+

Strategic Behavior and Industrial Organization Week 1 (Lectures on Dec 23 and 24, 2013) Lecturer: Mandar Oak (University of Adelaide)

SBIO

SBIO
!!

Strategic Behavior and Industrial organization

SBIO = SB + IO
!! !!

SB taught over first 5 weeks by me (Prof Oak) IO taught over next 5 weeks by Prof Gupta

!!

My part
!! !! !!

4 weeks of lectures Last week for case discussions (25%) Followed by a mid term exam (37.5%)

!!

Students to do either the case discussion following my part or the project following Prof Guptas part. Assignment to these two tasks is randomly determined.

Strategic Behavior
!!

Behavior under situations of Strategic Interdependence

Person is payoff explicitly dependent on the actions of person j.

!!

Forms of Strategic Interdependence:

Cooperation, Coordination, Competition, Conflict

!!

Game Theory: Science of strategic interactions

Strategic situations are modeled as games Participants are called players Games of strategy vs. games of skill Cooperative vs. Non-cooperative Game Theory (we study the latter)

Learning Outcomes for SB

After the SB part of this course you should

Be able to represent strategic situations as games Be able to use appropriate solution concepts to find the outcome of a game Understand how the outcome changes with the rules of the game

Uses of Game Theory for Business Students

Develops an awareness of the strategic situations that occur in day to day running of any business Provides tools that will help you get better at playing a game Makes you better at predicting the outcome of a game played by others Provides insights that will help you design games that will lead to a desired outcome (mechanism design)

!! !!

!!

In the context of this course, Game Theory will develop tools for studying Industrial Organization, an important subject for business students

Strategic Thinking: Informal Introduction

Tic-Tac-Toe, Rock-Paper-Scissors A slightly more complicated game (to be played on the board)

Some Principles of Strategic Thinking

!!

!!

!!

look forward and reason backward Put yourself in your rivals shoes; infer what he knows from his actions Anticipate that your rival will also be doing 1, 2 (and 3)!

The Hat Puzzle

3 students are sitting in a row. The teacher brings in 5 hats - 3 green and 2 red - and places one hat on each students head. Student 3 can only observe the color of student 1 and 2s hats, student 2 can only observe the color of 1s hat. The teacher (publicly) asks these students, starting with student 3, can you guess the color of the hat on your head? Student 3 answers: No; Then Student 2 also answers: No; But Student 1 (even though he cannot see anyones hat) answers: Yes. Question: What color is his hat?

Representing a Game
!! !!

Set of Players (who are the players) Set of actions available to each player at different stages (what can they do and when) Players information at each stage (what do they know) Outcome following each action profile (what happens) Preferences of players over the outcomes (what do they want to achieve)

!! !! !!

R-P-S
!! !! !! !!

Two players: 1 and 2 Each can choose an action from the set {R,P,S} At the time of choosing a player does not know others action Possible outcomes are: (RR) (RS) (RP) and so on R beats S, S beats P, P beats R Each player want to choose the action that beats the rivals action (e.g. payoffs could be winner gets 1, loser gets -1; both get 0 if tie) This game is best summarized in a game matrix (show on the board)

!!

!!

Dating Game
!!

Two boys B1 and B2 want to ask a girl G out on a date. Each boy chooses an action from {Propose, Dont Propose}. G chooses an action from {Accept, Dont Accept}; she can accept at most one proposal. Each boy gets utility 1 if the girl accepts the proposal, 0 if she does not; moreover, when both propose, the boy whose proposal is rejected gets utility -1 if she accepts the other boys proposal. G gets utility bi from dating i. She also gets ego utility e for each proposal she receives.

!!

!!

Dating Game contd.

Variant 1: B1 decides whether to propose, B2 observes B1s actions and then decides whether to propose. Then, G chooses her action Variant 2: B1 and B2 simultaneously (i.e. without know each others actions) choose whether to propose, Then G chooses her action This game is best summarized using a game tree (show on the board)

!!

!!

Two Representations
!!

The game matrix is called the Normal Form Representation (aka Strategic Form) The game tree is called the Extensive Form Representation Typically Simultaneous Move games are better represented in the Normal Form, and Sequential Move games are better represented in the Extensive Form However any game can be represented in a normal or an extensive form Represent R-P-S in Extensive Form (next lecture we will represent the Dating Game in Normal Form)

!! !!

!!

!!

Blind Dating Game

Consider the Dating Game as before but now G only knows if she got proposals, one proposal or none.

(work out on the board) Blind date variant 1: B1 and B2 move sequentially

Blind date variant 2: B1 and B2 move simultaneously

Types of Games
!! !!

Simultaneous vs. Sequential Games Perfect vs. Imperfect Information Games

(Imperfect Information: players do not know the actions chosen by other players, e.g. variant 2 of the dating game)
!!

Complete vs. Incomplete Information Games

(Incomplete Information: Structure of the gam, e.g. what payoff is earned after which action is not known to some player, e.g. value of b1 is not known to B2)
!!

For future reference: http://www.econ.ucsb.edu/~garratt/ Econ171/Lect14_Slides.pdf

Assumptions
!! 1.!

Unless otherwise stated we will assume that Players are rational, i.e. they will choose the strategy that will maximize their expected payoff There is common knowledge of rationality Payoffs contain all relevant aspects of utility (not just money) Players use Bayesian updating to incorporate new information (becomes relevant later in the course)

2.! 3.!

4.!

Interesting Stuff
In case there is spare time
!! !!

http://gametheory.net/media/Beautiful.mov World RPS Championship: http://www.youtube.com/watch?v=yUk-WnjNDYQ Can self-goal be rational? Yes, if the tournament rules are ill thought out! http://www.youtube.com/watch?v=4QVXQD6MySw

!!

Normal Form Games

Relationship between Actions and Strategies

Strategy is a complete plan of action Si=Ai1Ai2AiM

Player i has M information sets and Aim is the set of available actions at information set m Remember: information set can contain multiple nodes, but same action must be chosen at these nodes Example of the board

!!

!!

!!

Strategy Profile
!! !!

Collection of strategies, one for each player s = ( s1, , sN )

Normal Form Games contd.

Payoff Functions
!! !!

!!

ui( s ) : player is payoff from strategy profile s Payoff functions capture players preference over the outcomes following different strategies Sometimes we denote by s-i the strategy sub-profile of everyone other than i ui (si , s-i )

!!

!!

Formal representation of a Normal Form Game

(N; {S1, S2, , SN}; {u1, u2, , uN})

Example 1
!!

A potential entrant (player 1) must decide whether to enter the market at the high end (H), low end (L) or to not enter (N) The incumbent (player 2) must decide, after observing 1s action whether to accommodate the entrant (A) or to fight him off (F) If there is no entry, 1 gets 0, two gets 5 If 1 enters at the low end and 2 fights they get -1, 3 If 1 enters at the high end and 2 fights they get -1,1 If 2 accommodates the entry they get 3,3 (irrespective of 1s type of entry)

!!

!! !! !! !!

Example 2,3
!!

Centipede Game (drawn on the board)

Used to test cooperative behavior Each pass increases the amount in the pot A keep ends the game If the game reaches the end, they equally divide the pot

!!

A Strategy must specify action even when those nodes will not be reached under it! Example 3: Example 1 with imperfect information about entry type

!!

Mixed Strategies and Expected Payoffs

Mixed Strategy: Strategy obtained by randomizing between pure strategies.

Examples on the board

!!

Expected Payoff: Payoff obtained from mixed strategies by weighting each payoff by the probability of obtaining it
!!

Examples on the board

!!

Mixed Strategies by other players need not be objective, they may reflect the beliefs held by a player about others behavior

Some Typical Games

These games reflect some common themes in the social interactions one encounters in day to day life (fig 3.4, page 31 of the book)
!! !! !!

Prisoners Dilemma Battle of the Sexes (Coordination Game) Hawk/Dove Game

!!

Discuss the significance of these games

Solving A Normal Form Game

Having represented a strategic interaction as a game we need to be able to predict the outcome of the game
!! !! !!

Depends on how rational players are But also what beliefs they hold about each other Rationality and common knowledge of rationality

!! !!

Players hold beliefs about each others choices Some Notation

s : pure strategy, : mixed strategy, : beliefs Beliefs can be more general than mixed strategies!

Solution Concepts contd.

Dominated strategy: A strategy that is strictly worse than another strategy for any strategy profile s-i played by opponents (formal defn. on page 52)
!!

Example on board (fig. 6.1 page 52; fig. 6.5, page 61)

!!

A rational player will never play a dominated strategy

Prisoners Dilemma is no dilemma!

!!

Clarification: dominated means strictly worse than some other strategy. We use term weakly dominated for weakly worse strategy. Technical Question: Why is the above definition in terms of s-i and not -i ?

!!

Undominated Strategies and Best Response

A rational player will always choose from undominated (i.e. not strictly dominated strategies) Rational player will play the best response given her beliefs (formal defn. on page 56, example on 57)
!!

!!

Like dominated strategy, best response is also defined over the domain of pure strategies

!! !!

UDi : Set of all (pure) strategies that are undominated BRi : Set of all (pure) strategies that are best response to some beliefs For two player games BRi = UDi whereas for multi player games we have BRi is subset of UDi

!!

Iterated Deletion of Dominated Strategies

Rational player will not play a dominated strategy If the rival is known to be rational, you will expect him to not play a dominated strategy either When rationality is common knowledge this thought process may lead to a unique outcome of the game via iterated elimination of dominated strategies (IEDS)
!!

!!

Example on board (pg. 68)

Nash Equilibrium
!!

A Strategy Profile is Nash Equilibrium if each strategy is Best Response to the others strategies

Homework and Readings for Week 1

Closely read Chapters 2, 3, 4 and 6 especially make sure you understand the solutions to the Guided Exercise for each chapter Try exercises 5, 7 and 8 in Ch. 2 Try exercises 1, 3 and 7 in Ch. 3 Try exercises 1, 6 and 7 in Ch. 6 If you want to read ahead, skim through Chapters 7 and 9!

!! !! !! !!