On exiting after voting (Q2491086)

The authors assume that members of a society have preference orderings on the set of final societies, where a final society consists of an alternative, and a subset of initial members. They consider final societies to be the outcomes of a two-stage game. First, members choose an alternative \(x \in X\) by a given voting procedure. Second, after knowing the chosen alternative, members of the initial society decide whether to stay or exit. The authors first model exit procedures by a generic family of games \(\{\Gamma(x)\}_{x \in X}\), parametrized by the chosen alternative, where \(\Gamma(x)\) describes the rules under which the members have to decide their membership after \(x\) has been chosen. There are many societies whose members consider undesirable the exit of other members, independently of the chosen alternative. Preference relations that satisfy this general condition are called monotonic. Under this domain restriction the authors identify, for each chosen alternative \(x\), a final society consisting of \(x\) and the complimentary set of what they call the exit set after \(x\) is chosen, \(EA(x)\). An equilibrium of the exit procedure is called a panic equilibrium if it exhibits the feature that some members exit only because they expect that other members will exit as well, although all of them would be better off staying. The authors show that, independently of the exit procedure \(\Gamma(x)\), there is at least one equilibrium where agents in \(EA(x)\) exit and agents in \(N - EA(x)\) stay. They also show that the remaining equilibria of \(\Gamma(x)\) are panic equilibria, in which members in \(EA(x)\) exit. The authors prove that in simultaneous exit, panic equilibria can exist. If exit is sequential, for each alternative \(x \in X\), \(\Gamma(x)\) has a unique subgame perfect equilibrium outcome which corresponds to the unique non-panic equilibrium. This outcome is independent of the order in which the members reconsider their membership. In the last section it is shown that, even when preference profiles are monotonic, the two-stage game may not have equilibria. But for voting by quota, the authors present two results guaranteeing the existence of equilibria.