Page 86 - IJPS-10-4

International Journal of
Population Studies Migration and child mortality estimation

are common in data from small areas, such as rural or the data from non-migrant women were different from
urban regions within a country. In addition, Trussell’s those computed using data from both in-migrant women
regression equations are derived from a two parameter and non-migrant women. During comparisons of the two
model, as opposed to Palloni-Heligman equations, which sets of mortality estimates, the estimates from the non-
are derived from one parameter model. For the choice of migrant population were assumed to optimally describe
the estimation model, we proposed the use of the mean the mortality for each of the two regions.
standard error of under-five mortality based on six sets of
estimates of the non-migrant population. This was done 3. Results
after omitting estimates associated with women aged 15 Tables 2 and 3 show the infant ( ), one-to-four ( ), and
1 0
4 1
– 19 years because they are highly erratic (Neupert et al., under-five ( ) mortality rates based on the North life
5 0
2019). The model with the least mean standard error was table model. This model provided estimates with the least
considered to produce less erratic and more consistent mean standard error of under-five mortality, as described
trends in mortality estimates. Mortality estimates from in Section 2.3. The mortality estimates computed using
non-migrant women were considered because they were data from non-migrant women are given in the column
not affected by the deaths of children that occurred in labeled “non-migrant mortality estimate”, while the
another region. mortality estimates computed using both non-migrant

2.4. Analysis procedure and in-migrant women are given in the column labeled
“combined mortality estimate”. The column labeled “%
Our study focused on establishing the impact of migration Diff” includes the relative percent difference in mortality
on indirect estimates of child mortality from a statistical estimates for various age groups due to the inclusion of
perspective. To describe this impact, we computed two deceased children born to in-migrant women. Except for
sets of child mortality estimates per region: those based urban mortality estimates from the 1993 survey, the overall
on deceased children born to all women interviewed in effect of migration was an overestimation of the rates in
the region r, denoted as *r, and those based on deceased both regions.

children born to non-migrant women, denoted as .

Since death is a rare occurrence, there were few cases of We now answer the question of whether the observed
children reported dead born to migrant women when overestimation of childhood mortality is statistically
classified by their mothers’ 5-year age group. When the significant. In other words, we examine if the inclusion
number of dead children is few, the mortality estimates of deceased children born to in-migrant women altered
have large variances, making the estimates less robust. For the child mortality estimates in the recipient region.
this reason, mortality estimates from both in-migrant and Specifically, we tested the hypothesis that “There is no
out-migrant women were not considered for analysis. The significant difference in child mortality estimates computed
mortality estimates from non-migrant women described using data from non-migrant women and from combined
the mortality levels of a given region if the assumption of data of non-migrant and in-migrant women.” To do this, we
closed migration holds. On the other hand, the estimates compared the estimates under the NME and CME columns
from combined women provide child mortality levels of Tables 2 and 3 using a paired sample t-test. The results of
when the assumption is violated due to migration from this test are presented in Table 4. These results indicate that
regions with different levels of mortality regimes. in the urban region, all three child mortality rates obtained
using data from non-migrant and in-migrant women
The difference between the two mortality rates,
expressed as a percentage of the mortality rate of combined were statistically higher compared to those
deceased children born to non-migrant women, is the obtained using data from non-migrant women (p = 0.02).
This statistical significance means that the inclusion of
relative percent effect due to migration. Thus, the child deceased children born to rural-to-urban migrant women
mortality estimate effect due to migration, ℓ, is given by led to an overestimation of the infant, child, and under-five
Equation VIII:
mortality rates in the urban region. On the other hand, there
ℓ = [( * − )/ ] × 100 (VIII) was no sufficient evidence to conclude that urban-to-rural

Next, we assessed whether the effect or differences in migrant women altered any of the three child mortality
child mortality estimates were significant when describing estimates in the rural region (p > 0.05).
the child mortality of any of the two regions. Implicitly, the 4. Discussion
effect is significant if the two sets of mortality estimates are
statistically different. For statistical evidence, we used a A major issue associated with including in-migrant women
pairwise t-test to examine if the estimates computed using in the computation of child mortality for any region is the

Volume 10 Issue 4 (2024) 80 https://doi.org/10.36922/ijps.1837