We present theory and evidence to suggest that, in the context of analysing global poverty, the EKS approach to estimating purchasing power parities yields more appropriate international comparison of real incomes than the Geary-Khamis approach. Our analysis of the 1996 International Comparison Project data confirms that the Geary-Khamis approach leads to substantial overstatement of the relative incomes of the world’s poorest nations and to misleading comparisons of poverty rates across regions. Similar bias is found in the Penn World Table which uses a modified version of the Geary-Khamis approach. Estimation of both the level of global poverty and its location is very sensitive to the choice of index. The EKS index of real income is much closer to being a true index of economic welfare and is therefore to be preferred for assessment of global poverty.