Reputation and Equilibrium Selection in Games with a Patient Player (Q3468907)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 4135746
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Reputation and Equilibrium Selection in Games with a Patient Player |
scientific article; zbMATH DE number 4135746 |
Statements
Reputation and Equilibrium Selection in Games with a Patient Player (English)
0 references
1989
0 references
repeated game
0 references
incomplete information
0 references
long-run player
0 references
Stackelberg strategy
0 references
sequential equilibria
0 references
0.90866881608963
0 references
0.8995962738990784
0 references
0.8871212601661682
0 references
0.8849142789840698
0 references
0.8835822939872742
0 references
A certain type of repeated game with incomplete information is considered. There are two players: player 1 is a long-run player, completely informed about his own payoff function as well as player 2's payoff function. Player 2 has only probabilistic information about player 1's payoff function. The payoff function of player 1 is determined by his `type' \(\omega\in \Omega\) (\(\Omega\) is a finite set) and the players 2 have an initial belief about player 1's type given by a probability \(\mu\) on \(\Omega\). Player 1 is interested in the discounted payoff \((1- \delta)\sum^{\infty}_{t=0}\delta^ tg^ t_ 1\) where \(\delta\in (0,1)\) is the discount factor and \(g^ t_ 1\) is the payoff of player 1 in the t-th stage of the game. Player 2 is only interested in a high payoff in every stage. Let \(\omega_ 0\) be player 1's real type and \(s^*_ 1\) be the strategy which player 1 appreciates most (his Stackelberg strategy). If the event \(\omega_ 0\) has positive a priori probability, the event \(\omega^*\) that the Stackelberg strategy \(s^*_ 1\) is a strictly dominant stategy has positive a priori probability \(\mu *>0\) and the stage game is played with simultaneous moves, then there is a number \(k(\mu*)\) such that every equilibrium payoff is at least \(\delta^{k(\mu*)}g^*_ 1+(1-\delta^{k(\mu *)})\) \(\min g_ 1\) where \(g^*_ 1\) is the Stackelberg payoff \(\min_{s_ 2\in B(s^*_ 1)}g_ 1(s^*_ 1,s_ 2)\) (i.e., player 2 is a Stackelberg follower but player 1 is pessimistic about the choice of player 2 in the set \(B(s^*_ 1)\), the best replies of player 2 to the strategy \(s^*_ 1)\). As a consequence of this result there is for every given \(\mu *>0\) a bound \({\bar \delta}<1\) such that \[ (g^*_ 1- g_ 1)/(g^*_ 1-\min g_ 1)<\epsilon \] if \({\bar \delta}<\delta <1\) and \(g_ 1\) is an equilibrium payoff. This means that if player 2 gives some belief to the event that the strategy \(s^*_ 1\) is a strictly dominant strategy for player 1 and player 1 behaves like a player of such a type, he can build a reputation that enforces player 2 after a while to become a Stackelberg follower. This is the main result of this paper. Further the paper contains a number of nice examples (Selten's chain store e.g.) and extensions to games with more than one short-run player and a compact metric `type' space \(\Omega\). In case the stage game has sequential moves the authors show - by an example - that player 1 may not be able to build his reputation (since his actions are observationally equivalent) and their are even sequential equilibria with a much lower payoff than the theorem predicts. This case is solved by a technical change in the main theorem.
0 references