Power Relations in the IMF: A Study of the Political Economy of A Priori Voting Power Using the Theory of Simple Games

In general in organisations whose system of governance involves weighted majority voting, power and voting weight differ. Power indices are a value concept for majority voting games which provide a means of analysing this difference. This paper provides new algorithms for computing the two classical power indices (the Banzhaf index and the Shapley-Shubik index) and applies them to the voting distribution in the two governing bodies of the IMF in each year since its foundation. The focus is both substantive, being an analysis of the political economy of the IMF, and methodological, as a study of the use of the power indices. Power relations are studied with respect to two types of decisions: ordinary decisions requiring a simple majority and decisions requiring a special majority of 80% or 85%. Clear cut results are obtained for the former: among the G5 countries discrepancies between power and voting weight have declined over time with the exception of the United States which continues to have much more power than its weight even though that weight has declined. In the nineteen forties the United Kingdom?s power was considerably below its relatively large nominal voting power, similarly to some extent for France. Both power indices give results which are qualitatively comparable. For decisions requiring special majorities, however, few general results emerge because of conflict between the indices. We examine the effect of the size of the majority requirement on the power of the leading members and find that the power of the US declines as the majority requirement increases. This result confirms the warnings of Keynes that the US insistence on retaining a national veto for itself might be counterproductive. We conclude that the special majority requirement creates a distortion in the voting system which can be regarded as a serious lack of transparency. We also examine the effect of the EU countries voting as a block rather than individually and show that it would be dominant and the US power would fall considerably. We conclude that it is not possible to make a clear choice between the two power indices used but that there is some indication that the Banzhaf index may be more plausible.