The Effect of the Changing Composition of Births on Infant Mortality in Africa

Mark S. Mather, Population Reference Bureau
Peter J. Donaldson, Population Reference Bureau

After 50 years of more or less steady decline, infant mortality rates in some African countries appear to have reached a plateau and may have increased recently (Rutstein 2000). The report of the 1998 Kenya Demographic and Health Survey (NCPD 1998:91), for example, notes that "The 1998 KDHS data, in combination with similarly collected data from the 1993 KDHS, provide evidence of a worsening [infant and under-5] mortality situation in the 1990s." Similar trends have been observed in the Demographic and Health Surveys completed in Burkina Faso, Ivory Coast, Cameroon, Mali, Senegal, and Zimbabwe. Reasons for the Worsening Mortality The spread of HIV/AIDS is the factor most frequently mentioned as precipitating the worsening of mortality among infant and young children in Africa. Given the high and in most African countries growing HIV infection rates, some experts think that mother-to-child transmission of HIV might be an important cause of the increase. Increases in the incidence of malaria and other infectious diseases are another possible explanation for the observed changes. Other explanations for the apparent worsening of infant mortality include increasing poverty and therefore the lessening of resources for families and health institutions to safeguard the health of children. This situation is made worse for almost all African health institutions and for many families by the HIV/AIDS pandemic, which is taking resources that might otherwise be devoted to addressing child health issues. Some observers cite the negative impact of structural adjustment programs, which they say have required government to reduce expenditures on public health and medical care that could help reduce infant mortality. Others mention increases in the problems of governance that limit what healthcare systems can do to protect children's health. There is another possible explanation for the changes in infant mortality that has not been frequently discussed. Knowles (2001) has suggested that infant mortality rates may have increased at least in part because the composition of births has changed over time. To the extent that fertility has declined in Kenya and elsewhere in Africa, the change has been greatest among economically better off women. Because there are large differences in infant mortality rates by economic status, it may be that infant mortality is increasing because a larger proportion of births are to poor women whose children have higher infant mortality rates. If there were a large difference in infant mortality between rich and poor and a shift in births to poor women, a nation's average infant mortality rate may increase, even if the wealth-specific infant mortality rates among less well off women were improving. Potter (1988: 181) anticipated such an increase in infant mortality because of changes in what he termed "the social composition of births," noting that a decline in fertility among more highly educated, better-off mothers may be unfavorable because "[a] smaller proportion of children would be born into the families or households best equipped to take care of them, and thus infant mortality would increase if other factors remained unchanged." In this paper, we investigate the effects of the changing composition of births on infant mortality in Burkina Faso, Ivory Coast, and Kenya. Data and Methods We will use data from selected African Demographic and Health Surveys for the analysis. Each of the countries studied has conducted at least two DHS surveys. The DHS surveys do not collect data on household income or consumption expenditures but do ask about the ownership of various assets and the characteristics of housing. We will use principal components factor analysis to construct an index of household wealth, following procedures developed by Filmer and Pritchett (1998; 2001) and used by Gwatkin et al (2000a and b). First, we will replicated the calculations done by Gwatkin et al (2000) and then will calculate a new index using the same procedures using data from earlier surveys. We will use the index to create wealth quintiles. We will calculate standardized infant mortality rates, controlling for fertility and wealth in order to estimate the impact of changing fertility on infant mortality. We will also examine differences by education and, where appropriate, by region. Preliminary results indicate that the changing composition of births was not a major factor in the increasing infant mortality rates in the Ivory Coast. Instead, increases in infant mortality among poor families seem most important.