Joel Watson - Perfect Bayesian Equilibrium: Consistency Conditions for Practical Definitions

Seminars
(joint with Department of Decision Sciences)
Speakers
Joel Watson
12:45pm - 2:00pm
Room 3-E4-SR03

Abstract: 

"This paper develops and compares consistency conditions for belief-based solution concepts of noncooperative games, particularly perfect Bayesian equilibrium (PBE).  The paper promotes flexible application of a condition called plain consistency, which requires independent updating across dimensions of the strategy space conditional on “conjunctive” events.  For practicality, plain consistency constrains updating only on pairs of consecutive information sets for a player, and it can be imposed narrowly.  The concept is defined for infinite games, strengthening the foundation of PBE in settings where sequential equilibrium is not defined.  Implications of plain consistency are worked out, including the relation to subgame perfection, a version of “no signaling what you don't know,” and the structure of beliefs in a wide class of sender/designer games.  A key element of the approach taken herein is to express a player's belief at an information set as a probability distribution on strategy profiles."

For further information please contact michela.raganato@unibocconi.it