Magisterial district childhood mortality in South Africa: an application of geographically weighted regression approach

Martin Bangha, University of Pennsylvania

In this paper, we examine magisterial district differences in childhood mortality. Using data from the 2001 census, we employ exploratory spatial analysis and geographically weighted regression (GWR) approach. The GWR approach is particularly useful in this case because it provides additional important information on the nature of the observed disparities that would otherwise be ignored by the simple regression model. Consistent with the complex geographical area hierarchy maintained by South Africa, exploratory analysis are conducted at the level of the 354 magisterial districts (MDs) and 54 district municipal councils (DCs). Geographically weighted regression models are fitted at the Magisterial District level to examine the contribution of household poverty and other attributes to the observed district variation in childhood mortality. Childhood mortality risk is measured by the probability of dying before age 5.