An association between ethnic diversity and HIV prevalence in sub-Saharan Africa.

This paper investigates whether ethnic diversity at the Demographic and Health Surveys (DHS) cluster level predicts HIV serostatus in three sub-Saharan African countries (Kenya, Malawi and Zambia), using DHS household survey and HIV biomarker data for men and women aged 15–59 collected since 2006. The analysis relates a binary dependent variable (HIV positive serostatus) and a weighted aggregate predictor variable representing the number of different ethnic groups within a DHS Statistical Enumeration Area (SEA) or cluster, which roughly corresponds to a neighbourhood. Multilevel logistic regression is used to predict HIV prevalence within each SEA, controlling for known demographic, social and behavioural predictors of HIV serostatus. The key finding was that the cluster-level ethnic diversity measure was a significant predictor of HIV serostatus in Malawi and Zambia but not in Kenya. Additional results reflected the heterogeneity of the epidemics: male gender, marriage (Kenya), number of extramarital partners in the past year (Kenya and Malawi, but probably confounded with younger age) and Muslim religion (Zambia) were associated with lower odds of positive HIV serostatus. Condom use at last intercourse (a spurious result probably reflecting endogeneity), STD in the past year, number of lifetime sexual partners, age (Malawi and Zambia), education (Zambia), urban residence (Malawi and Zambia) and employment (Kenya and Malawi) were associated with higher odds of positive serostatus. Future studies might continue to employ multilevel models and incorporate additional, more robust, controls for individual behavioural risk factors and for higher-level social and economic factors, in order to verify and further clarify the association between neighbourhood ethnic diversity and HIV serostatus.