The United Nations publishes estimates of HIV prevalence, AIDS mortality, and orphan numbers for all countries of the world. It is important to assess the validity of these model-based estimates since they underpin much policy concerned with care and prevention. Household surveys that ask questions about the survival of children's parents (orphanhood) offer an independent source of data with which these estimates can be compared. Survey estimates of maternal and paternal orphans are significantly lower than model estimates for 40 surveys in 36 countries of sub-Saharan Africa (p < 0.001, p = 0.002). This is probably because adult mortality from causes other than AIDS is lower than assumed in the models, although under-reporting of orphanhood in surveys may also play a role. Reducing adult mortality from causes other than AIDS brings the model estimates into close agreement with the surveys. This suggests that the fraction of orphans attributable to AIDS is greater than estimated previously.