Static ggames: Rationalizabilityy Lectures in Game Theory Spring 2011, Part 2 10.12.2010 G.B. Asheim, ECON4240-2 1 Questions to answer: How do players behave in static strategic situations? How should players behave in such situations? Analysis: Apply the normal form (as a representation of games where all actions are taken simultaneously and independently). Model ‘rational behavior’. 10.12.2010 G.B. Asheim, ECON4240-2 2 1 Strict dominance Some examples: Definition : A p pure strategy gy si of p player y i is strictlyy dominated if there is a stragegy (pure or mixed) i Si such that ui ( i , si ) ui ( si , si ) for all strategy profile si S i of his opponents. Write UDi Set of undominate d strategies . gy is strictlyy dominated if there Comments: — A strategy is another strategy that is better for all opponent choices. — The dominating strategy may be mixed. — Weak dominance is not sufficient. 10.12.2010 G.B. Asheim, ECON4240-2 3 Simple theory of individually rational behavior: Players do not use strictly dominated strategies. Player i chooses some si UDi First tension: Individually rational behavior may not lead to collectively efficient outcomes. The concept of efficiency Definition : A strategy profile s ( s1 , , si , , sn ) is (Pareto) efficient if there is no other strategy profile s ( s1, , si, , sn ) such that ui ( s) ui ( s ) for every player i and u j ( s) u j ( s ) for some player j. 10.12.2010 G.B. Asheim, ECON4240-2 4 2 Equivalent Alternative theory of individually rational behavior: Players use strategies that can be best responses. Definition : Suppose player i has a belief i S i about the strategies played by his opponents. Player i' s strategy si Si is a best response to i if ui ( si , i ) ui ( si, i ) for every si Si . Write BRi ( i ) Set of best respo nses to i , and Bi {si | there h is a belief b l f i S i such h that h si BRi ( i )}. Player i chooses some si Bi 10.12.2010 G.B. Asheim, ECON4240-2 5 Strict dominance and best response compared Observations for two-player games: – If a strategy is not strictly dominated, then it is a best response to some belief. – If a strategy is strictly dominated, then it is not a best response to some belief. Result : In a finite two - player gam e, B1 UD1 and B2 UD2 . In ggames with more than two players, p y the equivalence between strict domianance and best response is obtained if beliefs are correlated. Result : In a finite game, Bi Bic UDi for each i 1,2, , n. 10.12.2010 G.B. Asheim, ECON4240-2 6 3 Algorithm for finding Bi UDi in two-player games Step 1: Strategies that are best b responses to simple (point mass) beliefs are in Bi. Step 2: Strategies that are dominated by other pure strategies are not in Bi. Step 3: Other strategies can be tested for mixed strategy dominance to see whether they are in Bi. 10.12.2010 G.B. Asheim, ECON4240-2 7 Sophisticated theory of individually rational behavior: Iterated strict dominance (Iterated elimination of strictly dominated strategies) Some examples: In two-player games, iterated strict dominance is equivalent to the procedure in which strategies that are never best responses are removed in each round. Rationalizability SStrategies i that h survive i iiterated d strict i ddominance i are called rationalizable strategies. Weak dominance is not sufficient. 10.12.2010 G.B. Asheim, ECON4240-2 8 4 Two problems with iterated strict dominance 1. It is based on an assumption that it is commonly players y choose rationally. y believed that p 2. In many games there are no strictly dominated strategies. Second tension: Common belief of individually rational behavior b may not lead d to coordination. d Question: How to analyze games where iterated strict dominance has no bite? 10.12.2010 G.B. Asheim, ECON4240-2 9 5