INFORMAL CARE IN THE UK: CONSTRAINTS ON CHOICE Fiona … · 2011-09-30 · 2 INFORMAL CARE IN THE...
Transcript of INFORMAL CARE IN THE UK: CONSTRAINTS ON CHOICE Fiona … · 2011-09-30 · 2 INFORMAL CARE IN THE...
1
INFORMAL CARE IN THE UK: CONSTRAINTS ON CHOICE
Fiona Carmichael*
ABSTRACT
This paper analyses the relationship between informal care provision and the
characteristics and needs of the people they care for. To do this it uses a unique
regional data set containing detailed information on 1,985 informal care-givers. The
data were provided by charitable organization providing support for carers in the
midlands region of the UK. The records provide information on carers‟
characteristics and the characteristics and needs of the people they care for. They also
contain data on caregiving provision including of hours of care, duration of the caring
episode and type of help given (personal, physical, practical or emotional). The
analysis incorporates an initial descriptive analysis of informal caregiving provision
by carers‟ age-group and gender and some contextual national evidence from the
British Household Panel Survey. The descriptive analysis is extended using
multivariate analysis to explore the interrelationships between caregiving provision
and the characteristics of both caregivers and the people they care for. The
relationship between caregiving and employment is also investigated using
instrumental-variables estimators to control for endogeneity. The results indicate that
the exogenous needs and characteristics of the cared-for play an important role in
shaping informal care provision and, indirectly, carers‟ employment. This implies that
individual decisions about caring provision are not simply a matter of choice.
Decisions are constrained, particularly after a caring episode has begun.
KEYWORDS: Caregiving, health, labor supply, unpaid work
JEL Codes: D13, J14, J17
ACKNOWLEDGMENTS
This research was supported by a grant from the British Academy. I am grateful to the
regional carers‟ support service for supplying the data and to Emma Silvester for
creating the anonymised data set. British Household Panel Survey was made available
through the Office of National Statistics and the ESRC Data Archive
*Birmingham Business School, University of Birmingham, Edgbaston,
Birmingham, B15 2TT, UK
e-mail: [email protected]
2
INFORMAL CARE IN THE UK: CONSTRAINTS ON CHOICE
1. Introduction
Informal carers look after someone who needs support because of frailty in older age,
physical or mental health needs or illness (UK Department of Health 2006(a)).
Parents who look after children with disabilities, children caring for elderly parents
and young people providing care for siblings all fall into this category of caregiving.
Informal carers are essentially a free source of labour from the perspective of health
and social services. The costs they incur are often ignored in policy decisions, in there
is a general lack of recognition of their contribution to the health and wellbeing of
others and the difficulties they face. This is in some ways surprising since community
health and social care systems rely on family carers providing support for people who
need care but want to remain in their homes (Michael Nolan 2001). Furthermore, the
numbers involved in care are substantial. According to official estimates,
approximately one in ten of the population of England and Wales are involved in
informal care (National Statistics Online 2003). Yeandle et al. (2006) have also
additionally estimated that nearly three million workers in the UK provide informal
care. The percentage of the population who will, at some point in time, participate in
informal care is higher; among respondents involved in the first wave of the British
Household Panel Survey (BHPS) in 1991, over half (51.8 percent) of those still of
working age 10 years later had participated in informal care. These relatively high
figures for caring incidence are in line with evidence from previous research in the
UK (Heitmeuller and Inglis, 2007; Hirst, 2002) and from Australia where 3-4 percent
of employees become carers each year (Hill et al. 2008).
Nevertheless, demographic changes have prompted a growing interest in
issues related to informal care provision. This is largely because population ageing
3
imposes extra demands on the health and caring services and raises the demand for
care (Hancock et al. 2003); according to some estimates, spending on personal care
alone will need to triple to £30 billion by 2026 in order to meet the needs of the aging
baby boomer generation (John Carvel 2006). At the same time, population ageing is
putting increasing pressure on pensions which has created an urgent imperative to
extend working lives. Consequently, the demand for informal care is rising at the
same time as the people‟s availability to supply care is being constrained.
In the light of these trends, it is not surprising successive UK governments
have initiated a series of policy measures that have aimed to address the needs of
carers (see Table A1 in the Appendix). These initiatives have focused mainly on
support for carers through formal services, financial support and support for working
carers. The latter reflecting government recognition that caring responsibilities can
constrain employment. For a discussion of some of the complexities underlying these
policies see for example, Twigg (2009). The impact of these policy initiatives is
difficult to judge, however, the available evidence suggests that the difficulties faced
by carers have not changed significantly (Department of Health, 2001; Challis et al,
2005; Carmichael and Hulme, 2008). Most recently, the government elected in 2010
stated an early intention to „refresh‟ the National Carers Strategy and produce a plan
of action for 2011-15. Paul Barstow, the minister for Care Services, issued a call for
views on this process (Department of Health, 2010) and a response to the views
submitted was published in November (Carers Policy Team, 2010). In the
„refreshment‟ process there is a focus on „effective early intervention‟ and
„personalisation‟. To improve the early identification of carers the Reaching out to
Carers Innovation Fund has awarded £1.35m to 79 projects run by voluntary groups
4
who are „keen to support carers‟ (Department of Health, 2011). At the time of
writing, no evidence is available on the success of these projects.
Alongside these policy initiatives, research on informal caregiving provision
is also growing. This research has tended to focus on how care provision impacts on
carers‟ health, wellbeing, income employment. The latter in particular has been the
subject of considerable quantitative research (e.g. Ettner 1996; Pavalko and Artis
1997; Spiess and Schneider 2003; Heitmueller 2007). There has also been some
consideration of the related effects of employment on care (Carmichael et al., 2010;
Michaud et al. 2011). Research is also being conducted on the effect on the trade-off
between employment and caregiving of employment and workplace policies (Hill et
al., 2008; Bryan, 2011).
The underlying complexity of these relationships has also been highlighted by
in qualitative studies (Baldwin, 1985; Arksey et al. 2005; Vickerstaff et al., 2009).
For example, Arksey et al. (2005) emphasise that care-giving relationships are
characterized by uncertainty and that carers often need to respond to changes that are
beyond their control. For example, care needs can change after „critical transition
points‟ such as a medical intervention (Arksey et al. 2005:53). Relatedly, Vickerstaff
et al. (2009:27) notes that the demands of caring tend to increase over time; beginning
with relatively low level types of care such as helping with the laundry and paperwork
but ending up with more intensive care including personal care. These studies
highlight that decisions around caring and employment are constrained by myriad
factors. In particular, and once a caring episode has begun, the needs of the cared-for
and the consequent type of care provided are of particular importance. In contrast,
most quantitative research has little to say about such factors probably because of lack
of data. One exception is the study by Hassink and Van den Berg (2011) who use
5
diary data to consider allocations of time to different caring tasks. Their results
indicate that the time constraints imposed by informal care vary with the type of
caring activity undertaken. Since the latter depend on the needs of the cared-person
these are an implicated in decisions about the allocation of time.
This paper contributes to this literature by analyzing the relationship between
the amount and type of care provided and the needs and characteristics of the cared-
for. The results of this analysis are then used to investigate the relationship between
hours of care and employment. To do this the study utilizes a unique regional data set
compiled by a charitable organization in the midlands region of the UK. The
organisation is engaged in the delivery of support and advice to carers (hereafter
CSAM). The dataset contains a record for all 1,985carers who accessed the services
of CSAM between 1998 and 2008. The records provide information on caregiving
provision including indicators of hours of care supplied and duration of the caring
episode. More importantly for this study, they also provide data on the type of help
given (personal, physical, practical or emotional) and the characteristics and needs of
the people being cared for. This enables a quantitative exploration of the constraints
imposed on carers‟ time by these largely exogenous factors. This is important since
previous research, particularly in economics, has tended to focus on the „choice‟
aspect of decisions about caring, particularly in relation to employment.
The plan of the paper is as follows. First, we draw on the BHPS in order to
provide some national context on caring incidence, the characteristics of carers and
the provision of informal care. We then analyse the CSAM data base. A preliminary
descriptive analysis is reported for the whole sample by age-group and gender. We
then explore the determination of hours of care and the provision of personal care
within a multivariate framework. The analysis is extended by estimating a model of
6
employment participation and hours of care using the method of instrumental
variables to control for potential endogeneity. The final section interprets the main
findings and discusses policy implications.
2. Contextual evidence from national data
In this section data from seventeen waves of the BHPS is used to provide some
contextual evidence on informal care-giving in the UK. The BHPS is an annual
survey consisting of a nationally representative sample initially composed of about
5,500 households recruited in 1991. Extension samples have since been added and the
BHPS currently covers around 10,000 households across the UK. The BHPS was
recently replaced by and incorporated into the Understanding Society Study which
follows 40,000 household in the UK. The data reported in Table 1 is derived by
pooling the first seventeen waves of the BHPS (1991-2007). The sample contains
36,536 observations for people involved in informal care and 163,825 observations
for people who were not caring in the year the data was collected.
Table 1 reports means and sample percentages by gender and caring status.
The data show that carers are significantly more likely to be female and their average
age is significantly higher than that of non-carers. Female carers are more likely to be
25 or over than female non-carers and male carers are statistically more likely to be
65 or over. Data from the Office of National Statistics (National Statistics Online,
2006(a)) indicate that the peak age for becoming an informal carer is between 45 and
64 years in the UK although just over one per cent of carers are children (aged
between 5 and 15) and five per cent of carers are aged 85 and over. Interestingly, the
7
predominance of females is only a feature among carers under 75 years old, among
people 75 and over a larger proportion of men provide care. The BHPS data indicate
that 50.19 percent of informal carers 75 and over are male (see also National Statistics
Online, 2006(a)). This is because spousal care is the most frequent type of care given
among older people and more older men have living spouses than older women.
Overall, however, women provide the bulk of informal care in the UK, as they do in
other European countries, the US, and in developing countries, such as China (Eaton
2005).
The figures in Table 1 show that both male and female carers are less likely to
be employed or be in training/education than than non-carers. They are more likely to
be either unemployed, retired, on maternity leave, involved in family care or long-
term sick/disabled. Overall, these figures would be consistent with the hypotheses
that informal care responsibilities impact negatively on employment. However the
figures in Table 1 also show that compared with non-carers, carers are less likely to
have a higher educational qualification and more likely to report that they have a
health problem. Both factors could be expected to contribute to the lower employment
participation of carers as could age and gender although caring has also been
implicated in ill-health. In addition, there is a potential source of endogeneity in the
caring-employment relationship since the opportunity costs of caring are likely to be
higher for people who have a higher propensity to undertake paid work e.g. because
they have higher unobserved ability and can earn higher wages. However, previous
research in the UK and elsewhere has shown that some of the differences in
employment participation remain when carers‟ characteristics are controlled for and
endogeneity is addressed (e.g. Lilly et al. 2010; Carmichael et al, 2010; Heitmueller
and Inglis 2007; Lázaro, Moltó, and Sanchez 2004; Carmichael and Charles 1998,
8
2003a, 2003b; Pickard et al. 2000; Hutton 1999; and see Lilly et al 2007 for a review
of this literature).
The figures in Table 1 also provide some information on the hours of caring
undertaken. The figure show that the majority of people categorised as carers in the
UK care for only a few hours a week; median hours of caring are between 5 and 9 and
modal weekly hours are less than 5. The majority of carers care for fewer than 20
hours a week with only 20.60 percent of male carers and 23.62 percent of female
carers caring for more than 20 hours weekly. Nevertheless, just under 10 percent of
male carers and just under 11 percent of female carers are intensive carers who are
caring for more than 50 hours a week (Vickerstaff et al., 2009). Between 8 and 9
percent are caring for very long hours of 100 or more a week. The only other direct
information on caregiving within the BHPS relates to residency. The figures show
that 40.36 of male carers and 32.82 percent of female carers are co-resident with the
people they care for.
The figures show that carers who perform longer hours of care are more likely
to be female; 62.92 percent of those who care for twenty or more hours per week are
female. This finding is consistent with previous research in the UK and research
undertaken in other countries (e.g. Lilly et al, 2010; Arksey et al, 2005; Lázaro,
Moltó, and Sanchez 2004; Maher and Green 2002). Hours of care are also related to
age with a higher percentage of older carers providing 50 or more hours of care a
week. The BHPS data indicate that among carers 70 and over the percentage caring
for at least 50 hours a week is 15.72 percent compared with only 10.10 percent among
those between 25 and 69 (see also National Statistics Online, 2006(a)).
To summarise, the national data from the BHPS shows that compared with
non-carers, carers in the UK are older, more likely to be female, less likely to be
9
employed and more likely to suffer from some health problem, they have also
invested in less human capital. The majority of carers are caring for fewer than 20
hours a week with only under 23 percent caring for at least 20 hours a week and only
around 10 percent are intensive carers who care for 50 or more hours a week. A
minority of carers are co-resident with the people they care for.
3. The regional data set
The charity providing the regional data, CSAM, was founded over 25 years ago and is
an established provider of support to carers at the sub-county and county level. During
this time it has supported a broad range of carers including parent carers, carers of
elderly people, carers of adults with physical disabilities or learning disabilities and
young carers. CSAM‟s overall aims are to enrich carers‟ lives and promote their
wellbeing. To meet these aims CSAM employs specialist staff to address the needs of
carers and offers a wide range of services including one-to-one advice and support
and the provision of information and advice (via a newsletter and a website with links
to other agencies). It also facilitates support groups that help carers to meet each other
and by doing so feel less isolated and give mutual support. In addition, the
organisation promotes carers' rights and aims to raise the profile of carers and carer
issues by ensuring that carers have a voice that is heard by decision makers. CSAM
has recently been restructured and at the time of writing services to young carers and
adult carers are provided by separate organisations.
Since 1998 CSAM has held a database of all the carers who contacted the
organisation. The organisation kindly allowed access to this database after the entries
had been anonymised. The data covers 1998-2000 during which time 1985 carers
contacted CSAM; just under 200 carers a year. Table 2 summarises the
10
characteristics of the carers who contacted CSAM over this 10 year period. Table 3
summarises the information that is available on the people they cared for and Table 4
provides some detail on the type of care provided. In each case the data is broken
down by both age-group and gender since previous research and the BHPS data
indicate that older people and women are the most intensive carers.
Characteristics of carers
In line with the national data, the majority of the carers in the sample are
female. However, the percentage of females in the regional sample (70 percent) is
higher than the national average for carers of just under 60 percent (Table 1).
Compared with the national average for carers , the CSAM carers are also older; their
average age is 59 and 46.53 percent are over 65. A somewhat larger percentage (62.8
percent) are over the statutory state pension age (SPA) since this only 60 for females
and 59.5 percent of the sample are over 60.
There is also a large group of mid-life carers between 25 and 64 years old
(41.36 percent) and a smaller group of relatively young carers; 10.7 percent are under
18 and 12.1 are under 25. The ethnicity of the sample members is predominantly
white (91.7 percent) although just fewer than 8 percent of the sample are categorised
as either Asian or Black. Although we do not have data on income, we know that 767
sample members (or the person they care for) are in receipt of some state benefit
(38.64 percent) and 173 carers (8.72 percent) are in receipt of Carer‟s Allowance. 46
others receive income support and 31 receive pension credits. 449 care recipients
(23.1 percent) are in receipt of either attendance allowance or disability living
allowance (Table 2)
11
Data on employment status is available for 954 sample members. 21.28
percent of this sub-sample are in employment of some kind and just under 17 percent
are involved in training/education. Among those aged 25-64, 41.12 percent are
employed; 39.35 percent of females and 49.32 percent of males. Among females
between 25 and 59 (the SPA for this cohort) a somewhat higher percentage, 44.7
percent are in employment. Among the sub-sample aged 65 and over, 5.02 percent are
in employment (4.17 percent of women and 6.5 percent of men). None of the carers
under 25 are employed, the vast majority are involved in training/education. Overall,
females are more likely to be employed than males but this result is driven by the
higher percentage of women who are mid-life carers.
Overall, the employment participation rates among those under 65 are very
low relative to the national averages reported in Table 1. As discussed in more detail
below, part of the explanation is likely to lie in the relatively long hours of care the
CSAM sample supplies. The relatively high training/education figures among the
younger sub-sample bolster their labour market participation rates but may mask an
inability to secure employment. Some carers may also be undertaking training in
order to fulfill benefit requirements. However, there is no detail on the kind of
training being undertaken and therefore this remains conjecture.
Characteristics of the people cared-for
Table 3 provides summary data on the characteristics of 1294 of the people
cared for. However, there are fewer observations for some measures as indicated in
the Table. The data show that the gender split among the cared-for is very equal with
only fractionally more males among those cared for. Older carers and women are
more likely to be caring for men (presumably their spouses).
12
The average age of those cared for is just under 61 which is only two years
older than the mean age for the carers. The small age gap is consistent with spousal
care being a large part of the care provided. Most of the 84.17 percent of those 65 and
over who are also caring for someone in the age group are also likely to be providing
spousal care. However, nearly half (49.71 percent) of midlife carers are providing
care for someone over 65 and this figure provides a very rough upper limit to the
amount of elder care being undertaken. Mid-life males in particular are more likely to
be caring for someone 65 or over than someone younger than 65. A more considered
proxy for the amount of elder care being provided is discussed below.
25.14 percent of the sample care for someone whose needs were categorised in
in terms of learning needs or mental health problems. This group includes older
person with mental health problems (n=114). 13.21 percent of carers look after
someone with physical disabilities and 15.2 percent are caring for disabled children.
In line with the age of the cared-for, the health problems of the majority (42.64
percent) were classified simply in terms of their status as an older person (65 or over).
Not surprisingly given their age, the majority of the sample (61 percent) care for
someone who is retired, however 18.9 percent care for someone who is in work and
16.2 percent look after someone who is a student.
A little less than 20 percent of the sample care for someone who receives
Attendance Allowance although this figure rises to 30 percent among older carers.
Fewer than 5 percent look after someone who receives Disability benefit. The
majority of those cared for live in their own home or a home they share jointly with
their carer. Interestingly, male carers are more likely to be looking after someone who
lives in their own (the cared-for‟s) home or a joint home and females are more likely
to care for someone who lives in their (the carer‟s) home. This is consistent with the
13
male caring role predominantly being spousal care or care for relatively independent
elders. Women, on the other hand, are more likely to be caring for disabled children.
From this data we can also tentatively conjecture that they women are more likely to
be caring for less independent elders who have moved into the carer‟s home.
Type and amount of care provided
The low employment rates of caregivers reported in Table 2 are not altogether
surprising given that the carers in the sample have taken the step of contacting a
support service for help or advice. As shown in Table 4, many are caring for long
hours and some have cared for many years. Among the 869 carers for whom we have
this information the mean number of care hours per week is between 35 and 49, the
modal category is at least 100 hours and the median category is between 50 and 99.
These figures indicate that the carers in the sample are representative of the minority
of carers who provide intensive care (50 or more hours a week). Furthermore, 72
percent of the sample are co-resident (in the caregiver‟s home or a joint home) with
the person they care for and previous research suggests that these carers have more
significant caring responsibilities and care for longer hours (Vickerstaff et al.. 2009;
Heitmueller, 2007).
Within the sample, carers 65 and over care for the longest hours and have been
caring for the longest duration; among the 25-64 age group the percentage caring for
longer hours (more than 50 and more than 100 a week) is significantly lower
compared with the older age group. Nevertheless a large majority of mid-life carers
are caring for 50 or more hours a week. Among the sub-sample of carers younger
than 25 the vast majority are involved in fewer than 20 hours of care a week although
a small minority are caring for between 20 and 49 hours a week.
14
The mean caring episode duration for the bulk of the sample (1926 sample
members) is between 2 and 3 years, the modal duration is 2 years and the median
duration is 3 years. However, just under 25 percent of the sample have been caring for
more than 5 years and 79 had cared for 9 or more years. The mean caring duration is
unsurprisingly higher for older carers. It is lower for women probably reflecting the
lower mean age of the people they care for. The average caring duration for younger
carers is only 2.32 years but this still indicates a considerable ongoing commitment
that could potentially impact on their future employment opportunities.
While intensive carers are usually defined as those who supply 50 or more
hours of care a week, the intensity of care in terms of how demanding it is for the
carer could be better indicated by other, more qualitative measures. For example, in
terms of the type of care provided. This would reflect the needs of the cared-for
person to some extent and would vary according to the ability of the cared-for person
to conduct core daily activities. The greater the difficulty of the person cared for to
perform these activities, the greater the need for assistance and the more demanding
or intense the associated carer responsibility (Hill et al., 2008). The provision of
personal care, for instance, implies that the care recipient faces profound limitations
on their capacity to perform basic personal care activities such as getting ready in the
morning, dressing, washing and feeding. In contrast, the provision of practical care
implies a need for help with activities such as household work, organizational
activities and transport. Other types of care could involve giving emotional support or
physical care such as physical therapy and help with mobility.
Vickerstaff et al. (2009) note that most of the carers in their sample who cared
for 50 or more hours a week also provided personal care while those caring for fewer
hours did not. This suggests that the provision of long hours of care and intimate
15
personal care are complements. This may reflect the long-term nature of many caring
relationships which, as observed by Vickerstaff et al. (2009:27) gradually “become
more intensive, time consuming and intimate.” Recent research additionally suggests
that personal care may also be the most difficult type of care to fit around other
activities such as paid work since related tasks often need to be completed at
particular times of the day (Hassink and Van den Berg, 2011).
The figures in Table 4 show that over 40 percent of the sub-sample for whom
we have the relevant information (n=988) provide personal care (the vast majority
with physical care and only 1.72 percent on its own). Women are significantly more
likely to provide personal care but the gender difference is only significant among the
older cohort who are anyway more likely to providing this type of care. Nearly twenty
percent provide only practical care (including drug administration, n= 2) and just over
17 percent provide emotional or „other‟ care. 17.11 percent of this sub-sample of 988
are categorised as young carers providing either excess chores (the majority) or
sibling care.
It would be useful to have a measure of intergenerational caring provision, in
particular, the degree to which the sample members are providing elder, spousal, or
child care. Unfortunately, there is no explicit data on the familial relationships
between carers and cared-for. However, we do know that 15.2 percent of the cared-for
are disabled children of a sample member (Table 3). 12.10 percent of the sample are
also under 25 and these relatively young carers are likely to providing care for a
sibling or a parental. We also know that a majority of the cared-for people are 65 and
over. As noted above, this doesn't necessarily mean that the majority of care
undertaken by the sample is elder care since many of the carers are older people
themselves and are probably giving spousal care. Some of the mid-life carers who are
16
caring for people 65 and over are also providing spousal care. However, 26 percent
of the sample for whom we have data on both the carer‟s and the cared-for person‟s
age (n= 1103) are caring for someone who is both at least 60 years old and at least 17
years old than the carer (Table 4). This figure gives an approximate indication of the
extent of elder care within the sample. The majority of carers who are 65 or over are
caring for older people who are less than 17 years older than themselves (n= 411).
The group is likely to be undertaking spousal care. Only two sample members who
are 65 or over are caring for people who are 20 or younger, these carers are probably
caring for their grandchildren. Taken together these estimates suggest that
approximately 47% of the care provided is spousal or „other‟ care.
4. Multivariate analysis
4.1 Empirical specification
Caring provision
In this part of the paper we explore the determinants of informal caring provision. We
consider how and to what extent caring provision is influenced by the characteristics
and status of the caregiver (including their employment status and whether they are in
receipt of Carer‟s Allowance) the characteristics of the cared-for person and the type
of care given (e.g. personal, physical or practical). Initially the focus is on caring
intensity measured by time involved in care, T and the estimated relationship is given
by:
T = β0 + β1X1 + β2 X2 + β2 X3 + ε (1)
where T is hours of care provided, X1 is a vector of covariates reflecting the
characteristics and status of the caregiver, X2 is a vector of covariates reflecting the
characteristics of the cared-for person, X3 is a vector of covariates indicating the type
17
of care given and the duration of the caring episode, ε is the error term. This
specification tests whether and how the largely exogenous characteristics of the
cared-for person impact on care provision. In addition, the relationship between type
of care provided and hours of caring is explored.
As discussed, provision of personal care is an alternative indicator of the
intensity of caring. We therefore estimate a version of (1) with a dependent variable
indicating whether or not the caregiver provides personal care (P). In this estimation
the vector X3 includes a more restricted set of covariates indicating type and
duration of care.
Four dichotomous indicators of the time involved in caregiving are used in the
estimations, three are dichotomous and one is an ordered variable. HOURS>50pw
(No=0, Yes=1) records whether an individual cares for at least 50 hours a week;
HOURS>100pw (=0,1) records whether an individual cares for at least 100 hours a
week; CAREHOURScat (=0, 1,2,3,4,5,6) is an ordered variable and records weekly
hours of caring by category (less than 5, 5-9, 10-19, 20-34, 35-49, 50-99, 100 or
more); HOURS<20pw (=0,1) is an alternative, negative measure of time committed
to care provision. It records whether an individual cares for fewer than 20 hours a
week. The indicator of personal care provision (PERSONAL CARE) is also a
dichotomous variable.
For the dichotomous variables, we model the probability of undertaking at
least T hours of care (T =50, 100) less than T hours (T = 20) or providing personal
care (P) using logit and for ease of interpretation report odds ratios. In the case of the
measures of hours of care provided, logit models the mean probability of undertaking
at least/less than T hours of care ( 0,1iT ) by maximising:
18
, ,
1
ln ln ( ) (1 ) ln(1 ( ))N
i j j i i j j i
i
L T F x T F x (1)
where 1
, ,1( ) (1 exp( ))
k
j j i j j ijF x x is the logistic function. The xjs are the
individual carer and cared-for characteristics and indicators of type of care that are
available in the dataset (Tables 2-4). The βjs are the corresponding coefficients (log-
odds ratios). An equivalent specification corresponds to the logit estimation for the
supply of personal care. For the regressions in which the dependent variable is the
ordered variable CAREHOURScat we use ordered logit. The ordered logit model
generalises the approach of the binary-choice logit model to the notion of multiple
thresholds where larger values are taken to correspond to higher outcomes.
For each of these estimations we adopt a stepped approach that involves
estimating 4 models. The stepped approach enables identification of significant
influences that would otherwise be obscured because of collinearity with other
variables. Model 1 regresses the measure of caring on only those variables reflecting
the characteristics or circumstances of individual carers: gender (MALE); age (AGE);
whether the carer is employed (EMPLOYED); whether the carer receives Carer‟s
Allowance (CARERS ALLOWANCE); ethnicity (ETHNIC BLACK/ASIAN,
ETHNIC OTHER, the reference group is white ethnic).
Model 2 regresses the measure of caring on only those variables that reflect
the characteristics and needs of the cared-for (CF) person. These include gender
(CF_MALE) and age (CF_AGE). Two variables indicating whether the cared-for
person receives Attendance Allowance or Disability Living Allowance are also
included. Model 2 includes the indicators of the needs of the cared for person
classified in terms of their health condition as either a disabled child, an adult with a
physical disability, and adult with mental health or learning needs (includes older
19
adults) or as an elderly person. Other included variables indicate the living
arrangements of the cared-for and whether they are employed (CF_EMPLOYED) or
retired (CF_RETIRED). For cared-for people of working age, whether they are
working or not provides some indication of their degree of independence. As such,
CF_EMPLOYED is expected to be negatively related to hours of care provided
although it may also be correlated negatively with CF_AGE.
Model 3 regresses the dependent variable only on indicators of the duration of
the caring episode and the type of care given and. The measure of the duration of the
caring episode is in years (DURATION). The indicators of type of care cover co-
residential care (CO-RESIDENTIAL) and a proxy measure for elder care
(ELDERCARE). The latter indicates whether care is given to an elderly person over
60 and at least 17 years older than the care-giver. Also included are an indicator of
personal care provision (PERSONAL CARE) and a variable indicating that the only
care given is emotional or „other‟ care (EMOTIONAL/OTHER CARE). The
reference category for these variables is the provision of only physical or only
practical care. These two measures are excluded when the dependent variable is
PERSONAL CARE.
Model 4 is the last step in the procedure and estimates a best-fit equation that
regresses the dependent variable on only those independent variables that are
significant in either Model 1, 2 or 3. Table A2 in the Appendix provides definitions of
all the variables used in the analysis.
Caring provision and employment
We additionally explore the relationship between caregiving and employment
participation for a restricted sample of carers aged 19-69 (among those older than 65
20
two sample members younger than 70 are working but only five carers aged 70-80 are
employed). The oldest employed sample member is 80). In the analysis we take steps
to control for possible endogeneity in the relationship between caring and
employment. As already referred to, the issue of endogeneity arises because time
spent caring incurs an opportunity cost in terms of the monetary value of forgone time
(see for example, Heitmueller, 2007; Carmichael et al., 2010). Decisions about caring
and employment are likely to be jointly determined because the shadow price of time
involved in care will be higher for some people (e.g. those employed in full-time jobs,
those working inflexible hours and those earning a relatively high income). Even
after a caring episode has begun, a variety of “uncertainties and unknown factors
and/or external constraints” including the carer‟s own health and job insecurity are
likely to influence decisions about both paid work and care (Arksey et al., 2005: 150).
More specifically, endogeneity arises because of omitted variables in the
estimated regression of employment participation, E (or other measure employment
status) against informal care commitment, C, given by:
E = β0 + β1C + β2X + ε (2)
where X is a vector of covariates reflecting individual characteristics and ε is the error
term. In this estimation there are unobserved covariates such as unobserved ability
that potentially impact on both employment participation and caring commitment.
The omission of these covariates violates the zero conditional mean assumption
E(ε|X) = 0 since changes in such variables will alter both C and E. This implies that
OLS estimates of (2) will be inconsistent. For example, inasmuch as people with
higher unobserved ability have a higher propensity for employment
(βUnobservedAbility > 0) and a lower propensity for caregiving then (Cov(ε, C) < 0)
OLS estimates of β1 will be biased downward.
21
Researchers have adopted a variety of approaches in an attempt to control for
endogeneity and selection into caring. For instance, they have used panel data in
conjunction with fixed or random effects models to allow for the effect of individual
characteristics on labor supply (Bolin et al, 2008; Heitmueller, 2007). Another
approach has been to control for individual characteristics by using longitudinal data
to estimate a relationship between changes in informal care provision and
simultaneous changes in employment status (Carmichael et al. 2010; Spiess and A.
Schneider 2003; Pavalko and Artis 1997).
The most common method used to address endogeneity in the caring-
employment relationship is to specify an instrument for caring commitment, C
(Ettner, 1996). This method is adopted here. In order to address the identification
problem, the selected instrument for caring commitment, z, needs to satisfy two
criteria. First, it needs to be strongly correlated with the measure of care. Second it
needs to be exogenous so that is uncorrelated with the error term in the estimated
employment relationship (2). To satisfy the first criteria, we construct instruments
from those variables that are significant in the estimated regressions in which the
relevant measure of caring commitment is the dependent variable. We then use 2SLS
to combine these multiple instruments into one optimal instrument and generate
consistent estimation of β1 in (2).1
We cannot observe ε, and therefore we cannot directly test for the second
assumption of zero correlation between z and ε (Wooldridge, 2002). However, to
address this point the proposed instrument is constructed only from variables
reflecting the characteristics of the cared-for and their care-needs. It is unlikely that
these measures are correlated with unobserved determinants such as the carer‟s
1 Conceptually, the procedure involves running an auxiliary regression with the dependent variable C
and the multiple instruments as independent variables. The optimal instrument is generated by the
predicted values of this estimation although the „two-stage‟ estimator is calculated in one computation.
22
unobserved ability. The latter are likely to be determined in the carer‟s formative
years not by the onset or nature of the cared-for person‟s illness or disability.
4.2 Results
Hours of care supplied: Tables 5-8
In table 5 the dependent variable is HOURS>50. The figures reported are odds-ratios.
MALE is insignificant in Model 1 suggesting that among this sample of carers,
females are no more or less likely to supply longer hours of care. While employment
status (EMPLOYED) is weakly and negatively significant in Model 1 it is not
significant in Model 4, the best fit estimation, which includes variables reflecting
cared-for characteristics and indicators of type/duration of care. However, carer‟s age
retains significance in Model 4 confirming that longer hours of care are provided by
older carers even after allowing for other factors. The reported odds ratio in Model 4
suggests that the odds of providing at least 50 hours of care a week increase by 12
percent with each year of a carer‟s life. Receipt of Carer‟s Allowance also retains
positive significance in Model 4 which is not surprising given the eligibility
conditions for receipt of this allowance (see Appendix); the odds of supplying at least
50 hours of care a week are higher by a factor of 7.89 for those in receipt of Carer‟s
Allowance.
The cared-for characteristics that retain significance in Model 4 are age
(CF_AGE) and receipt of either Attendance Allowance or Disability Living
Allowance and living arrangments. Age of the cared-for person is negatively related
to hours of care supplied in contrast to carer‟s age. The positive significance of either
state benefit is in line with the stringent needs-based eligibility conditions for receipt
of these benefits. As such receipt of these benefits is likely to identify those cared-for
23
people with the greatest care needs. It is therefore not altogether surprising that after
allowing for receipt of benefits, none of the variables indicating the specific health
needs of the cared-for are significant. Living arrangements appear to be important,
particularly whether the cared-for person lives in institutionalized accommodation
which retains negative significance in Model 4 (relative to the reference category of
living in the cared-for‟s own/joint home).
In Model 3, the provision of co-residential care, elder care, personal care and
emotional or other care are all positively related to hours of caring, but only co-
residential care and personal care retain significance in Model 4, both positively. The
odds ratio associated with personal care in Model 4 indicates that supplying this type
of care raises the odds of caring for at least 50 hours a week by a factor of 4.47 (or
347 percent). Providing co-residential care raises the odds by a somewhat smaller
factor of 2.43 or 143 percent.
The results in Tables 6-8 are largely in line with those in Table 5. However,
there are some exceptions suggesting that there are threshold effects related to hours
of care. For instance, in Table 6 where the weekly hours threshold is 100 hours
(HOURS>100pw= 1) the employment status variable (EMPLOYED) is negatively
significant in both Models 1 and 4. The odds ratio in Model 4 indicates that the odds
of caring for 100 or more hours a week are lower by a factor of 0.39 (61 percent) if
the carer is employed. In contrast, employment status is insignificant in Table 5 where
the weekly hours threshold is only 50 hours. This suggests that only the most
intensive carers trade-off care hours and hours of work. Another difference is that the
age of the cared-for person is not significantly related to the 100 hours threshold while
the proxy for eldercare is negatively significant. Since carer age is positively
significant in both sets of estimations, this difference can be interpreted as suggesting
24
that the most intensive carers are older and caring for someone who also elderly.
Presumably this is mostly spousal as opposed to eldercare. Personal care is positively
and significantly related to both the 50 and 100 hours thresholds. However the
positive significance of providing emotional/other care in Table 6 is difficult to
interpret.
The 50 and 100 hour thresholds also appear to be related to living
arrangements in different ways; when the cared-for person lives in the carer‟s home
the odds of caring for at least 100 hours a week are higher by a factor of 2.72 (172
percent). Interestingly, the Model 2 estimates in Table 6 indicate that carers looking
after an adult with a physical disability are significantly more likely to provide 100 or
more hours of care a week. However, this indicator of specific care need does not
retain significance in Model 4 when carer characteristics and type/duration of care
variables are included. As indicated in Table 9, this could be because the provision of
personal care is strongly related to care need.
Another notable difference is that when the dependent variable marks the 100
hours threshold, duration of the caring episode (DURATION) is negatively
significant; the odds of caring for 100 or more hours a week are lower by a factor of
0.58 (42 percent) for each extra year involved in caregiving. This result may be
interpreted as suggesting attrition or possibly a tradeoff between time spent caring
and the duration of the caring episode.
In Table 7 the dependent variable is the ordered variable CAREHOURScat
and the estimation procedure is ordered logit. The results are largely consistent with
those in Tables 5-6 but there are some differences. For instance, the variables
indicating that the cared-for person lives either in institutionalised accommodation
and the carer‟s home retain significance in Model 4, the former negatively, the latter
25
positively. The significance of these variables in Table 7 (but not Tables 5-6) suggests
that the living arrangements of the cared-for person impact on hours of caring in a
graduated way that is captured by the ordered measure of care hours but not the
dichotomous measures; the results imply that the odds of providing longer hours of
care are 3.95 times higher if the cared-for person is living in the carer‟s home and
0.30 times smaller if they live in insitutionalised accommodation. As in Table 5 (but
not Table 6) employment participation is insignificant. As in Table 6 but not Table 5,
the cared-for person‟s age and co-residential care are insignificant while
DURATION and the proxy for eldercare are both negatively significant in Model 4,
although the latter only weakly so.
In Table 8 where the dependent variable, HOURS<20pw, is a negative
measure of caring commitment, the estimates provide a consistency test of the
alternative specifications in Tables 5-7. In these estimations the variables indicating
receipt of either Attendance Allowance of Disability Living Allowance cannot be
included. This is because no carers with a dependent in receipt of these benefits cared
for fewer than 20 hours per week. As would be expected the results are otherwise
almost a mirror image of those in Tables 5 and 6. For example, younger carers and
those looking after someone living in institutionalized accommodation are more likely
to be caring for fewer than 20 hours a week. An exception is the negative significance
of EMPLOYED however this result is probably being driven by the very low
employment participation rates of younger carers who also care for fewer hours. In
Model 3, the indicators of type or amount of care provided are all negatively
significant with the exception of CO-RESIDENTIAL. However, only the provision of
emotional or other care retains significance in Model 4 and positively so: the odds of
26
supplying fewer than 20 hours a week are higher by a factor of 10.58 if the carer
provides only emotional or other care.
Provision of personal care: Table 9
In Table 9 the dependent variable indicates whether the carers provides personal care
(PERSONAL_CARE). As discussed the provision of personal care is an alternative
measure of the intensity of caregiving since it implies that the cared-for person has
profound limitations. The results indicate that gender (MALE) is an important
determinant of the provision of personal care; the odds of supplying personal care are
lower for males by a factor of 0.56 (44 percent). This is an interesting result as gender
is insignificantly related to hours of care (Tables 5-8) while PERSONAL_CARE is
positively significant. Thus gender is implicated indirectly in the time devoted to
caregiving. Equally interesting is the lack of significance of the receipt of either
Attendance Allowance or Disability Living Allowance in Table 9. In contrast, carers
of an adult with a physical disability are more likely to supply personal care. As in
Tables 5-8, living arrangements are important and co-residential carers (particularly
those sharing their home with the cared-for person) are more likely to provide
personal care. One possible interpretation is that the least independent people live in
their carer‟s home, and interesting their carers are more likely to be women.
Employment participation and hours caring: Table 10
The results of estimating the employment participation equation (2) are shown in
Table 10. In these estimations the dependent variable, EMPLOYED (N0 =0,Yes=1)
indicates whether the sample member is in paid work . Since the dependent variable is
dichotomous, logit would be an appropriate estimator. However, we additionally wish
27
to address endogeneity using the instrumental variables specification and in Stata
this is only available with probit or tobit. We therefore estimate equation (2) using
probit.
In these estimations EMPLOYED is regressed on either HOURS>100pw
(Model 6) or CAREHOURScat (Model 7). The other independent variables reflect
carer characteristics. The selection of the latter is restricted by availability to age,
gender and ethnicity. To capture diminishing returns to experience, proxied by age,
the square of age (AGE_squared) is also included in these estimations. The only other
available carer characteristic variable records receipt of Carer‟s Allowance. This
variable is excluded from this analysis because the earnings eligibility condition for
receipt of this benefit imply that it is not exogenously determined. Because of the
very low employment participation rates of younger and very elderly sample
members, the sample for these estimations is restricted to those aged 19 to 69.
The first estimation in Table 10 (Model 5) is included only for comparative
purposes and does not include the hours of caring measures. The inclusion of the
latter in estimations in Models 6 and 7 has very little effect on the carer
characteristics. The estimates show that employment participation is significantly and
negatively related to both HOURS>100 and CAREHOURScat. The marginal effect of
HOURS>100pw indicate that caring for 100 hours or more reduces the probability of
being employed by 0.26 (holding all variables at their mean value). The marginal
effect associated with the ordered variable CAREHOURScat is smaller and indicates
that longer hours of care reduce the probability of being employed by 0.1.
However, as already discussed there are unobserved covariates in these
estimations, such as unobserved ability, that potentially impact on both employment
participation and caring commitment. The results in Models 6 and 7 are therefore
28
likely to be biased due to the endogeneity of the caring measures. To address the
potential endogeneity, Models 6 and 7 were re-estimated using the probit conditional
maximum likelihood estimator with the instrumental variables specification. An
instrument for caregiving was constructed from indicators reflecting the
characteristics of the cared-for and their care-needs. The instruments were selected on
the basis of their significance in the estimations in Models 2-3 in Tables 6 and 8. As
argued above, it is unlikely that these measures are correlated with unobserved
determinants such as the carer‟s unobserved ability since the latter is likely to be
determined prior to the onset of illness of the cared-for person. However, this may not
be true for carers who have continued to care from a young age. Young carers are
excluded from this analysis but it is possible that the sub-sample contains adult carers
we were also young carers.
To test whether the chosen instruments were appropriately uncorrelated with
the error term we re-ran the probit estimations using Newey's two-step estimator for
probit with instrumental variables. This enabled us to obtain Amemiya-Lee-Newey's
Chi-squared test statistic to test for the validity of the instruments (using the Stata
command overid). This test is the equivalent of the Sargan test for TSLS. Use of the
Amemiya-Lee-Newey's Chi-squared test statistic suggested that the it was not
appropriate to use indicators for receipt of Attendance Allowance or Disability Living
Allowance as instruments in the analysis. This is perhaps unsurprising since receipt
of these benefits is an eligibility condition for receipt of Carer‟s Allowance which as
argued above is likely to be jointly determined with employment participation. Use of
the dichotomous indicators of co-residential status and whether the cared-for person
lived in institutional accommodation were also ruled out by this test. This suggests
that decisions about residential status, caring commitment and employment are jointly
29
determined. The indicators of personal care and emotional/other care provision were
not used as instruments since the evidence of Hassink and Van den Berg (2011)
suggests that this would be inappropriate.2 With the included instruments as listed in
the notes to Table 10, the Amemiya-Lee-Newey's Chi-squared test statistic is only
weakly significant (at the 10 percent level) in Model 6a in Table 10 and it is
insignificant in Model 7a (see notes to Table 10). Therefore it is possible to accept the
null hypothesis that the included instruments are uncorrelated with the error term.
A further complication is that in Stata, the probit estimator with instrumental
variables (ivprobit) assumes that the endogenous regressors are continuous and
neither HOURS >100 nor CAREHOURScat are continuous variables. As an
alternative to probit we estimated linear probability models using GMM to obtain
robust standard errors. The results from these estimations are reported in Table A3 in
the Appendix. For these estimations we report the Hansen J statistic which is the
GMM equivalent of the Sargan test for TSLS. Neither statistic is significant indicating
the null hypothesis of independence of the instruments and the disturbance process
need not be rejected.
After instrumenting, the marginal effects associated with HOURS>100 and
CAREHOURScat in Models 6a and 7a are absolutely larger than in Models 6 and 7.
This suggests that the endogeneity causes the influence of caring on employment to
be underestimated. The Wald test for endogeneity is weakly significant at the 10
percent level in estimation 6a and significant at the 5 percent level in estimation 7a.
This indicates that there is sufficient information to reject the null of no endogeneity
and that therefore the estimates in Models 6 and 7 are biased (downwards).
2 However, inclusion of these indicators as instruments was not ruled out by the Amemiy-Lee-Newey
test although their inclusion reduced the number of available observations and reduced the size of the
marginal effects associated with the instrumented care hours measures. Interestingly, inclusion of these
measures of instruments also resulted in the insignificance of the Wald test statistic for endogeneity
suggesting that indicators of type of care are potentially exogenous omitted variables.
30
Summary of results
The results indicate that time devoted to caregiving is influenced directly by the
characteristics of the cared-for person in terms of their age and their living
arrangements. Carers also care for longer hours when the needs of the people they
care for are severe enough for them to qualify for either Attendance Allowance or
Disability Allowance. Hours of care are also longer when the carer is providing
personal care. Not surprisingly, given the qualifying conditions, carers in receipt of a
Carer‟s Allowance provide longer hours of care.
The provision of personal care is more likely if the carer is an older female.
Co-residential carers are also more likely to supply personal care, particularly if the
shared home is the carer‟s home and the cared-for person is an adult with a physical
disability. The results relating to the employment-caregiving relationship are
consistent with those of previous research in that after addressing endogeneity, the
negative effect of longer hours of informal care on employment participation remains
statistically significant. Indeed, the size of the effect is larger after instrumenting.
5. Summary and implications
The research reported here is based on regional data relating to a sample of carers
living in the midlands region of the UK. As well as information on the carers
themselves and the number of hours of care they provided, the data include
information on the characteristics and needs of the people cared for as well as the type
of help given. The availability of this data allowed us to explore the relationships
between caregiving provision and these other factors in some depth. While the data
set is relatively small, the richness of the data in respect of care needs and the
31
characteristics of the cared-for leads to some new insights on the roles of these
factors in caring provision.
A limitation of the analysis is that the regional sample is composed of those
carers most heavily involved in informal care in terms of their time commitment, their
co-residency status and to a lesser extent the duration of their caring episodes. To the
extent that the results of this research are generalizable, this will only be true in
respect of carers in a similar position who represent only a minority of carers
nationally (Table 1). Nevertheless, it is important to examine the needs of the carers
who fall into this category as arguably it is this group that is in most need of support.
The results show that intensive carers providing longer hours of care are
older, more likely to be co-resident with the person they are caring for (particularly if
this is the carer‟s home) and giving personal care. Not surprisingly, those involved in
the longest hours are more likely to be in receipt of Carer‟s Allowance and the person
they care for is likely to be in receipt of either Attendance Allowance or Disability
Allowance.
The provision of personal care provides an alternative measure of the demands
or intensity of the caring role and the analysis indicates that this type of care is more
likely to be provided by older carers, those in receipt of Carer‟s Allowance, and co-
resident carers (particularly when the shared home is the carer‟s home). In addition
personal care is more likely to be given by women and more likely to be received by
older people. However, personal care is less likely to be given by those providing
elder care indicating that the bulk of personal care is likely to be spousal. We also find
a positive relationship between hours of caring and the provision of personal care.
This confirms that the provision of personal care and longer hours of care are
complements and possibly jointly determined. Nevertheless, whether personal care is
32
provided or not depends in part on the exogenously determined needs of the person
cared-for.
We find a negative relationship between the duration of the caring episode and
the provision of 100 or more hours of care. This suggests a potential tradeoff between
hours spent caring and the duration of a caring episode. Perhaps, those who manage to
supply care over the very long term find ways to reduce or contain the time they are
involved in care e.g. by organizing more help.
The analysis of the relationship between care-giving and employment
highlighted the negative relationship found in previous research between hours of
caring and employment participation. After instrumenting for hours of care the
negative coefficient on the measures of caring are absolutely larger suggesting that the
caring effect on employment is underestimated when endogeneity is not addressed. To
the extent that we are able to generalize from these results they confirm that carers
who care for substantial hours are less likely to be either willing or able to remain in
employment directly as a consequence of their caring responsibilities.
The evidence of this paper is that the care undertaken by informal carers is in
part determined by the characteristics and needs of the people they care for. Since all
the sample members were existing carers not much can be said about the effect on
employment of the initial decision to undertake care. However, the results can be
taken to imply that once the decision to provide care has been made, the amount and
type of care undertaken is at least to some extent exogenously determined by need.
Since this can change, some carers will have little discretion over their caring
provision. Notably, the results suggest that among carers who already care for quite
long hours, females are not significantly likely to be caring more hours. However,
female carers are more likely to be providing personal care, arguably the most
33
demanding or intensive type of care. This suggests that even among this group of
already time committed carers gendered norms of responsibility still apply.
REFERENCES
ACE National. 2006. “About ACE.” http://www.acecarers.org.uk/AboutACE
(accessed April 2006).
Age Concern and Help the Aged. (2007) “Parliamentary Briefing”
http://www.Ageconcern.org.uk/ageconcern/documents/general_committee_briefi
ng_lords_june07.pdf (accessed June 6 2007)
Anderson, Greg, Fiona Carmichael, Gemma Connell, Claire Hulme,, and Sally
Sheppard. 2005. “The Impact of Informal Care on Carers‟ Wellbeing; Health,
Wealth and Support Networks.” Management and Management Science Research
Institute Working Paper 207/05, University of Salford, UK.
Arber, Sara and Jay Ginn. 1995. “Gender Differences in the Relationship between
Paid Employment and Informal Care.” Work, Employment and Society, 9(3):
445–71
Arksey, H. 2002a. Rationed Care: Assessing the Support Needs of Informal Carers in English
Social Services Authorities. Journal of Social Policy, 31(1), 81-101
Arksey, H. 2002b. Combining Informal Care and Work: Supporting Carers in the Workplace.
Health and Social Care in the Community, 10(3), 151-161
Arksey, Hilary, Peter Kemp, Caroline Glendinning, Inna Kotchetkova, and Rosemary
Tozer. 2005. “Carers‟ Aspirations and Decisions around Work and Retirement.”
Department for Work and Pensions, Research Report 290, Her Majesty‟s
Stationery Office, Norwich
Baldwin, Sally. 1985. The Costs of Caring: Families with Disabled Children.
London: Routledge & Kegan Paul.
34
Bolin, K., Lindgren, B., Lundborg, P. 2008. Your Next of Kin or Your Own Career?
Caring and Working Among The 50+ of Europe. Journal of Health Economics,
27, 718-738
British Council. 2007. “Glossary of UK Educational Terms.”
http://www.britishcouncil.org/usa-education-uk-glossary.htm (accessed April
2006)
Bryan, M. 2011. Access to flexible working and informal care. Institute for Social
and Economic Research Working Paper, No. 2011-01, January
Carers UK. 2003. “Missed Opportunities: The Impact of New Rights for Carers”
Carers UK: London.
Carmichael F., Charles, S. and Hulme, C. T., (2010) “Who will care? Employment
status and willingness to supply informal care”, Journal of Health Economics, 29,
182-90
Carmichael, Fiona and Susan Charles. 1998. “The Labour Market Costs of
Community Care.” Journal of Health Economics 17(6): 645–795.
Carmichael, Fiona and Susan Charles. 2003a. “Benefit Payments, Informal Care and
Female Labour Supply.” Applied Economics Letters 10(7): 411–15.
Carmichael, Fiona and Susan Charles. 2003b. “The Opportunity Costs of Informal
Care; Does Gender Matter?” Journal of Health Economics 22(5): 781–803.
Carmichael F, Connell G, Hulme C, Sheppard S., (2008) Work life imbalance;
informal care and paid employment, Feminist Economics, April 14(2) 3-35
Carmichael F. and Hulme C., (2008) Are the Needs of Carers Being Met by
Government Policy? Journal of Community Nursing, September 22:, 4-12
Carers Policy Team, 2010. Recognised, valued and supported: next steps for the
Carers Strategy. Department of Health www.dh.gov.uk/publications (accessed
July 22 2011)
35
Carvel, John. 2006. “Baby Boomers Care in Old Age Set at £30 Billion.” Guardian,
30th
March.
Challis, David, Chengqiu Xie, Jane Hughes, Sally Jacobs, Siobhan Reilly, Karen
Stewart. 2005. “Social Care Services at the Beginning of the 21st Century.”
Discussion Paper M104, Personal Social Services Research Unit
Department of Health. 1998. Modernising Social Services Promoting Independence
Improving Protection Raising Standards: White Paper, Cm. 4169, London: HMSO.
Department of Health. 1999. Caring about Carers: a National Strategy for Carers.
http://www.dh.gov.uk/PublicationsAndStatistics/Publications/PublicationsPolicy
AndGuidance/PublicationsPolicyAndGuidanceArticle/fs/en?CONTENT_ID=400
6522&chk=yySBZ/ (accessed January 2006).
Department of Health. 2001. Carers and Employment: Report on Visits to Five Councils with
Social Service Responsibilities, London: HMSO
Department of Health. 2002. Fair Access to Care Services – Guidance on Eligibility Criteria
for Social Adult Care. Local Authority Circular (LAC(2002)14)
Department of Health. 2006(a). “Caring about Carers.” http://www.carers.gov.uk/
(accessed January 2005).
Department of Health. 2006(b) “Carers Grant 2006/07 and 2007/08 Guidance”
http://www.dh.gov.uk/en/publicationspolicyandguidance/DH_412840 (accessed
June 2007)
Department of Health 2008 Carers at the heart of 21st century families and
communities. www.dh.gov.uk/publications (accessed June 2008)
Department of Health 2010 Refreshing the national Carers Strategy – call for
evidence. Paul Barstow, Reference number 14557
http://www.dh.gov.uk/prod_consum_dh/groups/dh_digitalassets/documents/digi
talasset/dh_118046.pdf (accessed July 2011)
36
Department of Health 2011. Reaching Out to Carers Innovation Fund Department of
Health http://www.dh.gov.uk/en/SocialCare/Carers/DH_124163 22 (accessed
July 2011)
Department of Trade and Industry. 1998. Fairness at Work: White Paper, C3968,
London: HMSO
Directgov. 2006a. “Carer‟s Allowance.”
http://www.direct.gov.uk/en/CaringForSomeone/MoneyMatters/DG_1001252
(accessed January 2006)
Directgov. 2006b. “Direct Payments for Carers - Arranging Care and Services.”
http://www.direct.gov.uk/CaringForSomeone/MoneyMatters/MoneyArticles/fs/e
n?CONTENT_ID=10018517&chk=p0/b5B (accessed April 2006).
Eaton, Susan C. 2005. “Eldercare in the United States: Inadequate, Inequitable, but
Not a Lost Cause.” Feminist Economics 11(2): 37–51.
Ettner, Susan Louise. 1996. “The Opportunity Costs of Elder Care.” Journal of
Human Resources 31(1): 189–205.
Folbre, Nancy. 1995. “Holding Hands at Midnight: The Paradox of Caring Labor.”
Feminist Economics 1(1): 73–92.
Folbre, Nancy. 2004. “A Theory of the Misallocaion of Time” in Nancy Folbre and
Michael Bittman, eds., Family Time, The Social Organisation of Care, 7-24.
London: Routledge.
Hassink, Wolter H.J. and Van den Berg, Bernard, Time-Bound Opportunity Costs of
Informal Care: Consequences for Access to Professional Care, Caregiver
Support, and Labour Supply Estimates. IZA Discussion Paper No. 5433.
Available at SSRN: http://ssrn.com/abstract=1745708
37
Hancock, Ruth, Adelina Comas-Herrera, Raphael Wittenberg, and Linda Pickard.
2003. “Who Will Pay for Long-Term Care in the UK? Projections Linking
Macro- and Micro Simulation Models.” Fiscal Studies 24(4): 387–426.
Heitmueller, Axel and Kirsty Inglis. 2007. “The Earnings of Informal Carers: Wage
differentials and opportunity costs,” Journal of Health Economics 26(4): 821-841
Heitmueller, Axel. 2007. “The Chicken or the Egg? Endogeneity in the Labour
Market Participation of Informal Carers.” Journal of Health Economics 26(3):
536-559.
Hill, Trish, Cathy Thomson, Michael Bittman and Megan Griffiths, 2008. “What
kinds of jobs help carers combine care and employment?” Family Matters, 80,
27-32
Hirst, M. 2002. “Transitions to informal care in Great Britain during the 1990s.”
Journal of Epidemiology and Community Health, 56, 579-587
Hutton, Sandra. 1999. “The Employment of those Undertaking Informal Care: a
Longitudinal Analysis.” Social Policy Research Institute, Working Paper DH
1684 8.99 SH, University of York.
Lázaro, Nieves, Maria-Luisa Moltó, and Rosario Sanchez. 2004. “Paid Employment
and Unpaid Caring Work in Spain.” Applied Economics 36(9): 977–86.
Lewis, Suzan, Carolyn Kagan, Patricia Heaton, and Maureen Cranshaw. 1999.
“Economic and Psychological Benefits from Employment; the Experiences and
Perspectives of Mothers of Disabled Children.” Disability and Society 14(4): 561–
75.
Lilly, M. B., Laporte, A. and Coyte, P. C. (2007) Labor Market Work and Home
Care‟s Unpaid Caregivers: A Systematic Review of Labor Force Participation
38
Rates, Predictors of Labor Market Withdrawal, and Hours of Work. The Milbank
Quarterly, 85(4), 641-690
Lilly, M. B., Laporte, A. and Coyte, P. C. (2010) Do They Care Too Much To Work?
The Influence of Caregiving Intensity on the Labour Force Participation Of unpaid
Caregivers In Canada. Journal of Health Economics, 29, 895-903
Michaud, P., Heitmueller, A. and Nazarov, Z. 2010. A Dynamic analysis of informal
care and employment in England, Labour Economics, 17: 3, 455-465
Maher, Joanne and Hazel Green. 2002. Carers 2000. London: The Stationery Office.
National Statistics Online. 2003. “Census 2001: Carers.”
http://www.statistics.gov.uk/cci/nugget.asp?id=347 (accessed January 2006)
National Statistics Online. 2006(a) “Caring and carers - 6 million unpaid carers in the
UK”, http://www.statistics.gov.uk/cci/nugget.asp?id=1336 (accessed June 2007)
National Statistics Online. 2006(b). Patterns of Pay: Results of the ASHE
1997 to 2005. http://www.statistics.gov.uk/cci/article.asp?id=1348 (accessed
January 2006).
Nolan, Michael. 2001. “Supporting Family Carers in the UK: Overview of Issues
and Challenges.” British Journal of Nursing 10(9): 608–13.
Pavalko, Eliza K. and Julie E. Artis. 1997. “Women‟s Care-giving and Paid Work:
Casual Relationships in Late Midlife.” Journal of Gerontology: Social Sciences
52B(4): S170–9.
Phillips, J., Bernard, M., Chittenden, M. 2002. Juggling Work and Care. Bristol: The Policy
press and the Joseph Rowntree Foundation
Pickard, L. 2004. Caring for Older People and Employment. London: PSSRU
Pickard, Linda, Raphael Wittenberg, Adelina Comas-Herrera, Bleddyn Davies, and
Robin Darton. 2000. “Relying on Informal Care in the New Century? Informal
Care for Elderly People in England to 2031.” Ageing and Society 20(6): 745–72.
39
Ritchie, Jane and Liz Spencer. 1994. “Qualitative Data Analysis for Applied Policy
Research,” in Alan Bryman and Robert G. Burgess, eds. Analyzing Qualitative
Data, pp. 173–94. New York and London: Routledge.
Royal Commission. (1999) “With Respect to Old Age: Long Term Care – Rights and
Responsibilities”, (chairman: Professor Sir Stewart Sutherland) The Stationery
Office, http://www.archive.official-
documents.co.uk/document/cm41/4192/4192.htm (accessed June 2007)
Scourfield P. 2005. Implementing the Community Care (Direct Payments) Act: Will the
Supply of Personal Assistants Meet the Demand and at What Price?‟ Journal of Social
Policy, 34; 469-488
Spiess, C. Katharina and A. Ulrike Schneider. 2003. “Interactions between Care-
giving and Paid Work Hours among European Midlife Women, 1994 to 1996.”
Aging and Society 23(1): 41–68.
Stalker, Kirsten. 2003. “Carers.” Research, Policy and Planning 21(2): 57-61
Stark, Agneta. 2005. “Warm Hands in Cold Age – On the Need of a New World
Order of Care.” Feminist Economics 11(2): 7–36.
Twigg, Julia (2009) Models of Carers: How Do Social Care Agencies
Conceptualise Their Relationship with Informal Carers? Journal of Social Policy. 18,
1, 53-66 53
Van Houtven, Courtney Harold and Edward C. Norton. 2004. “Informal Care and
Health Care Use of Older Adults.” Journal of Health Economics 23(6): 1159–80.
Vetter, Rheba E.and Susan Myllykangas. 2006. “The Hour of Freedom: Using
Creative Movements to Facilitate Emotions Associated with Caregiving.” Paper
presented at the 2nd
International Conference on Physical Education, Coaching
and Health Fitness, June 29–30, Athens, Greece.
40
Vickerstaff, S., Loretto, W., Milne, A., Alden, E., Billings J., and White P. 2009.
Employment support for carers. Department for Work and Pensions Research
Report, No 597, September. HMSO
Women and Equality Unit. 2006. “Individual Income 1996/97 - 2004/05.”
http://www.womenandequalityunit.gov.uk/indiv_incomes/report2006.pdf
(accessed April 2007).
Women‟s Budget Group. 2005. “Women‟s Budget Group Response to Independence,
Wellbeing and Choice: Our Vision for the Future of Social Care for Adults in
England: Social Care Green Paper.”
http://www.wbg.org.uk/documents/WBGSOCIALCAREGreenPaperResponse.pd
f (accessed April 2007).
Women‟s Budget Group. 2006a. “A Gender Lens on Public Service Agreements
(PSAs).” http://www.wbg.org.uk/documents/WBG-
AGenderLensofPSAs2006.pdf (accessed April 2007).
Women‟s Budget Group. 2006b. “Women‟s Budget Group Response to the
Department of Work and Pensions White Paper – Security in Retirement:
Towards a New Pension System.”
http://www.wbg.org.uk/documents/WBGPensionsWhitePaperresponse_final11.0
9.06.pdf (accessed May 2007).
Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel Data.
Cambridge, Mass.:MIT Press
Yeandle, S., Bennett, C., Buckner, L., Shipton, L. and Suokas, A. 2006. Who cares
wins: The social and business benefits of supporting working carers. Report to
Carers UK
41
Table 1: National data: Carers and caregiving (British Household Panel Survey,
1991–2007)
Carers Non-carers
All respondents (no. of observations) 36,536 163,825
Percentage of all respondents 18.24 81.76
Percentage male 40.31*** 48.55
MEN (no. of observations) 14,726 79,534
Percentage of males 15.6 84.4
Average age 50.01*** 40.86
Percentage under 25 10.46*** 21.12
Percentage 25-64 67.14 66.68
Percentage 65 and over 22.40*** 12.31
Percentage of 25-64 year olds employed 68.36*** 81.54
Percentage of 25-64 year olds not employed
and not in training or educationb
27.67*** 14.48
Percentage with some health problem 49.85*** 34.31
Percentage with higher educational
qualification (higher or first degree, teaching
qualification, other)
37.19*** 38.17
Modal weekly hours of caring 0-4 -
Median weekly hours of caring 5-9 -
Percentage caring more than 20 hours a week 20.60 -
Percentage caring more than 50 hours a week 9.64 -
Percentage caring more than 100 hours a week 8.15 -
Percentage co-resident with cared-for person 40.36 -
WOMEN (no. of observations) 21.810 84.291
Percentage of females 20.6 79.4
Average age 49.06*** 39.81
Percentage under 25 8.00*** 21.20
Percentage 25-64 73.99*** 68.85
Percentage 65 and over 18.03*** 10.00
Percentage of 18-64 year olds employed 55.48*** 66.16
Percentage of 18-64 year olds not
employed/training/educationb 40.66*** 33.78
Percentage with some health problem 30.00*** 29.80
Percentage with higher educational
qualification (higher or first degree, teaching
qualification, other)
28.65*** 31.94
Modal weekly hours of caring 0-4 -
Median weekly hours of caring 5-9 -
Percentage caring more than 20 hours a week 23.62 -
Percentage caring more than 50 hours a week 10.98 -
Percentage caring more than 100 hours a week 9.19 -
Percentage co-resident with cared-for person 32.82 -
Notes: aEither unemployed, retired, maternity leave, family care, long-term sick/disabled other economic status
***; Independent samples test indicates that the null hypothesis value of a zero difference does not fall within
the 99%, 95% or 90% confidence interval implying that the reported mean is statistically significantly different
from the comparable figure for non-carers (at 1% level of significance) on basis of t statistic in two sample test.
Table 2: Characteristics of caregivers by age and gender
Carer characteristics
All carers
Younger carers: under 25
Mid-life carers: 25-64
Older carers: 65 and over
All Women Men All Women Men All Women Men All Women Men
Number of observations (sample %)
1985 (100%)
1383 602
206 (12.10%)
108
98 704 (41.36%)
561
143 792 (46.53%)
521
271
% female 69.67% 52.43%## 79.69%## 65.78%
Mean age n=1702
59.01 59.16 58.65 13.95%## 14.32 13.55 52.99## 52.69* 54.20 76.07 75.43** 77.30
% Employed n=954
21.28% 22.99%* 17.25% 0%## 0% 0% 41.12%## 39.35% 49.32% 5.02% 4.17% 6.5%
% In training or education n=954
16.88% 13.28%** 25.35% 99.37%## 99% 100% 0.24% 0.3% 0% 0% 0 0
% Not employed/ training/ Education n=954
60.48% 62.09% 56.69% 0.63%## 1% 0% 55.47%## 57.10% 47.95% 94.99% 95.83% 93.50%
% White ethnicity (British, Irish, European, other)
91.74% 92.77%*** 89.37% 91.75% 93.52% 89.80% 90.34% 92.16%** 83.22% 91.04% 91.36% 90.41%
% Asian or Black ethnicity
7.96% 6.94%** 10.30% 7.77% 5.56% 10.20% 9.09% 7.31%** 16.08% 8.97% 8.64% 9.59%
% other ethnic 0.30% 0.29% 0.33% 0.49%# 0.94% 0% 0.57%## 0.54% 0.7% 0 0% 0%
% in receipt of Carers Allowance
8.72% 9.76%** 6.31% 0%## 0% 0% 18.75%## 18.72% 18.89% 4.92% 5.57% 3.69%
% either carer and/or cared-for receives some state benefita
38.64% 40.13%* 35.22% 0%## 0% 0% 44.46% 45.10% 41.96% 47.98% 47.79% 48.34%
Notes: sample = 1985 unless n is given as less, percentages are by column **, *: Within age group and for row characteristic, mean of female sub-sample significantly different from mean of male sub-sample at 1% or 5% level ##
, #
: For row characteristic, mean within age group significantly different from mean of 65 and over age group (the modal category) a Either carer receives Carers Allowance, Income Support or Pension Credit and/or cared-for receives either Attendance or Disability Living Allowance
Table 3: Characteristics of cared-for person by caregivers’ gender and age
Cared-for characteristics All carers Younger carers: under 25 Mid-life carers: 25-64 Older carers: 65 and over
All Women Men All Women Men All Women Men All Women Men
% Male n= 1294 50.39% 61.61%** 23.9% 52.46% 50% 55.29% 45.25%## 51.50%** 19.23% 55.58% 77%** 11.83%
Mean age n= 1124 60.97 59.83** 63.98 22.03## 20.57 23.75 57.83## 54.96** 70.19 76.05 75.81 76.54
% older person (65 and over) n= 1124
58.27% 57.13% 61.29% 1.06## 0% 3.28% 49.71%## 47.70%** 58.33% 84.17% 82.96% 86.67%
% disabled child n=1257 15.20% 16.41%* 12.26% 5.15%## 49.46% 53.85% 16.57%## 19.85%** 19.61% 0.41% 0.62% 0%
% adult with physical disability n=1257
13.21% 11.12%** 18.26% 22.49%## 19.78% 25.64% 12.71% 10.66%** 21.57% 10.18% 9.57% 11.38%
% adult mental health or learning needs (includes older people) n= 1257
25.14% 25.71% 23.71% 17.75%# 21.98% 12.82% 28.18% 29.48% 22.55% 24.24% 21.61%** 29.34%
% employed n=1103 18.86% 18.91% 18.73% 32.89%## 32.53% 33.33% 22.04%## 22.14% 21.59% 11.12% 11.26% 10.96%
% retired n=1103 60.47% 59.14% 63.81% 2.01%## 0%** 4.55% 53.81%## 50.78%** 67.05% 86.56% 86.35% 87.00%
% adult not working/retired/student n= 1103
17.14% 16.50% 18.73% 32.89%## 32.53% 33.33% 19.28%## 18.75% 21.59% 9.80% 9.22% 10.96%
% student/pupil n= 1103 16.23% 17.40%* 13.33% 46.98%## 40.96% 54.55% 21.40%## 25.26%** 4.55% 1.14% 1.37% 0.68%
% receives Disability Benefit n=1985
4.23% 4.19% 4.32% 0##% 0% 0% 6.82%## 6.60% 7.69% 4.04% 3.26% 5.54%
% receives Attendance Allowance n= 1985
18.89% 18.94% 18.77% 0%## 0% 0% 11.79%## 12.30% 9.79% 30.05% 30.13% 29.89%
% live in own home or joint home (with carer) n= 1267
73.64% 68.64%** 85.56% 93.82%## 92.71% 95.12% 58.9%## 53.85%** 80.81% 81.76% 81.33% 82.63%
% live in caregivers home n= 1267
22.34% 27.10%** 10.96% 3.93%## 3.13% 4.88% 35.99%## 41.03%** 14.14% 15.23% 16.27% 13.17%
% live in medical/care/ residential/ sheltered/ supported home n= 1267
3.31% 3.46% 2.94% 0.56% 1.04% 0% 4.17% 4.2% 4.04% 3% 2.41% 4.19%
Notes: n = sub-sample size. Percentages are by column. **, *: Within age group and for row characteristic, mean of female sub-sample significantly different from mean
of male sub-sample at 1% or 5% level: ##
, #. For row characteristic, mean within age group significantly different from mean of 65 and over age group (the modal category)
Table 4: Type and amount of care-giving provided
Caring provision All carers Younger carers: under 25 Mid-life carers: 25-64 Older carers: 65 and over
All Women Men All Women Men All Women Men All Women Men
% fewer than 20 hrs per week n= 869
21.29% 17.00%** 30.86% 95.48%## 92.78%* 98.75% 2.69%## 2.61% 3.08% 1.04% 1.14% 0.88%
% between 20 and 49 hrs per week n= 869
12.66% 14.00%* 9.67% 45.20% 7.22%* 1.25% 20.16%## 20.85% 16.92% 6.57% 6.29% 7.02%
% at least 20 hrs per week 78.71% 83.00%** 69.15% 4.52%## 7.22%* 1.25% 97.31% 97.39% 96.92% 98.96% 98.86% 99.12%
% at least 50 hrs per week 66.05% 69.00%** 59.48% 0%## 0% 0% 77.15%## 76.55% 80% 92.39% 92.57% 92.11%
% at least 100 hrs per week n= 869
43.73% 46.33%* 37.92% 0%## 0% 0% 47.04%## 48.21% 41.54% 64.36% 65.14% 63.16%
Mean years caring n= 1926
2.77 2.66** 3.03 2.32## 2.27 2.39 2.33## 2.25* 2.65 3.30 3.23 3.44
% co-residential care n=1267
46.00% 46.06% 45.85% 43.69%## 42.59% 44.90% 50.28% 50.98% 47.55% 53.16% 53.17% 53.14%
% elder care (cared-for > 60 and min. 17 years older than carer) n=
26.02% 26.88% 23.78% 1.50%## 0% 3.28% 46.17%## 42.86%** 60.42% 10.85% 11.90% 8.67%
% gives personal care (with/without physical care) n= 988
41.70% 45.79%** 32.33% 2.82%## 4.17% 1.24% 47.42%# 48.71% 41.56% 53.45% 57.53%* 46.51%
% gives personal and physical care n= 988
39.98% 44.33%** 30.00% 2.26%## 3.13% 1.24% 46.01% 47.57% 38.96% 51.44% 55.71% 44.19%
% give only practical care (includes drug administration) n= 988
19.23% 19.33% 19.00% 0%## 0% 0% 24.65% 25.50% 20.78% 22.41% 19.64% 27.13%
% young carer excess chores/sibling care n= 988
17.11% 13.37%** 25.67% 94.35%## 93.75% 95.06% 0% 0% 0% 0% 0% 0%
% gives only emotional/’other ‘ care n= 988
18.42% 19.04% 17.00% 2.83%## 2.08% 3.70% 24.41%# 23.50% 28.57% 19.25% 19.18% 19.38%
Notes: n = sun-sample size, percentages are by column. **, *: Within age group and for row characteristic, mean of female sub-sample significantly different from mean of male sub-sample at 1% or 5% level.
##, #
: For row characteristic, mean within age group significantly different from mean of 65 and over age group (the modal category)
Table 5: Logit Regressions: Dependent variables is HOURS>50pw (1 =provides care for at least 50 hours a week; 0 = fewer than 50 hours) Independent variable Model 1
Carer characteristics
Model 2 Cared-for
characteristics
Model 3 Type of care
Model 4 Pooled
(significant variables only)
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Carer characteristics
MALE 0.75 0.5-1.2
AGE 1.09*** 1.1-1.1 1.12*** 1.1-1.5
EMPLOYED 0.64* 0.4-1.0 0.69 0.4-1.3
CARERS ALLOWANCE 5.23*** 2.7-10 7.89*** 3.5-17.9
ETHNIC BLACK/ASIAN 2.76 0.8-9.3
ETHNIC OTHER 4.00 0.1-114
Cared for characteristics
CF_MALE 1.29 0.9-1.9
CF_AGE 1.02* 1-1.1 0.96*** 0.9-1
ATTENDANCE ALLOWANCE
4.06*** 1.8-9.4 3.12** 1.2-8.4
DISABILITY LIVING ALLOWANCE
4.13** 1.2-14.2
4.38* 0.8-24.5
DISABLED CHILD 0.67 0.2-1.9
ADULT WITH PHYSICAL DISABILITY
1.40 0.7-2.8
ADULT WITH MENTAL HEALTH or LEARNING NEEDS
1.24 0.7-2.2
CF_EMPLOYED 1.54 0.6-3.7
CF_RETIRED 2.81* 0.9-8.9 1.03 0.3-3.2
INSTITUTIONALISED ACCOMMODATION
0.24*** 0.09-0.6
0.28** 0.8-1.0
LIVES IN CARER’S HOME
120.11*** 9.8-41.2 1.88 0.8-4.7
Type/amount of care provided
CO-RESIDENTIAL 6.88*** 4.3-11.0 2.43*** 1.2-5.0
DURATION (years) 1.01 0.8-1.2
ELDERCARE 1.62* 1.0-2.7 1.722 0.5-5.7
PERSONAL CARE 13.85*** 8.2-23.4 4.47*** 2.3-8.9
EMOTIONAL/OTHER CARE
3.79*** 2.3-6.3 0.58 0.3-1.2
No. of observations 763 687 683 638
Log likelihood -279.58 -314.54 --305.50 -180.91
Log likelihood Χ2
413.09*** 199.65*** 238.29*** 406.50***
Pseudo R2
0.425 0.241 0.281 0.529
Notes: Reference categories: Carer ethnic group; white ethnic. Carer economic/employment status; adult not working, adult/young person training/education. Cared-for need category; older person (65 and over) with mental or physical health needs. Cared-for economic status; not employed, student/pupil. Cared-for living arrangements; lives in own home. Type of help given; young carer chores /sibling care, only practical care (includes drug administration); only physical care ***, **, *: estimate significantly different from zero at 1%, 5% and 10% levels respectively
Table 6: Logit Regressions: dependent variable is HOURS>100pw (1 = provides care for at least 100 hours a week; 0 = fewer than 100 hours) Independent variable Model 1
Carer characteristics
Model 2 Cared-for
characteristics
Model 3 Type of care
Model 4 All significant variables in Models 1-3
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Carer characteristics
MALE 0.64*** 0.4-0.9 1.03 0.6-1.7
AGE 1.05*** 1.0-1.1 1.02*** 1.0-1.0
EMPLOYED 0.42*** 0.3-0.6 0.39*** 0.2-0.7
CARERS ALLOWANCE 2.78*** 1.8-4.3 4.22*** 2.3-7.7
ETHNIC_BLACK_ASIAN 1.87 0.7-5.3
ETHNIC_OTHER 7.04 0.5-92
Cared for characteristics
CF_MALE 1.32 0.9-1.8
CF_AGE 0.99 0.97-1
ATTENDANCE ALLOWANCE
3.14*** 1.9-5.3 3.16*** 1.6-6.1
DISABILITY LIVING ALLOWANCE
2.37** 1.1-5.1 3.38*** 1.3-8.7
DISABLED CHILD 1.18 0.5-3.0
ADULT WITH PHYSICAL DISABILITY
2.03** 1.1-3.7 0.85 0.4-1.8
ADULT WITH MENTAL HEALTH/LEARNING NEEDS
0.97 0.6-1.6
CF_EMPLOYED 1.66 0.8-3.7
CF_RETIRED 6.27*** 2.17-18 1.5 0.7-3.1
INSTITUTIONALISED ACCOMMODATION
0.87 0.3-2.3
LIVES IN CARER’S HOME
5.03*** 3.2-8 2.72*** 1.4-5.2
Type/amount of care provided
CO-RESIDENTIAL 3.05*** 1.9-4.8 1.26 0.7-2.3
DURATION (years) 0.68*** 1.9-4.8 0.58*** 0.5-0.7
ELDERCARE 0.49*** 0.3-0.8 0.29*** 0.1-0.6
PERSONAL CARE 8.78*** 5.8-13.3 6.13*** 3.7-10.2
EMOTIONAL/OTHER CARE
3.75*** 2.3-6.1 2.33*** 1.3-4.3
No. of observations 763 687 683 605
Log likelihood -423.01 -424.27 --362.81 -275.68
Log likelihood Χ2
198.85*** 99.76*** 211.11*** 283.05***
Pseudo R2
0.190 0.105 0.225 0.339
Notes: Reference categories as Table 4. ***, **, *: estimate significantly different from zero at 1%, 5% and 10% levels respectively
Table 7: Ordered Logit Regressions: dependent variable is CAREHOURScat (weekly caring hours ordered by category; less than 5; 5-9; 10-19; 20-34; 35-49; 50-99; 100+) Independent variable Model 1
Carer characteristics
Model 2 Cared-for
characteristics
Model 3 Type of care
Model 4 All significant variables in Models 1-3
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Carer characteristics
MALE 0.56*** 0.4-0.8 0.77 0.5-1.1
AGE 1.08*** 1.1-1.1 1.07*** 1.1-1.1
EMPLOYED 0.84 0.59-1.2
CARERS ALLOWANCE 4.14*** 2.7-6.3 5.01*** 3.0-8.3
ETHNIC_BLACK_ASIAN 2.84*** 1.3-6.1 1.69 0.7-4.3
ETHNIC_OTHER 7.92* 0.8-76.9 4.53 0.4-50.5
Cared for characteristics
CF_MALE 1.29* 1.0-1.7 0.84 0.6-1.2
CF_AGE 1.01 1.0-1.0
ATTENDANCE ALLOWANCE
3.15*** 1.9-5.2 2.94*** 1.6-5,2
DISABILITY LIVING ALLOWANCE
-3.20*** 1.5-6.6 3.10*** 1.4-7.0
DISABLED CHILD 0.69 0.3-1.4
ADULT WITH PHYSICAL DISABILITY
1.99*** 1.14-3.5
1.20 0.7-2.1
ADULT WITH MENTAL HEALTH/LEARNING NEEDS
0.99 0.65-1.5
CF_EMPLOYED 1.56 0.8-2.9
CF_RETIRED 4.79*** 2.1-11.2
1.09 0.6-2.1
INSTITUTIONALISED ACCOMMODATION
0.43* 0.2-1.0 0.30** 0.1-0.8
LIVES IN CARER’S HOME
13.75*** 8.4-22.6
3.95*** 2.3-6.8
Type/amount of care provided
CO-RESIDENTIAL 4.07*** 2.9-5.8 1.18 0.8-1.8
DURATION (years) 0.73*** 0.7-0.8 0.74*** 0.7-0.8
ELDERCARE 1.36* 1.0-2.0 0.57* 0.3-1.0
PERSONAL CARE 10.07*** 7.0-14.5 5.06*** 3.3-7.7
EMOTIONAL/OTHER CARE
4.05*** 2.7-6.1 1.60* 1.0-2.6
No. of observations 763 687 683 631
Log likelihood -921.77 -908.82 -904.87 -677.98
Log likelihood Χ2
533.18*** 235.20*** 290.40*** 553.15***
Pseudo R2
0.224 0.115 0.138 0.290
Notes: Reference categories as Table 4. ***, **, *: significantly different from zero at 1%, 5% and 10% levels respectively
Table 8: Logit Regressions: dependent variable is HOURS<20pw (1= provides care for less than 20 hours a week; 0 = at least 20 hours of care) Independent variable Model 1
Carer characteristics
Model 2 Cared-for
characteristics
Model 3 Type of care
Model 4 Pooled All significant
variables in Models 1-3
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Carer characteristics
MALE 2.92** 1.2-7.3 1.17 0.3-4.5
AGE 0.88*** 0.9-1.0 0.83*** 0.8-0.9
EMPLOYED 0.23*** 0.7-0.8 0.13** 0.02-0.8
CARERS ALLOWANCE 0.10*** 0.2-0.5 0.05*** 0.01-0.5
ETHNIC_BLACK_ASIAN 0.92 0.1-7.1
ETHNIC_OTHER 0.67 0.003-143
Cared for characteristics
CF_MALE 1.04 0.5-2.0
CF_AGE 0.94*** 0.9-0.97
1.03 1.0-1.1
ATTENDANCE ALLOWANCE
-- --
DISABILITY LIVING ALLOWANCE
-- --
DISABLED CHILD 2.51 0.6-10.4
ADULT WITH PHYSICAL DISABILITY
1.37 0.5-4.1
ADULT WITH MENTAL HEALTH/LEARNING NEEDS
1.07 0.4-3.1
CF_EMPLOYED 0.72 0.3-2.1
CF_RETIRED 0.31 0.1-1.7
INSTITUTIONALISED ACCOMMODATION
12.92***
3.3-51.4
70.08*** 8-614
LIVES IN CARER’S HOME 0.01*** 0.001-0.02
0.11** 0.01-0.9
Type/amount of care provided
CO-RESIDENTIAL 1.31*** 1.2-1.63
1.4 0.3-6.7
DURATION (years) 0.03*** 0.05-0.2
1.29 0.8-2.1
ELDERCARE 0.03*** 0.1-0.08
0.32 0.03-3.3
PERSONAL CARE 0.03*** 0.01-0.07
0.30 0.03-3.3
EMOTIONAL/OTHERCARE 0.07*** 0.03-0.2
10.58*** 1.8-61
No. of observations 763 687 683 626
Log likelihood -95.20 -135-56 -198.86 -46.40
Log likelihood Χ2
598.56*** 346.35*** 278.57*** 524.72***
Pseudo R2
0.759 0.561 0.412 0.850
Notes: Reference categories as Table 4. ***, **, *: significantly different from zero at 1%, 5% and 10% levels respectively
__ variable dropped as insufficient variation with respect to dependent variable; predicts failure perfectly
Table 9: Logit Regressions: dependent variable is PERSONAL CARE (1= provides personal care; 0 = physical, practical, emotional or other care) Independent variable Model 1
Carer characteristics
Model 2 Cared-for
characteristics
Model 3 Type of care
Model 4 All significant variables in Models 1-3
Odds ratio
95% CI Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
Carer characteristics
MALE 0.58*** 0.4-0.8 0.56*** 0.4-0.8
AGE 1.04*** 1-0-1.1 1.03*** 1.0-1.0
EMPLOYED 1.17 0.8-1.7
CARERS ALLOWANCE 2.33*** 1.6-3.5 1.87*** 1.2-2.8
ETHNIC_BLACK_ASIAN 3.29*** 1.6-3.5 2.05 0.8-5.2
ETHNIC_OTHER 1.33 0.1-16.9
Cared for characteristics
CF_MALE 1.45** 1.1-2 1.07 0.8-1.5
CF_AGE 1.03*** 1.01-1.04
1.01* 1.0-1.0
ATTENDANCE ALLOWANCE
1.30 0.8-2.1
DISABILITY LIVING ALLOWANCE
1.49 0.7-3
DISABLED CHILD 1.89 0.8-4.6
ADULT WITH PHYSICAL DISABILITY
3.84*** 2.1-7 3.25*** 2.0-5.3
ADULT WITH MENTAL HEALTH/LEARNING NEEDS
0.79 0.5-1.3
CF_EMPLOYED 0.59 0.3-1.2
CF_RETIRED 1.18 0.5-3.1
INSTITUTIONALISED ACCOMMODATION
1.52 0.7-3.5
LIVES IN CARER’S HOME
6.18*** 3.9-9.8 3.38*** 2.1-5.6
Type/amount of care provided
CO-RESIDENTIAL 2.75*** 1.9-4 1.48* 1.0-2.3
DURATION (years) 1.00 0.9-1.1
ELDERCARE 1.58** 1.1-2.3 0.92 0.5-1.7
EMOTIONAL/OTHER CARE
__ __
No. of observations 809 768 802 800
Log likelihood -482.15 -463.77 -526.59 -466.33
Log likelihood Χ2
121.64*** 118.30*** 32.63*** 159.50***
Pseudo R2
0.112 0.113 0.03 0.146
Notes: Reference categories as Table 4. ***, **, *: significantly different from zero at 1%, 5% and 10% levels respectively
__ variable dropped as insufficient variation with respect to dependent variable; predicts failure perfectly
Table 10: Probit Regressions of employment on caregiving: dependent variable is EMPLOYED (1= employed; 0 = not working; either not employed, in training or education, looking for work or training)
Model 5 Carer characteristics only
Model 6 Carer characteristics and
HOURS>100
Model 7 Carer characteristics and
CAREHOURScat
Model 6a Probit IV (CMLE)
HOURS>100 instrumented
Model 7a Probit IV (CMLE) CAREHOURScat
instrumented
Independent variable
Marginal Effect
95% CI S.E. Marginal Effect
95% CI S.E. Marginal Effect
95% CI S.E. Marginal Effect
95% CI S.E
. Marginal
Effect 95% CI S.E.
Carer characteristics
MALE 0.11**
0.005-0.210
0.05 0.122** 0.007-0.239
0.6 0.13** 0.009-0.241
0.06 0.13** 0.001-0.26
0.07 0.35** 0.009-0.693
0.18
AGE 0.07*** 0.047-0.100
0.14 0.071*** 0.042-0.1
0.015 0.08*** 0.055-0.113
0.02 0.05*** 0.016-0.081
0.02 0.20*** 0.101-0.289
0.05
AGE_squared -0.001*** -0.001- -0.0005
0.0001 -0.001*** -0.001- -0.0005
0.0001 -0.001*** -0.001 – -0.0006
0.0001
-0.001*** -0.0009- -0.0003
0.0002 -0.002*** -0.003- -0.001
0.0005
ETHNIC BLACK or ASIAN
0.07 -0.147- -0.291
0.11 0.048 -0.234-0.33
0.144 0.07 -0.221- 0.351
0.15 0.22 -0.127-0.57
0.18 0.61 -0.241-1.47
0.44
Indicators of hours of caring provision
HOURS>100pw -0.26*** -0.345- -0.18
0.04 -0.43*** -0.617- -0.26
0.09
CAREHOURScat -0.10*** -0.139 - -0.062
0.02 -0.20*** -0.86 -0.316
0.05
No. of observations
580 466 466 398 385
Log likelihood -321.96 -241.78 -247.10 -456.58 -736.2
Log likelihood Χ2
82.34*** 102.01*** 91.37*** 79.66*** 78.43***
Pseudo R2
0.113 0.174 0.156 - -
Wald exogeneity Χ
2
3.08* 4.25**
Notes: CMLE is the conditional maximum likelihood estimator. ***, **, *: significantly different from zero at 1%, 5% and 10% levels respectively Sample restricted to carers over 18 and younger than 70 (the oldest employed sample member is 80) Marginal effect is for a discrete change of a dummy variable from 0 to 1 Additional instruments for HOURS>100 and CAREHOURScat are ADULT WITH PHYSICAL DISABILITY, CF_RETIRED, LIVES IN CARER’S HOME, DURATION and ELDERCARE. CF_MALE is a an additional instrument for CAREHOURScat (CF_MALE is not significant in Model 2 in Table 6). Amemiya-Lee-Newey Χ
2 statistics for the Newey two-step estimation of Models 6a and 7a are 3.67 and 6.191. Neither is statistically significant signalling that the null hypothesis that the
instruments are uncorrelated with the error term can should not be rejected.
APPENDIX Definitions of allowance available to carers and cared-for Attendance Allowance is a benefit for disabled people aged 65 or over, who find it difficult to care for themselves because of a disability or long term health problem. You can get Attendance Allowance if you find things like dressing and washing very difficult, if you need someone to make sure that you are safe, or you have a terminal illness.
Disability Living Allowance (DLA) is a benefit paid to people aged under 65 who need to help to look after themselves and/or to get around because of a long-term health problem or disability. This might be because they: Need help to get washed, dressed, take medication etc; Need someone to keep an eye on them to make sure that they are safe; Have a terminal illness; Have problems walking when out of doors
Carer's Allowance is the main state benefit for carers. Carer's Allowance is currently worth £53.90 a week (2010-2011 rate). To qualify carers need to be 16 years old or over ; look after someone for at least 35 hours a week; the person they look after must receive a qualifying disability benefit (AA or DLA); must not earn more than £100 a week (2010-2011 rate); must not receive one of a list of other benefits; must be living in the UK; must not be a full-time student
Table A1: UK policy initiatives supporting and responding to the needs of informal
carers
Year Title Issues addressed (indicative)
1995 Carers
(Recognition and
Service) Act (1995)
Recognition of the role of informal carers in society, raising
the profile of carers and creating awareness of carers‟ issues.
1998 Modernising
Social Services
Highlighted a need to support carers
1998 Fairness at Work Proposals to give all employees the right to time-off to
deal with family emergencies
1999 National Strategy
for carers
Recognition of carers‟ role and carers‟ needs.
1999 Working Families
Tax Credit
Parents with disabled children entitled to claim
childcare costs for children up to the age of 16
(previously 12)
2000 Carers and
Disabled Children
Act
Entitled carers to an assessment in their own right and
required local authorities to provide direct services to
carers to meet their assessed needs.
2001 Health and Social
Care Act
Direct Payment Scheme entitled carers to direct cash
payments from their local council to pay for short
breaks, nursery placement providing specialist support
for children, assistance to attend an activity, and
personal care
2002 The State Second
Pension
Provided a more generous additional State Pension for
low and moderate earners, some carers, and people
with long-term illness or disability.
2002 Amendment to
The Employment
Relations Act
Parents of disabled children under 18 entitled to request
flexible working arrangements and unpaid time off in
an emergency
2003 Fair Access to
Care
Provided councils with a framework for setting their
eligibility criteria for adult social care to create fairer
and more consistent eligibility decisions across the
country.
2004 Carers (Equal
Opportunities)
Act
The Act came into force in April 2005. It focuses on
health, employment, and life-long learning issues for
carers.
2006 Work and
Families Act
Carers of some adults given the right to request flexible
working arrangements
2006 White Paper More support for carers promised in Our Health, Our
Care, Our Say: A New Direction for Community
Service
2007 New Deal for
Carers
Extra funding of £33 million in the Carers grant
dedicated to respite care, a carer training programme
and national helpline
2008 New Carers‟
Strategy
Strategy incorporating the changes in the 2006 White
Paper
2010 Next steps for the
carers strategy:
Response to the
call for views
Identifies the actions that the Government will take
over the next four years to support carers. Priority areas
identified as: early identification; education and
employment; personalised support; carers‟ health
2010/11 Reaching out to
Carers Innovation
Fund
£1.35m funding for 79 projects „to improve early
identification of carers so that they can be supported in
considering their various options and make informed
choices about their lives‟.
Updated from Carmichael et al. (2008)
Table A2: Definitions of variables used in the analysis
Variable Definition
Type/amount of care provided
HOURS>50pw Caregiver provides at least 50 hours of care a week (=0,1).
HOURS>100pw Caregiver provides at least 100 hours of care a week (=0,1).
HOUR<20pw Caregiver provides fewer than 20 hours of care a week (=0,1).
CAREHOURScat Ordered variable indicating whether caregiver cares for less than 5, 5-9, 10-19, 20-34, 35-49, 50-99 or 100 or more hours a week (=0, 1,2,3,4,5,6).
CO-RESIDENTIAL Caregiver provides co-residential care, either in own home or home of cared-for (=0,1).
DURATION (years) Number of years has been registered with CSAM (proxy for number of years has provided care).
ELDERCARE Caregiver provides care for a person who is 60 or over and at least 17 years older than the caregiver (=0,1, proxy for elder care).
PERSONAL CARE Caregiver provides personal care (=0,1). Reference category is caregiver is either a young carer (chores or sibling care), or provides only practical care (include drug administration) or only physical care.
EMOTIONAL/OTHER CARE
Caregiver provides emotional or ‘other’ care (=0,1). Reference category is caregiver is either a young carer (chores or sibling care), or provides only practical care (include drug administration) or only physical care.
Carer
characteristics/status
MALE Caregiver is male (=0,1)
AGE Caregiver’s age
EMPLOYED Caregiver is employed (=0,1). Reference category is adult not working or young person in training or education.
CARERS ALLOWANCE Caregiver receives Carers Allowance (=0,1)
ETHNIC_BLACK_ASIAN Caregiver’s ethnic group is ‘black or Asian’ (=0,1). Reference category is white.
ETHNIC_OTHER Caregiver’s ethnic group is ‘other’ (=0,1). Reference category is white.
Cared for characteristics
CF_MALE Cared-for is male (=0,1).
CF_AGE Age of cared-for person.
ATTENDANCE ALLOWANCE
Cared for receives Attendance Allowance (=0,1).
DISABILITY LIVING ALLOWANCE
Cared for receives Disability Living Allowance (=0,1).
DISABLED CHILD Cared-for is a disabled child of the caregiver (=0,1). Reference category is older person (65 and over) not categorised as having mental health or learning needs.
ADULT WITH PHYSICAL DISABILITY
Cared-for is an adult with a physical disability (=0,1). Reference category is older person (65 and over) not categorised as having mental health or learning needs.
ADULT WITH MENTAL HEALTH/LEARNING NEEDS
Cared-for is an adult with mental health or learning needs, includes older people with mental health or learning needs (=0,1). Reference category is older person (65 and over) not categorised as having mental health or learning needs.
CF_EMPLOYED Cared-for is employed (=0,1). Reference category is adult not working or young person in training or education.
CF_RETIRED Cared-for is retired (=0,1). Reference category is adult not working or young person in training or education.
INSTITUTIONALISED ACCOMMODATION
Cared-for lives in institutionalised accommodation e.g. nursing home (=0,1). Reference category is cared-for lives in own home
LIVES IN CARER’S HOME
Cared-for lives in caregiver’s home (=0,1). Reference category is cared-for lives in own home
Table A3: GMM Linear Probability Model Regressions of employment on caregiving: dependent variable is EMPLOYED (1= employed; 0 = not working; either not employed, in training or education, looking for work or training)
Model 6a IV (GMM)
HOURS>100 instrumented
Model 7a IV (GMM)
CAREHOURScat instrumented
Independent variable Coefficient Robust S.E. Coefficient Robust S.E.
HOURS>100pw -0.438*** 0.116
CAREHOURScat -0.215*** 0.056
MALE 0.9085 0.06 0.094 0.061 AGE 0.035*** 0.013 0.05*** 0.016
AGE_squared -0.0004*** 0.0001 -0.0006*** 0.0001
ETHNIC BLACK or ASIAN -0.203 0.138 0.699 0.538
Constant -0.00003 0.362 0.699 0.538
No. of observations 398 385
F 18.19*** 18.98***
Centered R2
0.161 0.107
Uncentered R2 0.431 0.394
Root MSE 0.428 0.441
Hansen j statistic ( Χ2 for test
of overidentification of instruments)
4.776 4.663
Notes: ***, **, *: significantly different from zero at 1%, 5% and 10% levels respectively Sample restricted to carers over 18 and younger than 70 (the oldest employed sample member is 80) Additional instruments for HOURS>100 and CAREHOURScat are as for Table 10