Game Theory and Strategic Behaviour

Embed Size (px)

Citation preview

  • 8/3/2019 Game Theory and Strategic Behaviour

    1/27

    Game Theory and Strategic

    BehaviourVani Borooah

    University of Ulster

  • 8/3/2019 Game Theory and Strategic Behaviour

    2/27

    Confess Deny

    Confess -5,-5 0,-10

    Deny -10,0 -1,-1PrisonerX

    Prisoner Y

  • 8/3/2019 Game Theory and Strategic Behaviour

    3/27

    Elements of a Game:

    Players: agents participating in the game (X and Y))

    Strategy Set: Actions that each player may take under anypossible circumstance (Confess, Deny)

    Strategy: An action that a player takes

    Outcomes: The various possible results of the game (four,each represented by one cell of matrix)

    Payoffs: The cost/benefit that each player gets from eachpossible outcome of the game (the prison sentences enteredin each cell of the matrix)

  • 8/3/2019 Game Theory and Strategic Behaviour

    4/27

    Points to note

    1. The Payoff to a player depends not juston his chosen action but also on the actionchosen by the other player

    So, there is inter-dependence2. Each player chooses his action, without

    knowing the others choice

    It is a non-co-operative game

    3. It is played once: it is a one-shot game

  • 8/3/2019 Game Theory and Strategic Behaviour

    5/27

    A Solution to a Game

    Each player chooses a strategy that gives himthe highest payoff, given the choices of others

    If no player wishes to change his strategy, given

    the choice of the others, the game has a solutionand the players arrive at an equilibrium

    The solution is the set of strategies chosen bythe players

    Question is: will a game have a solution?

  • 8/3/2019 Game Theory and Strategic Behaviour

    6/27

    Dominant Strategy

    A dominating strategy is one which gives theplayers his highest payoff, regardless of theother players choice of action

    In the PD game, confess is the dominating

    strategy for X:1. If Y denies, X is better off confessing (0 years)2. If Y confesses, X is better off confessing (5

    years) Confess is also the dominating strategy for Y

    Solution to the game is for both to confessand spend 5 years in jail

  • 8/3/2019 Game Theory and Strategic Behaviour

    7/27

    Collective versus self-interest

    For both to confess (5 years in jail foreach) is not the optimal strategy

    The optimal strategy (1 year in jail for

    each) is for both to deny But the optimal strategy is inaccessible,

    because each cannot be sure what theother will choose: there is an absence oftrust!

  • 8/3/2019 Game Theory and Strategic Behaviour

    8/27

    Build Dont Build

    Build 16, 16 20, 15

    Dont Build 15, 20 18, 18GeneralMotors

    Ford

  • 8/3/2019 Game Theory and Strategic Behaviour

    9/27

    Solution to the Ford-GM Game

    The strategy build dominates the strategydont build

    But the optimal strategy is dont build

    Think of India and Pakistan building a nuclearbomb

    Same conclusion: build a bomb is thedominant strategy, but dont build is the beststrategy

    But absence of trust means that best strategyis not chosen

  • 8/3/2019 Game Theory and Strategic Behaviour

    10/27

    Dominated Strategy

    The opposite of a dominant strategy is adominated strategy

    A dominated strategy will always give a

    player a lower payoff than the alternativestrategies, regardless of the strategychoice of the other players

    A rational player will never play adominated strategy

  • 8/3/2019 Game Theory and Strategic Behaviour

    11/27

    In the game alongside, Hondadoes not have a dominantstrategy: if Toyota builds, it isbetter for Honda to not build; if

    Toyta does not build, it is betterfor Honda to build

    But Toyata has a dominantstrategy: whether Honda buildsor not, it is better for Toyota tobuild

    So, Toyota will build and Hondawill not build: SOLUTION!

    1. Both players do not need adominant strategy. If one playerhas a dominant strategy, itsother (dominated) strategy canbe eliminated

    2. Game theory teaches us to putourselves in the mind of ourrival Honda argues thatToyota will build and so Hondashould not build

    18,515,6

    20,312,4

    Toyota

    Build

    Builld Do not Build

    Donotbuild

    Honda

    Elimination of Dominated strategies

  • 8/3/2019 Game Theory and Strategic Behaviour

    12/27

    18,1815,209,18

    20,1516,168,12

    18,912,80,0

    Build large Build small Do not build

    Build large

    Build small

    Do not build

    Toyota

    Honda

    Neither Toyota nor Honda has a dominant strategy

    But, for Toyota, build large is dominated by other strategies:Toyota will never choose build large

    For Honda, build large is dominated by other strategies:Honda will never choose build large

    So build large can be eliminated for both

    Elimination of Dominated Strategies

  • 8/3/2019 Game Theory and Strategic Behaviour

    13/27

    In the game alongside, Hondadoes have a dominant strategy:if Toyota builds small, it isbetter for Honda to build small;

    if Toyota does not build, it isbetter for Honda to build small

    Toyota has a dominantstrategy: whether Honda buildssmall or does not build, it isbetter for Toyota to build small

    So, Toyota and Honda will bothbuild small: SOLUTION!

    1. Both players do not need adominant strategy to start with.

    2. However, after eliminating theirdominated strategies, they do

    have dominant strategies

    18,1815,20

    20,1516,16

    Toyota

    Build

    Small Build Small Do not Build

    Donotbuild

    Honda

    Elimination of Dominated strategies

  • 8/3/2019 Game Theory and Strategic Behaviour

    14/27

    Obtaining a Solution to a Game

    If both players have a dominantstrategy, theywill choose it: Prisoners Dilemma

    If one player has a dominantstrategy, he will

    choose it, the other players strategy will be hisbest response to this (Toyota-Honda)

    If neither player has a dominant strategy,successively eliminate each players dominatedstrategies to simplify the game and look fordominantstrategies

  • 8/3/2019 Game Theory and Strategic Behaviour

    15/27

    Nash Equilibrium

    A dominant strategy for a player has to be hisbest strategy for every choice of strategy by therival

    This is a very demanding requirement and very

    often it will not be met, even after eliminatingdominated strategies A weaker requirement is that a players choice is

    best for the best choice of his rival A pair of strategies is a Nash equilibrium if

    player 1s choice is optimal, given 2s optimalchoice and player 2s choice is optimal, given 1soptimal choice

  • 8/3/2019 Game Theory and Strategic Behaviour

    16/27

    Finding a Nash Equilibrium

    If a dominant cant be found, identify eachplayers best strategy, given the strategyof the other player

    The combination of best strategies is theNash equilibrium

  • 8/3/2019 Game Theory and Strategic Behaviour

    17/27

    5,44,1112,14

    15,20,03,10

    1,313,64,2

    Player 2

    Strategy D Strategy E Strategy F

    Strategy A

    Strategy B

    Strategy C

    Player1

    Neither player has a dominating nor a dominated strategy

    For each player we identify his best strategy, given the otherplayers strategy: red for player 1, green for player 2

    Two equilibrium points: (A,E) or (C,D)

  • 8/3/2019 Game Theory and Strategic Behaviour

    18/27

    1,20,0

    0,02,1

    Player 2

    Left Right

    Top

    BottomPlayer1

    Neither player has a dominating nor a dominated strategy

    For each player we identify his best strategy, given the otherplayers strategy: red for player 1, green for player 2

    Two equilibrium points: (Top, Left) or (Bottom, Right)

  • 8/3/2019 Game Theory and Strategic Behaviour

    19/27

    There are two firms: y1 and y2 are their respectiveoutputsY=y1+y2 is industry output: price (p) depends on YThe payoffs are firm profits:1=p(Y)y1-C(y1)=p(y1+y2)y1 - C(y1) = 1(y1, y2)2=p(Y)y2-C(y2)=p(y1+y2)y2 - C(y2) = 2(y1, y2)

    The payoffs depend upon own output and output of rival

    Firm 1 chooses y1 to maximise 1, given y2: y1*= R1(y2)

    Firm 2 chooses y2 to maximise 2, given y1 :y2*= R2(y1)

    Cournot Duopoly

  • 8/3/2019 Game Theory and Strategic Behaviour

    20/27

    Cournot-Nash Equilibrium

    Two Firms The dashed arrowsrepresent the pathtowards the Cournot-Nash equilibrium

    At the Cournot-Nashequilibrium, no firmhas an incentive tochange

    Firm

    2sOutpu

    t:y2

    Firm 1s Output: y1

    Firm 1s reaction function:

    y1*

    =R1(y2)

    Firm 2s reaction function:

    y2*

    =R2(y1)

    Nash Equilibrium

  • 8/3/2019 Game Theory and Strategic Behaviour

    21/27

    Betrand Duopoly

    In Cournot-Nash firms compete on output

    In a Betrand model, firms compete onprice

    There are two firms, producing ahomogenous product with marginal cost(MC)

    They price this at p1 and p2, respectivelyD(p) is the aggregate demand curve

  • 8/3/2019 Game Theory and Strategic Behaviour

    22/27

    P2=40

    If p2=40, firm 1 will sellnothing if p1>p2 and willcapture the market if p1X, it willundercut

    Then, firm 2 will undercut 1,by reducing price below$39

    Process continues untilp1=p2=MC

    Unprofitable to lower pricebelow MC

    A

    B C

    E

    30 60

    P1=39X

    Y MC

    P=MC

    Marginal Cost Pricing Under Betrand Equilibrium

  • 8/3/2019 Game Theory and Strategic Behaviour

    23/27

    Reaction Functions Under Betrand Duopoly with Homogenous Product

    Price charged

    by firm 1

    Price charged by firm 245 line

    p2*

    Reaction function of firm 1

    Reaction function of firm 2

    p1*0 40

    40

    MarginalCost

  • 8/3/2019 Game Theory and Strategic Behaviour

    24/27

    Betrand Duopoly and CompetitiveEquilibrium

    1. Firms price at marginal cost

    2. Firms make zero profits

    3. The number of firms is irrelevant to the price levelas long as more than one firm is present: two firmsare enough to replicate the perfectly competitiveoutcome!

    essentially, the assumption of no capacity constraintscombined with a constant average and marginal costtakes the place of free entry

  • 8/3/2019 Game Theory and Strategic Behaviour

    25/27

    Betrand Duopoly with Differentiated Products

    There are two products: Hondas

    Jazz and Renualts Clio. For eachproduct, demand depends on ownprice and rivals price:

    DClio=f(pC, pJ) and DJazz=g(pC, pJ)

    For each price of Jazz, Clio willhave a profit maximising price, pC

    We can compute, for every Jazzprice pJ, the corresponding profitmaximising Clio price, pC: this is theClio reaction function

    We can compute, for every Clioprice p

    C

    , the corresponding profitmaximising Jazz price, pJ: this is theJazz reaction function

    Betrand equilibrium is atintersection of reaction functions

    Demand Curve for Clioswhen pJ=10,000

    MR MC

    pC

    Quantity Clios

    Clio reaction function

    Jazz reactionfunction

    Betrandequilibrium

    Price Clio

    PriceJazz

  • 8/3/2019 Game Theory and Strategic Behaviour

    26/27

    -1,31,0

    0,-10,0

    Player 2

    Left Right

    Top

    BottomPlayer1

    Neither player has a dominating nor a dominated strategy

    For each player we identify his best strategy, given the otherplayers strategy: red for player 1, green for player 2

    There is no Nash equilibrium

  • 8/3/2019 Game Theory and Strategic Behaviour

    27/27

    Mixed Strategies

    Suppose A plays top with probability and bottom with probability

    Suppose B plays left and right with

    probabilities Then p(top, left)=p(top,right)=3/8

    And p(bottom, left)=p(bottom,right)=1/8

    This yields a Nash equilibrium