Putting Children First: Identifying solutions and taking action to tackle poverty and inequality in Africa.
Addis Ababa, Ethiopia, 23-25 October 2017
This three-day international conference aimed to engage policy makers, practitioners and researchers in identifying solutions for fighting child poverty and inequality in Africa, and in inspiring action towards change. The conference offered a platform for bridging divides across sectors, disciplines and policy, practice and research.
How to design healthy team dynamics to deliver successful digital projects.pptx
Putting Children First: Session 3.1.B Saurabh Sinha & Kalkidan Assefa - Analysing child stunting in Africa [25-Oct-17]
1. ANALYSING CHILD STUNTING IN AFRICA:
Does urbanization make a difference?
Presentation by
Saurabh Sinha
Kalkidan Assefa
Employment and Social Protection
Social Development Policy Division, UNECA
International Conference ‘Putting Children First: Identifying
Solutions and Taking Action to Tackle Child Poverty and Inequality
in Africa’
23-25 October 2017, Addis Ababa, Ethiopia
2. Structure of the presentation
• Objective of the research
• What does the research do?
• Urbanisation trends
• Hypotheses testing – methodology
• Results
• Conclusion
2
3. Objective of the research
• Analyse the linkage of urbanization with nutrition outcomes, focusing
on child stunting.
• The research recognizes the heterogeneous nature of urbanization in
Africa and introduces the average annual rate of change of the urban
population as a determining factor for the difference in nutritional
outcomes between urban and rural areas.
3
4. What does the research do?
The research tests two hypotheses:
• Even though the prevalence of childhood stunting is equally high
across countries, urban populations are significantly better off than
rural populations as in almost every country in Africa, average living
standards in urban areas are superior to those in rural areas
regardless of national income levels.
• In Africa, the size of urban-rural welfare gaps within countries varies
a great deal across countries, with higher gaps in less urbanized
countries so that in most countries, urban-rural differentials in child
stunting converge with increased urbanization.
4
5. Urbanization trends in Africa
• Africa is the fastest urbanizing region in the world.
• Richer countries are more urbanized
5
Average annual rate of
growth of urban
population (%)
Number of countries
< 1 0
1 – 2 3
2.1 – 3 5
3.1 – 4 13
4.1 – 5 12
> 5 6
• Though close to 60% of people still live in rural areas, it is
estimated that by 2030, more than half of Africa’s population
will live and work in urban areas.
6. Higher propn. of children stunted at lower levels of urbanization
Cat.
Urban
popn. as
% of total
popn.
No. of
countries
Countries Avg. GNI
per capita
(2011 PPP $)
Propn. of
children
stuntedResource-rich Non-resource-rich
I >60 10
Algeria, Libya (NA); Djibouti
(EA); Congo, Gabon (CA); South
Africa (SA)
Morocco, Tunisia (NA); Cabo
Verde (WA); Sao Tome and
Principe (CA)
9,201
26.3
(4)
II 51-60 7
Mauritania (NA); Cote d’Ivoire;
Ghana (WA); Cameroon (CA);
Botswana (SA)
Seychelles (EA); Gambia
(WA)
7,834
28.9
(5)
III 41-50 10
DR Congo (EA); Benin, Liberia,
Nigeria (WA); Angola, Namibia,
Zambia (SA)
Egypt (NA); Guinea-Bissau,
Senegal (WA)
4,263
33.1
(9)
IV 31-40 13
Sudan (NA); Madagascar,
Tanzania (EA); Guinea, Mali,
Sierra Leone, Togo (WA); CAR,
Eq. Guinea (CA); Mozambique,
Zimbabwe (SA)
Somalia (EA); Mauritius (SA) 4,590
35.8
(9)
V <30 14
Eritrea, Rwanda, South Sudan,
(EA); Burkina Faso, Niger (WA);
Chad (CA); Lesotho (SA)
Burundi, Comoros, Ethiopia,
Kenya, Uganda (EA);
Malawi, Swaziland (SA)
1,937
41.9
(13)
54 36 18 5031
6
8. 0
10
20
30
40
50
60
10 20 30 40 50 60 70 80 90
Stunting(%ofchildrenunder5)
% of urban population
LOWER LEVELS OF CHILD STUNTING
Source: UNDP (2015), Human Development Report; ECA (2015), Demographic Profile of Africa
Countries with large urban populations exhibit improved
social outcomes…
8
9. BUT, child stunting in lowest quintile has either increased,
stagnated, or declined only slowly
Sub-region Country Year Q1 Q5 Q1/Q5
National
Average
Central Africa
Cameroon
1991 39.1 14.8 2.6 31.3
2011 48.6 12.3 4 32.5
Chad
1996-97 50.9 35.5 1.4 44.5
2014-15 41.2 31.5 1.3 39.9
Congo
2005 37.4 24.8 1.5 30.8
2011-12 34.5 9.3 3.7 24.4
Gabon
2000 41 14 2.9 25.1
2012 29.9 5.8 5.2 16.5
East Africa
DRC
2007 46.6 25.8 1.8 45.5
2013-14 49.7 22.9 2.2 42.7
Eritrea
1995 53.9 28.8 1.9 44.4
2002 50.1 21 2.4 42.9
Ethiopia
2000 60.6 48.5 1.2 57.7
2016 44.6 25.6 1.7 38.4
Kenya
1993 47.8 23.3 2.1 39.9
2014 30.2 13.8 2.6 26
9
10. Sub-region Country Year Q1 Q5 Q1/Q5
National
Average
Southern
Africa
Lesotho
2009 45.6 28.3 1.6 39.2
2014 45.6 13.4 3.4 33.2
Malawi
1992 60.6 42.6 1.4 55.2
2015 45.7 24.3 1.9 37.1
Namibia
1992 44.9 22.2 2.0 34.8
2013 31.3 8.7 3.6 23.7
Zimbabwe
1994 31.6 19.4 1.6 28.6
2015 33 16.6 2.0 26.8
West Africa
Burkina Faso
1993 43 27 1.6 38.8
2010 41.9 18.6 2.3 34.6
Guinea
1999 36.9 18 2.1 30.5
2012 35.2 15.2 2.3 31.2
Liberia
2007 44.5 26.4 1.7 39.4
2013 35.3 19.9 1.8 31.6
Nigeria
1990 53.7 36.7 1.5 48.7
2013 53.8 18 3.0 36.8
10
BUT, child stunting in in lowest quintile has either
increased, stagnated, or declined only slowly…
11. North Africa* bucks the trend:
Decline in child stunting is pro-poor
Country Year Q1 Q5 Q1/Q5
National
Average
Egypt
1995 41.7 23.6 1.8 33.7
2014 24.1 23.4 1.0 21.4
Morocco
1992 44.3 12.6 3.5 29.9
2003-04 35 12.5 2.8 22.4
11
* Comparable DHS data available only for 2 countries
12. SIMILARLY, child stunting in rural areas has either
increased, stagnated, or declined only slowly
Sub-region Country Year Rural Urban
Rural:urban
ratio
National
Average
Central Africa
Cameroon
1991 37.1 22.7 1.63 31.3
2011 40.5 21.9 1.85 32.5
Chad
1996-97 46.6 36.5 1.28 44.5
2014-15 46.8 35.4 1.32 39.9
Congo
2005 34.2 26.7 1.28 30.8
2011-12 30.4 20.3 1.5 24.4
Gabon
2000 36.2 20.5 1.77 25.1
2012 28.5 14.1 2.02 16.5
East Africa
Burundi
1987 53.2 32.5 1.64 52.5
2010 59.5 37.8 1.57 57.7
DRC
2007 51.5 36.7 1.40 45.5
2013-14 47.1 32.5 1.45 42.7
Eritrea
1995 47 34.5 1.36 44.4
2002 48.6 31.7 1.53 42.9
Ethiopia
2000 58.9 47.6 1.24 57.7
2016 39.9 25.4 1.57 38.4
12
13. Sub-region Country Year Rural Urban
Rural:Urban
ratio
National
Average
Southern
Africa
Malawi
1992 56.9 40.7 1.40 55.2
2015-16 38.9 25 1.56 37.1
Mozambique
1997 45.3 33.8 1.34 42.4
2003 51.7 35.6 1.45 42.6
Zambia
1992 52.8 39.2 1.35 46.4
2013-14 42.1 36 1.17 40.1
Zimbabwe
1994 30.2 24.1 1.25 28.6
2015 28.5 22.1 1.29 26.8
West Africa
Benin
1996 34 29.1 1.17 32.5
2011-12 46.1 42.3 1.09 44.6
Burkina Faso
1993 41.5 24.2 1.71 38.8
2010 37.3 21.3 1.75 34.6
Mali
1987 30.6 22.3 1.37 27.8
2012-13 41.9 23.2 1.81 38.3
Niger
1992 48 32.8 1.46 45.2
2012 45.9 29.6 1.55 43.9
13
SIMILARLY, child stunting in rural areas has either
increased, stagnated, or declined only slowly…
14. Again, North Africa* bucks the trend:
Egypt the only country where urban stunting higher than rural stunting
Country Year Rural Urban
Rural:Urban
ratio
National
Average
Egypt
19 39.5 28.8 1.37 34.8
2014 20.7 23 0.90 21.4
Morocco
1992 36.2 23.3 1.55 29.9
2003-04 28.8 16.2 1.78 22.4
14
* Comparable DHS data available only for 2 countries
15. KEY OBSERVATIONS, SO FAR
• Five child stunting hotspots
• Burundi
• Benin
• Cameroon
• Mali
• Mozambique
• Nutrition policy can be pro-poor, so that the lower quintiles (and rural areas)
reduce stunting independent of the top quintiles (and urban areas) – Egypt;
Morocco
• Lessons can be learnt from Ethiopia, Kenya, Liberia, Malawi, Namibia, Tanzania,
Uganda which have achieved appreciable and uniform declines in child stunting,
but also from Burkina Faso, Niger, Rwanda, Zambia where progress has been
uneven.
15
16. HYPOTHESES TESTING
1. An independent t-test used to determine whether the mean prevalence
rate of two samples are significantly different from each other
Mean significance test between the Rural and Urban stunting prevalence rate
Mean significance test between the Lowest quintile and Highest quintile in stunting
prevalence rate
Hypothesis testing
• Ho: Rural-Urban = 0
• Ha: (Rural-Urban) > 0 or Rural-Urban < 0
2. Pearson correlation coefficient (r) to test the relationship between
average annual rate of change of urban population (the urban popn.
growth rate) and the Rural-Urban stunting differential (“stunting gap”)
magnitude of the Pearson correlation coefficient determines the strength of the correlation
Coefficient Value Strength of Association
0.1 < | r | < .3 small correlation
0.3 < | r | < .5 medium/moderate correlation
| r | > .5 large/strong correlation
17. Result: Normality test
• Assessment of the normality of the pooled sample data
Distribution of Rural and Urban stunting prevalence:
17
• Based on the available data and normality test, we can conclude that the distribution of both samples
is normal
18. • The result using both paired and unpaired sample t-test indicates a
significant difference between rural and urban child stunting
prevalence rate of 12.2 percentage points (significant at 99%)
Results: Rural-Urban difference (mean significance test)
Variable Obs. Mean Std. Err. Std. Dev. [95% Conf. Interval]
Rural stunting Prev. 136 40.68824 .8004105 9.33431 39.10527 42.2712
Urban Stunting Prev. 136 28.47721 .7153849 8.34275 27.06239 29.89202
combined 272 34.58272 .6516131 10.74668 33.29985 35.86559
diff 12.21103 1.073514 10.09751 14.32455
diff = mean (Rural Stunting Prevalence) - mean (Urban stunting Prevalence) t = 11.3748
Ho: mean(diff) = 0 degrees of freedom = 270
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 1.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 0.0000
19. Result-normality test
• Assessment of the normality of the pooled sample data
Distribution of Lowest Quintile and Highest Quintile in stunting prevalence:
• Based on the available data and normality test, we can conclude that the distribution of both samples
is also normal
20. • The result using both paired and unpaired sample t-test indicates a
significant difference between child stunting prevalence rate in the
lowest quintile (Q1) and the highest quintile (Q5) of 20 percentage
points (significant at 99%).
Results: Lowest and Highest Wealth Quintile
(mean significance test)
Variable Obs. Mean Std. Err. Std. Dev. [95% Conf. Interval]
lowest 119 44.43025 .8568113 9.346707 42.73353 46.12697
highest 119 24.27395 .8669422 9.457223 22.55717 25.99073
combined 238 34.3521 .8935452 13.78494 32.5918 36.11241
diff 20.1563 1.218899 17.75499 22.55761
diff = mean(lowest) - mean(highest) t = 16.5365
Ho: mean(diff) = 0 degrees of freedom = 236
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 1.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 0.0000
21. • Between average annual rate of change of urban population (urban
growth rate) and Rural-Urban differential in stunting prevalence
• Based on the our data, the Pearson correlation coefficient (r) = 0.4517
• Coefficient of determination (r2) = 0.20, or 20% is explained by the model
• The level of statistical significance (p-value) of the correlation coefficient
is 0.0 (99% significance) – suggests a strong and significant relationship
between the two variables: urban growth rate and Rural-Urban stunting
difference.
• That is, as urban population growth rate increases, the rural-urban
stunting differential (“stunting gap”) also increases.
Results: Correlation analysis
22. Conclusions
Important to undertake a more dynamic analysis – ‘annual rate of growth of
urban population’ better predictor than ‘extent of urbanization’.
As Africa urbanises rapidly, child stunting declining faster in urban areas and
among upper quintiles.
As the rural-urban differential (“stunting gap”) increases with increasing rate of
growth of urban population, a large proportion of children in rural areas from
lower income quintiles likely to be left behind.
Urgent need to extend universal access to basic services to address child
stunting in rural areas. Esp. focus on the stunting hotspots.
More rigorous research required, esp. to better analyse the differentials
among socioeconomic groups within urban areas – whether the poorest
urban quintiles are better or worse off compared to the rural mean.
22