Applied Research Branch Strategic Policy Human Resources Development Canada

Applied Research Branch Strategic Policy Human Resources Development Canada
Applied Research Branch
Strategic Policy
Human Resources Development Canada
Direction générale de la recherche appliquée
Politique stratégique
Développement des ressources humaines Canada
Economic Resources and Children’s Health and
Success at School
An Analysis Using the NLSCY
W-01-1-4E
by
Lori Curtis and Shelley Phipps
September 2000
The views expressed in Applied Research Branch papers are the authors’ and do not necessarily reflect the opinions
of Human Resources Development Canada or of the federal government.
Les opinions exprimées dans les documents de la Direction générale appliquée sont celles des auteurs et ne reflètent
pas nécessairement le point de vue de Développement des ressources humaines Canada ou du gouvernement
fédéral.
n
The Working Paper Series includes analytical studies and research conducted under the auspices of the Applied
Research Branch of Strategic Policy. Papers published in this series incorporate primary research with an empirical
or original conceptual orientation, generally forming part of a broader or longer-term program of research in
progress. Readers of the series are encouraged to contact the authors with comments and suggestions.
La série des documents de travail comprend des études analytiques et des travaux de recherche réalisés sous l’égide
de la Direction générale de la recherche appliquée, Politique stratégique. Il s’agit notamment de recherches
primaires, soit empiriques ou originales et parfois conceptuelles, généralement menées dans le cadre d’un
programme de recherche plus vaste ou de plus longue durée. Les lecteurs de cette série sont encouragés à faire part
de leurs observations et de leurs suggestions aux auteurs.
This report is part of a set of research studies on the National Longitudinal Survey of Children and Youth. /
Le présent rapport fait partie d’un ensemble d’études sur l’Enquête longitudinale nationale sur les enfants et les
jeunes.
n
Publication date/Date de parution-Internet 2002
ISBN: 0-662-32011-5
Cat. No./No. de cat. MP32-28/01-1-4E-IN
n
General enquiries regarding the documents
published by the Applied Research Branch should
be addressed to:
Si vous avez des questions concernant les documents
publiés par la Direction générale de la recherche
appliquée, veuillez communiquer avec :
HRDC Publication Centre
Human Resources Development Canada
140 Promenade du Portage
Phase IV, Level 0
Hull, Quebec, Canada
K1A 0J9
Centre des publications de DRHC
Développement des ressources humaines Canada
140, Promenade du Portage
Phase IV, niveau 0
Hull (Québec) Canada
K1A 0J9
Facsimile: (819) 953-7260
http://www.hrdc-drhc.gc.ca/arb
Télécopieur : (819) 953-7260
http://www.hrdc-drhc.gc.ca/dgra
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
Abstract
This paper asks whether economic resources matter for children’s outcomes. Both economic
theory and public concern over high levels of child poverty suggest that there is an important
association. Yet, research utilizing the first wave of the National Longitudinal Survey of Children
and Youth (1994) suggested, surprisingly, that low-income status is a relatively unimportant
correlate of children’s outcomes. If true, the policy implication is that income transfers are
relatively unimportant for children.
The goal of this paper is to re-examine the association between economic resources and children’s
health and success at school, two particularly “economic” outcomes (i.e., key elements of
children’s “human capital ”; both have important implications for children’s eventual labour
market success). We move beyond current income and/or poverty status as a measure of the
economic resources available to the child.
Economists would argue that wealth and income flows are also vital components of the economic
resources available to a family. As well, traditional economic reasoning also suggests that, income
constant, families with more time are better off than those with less. When we control for both
housing and available parental time per week, we find that children who live in owner-occupied
housing have better outcomes than children who do not; children who live in housing in need of
major repairs have worse outcomes. This finding represents an additional channel through which
economic resources can influence outcomes for children. Weekly hours of parental time available
has no statistically significant association with child health; however, income constant, more hours
of parental time available each week significantly improves a child’s success at school.
These results indicate that while “longer-term” income is an important factor in child well-being,
measured by health and educational success, other measures of economic resources are also
important. The policy conclusions associated with “low or moderate association between income
and child well-being ” may be misleading. Income transfers may have an additional relationship
with child well-being if they assist families with children to accumulate assets such as housing or if
they increase parental time spent with children. These attributes are associated with better
outcomes for children even after controlling for income. Other policy instruments which may
improve outcomes are extended parental leave, home ownership assistance plans or assisting lowincome families with the completion of housing repairs.
Applied Research Branch
3
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Résumé
L’auteur du présent document se demande si les ressources économiques comptent pour les
résultats des enfants. Tant la théorie économique que les préoccupations du public concernant les
niveaux élevés de pauvreté chez les enfants laissent supposer qu’il existe un lien important.
Cependant, il ressort curieusement de la recherche effectuée en utilisant la première série de
résultats de l’Enquête longitudinale nationale sur les enfants et les jeunes (1994) que le fait d’avoir
un faible revenu est un corrélat relativement sans importance pour les résultats des enfants. Si cela
est vrai, l’incidence sur les politiques générales est que les transferts de revenu sont relativement
sans importance pour les enfants.
L’objectif du présent document est de réexaminer le lien entre les ressources économiques et la
santé et la réussite scolaire des enfants, qui sont deux résultats particulièrement « économiques »
(c.-à-d. des éléments essentiels du « capital humain » des enfants, qui ont tous deux des
conséquences importantes pour la réussite éventuelle des enfants sur le marché du travail). Nous
allons au-delà du revenu actuel et/ou de la pauvreté en tant que mesure des ressources
économiques disponibles à l’enfant.
Les économistes soutiendraient que la richesse et les flux de revenu sont également des
composantes vitales des ressources économiques disponibles à une famille. En outre, la pensée
économique traditionnelle laisse supposer que, à revenu égal, les familles qui disposent de plus de
temps s’en sortent mieux que celles qui en ont moins. Lorsque nous tenons compte du logement
et du temps que les parents consacrent hebdomadairement à leurs enfants, nous constatons que les
enfants qui habitent dans un logement occupé par son propriétaire obtiennent de meilleurs
résultats que les autres; les enfants qui habitent dans un logement nécessitant des réparations
majeures obtiennent les plus mauvais résultats. Cette constatation représente une voie
supplémentaire par laquelle les ressources économiques peuvent influer sur les résultats des
enfants. Il n’existe aucun lien important du point de vue statistique entre le nombre d’heures que
les parents consacrent hebdomadairement à leurs enfants et la santé de ceux-ci; toutefois, à revenu
égal, lorsque les parents consacrent hebdomadairement plus d’heures à leurs enfants, la réussite
scolaire de ceux-ci s’accroît considérablement.
Ces résultats indiquent que, bien que le revenu « à long terme » soit un facteur important du bienêtre de l’enfant, mesuré par son état de santé et sa réussite scolaire, d’autres mesures des
ressources économiques sont tout aussi importantes. Les conclusions politiques selon lesquelles il
existe « un lien faible ou modéré entre le revenu et le bien-être de l’enfant peut être tendancieux.
Les transferts de revenu peuvent avoir un lien supplémentaire avec le bien-être de l’enfant », s’ils
aident les familles avec enfants à accumuler des biens, comme un logement, ou s’ils accroissent le
temps que les parents consacrent à leurs enfants. Ces attributs sont liés aux meilleurs résultats des
enfants, même lorsqu’on tient compte du revenu. Les autres instruments de politique qui peuvent
améliorer les résultats sont le congé parental prolongé, les régimes d’aide à l’accession à la
propriété, ou l’aide aux familles à faible revenu afin de réparer leur logement.
4
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
Acknowledgements
We would like to thank Lynn Lethbridge for her excellent work as research assistant and Peter
Burton, Martin Dooley and Lars Osberg for their extremely helpful comments. Funding for this
research was received from Human Resources Development Canada and is gratefully
acknowledged.
Applied Research Branch
5
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. A Review of Economic Models of the Link Between Economic Resources and
Children’s Well-being . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Review of Empirical Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4. Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5. Multivariate Analysis of the Association Between Child Health and Success at School
and Current Versus Longer-Term Measures of Family Income and Poverty Status . .11
6. Multivariate Analysis of the Association Between Child Health and Success at School
and Economic Resources, Adding Controls for Assets and for Time . . . . . . . . . . . . . . .20
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
Foreword
The National Longitudinal Survey of Children and Youth (NLSCY) is a unique Canadian survey
designed to follow a representative sample of children from birth to early adulthood. It is
conducted in partnership by Human Resources Development Canada (HRDC) and Statistics
Canada. Statistics Canada is responsible for data collection, while HRDC, the major funder,
directs and disseminates research. Data collection began in 1994 and continues at two-year
intervals.
The survey for the first time provides a single source of data for the examination of child
development in context, including the diverse life paths of normal development. The survey and
the research program were developed to support evidence-based policy, using a human
development view of the early decades of life. This research paper is part of an ongoing series of
papers emanating from a program of research that examines NLSCY data collected in the first
two cycles (1994, 1996) of the survey.
Applied Research Branch
7
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
1.
W-01-1-4E
Introduction
Despite the large literature indicating a strong association between low income and the well-being
of adults (see for example Deaton and Paxon, 1999; Kephart 1998; Lantz 1998; Curtis et al. 1998,
Smith et al 1990a,b) the evidence is not as convincing for children. Recent Canadian studies that
investigate the link between current household income or poverty status and child well-being find
relationships that are small in magnitude or even sometimes insignificant (for example, see Curtis
et al. 1998, Dooley et al. 1998a,b). These results appear contrary both to economic theory (e.g.,
Becker, 1974; Haveman and Wolfe, 1995) which clearly makes the case that income is a key input
to children’s well-being, and to general public discourse which has been much concerned with
levels of child poverty in recent years. Understanding the link between income and children’s wellbeing is vital for policy formulation. If it is true that income is a relatively unimportant determinant
of children’s well-being, then the policy implication is that transfer programmes for low-income
children are also relatively unimportant and thus that other forms of policy intervention should be
pursued.
The goal of this paper is to re-examine the hypothesis that economic resources are important
inputs to children’s well-being (specifically, their health and success at school) using the newly
released second wave of the National Longitudinal Survey of Children and Youth (NLSCY). We
argue that, on their own, neither current income nor current poverty status provide a very
accurate measure of the economic resources available to a child. First, since there can be
considerable volatility in income, today’s income does not always provide a good indication of the
average level of resources which have generally been available to the child (e.g., a parent may
have just lost his or her job or gone back to school, etc). Second, it may be that the consequences
of low income for children’s well-being only appear with a lag. In the Becker (1974) framework,
for example, parents are assumed to invest resources in their children today with the expectation
of increased attainment in the future. Third, current income takes no account of available family
assets (e.g., owner-occupied housing). Finally, two families may have the same income, but
earning the income may have taken quite different total amounts of time (e.g., one- versus twoearner families; income from assets versus income from the labour market). According to Becker
8
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
(1974), the two key inputs to children’s attainments provided by parents, aside from genetic
endowment, are income and time. Thus, if we look just at the association between income and
children’s well-being without controlling for the time required to earn the income, we may have an
incomplete picture of the economic resources available for children.1
Section 2 of the paper outlines some theoretical models of the determinants of children’s wellbeing. Section 3 reviews the relevant empirical work. Section 4 discusses the data used (the 2ndwave of the National Longitudinal Survey of Children and Youth, or NLSCY), and provides an
initial descriptive analysis of the two child outcomes upon which the paper is focussed—
children’s over-all health and success at school. Section 5 examines the hypothesis that child wellbeing may be more closely associated with a) longer-term measures of income and poverty status
(i.e., using information from 1994 and 1996); b) lagged income or poverty status. In this section,
we also consider the possibility that income is endogenous to child well-being. Section 6 provides
an empirical examination of two other dimensions of family economic resources: a) home
ownership status and housing characteristics; b) time involved in acquiring income. Section 7
offers some conclusions.
1
Alternative explanations of current findings of the relative empirical unimportance of income for children’s
well-being include: a) income and/or poverty may be measured with error thus biasing the coefficient towards
zero; b) there may be problems of endogeneity (e.g., a lone mother with a seriously ill child may find it hard
to work full-time); c) income and/or poverty may be highly correlated with other socioeconomic variables
typically included in regression models of the determinants of child well-being, leading to low significance
levels (e.g., low income and lone-mother status are very highly correlated). These are extremely valid
concerns but difficult to address when studying socioeconomic relationships with health outcomes in a crosssectional data set. The inconsistencies created by the presence of variables measured with error or endogenous
variables can be remedied by the use of instrumental variables but, given the broad determinants of health,
researchers find it particularly difficult to identify instruments within available health data sets. Longitudinal
data sets just becoming available in Canada, like the National Longitudinal Survey of Children and Youth
(NLSCY), should help researchers address these econometric issues when several waves are available.
Finally, a point not emphasized here, but which we pursue in other research with P. Burton, is that family
resources are not always shared equally among all family members. See, for example, Phipps and Burton,
1995. Thus, it may be true in some families that children receive less than an equal share of family-level
resources; in other families they may receive more.
Applied Research Branch
9
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
2.
W-01-1-4E
A Review of Economic Models of the Link Between Economic
Resources and Children’s Well-being
Becker (1991, chapters 5 and 6) and Becker and Tomes (1979; 1986), for example, assume that
children’s well-being essentially depends upon investment decisions made by their parents.2 This
approach supposes that each individual lives during two periods—childhood and adulthood. The
utility of the parent today is assumed to be a function of own consumption today and child’s
income tomorrow:3
u1 = U(c1, I2)
where c1 is adult consumption today and I2 is child’s income tomorrow, when the child is an adult.
Thus, the parent is assumed to care about the well-being of his/her child as an adult. In the second
period, when the child becomes an adult, his or her utility will be
u2 = U(c2, I3).
Given this framework, Becker assumes that parents allocate resources between personal
consumption today and investment in the future of their children in order to maximize parents’
utility today. Utility maximization occurs subject to the constraint of available income, and the
relative prices of consumer goods versus investment in children. Children’s well-being tomorrow
will depend upon how much parents choose to invest in them today (as well as upon the genetic
and possibly material asset endowments which they may have inherited from their parents and
upon any “pure luck” which they may experience). Investing in children means making
“expenditures on their skills, health, learning, motivation, “credentials,” and many other
characteristics” (Becker and Tomes, 1986, p. S5). The prediction of this framework is that
children’s incomes will depend upon parents’ incomes (positively) and number of other children in
the family (negatively, since additional children mean less money to spend on any one individual
2
Behrman, Pollak and Taubman (1995) for example, adopt a similar perspective. This is the dominant
framework employed in discussion by economists. In fact, we are not familiar with any serious contenders.
3
The focus of theoretical economics literature, and much of the empirical literature, is on outcomes for
children after they grow up rather than upon outcomes for children while they are children. Phipps (1999)
argues that this is not always appropriate.
10
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
child.
Leibowitz (1974) adds the idea that investments in children depend upon both the amount and
quality of time parents spend with them as well as upon material investments. (“Quality” of time
with children is assumed to increase with the education level of the parents.) Thus, parental
choices about, for example, labour supply, will determine both how much money and how much
time is available for children.
The “human capital” idea has been expanded by Coleman (1988) to include “social capital.” Social
capital exists when relationships among persons function as resources which can be used to
“facilitate action” or to “achieve the interests” of the persons involved (pp. S100-101). According
to Coleman, social capital helps to create human capital in the “next generation.” He highlights
roles both for social capital within the family and for social capital outside the family. Within the
family, social capital exists in `the relations between children and parents’ (p. S110). Coleman
argues, for example, that if highly educated parents spend little time with their children, then the
high levels of human capital possessed by the parents can be of little benefit to the children—and
little new human capital will be produced.
Social capital outside the family is defined by Coleman (p. S113) to exist in the relations among
parents (of different children) and in the relations of parents with institutions of the community.
He argues that social capital is likely to be greatest in situations where parents interact with other
parents in a variety of different settings (e.g., at school meetings, in social clubs, at church—all of
which take time).
Thus, the major line of reasoning apparent in the economics literature is that more economic
resources are better for children because they allow for greater investment in human capital.4 But,
time as well as money matters—parents with the same income but less available time will not be
4
Grossman (1972, 1072a) and Grossman et.al. (1989) also argue that additional economic resources will allow
individuals to “produce” more health for themselves, or presumably for their children. These authors
postulate a health production function, which includes income as an argument.
Applied Research Branch
11
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
able to make the same investments in their children. Finally, as with any other “investment” it is
possible that returns are only realized in the future.5
5
12
Non-economists have worked more extensively than economists on the subject of child development and offer
several alternative theoretical perspectives, surveyed by Haveman and Wolfe (1995). These include: 1) the
“socialization/role model perspective” which focuses upon the important influences of parents, siblings and
peers on the development of children’s aspirations, values and behaviour (e.g., Seltzer, 1994; Jencks and
Mayer, 1990); 2) the “ecological systems” approach favoured by many developmental psychologists which
argues that development occurs throughout life, and that the timing and context of any significant life event
(e.g., parental divorce) will modify its impact on that particular individual (e.g., Bronfenbrenner, 1989); 3)
stress theory and coping strategy perspectives argue that a particular stressful event (again, for example,
parental divorce) may change a child’s equilibrium path of development though the impact of such a stressful
event can be mitigated, or not, depending upon parental coping capacities (e.g., Hamilton and McCubbin,
1980). As Haveman and Wolfe argue, these psychological and sociological perspectives emphasize
environmental/cultural factors rather than the individual choices/characteristics upon which economists
focus. Empirically, however, it may not always be easy to distinguish the various perspectives. For example, is
it higher parental income as an input as economists might argue or better role models in the neighbourhood
as sociologists might argue which is the key factor associated with better outcomes for children? Empirically,
these two hypotheses would be very difficult to disentangle (though see Corak and Heisz, 1998).
Applied Research Branch
W-01-1-4E
3.
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
Review of Empirical Literature
The framework discussed above is very clear that income should matter for children’s well-being,
yet as noted in the introduction, Canadian studies using the first wave of the National Longitudinal
Survey of Children and Youth have found small, and sometimes even statistically insignificant
associations, depending upon the measure of income/poverty and the component of children’s
well-being studied (see Curtis, et.al., 1998; Dooley, et al., 1998a,b)
Several excellent studies in the United States have already made the point that “permanent”
income may be a better measure of economic resources than current income, in the context of
studying the role of income as a determinant of children’s well-being (see, especially, Blau 1998,
Korenman et al. 1995, Mayer, 1997). The primary data source for this work has been the National
Longitudinal Survey of Youth’s (NLSY) Mother and Child Supplement which provides very long
income histories. At this stage, although there is consensus that permanent income matters more
than current income, there is disagreement as to the magnitude of the effects.6
Korenman, et al. (1995) interpret their results to indicate a “moderate to large” effect of changes
in long-term poverty status on children’s cognitive development. Mayer (1997) reviews existing
literature and utilizes several different US data sets and methodologies to conclude that the effect
of increases in parental income on child outcomes, ceteris paribus, “is nowhere near as large as
many political liberals imagine, neither is it zero as many political conservatives seem to believe”
(p. 143). She goes on to say that although the effect on any single outcome may be small that most
outcomes seem to be affected by income to some extent, thus increasing income may have a
substantial cumulative impact. Therefore, changes in the distribution of income (increasing income
to the poor) may be as cost-effective as any other policy.
Blau (1999) finds only small effects. In fact he states that the income effects are so small that
income transfers to poor families are likely to have very little impact on child development;
“Policies that affect family income will have little direct impact on child development unless they
result in very large and permanent changes in income.” (P. 261).
6
A more detailed literature review is provided in Curtis and Phipps (2000).
Applied Research Branch
13
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
The limited permanent income and moderate poverty effects found thus far in the US NLSY must
be taken in context. Mothers in the survey are all very young, aged 14 to 21 in 1979. Thus, by the
late 1980s or early 1990s, the time frame for the data used in the studies discussed above, mothers
would be in their mid twenties to early thirties (mean age = 27.7).7 As a consequence of the young
age of the mothers the children tend to be young as well (mean 5.7). The measure of permanent
income used by both Blau (1999) and Korenman, et al. (1995) is an average of income or income
/needs ratio measured over the available 13 years of data. This measure is averaged over the very
early portion of these women’s lifetime-income paths, and it is possible that this is not very
reflective of future earnings capacity and thus permanent income or poverty. As well, of course,
these studies pertain to the United States not Canada.
To date, two longitudinal data sets are available to investigate the association between child health
and well-being and socioeconomic status in Canada. The Ontario Child Health Survey (OCHS)
conducted in 1983 and 1987 and the 1994 and National Longitudinal Survey of Children and
Youth (NLSCY), conducted in 1994 and 1996.
Findings using the OCHS indicate a consistently significant association between low income or
poverty and psychiatric disorders (Offord, Boyle and Jones, 1987), social and educational
functioning (Lipman and Offord 1994), and chronic physical health problems (Cadman et al.,
1986a) in children. Studies using the longitudinal nature of the OCHS find that changes in income
levels are very weakly correlated with changes in the levels of child health (Lipman and Offord
1996, Lipman, Offord and Boyle 1994 and Boyle et al. 1998). The studies that find a significant
relationship between income and child health tend to limit the use of other explanatory variables.
Curtis et al. (2000) investigate the relationship between current low-income and low-averageincome using the OCHS. Fifteen percent of families were poor in 1983 and 13% in 1987.
7
14
The means are taken from Table A-1 of the Blau paper. No summary statistics for the entire sample were
presented in the Korenman et al. paper.
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
13% of children lived with families whose average incomes (between 1983 and 1987) were less
the LICO8 and 7% of the children were from households with income below the LICO in both
years. The study investigated the presence of emotional problems, cognitive problems, any health
problems and an over-all health related quality of life score, the Health Utilities Index Mark 2
(HUI2).
As in the Koreman study, children from low-income families suffered from substantially more
problems than did children from non-low-income families. Although current low-income had no
statistically significant relationship with emotional problems in either 1983 or 1987, living in a
family whose average income for the two years was less than poverty level increased a child’s
probability of having an emotional problem by 9 percentage points (from 44 to 53%) which was
roughly comparable to the lone-mother association. Results were very comparable for HUI2
scores. For cognitive problems, both current and average low-income were negatively associated,
though the effect was larger for average low-income. Curtis et. al. (2000) conclude that, like many
of the NLSY studies, child outcomes are more strongly related to low-average income than lowcurrent income. Contrary to some of the results coming out of the NLSY they find that the
magnitude of income effects to be “large” for some outcomes.
8
See companion paper to Curtis et al. (2000) for description of income measures.
Applied Research Branch
15
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
4.
W-01-1-4E
Data
The data used in this analysis are drawn from Cycle Two of the NLSCY, which is a representative
national sample of children who were aged 0 to 11 years in 1994 and 0 to 13 years in 1996. The
main component of the survey consists of children living in households who had recently been part
of the Labour Force Survey (thus households living in the North, on Indian Reserves or in
institutions are excluded). We use information obtained from the “person most knowledgable
about the child” (or pmk)—the mother in 90 percent of cases. Since we want to have longer-term
measures of poverty status and income, we select only children present in both years of the survey.
Since we rely upon the reports of the pmk, we require this to be the same person for both years.
These two restrictions limit the sample to 12,824 children aged 2 to 13. Since one of our key
outcome variables is “success at school” we must further restrict the sample to children aged 6 to
13 (7577 children). Finally, exclusion of observations with non-response to any questions used in
our analysis results in an estimating sample of 7337 observations.
Although we recognize that “child well-being” is a multi-dimensional construct, we have chosen to
limit our analysis in this paper to just two outcomes—over-all health and success at school. These
two outcomes seem particularly relevant to the economic models of “human capital formation”
discussed above. Moreover, health and success at school are child outcomes with significant
economic implications both in terms of health care and education costs and in terms of the child’s
eventual labour market success.9 Finally, we have chosen to focus on 1996 levels of these child
outcomes (rather than upon changes in child outcomes between 1994 and 1996). We argue that,
substantively, it is a different question to ask why some children have good or bad outcomes than
to ask why some children have changed outcomes. For example, a majority of children in the
NLSCY do well in both periods; some children do very badly in both periods—we want to
understand the correlates of these good or bad outcomes even if there was no change between the
two time periods.10
9
However, see Phipps (1999) who argues that the child’s well-being today as well as their attainments
tomorrow are important if we are concerned about social welfare.
10
Curtis and Phipps (2000) use both 1994 and 1996 child outcomes to look at changes in outcomes as a
function of changes in explanatory variables. However, for the outcomes studied here, in particular, there are
some associated technical difficulties since the outcomes studied have only 5 possible categories and a
majority of respondents are already in the top category (i.e., the only change possible for a majority of the
children we study is a movement down).
16
Applied Research Branch
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Table 1 reports frequencies for our two child outcome measures for 1996. The first of these is a
pmk assessment of the child’s over-all health status. “In general, would you say your child’s health
is: Excellent, Very good, Good, Fair or Poor.” Fortunately, it is clear from Table 1 that a majority
(59.6 percent) of parents rate their children’s health as “excellent;” an additional 28 percent of
children are assessed as having “very good” health. This seems reasonable for a sample of young
children.11 Only 0.2 percent (16 observations) are assessed as having “poor” health.
Our second outcome is a measure of the child’s “success at school:” “Based on your knowledge of
your child’s school work, including his or her report cards, how is your child doing overall?
Possible answers include: “Very Well, Well, Average, Poorly, Very Poorly.” It is again clear from
Table 1 that most parents perceive their children to be very successful at school over-all—46.6
percent of children are ranked as doing “very well” and 25.6 percent are ranked as doing “well.”12
Thus, 72.2 percent of children are apparently doing “better than average!” However, to put these
rather optimistic figures in perspective, only 1 percent of married men and 2 percent of married
women report themselves to be unhappy with life over-all (see Phipps, Burton and Osberg, 2000).
Table 1
Frequencies of General Health and Success at School
Health
Success at School
In general, would you say (your child’s) health is:
Based on your knowledge of (your child’s) school
work, including his/her report cards, how is your
child doing overall ?
Poor
0.2%
Very poorly
0.3%
Fair
1.7%
Poorly
3.1%
Good
10.5%
Average
24.4%
Very good
28.0%
Well
25.6%
Excellent
59.6%
Very well
46.6%
11
It is possible that the choice of the label “average” for the middle category was a bad one. If the middle
category is “average” this suggests a normal distribution for children’s health, which may not be appropriate
if in fact most young children are very healthy.
12
In other work (see Curtis, Dooley and Phipps, 2000), we present evidence of relatively low correlations
between parental and child reports of child outcomes for 10 and 11-year old children. “Success at school” was
the outcome with the highest correlation between child and parent reports. This provides an additional
motivation for choosing this outcome to study. (Children were not asked to assess their own health status.)
Applied Research Branch
17
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
5.
W-01-1-4E
Multivariate Analysis of the Association Between Child Health
and Success at School and Current Versus Longer-Term
Measures of Family Income and Poverty Status
In this section of the paper, we begin our re-examination of the hypothesis that economic
resources are important inputs to children’s health and success at school. Since each outcome
measure is reported in 5 categories, we estimate ordered probit models.13 Our basic specification
for these regressions follows Dooley et. al., 1998 or Curtis, Dooley and Phipps, 1999. That is, we
employ a relatively pared-down specification, controlling for pmk having less than high-school
education, child age, a dummy for gender of the child (=1 if the child is female); number of
children in the household and age of the pmk. These controls are employed in all specifications,
with our focus being upon the impact of varying our measure of poverty/income. 14
To provide a benchmark, we begin by re-estimating, using the 1996 data, models which are very
similar to those available using the first wave of the NLSCY. Thus, we regress current (1996)
outcome measures on current (1996) measures of poverty and annual income. However, as argued
above, current measures of either income or poverty status are arguably not the most appropriate
indicators of the economic resources available to the child.
First, given the volatility of current annual income for some families, particularly, lower-income
families, a two-period average measure of income is presumably preferable to current annual
income purely as an indicator of family “permanent income.” Thus, we include average income
and “average poverty” (i.e., income for 1994 and 1996 less than a two-period average poverty
line).15 However, it is not necessarily the same experience to have an average income which is low,
13
We were concerned about the statistical validity of using all 5 categories when there are so few children
reported to have, for example, poor health. Thus, we re-ran all of the models reported in this paper
aggregating to 4 and also to just 3 categories. Qualitative results were unaffected. We thus maintain the 5
categories on the grounds that the small number of children reported in the worst health category are
presumably those with very serious medical conditions, whose poor health status should not be aggregated
with those of other children.
14
Since some of our observations are children from the same family, we adjust all standard errors for nonindependence of observations using the “cluster” procedure of Stata.
15
18
Our measure of poverty is 50 percent of median equivalent income using the OECD equivalence scale. Median
equivalent incomes are calculated using the 1994 and 1996 SCF’s, respectively. The “average poverty” line is
calculated as the average of the 1994 and 1996 poverty lines (with the 1994 poverty line expressed in 1996
Applied Research Branch
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
and to have low income which lasts a long time. Thus, we also test the association between twoperiod poverty (i.e., poor in both 1994 and 1996) and children’s health and success at school. This
measure begins to incorporate the duration of low-income experience, but it is also possible that
the dynamics of low-income experience are important for children’s well-being. Becker’s
framework, for example, allows for the possibility that returns to investments in human capital
may only be apparent with a lag (e.g., tutoring may result in improved school performance next
period rather than immediately). Hence, we also consider lagged rather than current measures of
poverty and income. On the other hand, coming full circle, it may be more important that you are
hungry today than that you were hungry yesterday for how well you are able to perform at school
currently. Thus, it is not a priori obvious which measures of income/poverty are likely to have the
largest association with child outcomes, and this may differ across outcomes.
Table 2
Means for the Estimating Sample—Analysis Variables
Poor in 1996
24.9%
Poor in 1994
25.5%
Poor in 1994 and 1996
17.8%
Poor using an average over 1994 and 1996
23.3%
Equivalent income 1996 (1996 $)
19,149
Equivalent income 1994 (1996 $)
18,933
Average equivalent income (1996 $)
19,041
Household member owns home
78.1%
House needs major repairs
6.4%
Weekly available parental hours
147.5
Number of observations
7337
Table 2 reports means for our first set of “economic resource” variables (means for other controls
are reported in Appendix Table 1). In 1994, 25.5 percent of children aged 6 to 13 years lived in
poor families; in 1996, 24.9 percent were poor.16 23.3 percent of children lived in families whose
dollars). Average income is the family’s average income for the two periods, expressed in 1996 dollars.
16
Recall that these figures are for our estimating sample. However, poverty rates for the full sample of children
aged 2 to 13 years in 1996 were very similar: 26.9 percent in 1994 and 26.1 percent in 1996.
Applied Research Branch
19
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
average incomes across the two periods was low enough to be counted as poor “on average;” 17.8
percent of children were poor in both periods. While we do not know whether children who were
poor in both 1994 and 1996 were poor throughout the entire time period, these data do indicate
significant persistence of low-income status.
Table 3 presents results for ordered probit models17 of the association between child health and
alternative measures of income and poverty status. Consider, first, results for alternative measures
of poverty. Interestingly, given the motivation for this paper, it is current poverty status which has
the largest association with child health, though all measures considered are statistically significant,
and fairly similar in magnitude. The second largest association is with having been poor in both
periods. Note that low-income status is the most important factor associated with child health in
these regressions, followed by low-education status and lone-parent status.18
The last 3 specifications reported in Table 3 use alternative measures of income rather than dummy
variables for poverty status. We use “equivalent” income rather than “actual” income, where
equivalent income is actual income divided by an equivalence scale to adjust for the economies of
scale available to individuals who live together (e.g., savings on housing, utilities, transportation
costs).19 Since we also include a direct control for number of children in the household, our intent
is to capture the financial implications of additional children through the “equivalent” income
variable and other implications of siblings in the direct measure (e.g., playing, fighting, comforting,
etc.). Since economists generally argue that a marginal dollar is of more value to a low-income
family than to a high-income family (i.e., that there is diminishing marginal utility of income), we
include both equivalent income and its square in these models.
17
The ordered probit results cannot be interpreted in the same way as OLS results can be. A positive coefficient
means the distribution of answers shifts to the right. From this we know that the probability of being in the
lowest (best in outcomes in this case) category decreases and being in the highest category increases (worst
outcome here). It is not possible to discern the changes in the distribution of the middle categories from
“eyeballing” the coefficients.
18
In this paper, “lone-parents” include both lone mothers and lone fathers.
19
As for our poverty calculations, we employ the OECD equivalence scale for these calculations. Thus, the first
adult is assigned an equivalence value of 1.0; subsequent adults are assumed to add 0.7 to household needs;
subsequent children are assumed to add 0.5.
20
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
When alternative measures of equivalent income are used in place of the poverty dummies, it is the
two-period average of equivalent income which has the largest association with child health
(though again all income variables tested are significant at the 99 percent level). For both current
and lagged income, “declining marginal utility of income” is apparent (i.e., the quadratic term is
statistically significant and negative—inflection points are at about $26,000 1994 equivalent dollars
and at $35,000 equivalent dollars respectively).
Table 4 changes the focus from health to success at school, but otherwise repeats exactly the same
exercise. In this case, the poverty measure with the largest association with success at school is
“poor in both periods ”; lagged poverty has the second largest effect. Again, aside from child
gender (girls do significantly better at school), the poverty associations are the largest observed in
these models. In terms of other controls, having a pmk with low education is next most important
followed by lone-parent status.
For the models which include alternative measures of equivalent income and equivalent income
squared (reported in the final 3 columns of Table 4), it is again true that lagged and average
equivalent income have larger associations with children’s success at school than current income.
And, for both lagged and average equivalent income, the quadratic term is statistically significant
and negative (indicating that in terms of association with success at school, dollars to poorer
children are more important than dollars to richer children).20
Throughout this paper, some readers may feel concern about our treatment of income as an
exogenous variable. Specifically, it is possible that in fact the causality is in the other direction: it
may be that poor child health is associated with low-income status because, for example,
possibilities for parental participation in the paid labour market are limited by the need to care for
the child with health problems. Regressing change in dependent variables on changes in
independent variables is one way to address the issue of endogeneity between income/poverty and
health; if we could identify a change in income in one period followed by a change in health status
the next period we could argue convincingly that the poverty “causes” decreases in health status.
20
Inflection points for these quadratics are at $22,000 1994 dollars and $29,000 1996 dollars, respectively.
Applied Research Branch
21
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
However, as there are only two waves of the NLSCY available to date we still cannot trace in
which order the changes are occurring. Moreover, as argued earlier, explaining changes in
outcomes is not the same thing as explaining levels of outcomes, which is our emphasis in this
paper.21
We present here the results of an informal investigation of the issue of potential endogeneity of
income to child health status. The key line of reasoning outlined earlier is that it is possible that
having an unhealthy child might mean parents do not work for pay, or work fewer hours. It does
not seem reasonable to argue that having an unhealthy child would limit other forms of income
(e.g., transfers, asset income). Thus, if there is validity in the idea that income is endogenous to
child health, we should see parents with less healthy children engaging in less paid work.
We investigate this idea informally in Table 5, which compares parental labour market outcomes
for the “best” and “worst” outcomes, where best and worst are identified by the bottom two and
top categories for schooling and health, respectively. Note, first, that father’s rates of labour-force
participation and hours of paid work have essentially no association with the child outcomes (i.e.,
outcomes appear very similar for fathers with most and least healthy children). Mother’s labourforce participation is lower for health, but much the same for schooling outcomes (in fact, labour
force participation is slightly higher for the children with the best outcomes). Thus, there is
possibly a connection for general health, but the issue seems less important for success at school.
This seems reasonable.
21
22
In other work (Curtis and Phipps, 2000), we present the results of a multivariate analysis of changes in child
health on changes in poverty status. Unfortunately, results were considerably less precise than those obtained
for levels of child outcomes.
Applied Research Branch
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Table 3
Ordered Probit Estimates of 1996 Health Using Current, Lagged and Two-Period Measures of Poverty
and Equivalent Income
Poor in
1996
Poor in
1994
Poor in
both 1996
and 1994
Average income
less than poverty
line
1994
equivalent
income
1996
equivalent
income
Average
equivalent
income
0.222*
(0.061)
0.191*
(0.060)
0.217*
(0.069)
0.205*
(0.062)
—
—
—
Equivalent income
—
—
—
—
Equivalent income squared
—
—
—
—
0.145**
(0.067)
0.161**
(0.079)
0.166*
(0.067)
0.171**
(0.078)
0.159**
(0.068)
0.170**
(0.079)
0.155**
(0.068)
0.163**
(0.080)
-0.206*
(0.040)
0.008**
(0.004)
0.121***
(0.065)
0.126
(0.080)
-0.209*
(0.037)
0.006***
(0.004)
0.98
(0.066)
0.110
(0.081)
-0.225*
(0.042)
0.007
(0.004)
0.095
(0.065)
0.108
(0.081)
-0.015
(0.010)
-0.001
(0.043)
-0.069*
(0.027)
0.0003
(0.005)
0.055
(0.197)
0.980*
(0.197)
1.926*
(0.204)
2.736*
(0.227)
-0.015
(0.010)
0.001
(0.043)
-0.065**
(0.027)
0.001
(0.005)
0.090
(0.198)
1.014*
(0.199)
1.961*
(0.205)
2.772*
(0.229)
-0.014
(0.010)
-0.002
(0.043)
-0.067*
(0.027)
0.0001
(0.005)
0.041
(0.197)
0.965*
(0.198)
1.910*
(0.204)
2.718*
(0.228)
-0.015
( 0.010)
0.002
(0.043)
-0.068*
(0.027)
0.001
(0.005)
0.056
(0.197)
0.980*
(0.197)
1.927*
(0.203)
2.738*
(0.226)
- 0.017***
(0.010)
0.004
(0.043)
-0.090*
(0.027)
0.005
(0.005)
-0.236
(0.198)
0.697*
(0.199)
1.648*
(0.205)
2.458*
(0.226)
-0.018***
(0.010)
0.007
(0.043)
-0.095*
(0.027)
0.006
(0.005)
-0.247
(0.204)
0.688*
(0.204)
1.640*
(0.209)
2.449*
(0.231)
-0.018***
(0.010)
0.006
(0.043)
-0.098*
(0.027)
0.006
(0.005)
-0.266
(0.202)
0.670*
(0.202)
1.623*
(0.208)
2.432*
(0.229)
Variable
Dummy=1 if family is poor
Dummy=1 if a lone parent family
Dummy=1 is the PMK has< high
school
Age of the child
Dummy=1 if the child is female
Number of children in household
Age of the PMK
Intercept 1
Intercept 2
Intercept 3
Intercept 4
Note: Equivalent income is family income divided by the OECD equivalence scale to account for the economies of scale available to individuals who live together. Equivalent income is measured in tens of thousands Canadian dollars. All
income measures are expressed in 1996 dollars (i.e. 1994 dollars are inflated to 1996 values).
* significant with 99% confidence
** significant with 95% confidence
*** significant with 90% confidence
Applied Research Branch
23
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Table 4
Ordered Probit Estimates of 1996 Success at School Using Current, Lagged and Two-Period Measures
of Poverty and Equivalent Income
Poor in
1996
Poor in
1994
Poor in
both 1996 and
1994
Average income
less than poverty
line
1994
equivalent
income
1996
equivalent
income
Average
equivalent
income
0.236*
(0.058)
0.244*
(0.052)
0.300*
(0.062)
0.225*
(0.056)
—
—
—-
Equivalent income
—
—
—
—
Equivalent income squared
—
—
—
—
-0.177*
(0.037)
0.008**
(0.004)
-0.143*
(0.034)
0.004
(0.003)
-0.176*
(0.038)
0.006***
(0.004)
Dummy=1 if a lone parent
family
Dummy=1 is the PMK has<
high school
Age of the child
0.138**
(0.065)
0.143**
(0.062)
0.125**
(0.063)
0.143**
(0.064)
0.137**
(0.062)
0.138**
(0.063)
0.126**
(0.063)
0.176*
(0.071)
0.180*
(0.070)
0.174*
(0.071)
0.176*
(0.071)
0.155**
(0.070)
0.156**
(0.071)
0.147**
(0.071)
0.029*
(0.010)
-0.238*
(0.041)
0.030*
(0.010)
-0.234*
(0.041)
0.030*
(0.010)
-0.239*
(0.041)
0.029*
(0.010)
-0.223*
(0.041)
0.028*
(0.010)
-0.233*
(0.041)
0.027*
(0.010)
-0.230*
(0.041)
0.027*
(0.010)
-0.232*
(0.041)
-0.021
(0.026)
-0.020
(0.026)
-0.025
(0.030)
-0.021
(0.026)
-0.033
(0.026)
-0.029
(0.026)
-0.035
(0.026)
-0.006
(0.005)
-0.094
(0.181)
0.594*
(0.182)
-0.005
(0.005)
-0.036
(0.184)
0.652*
(0.186)
-0.006
(0.005)
-0.098
(0.182)
0.590*
(0.183)
-0.006
(0.005)
-0.094
(0.182)
0.594*
(0.183)
-0.002
(0.005)
-0.341***
(0.179)
0.350**
(0.181)
-0.002
(0.005)
-0.301***
(0.185)
0.389**
(0.186)
-0.001
(0.005)
-0.341***
(0.182)
0.351***
(0.184)
1.852*
(0.187)
2.770*
(0.202)
1.910*
(0.1907)
2.831*
(0.205)
1.850*
(0.189)
2.771*
(0.203)
1.850*
(0.188)
2.770*
(0.203)
1.615*
(0.186)
2.532*
(0.199)
1.651*
(0.190)
2.569*
(0.204)
1.616*
(0.188)
2.534*
(0.202)
Variable
Dummy=1 if family is poor
Dummy=1 if the child is
female
Number of children in
household
Age of the PMK
Intercept 1
Intercept 2
Intercept 3
Intercept 4
Note: Equivalent income is family income divided by the OECD equivalence scale to account for the economies of scale available to individuals who live together. Equivalent income is measured in tens of
thousands Canadian dollars. All income measures are expressed in 1996 dollars (i.e. 1994 dollars are inflated to 1996 values).
* significant with 99% confidence
** significant with 95% confidence
***significant with 90% confidence
Applied Research Branch
24
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Table 5
Average Labour Force Participation and Weeks Worked for Children in
Top and Bottom Decile/Category for Each Outcome, Mothers and
Fathers
Health
Top two categories
(fair & poor)
Mother
Labour force
participation
Weeks paid work
last year
Father
Labour force
participation
Weeks paid work
last year
Applied Research Branch
Success at School
Bottom category
( excellent)
Top two categories
(poorly & very
poorly)
Bottom category
(very well)
65.3%
73.0%
72.6%
75.4%
24.8
31.7
30.8
32.9
94.8%
96.4%
96.7%
95.9%
46.3
47
45.3
47
25
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Table 6
Ordered Probit Estimates of 1996 Health, Adding Controls for Home Ownership, Housing State of
Repair and Available Parental Time
Variable
Dummy=1 if family is poor
Poor in both Poor in both
1996 and 1994 1996 and 1994
0.141**
(0.068)
Equivalent Income
—
Equivalent Income Squared
—
0.208*
(0.072)
—
Poor in both Average equivalent Average equivalent Average equivalent
1996 and 1994
income
income
income
0.140**
(0.071)
—
—
—
-0.187*
(0.046)
-0.239*
(0.042)
-0.202*
(0.045)
—
0.004
(0.005)
0.008**
(0.004)
0.005
(0.005)
Dummy=1 if a lone parent family
0.083
(0.073)
0.222***
(0.117)
0.114
(0.121)
0.046
(0.072)
0.023
(0.113)
-0.046
(0.117)
Dummy=1 is the PMK has< high
school
0.147***
(0.080)
0.171**
(0.080)
0.151***
(0.081)
0.096
(0.082)
0.116
(0.081)
0.105
(0.082)
Age of the child
-0.016
(0.010)
-0.140
(0.010)
-0.015
(0.010)
-0.018***
(0.010)
-0.018***
(0.010)
-0.019***
(0.010)
Dummy=1 if the child is female
0.003
(0.042)
-0.002
(0.043)
0.002
(0.043)
0.008
(0.043)
0.006
(0.043)
0.008
(0.043)
Number of children in household
-0.063**
(0.027)
-0.070*
(0.027)
-0.063**
(0.027)
-0.093*
(0.027)
-0.097*
(0.027)
-0.092*
(0.027)
Age of the PMK
0.002
(0.005)
0.00002
(0.005)
0.002
(0.005)
0.007
(0.005)
0.007
(0.005)
0.008
(0.005)
Dummy=1 if dwelling is owned
by household member
-0.218*
(0.069)
—
-0.218*
(0.069)
-0.141**
(0.069)
—
-0.151**
(0.069)
Dummy=1 if major repairs
required
0.275*
(0.086)
—
0.280*
(0.086)
0.226*
(0.083)
—
0.227*
(0.084)
—
0.001
(0.001)
0.0004
(0.001)
—
-0.001
(0.001)
-0.001
(0.001)
Intercept 1
-0.059
(0.199)
0.171
(0.280)
0.010
(0.280)
-0.274
(0.205)
-0.419
(0.288)
-0.464
(0.290)
Intercept 2
0.870*
(0.199)
1.093*
(0.280)
0.937*
(0.279)
0.665*
(0.205)
0.514***
(0.287)
0.472***
(0.289)
Intercept 3
1.822*
(0.206)
2.038*
(0.288)
1.889*
(0.289)
1.622*
(0.211)
1.466*
(0.295)
1.429*
(0.297)
Intercept 4
2.635*
(0.227)
2.847*
(0.300)
2.704*
(0.298)
2.436*
(0.231)
2.276*
(0.305)
2.243*
(0.305)
Weekly available parental hours
* significant with 99% confidence
** significant with 95% confidence
*** significant with 90% confidence
Applied Research Branch
26
W-01-1-4E
6.
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
Multivariate Analysis of the Association Between Child
Health and Success at School and Economic Resources,
Adding Controls for Assets and for Time
This section of the paper moves beyond measuring economic resources simply as income flows in
an attempt to control for: 1) assets available to the family and; 2) time associated with acquiring
income. Economists would not argue with the proposition that two families with the same income
but different levels of wealth are not in the same economic circumstances. The problem is that few
microdata sets provide information about assets, and the NLSCY is no exception. However, since
housing is often the major form in which families with children hold assets, we decided to include
a dummy variable indicating that the house in which the child lives is owned by a member of the
household (not necessarily mortgage free). In our estimating sample, 78 percent of children live in
owner-occupied dwellings. Of course, houses can be big or small, fancy or plain. We have no
indication of the assess value of the home, so we attempted a variety of measures to proxy for this.
Our best alternative22 appears to be that the child lives in a home “in need of major repair” (i.e.,
defective pluming or electrical wiring, structural repairs to walls, floors or ceilings, etc.).23 Only 6
percent of children in our estimating sample live in homes which require major repairs.24
The other new control which we add in this section is for the amount of time required to acquire
family income. Particularly for children, parental time can be an important input to well-being. A
Becker-style human capital model predicts that, income equal, two families with more time
available for children should be able to make greater investments in their children’s human capital.
The Coleman perspective adds the idea that parental time is required to develop “social capital”
(e.g., by attending Home and School meetings or helping to coach soccer teams). Thus, we create
a “parental time available per week.”25 For lone parents, this is total weekly hours, less weekly
22
“Number of bedrooms per person” was not statistically significant in any case.
23
Of course, some homes can be both very expensive and in need of major repairs.
24
Interestingly, correlations of these variables with the incidence of poverty is relatively low. For example, the
correlation between being poor in both 1994 and 1996 and owning a home in 1996 is -0.41; for housing
requiring major repairs the correlation is 0.07.
25
We choose to focus upon weekly rather than annual hours as this seems more relevant for children. Care can’t
be deferred until a later point in the year. See also Phipps, Burton and Osberg who find it is weekly hours
which generate most time stress for adults.
Applied Research Branch
27
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
“sleep time” (i.e., 8X7),26 less usual weekly hours of paid work.27 For two-parent families, we
create this measure for each parent, and add them. Thus, it is clear that we are in some sense
muddling lone-parent status with the “available time” variable.28 That is, the maximum available
weekly parental hours for a lone-parent with zero hours of paid work is 112 [24X7 - 8X7]. Both
parents in a two-parent household would have to work at least 56 hours per week in the paid
labour force to have as little available time as a lone parent with the maximum available time. In
fact, in our data, 32 percent of lone parents do not have paid work and thus have 112 available
weekly hours. And, a handful of two-parent households do have fewer free hours than this (0.4
percent of all two-parent families). However, it is very clear that mean available parental time is
much greater for two-parent families (161 hours) than for lone-parent families (90 hours).
The major sources of variation in the available time variable are thus marital status of parents in
combination with patterns of labour-force participation. As noted above, 68 percent of lone
parents participate in the paid labour force. For couples, in 73 percent of cases, both parents have
paid employment; 25 percent of two-parent families have one earner. And, of course, hours of
paid employment vary significantly for those with paid employment. Figure 1 illustrates the
distributions of the available time variable for one- and two-parent families.
In Table 6, we add, separately and together, the home and time variables to the child health
ordered probit equations with “poor in both 1994 and 1996” and “average equivalent income” as
our currently “longest term” measures of financial flows. Table 7 presents the same information
for success at school. We are interested both to see what these additional measures of economic
resources add to our story, and to see how they affect the estimated association between
poverty/income and children’s outcomes.
26
We all know that parents are not always able to sleep a full 8 hours. And, it seems likely that this is especially
difficult for lone parents. However, we do not actually know how many hours parents sleep. Subtracting the
same quantity of “sleep hours” for each parent will have no real impact on our results, but it makes
interpretation more logical.
27
Parents who are self employed are not asked their usual weekly hours. However, we have a report of annual
hours, which we use to construct the weekly measure. For anyone who reported no paid work during the
survey year, “zero weekly hours” were assigned.
28
The correlation between the “lone-parent” dummy and the available parental time variable is -0.80.
28
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
Consider, first, the role of home ownership as a proxy for assets. This variable is statistically
significant, and associated with better outcomes for children in all of the health models estimated.
For over-all health status, the quantitative magnitude of the variable is rather large—considerably
larger than the dummy variable indicating poverty in both periods, for example. For success at
school, it is not significant when entered without controlling for time, but significant at the 90
percent level when the time variable is added to the model.
The dummy variable indicating that the family residence is in need of major repairs is statistically
significant for both outcomes in all specifications (at the 99 percent level) and always associated
with worse outcomes for children. The association between “major repairs” and child health status
is one of the most important observed (larger than the two-period poverty and the home
ownership associations). For schooling, two-period poverty status still has the larger association,
but “major repairs” dominates lone-parent status (which, in fact, is not statistically significant in
this specification).
Note that with the addition of the home ownership dummy, the apparent association of both loneparent status and poverty status with child health status diminish. Poverty status remains
statistically significant and a relatively important explanatory variable, but coefficient size falls
from 0.217 to 0.141 when the housing variables are added to the specification. Lone-parent status
becomes insignificant. However, the finding of a smaller association between low-income and
child health does not mean that economic resources are less important than we had previously
thought. Rather, we have added another channel through which economic resources may influence
child health. Thus, for example, a lone mother living with a low income but who owns her own
home is in rather different circumstances than a lone mother with the same income who does not.
We should not just look at income when attempting to understand the importance of economic
variables for child health.
Finally, consider the role of available parental time as a correlate of child outcomes. This variable
is not statistically significant in any of the child health equations, but, controlling for income, more
available parental time is consistently significant at the 99 percent and associated with greater
success at school. Presumably, time is a particularly important input both for direct help with
homework, enriching outings, and reading to children, for example, and for indirect support of
Applied Research Branch
29
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
school activities (e.g., home and school work, volunteering at the school, going along on school
outings). Notice that when “available time” is added without also controlling for the housing
variables, lone parent status is statistically insignificant (though coefficient estimates are not
noticeably different in actual magnitude). In the specification which includes both available hours
and the housing variables, lone-parent status is again statistically significant.29
Once we have controlled for available time in the schooling equations, the estimated association
between poverty and equivalent income and success at school increases in magnitude. This
remains true after we add in the housing variables.
29
30
Given the high correlation (-0.8) between lone-parent status and available time, it seems natural to suppose
that what is going on here is imprecision due to multicollinearity. Given this concern, we also re-ran these
models separately for two-parent families and obtain qualitatively the same results—available time is
statistically insignificant in the health equations, but significant at the 99 percent level in the schooling
equations. Moreover, the observed poverty and income effects increase in magnitude when we control for
time.
Applied Research Branch
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Table 7
Ordered Probit Estimates of 1996 Success at School, Adding Controls for Home Ownership and
Housing State of Repair
Variable
Poor in both
Poor in both
Poor in both
Average
Average
Average
1996 and 1994 1996 and 1994 1996 and 1994 equivalent income equivalent income equivalent income
0.261*
(0.065)
—
0.358*
(0.064)
—
0.317*
(0.066)
—
Equivalent income squared
—
—
—
Dummy=1 if a lone parent
family
Dummy=1 is the PMK has<
high school
Age of the child
0.086
(0.069)
-0.121
(0.113)
0.162**
(0.072)
Dummy=1 if family is poor
Equivalent Income
Dummy=1 if the child is female
Number of children in
household
Age of the PMK
Dummy=1 if dwelling is owned
by household member
Dummy=1 if major repairs
required
Weekly available parental
hours
Intercept 1
Intercept 2
Intercept 3
Intercept 4
—
—
—
0.187***
(0.115)
-0.154*
(0.038)
0.005
(0.004)
0.094
(0.069)
-0.213*
(0.040
0.009**
(0.004)
-0.154
(0.112)
-0.190*
(0.039)
0.007***
(0.004)
-0.211***
(0.115)
0.191*
(0.071)
0.177*
(0.072)
0.138**
(0.071)
0.163**
(0.071)
0.154**
(0.071)
0.297*
(0.010)
-0.238*
(0.041)
-0.022
(0.026)
0.029*
(0.010)
-0.239*
(0.041)
-0.019
(0.016)
0.029*
(0.010)
-0.237*
(0.041)
-0.016
(0.026)
0.027*
(0.010)
-0.231*
(0.040)
-0.032
(0.026)
0.255*
(0.010)
-0.230*
(0.041)
-0.030
(0.027)
0.025*
(0.010)
-0.229*
(0.041)
-0.016
(0.026)
-0.005
(0.005)
-0.112
(0.071)
0.005
(0.005)
—
-0.004
(0.005)
-0.134***
(0.070)
-0.001
(0.005)
-0.091
(0.068)
-0.0001
(0.005)
—
0.001
(0.005)
-0.116***
(0.067)
0.193**
(0.087)
—
0.188**
(0.086)
0.164***
(0.088)
—
0.153***
(0.086)
—
-0.003*
(0.001)
-0.003*
(0.001)
-0.147
(0.186)
0.543*
(0.188)
1.806*
(0.192)
2.730*
(0.206)
-0.563**
(0.266)
0.124
(0.267)
1.388*
(0.273)
2.306*
(0.281)
-0.663*
(0.270)
0.025
(0.271)
1.294*
(0.276)
2.215*
(0.283)
—
-0.351***
(0.184)
0.342***
(0.185)
1.609*
(0.189)
2.530*
(0.203)
-0.004*
(0.001)
-0.004*
(0.001)
-0.937*
(0.276)
-0.246
(0.277)
1.024*
(0.282)
1.939*
(0.288)
-0.987*
(0.278)
-0.296
(0.279)
0.978*
(0.283)
1.895*
(0.288)
* significant with 99% confidence
** significant with 95% confidence
***significant with 90% confidence
Applied Research Branch
31
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Figure 1
Available Parental Hours
Lone Parents and Couples
35
30
lone parents
couples
percentage
25
20
15
10
Applied Research Branch
223
220
217
214
211
208
205
202
199
196
193
190
187
184
181
178
175
172
169
166
163
160
157
154
151
148
145
142
139
136
133
130
127
124
121
118
115
112
109
106
103
97
100
94
91
88
85
82
77
74
71
68
64
57
28
25
0
22
5
32
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
7. Conclusions
This paper asks whether economic resources matter for children’s outcomes. Both economic
theory and public concern over high levels of child poverty suggest that there is an important
association. Yet, research utilizing the first wave of the National Longitudinal Survey of Children
and Youth (1994) suggested, surprisingly, that low-income status is a relatively unimportant
correlate of children’s outcomes. If true, the policy implication is that income transfers are
relatively unimportant for children.
The goal of this paper is to re-examine the association between economic resources and children’s
health and success at school. Although we recognize that the well-being of children is multidimensional, we limit ourselves in this paper to health and success at school as two particularly
“economic” outcomes (i.e., they may be viewed as key elements of children’s “human capital” and
hence may be studied within the framework of economic models; both have important implications
for children’s eventual labour market success).
A key advantage of the 2nd wave of NLSCY data is that it allows us to move beyond current
income and/or poverty status as a measure of the economic resources available to the child. Given
the ups and downs of family income, particularly for lower-income families, a longer-term average
should be a more reliable indication of family economic resources than any current year measure.
It is also possible that the effects of economic resources only appear with a lag, hence it may be
that previous year’s income is more important than current income. Finally, it is also possible that
duration of low-income status is important. We examine these hypotheses and conclude that for
children’s health status, current poverty status and two-period (i.e., longer-duration) poverty have
the largest associations with current child health, though poverty matters regardless of the way in
which we measure it. A two-period average of income has the largest association with child
health. For success at school, it is clearly the longer-term poverty and the two-period average of
income which have the largest associations.
However, economists would argue that wealth as well as income flows are also a vital component
of the economic resources available to a family. While we do not have any direct information
about family assets, we include a proxy for home ownership and for the state of repair of the
Applied Research Branch
33
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
family dwelling. Finally, traditional economic reasoning also suggests that, income constant,
families with more time are better off than those with less. When we control for both housing and
available parental time per week, we find that children who live in owner-occupied housing have
better outcomes than children who do not; children who live in housing in need of major repairs
have worse outcomes. This represents an additional channel through which economic resources
can influence outcomes for children. Weekly hours of parental time available has no statistically
significant association with child health; however, income constant, more hours of parental time
available each week significantly improves a child’s success at school.
These results indicate that while “longer-term” income is an important factor in child well-being,
measured by health and educational success, that other measures of economic resources are also
important. The policy conclusions associated with “low or moderate association between income
and child well-being may be misleading (for example Blau (1999) “Policies that affect family
income will have little direct impact on child development unless they result in very large and
permanent changes in income.” (P.261)). Income transfers may assist families with children to
accumulate assets such as housing or increase parental time spent with children. These attributes
are associated with better outcomes for children even after controlling for income. Other policy
instruments may be extended parental leave, home ownership assistance plans or assisting lowincome families with the completion of housing repairs.
34
Applied Research Branch
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
Appendix
Table 1
Means for the Estimating Sample - Control Variables
PMK has less than high school education
12.1%
Age of the child
9.56
Child is female
49.1%
Lone parent household
17.7%
Number of children in the household
2.37
Age of the PMK
37.8
Number of observations
7337
Applied Research Branch
35
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
W-01-1-4E
References
Boyle, M.H. and Lipman E. (1998) Do Places Matter? A Multilevel Analysis of
GeographicVariations in Child Behaviour in Canada. Working Paper W-98-16E, Applied
Research Branch, Strategic Policy, Human Resources Development Canada.
Cadman, D., Boyle, M.H., Offord, D.R., Szatmari, P., Rae-Grant, N.I., Crawford, J., and Byles, J.
(1986a) “Chronic Illness and Functional Limitation in Ontario Children: Findings of the
Ontario Child Health Study.” Canadian Medical Association Journal (135), pp. 761-767.
Curtis, L.J., Dooley M., Lipman E, and Feeny D. (1998) The Role of Permanent Income and
Family Structure in the Determination of Child Health in the Ontario Child Health Study.
Mimeo, Dalhousie University (Submitted to Journal of Health Economics).
Curtis, L.J., and Phipps S. (2000) The Missing Link Between Income and Child Health and Wellbeing: Are We Measuring Income Adequately? Mimeo, Dalhousie University.
Curtis, L.J., Dooley M., and Phipps S. (2000) Does Mother or Father Know Best? An Assessment
of Parent/Child Agreement in the Canadian National Longitudinal Survey of Children and
Youth. Mimeo, Dalhousie University.
Deaton, A., and Paxson, C. (1999) Mortality, Education, Income and Inequality Among
American Cohorts. National Bureau of Economic Research working paper series no. 7140,
Cambridge Massachusetts.
Dooley, M.D., and Lipman, E.L. (1996) “Child Psychiatric Disorders and Poor School
Performance: The Roles of Family Type, Maternal Market Work and Low Income.” in
Towards the XXIst Century: Emerging Sociodemongraphic Trends and Policy Issues in
Canada. Proceedings of a Symposium of the Federation of Canadian Demographers.
Dooley, M.D., Curtis, L., Lipman, E. and Feeny, D. (1998a) “Child Behaviour Problems, Poor
School Performance and Social Problems: The Roles of Family Structure and Low Income in
Cycle One of the National Longitudinal Survey of Children and Youth.” in Labour Markets,
Social Institutions, and the Future of Canada's Children, edited by Miles Corak, Statistics
Canada.
Dooley, M.D., and Curtis, L. (1998b) The Role of Income, Family Structure and Parental Market
Work in the Determination of Child Health in the National Longitudinal Survey of Children
and Youth. Canadian International Labour Network Working Paper, McMaster University.
Grossman, M. (1972) “On the Concept of Health Capital and the Demand for Health.” Journal of
Political Economy (80), no. 2, pp. 223-255.
_______ (1972a) The Demand for Health: a Theoretical and Empirical Investigation, Occasional
paper (119), National Bureau of Economic Research, Columbia University Press, New York.
Grossman, M. and Joyce T. (1989) “Socioeconomic Status and Health: A Personal Research
Perspective.” In Pathways to Health: The Role of Social Factors, Gomby D.S., & Kehrer B.
H., editors. Mento Park, The Henry J. Kaiser Family Foundation, pp. 138-163.
Kephart, G., Thomas V.S. and MacLean D.R. (1998) “Socioeconomic Differences in the Use of
Physician Services in Nova Scotia.” American Journal of Public Health (88), no. 5,
pp. 800-803.
36
Applied Research Branch
W-01-1-4E
Economic Resources and Children’s Health and Success at School:
An Analysis Using the NLSCY
Korenman, S., Miller J. and Sjaastad J. (1995) “Long-term Povery and Child Development in the
United States: Results from the NLSY.” Children and Youth Services Review (17), no.1-2,
pp.127-155.
Lantz, P.M., House J.H., Lepkowski J.M., Williams D.R., Mero R.P., and Chen J. (1998)
“Socioeconomic Factors, Health Behaviours, and Mortality: Results from a Nationally
Representative Prospective Study of U.S. Adults.” JAMA (279), no. 21, pp.1703-1708.
Laroche, M. (1998). “In and Out of Low Income.” Canadian Social Trends (50), pp. 20-24.
Lipman, E.L., and Offord, D.R. (1994) Disadvantaged Children. In The Canadian Guide to
Clinical Preventive Health Care. The Canadian Task Force on the Periodic Health
Examination, Minister of Supply and Services Canada, Ottawa, Ontario,
(Cat. No. H21-117/1994E), pp. 356-368.
_______ (1996). Psychosocial Morbidity Among Poor Children in Ontario. in Duncan G.,
Brooks-Gunn J. (eds.), Growing Up Poor. Russell Sage Foundation.
Lipman, E.L., and Offord, D.R., Boyle, M.H. (1994) Economic Disadvantage and Child PsychoSocial Morbidity. Canadian Medical Association Journal (151), pp.431-437.
Munroe Blum, H., Boyle, M.H., and Offord, D.R. (1988) Single Parent Families: Child Psychiatric
Disorder and School Performance. Journal of the American Academy of Child and
Adolescent Psychiatry (27), pp. 214-219.
Offord, D.R., Boyle, M.H., and Jones, B.R.. (1987) “Psychiatric Disorder and Poor School
Performance Among Welfare Children in Ontario.” Canadian Journal of Psychiatry (32), pp.
518-525.
Phipps, S.A. (1998a) The Well-Being of Canadian Children in International Perspective. Paper
presented at the conference The State of Living Standards and the Quality of Life in Canada.
Ottawa, October 30, 1998.
_______ (1998b) What Is the Best Mix of Policies for Canada’s Children? An International
Comparison of Policies and Outcomes for Young Children. Study prepared for the Canadian
Policy Research Network and Human Resources Development Canada.
Smith, D.G., Bartley, M., and Blane, D. (1990) “The Black Report on Socioeconomic Inequalities
in Health 10 years On.” British Medical Journal (301), pp. 373-377.
Smith, D.G., Shipley, M.J., and Rose, G. (1990b) “ Magnitude and Causes of Socioeconomic
Differentials in Mortality: Further Evidence from the Whitehall Study.” Journal of
Epidemiology and Community Health (44), pp. 265-270.
Williamson, D.L., and Fast, J.E. (1998) “Poverty Status, Health Behaviours and Health
Implications for Social Assistance Policy” Canadian Public Policy (24), no.1, pp. 1-25.
Applied Research Branch
37
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement