Child Mortality, Migration and Household Composition in Sub-Saharan Africa
Philippe Bocquier, Université catholique de Louvain
Carren Ginsburg, University of the Witwatersrand
Yacouba Compaoré, Université Catholique de Louvain
Mark Collinson, University of the Witwatersrand
Our objective is to evaluate the nature and extent of the impacts of migration, wealth, and living arrangements on under-5 child mortality. We will analyse the extensive Health and Demographic Surveillance Sites (HDSS) data gathered by the INDEPTH Network a previously untapped source of data. To better understand child mortality we will control for the first time in longitudinal analyses for the impacts of children’s migration, parents’ and siblings’ migration, parents’ and siblings’ death, and household composition and socio-economic environment. We will then use the findings to test key hypotheses regarding the extent of the differential impact of migration and household shocks on child health.
Our first results show positive effects of child migration on their survival, at the opposite of migration effect on adult mortality. Mother’s and also siblings’ death, but not their migration, increase child death risks. A complex data management procedure has been tested with success on Ouagadougou HDSS (Burkina Faso) and Agincourt HDSS (South Africa) which will enable us to construct a household typology. It will be used to analyse data obtained in September 2017 on more than 20 other HDSS in SSA, covering at least 1 million children in 8 or more African countries.
The lives of a high proportion ofchildren under-5 in Africa are affected by the rise in internal andinternational migration. Empirical studies show that changes in familystructure, household wealth, and place of residence impact on healthcare,access to health services, health status, and the survival of both the childrenwho migrate and those who are left behind. Most past research has focused onhealth at destination and has not taken account of the variation in wealth andliving arrangements before and after relatives migration or death, and has notcontrolled for independent migration of children and other family members. Theconsequences of death or migration-related fostering on child health are ofparticular importance in Africa given the AIDS epidemic, violent conflicts, andpoverty.
The objective of the project isto evaluate the nature and extent of the impacts of migration, wealth, andliving arrangements on under-5 child mortality. We will analyse the extensive datagathered by the INDEPTH Network, a yet to be exploited major informationresource on these issues. This will be used to reconstruct family relationshipsusing multi-year longitudinal data covering more than 18,000 child deathsamongst 1 million children under-5 in 12 Health and Demographic SurveillanceSites (HDSS) in 8 African countries (Burkina Faso, Côte dIvoire, Ethiopia,Kenya, South Africa, Malawi, Mozambique, Senegal), but the maximum could reach20 in 11 African countries if data from other HDSSs are validated.
To better understand childmortality we will control for the first time in longitudinal analyses for theimpacts of childrens migration, parents and siblings migration, parents andsiblings death, and household composition and socio-economic environment. Wewill then use the findings to test key hypotheses regarding the extent of thedifferential impact of childs, mothers and fathers migration and householdshocks and composition (presence of grand-parents) on child health. The newinsights gained will stimulate new thinking about the strategies and measuresneeded to improve child health support in highly mobile and vulnerablesub-Saharan African populations.
Our first results show positiveeffects of child migration on their survival, at the opposite of migrationeffect on adult mortality. Mothers and also siblings death, but not theirmigration, increase child death risks. A complex data management procedure hasbeen tested with success on Ouagadougou HDSS (Burkina Faso) and Agincourt HDSS(South Africa). A household typology is under construction which will enable usto identify grand-parents. It will be used to analyse data obtained inSeptember 2017 on more than 20 other HDSSs in SSA.
As part of a 5-year project, thispaper will be the first to use systematic data management procedures on HDSSdata to produce strictly comparable indicators. For example, five eligibilityquantitative criteria will be applied in the selection of the HDSS forinclusion in the analysis. The paper will include a thorough methodologicaldiscussion on procedures and analytical limitations. In this first paper, onlythe readily available data will be analysed, and it is expected that furtherhousehold data will be collected to pursue the objectives of the project.
Difference inadult and child mortality by migration status (23 HDSSs in SSA)
Preliminaryunder-5 child mortality analysis (Agincourt HDSS)
No.of subjects = 44,169 Number of obs = 232,014
No.of failures = 1,578
Timeat risk = 119751.4847
LR chi2(23) = 215.43
Loglikelihood = -15877.175 Prob > chi2 = 0.0000
-------------------------------------------------------------------------------------
_t | Haz. Ratio Std.Err. z P>|z| [95% Conf. Interval]
child_rank | REF=1
2 | .4382341 .2197569 -1.65 0.100 .1640071 1.170981
3 | .4100676 .2087486 -1.75 0.080 .1511972 1.11216
4 | .3794676 .1945061 -1.89 0.059 .1389533 1.036288
5 | .3502097 .1812705 -2.03 0.043 .1269818 .9658616
6 | .2769887 .1464696 -2.43 0.015 .0982546 .7808566
7 | .3876159 .2053056 -1.79 0.074 .1372625 1.094589
8 | .2391465 .1358139 -2.52 0.012 .0785701 .7278982
9 | .3060686 .1782928 -2.03 0.042 .0977174 .9586621
10 | .147737 .113404 -2.49 0.013 .0328171 .6650857
11 | .1109004 .1242863 -1.96 0.050 .0123308 .9974155
12 | .2688244 .2337006 -1.51 0.131 .0489196 1.477253
EducationMother | REF=Secondary
No educ | 1.374413 .1239197 3.53 0.000 1.151785 1.640073
Primary | 1.421881 .0975215 5.13 0.000 1.243033 1.626463
Tertiary | .5686868 .0798323 -4.02 0.000 .4318983 .7487983
DK | .6584451 .1363105 -2.02 0.044 .4388387 .9879482
Mother | REF=mother resident
mother non-res | 1.061693 .0743917 0.85 0.393 .9254564 1.217984
mother dead | 9.675754 1.942289 11.31 0.000 6.528547 14.34013
OlderSibling | REF=no older sibling
o sibling resident | 2.040887 1.02661 1.42 0.156 .7614588 5.470054
o sibling non-res | 3.429269 1.743713 2.42 0.015 1.265852 9.290093
o sibling dead | 13.87714 8.240796 4.43 0.000 4.333342 44.44032
YoungerSibling | REF=no younger sibling
y sibling resident | .9813923 .1693335 -0.11 0.913 .6997982 1.376298
y sibling non-res | 1.134426 .0627541 2.28 0.023 1.017863 1.264337
y sibling dead | 3.75815 1.908307 2.61 0.009 1.389161 10.16707
Presented in Session 1185: Mortality and Longevity