This presentation by Derek Headey, IFPRI was shown at the Transform Nutrition - Evidence for Action regional meeting in Kathmandu, Nepal on 8 July 2017. This one-day event shared Transform Nutrition evidence on key issues related to nutrition policy in Nepal, Bangladesh and India, lessons on strategies for change from other contexts and discuss the relevance and applicability of the research findings to policies/programmes that aim to address nutrition in South Asia.
Drivers of nutritional change in South Asia: Insights from empirical analyses of national survey data
1. Drivers of nutritional change in South Asia:
Insights from quantitative analyses
Derek Headey
d.headey@cgiar.org
Poverty, Health and Nutrition Division
International Food Policy Research Institute (IFPRI)
1
2. Quantitative research on nutritional change in
South Asia
• Headey, D., Hoddinott, J., Ali, D., Tesfaye, R., Dereje, M., 2015. The Other
Asian Enigma: Explaining the Rapid Reduction of Undernutrition in
Bangladesh. World Development 66, 749-761.
• Headey, D.D., Hoddinott, J., 2015. Understanding the Rapid Reduction of
Undernutrition in Nepal, 2001–2011. PLoS ONE 10, e0145738.
• Headey, D., Hoddinott, J., 2016. Agriculture, nutrition and the green
revolution in Bangladesh. Agricultural Systems 149, 122-131.
• Headey, D., Hoddinott, J., Park, S., 2016. Drivers of nutritional change in
four South Asian countries: a dynamic observational analysis. Maternal &
Child Nutrition 12, 210-218.
• Headey, D., Hoddinott, J., Park, S., 2017. Accounting for nutritional
changes in six success stories: A regression-decomposition approach.
Global Food Security 13, 12-20.
• Cunningham, K., Headey, D., Singh, A., Karmacharya, C., Rana, P.P., 2017.
Maternal and Child Nutrition in Nepal: Examining drivers of progress from
the mid-1990s to 2010s. Global Food Security 13, 30-37.
3. Introduction
Why do all this work on nutritional change in South Asia?
Notorious for undernutrition: 50-60% stunting in 1990s
This “Asian Enigma” produced many interesting hypotheses:
intrahousehold biases (Jayachandran & Pande, 2013)
low status of women (Shroff et al., 2009, Menon, 2012),
early marriage and child birth (Raj et al., 2010),
genetic predispositions (Nubé, 2009),
poor diets (Deaton & Dreze, 2008, Headey et al., 2012)
poor child feeding practices (Menon, 2012)
inefficacy of nutrition programs (World Bank, 2005)
open defecation (Spears 2012)
4. Figure 1A. The Asian Enigma in the 1990s
Bangladesh-1997
India-1998
Nepal-1997
Pakistan-1991
Sri Lanka-1993
0
204060
0 2000 4000 6000 8000 10000
GDP per capita (constant international dollars)
5. Year Under-
weight
Stunting Year Under-
weight
Stunting
Bangladesh 1997 51.2 58.9 Nepal 1996 42.1 56.3
2011 36 41 2011 28.1 39.6
Speed -1.1 -1.3 Speed -0.9 -1.1
India 1998 42.3 50.7 Pakistan 1991 35.3 53.5
2005 42.1 47.5 2012 28.6 44.8
Speed 0.0 -0.5 Speed -0.3 -0.4
Table 1.
Nutritional change in 4 South Asian countries
Source: DHS. Notes: Speed is prevalence points per annum.
6. Figure 1B. Two Asian Enigmas in the late 2000s
BGD-2011
India-2006
Nepal-2011
Pakistan-2012
Sri Lanka-2006
0
204060
0 2000 4000 6000 8000 10000
GDP per capita (constant international dollars)
7. Introduction
So in reality, nutritional progress has been solid to impressive:
HAZ scores improved by 15.6% to 25.6% since 1990s
Nepal & Bangladesh highly successful in reducing stunting;
Nepal did so during conflict!!!
Goal of this body of research was to break away from
pessimistic narrative & identify:
1. What has been working?
2. Where is more progress needed?
3. What can South Asian countries learn from each other?
8. Introduction
We use a quantitative regression-decomposition approach
Several advantages:
Can standardize and compare across countries
Avoids cherry-picking: let’s the data to the talking
Can be easily replicated and is highly transparent
Several limitations:
Difficult to say much about nutrition-specific programs,
though we can look at changes in IYCF practices
Cannot fully ascribe causality, only plausibility
Does not tell us about political economy of nutritional
change or directly about effectiveness of specific policies
Qualitative work helps to bridge these limitations
9. What does regression-decomposition approach entail?
1. Requires multiple rounds of same survey instrument (DHS)
2. Identify set of explanatory (X) factors that may explain HAZ
change, and measure them consistently over time
3. Regression model linking change in X to change in HAZ (b)
4. Decompose historical change across survey rounds as the
product of changes in Xs and corresponding bs
5. Predict future change based on targeted changes in Xs
Example of antenatal care (ANC)
1. ANC in Nepal: 10% to 50% over 1996-2011, or 40 points
2. Regression model says ANC improves HAZ by 0.1
3. Estimated contribution of ANC to HAZ improvement over
1996-2011 is 40% multiplied by 0.1 = 0.04 std. deviations
Data & methods
10. To look at nutritional change across South Asia, we used all
available DHS:
Bangladesh: 1996/1997, 1999/2000, 2004, 2007, and 2011
Nepal: 1996, 2001, 2006, and 2011
India: 1992/1993 and 2005/2006
Pakistan: 1991 and 2013
For all rounds in all countries we construct standard indicators of:
Asset scores to capture economic growth/poverty reduction
Parental education
WASH: open defecation, improved water
Health: ANC, born in medical facility (general health proxies)
Demographics: Children ever born, birth intervals
Data & methods
11. Figure 1. The distribution of HAZ scores in Nepal, 2001 and 2011
0
.1.2.3
Density
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4
HAZ score (children 0-59 months)
2001 2011
12. Figure 2. HAZ scores by child age in Nepal, 2001 and 2011
-3-2-1
0
0 20 40 60
Child's age (months)
95% CI 2001
95% CI 2011
Major increase in
birth sizes
Very small abatement of
growth faltering
14. Which explanatory factors have large marginal
effects on HAZ?
Model
(1)
Bangladesh
(2)
Nepal
(3)
India
(4)
Pakistan
Asset index, 1-10 0.069*** 0.047*** 0.071*** 0.054***
Maternal education (years) 0.023*** 0.028*** 0.022*** 0.032***
Paternal education (years) 0.024*** 0.018*** 0.014*** 0.021***
4 or more antenatal visits 0.053 0.095** 0.069*** 0.137**
Born in a medical facility 0.114*** 0.146*** 0.071*** 0.254***
Open defecationa -0.150*** -0.172*** -0.165*** -0.185*
Piped waterb 0.0201 -0.149*** -0.0790* -0.138
Tube well waterb 0.282*** 0.198*** -0.0239 -0.0172
Birth interval 0.049*** 0.032*** 0.044*** 0.043**
Number of children -0.030*** -0.022** -0.035*** -0.031**
Female child 0.005 0.017 0.041*** 0.062
R-squared 0.183 0.236 0.175 0.223
N 16279 9852 39568 4865
15. Estimated contributions to HAZ change over
time in four South Asian countries
0.14 0.14
0.07
0.10
0.07 0.07
0.03
0.05
0.05
0.07
0.03
0.05
0.03
0.03
0.01
0.03
0.06 0.02
0.01
0.02
0.03 0.08
0.02
0.12
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
Bangladesh Nepal India Pakistan
EstimatedchangeinHAZscore
Healthcare
Fertility
Paternal
schooling
Open
defecation
Women's
schooling
Assets
16. Commonalities across models:
Similar effect sizes and contributions of assets and education
Reductions in open defecation important (not silver bullet)
Some differences:
Family planning only sizeable in Bangladesh
Improvements in access to health services big factors in
Pakistan and Nepal (major policy success in Nepal?)
Explanatory power of decomposition varies: 52% in India,
65% in Bangladesh, 84% in Nepal and 121% in Pakistan
Results
17. What else may have been driving nutritional change?
Little evidence of nutrition-specific programs driving change :
Few notable improvements in IYCF practices
One exception is earlier introduction of weaning foods in
Bangladesh, which may be related to rice productivity growth
Also improved consumption of fish in Bangladesh but no
improvement in dairy consumption
Diets mostly remain highly monotonous
Results
Exclusive
breastfeeding,
kids 0-6 months
Fed any solid foods
yesterday,
kids 6-8 months
Fed 4 types of food
yesterday,
kids 6-24 months
Fed 3 or more
times yesterday,
kids 6-23 months
2001 79.7% 61.1% 18.4% 72.7%
2011 78.6% 61.9% 17.7% 85.5%
Change (%) -1.1% 0.8% -0.7% 12.8%
18. What about the future?
Women still catching up on education, but evidence that it
has wide-ranging nutritional benefits (Headey & Alderman)
Huge gaps in antenatal, neonatal & postnatal care, but
perhaps some valuable lessons from Nepal
Open defecation persists, but also broader hygiene problems
Bangladesh mostly eliminated open defecation, but still
massive seasonal problems in wasting
Mounting evidence that animal feces are a problem
WASH problems compounded by high population density
South Asian children still face terrible IYCF practices, which
are still not well understood
But emerging evidence from Bangladesh that both
knowledge and financial constraints play a role
Conclusions
Number of units: 1: Unit of measurement: Kilogram: Min: 0.5: Max: 1.2: Brand: Well known: Type: Long grain, white rice (milled rice): Packaging: Pre-packed; paper or plastic bag: Quality: High grade: Preparation: Parboiled : Share of broken rice: Very low (not more than 5%): Other features: Not enriched, not aromatic (fragrant), not sticky: Exclude: Premium rice e.g. Basmati rice, Jasmine rice