To effectively target projects and monitor progress, HIV prevalence data are ideally needed at the sub-national level. A few countries in Africa, including Malawi, have recently implemented DHS surveys that included the testing of adult population for HIV, and provide HIV prevalence rates at the sub-national level. However, such surveys remain very rare, and will not be conducted on a repeated basis given the challenges and costs of data collection. This paper applies, using various data sources for Malawi, a methodology that allows one to produce sub-national estimates of HIV prevalence in the absence of population-based surveys. The prediction errors of prevalence rates are estimated using bootstrap. The obtained estimates are compared with the “golden standard” of the prevalence rates derived directly from the 2004 Malawi DHS which tested adults for HIV.