Page 27 - IJPS-3-2

Upadhyay AK et al.

obtained by children from non-migrant households.
Table 1 describes the bivariate analysis of child characteristics and maternal/household characteristics according to the
status of forced migration in Andhra Pradesh (India). Results show that about 6.2% households had migrated since the
pregnancy of the index child due to some catastrophic event following Wave 1. The information on the migration status
of 1913 children interviewed in Wave 3 showed that around 29.4% of the children from the migrant households were
below average size at birth, whereas only 25.5% of the children from the non-migrant households were below average
size at birth. 87.4% of the migrant households and 79.7% of the non-migrant households had received medium/high
social support. A higher percentage of children of non-migrant mothers had been breastfed compared to those of migrant
mothers. Around 28.6% children from migrant households had suffered a serious illness/injury in wave 1, whereas only
21.7% children had done so in case of non-migrant households. About 33.9% migrant children were stunted, while only
29.7% non-migrant children experienced stunting. Migrant and non-migrant households enjoyed equal coverage of
immunization and pre-schooling. The inter-class comparison of mother’s education for both migrant and non-migrant
households revealed that around 85.7% of the migrant mothers had below primary level education compared to only
58.7% of the non-migrant mothers. The migrant mothers were primarily engaged in agricultural work, while the majority
of the non-migrant mothers were non-working. The majority of the migrant households lacked access to improved
drinking water and improved toilet facility. About 94.1% of the migrant households belonged to a rural area compared to
a lower figure of 72.4% in the case of non-migrant rural households.
Table 2 describes the results of multivariate linear regression analysis examining the effect of forced migration on child
cognitive well-being in Andhra Pradesh, India. Model I presents the adjusted regression coefficients of all the scores for
all the predictor variables except social support. Findings of Model I show that children born to migrant households were
significantly less likely to achieve higher math (Coefficient: -2.008, 95% CI: -3.108, -0.908), EGRA (Coefficient: -0.746,
95% CI: -1.366, -0.126) and memory scores (Coefficient: -0.503, 95% CI: -0.834, -0.173) compared to the children of
non-migrant households. Model II of Table 2 presents the adjusted regression coefficients of all the scores for all the
predictor variables including social support. The effect of social support was visible only in the case of PPVT scores. The
children from households that had received medium/high social support were significantly more likely to achieve higher
PPVT scores (Coefficient: 0.081, 95% CI: 0.030, 0.133) compared to those from households that had received low social
support. Notably, in this model also, the children born to migrant households were significantly less likely to achieve
higher math, EGRA and memory scores compared to the children of non-migrant households. By comparing Model
I and Model II, we found that the magnitude of the effect of forced migration on child cognitive well-being remained
unchanged even after controlling the social support.
A number of other variables were found to be statistically associated with the cognitive well-being of children. Preterm
born children were significantly less likely to achieve higher PPVT scores (Coefficient: -0.101, 95% CI: -0.173, -0.030)
compared to the children of full-term pregnancy. Age of children was also found to be significantly associated with the
cognitive well-being, with the older children being more likely to achieve higher PPVT, math, EGRA and memory scores.
Interestingly, female children were significantly less likely to achieve higher PPVT scores (Coefficient: -0.100, 95% CI:
-0.140, -0.059), math scores (Coefficient: -0.539, 95% CI: -1.070, -0.008) and memory scores (Coefficient: -0.198, 95%
CI: -0.358, -0.037) compared to the male children. Children who had suffered from a serious illness/injury during Wave
1 were significantly less likely to achieve higher PPVT scores (Coefficient: -0.062, 95% CI: -0.111, -0.014), math scores
(Coefficient: -0.647, 95% CI: -1.283, -0.012) and EGRA scores (Coefficient: -0.388, 95% CI: -0.746, -0.029) than their
counterparts. Stunted children during Wave 1 were significantly less likely to achieve higher PPVT, math, EGRA and
memory scores compared to the non-stunted children during the same wave. It is worth noticing that the fully immunized
children were significantly more likely to obtain higher math scores (Coefficient: 1.768, 95% CI: 0.757, 2.780) and EGRA
scores (Coefficient: 0.577, 95% CI: 0.005, 1.149) compared to the not/partially immunized children. Children studying
in government schools were significantly more likely to obtain higher EGRA and memory scores compared to the ones
studying in private schools. Children whose mothers had primary and higher levels of education were significantly more
likely to achieve higher PPVT, math and EGRA scores than those children whose mothers had below primary level
education. The math, EGRA and memory scores were significantly higher among children of urban settings than those
belonging to rural settings.
The estimates obtained from multivariate linear regression may have a potential selection bias due to the differences in

International Journal of Population Studies | 2017, Volume 3, Issue 2 21