Measuring substitution bias in international comparisons based on additivepurchasing power parity methods

This paper shows that additive purchasing power parity (PPP) methods, such as Geary-Khamis, are subject to substitution bias. As a result, Geary-Khami s comparisons tend to underestimate per capita income differentials across countries. The magnitude of the bias is measured for a cross-section of 64 countries, using International Comparison Programme (ICP) data. Geary-Khami s overestimates some poorer countries' per capita incomes, relative to rich er countries, by as much as 70%. These results have important implications for much empirical work in development and international economics, since G eary-Khamis is the PPP method used by the World Bank, IMF, OECD, ICP and th e Penn World Table. (C) 2000 Elsevier Science B.V. All rights reserved.