Effects of Family Composition on Mortality Differentials by Sex Among Children in Matlab, Bangladesh

Abstract

The Matlab Demographic Surveillance System of the International Center for Diarrheal Disease Research Bangladesh (ICDDRB) in 1981-82 was used to determine whether sex differences in mortality varied with socioeconomic family conditions or whether major maternal and child health interventions reduced the effect. It was hypothesized that large numbers of children of the same sex in families will be associated with higher child mortality of that sex. Socioeconomic data were obtained from the 1982 Matlab Census. Univariate models were developed and parameters estimated with the PROC LOGIST procedure of the Statistical Analysis System (SAS). Sex differentials appear at >6 months. The total number of observations pooled across age intervals (.5-1 and annually to 5 years) is 57937. 50% of the births occurred in the treatment area. The results for the univariate tabulations show that girls with surviving older sisters have at every age higher death rates than girls without older siblings or with older sons. Having >2 sisters is not a greater disadvantage but having mixed-sex sibships does increase girls mortality. Boys with older male siblings follow the same pattern but to a lesser extent. Girls mortality is more than double that of boys at all 3 ages. There is a high mortality rate for males up to 1 year. Having an older brother in addition to an older sister increases mortality in only 2 out of 6 cases but having more than 1 older brother is worse than having only 1. In the multivariate analysis the log odds of dying are higher for females by .436 and the presence of 1 older sister raises mortality significantly to .558 vs. no siblings (>2 sisters is .573). The log odds of dying are higher for sisters having older brothers vs. no siblings or boys having older brothers. Boys with older sisters have lower mortality. The addition of socioeconomic and residential variables has little effect on the family composition coefficients. Child mortality is higher among the following groups: families with very little dwelling space per capita; families that do not own a radio bicycle watch quilt and lamp or received remittances; and families with an unschooled mother. Treatment area residence significantly lowers child mortality for both sexes equally. Mothers with some schooling offer the same advantage as living in a treatment area. But the presence of an older sister is the most important determinant. Female mortality is somewhat lower in poorer families. The effects of the family planning program on birth order distributions are likely to accumulate over time. The major conclusion is that high female mortality is due to conscious selective neglect.