Autor: Miranda Zanetti Maximiliano, Tohmé Fernando Abel
JEL: C7, C5
The concept of Perfect Bayesian Equilibrium [PBE] refines Bayes-Nash equilibria embodying notions of sequential rationality. However, on one hand it leaves beliefs totally unrestricted for information sets off the equilibrium path. On the other hand, it lacks structural robustness, as even the same game modelled with different extensive forms can have a different set of PBE. The sequential equilibrium [SE] stands out as a suitable refinement that has more structural properties and gives place to reasonable off-path beliefs. However, SE requires different agents using exactly the same perturbations to generate consistent beliefs off equilibrium path, which may be too strict a requirement. Furthermore, testing whether a certain assesment could be a potential SE is cumbersome, and involves in principle finding the corresponding sequence of perturbations or completely mixed strategies. In this paper, we develop a generalization of SE, the robust equilibrium [RE], which refines PBE in a less restrictive way, and which computation can be algoritmically carried out solving a relatively simple set of inequalities. Although it can be described as an assesment consistent in a more general way by the use of sequences of perturbations, it can also be characterized by an assortment of nodes according to equilibrium behavior strategies.