Co-ordination, spillovers, and cheap talk (Q697848)

The author analyzes a two-person non-zero-sum cheap talk game described by the following model: Each Player \(i\) has a finite set \(A_i\) of actions, and the game depends also on some parameter \(t\) (known only to Player 1), chosen with a fixed probability distribution \(\pi\) over a finite type set \(T\) (\(\pi \in \Delta(T))\). The utility function of Player 1 (informed player) is \(u_1: A\times T \to R\) (here \(A = A_1\times A_2\)), while for Player 2 (uninformed player) is \(u_2: A \to R\). A cheap talk stage is added, that is, there is a discrete message space \(M\) for Player 1, such that his general strategy is every pair \((\mu,\alpha_1)\), where \(\mu: T \to \Delta(M)\) and \(\alpha_1: M \times T \to \Delta(A_1)\) are his talking strategy and action strategy, respectively. Distribution \(\mu (t)\) and any Player 1's strategy \((\mu, \alpha_1)\) result in a triplet \((m,a_1,t) \in M\times A_1 \times T\), where the realization \(m\) of \(\mu (t)\) over \(M\) is announced to Player 2. General strategies of Player 2 are the pairs \((\alpha_2,\lambda)\), where \(\alpha_2: M \to \Delta(A_2)\) and \(\lambda_1: M \to \Delta(T)\). For such a game, perfect Bayesian equilibria \((\mu, \alpha_1, \alpha_2, \lambda)\) are defined and studied in two versions of the model: with full communication (type \(t\) is known also to Player 2) and without communication. The author gives relationships between perfect Bayesian equilibria, Nash equilibria and the so-called full revelation equilibria. Several examples illustrate the obtained results.