Publisher/Institution: The Johns Hopkins University

Abstract: This dissertation is comprised of three studies that investigate potential issues in the assessment of the relationship between economic status and HIV infection in sub-Saharan Africa. The first study explores in what ways selection bias affects the association between economic status and HIV infection and assesses whether this effect differs by country and by sex using DHS data from Burkina Faso (2003), Cameroon (2004), Ethiopia (2005), Ghana (2003), Kenya (2003), Lesotho (2004), Malawi (2004), Rwanda (2005), and Zimbabwe (2005-2006) and AIS data from Côte d’Ivoire (2005) and Tanzania (2003-2004). Manski worst-case bounds are constructed around the proportion of HIV infected. These bounds are then used to study the potential effects of selection into HIV testing on the association between economic status and HIV infection using logistic regression. The second study investigates the association between economic status and HIV infection in rural and urban areas using asset indices constructed for rural and urban households separately using DHS data from Cameroon (2004), Kenya (2003), Lesotho (2004), Malawi (2004) and Zimbabwe (2005-2006) and AIS data from Tanzania (2003-2004). Three separate asset indices are derived for each country—a national index, an urban-specific index and a rural-specific index—and logistic regression models for the odds of HIV infection are run for urban areas, rural areas and for urban and rural areas combined. In the third study, we measure economic inequality among prime aged men and women at baseline and deaths during a thirteen-year period using data from the 1991-2004 Kagera Health and Development Survey (KHDS).