(PBE) is a game theory concept that combines strategic thinking with belief updating. It's used in games where players have incomplete information about each other, like in job market signaling or .
PBE requires players' strategies to be optimal given their beliefs, and beliefs to be consistent with strategies. This concept helps analyze complex scenarios where information is asymmetric and actions can reveal hidden information about players' types.
Perfect Bayesian Equilibrium
Definition and Requirements
Top images from around the web for Definition and Requirements
Bayes’ Theorem | Mathematics for the Liberal Arts View original
Is this image relevant?
Bayesian Probability Illustration Diagram | TikZ example View original
Is this image relevant?
Bayes’ Theorem | Mathematics for the Liberal Arts View original
Is this image relevant?
Bayesian Probability Illustration Diagram | TikZ example View original
Is this image relevant?
1 of 2
Top images from around the web for Definition and Requirements
Bayes’ Theorem | Mathematics for the Liberal Arts View original
Is this image relevant?
Bayesian Probability Illustration Diagram | TikZ example View original
Is this image relevant?
Bayes’ Theorem | Mathematics for the Liberal Arts View original
Is this image relevant?
Bayesian Probability Illustration Diagram | TikZ example View original
Is this image relevant?
1 of 2
Perfect Bayesian equilibrium (PBE) extends to games with incomplete information
PBE requires players' strategies and beliefs about other players' types to be consistent with each other and the equilibrium being played
In a PBE, players' strategies must be sequentially rational
Strategies are optimal given their beliefs at each information set
Players' beliefs are determined by Bayes' rule and the equilibrium strategies whenever possible
At information sets off the equilibrium path, beliefs must be consistent with the equilibrium strategies but are otherwise unrestricted
Applications and Examples
PBE is commonly used to analyze with asymmetric information (auctions, , )
Example: In a job market signaling game, a job seeker's education level (type) is unknown to employers
The job seeker's strategy is the level of education to obtain, while the employer's strategy is the wage to offer based on the observed education
In a PBE, the job seeker's education choice must be optimal given the employer's wage strategy, and the employer's beliefs about the job seeker's type must be consistent with the education choice and Bayes' rule
Beliefs and Bayesian Updating
Forming and Updating Beliefs
In dynamic games with incomplete information, players form beliefs about the types of other players based on their observed actions
Beliefs are probability distributions over the possible types of other players at each information set
Players update their beliefs using Bayes' rule whenever they receive new information, such as observing the actions of other players
Bayesian updating involves calculating the posterior probability of a player's type given the prior probability and the observed action
If an observed action has zero probability under the equilibrium strategies, Bayes' rule cannot be applied
In this case, beliefs are unrestricted but must be consistent with the equilibrium
Consistency requires that beliefs assign positive probability only to types that could have taken the observed action under the equilibrium strategies
Example: In a signaling game, if the equilibrium specifies that only high types take a certain action, then observing that action must lead to the belief that the player is a high type with probability 1
Perfect Bayesian Equilibria in Dynamic Games
Solving for PBE
To find a PBE, first determine the possible types of players and their probabilities, as well as the available strategies and payoffs for each type
Represent the game in extensive form, showing the moves of nature that determine players' types, the actions available to each player at each information set, and the payoffs for each outcome
Use backward induction to solve for the optimal strategies at each information set, starting from the end of the game and working backwards
At each information set, calculate the expected payoff of each action based on the player's beliefs about the other players' types and their expected future actions
Choose the action that maximizes the player's expected payoff at each information set, given their beliefs
Verifying PBE
Check that the resulting strategies and beliefs form a PBE by verifying that they are sequentially rational and consistent with each other and Bayes' rule
Sequential rationality: Players' strategies must be optimal at every information set, given their beliefs
Consistency: Beliefs must be updated according to Bayes' rule whenever possible, and must assign positive probability only to types that could have taken the observed actions under the equilibrium strategies
Example: In a two-stage signaling game, verify that the sender's strategy is optimal given the receiver's response strategy, and that the receiver's beliefs and response strategy are optimal given the sender's strategy and Bayesian updating
Perfect Bayesian Equilibrium vs Other Concepts
Comparison to Subgame Perfect Equilibrium
PBE is an extension of subgame perfect equilibrium (SPE) to games with incomplete information, where players have beliefs about each other's types
Like SPE, PBE requires that strategies are optimal at every information set, not just at the beginning of the game
However, PBE allows for imperfect information and Bayesian updating of beliefs, while SPE assumes perfect information
Relationship to Bayesian Nash Equilibrium
PBE is a refinement of (BNE), which is the Nash equilibrium of a game with incomplete information
BNE requires that strategies are optimal given beliefs, but does not require sequential rationality or off the equilibrium path
PBE is a stronger solution concept than BNE, as it imposes additional requirements on beliefs and strategies
PBE can be seen as a combination of SPE and BNE, incorporating elements of both sequential rationality and Bayesian updating in games with imperfect information