A MATHEMATICAL ANALYSIS OF CONFLICTS IN VOTING SYSTEMS

Game Theory applied to voting gives rise to the study of many different topics. In this thesis we present some results in order to evaluate the power share inside a Parliament, to analyze the criteria for the assessment of the voters’ preferences and to provide more efficient ways to compute the existing instruments. The first part of the thesis deals with the issue of power in a voting system. In order to better analyze it, we investigate the way of combining a communication structure with the already existing indices of power. Moreover, we present the possibility of assuming some coalitions as less probable, but not infeasible, as the political scenario is complex enough to admit the possibility of very unlikely coalitions to form. Then, we add another important aspect to the problem; the alliances inside a decisional situation are not stable, but may evolve with time, mainly due to the fact that each agent aims at getting a higher power. We refer to an already existing instrument which, theoretically, may perfectly describe this situation. Unfortunately, the current computational complexity does not permit to adopt it in many real cases. Another important topic is the evaluation of the power to block instead of the power to win. We provide an index which evaluates the veto power and which can be extended in order to catch other characteristics of the game; for example, the probability that a party takes a particular decision, or that not every member of a party is present at the vote. The second part of this thesis considers a previous step of the formation of a democracy: the evaluation of how much the resulting Parliament reflects the preferences expressed by the electors. We think that the goodness of a Parliament mainly depends on the power share between the parties and not only on the number of seats. Finally, the last part shows a new exact method to improve the efficiency of the evaluation of one of the existing indices of power, the Public Good index.