Despite the many useful applications of power indices, the literature on power indices is raft with counterintuitive results or paradoxes, as well as real-life institutions that exhibit these behaviors. This has led to a cataloging of sorts where new and different paradoxes are calculated and then shown to exist in nature. Even though the paradoxes sound different from one another with names like the paradox of redistribution, the donor and transfer paradoxes, the paradox of quarreling members, the paradox of a new member, and the paradox of large size, they can be classified by the underlying geometric properties that induce the counterintuitive results. Perhaps surprisingly, analyzing the geometry behind the paradoxes for three voters is sufficient to understand the geometry behind the paradoxes. Voting power induces a partition on games where two games are in the same part if each player i has the same power in each game. The paradoxes are a result of three geometric ideas and how they interact with the partition: a point passing a hyperplane thereby changing parts, moving hyperplanes that change the size or number of parts in a partition, and changing the dimension of the space by adding or subtracting a voter.Key words : Voting Power, Paradoxes, Geometry Power indices are used to measure the a priori distribution of power among voters under a given voting rule. Many of these power indices uniquely satisfy different sets of axioms, including the most commonly used indices by Penrose (1946), Shapley and Shubik (1954) and Banzhaf (1965), as well as others. Such axiomatic approaches have been used to create new indices, as well as to champion one power index over others. Generalized power indices, such as semivalues (Carreras, Freixas, and Puente, 2003, Laruelle and Valenciano, 2003a, and Saari and Sieberg, 2001, measure power in broader classes of cooperative games, often following the same axiomatic development. * Department of Mathematical Sciences, Montclair State University, Montclair, NJ 07013 USA. jonesm@mail.montclair.edu
193The Geometry Behind Paradoxes of Voting Power As a tool, power indices have been used to examine weighted voting in institutions including the International Monetary Fund (Dreyer andSchotter, 1980 andLeech, 2002c), the Electoral College (Mann and Shapley, 1964), the European Union Council of Ministers (Johnston, 1995 andLeech, 2002b), and the Israeli Knesset (Laruelle, 2001). Not only have power indices been used to analyze existing institutions, but they have been part of the debate about the design of new institutions. For example, Turnovec (1996) and Widgren (1994) use power indices to model the effects of institutional reforms on, and the introduction of new members into, the European Union. Because power indices rarely agree on the measure of power for a voter, let alone on the ranking of the power of voters (cf. Saari and Sieberg, 2001), the selection of a power index is paramount. Although a productive way to generate power indices, the axiomatic approach has not bee...