IV Probit OLS 2SLS

Intercept 1.945*** (0.177) 6.643*** (0.350) 7.894*** (0.517) 0.662 (0.566) 0.808 (0.609)
Informal care 0.050*** (0.018) 0.594*** (0.185) 0.121*** (0.025) 0.184* (0.106)
Woman 0.091*** (0.028) 0.259*** (0.058) 0.314*** (0.063) 0.105 (0.094) 0.112 (0.093)
Age 0.032*** (0.002) 0.058*** (0.004) 0.078*** (0.007) 0.010 (0.006) 0.008 (0.008)
Years of education 0.019*** (0.004) 0.009 (0.008) 0.006 (0.009) 0.027** (0.013) 0.029** (0.013)
Single household 0.378*** (0.036) 0.503*** (0.065) 0.743*** (0.098) 0.057 (0.101) 0.049 (0.100)

Income quartile
1st
2nd 0.070* (0.036) 0.107 (0.070) 0.145* (0.074) 0.057 (0.107) 0.064 (0.106)
3rd 0.043 (0.041) 0.017 (0.087) 0.038 (0.092) 0.359** (0.145) 0.356** (0.143)
4th 0.121*** (0.043) 0.180** (0.087) 0.246*** (0.093) 0.305** (0.139) 0.334** (0.145)

Wealth quartile
1st
2nd 0.011 (0.036) 0.026 (0.068) 0.035 (0.072) 0.017 (0.105) 0.025 (0.104)
3rd 0.031 (0.038) 0.108 (0.077) 0.143* (0.082) 0.094 (0.127) 0.102 (0.126)
4th 0.009 (0.041) 0.023 (0.087) 0.040 (0.091) 0.261* (0.147) 0.277* (0.147)

Disability index 0.218*** (0.019) 0.525*** (0.041) 0.671*** (0.060) 0.069 (0.069) 0.047 (0.078)
(Disability index)2 0.039*** (0.010) 0.046*** (0.017) 0.022 (0.020) 0.060** (0.025) 0.063** (0.025)

Instrumental variables
Proportion of daughters 0.086** (0.035)
Distance to the nearest child 0.310*** (0.045)
(Distance to the nearest child)2 0.045*** (0.010)

Country dummies Yes Yes Yes Yes Yes

(Pseudo-) R2 0.178 0.336 0.213 0.232

Number of observations 7,329 7,329 7,329 635 635

Notes: SHARE 2004. Sample includes all individuals being 65 year old or over having at least one child and not living with them. Asterisks (*), (**), (***)
means that the coefcient estimate is signicantly different from zero at the 10%, 5%, 1% level, respectively. Standard errors are in parentheses. Informal
care corresponds to the logarithm of the number of hours of informal care provided to the parents (plus one).

equation of paid domestic help, the null hypothesis of the Wu-Hausman test is not rejected (F(1, 612) = 0.37) indicating that
the exogeneity of informal care on the intensity of paid domestic help cannot be rejected. However, as endogeneity has been
detected in part one of the utilization of paid domestic help model, we focus on the results assuming endogeneity even where
the data failed to reject the hypothesis of exogenity for the intensity equation. The effect of informal care on the number of
hours of paid domestic help is positive, but only weakly signicant (at the 10% level). From these results, we compute the
elasticity of paid domestic use with respect to informal care for the average individual taking into account the fact that the
parameters of the IV probit are only estimated up to scale. A 10% increase in informal care leads to a decrease by 6.8% in the
use of paid domestic help (see Table 6).
It is worth noting several other results obtained from this analysis. Regarding individual characteristics, elderly women
are more likely to use paid domestic help than men. Age is also a factor that increases the need for more paid domestic help.
Education plays no role in the decision to use such care but it has a signicant positive impact regarding the quantity of
domestic help conditional on having any. The coefcients on income quartiles indicate a positive and signicant relationship
between income and paid domestic help. The disability index is, as expected, an important factor in the decision to obtain
formal domestic help among the elderly in Europe. Moreover, elderly parents living alone are much more likely to use this
kind of formal care, suggesting substitution effect of informal care received from other individuals (mainly the spouse) living
in the same household.

5.3. Nursing care utilization equation

Table 3 presents the results from the two-part model of nursing care utilization. Regarding the decision to receive such
care, the Wald test does not reject the exogeneity of informal care on the dependent variable (21 = 0.06). Moreover, the
null hypothesis of the Wu-Hausman test of exogeneity of informal care on the intensity of nursing care is not rejected (F(1,
425) = 0.00). As a result, the simple two-part model is preferred to the model assuming endogeneity of informal care on
nursing care. The simple two-part model exhibits a positive and signicant relationship between informal care and the
decision to receive nursing care but it is not signicant regarding the quantity of nursing care received, conditional on
receiving any. This suggests that informal care from the children is complementary to the utilization of nursing care in the
decision to resort to such care. However, the magnitude of this effect is limited: a 10% increase in informal care leads to an
increase by 2% in nursing care use for the average individual (see Table 6).
150 E. Bonsang / Journal of Health Economics 28 (2009) 143154

Table 3
Two-part model of nursing or personal care utilization.

Dependent variable Nursing/personal care utilization

Probit IV Probit OLS 2SLS

Intercept 3.753*** (0.379) 3.857 (0.538) 0.325 (0.688) 0.310 (0.728)

Informal care 0.092*** (0.020) 0.038 (0.200) 0.036 (0.032) 0.028 (0.147)
Woman 0.030 (0.062) 0.035 (0.064) 0.021 (0.117) 0.021 (0.114)
Age 0.017*** (0.005) 0.018** (0.008) 0.006 (0.008) 0.006 (0.010)
Years of education 0.003 (0.008) 0.002 (0.009) 0.008 (0.013) 0.008 (0.014)
Single household 0.111 (0.074) 0.131 (0.104) 0.174 (0.141) 0.180 (0.186)

Income quartile
1st
2nd 0.048 (0.075) 0.045 (0.076) 0.012 (0.139) 0.012 (0.136)
3rd 0.035 (0.089) 0.034 (0.090) 0.031 (0.162) 0.031 (0.158)
4th 0.008 (0.091) 0.003 (0.093) 0.095 (0.167) 0.093 (0.167)

Wealth quartile
1st
2nd 0.093 (0.075) 0.093 (0.075) 0.146 (0.134) 0.143 (0.140)
3rd 0.075 (0.082) 0.077 (0.083) 0.200 (0.150) 0.205 (0.166)
4th 0.020 (0.090) 0.020 (0.091) 0.002 (0.169) 0.003 (0.167)

Disability index 0.440*** (0.044) 0.452*** (0.062) 0.365*** (0.087) 0.368*** (0.100)
(Disability index)2 0.009 (0.018) 0.011 (0.019) 0.023 (0.030) 0.023 (0.031)

Country dummies Yes Yes Yes Yes

(Pseudo-) R2 0.275 0.318 0.350

Number of observations 7,329 7,329 448 448

Notes: SHARE 2004. Sample includes all individuals being 65 year old or over having at least one child and not living with them. Asterisks (*), (**), (***)
means that the coefcient estimate is signicantly different from zero at the 10%, 5%, 1% level, respectively. Standard errors are in parentheses. Informal
care corresponds to the logarithm of the number of hours of informal care provided to the parents plus one. The instruments for informal care include the
proportion of daughters and the distance from the nearest child.

It is worth noting that, except for age and disability, all explanatory variables included in the model have no signicant
effect on the utilization of nursing care.

5.4. The relationship between informal care and formal care utilization according to the level of disability

This section analyzes in deeper details the effect of informal care on formal care according to the level of disability of
the elderly parent. In order to carry out this, we introduce into the empirical model the interaction between informal care
and the disability level. The model now has one additional endogenous variable and thus requires additional instruments:
these are naturally the interaction of the distance from the nearest child and the disability index and the interaction of the
proportion of daughters and the disability index.
Table 4 presents the results of the paid domestic help model including the interaction term between informal care and
the level of disability of the elderly parent. The exogeneity hypothesis of the Wald test is signicantly rejected regarding the
decision to use paid domestic help (22 = 14.86) and the instruments pass the over-identication tests (23 = 1.08). However,
the hypothesis of exogeneity of informal care is not rejected for the intensity of paid domestic help (F(2, 610) = 1.88). As
expected, the interaction term between informal care and the disability index is positive and signicant suggesting that the
substitution effect is lower for elderly suffering from high disability level. Regarding the intensity of paid domestic help,
the coefcient on informal care is negative but not signicant, and the coefcient on the interaction term is both positive
and signicant, conrming our results. The elasticity of paid domestic help with respect to informal care for the average
individual is computed at three different values of the disability index corresponding to individuals suffering from no limita-
tion (disability index = 0.57), those being limited but not severely (disability index = 0.19) and those being severely limited
(disability index = 1.18). The elasticity estimates are 1.62, 0.94 and 0.23, respectively, the latter being not signicant at
the 5% level (see Table 6).
The results from the nursing care equation are not presented because the inclusion of the interaction term between
informal care and the level of disability has no signicant effect and does not change the results from the previous section.

6. Sensitivity analysis

6.1. The instruments

One crucial identication assumption of our model is that geographical proximity is independent to the error term of
the formal care equation. However, children may live closer to their parents when the latter are in worse health, or they
 E. Bonsang / Journal of Health Economics 28 (2009) 143154 151

Table 4
Two-part model of paid domestic help utilization (extended).

Dependent variable Paid domestic help

Probit IV Probit OLS 2SLS

Intercept 6.694*** (0.351) 7.882*** (0.523) 0.731 (0.569) 1.271 (0.650)

Informal care 0.087*** (0.023) 0.941*** (0.302) 0.090** (0.036) 0.019 (0.202)
Disability index informal care 0.039*** (0.015) 0.429** (0.173) 0.026 (0.021) 0.169 (0.107)
Woman 0.263*** (0.059) 0.312*** (0.066) 0.108 (0.094) 0.132 (0.097)
Age 0.058*** (0.004) 0.080*** (0.008) 0.010 (0.006) 0.003 (0.008)
Years of education 0.009 (0.008) 0.012 (0.011) 0.027** (0.013) 0.032** (0.014)
Single household 0.498*** (0.066) 0.800*** (0.114) 0.060 (0.101) 0.071 (0.106)

Income quartile
1st
2nd 0.106 (0.070) 0.153* (0.079) 0.060 (0.107) 0.080 (0.109)
3rd 0.017 (0.087) 0.044 (0.096) 0.352** (0.145) 0.314** (0.150)
4th 0.176** (0.087) 0.277*** (0.101) 0.307** (0.139) 0.350** (0.149)

Wealth quartile
1st
2nd 0.029 (0.068) 0.020 (0.076) 0.016 (0.105) 0.019 (0.108)
3rd 0.108 (0.077) 0.139 (0.086) 0.096 (0.127) 0.113 (0.131)
4th 0.030 (0.087) 0.030 (0.096) 0.252* (0.147) 0.217 (0.160)

Disability index 0.543*** (0.041) 0.540*** (0.060) 0.051 (0.071) 0.077 (0.100)
(Disability index)2 0.034* (0.018) 0.169*** (0.058) 0.056** (0.025) 0.033 (0.033)

Country dummies Yes Yes Yes Yes

(Pseudo-) R2 0.337 0.214 0.177

Number of observations 7,329 7,329 635 635

Notes: SHARE 2004. Sample includes all individuals being 65 year old or over having at least one child and not living with them. Asterisks (*), (**), (***)
means that the coefcient estimate is signicantly different from zero at the 10%, 5%, 1% level, respectively. Standard errors are in parentheses. Informal
care corresponds to the logarithm of the number of hours of informal care provided to the parents plus one. The instruments for informal care include the
proportion of daughters, the distance from the nearest child and their interactions with the disability index.

may choose to live nearer to their parents if the availability of formal care is scarce in the region where the parents live.
In this case, geographical distance may not be a valid instrument for informal care in a model of formal care utilization.
As a check, we estimate the model after having dropped geographical distance from the model. Moreover, we add to our
sample all individuals who are aged 65 years or older and who have no children (1291 individuals). The instruments for
informal care are now the number of sons and the number of daughters of the respondents, and their interactions with the
disability index. Table 5 presents the results of the extended two-part model of paid domestic help.9 Results are consistent
with those obtained from the model using geographical distance as additional instrument. The coefcient on informal care
is still negative and signicant at the 10% level and the coefcient on the interaction term between informal care and the
disability level is still positive but no more signicant for the decision equation. Regarding the intensity of the use of paid
domestic help, the coefcient on informal care is also negative and signicant while the coefcient on the interaction term is
both positive and signicant. The estimated elasticity of the use of paid domestic help with respect to informal care is 1.26
for individuals with no limitations, 0.72 for respondents who are limited, but not severely so, and 0.18 for those who are
severely limited with their daily activities (see Table 6). These results suggest that the use of geographical distance from the
nearest child as an additional instrument for informal care does not affect the main results of this analysis but improves their
accuracy.

6.2. The effect of co-residence with adult children on formal home care

The previous analysis does not include individuals living in institutions or co-residing with any of their children. In
particular, co-residence represents an important source of informal care, especially in Southern European countries, while
institutional care represent an important source of formal care for the oldest-old, especially in Northern European countries.
These restrictions are likely to affect our results, especially for highly handicapped individuals, which may be more likely to
use these types of formal and informal care arrangement. Unfortunately, the rst wave of SHARE does not include institu-
tionalized individuals and thus precludes any analysis on the effect of informal care on this type of formal care. However,
SHARE allows controlling the effect of living arrangement on the utilization of formal home care. For this purpose, we include
in our sample all >65 individuals living with any of their children, but have at least one extra-resident child. This sample
selection allows measuring both the effect of informal care provided by extra-residents and the effect of co-residence on the

9
The results from the nursing care equation are not shown as the hypothesis of exogeneity of informal care is not rejected. The results from the simple
two-part model from Table 3 still hold.
152 E. Bonsang / Journal of Health Economics 28 (2009) 143154

Table 5
Two-part model of paid domestic help utilization (extended) without using distance as an instrument.

Dependent variable Paid domestic help

Probit IV Probit OLS 2SLS

Intercept 6.219*** (0.302) 6.669*** (0.364) 0.787 (0.484) 1.002* (0.515)

Informal care 0.074*** (0.022) 0.447* (0.230) 0.069** (0.035) 0.162 (0.132)
Disability index informal care 0.036** (0.014) 0.217 (0.143) 0.033 (0.020) 0.200*** (0.072)
Woman 0.242*** (0.052) 0.269*** (0.054) 0.112 (0.084) 0.121 (0.086)
Age 0.053*** (0.004) 0.062*** (0.005) 0.009 (0.006) 0.007 (0.006)
Years of education 0.010 (0.007) 0.000 (0.008) 0.022* (0.011) 0.024** (0.012)
Single household 0.547*** (0.058) 0.647*** (0.073) 0.157* (0.091) 0.176* (0.094)

Income quartile
1st
2nd 0.087 (0.062) 0.113* (0.065) 0.111 (0.097) 0.120 (0.100)
3rd 0.082 (0.077) 0.091 (0.079) 0.407*** (0.127) 0.375*** (0.131)
4th 0.202** (0.078) 0.245*** (0.083) 0.210* (0.127) 0.201 (0.133)

Wealth quartile
1st
2nd 0.048 (0.060) 0.040 (0.062) 0.054 (0.092) 0.049 (0.094)
3rd 0.110 (0.068) 0.118* (0.070) 0.083 (0.111) 0.078 (0.114)
4th 0.089 (0.079) 0.077 (0.081) 0.043 (0.133) 0.048 (0.142)

Disability index 0.552*** (0.037) 0.546*** (0.045) 0.073 (0.063) 0.019 (0.077)
(Disability index)2 0.042*** (0.016) 0.105** (0.041) 0.041* (0.022) 0.020 (0.025)

Country dummies Yes Yes Yes Yes

(Pseudo-) R2 0.325 0.192 0.146

Number of observations 8,620 8,620 804 804

Notes: SHARE 2004. Sample includes all individuals being 65 year old and not living with any child, if any. Asterisks (*), (**), (***) means that the coefcient
estimate is signicantly different from zero at the 10%, 5%, 1% level, respectively. Standard errors are in parentheses. Informal care corresponds to the
logarithm of the number of hours of informal care provided to the parents plus one. The instruments for informal care include the number of sons and
daughters and their interactions with the disability index.

utilization of paid domestic help and nursing care. Living arrangement might also be endogenous to the utilization of formal
home care. As a result, we use instrumental variables techniques to deal with this issue. We use the age of the youngest child,
the proportion of children living in a couple, and the proportion of children being parents as instrumental variable for living
arrangement. These variables are signicantly correlated to living arrangements and can reasonably be assumed not to be
correlated to the error term of the formal care utilization equation. This set of instruments passed the overidentication
restriction test. Note also that the explanatory variable living in a single household is replaced by living in a couple to
take into account the potential informal care of the spouse. Table 7 reports the coefcient estimates of the extended two-part
model of paid domestic help including the living arrangement dummy and its interaction term with the disability index.
Results for informal care provided by extra-residents children conrm previous analysis: the coefcient on informal care is
negative and signicant and its interaction term with the disability index is positive and signicant. Moreover, co-residence
has a negative and signicant effect on the utilization of paid domestic help. However, the interaction term with the disability
index is positive but not signicant, once we account for endogeneity, suggesting that co-residence is an effective substitute
for paid domestic help whatever the disability level of the elderly.

Table 6
Elasticity of formal care with respect to informal care.

Limitation with daily activities Paid domestic help Nursing care

Elasticity [95% C.I.] Elasticity [95% C.I.]

Baseline model 0.68 a

[1.08; 0.32] 0.22a
[0.09; 0.31]

Extended model All 1.11a [1.68; 0.66]

No limitation 1.62a [2.13; 0.97]
Not severely limited 0.94a [1.52; 0.65]
Severely limited 0.23 [0.52; 0.02]

Extended model without distance All 0.85a [1.44; 0.26]

No limitation 1.26a [1.92; 0.32]
Not severely limited 0.72a [1.08; 0.14]
Severely limited 0.18 [0.39; 0.04]

Note: The 95% condence intervals are bootstrapped and bias-corrected.

a
The coefcient is signicant at the 5% level.
 E. Bonsang / Journal of Health Economics 28 (2009) 143154 153

Table 7
Two-part model of paid domestic help use (extended).

Dependent variable Paid domestic help

Probit IV Probit OLS 2SLS

Intercept 6.000*** (0.348) 6.478*** (0.435) 1.176** (0.581) 1.822*** (0.670)

Co-residence 0.259* (0.136) 1.154*** (0.414) 0.159 (0.301) 0.532 (1.195)
Disability index co-residence 0.212*** (0.077) 0.544 (0.331) 0.081 (0.147) 0.544 (0.679)
Informal care 0.090*** (0.022) 0.773*** (0.273) 0.090** (0.037) 0.058 (0.192)
Disability index informal care 0.041*** (0.014) 0.267** (0.129) 0.023 (0.021) 0.139 (0.089)
Woman 0.262*** (0.057) 0.283*** (0.062) 0.160* (0.093) 0.190* (0.098)
Age 0.055*** (0.004) 0.071*** (0.007) 0.007 (0.006) 0.002 (0.008)
Years of education 0.013* (0.007) 0.007 (0.010) 0.022* (0.013) 0.032** (0.016)
Living in a couple 0.470*** (0.063) 0.723*** (0.105) 0.165* (0.100) 0.090 (0.127)

Income quartile
1st
2nd 0.093 (0.067) 0.112 (0.074) 0.078 (0.106) 0.041 (0.115)
3rd 0.033 (0.082) 0.045 (0.091) 0.388*** (0.141) 0.264 (0.177)
4th 0.200** (0.082) 0.271*** (0.093) 0.296** (0.135) 0.259 (0.176)

Wealth quartile
1st
2nd 0.023 (0.065) 0.041 (0.071) 0.004 (0.104) 0.028 (0.113)
3rd 0.068 (0.073) 0.119 (0.081) 0.067 (0.124) 0.123 (0.139)
4th 0.059 (0.083) 0.076 (0.090) 0.197 (0.143) 0.214 (0.158)

Disability index 0.556*** (0.040) 0.543*** (0.064) 0.010 (0.071) 0.132 (0.098)
(Disability index)2 0.040** (0.017) 0.143*** (0.047) 0.075*** (0.025) 0.048 (0.036)

Country dummies Yes Yes Yes Yes

2
(Pseudo-) R 0.331 0.204 0.144
Number of observations 8,064 8,064 674 674

Notes: SHARE 2004. Sample includes all individuals being 65 year old and having at least one extra-resident child. Asterisks (*), (**), (***) means that the
coefcient estimate is signicantly different from zero at the 10%, 5%, 1% level, respectively. Standard errors are in parentheses. Informal care corresponds
to the logarithm of the number of hours of informal care provided to the parents plus one. The instruments for informal care and co-residence include the
proportion of daughters, the distance from the nearest child, the proportion of children living in a couple, the proportion of children being parent, the age
of the youngest child, and their interactions with the disability index.

7. Conclusion

This paper examines the effect of informal care from the children on the use of formal home care of the elderly in Europe.
Using data from the rst wave of SHARE (2004), we construct a two-part utilization model estimating the effects of informal
care receipt from the children on the utilization of paid domestic help and nursing/personal care among the >65 individuals in
nine European countries. The model takes into account the potential endogenous relationship between formal and informal
care by using children characteristics as instruments. Results are consistent results with ndings of Bolin et al. (2008) and
Van Houtven and Norton (2004), who identied a substitution relationship between informal and formal home care. We
extend their ndings by allowing the effect of informal care to differ according to low and high level skills care. Informal
care is found to decrease low-skilled home care use (paid domestic help) while it is a complement to high-skilled home care
(nursing/personal care). Furthermore, we extend the previous results by taking into account the effect of the disability level
on the relationship between formal and informal care. We show that the substitution effect vanishes for elderly suffering
from heavy disability.
As co-residence with the children may also be an important source of informal care in Europe, we extend the analysis to
older individuals living with one of their children as another source of informal care. They conrm the relationship between
extra-resident informal care and formal home care. Moreover, they show that co-residence is a substitute for paid domestic
help whatever the disability level of the elderly, while it is not related to the utilization of nursing care.
It is worth noting some caveats from this analysis. In particular, it focuses on elderly living in the community. This
may affect the estimates if highly disabled individuals move into institutions. In particular, it may obscure the substitution
relationship at higher level of disability. Next waves of SHARE will allow analyzing the relationship between informal care
and nursing home entry. Moreover, this study does not distinguish private and publicly funded formal services due to the
lack of information in SHARE data. Assessing the effects of informal care on the use of public and private care is of high
relevance for the organization of long-term care and the trade-off between the state, the market, or the family, but is beyond
the scope of this paper.
Our results highlight the complexity of the relationship between formal and informal care. While informal care is
an important source of care for the elderly, it provides an alternative to formal services as long as the burden and the
complexity of the dependency is limited. As long as the elderly is not too severely disabled the family is able to take in
charge the care responsibilities. However, once the disability level increases and the burden becomes heavy and requires
154 E. Bonsang / Journal of Health Economics 28 (2009) 143154

more skilled support, children and the formal services tend to be used together in order to cope with loss of autonomy of the
elderly.
In Europe, several countries are thinking about the way to limit the expected increase in long-term care expenditures.
Among the possible solutions, some policy-makers have suggested to encourage informal care provided by the family, as an
alternative for more costly formal care. The effectiveness of such policies on long-term care expenditures heavily depends
on the assumption of substitutability between informal and formal care. This paper suggests that substitution between
informal and formal care only holds for unskilled care for elderly suffering from low disability level, limiting the potential
role of informal care to cope with the future needs of the growing share of older individuals in the population.

Acknowledgements

I am especially grateful to Izabela Jelovac and Claire Marchal for helpful discussions. Moreover, I wish to thank Sergio
Perelman, Pierre Pestieau, Courtney Van Houtven, Arthur Van Soest, Jrme Wittwer and two anonymous referees for their
precious comments and suggestions. I also thank the participants of the SHARE-I3 and COMPARE Midterm Meeting (Chania,
Greece), the Netspar seminar (Tilburg University, the Netherlands), the ARC Meeting (University of Lige, Belgium) and the
XXII European Society for Population Economics Conference (University College London, United Kingdom). I also acknowledge
the Communaut Francaise de Belgique for nancial support under ARC contract (ARC 05/10-332).

References

Baker, M., Stabile, M., Deri, C., 2004. What do self-reported, objective, measures of health measure? The Journal of Human Resources 39 (4), 10671093.
Bolin, K., Lindgren, B., Lundborg, P., 2008. Informal and formal care among single-living elderly in Europe. Health Economics 17 (3), 393409.
Brsch-Supan, A., Jrges, H. (Eds.), 2005. The Survey of Health, Ageing and Retirement in EuropeMethodology. MEA, Mannheim.
Bound, J., 1991. Self-reported versus objective measures of health in retirement models. The Journal of Human Resources 26, 106138.
Bound, J., Schoenbaum, M., Stinebrickner, T., Waidmann, T., 1999. The dynamic effects of health on the labor force transitions of older workers. Labour
Economics (6), 179202.
Bound, J., Jaeger, D.A., Baker, R.M., 1995. Problems with instrumental variables estimation when the correlation between the instruments and the endogenous
explanatory variable is weak. Journal of the American Statistical Association 90, 443450.
Carmichael, F., Charles, S., 2003. The opportunity costs of informal care: does gender matter? Journal of Health Economics 22, 781803.
Charles, K.K., Sevak, P., 2005. Can family caregiving substitute for nursing home care? Journal of Health Economics 24, 11741190.
Christianson, J.B., 1988. The evaluation of the national long-term care demonstration: the effect of channelling on informal caregiving. Health Services
Research 23 (1), 99117.
Comas-Herrera, A., Costa-Font, J., Gori, C., di Maio, A., Patxot, C., Pickard, L., Pozzi, A., Rothgang, H., Wittenberg, R., 2003. European study of long-term
care expenditure: investigating the sensitivity of projections of future long-term care expenditure in Germany, Spain, Italy and the United Kingdom to
changes in assumptions about demography, dependency, informal care, formal care and unit costs. Report to the European Commission, Employment
and Social Affairs DG.PSSRU Discussion Paper 1840.
Duan, N., Manning Jr., W.G., Morris, C.N., Newhouse, J.P., 1983. A comparison of alternative models for the demand for medical care. Journal of Business and
Economic Statistics 1 (2), 115126.
Dwyer, D.S., Mitchell, O.S., 1999. Health problems as determinants of retirement: Are self-rated measures endogenous? Journal of Health Economics 18 (2),
173193.
Ettner, S.L., 1994. The effect of the medicaid home care benet on long-term care choices of the elderly. Economic Inquiry 32, 103127.
Grossman, M., 1972. On the concept of health capital and the demand for health. Journal of Political Economy 80, 223255.
Horowitz, A., 1985. Family caregiving to the frail elderly. Annual Review of Gerontology and Geriatrics 6, 194246.
Katz, S.J., Down, T.D., Cash, H.R., et al., 1970. Progress in the development of the index of ADL. Gerontologist 10, 2030.
Langa, K.M., Chernew, M.E., Kabeto, M.U., Katz, S.J., 2001. The explosion in paid home care in the 1990s: who received the additional services? Medical Care
39 (2), 147157.
Lawton, M.P., Brody, E.M., 1969. Assessment of older people: self-maintaining and instrumental activities of daily living. Gerontologist 9, 179186.
McGarry, K., 1999. Inter vivos transfers and intended bequests. Journal of Public Economics 73, 321351.
Motel-Klingebiel, A., Tesch-Roemer, C., v. Kondratowitz, H.-J., 2005. Welfare states do not crowd out the family: evidence for mixed responsibility from
comparative analyses. Ageing and Society 25, 863882.
Newey, W.K., 1987. Efcient estimation of limited dependent variable models with endogenous explanatory variables. Journal of Econometrics 36, 231250.
Norton, E.C., Van Houtven, C.H., 2006. Inter-vivos transfers and exchange. Southern Economic Journal 73 (1), 157172.
OECD, 1999. Classifying educational programmes. Manual for ISCED-97 Implementation in OECD Countries.
Pezzin, L.E., Schone, B., 1999. Intergenerational household formation, female labour supply, and informal caregiving: a bargaining approach. Journal of
Human Resources 34, 475503.
Pezzin, L.E., Kemper, P., Reschovsky, J., 1996. Does publicly provided home care substitute for family care? Experimental evidence with endogenous living
arrangements. The Journal of Human Resources 31 (3), 650676.
Rivers, D., Vuong, Q.H., 1988. Limited information estimators and exogeneity tests for simultaneous probit models. Journal of Econometrics 39 (3), 347366.
Romoren, T.I., 2003. The carer careers of son and daughter primary carers of their very old parents in Norway. Ageing and Society 23, 471485.
Stabile, M., Laporte, A., Coyte, P.C., 2006. Household responses to public home care programs. Journal of Health Economics 25, 674701.
Stern, S., 1995. Estimating family long-term care decisions in the presence of endogenous child characteristics. The Journal of Human Resources 30 (3),
551580.
Terza, J.V., Basu, A., Rathouz, P.J., 2008. Two-stage residual inclusion estimation: addressing endogeneity in health econometric modelling. Journal of Health
Economics 27, 531543.
Van Houtven, C.H., Norton, E.C., 2004. Informal care and health care use of older adults. Journal of Health Economics 23, 11591180.
Van Houtven, C.H., Norton, E.C., 2008. Informal care and Medicare expenditures: testing for heterogeneous treatment effects. Journal of Health Economics
27, 134156.
Viitanen, T.K., 2007. Informal and formal care in Europe. IZA Discussion Paper No. 2648.
Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press.
Yang, Z., Norton, E.C., Stearns, S.C., 2003. Longevity and health care expenditures: the real reasons older people spend more. The Journals of Gerontology.
Series B, Psychological Sciences and Social Sciences 58 (1), 210.