And Middle-Income Countries PPT - Development Studies Association

Report
Children’s Multidimensional
Health and Medium-Run
Cognitive Skills in Low- and
Middle-Income Countries
Elisabetta Aurino
Imperial College, London
Francesco Burchi
German Development Institute, Bonn
Background & Motivation
• Child development intrinsic and instrumental
objective of development policy
• Literature on Early Child Development (ECD)
– Children’s health in the first 1000 days on
lifecourse outcomes
• Limitations
– Height as key indicator of health
– Enrolment as most used proxy education
– Often cross-sectional data
Objectives
• To examine the effect of child’s height, the most commonly
used indicator of health, on cognitive abilities at preschool
and primary school age in Ethiopia, India, Peru and
Vietnam.
• To investigate whether cognitive abilities are better
explained by a “suite of health indicators”.
• To analyse whether a composite index of health
deprivation in early childhood can synthesise adequately
the overall effect of early childhood deprivation in health
on children’s later cognitive outcomes
• To explore a few possible channels through which child
health may affect cognitive skills.
Child health and mid-run educational
achievements: direct effects
• Early childhood health can affect cognitive development in
different ways (Jukes 2005).
• Child health typically proxied by low height-for-age
•
Associations with:
– School completion (e.g. Moock and Leslie 1986 in Nepal;
Clark et al. 1990 in Jamaica; Glewwe and Jacoby 1995 in
Ghana; Shariff et al. 2000 in Malaysia; Alderman et al. 2006
in Zimbabwe etc).
– Cognitive abilities (Hoddinott et al. 2008; Behrman et al.
2008)
•
Some studies analysed the relationship between child height/stunting and
cognitive outcomes of pre-school children (Sanchez 2009; Outes-Leon et al 2011)
or of children at age 8 (Crookston et al. 2010, 2013) using Young Lives data.
Child health and mid-run educational
achievements: indirect effects
• Child health could have an indirect effect on
learning abilities, by increasing educational
aspirations.
– Dercon and Sanchez (2013) find a positive and large
effect of height-for-age at the age 7-8 on educational
aspirations for children of age 11-12
• Ailments or undernutrition may affect long-run
cognitive functioning by impacting on short- middlerun health outcomes (i.e. brain development) or
affecting other psychosocial competencies such as
agency or self-efficacy (Jukes 2005)
Conceptual framework
• Child health is viewed as a complex, multi-faceted
phenomenon
– About 15% of “healthy years of life” among children aged 0-4 in
less developed countries is lost due to mortality and morbidity, and
half of the burden of disease is due to communicable diseases
(Glewwe & Miguel 2008)
– Most of the infectious diseases affecting children in developing
countries are unlikely to affect height-for-age
– Malnutrition itself is a multidimensional phenomenon (i.e. chronic,
concurrent, micronutrient deficiencies etc)
• We move from one single measure (height) to a
multidimensional assessment.
– Given the complexity of health, and the complexity of its
relationship with other human development outcomes, a set of
health indicators is likely to be more informative (Strauss &
Thomas 2008).
Econometric strategy
• We model cognitive outcomes at preschool
and primary school age as a function of
children’s early childhood health, and a set of
other child- and household- characteristics
• OLS estimation: long-term associations, not
causal impact
• Omitted variables bias alleviated by the use
of cluster fixed effects and many relevant
controls available in the YL study
Econometric strategy (ctd.)
• Several different models for every dependent
variable (2), country (4) and round of survey
(2):
1. With only height-for-age
2. With all health indicators separately
3. With composite health deprivation index
• Robustness check: addition of variables
related to child’s concurrent nutrition and
schooling, and to her cognitive test scores
(round 2) were added to explore possible
channels.
The Young Lives data
• Longitudinal study of childhood poverty in
Ethiopia, India (Andhra Pradesh), Peru and
Vietnam.
• Younger cohort: 1 y.o. in round 1 (2002), 5 y.o. in
round 2 (2006) and 8 y.o. in round 3 (2009).
• Sampling: non-random selection of “sentinel
sites”, and then random selection of children in
these sites.
• Attrition in the sample is extraordinarily low thanks
to a particular effort in tracking children when they
move (Outes-Leon and Dercon 2008).
Outcome indicators of learning
• At preschool age (5 years, round 2):
– Peabody Picture Vocabulary Test (PPVT): test of
vocabulary acquisition.
– Cognitive Developmental Assessment (CDA) test:
measures children’s grasp about quantity-related
concepts.
• At primary school age (8 years, round 3):
– PPVT
– Mathematics Achievement Test: measures basic
numeracy skills. 29 simple arithmetical problems.
Indicators of early childhood health
• 2 indicators of nutrition (in z-scores):
– Height-for-age (HAZ), proxy for chronic nutritional status,
– Weight-for-height (WHZ), proxy for acute nutritional status.
• 1 indicator of morbidity (binary): whether the child has
experienced a life-threatening illness or injury since
birth (reported by the main caregiver).
• Composite index “Multidimensional Health Poverty
Index” (MHPI), a multidimensional score of children’s
joint deprivations in the health dimension.
• 3 dimensional cut-offs, equal weights: ranges from 0
(child is not deprived in any dimension) to 1 (deprived
in all 3 dimensions)
• ...a possible proxy for child multidimensional poverty!
Descriptive stats of EC health
Height-for-age
z-scores
Weight-for-
Life-
height z-
threatening
scores
illness
MHPI
Mean
-1.48
-0.73
0.30
0.29
SD
-1.83
1.39
0.46
0.27
Mean
-1.3
-1.21
0.22
0.24
SD
1.47
1.06
0.42
0.26
Mean
-1.28
0.6
0.32
0.2
SD
1.27
1.11
0.47
0.23
Mean
-1.12
-0.62
0.13
0.13
SD
1.25
0.95
0.34
0.21
Ethiopia
India
Peru
Vietnam
Controls
• Child’s gender; age (in months); ethnicity (or caste
in India); child’s mother tongue; disability status;
whether the child is first-born; born in a health
facility; vaccinations; preschool attendance;
• Caregiver’s sex, age, level of education;
• HH size; sex, age and education of the head of
the household; mother’s and father’s presence, 3
composite indicators for HH economic status;
• Robustness check: (1) child’s contemporaneous
nutritional status (round 3), as measured by BMI;
(2) child school grade completed (round 3); (3)
child cognitive attainments (round 2).
Results (I)
Results (II)
• As expected, child HAZ has a positive and significant
effect on all cognitive skills, throughout the 4 countries
• Only exception is for PPVT scores in round 3 in
Vietnam
• The magnitude of the coefficient varies across
countries, outcomes and child developmental stages
• WHZ is significantly associated with all the cognitive
outcomes in Andhra Pradesh and with Maths scores in
Peru. When significant, the magnitudes of the WHZ
coefficients are similar, or slightly smaller, to the ones of
the HAZ indicator.
• Especially in India, both forms of malnutrition work in
concert to hinder children’s cognitive development
Results (III)
• The indicator related to early childhood lifethreatening disease is not significantly associated to
any outcomes in any countries, with the exception of
Vietnam in Maths.
• MHPI is significant in all the outcomes and rounds
considered in Andhra Pradesh and Peru, while it is
significant in the case PPVT round 2, CDA and
Maths in Vietnam, and only in the case of Maths
scores in Ethiopia.
• With MHPI the adjusted R2 falls and we lose
information about the specific health dimension that
has an influence on cognition
• The suite of indicators is more informative
Robustness checks (I)
Table 5. Results from the extended models, Ethiopia Round 3
VARIABLES
Height-for-age z-scores
Weight-for-height z-scores
Illness (z-scores)
(1)
PPVT
(2)
PPVT
(3)
PPVT
(4)
PPVT
(5)
MATHS
(6)
MATHS
(7)
MATHS
(8)
MATHS
0.030
(2.470)**
0.009
(0.587)
-0.009
(-0.888)
0.030
(2.461)**
0.016
(1.048)
-0.008
(-0.791)
0.018
(1.753)*
0.005
(0.344)
-0.016
(-1.579)
0.063
0.025
(2.110)**
0.007
(0.455)
-0.010
(-0.833)
0.037
(3.553)***
0.027
(1.688)
0.014
(0.842)
0.037
(3.540)***
0.030
(1.790)*
0.014
(0.832)
0.014
(1.540)
0.019
(1.327)
-0.002
(-0.141)
0.118
0.036
(3.477)***
0.027
(1.624)
0.012
(0.671)
Enrolled but not completed
Grade 1
(1.038)
0.283
(3.972)***
0.537
Enrolled & completed Grade1
Enrolled & completed Grades 2
or 3
(2.583)**
0.678
(10.085)***
0.970
(5.742)***
BMI-for-age z-score
-0.016
(-0.860)
CDA raw score R2 z scores
PPVT raw score R2 z scores
Constant
(7.913)***
-0.038
(-2.155)**
-3.613
-3.645
(-7.579)***
(-7.834)***
Observations
1,639
1,638
Number of clusters
20
20
Adj. R-squared
0.140
0.141
R3 core controls
YES
YES
Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1
-3.077
(-7.231)***
1,628
20
0.174
YES
0.118
(5.663)***
0.105
(2.918)***
-2.989
(-6.758)***
1,599
20
0.179
YES
-2.569
(-5.102)***
1,596
20
0.126
YES
-2.591
(-5.110)***
1,595
20
0.126
YES
-1.723
(-4.700)***
1,584
20
0.263
YES
0.078
(2.759)**
0.079
(2.478)**
-2.111
(-4.550)***
1,560
20
0.144
YES
17
Robustness checks (II)
Table 7. Results from the extended models, Peru Round 3
VARIABLES
Height-for-age
scores
z-
Weight-for-height zscores
Illness (z-scores)
(1)
PPVT
(2)
PPVT
(3)
PPVT
(4)
PPVT
(5)
MATHS
(6)
MATHS
(7)
MATHS
(8)
MATHS
0.037
0.036
0.033
0.026
0.057
0.055
0.043
0.052
(3.789)***
0.008
(3.583)***
0.005
(3.449)***
0.005
(3.490)***
0.005
(5.919)***
0.044
(5.832)***
0.037
(4.232)***
0.034
(5.349)***
0.039
(0.818)
-0.004
(-0.434)
(0.605)
-0.005
(-0.450)
(0.537)
-0.004
(-0.390)
0.293
(7.146)***
0.368
(0.591)
-0.001
(-0.148)
(3.423)***
-0.010
(-0.896)
(3.030)***
-0.010
(-0.886)
(2.678)**
-0.009
(-0.785)
0.631
(7.492)***
1.061
(2.757)**
-0.009
(-0.713)
Completed Grade 2
Completed Grade 3 or
4
(7.318)***
Bmi-for-age z-scores
0.022
(1.435)
CDA raw score R2 zscores
PPVT raw score R2
z-scores
Constant
Observations
Number of clusters
R3 core controls
Adj. R-squared
(11.411)***
0.009
(0.836)
0.041
0.106
(4.576)***
0.198
(4.474)***
0.170
-3.223
(-10.882)***
-3.217
(-10.709)***
-2.761
(-6.097)***
(10.046)***
-2.131
(-6.626)***
-5.919
(-11.809)***
-5.901
(-11.847)***
-2.849
(-5.516)***
(7.813)***
-4.849
(-9.466)***
1,762
20
YES
0.265
1,760
20
YES
0.265
1,752
20
YES
0.286
1,720
20
YES
0.374
1,800
20
YES
0.251
1,798
20
YES
0.251
1,790
20
YES
0.322
1,756
20
YES
0.287
Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1
18
Conclusions (1)
1. Does child height affect cognitive skills?
 With the exception of PPVT scores in round 3 in
Vietnam, the estimates show a positive, highly
significant effect
 In India a 40% increase of a standard deviation in
HAZ would translate into equalising the
performances of rural and urban children in CDA
scores, while an increase of a standard deviation
in HAZ would be equivalent to closing half of the
gender gap in Maths scores in Peru in round 3.
19
Conclusions (2)
2. Does a suite of indicators help understanding the
health effect on learning?
 Evidence of the relevance of WHZ, proxy for acute
malnutrition, particularly in India, where its coefficient
is always significant, and in the case of Maths scores
in Peru. However, large heterogeneity.
 The morbidity indicator, instead, contributes to
explain only Maths scores in Vietnam.
 While height remains the key indicator, recognizing
the multidimensional nature of health provides
additional policy-relevant information, not necessarily
at higher costs.
 Improve the quality of morbidity indicators
20
Conclusions (3)
3. Can the whole contribution of children’s health be
summarized by a composite health deprivation index?
 “Suite of indicators” approach provides substantially
more information.
 Though, the index can provide a “quick and dirty”
measure for assessing overall health status
4. What are the possible channels?
 A large part of the early childhood health-cognition
nexus is mediated by variation in grade attainment,
especially in Ethiopia, India and for maths skills in
Vietnam. However, other channels may play an
additional role.
 Peru is an outlier
21
• Thank you for your attention!
– [email protected]
– And thanks to the thousand Young Lives
children, families and communities that
rendered possible this study
Findout
outmore
more@
@
Find
www.schoolsandhealth.org
www.child-development.org
ForFor
School
& Nutrition (SHN):
ourHealth
latest
• Research
• Documents
& resources
• News
& Events
• Case
studies
• SHN topic information
• Toolkitsand views
• Country specific data
Follow SHN @schoolhealth
www.facebook.com/PartnershipforChildDevelopment
@ schoolhealth

similar documents