Child health and parental paid work
Peter Burton • Kelly Chen • Lynn Lethbridge •
Shelley Phipps
Received: 16 September 2013 / Accepted: 14 May 2014
� Springer Science+Business Media New York 2014
Abstract We ask how the paid work of Canadian married mothers and fathers is
affected when a child has a physical/mental condition or health problem that leads
to restrictions in daily activities. Using the Statistics Canada National Longitudinal
Survey of Children and Youth, we find that married mothers of children with
disabilities are less likely to engage in paid work and/or work fewer paid hours per
week. No statistically significant changes in paid work participation or hours are
apparent for fathers of the same children. We find, moreover, evidence that the
degree of specialization within families increases when there is a child with a
disability. These responses are consistent with traditional gender roles within
families, and may make sense as a ‘household’ coping strategy. However, such a
division of labor may generate economic vulnerability for mothers compared to
fathers.
P. Burton � S. Phipps
Department of Economics, Dalhousie University, Halifax, NS B3H 3JH, Canada
e-mail: [email protected]
S. Phipps
e-mail: [email protected]
K. Chen (&)
Digonex Technologies Inc., 150 West Market Street, Indianapolis, IN 46204, USA
e-mail: [email protected]
L. Lethbridge
Community Health and Epidemiology, Dalhousie University, 5790 University Ave, Halifax,
NS B3H 1V7, Canada
e-mail: [email protected]
S. Phipps
Department of Economics, Dalhousie University and the Canadian Institute for Advanced Research,
Halifax, NS, Canada
123
Rev Econ Household
DOI 10.1007/s11150-014-9251-z
Keywords Child disability � Maternal labor force participation � Within-
household � Intra-household � Family � Gender
JEL Classification I14 � J14 � J16
1 Introduction
In this paper, we ask how the paid work of Canadian married1 mothers and fathers is
affected when a child has a physical/mental condition or health problem that leads
to an activity limitation.2 This is an important question given both the growth in the
number of children living at home with disabilities3 and the growth in the number of
married-couple families in which both mother and father are engaged in paid work.
Although demands placed upon families vary with the nature and severity of the
child’s health problem, it is often the case that there are both time and financial
pressures to be faced (Gould 2004; Stabile and Allin 2012).
Children with disabilities often have frequent hospitalizations, may need to
attend many medical appointments or therapy sessions, require extra help with
schoolwork or parental advocacy for their special needs and/or require additional
physical care (Beagan et al. 2005). At the same time, supports commonly available
to other parents engaged in paid work may not be available when a child has a
disability or serious health problem (Thyen et al. 1999). For example, daycare for
children with disabilities is often very limited; even friends and family may be
reluctant to take on the care of a child who requires specialized treatment. Some
parents may thus be unable to continue in paid work (or in paid work of the same
kind and quantity) if a child has serious health problems, with negative implications
for family financial well-being.
At the same time, even in Canada where doctor or hospital bills are not an issue,
many other expenses are not covered by public health insurance. For example, extra
money may be needed to build wheelchair ramps, buy hearing aids, travel from a
rural area to visit specialists in the city, or pay ‘deductibles’ on drugs. US studies
document extra financial costs incurred by families of children with disabilities or
chronic conditions (e.g., Hobbs and Perrin 1985; Lukemeyer et al. 2000; Meyers
et al. 1998). While these may be lower in Canada, data from the Statistics Canada
Participations and Activity Limitations Survey nonetheless indicate that 67 % of
1 Throughout the paper, ‘married’ refers to both legal and common-law marriages.2 Starting from 2000, there has been a change in the definition of ‘disability’ in Canadian national
surveys (Human Resources Development Canada 2003). In this paper, we focus on children reported to
experience functional limitations at home, at childcare, at school or in any other activities such as
transportation, play, sports or games, for a child of his/her age, in order to construct consistent measure
across survey years.3 Powers (2003) presents US data. Trends across time in child disability rates are harder to identify with
Canadian data, given changes in definitions used by nationally representative surveys (Human Resources
Development Canada 2003), but are likely to follow a similar trend. This may, paradoxically, be partially
due to advances in medical science which mean that children survive but live with health problems; it is
also the case that fewer children with serious disabilities are institutionalized than was previously the case
(Salkever 1982a).
P. Burton et al.
123
children with severe to very severe disabilities have ‘unmet needs’ for specialized
aids; in the majority of cases, needs were unmet due to cost (Burton and Phipps
2009; Statistics Canada 2001). Thus, families in which a child is not well may be
caught in a ‘double bind’ of needing both more time at home and more money.
1.1 Theoretical perspectives on household behavior
Since the work of Mincer (1962) and Becker (1965), theoretical models of
household behavior have recognized that families combine time and money to
produce domestic goods and services (e.g., clean clothes and hot meals) that
ultimately generate utility for family members. Both authors argue that the parent
with the higher market wage (typically the father) is likely to specialize in the paid
work necessary to earn money needed to purchase inputs to household production
while the parent with the lower market wage (typically the mother) devotes her time
to work at home. How non-market production is actually carried out by any given
family is hypothesized to depend upon available technology as well as the relative
costs of ‘production inputs’ (e.g., mother’s time versus purchased market
substitutes).
Extending these initial insights, Leibowitz (1974), in the context of child health
production and Browning (1992), in the context of parents deriving utility from the
well-being of their children both argue that the health and well-being of children
also depends upon both purchased inputs (food, clothing) and parental time. If a
child becomes disabled or develops a chronic condition, time and money needs will
both increase and one solution to the increased pressure may be increased
specialization within the family, even if this was not previously the case.4 Since
most wives earn less than their husbands, relative opportunity costs are likely to
mean that, from a family perspective, it makes most sense for the mother to reduce
paid work. Women who are mothers may also have chosen more ‘family-friendly’
jobs in order to accommodate ‘regular’ care-giving needs (e.g., chicken pox or the
‘flu) which could also make it relatively easier for them to reduce hours when faced
with more serious child health problems.
Even if the mother is not the parent with lower earnings or a more flexible job,
the ‘identity’ model proposed by Akerlof and Kranton (2000) suggests that she may
be the one who reduces paid work to do the care-giving. In this context, behavior
consistent with a ‘good mother’ identity might be to prioritize the care-giving role.
The mother of a child who is seriously ill may feel she ‘should’ be his or her
principal care-giver (and this may be reinforced if extended family and health-care
workers also think this should be the mother’s role). Qualitative evidence is
consistent with such reasoning. In a study of parents of children with high-
functioning autism, Gray (2003) quotes a mother as saying: ‘‘Yes, I do work but…of course, again [my son] dominated that’’ (p. 636).
4 Lone parents obviously do not have this option and so are faced with extraordinarily difficult
circumstances. However, we do not study lone mothers in this paper given our focus on ‘household’
responses with potentially different roles for mothers and fathers connected to gendered norms.
Child health and parental paid work
123
On the other hand, behavior consistent with a ‘good father’ identity when a child
is very ill or disabled might be to prioritize bread-winning (e.g., by not reducing
current paid hours or even possibly increasing them, though options for increasing
hours are likely to be limited if most fathers are already working full time). Again,
such behavior may be reinforced through the re-actions of outsiders, including
extended family and health-care professionals. Evidence from qualitative studies is
also consistent with this idea: ‘‘So, I’d basically come home and have my tea,
shower, bit of rest, change, go back to the office and do another 3 h of work, which
was quite stressful’’ (father quoted by Gray 2003, p. 635).
Notice, too, that policies and institutions may, perhaps inadvertently, serve to
reinforce adherence to traditional norms of behavior. For example, if standard paid
work weeks involve very high hours or if there is no daycare available for children
with special needs, it will be difficult for parents to share bread-winning and care-
giving; to behave as a ‘dual-earner/dual-carer’ couple (Gornick and Meyers 2003).
While specialization may make sense as a family coping strategy, it can have
negative implications for the personal economic well-being of the mothers. This
point is also noted in the literature on bargaining models of household behavior
which emphasize the relative earnings of husband and wife as key predictors of
bargaining power (e.g., Chen and Woolley 2001; Lundberg et al. 1997). Thus, role
specialization as a means of coping with a child disability has the potential of
reducing the mother’s bargaining power within marriage if it leads to the erosion of
job-related human capital and hence earnings potential over time. Second, reducing
paid hours or withdrawing completely from paid work may have negative long-term
implications for the mother’s financial well-being if the couple should divorce,5 or
even if they remain married but she does not have the opportunity to gain labor
market experience, pension entitlements, etc. This might be viewed as a more
extreme version of the ‘child penalty’ documented for women with children
compared to women without, regardless of the child’s health (see, for example,
Waldfogel 1998).
In summary, we hypothesize that married mothers and fathers will not respond in
the same way to a reduction in child health status. Instead, it seems likely that: (1)
mothers will reduce paid hours or even withdraw from the labor market; (2) fathers
will not reduce paid hours (and indeed may, if anything, do more hours of paid
work); (3) there will, as a result, be increased specialization according to traditional
gender roles within the family. To investigate these hypotheses, we use a sample of
children with married-couple parents drawn from Statistics Canada’s National
Longitudinal Survey on Children and Youth (NLSCY).
Our main contributions to the literature are as follows: first, since we have labor
market data for both mother and father, we can study differences in their responses.
Results suggest increased specialization within the family if there is a child with a
disability present. We thus highlight the ‘household’ nature of parental decision
5 Pollak (1985) notes that complete specialization in non-market production might be regarded as an
extreme investment in ‘marriage-specific’ human capital, which would both increase the ‘payoff’ to
remaining married, but also reduce the ‘payoff’ to leaving the marriage if no market human capital is
acquired (Pollak 1985). .
P. Burton et al.
123
making when a child develops a disability, an aspect of the situation which has thus
far received very little attention in the literature.
Second, we use longitudinal data to study the onset of child disability on parental
paid work in order to compare families in which a child develops a disability with
other families with similar observable characteristics. We find that mother’s
participation in paid work and usual weekly hours fall, controlling for her labor
market behavior before the activity limitation appeared; no association appears to
exist for fathers of the same children.
Third, since much of the existing literature on child disability and parental paid
work uses US data, but policies and institutions for families of children with
disabilities differ across countries, it is important to examine the impact of child
disability on parental paid work in other contexts. Policy differences between
Canada and the US which may be relevant in this context are: (1) medical expenses
connected with child disability are likely lower in Canada given universal public
health insurance; (2) nearly all families with children receive cash transfers and
families of children with disabilities receive extra benefits (see Burton and Phipps
2009).6 Thus, the need to increase paid work in order to cover additional expenses
may be somewhat less urgent for Canadian families.7
The remainder of the paper is organized in the following way: Sect. 2 provides a
brief review of the relevant literature. Section 3 describes the data. Contempora-
neous estimates of the association between child disability and parental paid work
are presented in Sect. 4. Section 5 discusses the onset models. Section 6 provides
discussion and conclusions.
2 Previous empirical literature
A small early literature using primarily US data studied cross-sectional associations
between maternal paid work and child disability, thus mingling health conditions
that have just appeared with those that have existed for many years. Findings from
these early studies are fairly consistent, indicating reduced probabilities of
participating in paid work by married mothers of children with disabilities (e.g.,
Breslau et al. 1982; Gould 2004; Kimmel 1998; Powers 2003; Salkever 1982a, b) as
well as lower paid hours, given participation (Gould 2004; Powers 2003; Salkever
1982b).8 Larger negative impacts are apparent for lower-income mothers (Breslau
et al. 1982; Salkever 1982b); smaller associations are apparent when young children
6 As well, some medical expenses can be deducted from taxes owing and tax credits are also available.
See Burton and Phipps (2009).7 Six weeks of ‘Compassionate Care’ benefits as part of the Canadian Employment Insurance program
were also available during our study period; however, parents in our sample would not generally be
eligible since these benefits were only available if the child is ‘was at significant risk of death’.
Compassionate Care take-up rates were thus very low. .8 In contrast, a slightly larger literature focussed on lone mothers generates more ‘mixed’ results,
sometimes finding no impact on paid work (e.g., Salkever 1982a; Kimmel 1997, 1998), sometimes finding
a negative impact (e.g., Baydar et al. 2007; Breslau et al. 1982; Lukemeyer et al. 2000; Salkever 1990;
Wolfe and Hill 1995).
Child health and parental paid work
123
are present in the family (Powers 2001; Salkever 1982b). Several studies point out
that size of the estimated association between child disability and mother’s paid
work will vary with the nature and severity of the child’s health problem, with larger
reductions when the condition is more severe or when multiple disabilities are
present (e.g., Powers 2003; Salkever 1982b).
More recent studies have examined the dynamics of maternal response to a
child’s disability (see also Stabile and Allin 2012 for a recent review). For example,
Powers (2003) uses pooled SIPP panels to estimate changes in work activity
(‘dropping out’ of paid work; changes in paid hours) for mothers whose children
have disabilities in the starting year of the analysis. All explanatory variables take
starting-year values. Powers is able to estimate changes in maternal labor market
activity one year later and two years later. Results for her dynamic models are
somewhat weaker than for her static, cross-sectional models in the sense that fewer
variables are statistically significant. For married mothers, she finds little evidence
that having a child with a disability affects hours or participation in paid work (in
fact, the only cases of statistical significance for child disability variables indicate,
somewhat non-intuitively, increases in hours).
Using the U.S. ‘Fragile Families’ data, focused on low-income unmarried
mothers, Corman et al. (2005) find that mothers of children born with health
problems are less likely to be in the labor force when the child is one year old,
controlling for baseline characteristics of both mother and father. Baydar et al.
(2007) estimate probabilities of withdrawing from full-time employment for
mothers of children with asthma and find that single mothers (though not married
mothers) of children with asthma had increased odds of leaving full-time work.
Finally, Kvist et al. (2013) use Danish register data to estimate labor supply for
parents of first-born ten-year old children diagnosed with ADHD. The particular
strengths of their study are that they need not rely on parental assessments of child
health and can condition on a rich set of covariates measured prior to the birth of the
child. They find, for both mothers and fathers, a reduction in paid work. The size of
the association is larger for mothers than fathers (e.g., 5–8 days per year vs. 4–6).
Mothers of children with ADHD also have a 2 % point lower probability of
participation in paid work.
Gould (2004) very usefully emphasizes that some disabilities are demanding of
parental time, others are very expensive and some require both time and money.9
She develops an individual model of mother’s behavior reflecting the different
implications of these aspects of child disability, conducts focus groups with medical
experts to ascertain which sorts of disabilities will require more time compared to
money and supports her model with an empirical analysis using the PSID.
As is evident from the discussion above, much of the research on the paid work
implications of parenting a child with a disability has focussed on implications for
the mother. However, Salkever (1982b) symmetrically studies father’s responses,
and finds no statistically significant association with child disability status; Noonan
9 Hobbs and Perrin (1985) provide discussions of the nature and implications of individual childhood
chronic conditions.
P. Burton et al.
123
et al. (2005) find reductions in the labor supply of fathers of newborns with health
problems in a US population of mostly unwed parents.
There has thus far been relatively little explicit discussion in the economics
literature of the possible ‘household’ nature of responses to the onset of child health
problems. Although most empirical studies of the paid work of married mothers of
children with disabilities control for the husband’s characteristics (e.g., his earnings,
health status), the idea that both parents may respond, while perhaps specializing
according to traditional roles is not explicitly considered. The relative lack of
attention to what happens to the paid work of fathers of children10 with disabilities
may also in itself be indicative of implicit assumptions that married parents will
behave in accordance with traditional roles—that fathers will specialize as bread-
winners and not reduce paid hours or withdraw from paid work when a child has
health problems.
Following on from Gould’s (2004) argument that child health problems can be
demanding in terms of both time and/or money, it is plausible to suppose that for
married-couple parents, deciding how to cope with the onset of a child’s serious
health problem will require household decision-making. Although not the focus of
her paper, Gould recognizes the possibility of ‘household’ responses to a child’s
health problem by married-couple parents by summing mother’s and father’s hours
and estimating the impact on total family hours (Table 6, p. 536). She finds no
statistically significant relationships between total family hours and her child
disability measures. Her interpretation, not pursued, is that there is ‘potential
substitution between mother’s and father’s work hours’ (p. 538).
3 Data
The data set employed is the Statistics Canada National Longitudinal Survey of
Children and Youth (NLSCY), a nationally representative survey of Canadian
children conducted every 2 years from 1994 to 2008). The surveys we use were
answered by the ‘person most knowledgeable’ about the child (or, pmk). Children in
our sample range from age 0 to 17 (18 is the age of majority in Canada).
Our parental paid work variables are derived from a question about ‘usual’ hours
in the year before the survey: ‘‘During the past 12 months, how many weeks did %
you/he/she % do any work at a job or business? Include weeks on paid vacation
leave, paid maternity or parental leave,11 paid sick leave. About how many hours a
week did % you/he/she % usually work?’’
These questions are answered by the pmk, for both parents. Since in 96 % of our
sample, the pmk is the mother of the child, it is possible that mother’s paid work
time is more accurately reported than father’s, though we would argue that
10 Some surveys may not provide details on father’s labor market behaviour, if, for example he is not
present in the household (i.e., does not reside with the child).11 Since maternity and parental leaves are not counted as time outside the labor market, we should not
expect any impact of the 2001 extension of parental benefits on the reported paid work of new parents.
Child health and parental paid work
123
participation in paid work is easily observable and even paid hours per week can be
quite accurately reported by one’s spouse.
We define a child to have a disability if he or she has an activity limitation that
prevents his/her activities at home, at childcare, at school or in any other activities,
for example, transportation, play, sports or games, for a child of his/her age.12 Note
that these are not mutually exclusive categories—a child could be limited in
multiple functional domains. Averaged across all cycles, 6.1 % of children meet one
or more of these criteria.
As noted earlier, the ‘usual weekly hours of paid work’ refer to the year
preceding the survey while child activity limitations are reported for the survey
year. Since we want to be certain that the child’s disability status is contempo-
raneous with reported labor market status, we use child’s activity limitation status
from 1 cycle earlier (and exclude children who ‘recovered’). Thus, for example, a
survey conducted in 2008 would report labor force status for calendar year 2007.
We would then use child’s disability status as reported for 2006 (with any children
activity limited in 2006 but no longer limited in 2008 excluded).13
Since all information about the child’s health is reported by the pmk, the question
of reliability of such reports may arise. While there is some evidence of
inconsistencies between medical records and mother reports (e.g., Miller et al.
2001), in general, consistency between medical records and self reports appears to
increase with the severity of the condition (e.g., Baker et al. 2001). We argue that
the activity limitation measure we study is both relatively severe and that the
questions are easier to answer than questions such as ‘how healthy is your child?’
Powers (2001) argues that reports of the severity of child disability status may be
endogenous to maternal labor market behavior (e.g., because mothers who withdraw
attempt to justify their behavior by emphasizing the severity of the child’s
condition). However, Powers also argues that maternal reports of specific
impairments are more likely objective. Since the NLSCY does not ask pmks to
assess the severity of the condition, but only to report if there is any restriction of
activities, we believe that reporter bias is likely to be fairly small in our case.14
Given our interest in comparing labor market responses of both mothers and
fathers, we also select only children in married-couple households. It is, however,
important to recognize the possibility that parenting a child with a disability or
chronic condition may increase the probability of parental divorce, and that, if this
12 The survey questions that we used to construct child disability variable are as follows: ‘‘Does child
have any long term conditions or health problems which prevent or limit % his/her % participation in
school, at play, or in any other activity for a child of % his/her % age?’’ (cycles 1–3), or ‘‘Does a physical
condition or mental condition or health problem reduce the amount or the kind of activity this child can
do: (1) at home? (2) at childcare? (3) at school? (4) in other activities, for example, transportation, play,
sports or games? (cycles 4–8)’’.13 We have also estimated all models retaining the ‘recovering’ children. Results are not affected.14 A limitation of the NLSCY is that it is not possible to provide separate estimates for children more or
less severe activity limitations or with specific conditions, though as emphasized by Salkever (1982a, b),
Powers (2003) and Gould (2004), results may be sensitive to the definition of child disability employed,
since both time and financial demands will vary with the nature and severity of the disability. Observed
patterns of specialization could differ depending upon whether the child’s health problem is more
demanding of time or money (Gould 2004).
P. Burton et al.
123
is so, then we are probably under-stating some of the potential negative
consequences of parenting a child with a disability. Evidence on this point appears
mixed. Hoddap and Krasner (1994); Lehrer (2003) and Mauldron (1992) find
evidence of reduced family functioning and/or increased probability of divorce for
families with a disabled child; whereas, Haven (2005) and Seltzer et al. (2001) do
not.
After excluding children with missing information for any analysis variable, our
basic sample consists of 40,656 observations.
All estimates are carried out using longitudinal survey weights. Since some
children can appear twice and/or siblings can be present in the data, standard errors
are adjusted to take account of the non-independence of these observations (i.e.,
clustering at the household level).
3.1 Descriptive statistics for ‘contemporaneous’ sample
Table 1 provides a first indication of the associations between parental labor market
activity and child functional status. Unconditionally, mothers15 of children with
disabilities, are slightly less likely to engage in paid work than mothers of children
with no reported disabilities (82 % compared to 84 %—a small but statistically
significant difference); mean current weekly hours are also lower (27.5 compared to
28.2). There are no statistically significant differences in participation or paid hours
for fathers of the same children.
Finally, for each family we construct a measure of the difference between father
and mother paid hours (father hours minus mother hours) and find that this
difference is higher for families with a disabled child (15.3 vs 13.8).
Table 1 Parental paid work by contemporaneous child disability statusa
Child without
activity limitation
Child with activity
limitation
Mothers
In labor force participation 0.84 (0.003) 0.82 (0.01)
Usual weekly hours (including zeros) 28.15 (0.13) 27.49 (0.53)
Fathers
In labor force 0.96 (0.002) 0.96 (0.005)
Usual weekly hours (including zeros) 43.01 (0.09) 43.14 (0.31)
Difference in usual weekly parental paid
hours (father minus mother, including zeros)
13.78 (0.17) 15.32 (0.62)
Number of observations 38,186 2,470
Standard errors are in parentheses
15 Recall that we only study children with two parents. If the pmk is female, we code her as the mother
and her spouse as the father. If the pmk is male, we code him as the father and his spouse as the mother.
We did not identify any same-sex couples in the data. .
Child health and parental paid work
123
4 ‘Contemporaneous’ models
We first use probit models to estimate the probability that the mother is engaged in
paid work and GLM models with gamma error distributions for her usual weekly
paid hours.16 Our key explanatory variable is the dummy indicating that the child
has a disability. Consistent with other empirical studies in this area (e.g., Powers
2003; Gould 2004), we control for personal characteristics of the mother as reported
in the same cycle as the labor force variables. These include: age (mean of 39) and
age squared; education level (52.9 % have at least some post-secondary education);
immigrant status (15.2 % are born outside of Canada); own self-assessed health
status (19.6 % have an activity limitation, chronic condition, or fair/poor health).
We also control for family characteristics likely to affect a mother’s reservation
wage: father’s health status (18.3 % have an activity limitation, chronic condition,
or fair/poor health status); and number of children (mean of 2.3). Finally, we control
for whether the family resides in a rural area (14.5 %) and for the provincial
unemployment rate (mean of 7.6 %), both likely indicative of local employment
opportunities.
The same models of participation and hours are estimated for fathers (substituting
‘mother’ for ‘father’ as appropriate); and, OLS models of differences in parental
paid hours are estimated, again using the same explanatory variables. Means and
frequencies of control variables are reported in Table 2.
Estimation results for participation in paid work are reported in Table 3;
estimates for weekly hours are reported in Table 4; and, estimates for differences
between father and mother hours are reported in Table 5. In each case, we estimate
three model specifications, a basic model [column (1)] that controls only for the
child’s functional status and survey year fixed effects, an enhanced model [column
(1)] that additionally controls for parental and family characteristics except for self-
reported health status and a most comprehensive model [column (3)] that includes
everything. Although controlling for own and spouse health is common in the this
literature, we report separate estimates for models that include these variables given
that parental health is also be affected by child disability (see, for example, Burton
et al. 2008a, b).
For mothers, both participation in paid work and weekly hours of paid work are,
other things equal, lower for mothers of children with activity limitations.17 For
fathers of the same children, no statistically significant association between child
health and labor market behavior is evident. Finally, the difference between father
and mother hours is larger if the child is activity limited.
16 Given the non-negative and right skewed nature of the parental paid hour variable, the GLM models
with a gamma distribution appear most appropriate. We also estimated Tobit models. Despite yielding
qualitatively similar results, the Tobit assumption of normality is strongly rejected by the Lagrange
multiplier test (at 1 % significance level) in all cases. Since there is no clear theoretically best choice for
the link function, we follow Hardin and Hilbe (2012) and used a power analysis to determine the optimal
link. The result suggests that the preferred one is the canonical inverse (power = -1). We thus present
results only from inverse-gamma models in the paper, although they are generally similar among different
links such as the log (power = 0).17 The coefficient on child disability in the labor force participation model for mothers is, however, no
longer statistically significant when we control for mother’s health status.
P. Burton et al.
123
5 Estimation of ‘onset’ models
There are several potential criticisms of contemporaneous estimates of the
association between child disability and parental labor market behavior. First, child
disability status in any given year mingles together children who have always had a
disability with children who have just developed a problem. Thus, a first criticism is
that labor market implications may be different during the initial adjustment period
than over the longer-term (and it isn’t obvious which would be larger—daycare for a
special needs child might be found with enough time, for example, but the parent
might ‘burn out’ and feel less able to continue with both paid work and care-giving).
Also, since some studies suggest that women may increase labor supply prior to a
divorce (e.g., Johnson and Skinner 1986), and, as noted earlier, child disability has
been found to increase the probability of divorce, this may also serve to obscure the
connection between child disability and labor supply in a cross-sectional analysis.
Since our data set does not provide health histories from birth for most children,
we cannot know when a currently existing problem began. It is thus not an option to
study labor market implications of complete child health histories. We can,
however, start with a sample of children who all have a ‘clean bill of health’ and
trace the implications for parental labor market activity of a health problem that first
emerges during our study period. We refer to this as our ‘onset’ model.
A second limitation of estimates of contemporaneous associations between child
health and parental labor market behavior is that we know there are likely to be
unobservable differences between parents in degree of career motivation,
Table 2 Means for independent variables
All
children
No activity
limitation
Activity
limited
Child’s age 9.7 9.7 10.9
Mother’s age 39 38.8 40
Father’s age 41.1 41.0 42.1
Mother has post-secondary diploma or university degree (%) 52.9 52.7 56.1
Father has post-secondary diploma or university degree (%) 52.2 52.0 55.5
Mother has an activity limitation, chronic condition or fair/
poor health status (%)
19.6 18.6 35
Father has an activity limitation, chronic condition or fair/
poor health status (%)
18.3 17.6 28.1
Pmk is an immigrant (%) 15.2 15.6 9.9
Father’s total earnings (2008 constant dollars) 59,718 59,872 57,363
Mother’s total earnings (2008 constant dollars) 30,979 31,112 29,002
Number of children in the family 2.3 2.3 2.3
Rural residence (%) 14.5 14.4 15.3
Provincial unemployment rate (%) 7.6 7.6 7
Number of observations 40,656 38,186 2,470
Contemporaneous sample
Child health and parental paid work
123
Ta
ble
3M
argin
alef
fect
sfr
om
pro
bit
model
sof
par
enta
lla
bor
mar
ket
par
tici
pat
ion
Moth
erF
ather
(1)
(2)
(3)
(1)
(2)
(3)
Dum
my=
1if
chil
d
has
adis
abil
ity
-0.0
243*
(0.0
88)
-0.0
310**
(0.0
34)
-0.0
187
(0.1
81)
0.0
00192
(0.9
78)
0.0
00494
(0.9
35)
0.0
0375
(0.4
49)
Moth
er’s
age
0.0
226***
(0.0
05)
0.0
223***
(0.0
05)
0.0
0620*
(0.0
86)
0.0
0610*
(0.0
57)
Moth
er’s
age
squar
ed-
0.0
00288***
(0.0
05)
-0.0
00285***
(0.0
06)
-0.0
000821*
(0.0
70)
-0.0
000832**
(0.0
37)
Fat
her
’sag
e0.0
145**
(0.0
38)
0.0
139**
(0.0
48)
0.0
131***
(0.0
00)
0.0
118***
(0.0
00)
Fat
her
’sag
esq
uar
ed-
0.0
00212***
(0.0
10)
-0.0
00203**
(0.0
14)
-0.0
00175***
(0.0
00)
-0.0
00152***
(0.0
00)
Moth
erhas
post
-sec
ondar
y
dip
lom
aor
univ
ersi
ty
deg
ree
0.0
850***
(0.0
00)
0.0
828***
(0.0
00)
0.0
0692
(0.1
62)
0.0
0570
(0.1
87)
Fat
her
has
post
-sec
ondar
y
dip
lom
aor
univ
ersi
ty
deg
ree
0.0
0446
(0.6
43)
0.0
0353
(0.7
10)
0.0
199***
(0.0
00)
0.0
166***
(0.0
00)
Pm
kis
anim
mig
rant
-0.0
457**
(0.0
49)
-0.0
448*
(0.0
50)
-0.0
108
(0.0
93)
-0.0
103
(0.1
10)
Rura
lre
siden
ce-
0.0
234**
(0.0
41)
-0.0
253**
(0.0
28)
0.0
0766*
(0.0
71)
0.0
0633*
(0.0
83)
Num
ber
of
chil
dre
nin
the
fam
ily
-0.0
415***
(0.0
00)
-0.0
428***
(0.0
00)
-0.0
0415*
-0.0
0379*
(0.0
78)
Chil
d’s
age
0.0
0726***
(0.0
00)
0.0
0755***
(0.0
00)
0.0
0233***
(0.0
01)
0.0
0234***
(0.0
00)
Pro
vin
cial
unem
plo
ym
ent
rate
-0.0
0866***
(0.0
00)
-0.0
0892***
(0.0
00)
-0.0
0335***
(0.0
00)
-0.0
0315***
(0.0
00)
Hea
lth
stat
us
of
moth
er-
0.0
938***
(0.0
00)
-0.0
0234
(0.6
35)
Hea
lth
stat
us
of
fath
er-
0.0
137
(0.2
70)
-0.0
675***
(0.0
00)
N40,6
56
40,6
56
40,6
56
40,6
56
40,6
56
40,6
56
Conte
mpora
neo
us
chil
ddis
abil
ity
stat
us
(1)
Mar
gin
alef
fect
s(i
.e.
inst
anta
neo
us
rate
sof
chan
ge
for
conti
nuous
var
iable
san
ddis
cret
ech
ange
of
dum
my
var
iable
from
0to
1)
from
pro
bit
regre
ssio
ns
are
report
ed;
(2)
two-
tail
edp
val
ues
inpar
enth
eses
wit
hS
Ecl
ust
ered
atth
ehouse
hold
level
.(3
)C
ycl
ean
dco
hort
dum
mie
sar
ein
cluded
but
not
report
ed
*p
\0.1
,**
p\
0.0
5,
***
p\
0.0
1
P. Burton et al.
123
attachment to paid work, etc. In addition to sorting out the history/dynamics of the
child’s health condition, a key advantage of the ‘onset’ estimation approach is that
we are able to control for parental labor market activity before the child’s health
status falls, thus helping us deal with the issue of unobserved heterogeneity.
Finally, estimates using only contemporaneous information might be criticized
insofar as values for some control variables may, in part, reflect responses to the
child’s health condition. For example, families may have moved from the country to
the city in order to be nearer to needed specialists/therapists; parents’ own health
may have been compromised. The onset estimation approach allows us to set all
control variables at their ‘pre-child disability’ values.
To estimate the impact of the onset of child health problems on the labor market
behavior of their parents, we again pool panels of data constructed from cycles 1–8.
Sample selection criteria are as previously described with the addition that we now
select only children who had no reported disability at the first observation.
Although, as outlined above, they solve some technical problems, one important
limitation of the onset models is thus that we are excluding some of the most
seriously disabled children from our sample (e.g., those who have had disabilities
from birth or early life).
After the additional exclusion, we have a sample of 32,767 children who did not
have an activity limitation at the first observation.18
5.1 Descriptive statistics
From our sample of children in married-couple families who were healthy at the first
observation, 4.4 % experienced the onset of disabilities between the first and second
observations (i.e., an activity limitation not reported at the first observation appeared
at the second). For children who remained healthy, 87 % of mothers engaged in
some paid work compared to 79 % of mothers whose children developed health
problems (see Table 6). Mother’s current weekly hours are also lower following a
fall in child health status, 29.8 h per week compared to 27.3 h. For fathers, a slight
increase in the probability of engaging in paid work is evident (97 % compared to
96 % were engaged in paid work,); mean hours are also slightly higher for the
sample where child health status declined (43.5 h per week compared to 43 h per
week).
18 It is possible that the incidence of child disability is not a ‘random event’ that is equally likely to
happen to any child in the population. For example, if reductions in health status are more likely in rural
areas and labor force participation is also lower in rural areas, then we might observe an association
between incidence of child disability and low rates of participation without necessarily any causal
connection. To help address this concern, we use a ‘propensity score reweighting’ technique (Rosenbaum
1987; Hirano and Imbens 2001) that involves constructing a scalar weight based on estimated propensity
scores to create a ‘balanced’ sample in order to compare the labor market behavior of parents of children
with disabilities to parents whose children remain healthy but are otherwise as similar as possible in terms
of other observable characteristics. Results obtained using propensity score reweighting are qualitatively
very similar to the onset results reported here. They are available in an earlier version of the paper,
available on request.
Child health and parental paid work
123
Ta
ble
4M
argin
alef
fect
sfr
om
GL
M-g
amm
am
odel
sof
par
enta
lw
eekly
pai
dhours
Mo
ther
Fat
her
(1)
(2)
(3)
(1)
(2)
(3)
Du
mm
y=
1if
chil
d
has
ad
isab
ilit
y
-1
.735
**
*(0
.00
4)
-1
.730
**
*(0
.00
2)
-1
.45
1*
**
(0.0
10)
0.2
63
(0.5
04)
-0
.039
4(0
.91
4)
0.1
83
(0.6
19)
Mo
ther
’sag
e1
.372
**
*(0
.00
9)
1.3
41
**
(0.0
11)
-0
.134
(0.7
55)
-0
.10
6(0
.80
0)
Mo
ther
’sag
esq
uar
ed-
0.0
18
3*
**
(0.0
07)
-0
.01
79
***
(0.0
00)
0.0
00
94
0(0
.86
7)
0.0
00
49
1(0
.92
9)
Fat
her
’sag
e0
.903
*(0
.06
2)
0.9
08
*(0
.06
0)
2.5
72
**
*(0
.00
0)
2.5
13
**
*(0
.00
0)
Fat
her
’sag
esq
uar
ed-
0.0
12
6*
*(0
.03
1)
-0
.01
26
**
(0.0
30)
-0
.033
0*
**
(0.0
00
)-
0.0
32
0*
**
(0.0
00)
Mo
ther
has
po
st-s
eco
nd
ary
dip
lom
ao
ru
niv
ersi
tyd
egre
e
4.0
05
**
*(0
.00
0)
3.9
33
**
*(0
.00
0)
0.4
30
(0.1
31
)0
.405
(0.1
45)
Fat
her
has
po
st-s
eco
nd
ary
dip
lom
ao
ru
niv
ersi
tyd
egre
e
-0
.707
*(0
.09
8)
-0
.70
2*
(0.0
97
)0
.90
0*
**
(0.0
03)
0.7
91
**
*(0
.00
7)
Pm
kis
anim
mig
ran
t0
.247
(0.7
92
)0
.259
(0.7
80)
-0
.488
(0.3
38)
-0
.46
4(0
.36
5)
Rura
lre
sid
ence
-0
.681
(0.1
94)
-0
.71
4(0
.17
0)
1.3
60
**
*(0
.00
0)
1.3
22
**
*(0
.00
0)
Nu
mb
ero
fch
ild
ren
inth
efa
mil
y-
2.9
24
**
*(0
.00
0)
-2
.95
0*
**
(0.0
00)
-0
.281
(0.1
31)
-0
.28
1(0
.12
5)
Chil
d’s
age
0.3
25
**
*(0
.00
0)
0.3
33
**
*(0
.00
0)
0.2
58
**
*(0
.00
0)
0.2
58
**
*(0
.00
0)
Pro
vin
cial
un
emp
loy
men
tra
te-
0.1
42
*(0
.07
3)
-0
.15
1*
(0.0
57
)-
0.4
21
**
*(0
.00
0)
-0
.42
7*
**
(0.0
00)
Hea
lth
stat
us
of
mo
ther
-2
.33
4*
**
(0.0
00)
-0
.17
0(0
.63
9)
Hea
lth
stat
us
of
fath
er-
0.0
92
2(0
.84
3)
-3
.33
0*
**
(0.0
00)
N4
0,6
56
40
,65
64
0,6
56
40
,65
64
0,6
56
40
,65
6
Co
nte
mp
ora
neo
us
chil
dd
isab
ilit
yst
atu
s
(1)
Mar
gin
alef
fect
s(i
.e.
inst
anta
neo
us
rate
so
fch
ang
efo
rco
nti
nu
ou
sv
aria
ble
san
dd
iscr
ete
chan
ge
of
du
mm
yv
aria
ble
from
0to
1)
fro
mg
amm
are
gre
ssio
ns
are
report
ed;
(2)
two
-tai
led
pv
alu
esin
par
enth
eses
wit
hst
andar
der
rors
clu
ster
edat
the
ho
use
ho
ldle
vel
.(3
)C
ycl
ean
dco
ho
rtd
um
mie
sar
ein
clu
ded
bu
tn
ot
rep
ort
ed
*p
\0
.1,
**
p\
0.0
5,
**
*p\
0.0
1
P. Burton et al.
123
5.2 Multivariate results for onset models
Using our sample of children in married-couple families who, at the first observation,
did not have any reported disabilities, we again estimate probit models of the
probability that the mother (father) engaged in any paid work as well as GLM-gamma
models of weekly hours of paid work at the third observation. The key explanatory
variable is now a dummy equal to one if the child developed an activity limitation
between the first and second observations and we control both for whether or not the
parent engaged in paid work at the first observation and how many hours she/he
worked per week (before any child disabilities were apparent). Other explanatory
variables are as described above, except that we use characteristics of the parent,
family and region at the first rather than the second observation (means are reported in
Table 7). We also additionally control for spouse’s earnings at the first observation.
Marginal effects for estimated probit models of labor force participation are
reported in columns 2 and 3 of Table 8. We find that married mothers of children
who develop disabilities between the first and second observations are less likely to
be engaged in paid work, controlling for both labor market participation and weekly
Table 5 OLS models of difference between usual weekly hours of parents (father minus mother)
Difference in weekly hours
(1) (2) (3)
Dummy=1 if child has
a disability
2.050*** (0.004) 1.830** (0.010) 1.732** (0.016)
Mother’s age -1.212** (0.038) -1.153** (0.049)
Mother’s age squared 0.0154** (0.039) 0.0146* (0.051)
Father’s age 1.221*** (0.007) 1.165** (0.011)
Father’s age squared -0.0148*** (0.006) -0.0139*** (0.009)
Mother has post-secondary
diploma or university degree
-3.703*** (0.000) -3.668*** (0.000)
Father has post-secondary
diploma or university degree
1.751*** (0.001) 1.644*** (0.002)
Pmk is an immigrant -0.775 (0.500) -0.757 (0.511)
Rural residence 2.018*** (0.001) 2.032*** (0.001)
Number of children in the family 2.441*** (0.000) 2.474*** (0.000)
Child’s age -0.0694 (0.389) -0.0775 (0.332)
Provincial unemployment rate -0.272*** (0.003) -0.270*** (0.004)
Health status of mother 2.472*** (0.000)
Health status of father -3.252*** (0.000)
N 40,656 40,656 40,656
Contemporaneous Child Disability Status
(1) Regression coefficients from OLS regressions are reported; (2) two-tailed p values in parentheses with
SE clustered at the household level. (3) Cycle and Cohort dummies are included but not reported
* p \ 0.1, ** p \ 0.05, *** p \ 0.01
Child health and parental paid work
123
hours prior to the onset of the child’s condition.19 For married fathers, there is no
association between current labor force participation and the onset of health
problems for the child.
Results for estimated GLM-gamma models of weekly paid hours are reported in
columns 4 and 5 of Table 8, for married mothers and fathers, respectively. As was
also true for the contemporaneous models, for mothers we find that weekly paid
hours are lower, controlling for baseline hours, if a child develops a disability
between the first and second observations. Moreover, the size of this effect is
relatively large if we use having an additional child as a basis of comparison.
Specifically, the onset of an activity limitation is associated with a 2.2 % point
reduction in the probability that a mother will participate in paid work, versus a
3.1 % reduction with the addition of an additional child. Hours of work are
estimated to fall by 1.1 h if the child develops an activity limitation compared to
1.3 h for the addition of another child to the family.
For fathers of the same children, we do not find statistically significant
associations between paid work and the onset of a child’s activity limitation.
Column 6 of Table 8 reports results for estimated OLS model of differences in
weekly paid hours between married mothers and fathers, which is larger (about 2 h)
if the child develops an activity limitation, versus a 1.6 h increase in the gap when a
new child appears.
This is not consistent with our theoretical prediction that fathers of children who
develop disabilities are likely to increase specialization in market production.
Table 6 Parental paid work by child disability onset
No child disability develops
in the 2nd observation
Child disability develops
in the 2nd observation
Mothers
In labor force (%) 0.87 (0.004) 0.79 (0.019)
Usual weekly hours (including zeros) 29.79 (0.16) 27.25 (0.80)
Fathers
In labor force (%) 0.96 (0.002) 0.97 (0.006)
Usual weekly hours (including zeros) 42.98 (0.10) 43.47 (0.38)
Difference in hours (including zeros) 13.42 (0.19) 15.32 (0.89)
Number of observations 31,310 1,456
No children with disabilities in first period
Standard errors in parentheses
19 Although some authors have found larger negative impacts for lower-income mothers (Breslau et al.
1982; Salkever 1982b), we find no difference in effect for mothers with high school or less education,
controlling for prior labor market behavior (i.e., the interaction between low education and onset of child
disability is statistically insignificant). Also, we find no statistically significant difference for older versus
younger children (whereas Salkever 1982b found smaller associations for younger children using cross-
sectional US data). This may reflect higher rates of labor force participation for women with young
children in the late 1990’s and 2000’s.
P. Burton et al.
123
However, almost all the fathers in our sample already do many hours of paid work
per week regardless of child functional status. For example, 83 % of fathers in our
sample work more than 40 h per week before the onset of the child’s health
problem; indeed, 30 % already work more than 50 h per week. Thus, it may difficult
for them to obtain (or do) more hours. The key point is perhaps simply that fathers
do not appear to have lower participation in paid work when there is a child with a
disability in the family.
Our results of ‘no change’ for Canadian fathers are consistent with Salkever
(1982b) though not with Noonan et al. (2005). A potential explanation for the
difference between our findings and those of Noonan, Reichman and Corman could
be that they use the US ‘Fragile Families’ survey with a high proportion of younger
unwed fathers who may have more marginal attachment to the labor force than a
nationally representative sample of married Canadian fathers with strong labor force
attachment.
Table 7 Means for independent variables
Average No activity
limitation
Activity
limited
Child’s age 8.0 7.9 8.9
Mother’s age 37.3 37.2 38
Father’s age 39.5 39.5 40
Education of mother post-secondary diploma
or university degree %
46.4 46.6 42.4
Education of father post-secondary diploma
or university degree %
47.1 47.3 42.6
Mother has an activity limitation, chronic
condition or fair/poor health status %
21.8 21.5 27.4
Father has an activity limitation, chronic
condition or fair/poor health status %
20.4 20.3 22.5
Pmk is an immigrant % 15.1 15.3 11.8
Father’s total earnings (2008 constant dollars) $49,494 $49,381 $51,908
Mother’s total earnings (2008 constant dollars) $25,824 $25,782 $26,732
Number of children in the family 2.3 2.3 2.4
Change in the number of children -0.03 -0.03 -0.05
Rural residence % 14.5 14.5 14.7
Provincial unemployment rate % 8.2 8.2 7.5
Change in provincial unemployment rate -0.87 -0.88 -0.55
Dummy=1 if mother is in the labor force 82.6 82.7 79.1
Mother’s usual weekly hours of paid work 28.1 28.2 27.3
Dummy=1 if father is in the labor force 97.0 97.0 97.2
Father’s usual weekly hours of paid work 43.2 43.2 43
Number of observations 32,767 31,455 1,312
First period. Onset sample
All level covariates measured at the baseline year
Child health and parental paid work
123
Ta
ble
8P
aren
tal
lab
or
mar
ket
par
tici
pat
ion
and
the
on
set
of
chil
dd
isab
ilit
y
Mo
ther
LF
PF
ath
erL
FP
Mo
ther
ho
urs
Fat
her
ho
urs
Dif
fere
nce
inh
ou
rs
Du
mm
y=
1if
chil
dd
evel
op
s
dis
abil
ity
-0
.021
7*
(0.0
81)
0.0
01
64
(0.7
31)
-1
.08
0*
**
(0.0
09)
0.1
13
(0.7
45
)1
.720
**
*(0
.00
4)
Du
mm
y=
1if
moth
eris
inth
e
lab
or
forc
e(b
asel
ine)
0.2
41
**
*(0
.00
0)
12
.57
**
*(0
.00
0)
Mo
ther
’sw
eek
lyh
ou
rso
f
pai
dw
ork
(bas
elin
e)
0.0
02
46
**
*(0
.00
0)
0.3
00
**
*(0
.00
0)
Du
mm
y=
1if
fath
eris
inth
e
lab
or
forc
e(b
asel
ine)
0.1
05
**
*(0
.00
7)
16
.64
**
*(0
.00
0)
Fat
her
’sw
eek
lyh
ou
rso
fp
aid
wo
rk(b
asel
ine)
0.0
01
04
**
*(0
.00
0)
0.4
62
**
*(0
.00
0)
Dif
fere
nce
inm
oth
er’s
ver
sus
fath
er’s
wee
kly
ho
urs
(bas
elin
e)
0.4
25
**
*(0
.00
0)
Fat
her
’sto
tal
earn
ings
(20
08
con
stan
td
oll
ars)
(bas
elin
e)
-5
.30e-
08
(0.5
42)
-0
.00
00
086
0(0
.13
4)
0.0
00
03
43
**
*(0
.00
0)
Mo
ther
’sto
tal
earn
ings
(20
08
con
stan
td
oll
ars)
(bas
elin
e)
9.9
5e-
08
*(0
.05
6)
0.0
00
00
13
3(0
.59
2)
-0
.00
00
786
**
*(0
.00
0)
Nu
mb
ero
fch
ild
ren
inth
e
fam
ily
(bas
elin
e)
-0
.009
44
**
(0.0
27)
0.0
00
23
1(0
.88
8)
-0
.67
5*
**
(0.0
00)
-0
.047
0(0
.70
7)
0.6
20
**
*(0
.00
7)
Chan
ge
inn
um
ber
of
chil
dre
n-
0.0
31
0*
**
(0.0
00)
0.0
00
14
2(0
.96
4)
-1
.32
4*
**
(0.0
00)
-0
.096
0(0
.68
7)
1.6
13
**
*(0
.00
0)
Mo
ther
’sag
e(b
asel
ine)
0.0
04
37
(0.4
91
)0
.00
76
3**
(0.0
11)
0.0
52
5(0
.87
6)
0.0
47
1(0
.90
1)
0.2
61
(0.6
09)
Mo
ther
’sag
esq
uar
ed(b
asel
ine)
-0
.000
07
41
(0.3
70)
-0
.000
09
52
**
(0.0
17)
-0
.00
17
6(0
.69
1)
-0
.000
60
3(0
.90
8)
-0
.00
13
6(0
.84
1)
Fat
her
’sag
e(b
asel
ine)
0.0
10
2*
*(0
.03
2)
0.0
02
07
(0.2
72)
0.8
16
**
*(0
.00
4)
1.5
66
**
*(0
.00
0)
0.2
49
(0.4
25)
Fat
her
’sag
esq
uar
ed(b
asel
ine)
-0
.000
14
0*
*(0
.01
2)
-0
.000
04
49
**
(0.0
29)
-0
.01
08
***
(0.0
02)
-0
.021
8*
**
(0.0
00
)-
0.0
04
45
(0.2
36
)
Mo
ther
has
po
st-s
eco
nd
ary
dip
lom
ao
ru
niv
ersi
ty
deg
ree
(bas
elin
e)
0.0
25
1*
**
(0.0
01
)0
.00
44
7(0
.15
3)
1.0
29
**
*(0
.00
0)
0.2
17
(0.3
07
)-
0.6
16
(0.1
49
)
P. Burton et al.
123
Ta
ble
8co
nti
nu
ed
Mo
ther
LF
PF
ath
erL
FP
Mo
ther
ho
urs
Fat
her
ho
urs
Dif
fere
nce
inh
ou
rs
Fat
her
has
po
st-s
eco
nd
ary
dip
lom
ao
ru
niv
ersi
ty
deg
ree
(bas
elin
e)
0.0
08
50
(0.2
78
)0
.00
63
4**
(0.0
40)
-0
.23
8(0
.33
0)
0.3
08
(0.1
26
)0
.505
(0.2
00)
Pm
kis
anim
mig
ran
t(b
asel
ine)
-0
.021
9(0
.13
2)
-0
.003
82
(0.3
95
)0
.663
(0.1
30)
-0
.024
7(0
.94
8)
-0
.74
4(0
.31
2)
Rura
lre
sid
ence
(bas
elin
e)-
0.0
11
7(0
.12
1)
0.0
04
67
(0.1
03)
-0
.46
1*
(0.0
88)
0.8
77
**
*(0
.00
0)
1.3
79
**
*(0
.00
1)
Chil
d’s
age
(bas
elin
e)0
.002
29
**
(0.0
26
)0
.00
13
5**
*(0
.00
5)
0.0
78
9*
*(0
.03
3)
0.1
94
**
*(0
.00
0)
0.0
71
3(0
.27
6)
Pro
vin
cial
un
emp
loy
men
t
rate
(bas
elin
e)
-0
.002
53
*(0
.06
5)
-0
.000
84
2(0
.10
0)
0.0
57
8(0
.17
4)
-0
.172
**
*(0
.00
0)
-0
.24
8*
**
(0.0
01)
Chan
ges
inp
rov
inci
al
un
emp
loy
men
tra
te
0.0
02
16
(0.6
05
)-
0.0
00
02
62
(0.9
87
)-
0.0
50
9(0
.71
5)
-0
.219
**
(0.0
35)
-0
.29
2(0
.17
9)
Hea
lth
stat
us
of
mo
ther
(bas
elin
e)
-0
.031
7*
**
(0.0
01)
-0
.002
44
(0.4
92
)-
0.6
17
**
(0.0
15)
-0
.121
(0.5
92)
0.6
84
(0.1
37)
Hea
lth
stat
us
of
fath
er
(bas
elin
e)
-0
.012
1(0
.16
4)
-0
.020
1*
**
(0.0
00
)0
.030
2(0
.90
5)
-1
.449
**
*(0
.00
0)
-1
.72
6*
**
(0.0
00)
N3
2,7
67
32
,76
73
2,7
67
32
,76
73
2,7
67
(1)
Mar
gin
alef
fect
s(i
.e.
inst
anta
neo
us
rate
so
fch
ang
efo
rco
nti
nu
ou
sv
aria
ble
san
dd
iscr
ete
chan
ge
of
du
mm
yv
aria
ble
fro
m0
to1
)fr
om
pro
bit
and
gam
ma
regre
ssio
ns;
reg
ress
ion
coef
fici
ents
from
OL
Sre
gre
ssio
ns
are
rep
ort
ed;
(2)
Tw
o-t
aile
dp
val
ues
inp
aren
thes
esw
ith
stan
dar
der
rors
clu
ster
edat
the
ho
use
ho
ldle
vel
.(3
)C
ycl
ean
dco
ho
rt
du
mm
ies
are
incl
ud
edb
ut
no
tre
po
rted
*p
\0
.1,
**
p\
0.0
5,
**
*p\
0.0
1
Child health and parental paid work
123
Ta
ble
9E
stim
ates
of
the
Imp
act
of
gen
eral
mea
sure
so
fch
ild
dis
abil
ity
on
lab
or
mar
ket
beh
avio
ro
fm
arri
edm
oth
ers
usi
ng
lon
git
udin
ald
ata
Dis
abil
ity
defi
nit
ion
Sam
ple
Lab
or
forc
e
mea
sure
Est
imat
edim
pac
to
fch
ild
hea
lth
Th
is pap
er
Funct
ional
lim
itat
ion
athom
e,at
chil
dca
re,
atsc
hool
or
in
any
oth
erac
tiv
itie
ssu
chas
tran
spo
rtat
ion
,p
lay,sp
ort
so
r
gam
es,
rela
tiv
eto
oth
erch
ild
ren
of
sam
eag
e
Can
ada,
NL
SC
Y,
19
94
–2
00
8C
hil
dre
n,
0–
17
Usu
al
ho
urs
per
wee
k
last
yea
r
Cro
ss-s
ecti
on
alre
sult
s:3
.3%
pts
less
lik
ely
tob
ein
lab
or
forc
e;2
.3fe
wer
wee
kly
ho
urs
On
set
mo
del
:2
.7%
pt
red
uct
ion
inlf
pan
d
redu
ctio
no
f1
.7in
wee
kly
ho
urs
Po
wer
s
20
03
Defi
nit
ion
2:
chil
dre
nle
ssth
an6
:‘‘
any
lim
itat
ion
sat
all
in
usu
alkin
dof
acti
vit
ies
done
by
most
chil
dre
nof
thei
r
age
bec
ause
of
physi
cal,
lear
nin
gor
men
tal
hea
lth
con
dit
ion
?ch
ild
ren
6–
21
wit
h‘‘
alo
ng
-las
ting
con
dit
ion
that
lim
its
thei
rab
ilit
yto
wal
k,
run
or
use
stai
rs’’
or
ali
mit
atio
nin
the
abil
ity
tod
ore
gu
lar
sch
ool
work
bec
ause
of
‘‘a
physi
cal,
lear
nin
gor
men
tal
hea
lth
con
dit
ion
’’
US
SIP
P,
19
92/9
3ch
ild
ren
0–
21
Usu
al
wee
kly
ho
urs
Cro
ss-s
ecti
on
alre
sult
s:5
.9%
pt
less
lik
ely
tob
ein
lab
or
forc
e;3
.7h
few
erw
eek
lyp
aid
ho
urs
Fix
edef
fect
s:no
signifi
cant
effe
ct(t
hough
un
expec
ted
po
siti
ve
asso
ciat
ion
s)
Go
uld
20
04
28
spec
ific
illn
ess
or
med
ical
con
dit
ions
iden
tifi
ed
thro
ug
h:
‘‘H
asy
ou
rd
oct
or
or
hea
lth
pro
fess
ion
alev
er
said
yo
uch
ild
had
…’’
?‘‘
or
any
oth
eril
lnes
s/
condit
ion?’
’ch
arac
teri
zed
asm
oney
inte
nsi
ve
and
tim
e
inte
nsi
ve
US
19
97
PS
ID,
chil
d
dev
elo
pm
ent
sup
ple
men
t,
chil
dre
n0
–1
2
Av
erag
e
wee
kly
ho
urs
No
sign
ifica
nt
asso
ciat
ion
ifco
nd
itio
no
nly
‘mo
ney
inte
nsi
ve
‘or
on
ly‘t
ime
inte
nsi
ve
(th
oug
hp
osi
tiv
e
coef
fici
ents
);’
17
%p
tlo
wer
pro
bab
ilit
yo
f
par
tici
pat
ion
ifco
ndit
ion
both
tim
ean
dm
oney
inte
nsi
ve
Wee
kly
hours
1.7
low
erif
condit
ion
both
tim
ean
d
mon
eyin
ten
siv
e
Corm
an
etal
.
20
05
‘Po
or
hea
lth
’ch
ild
isco
nsi
der
edto
be
inp
oo
rh
ealt
hif
at
leas
t1
of
the
foll
ow
ing
istr
ue:
chil
dw
eig
hed
\4
po
und
s
atb
irth
;m
oth
erre
po
rted
ap
hy
sica
ld
isab
ilit
yat
12
mon
thfo
llo
w-u
p;
chil
dh
adn
eith
erw
alk
edn
or
craw
led
by
12
mo
nth
s
US
frag
ile
fam
ilie
ssu
rvey
,
sam
ple
of
most
lyu
nw
ed
par
ents
,1
99
8–
200
0;
Ch
ild
ren
age
0–
1.5
8%
po
int
red
uct
ion
inp
rob
abil
ity
moth
eris
emplo
yed
at1
2m
on
thfo
llo
w-u
p(c
on
troll
ing
init
ial
emplo
ym
ent)
;3
.2fe
wer
ho
urs
of
pai
dw
ork
Ple
ase
see
Po
wer
s(2
00
3)
for
asi
mil
ardis
cuss
ion
of
earl
ier
rese
arch
.E
stim
ates
for
single
,sp
ecifi
cco
ndit
ions
are
not
report
edher
e(e
.g.,
Kvis
tet
al.
20
13
for
AD
HD
;
Bay
dar
etal
.2
00
7fo
ras
thm
a)
P. Burton et al.
123
5.3 Canadian compared to US findings
As noted in the introduction, since most earlier research in this area uses US data,
one of our contributions is to provide estimates of the impact of child disability on
the labor supply of married mothers in a different institutional setting. Table 9
presents a summary comparison of key differences in data, methods and results for
several recent studies all of which use general definitions of child disability (rather
than condition-specific) and all of which use longitudinal data. In all cases, the size
of impact is slightly smaller when longitudinal rather than cross-sectional methods
are used. However, for Canadian mothers, results change very little whereas for two
of the US studies, the impact of child disability of mother hours mostly disappears
(or even becomes positive). This could reflect greater increases in income needs for
families of children with disabilities living in the US, both given that some families
will not have adequate health insurance and that families of children with
disabilities do not receive special cash transfers to help offset the additional costs in
the US as they do in Canada. Perhaps fewer US mothers can afford to reduce paid
hours under these circumstances?
6 Discussion and conclusions
Using longitudinal microdata from the Statistics Canada National Longitudinal
Survey of Children and Youth, we find that married mothers of children with
disabilities are less likely to engage in paid work and/or work fewer paid hours per
week than otherwise similar women whose children do not have health problems.
This finding is apparent in contemporaneous estimates of the association between
child health and mother’s labor market behavior and in models of the onset of child
health problems that control for prior labor market attachment of the mother. And,
these results seem large enough to have policy relevance (about three quarters the
size of estimates for the addition of another child to the family).
No statistically significant changes in paid work participation or hours are
apparent for fathers of the same children (though the point estimate for hours is
positive). Not surprisingly, then, we find a statistically significant increase in the
difference between father paid hours and mother paid hours in families where a
child develops a disability.
These results suggest a ‘household’ coping strategy, with increased specialization
according to traditional gender roles. While this probably makes sense for many
families, it is important to keep in mind that such a division of labor may generate
economic vulnerability for mothers compared to fathers. For example, some
household models emphasize relative earnings as important predictors of bargaining
power; the literature of ‘family gap’ demonstrates long-run earnings penalties
associated with labor-market withdrawal for mothers; such role specialization can
leave mothers particularly vulnerable in the event of divorce.
How could policy help? Certainly, more flexible job schedules as well as
childcare/afterschool care programs that accommodate children with special needs
could help parents balance care-giving and paid work. As well, income supplements
Child health and parental paid work
123
for families of children with special needs would help to relieve some of the
financial costs often associated with raising a child with a disability. In Canada, paid
maternity and parental leave to support the care-giving of new mothers and fathers
(including adoptive parents) are available. Since reductions in maternal paid work
associated with the onset of a child’s disability are in the same order of magnitude
as those estimated for women with a new child, we could provide similar benefits
for families caring for children with disabilities such as are now offered in many
European countries (see Gornick and Meyers 2003). Ideally, these benefits would be
available to mothers and fathers and would allow for periodic absences from paid
work to accommodate medical appointments, therapy sessions, etc. However,
regardless of the type of policy adopted, it would be important that it not
inadvertently entrench mothers as care-givers (e.g., a low benefit ceiling would
make effective replacement rates significantly lower for fathers) rather offering
parents the option of sharing bread-winning and care-giving responsibilities.
Acknowledgments We would like to thank both the Canadian Institute for Advanced Research and the
Canadian Institutes for Health Research through the ‘‘Healthy Balance Research Program: A Community
Alliance for Health Research on Women’s Unpaid Caregiving’’ for funding this work.
References
Akerlof, G. A., & Kranton, R. E. (2000). Economics and identity. The Quarterly Journal of Economics,
65(3), 715–753.
Baker, M., Stabile, M., and Deri, C. (2001). What Do Self-Reported, Objective, Measures of Health
Measure? National Bureau of Economic Research Working Paper. 8419.
Baydar, N., Joesch, J. J., Kieckhefer, G., Kim, H., & Greek, A. (2007). Employment behaviors of mothers
who have a child with asthma. Journal of Family Economic Issues, 28, 337–355.
Beagan, B., Stadnyk R., Loppie, C., MacDonald, N., Hamilton-Hinch, B., MacDonald, J. (2005). I do it
because I love her and I care: Snapshopts of the lives of caregivers. www.healthyb.da.ca/.
Becker, G. (1965). A theory of the allocation of time. The Economic Journal, 75(299), 493–517.
Breslau, N., Salkever, D., & Staruch, K. S. (1982). Women’s labor force activity and responsibilities for
disabled children. Journal of Health and Social Behavior, 23(2), 169–183.
Browning, M. (1992). Children and household economic behavior. Journal of Economic Literature,
30(September), 1434–1475.
Burton, P., Lethbridge, L., & Phipps, S. (2008a). Mothering children with disabilities and chronic
conditions: long-Term Implications for self-reported health. Canadian Public Policy, 34(3),
359–378.
Burton, P., Lethbridge, L., & Phipps, S. (2008b). Children with disabilities and chronic conditions and
longer-term parental health. Journal of Socioeconomics, 37(3), 1168–1186.
Burton, P., & Phipps, S. (2009). Economic costs of caring for children with disabilities in Canada.
Canadian Public Policy, 35(3), 269–290.
Chen, Z., & Woolley, F. (2001). A Cournot-Nash model of family decision making. Economic Journal,
111, 722–748.
Corman, H., Noonan, K., & Reichman, N. E. (2005). Mothers’ labor supply in fragile families: The role
of child health. Eastern Economic Journal, 31(4), 601–616.
Gornick, J., & Meyers, M. (2003). Families that work: Policies for reconciling parenthood and
employment. New York: Russell Sage Foundation.
Gould, E. (2004). Decomposing the effects of children’s health on mother’s labor supply: Is it time or
money? Health Economics, 13(October), 525–541.
Gray, D. E. (2003). Gender and coping: The parents of children with high functioning autism. Social
Science and Medicine, 56, 631–642.
P. Burton et al.
123
Hardin, J. W., & Hilbe, J. M. (2012). Generalized linear models and extensions. StataCorp LP: Third
Edition.
Haven, C.A. (2005). Becoming a resilient family: child disability and the family system. Access Today,
17 (Spring). http://www.ncaonline.org (Accessed March 10, 2006).
Hirano, K., & Imbens, G. W. (2001). Estimation of causal effects using propensity score weighting: An
application to data on right heart catheterization. Health Services and Outcomes Research
Methodology, 2(3), 259–278.
Hobbs, N., & Perrin, J. M. (Eds.). (1985). Issues in the care of children with chronic illness. San Franciso:
Jossey-Bass Publishers.
Hodapp, R. M., & Krasner, D. V. (1994). Families of children with disabilities: Findings from a National
Sample of Eighth-Grade Students. Exceptionality, 5(2), 71–81.
Human Resources Development Canada. (2003). Office for Disability Issues. Defining Disability: A
Complex Issue. http://publications.gc.ca/collections/Collection/RH37-4-3-2003E.pdf.
Johnson, W. R., & Skinner, J. (1986). Labor supply and marital separation. The American Economic
Review, 76(3), 455–469.
Kimmel, J. (1997). Reducing the welfare dependence of unmarried mothers: Health-related employment
barriers and policy responses. Eastern Economic Journal, 23(2), 151–163.
Kimmel, J. (1998). Child care costs as a barrier to employment for single and married mothers. Review of
Economics and Statistics, 80(2), 287–299.
Kvist, A. P., Nielsen, H. S., & Simonsen, M. (2013). The importance of children’s ADHD for parents’
relationship stability and labor supply. Social Science and Medicine, 88, 30–38.
Lehrer, E. (2003). The economics of divorce. In S. Grossbard (Ed.), Marriage and the economy (pp.
55–74). Cambridge: University Press.
Leibowitz, A. (1974). Home investments in children. Journal of Political Economy, 82(2), S111–S131.
Lukemeyer, A., Meyers, M., & Smeeding, T. (2000). Expensive children in poor families: Out-of-pocket
expenditures for the care of disabled and chronically ill children in welfare families. Journal of
Marriage and Family, 62(2), 399–415.
Lundberg, S., Pollak, R. A., & Wales, T. (1997). Do husbands and wives pool their resources? Evidence
from the United Kingdom child benefit. Journal of Human Resources, 32, 463–480.
Mauldron, J. (1992). Children’s risks of experiencing divorce and remarriage: Do disabled children
destabilize marriages? Population Studies, 46(2), 349–362.
Meyers, M., Lukemeyers, A., & Smeeding, T. (1998). The cost of caring: Childhood disability and poor
families. The Social Service Review, 72(2), 209–233.
Miller, J. E., Gaboda, D., Davis, D. (2001). Early childhood Chronic Illness: Comparability of Maternal
Reports and Medical Records. Department of Health and Human Services, Centers for Disease
Control and Prevention, National Center for Health Statistics. Publication No. 2001- 1331, Series 2,
No. 131.
Mincer, J. (1962). Labor Force Participation of Married Women: A Study of Labor Supply. in Aspects of
Labor Supply. http://nber.org/books/univ62-2.
Noonan, K., Reichman, N. E., & Corman, H. (2005). New fathers’ labor supply: Does child health matter?
Social Science Quarterly, 86(Supplement), 1399–1417.
Pollak, R. A. (1985). A transaction cost approach to families and households. Journal of Economic
Literature, 13(June), 581–608.
Powers, E. T. (2001). New estimates of the impact of child disability on maternal employment. American
Economic Review Papers and Proceedings, 91(2), 135–139.
Powers, E. T. (2003). Children’s health and maternal work activity: Estimates under alternative disability
definitions. The Journal of Human Resources, 38(3), 522–556.
Rosenbaum, P. R. (1987). Model-based direct adjustment. Journal of the American Statistical
Association, 82(398), 387–394.
Salkever, D. (1982a). Children’s health problems and maternal work status. Journal of Human Resources,
17(1), 94–109.
Salkever, D. (1982b). Children’s health problems: Implications for parental labor supply and earnings. In
Victor Fuchs (Ed.), Economic aspects of health (pp. 221–251). Chicago: University of Chicago
Press.
Salkever, D. (1990). Child health and other determinants of single mothers’ labor supply and earnings. In
Ismael Sivageldin, Alan Sorkin, & Richard Frank (Eds.), Research in human capital and
development 6 (pp. 147–181). London: JAI Press.
Child health and parental paid work
123
Seltzer, M. M,, Greenberg, J. S., Floyd, F. J., Pettee, Y., & Hong, J. (2001). Life course impacts of
parenting a child with a disability. American Journal on Mental Retardation, 106(3), 265–286.
Stabile, M., & Allin, S. (2012). The economic costs of childhood disability. The Future of Children,
22(1), 65–96.
Statistics Canada. (2001). Participation and Activity Limitation Survey. Children with disabilities and
their families. Cat No. 89-585-XIE.
Thyen, U., Kuhlthau, K., & Perrin, J. M. (1999). Employment, child care, and mental health of mothers
caring for children assisted by technology. Pediatrics, 103(6), 1235–1242.
Waldfogel, J. (1998). Understanding the family gap in pay for women with children. Journal of Economic
Perspectives, 12(1), 137–156.
Wolfe, B. L., & Hill, S. C. (1995). The effect of health on the work effort of single mothers. The Journal
of Human Resources, 30(1), 41–62.
P. Burton et al.
123