l(W I

~
ignou
THE PEOPLE'S
UNIVERSITY
MEC -101
MICROECONOMIC
Indira Gandhi National Open University
School of Social Sciences ANALYSIS

Volume 2
--~~
 MEC-IOl
1te:\11 ~~ ..~
~ 'UNIVERS1TY
Microeconomic
Indira Gandhi Analysis
National Open University
School of Social Sciences

Block

7
NON-COOPERATIVE GAME THEORY-I
UNIT 20
Static Games of Complete Information 5
UNIT 21
Static Games with Complete Information: Applications 45
UNIT 22
Dynamic Games with Complete Information 77
Expert Committee
Prof. Bhaswar Moitra
Prof. Gopinath Pradhan
Department of Economics
School of Social Sciences
Central Road, Jadavpur University
Indira Gandhi National Open University
Kolkata
New Delhi
Dr. Naresh Kumar Sharma
Prof. Narayan Prasad
University of Hyderabad
School of Social Sciences
School of Economics, Hyderabad
Indira Gandhi National Open University
Dr. Anirban Kar New Delhi
Deptt. of Economics
Prof. Kaustuva Barik
Delhi School of Economics
School of Social Sciences
University of Delhi
Indira Gandhi National Open University
Dr. Indrani Roy Chowdhury New Delhi
Economics Faculty
Prof. B.S. Prakash
Jamia Millia Islamia, New Delhi
School of Social Sciences
Prof. Sudhir Shah Indira Gandhi National Open University
Deptt. of Economics, Delhi School of Economics New Delhi
University of Delhi
Mr. Saugato Sen
Dr. Manish Gupta School of Social Sciences
NIPFP, New Delhi Indira Gandhi National Open University
New Delhi

Course Editor : Prof. Gopinath Pradhan

Programme Coordinators Course Coordinator

Prof. Gopinath Pradhan, IGNOU, New Delhi Prof. Gopinath Pradhan
Prof. Kaustuva Barik, IGNOU, New Delhi

Block Preparation Team

Unit Writer
20&21 . Dr. Indrani Roy Chowdhury
22 Prof. Gopinath Pradhan

PRINT PRODUCTION
Sh. Tilak Raj
Assistant Registrar
MPDD, IGNOU, New Delhi
September 2021 (Reprint)
© Indira Gandhi National Open University 1993
ISBN: 978-93-86375-97-1
All rights reserved. No part of this work may be reproduced in any form, by mimeograph or any other
means, without permission in writing from the lndira Gandhi National Open University.
Further information about the lndira Gandhi National Open University courses may be
obtained from the University's office at Maidan Garhi, New Delhi-ll0068
Printed and Published on behalf of the Indira Gandhi National 'Open University, by Registrar
MPPDD, IGNOU New Delhi.
Printed at : Mls Nova Publications and Printers (Pvt.) Ltd., Plot No. 09-10, Sector-59, Phase-2 Faridabad-121004,
Phone: 0129-4317645.
 BLOCK 7 NON-COOPERATIVE GAME
THEORY-I

I. Introduction
This block deals with the non-cooperative game theoretic approach often used in
seeking solutions to economic problems in general and microeconomic problems in
particular. Unit 20 introduces the static games with complete information along with
the basic concepts that helps understand the method of solution. The concept of
Nash equilibrium is introduced by considering pure and mixed strategies. The unit
also contains the description of the concept of equilibrium and representation of
games through normal form. Some of the applications of static games with complete
information are discussed in Unit 21. Insights thrown by game theory to problems
of tragedy of commons and duopoly markets are important themes covered in the
unit. The last unit of the block (Unit 22) contains solution to problems when we
have situation of dynamic games with complete information. Itintroduces the concept
of backward induction as a refinement ofNash equilibrium and presents the extensive
format of a game.
UNIT 20 STATIC GAMES OF
COMPLETE INFORMATION

Structure
20.0 Objectives
20.1 Introduction
20.2 Some Examples of Strategic Games
20.3 Classifications of Games
20.3.1 Cooperative and Non-Cooperative Game Theory
20.3.2 Games of Complete Information and Incomplete Information
20.3.3 Strategic (or Normal Farm) and Extensive Form Games
20.4 Rules of the Game
20.5 Normal Form of Game under Complete Information
20.5.1 Some Well-known Strategic Games
20.5.1.1 Matching of Pennies
20.5.1.2 Battle of Sexes
20.5.2.3 Prisoners' Dilemma Game
20.5.1.4 Hawk-Dove (or Chicken) Game
20.5.1.5 Coordination Game

20.6 Solution Concept under Dominant Strategy

20.6.1 Dominance
20.6.2 Dominant Strategy Equilibrium
20.6.3 Iterative Elimination of Dominant Strategy Equilibrium
20.7 Solution Concept under Nash Equilibrium in Pure Strategy
20.7.1 Minimax Method for Zero Sum Game
20.7.2 Nash Parables
20.7.3 Some Common Examples of Nash Equilibrium
20.7.4 Relation between the Solution Concepts

20.8 Mixed Strategy Nash Equilibrium

20.S.1 Definition and Example
20.9 Let Us Sum Up
20.10 KeyWords
20.11 Some Useful Books
20.12 Check Your Progress AnswerlHints
20.13 Exercises

20.0 OBJECTIVES
After going through this unit, you will be able to understand:
• various classifications of games like cooperative and non-cooperative
games along with that of simultaneous move games;
5
Non-Cooperative • rules of strategic game or normal form of game;
Game Theory-I
• basics of the normal form or the strategic form of game;
• solution concept under Dominant Strategy;
• solution concept under Nash equilibrium in pure strategy; and
• solution concept under Mixed Strategy Nash equilibrium.

20.1 INTRODUCTION
Game theory studies the general principles that explain how people and
organisations act in strategic situations in a multiple agent framework.
A strategic situation involves when two or more interacting players, who
make decisions by trying to anticipate the actions and reactions of others.
Thus, Game Theory can be described as a strategy theory, or theory of
interactive decision making. It is rather a formal study of interdependence,
conflict and cooperation. Game theory can be used to explain phenomena in
the areas of economics, political theory, evolutionary science and military
strategy, to name a few.
Game theory was established as a field in its own right after the 1944
publication of the monumental volume 'Theory of Games and Economic
Behavior' by von Neumann and the economist Oskar Morgenstem. It
provided much of the basic terminology and problem setup that is still in use
today. In 1950, John Nash demonstrated that [mite games have always have
an equilibrium point, at which all players choose actions which are best for
them, given their opponents' choices. This central concept of non-cooperative
game theory has been a focal point of analysis since then. Since the 1970s, it
has driven a revolution in .economic theory. Additionally, it has found
applications in sociology and psychology and established links with evolution
and biology.Game theory received special attention in 1994 with the awarding
of the Nobel Prize in economics to Nash, John Harsanyi and Reinhard Selten.

Game theory studies strategy mainly through the analysis of different

"games". A "game" in game theory is a fully explicit structure which
characterises players, their set of actions, payoffs and possible outcomes under
given rules of playing. Thus a game is described by the following ingredients:

Game Theory is a formal way to consider each of the following items:

a) Group: In any game there is more than one decision -rnaker, each
decision maker is referred to as 'player'.
b) Interaction: What anyone individual player does directly affects at
least other player in the group.
c) Strategy: An individual player accounts for her interdependence, each
player chooses her best action or the strategy.
d) Rational: A central assumption in many variants of game theory is that
the players are rational. A rational player is one who always chooses an
action which gives the outcome she most prefers, given what she expects
her opponents to do. The goal of game-theoretic analysis then is to
predict how the game will be played by rational players, or, to give
advice on how best to play the game against opponents who are rational.
Let me now illustrate these four concepts - group, interaction, strategic and
rational - by discussing in detail some of the examples given below.
6
 Static Games of
20.2 SOME EXAMPLES OF STRATEGIC GAME Complete
Information .
a) Working on a group project:A case study for the game theory class.
The group comprises the students jointly working on the case. The
interaction arises from the fact that the certain amount of work needs to
be done in order to write a paper. Hence if one students slacks off,
somebody else has to put in extra hours the night before the paper is due.
Strategic play involves estimating the likelihood of freeloaders in the
group and rational play requires a careful comparison of the benefits to a
better grade against the costs of extra work.

b) Random drug testing (at Olympics)

The group is made up of competitive athletes and the International
Olympic Committee (IOC). The interaction is both between the athletes
~ - who make decision on training regimens as well as whether or not to
use drugs - and with the IOC, which needs to preserve the reputation of
the sport. Rational strategic play requires the athletes to make decisions
based on their chances of winning and if they dope, their chances of
getting caught. Similarly, it requires the IOC to determine drug testing
procedures and punishments on the basis of testing procedures and
punishments on the basis of testing costs and the value of a clean whistle
reputation.

c) Rand Jj efforts by pharmaceutical companies

Some estimates suggest that research and development (R and D)
expenditures constitutes as 20% of annual sales of US pharmaceutical
companies and that on an average the development costs of a new drug
is about $350 million dollars. Companies are therefore very concerned
about ploughing back their investment in Rand D and mitigation of risk
in their investment. If we take the set of big drug companies, the
strategic interaction arises because the first developer of a drug makes
the most profits (thanks to the associated patents).

20.3 CLASSIFICATION OF GAMES

In this section we will discuss the various classifications of games.

20.3.1 Cooperative and Non-Cooperative Game Theory

Game theory comprises two branches: Cooperative game theory and non-
cooperative game theory. Note that 'cooperative' and 'non-cooperative'<are
technical terms and are not an assessment of the degree of cooperation among
agents in the model. A cooperative game can as much model extreme
competition as a non-cooperative game can model cooperation.

1) Cooperative Game: A game in which players/ agents negotiate and

enforce their contracts (contract is binding through third parties) so that
they can make joint strategies. Firms buying raw material or buying a
service can negotiate for price settlement. Cooperative game theory
models analyses situations where agents can cooperate to create value by
joining coalitions, but also where agents compete to capture value, in
unstructured interactions. It can be thus called coalitional as it abstracts
from the detailed procedures and focuses on how the value creation
7
Non-Cooperative abilities of each coalition of agents can bear on the agents' ability to
Game Theory-I capture value.

Its formalism comprises a set of agents and a function that returns the
value each subset of agent (i.e., each coalition) can create on its own (the
characteristic function). The characteristic function is the input to a
solution concept, which returns the value captured by each agent (their
imputation). Because no structure is put on the bargaining procedure, the
solution concept models the outcome of the negotiations occurring
among the agents, accounting for the value each coalition can create.
While the solution concepts are the Core and the Shapley value, but a
number of others have been developed (for further details see Osborne
and Rubinstein (1994) and Owen (1995)).

The core is the set of imputations such that each coalition receives at
least as much as it can create on its own. This concept formalizes a
notion of individual and coalitional self-interest. The core is thus
appropriate to model the competitive nature of free-form competitive
interactions. However, the core may not exist and is not necessarily
unique. Non-existence of the core signals the intrinsic instability of a
particular interaction. Non-uniqueness of the core implies that
competitive forces alone are not enough to determine each agent's value
capture and that negotiating abilities also come into play. An agent's
added value (or marginal contribution) is the value lost to the grand
coalition (the coalition of all agents) if the agent is not included. The
added-value principle, implied by the core, states that under competitive
free- form interaction an agent can capture no more than her added value.
Otherwise, the other agents would be better off to exclude her and trade
only among themselves.

The Shapley value is the average added value of an agent to all possible
coalitions. It always exists and is unique. It emphasizes the fairness of
the division of value among agents. It has been used to model the
allocation of costs among users of a common facility and to measure
political power.

2) Non-Cooperative Game: In game theory, a non-cooperative game

theory is concerned with the analysis of strategic choices where players
(or agents) make decisions independently, out of their own interest.
Non-Cooperative game theory models the actions of agents, maximizing
their utility in a defmed procedure, relying on a detailed description of
the moves and information available to each agent. Thus, in the
paradigm of non-cooperative game theory, the details of the ordering
and timing of players' choices are crucial to determine the outcome of a
game. In the non-cooperative game theory players cannot write any
binding contract. Cooperation can, and often does, arise in non-
cooperative models of games, when players find it in their owl). best
interests i.e., that strategy if played in equilibrium must be self-
enforcing.

Non-cooperative game is the natural way to start a lesson on Game

Theory.

8
 Static Games of
20.3.2 Games of Complete Information and Incomplete
Complete
Information Information
Non-Cooperative games can be classified into (i) Games of Complete
Information and (ii) Games of Incomplete information.
Under non-cooperative games, a category of game in which knowledge about
other players, utility functions, payoffs, strategies and "types" of players are
common knowledge i.e., those information are available to all participants.
Then such games are called Games of Complete Information.

On the other hand, in a game of incomplete information, players may not

possess full information about their opponents. Some players may possess
private information that the others should take into account when forming
expectations about how a player would behave. A typical example is an
auction: each player knows her own utility function (i.e., the valuation for the
item), but does not know the utility function of the other players. It is often
assumed that the players have some statistical information about the other
players. For example, in an auction, each player knows that the valuations of
the other players are drawn from some probability distribution. In this case,
the game is called a Bayesian game.

20.3.3 Strategic (or Normal Form) and Extensive Form Games

.Games of Complete Information again can be classified as Strategic and
Extensive form games.
The strategic form (also called normal form) is the basic type of game studied
in non-cooperative game theory, where players take their decision
simultaneously. Thus in a strategic game any player does not know the action
or the strategy of the rival players before taking her own move.
The rules.of the game are defined by the players, actions, possible strategies
and the payoffs of a game.
Players are the agents who are involved in the game of decision making.
A strategy is a complete contingent plan, or decision rule, that specifies how
the player will act in every possible distinguishable circumstance in which she
might be called upon to move. All the players have preferences and goals.
That is to say, like all normal people, these agents prefer some options over
another. The players will choose actions or strategies that increase the chances
of getting their preferred outcomes.

Outcomeslpayoffs: The key to game theory is that my payoff is contingent on

the action/strategy of the other players playing the game. Thus, a game has
different outcomes that result from each possible combination of choices of
the players. An outcome is represented by a separate payoff for each player,
which is a number (also called utility) that measures how much the player
likes the outcome.

The extensive form (also called the sequential form), also called a game tree,
is more detailed than the strategic form of a game. It is a complete description
of how the game is played over time. This includes the order in which players
take actions, the information that players have at the time they must take those
actions and the times at which any uncertainty in the situation is resolved. A
game in extensive form may be analysed directly, or can be converted into an
equivalent strategic form.
9
Non-Cooperative Check Your Progress 1
Game Theory-I
1) What is a cooperative game?
.........................................................................................

.........................................................................................

.........................................................................................

.........................................................................................

.........................................................................................

2) State the meaning of a sequential game.

.........................................................................................

.........................................................................................

.........................................................................................

.........................................................................................

.........................................................................................

.3) What do you mean by common knowledge in a game with complete

information?
.........................................................................................

.........................................................................................

.........................................................................................

.........................................................................................

.........................................................................................

4) Defme a strategy used in game theory.

.........................................................................................

.........................................................................................

.......•••••••....................• ! •••••••.......•••••••••••.....•••••.....••••.•.....••

.........................................................................................

.........................................................................................

.........................................................................................

5) State the meaning of Nash equilibrium.

.........................................................................................

.........................................................................................

.........................................................................................

.........................................................................................
10
6) What is a non-cooperative game? Static Games of
Complete
Information

7) State the meaning of payoff in a game.

.................. ' .

8) Defme a pure strategy.

.................................... . .. ;
, .

..........................................................................................

9) Give the meaning of weakly dominant strategy.

10)There are two players, 1 (kicker) and 2 (goalie). Each has two actions,
ai E {L, R} to denote left or right. The kicker wins when they choose
opposite directions while the goalie wins if they choose the same
direction. Using 1to denote a win and -1 to denote a loss, write the game
in a matrix form.

..........................................................................................

11
Non-Cooperative 11) Provide an example of a game in which there is more than one weakly
Game Theory-I dominant strategy equilibrium.

12) Give an example of a game in which there is no weakly dominant

strategy equilibrium .

............... ; .

13) Provide an example of a game in which there is more than one weakly
dominant strategy equilibrium.

20.4 RULES OF THE GAME

Every game is played by a set of rules which have to specify four issues.

1) Who is playing the group of players that strategically interacts.

2) What they are playing with - the alternative actions of choices, the
strategies, that each player has available.
3) When each players get to play (in what order)
4) How much they stand to gain (or lose) from the choices made in the game.

5) How much are the players in a game supposed to know about the rules?
In game theory it is standard to assume common knowledge about the
rules. Everybody has a knowledge about the rules means that if you asked
any two players in the game, a question about who, what, when or how
much, they would give you the same answer.

The two principal representations of (the rules of) a game are called,
respectively, the Normal (or strategic) form of a game and the Extensive form
ofa game.

12
 Static Games of
20.5 NORMAL FORM OF GAME UNDER Complete
COMPLETE INFORMATION Information

A normal form is a model of interacting decision makers (we call them

players or agents). Each player has a set of possible actions. The model
captures the interaction between the players by allowing each player to be
affected by the actions of all players, not only her own action. Each player
simultaneously chooses a strategy and the combination of strategies chosen by
the players, determine a payoff of each player. Each player has ordinal
preferences about the action profile - payoffs received by the player for each
combination of strategies that could be chosen by the players.

The Normal form of game thus specifies:

1) The players in the game
2) The strategies available to each player
I

3) The payoff received by each player, for each combination of strategies

that could be chosen by the players.

Suppose the game 'consisting of n players and let the representative player is
denoted by player I ( i = 1, 2, n). Let Si denotes the set of strategies
available to each player i and let si denotes the an arbitrary specific strategy
(Si E Sa, to indicate the strategy Si is the member of the set of strategy Si' Let
(Sv S2, S3' ... .;... sn)denotes a combination of strategies one for each player, let
Ui denote the player i's payoff Ui(SVS2, sn), where players choses the
strategies (sv S2' sn). Thus the normal form of n-players game is
specified as G = {Sv S2' Si, Uv U2, Ui} (20.1).

In the normal form of game player's choose their strategies simultaneously. It

implies that each player chooses her or her action without the knowledge of
the other's choices.

20.5.1 Some Well-known Strategic Games

20.5.1.1 Matching of Pennies
a) The game involves two players, child 1 and 2, simultaneously placing a
penny on the table with Head or tail at the top. The payoff depends on
whether the pennies match or not. If both pennies are heads or tails i.e., (h,
h) or (t, t), Child 1 wins and keeps Child 2's penny; if they do not match
i.e., (h, t) or (t, h), Child 2 wins and keeps Child 1's penny. This is a zero-
sum game because oneplayer's gain is the other's loss. The payoffs for
Child 1 and 2 are shown in the table below (see Table 20.1), with the first
numeral in the four payoff cells representing Childl's payoff and the
second numeral Child 2's playoff. As can be seen, the combined playoff
for Childl and 2 in all the four cells is zero.

Table 20.1: Example of Zero-Sum Game

Child 2
Child 1 Head (h) Tail(t)
Head (h) (h, h) = (1, -1) (h, t) = (-1, 1)
Tail(t) (t, h) = (-1,1) (t, t) = (1, -1)

13
r

Non-Cooperative 20.5.1.2 Battle of Sexes

Game Theory-I
A husband and wife are in a dilemma of going to movie or to a football match.
They each respectively prefer the football game and movie. Again at the same
time they prefer to go with the spouse than going alone. See Table 20.1.

Table 20.2: Battle of Sexes

Wife
Husband Football (F) Movie(M)
Football (F) (f, f) = (3, 1) (h, t) = (0, 0)
Movie (M) (t, h) = (0, 0) (m, m) = (1, 3)

20.5.1.3 Prisoners' Dilemma (PD) Game

The prisoners' dilemma is a popular example widely used in game theory that
shows why two completely rational individuals might not cooperate, even if it
appears that it is in their best interests to do so. The game is presented as
follows:

Two members «1 and 2) of a criminal gang are arrested and imprisoned. Each
prisoner is in solitary confinement with no means of communicating with the
other. Prosecutors lack sufficient evidence to convict the pair on the principal
charge. They are being interrogated separately about their crime. Each
prisoner is given the opportunity either to: betray the other by testifying that
the crime is committed, or to cooperate with the other by remaining silent.
The convicts have the options either to confess or not to confess their crime.
Based on their joint decision penalty will be decided. If both of them confess
the crime will be established and each will get 6 years of imprisonment. If
none of them confesses, the crime is not established. So, they will be
nominally punished for 1 year. However in case one confesses and the other
does not, the crime is established, so the convict who has confessed will be set
free for her honesty and the other one will be put under the bar for 9 years.
Table 20.3.

Table 20.3: Prisoners' Dilemma Game

Convict 2
Convict 1 Confess (c) Defect (d)
Confess (c) (c, c )=(-6,-6) (c, d) =(0, -9)
Defect (d) (d, c) =(-9,0) (d, d)=(-l, -1)

PD Game in a Generalized Form

The structure of the traditional Prisoner's Dilemma can be generalized from
its original prisoner setting. If both players cooperate, they both receive the
reward R for cooperating. If both players defect, they both receive the
punishment payoff P. If Player 1 defects while Player 2 cooperates, then
Player 1 receives the temptation payoff T, while Player 2 receives the
"sucker's" payoff, S. Similarly, if Player 1 cooperates while Player 2 defects,
then Player 1 receives the sucker's payoff S, while Player 2 receives the
temptation payoff T. This can be expressed in the following normal form (see
Table 20.4).

14
 Table 20.4: Prisoner's Dilemma in Generalized Form Static Games of
Complete
Player 2 Information

Player 1 Cooperate (c) Defect (d)

Cooperate (c) (R, R) (S, T)
Defect (d) (T,S) (P, P)

The condition that must hold for a prisoner's dilemma game, in the strong
sense, is that.if the payoffs follows the inequality: ·T > R > P > S. The payoff
relationship R > P implies that mutual cooperation is superior to mutual
defection, while the payoff relationships T > Rand P > S imply that defection
is the dominant strategy for both agents. That is, mutual defection is the
unique Nash equilibrium in the game (i.e., the only outcome from which each
player could only do worse by unilaterally changing strategy). The dilemma
then is that mutual cooperation yields a better outcome than mutual defection
but it is not the rational outcome because from a self-interested perspective,
,"
the choice to cooperate, at the individual level, is irrational. This sort of
outcome is described by Mancur Olson (1965) as - individual rationality
produces collective irrationality.

The prisoner's dilemma game is widely used in the study of game theory.

1) Payoff structure associated with PD game arises in many strategic

situations in economic, social, political, biological competitions. This
.game has a wide range of applicability from the strategic stand point.

2) Curious nature of equilibrium outcome achieved in such game. Both

players follow conventional wisdom in choosing their dominant strategy,
but the resulting equilibrium outcome yields them a payoff that are lower
than that could have achieved if they have chosen their dominated
strategy. In casual usage, it may be applied to situations in which two
entities could gain important benefits from cooperating or suffer from the
failure to do so, but find it merely difficult or expensive, to coordinate
their activities t? achieve cooperation.

20.5.1.4 Hawk-Dove (or, Chicken) Game

The chicken game, also known as the hawk-dove game, is a model of conflict
for two players in game theory. The principle of the game is that while it is for
both players beneficial if the other player yields, their own optimal choice
depends on what their opponent is doing: if their opponent yields, the player
should not, but if the opponent fails to yield, the player should.

The name "chicken" has its origins in a game in which two drivers drive
towards each other on a collision course: one must swerve, or both may die in
the crash, but if one driver swerves and the other does not, the one who
swerved will be called a "chicken", meaning a coward; this terminology is
most prevalent in political science and economics.

The name "Hawk-Dove" refers to a situation in which there is a competition

for a shared resource and the contestants can choose either conciliation or
conflict; this terminology is most commonly used in biology and evolutionary
game theory. From a game-theoretic point of view, "chicken" and "hawk-
dove" are identical; the different names stem from parallel development of the
basic principles in different research areas. See Table 20.5 and Table 20.6.
15
Non-Cooperative Table 20.5: Chicken Game
Game Theory-I
Player 2

Player 1 Swerve (Sw) Straight (St)

Swerve (Sw) Tie, Tie Lose, Win

Straight St) Win, Lose Crash, Crash

Table 20.6: Chicken Game with Numerical Payoffs

Player 2
Swerve (Sw) Straight (St)
Player 1 Swerve (Sw) 0,0 -I, +I(NE)

Straight (St) +1, -I (NE) -10, -10

Two versions of the payoff matrix for the Chicken game are presented here
(Tab1e-20.5 and Table 20.6). Two (young) players are engaged in a conflict
situation. They may be racing their car towards each other on Main Street,
while being egged on by their many friends. If player 1 hangs tough by racing
straight and stays in the Centre of the road while the other player swerves by
moving out of the way, then all glory is her and the other player eats humble
pie (St, Sw). If they both hang tough by racing straight (St, St) they end up
with broken bones, while if they both swerve they have their bodies- but not
their pride - intact.

In such game the outcomes are such that, where each player would prefer to
win over tying, prefer to tie over losing and prefer to lose over crashing. Here,
the benefit of winning is 1, the cost of losing is -I and the cost of crashing is -
10. It is mutually beneficial for the players to play different strategies.

Hawk-Dove Game
In the biological literature, this game is referred to as Hawk-Dove. The earliest
presentation of a form of the Hawk-Dove game was by John Maynard Smith
and George Price in their paper, ''The logic of animal conflict". Both in
Chicken and Hawk-Dove, it is mutually beneficial for the players to play
different strategies. It can be thought of as the opposite of a coordination game
(see below for Coordination game), where playing the same strategy Pareto
dominates playing different strategies. The underlying concept is that players
use a shared resource. In coordination games" sharing the resource creates a
.. benefit for all: the resource is non-rivalrous and the shared usage creates
positive externalities. In anti-coordination games the resource is rivalrous but
non-excludable and sharing comes at a cost (or negative externality).

The traditional payoff matrix for the Hawk-Dove game is given in Table 20.7,
where V is the value of the contested resource and C is the cost of an escalated
fight. It is (almost always) assumed that the value of the resource is less than
the cost of a fight, i.e., C > V > 0. If C :s V, the resulting game is not a game'
of Chicken but is instead a Prisoner's Dilemma. The exact value of the Dove
vs. Dove payoff varies between model formulations. Sometimes the players
are assumed to split the payoff equally (V 12 each), other times the payoff is
assumed to be zero. .
16
While the Hawk-Dove game is typically taught and discussed with the payoffs Static Games of
in terms of V and C, the solutions hold true for any matrix with the payoffs in Complete
Information
Table 20.8, where W > T > L > X.
Table 20.7: Hawk-Dove game
Player 2
Hawk (H) Dove (D)
Hawk (H) (V-C)/2, (V-C)/2 V,O
Player 1
Dove (D) 0, V V/2, V/2

Table 20.8: General Form of a Hawk-Dove game

Player 2
Hawk (H) Dove (D)
Player 1 Hawk (H) X,X W, L(NE)

Dove (D) L, W(NE) T,T

20.5.1.5 Coordination Game

Coordination games are a formalization of the idea of a coordination
problem, which is widespread in the social sciences, including economics,
meaning situations in which all parties can realize mutual gains, but only by
making mutually consistent decisions. A common application is
the choice of technological standards.

One way to model a minimum-effort coordination game is to let players

choose effort levels, 1 or 2, where each unit of effort results in a cost of c < 1.
The output per person is determined by the minimum of the effort levels
chosen. For example, if both choose an effort level of 1, then for a given per-
unit effort cost (c), each player receives a payoff of l-c, as shown in the
upper-left box of Table 20.9. Similarly, when both choose a high effort level
of 2, they each obtain 2-2c. But if they fail to coordinate with effort choices of
1 and 2, then the minimum is 1 and payoffs are l-e for the low-effort
individual and 1- 2c for the high-effort individual.

Table 20.9:-Minimum-effort Coordination Game

Player 2's Effort
1 2
Playerl's 1 (1 -c), (1 - c) (NE) (1 - c), (1 -2c)
Effort
2 (1 -2c), (1 - c) (2 -2c), (2 - 2c )(NE)

20.6 SOLUTION CONCEPT UNDER DOMINANT

STRATEGY
In this section we will discuss the fIrst solution concept of the Static Game
under Complete Information.

17
Non-Cooperative 20.6.1 Dominance
Game Theory-I

Consider a player have two strategies A and B so that, given any combination
of strategies of the other players, the outcome resulting from A is better than
the outcome resulting from B. Then strategy A is said to dominate strategy B.
Since all players are assumed to be rational, they make choices which result in
the outcome they prefer most, given what their opponents do. If out of N
strategies A, B, C, ... , N, the outcome from strategy A is better than any other
(N-I) strategies, irrespective of the strategies of the other players then there is
a compelling reason to think that the rational player will choose it. If a
strategy is bad for an agent no matter what the others might be doing there is
equally compelling reason to think that a rational player would avoid it. A
rational player will never choose to play a dominated strategy.

20.6.2 Dominant Strategy Equilibrium

In a normal. form of n players game specified asG = {Sv S2' ..... Si' Uv U2,
.....ud. Let s;ands;'be feasible the strategies for player i (i.e., s;ands;' are the
members of Si' Strategy s; is strictly dominated by strategy s;', if for
each feasible combination of the other player's strategies, player i's payoff
from playing s; is strictly less than i's payoff from playings.', i.e.,
Ui( Sv S2' S3, Si, s;, Si+V sn) < Ui (Sv S2'S3' Si, s;', Si+V sn), .
(20.2) for each combination of strategy (sv S2'S3' Si-V Si+V sn) that can
be constructed from the other player's strategy spaces
(Sv S2' ..... Si-i, Si+V .... Sn)·

If a strategy is not dominated by any other, it is called undominated strategy.

It is useful to think of a dominated strategy as a bad strategy and an
undominated strategy as a good one. Thus a rational player will not play a
strictly dominated strategy. Consider a game of two players with strategies
High- Low and North-South respectively (Table 20.10).

Table 20.10: Example of Undominated Strategy

Player 2
North (N) South (S)
Player 1 High (H) 1Ti (H, N),1T2(H, N) 1T1 (H, S) ,1T2 (H, S)
Low(L) 1Ti (L, N), 1T2 (L, N) 1Ti (L, S),1T2(L, S)

In this above example of strategic game the Player's 1 strategy Low (L) is
dominated by High (H) if:
a) 1Ti (H, N) ~ 1Ti (L, N) (20.3a)
b) 1Ti (H, S) ~ 1Ti (L, S), (20.3b)

The first inequality says that H yields a higher payoff than L to the player 1, if
the player 2 plays her first strategy i.e., North (N). The second inequality says
the same is true even if player 2 plays her second strategy.
With at least one of those inequalities being strict, Low is not dominated by
High (and vice versa).If it is better against, say, South but does worse against
North i.e.,
a) 1Ti (H, N) > 1Ti (L, N), (20.4a)
18 But b) 1Ti (H, S) < 1Ti (L, S), (20.4b).
 Consider the Prisoners Dilemma game of Table 20.3. Static Games of
Complete
1) Each agent has 2 strategies to cooperate (not confess) with one's rival and Information
the other to defect from cooperation (by confessing).
2) Each player has a dominant strategy (to confess) or to defect from the
cooperation strategy.
3) Dominant strategy equilibrium is worse for both the agent than the non
equilibrium situation or the dominated strategy to cooperate with the rival
(not confess, not confess).
The strategy confess is dominating the strategy defect, as it gives the Convict
.1 a higher payoff than Defect irrespective of the strategy played by the
Convict 2 i.e., whether she confesses or defects ~
(07rI(C,C) = (-6) > 7r1(d,c) = (-9)
(ii)7r1(c, d) = (0) > 7r1(d, d) = (-1)

Similarly, for the Convict 2 also, confession is dominating the strategy defect,
as it gives the Convict 2 a higher payoff than Defect, irrespective of the
strategy played by the Convict 1 whether she confesses or defects ~.
(iii)7r2(C, c) = (-6) > 7r2(d, c) = (-9)
(iv) 1r2(C, d) = (0) > 7r2(d, d) = (-1).
So, in this Prisoner's Dilemma game, confession is the Dominant strategy for
both the convict. Thus in this game, it is reasonable to predict that both the
players will end up confessing in the equilibrium. Hence {Confess, Confess}
. is the Dominant Strategy Equilibrium.

20.6.3 Iterated Elimination of Dominant Strategies Equilibrium

The problem with the dominant strategy equilibrium is that in many games a
player need not have a dominant or best strategy. Thus the modest objective is
instead of searching for the best strategy we will end up in successively
eliminating the dominated strategies (See Table- 20.11, Table -20. 11a, Table-
20. l lb and Table-20.lIc for the iterative elimination).
Consider the following game given in Table 20.11.

Player 2
Left Right
Player 1 Up (1,1) (0,1)
Middle (0,2) (1,0)
Down (0,-1) (0,0)

For the Player 1, the strategies Up and Middle dominate the strategy Down.
Hence Down is the dominated strategy for the player 1. So, for a rational
player, it will be irrational to play a dominated strategy Down. The reason is
that by playing any strategy that dominates (this dominated strategy), the
player can guarantee a payoff which is at least as high, irrespective of the
moves of the rival players. Hence the player 1 will ever play Down but will
rather play either Up or Middle.

The logic could then 'Set in motion a chain reaction. In a game once it is
known that the player 1will not play the dominated strategy Down, other
player might find that that certain of their strategies are in fact dominated.
19
Non-Cooperative This is because Player 2 for instance, no longer has to worry how her
Game Theory-I strategies would perform against player l' s dominated strategies. So, some of
the strategies of player 2's, which are only good against player 1's dominated
strategies, might turn out to be bad strategies. In that case player 2 will not
play those strategies. This might lead to a third round of discovery of bad
strategy by some of the other players and so on.
Table 20.11a

Player 2
Player 1 Left Right
Up (1,1) (0,1)
Middle (0,2) (1,0)

.Thus if it is known to Player 2 that Player 1 will never play Down, the Right
looks dominated to her. Therefore the rational Player 2 would never play
Right. But then the Player 1 should not worry about Player 2 playing Right.
. Hence she should choose the strategy Up.
Table 20.11b

Player 2
Left
Player 1 Up (1,1)
Middle (0,2)

The strategy (Up, Left) is said to be reached by iterated elimination of

dominated strategies (IEDS) and the game is said to be dominance solvable. If
we are able to reach a unique strategy by following this procedure, we call the
outcome the solution to IEDS and call the game dominance solvable.

Table 20.11c

Player 2
Left
Player 1 Up (1,1)

However the process of iterated elimination of dominated strategies (lEDS)

has two drawbacks. First, it is based on the assumption of common knowledge
that the players are rational. That is, we need to assume not only that all the
players are rational, but also that all the players know that all the players are
rational, and that all the players know that all the players know that all the
players are rational and so on (the formal defmition of common knowledge).
Second drawback of iterated elimination of dominated strategies (IEDS) is
that the process often produces a very imprecise prediction about the play of
the game. Consider the following game:
Table 20.12: Game Without Dominated Strategy

Player 2
Left (L) Centre (C) Right (R)
Player 1 Top (T) (0,4) (4,0) (5,3)
Middle (M) (4,0) (0,4) (5,3)
Bottom (B) (3,5) (3,5) (6,6)
20
In this above game (Table20.12) there are no strictly dominated strategies to Static Games of
be eliminated. Since all the strategies in the game survive IEDS, the process Complete
Information
produces no prediction whatsoever about the play of the game.

Check Your Progress 2

1) What is a strictly dominated strategy?

................ , .

2) When would a player fmds a dominant strategy equilibrium ?

3) Solve the following game using iterated elimination of strictly dominated

strategies.

Player 2
left middle right
Player 1 up 1,0 1,2 0,1
down 0,3 0,1 2,0

20.7 SOLUTION CONCEPT UNDER THE NASH

EQUI~mRIUM IN PURE STRATEGY
Many simultaneous games have no dominant strategy or dominated strategy
(like Battle of Sexes game). We take up as a next step to find a solution ofthe
game in which every player or agent does the best she can, given the strategies
of the other players.
A pair of strategies satisfies Nash equilibrium condition if players' strategies
are mutually the best response strategy. The outcome is strategically stable or
self enforcing because no single player has an incentive to deviate from her
strategy.
21
Non-Cooperative In a normal form of n players game specified asG = {Sv S2, ..... Si, Uv U2,
Game Theory-I
..... ud, the strategies (s;, si, ..... , s~) are a Nash Equilibrium if, for each
playeri, si is player I's best response to the strategies specified for the (n-I)
ot her payers,
1 * *
(S1'S2' * Si+1'
si-1' * , Sn*). .

The above equation (20.3) says that each player i, in playing si .is playing a
best response to the other players' strategy choice. This condition includes
two requirements of Nash equilibrium:
1) Each player must be playing a best response against a conjecture
2) The conjecture must be correct.

Thus Nash equilibrium includes the first requirement because si is the ith
player's best response against the conjecture s~i(optimal strategy choice of all
the other (n-l) players (except the ith player». It includes the second
requirement because no player has an incentive to change her strategy (from
si). Hence si is stable and each player's conjecture is correct.

For every feasible strategy Si in Si> that is si solves

Maxui(S;, S2'" si-1' Si, S1+1 , s~) (20.6)

Suppose game theory offers the strategies (s~, s;, ..... , s~) as a solution to the
normal form of game G = {Sv S2, ..... Si, Uv U2, ..... ud. Saying that
(s~, s;, ..... , s~) is not a Nash equilibrium of G is equivalent of saying that
there exists some player i such that s; is not a best response to
(s~, s;, ... S;-1' S;+1'" s~). That is there exists some s;'inSi, such that

Ui(S~,S;, S;_1,S;,S;+1 .. ,S~)<Ui(S~,S;, S;_1,S;',S;+1"'S~) (20.7)

Thus if the theory offers the strategy (s~, s;, ..... , s~), as solution but these
strategies are not a Nash equilibrium, then atleast one player will have an
incentive to deviate from the theory's prediction, so the theory will be
falsified by the actual play of the game.

Consider the case of two player 1 and.2, each with two strategies (av a2) for
the player 1 and (bv b2) for the player 2. For example, (az, bt) is a Nash
equilibrium if and only if :
(a). 1f1 (a2, bt) ~ 1f1 (av bt) (20.8a)
(b).1f2(az,bt) ~ 1fz(az,bz), (20.8b)

20.7.1 Minimax Method for Zero Sum Game

Zero-sum is a situation in game theory in which one person's gain is
equivalent to another's loss, so the net change in benefit is zero. A zero-sum
game may have as few as two players, or millions of participants. Zero-sum
games are found in game theory, but are less common than non-zero sum
games. Poker and gambling are popular examples of zero-sum games since
the sum of the amounts won by some players equals the combined losses of
the others. Games like chess and tennis, where there is one winner and one
loser, are also zero-sum games. In the fmancial markets, options and futures
are examples of zero-sum games, excluding transaction costs. For every
22 person who gains on a contract, there is a counter-party who loses. The game.
of matching pennies (See Figure 20.1) is as an example of a zero-sum game. Static Games of
Zero-sum games are most often solved with the minimax theorem which is Complete
Information
closely related to linear programming duality or with Nash equilibrium.
Another variant of zero sum game is the constant sum games where the sum
of the payoffs of the players from each outcome is always a constant.

In contrast, situations where participants can all gain or suffer loss together
are referred to as non-zero-sum. Thus in non-zero-sum games are games in
which the sum of gains and losses by the players are sometimes more or less
than what they began with.

In the equilibrium of the zero sum game, each player is choosing her or her
best response, given her beliefs about what the others will do. In anticipating
such best responses, each player will expect to receive the worst payoff
associated with each action and will choose her own action accordingly. She
is thus choosing her best payoff from among the set of worst pay offs.

In some game most notably in sports contexts, the interests of the two sides
are exactly opposite to each other. Then, for each combination of players'
choices, the payoff of one can be obtained by reversing the sign of the payoff
of the other. So, the payoff matrix under a zero sum game can be
simplistically represented by showing the payoff of one player. Consider the
following example of zero sum game where in a soccer game the two players
(Offense and Defense) are represented row wise and column wise with the
strategies respectively (Run, Short Pass, Medium Pass and Long Pass) and
(Run, Pass, Blitz). Suppose the payoff table shows the row player's (Offense)
payoff and therefore it is implicit that the Column player's (Defense) payoff
will be the negative of the same. The Offense wants the outcome to be a cell
with as high a number as possible in each column. The Defense (column)
wants the outcome of the cell with as Iowa number as possible. See Table 20-
20.13.

Using this logic the Offense figures that for each of her row, the defense or the
Column will choose the column with the lowest number in that row. Therefore
the Row should choose the row that gives her the highest among the lowest
numbers i.e., maximum among the minima - the maxi-min strategy.
Similarly, the Columns reckon that, for each of her columns, Row will choose
the row with the largest number in that column. The Column should choose
the column with the smallest number among the largest one or the minimum
among the maxima - called the mini-max strategy. If the Offense (row's)
maximin values and the defense's minimax values are in the same cell of the
payoff matrix, then the outcome is the Nash equilibrium of the zero sum
game - a concept which we will elaborate in details in the next section. The
method of finding equilibrium in the zero sum game is called maximin-
minima x method. The equilibrium will lead to Nash equilibrium in pure
strategies, if it exists.

We begin by finding the offense's worst payoff from each her strategy (which
is lowest number in each row) and defense's worst payoff in each strategy
(which should be the highest number from each column). Thus the offense's
worst payoff from Run is 2, worst payoff from Short Pass is 5.6, worst payoff
from Medium Pass is 1 and worst payoff from Long Pass is -2. We write the
minimum of each row at the last column. The Defense's worst payoff from
Run is 10, worst payoff from Pass is 5.6 and worst payoff from Blitz is 13.
We write the maximum of each column at the bottom of the column.
23
Non-Cooperative The next step is to find out the best of each player's worst possible outcomes,
Game Theory-I i.e., the largest row minimum and the smallest column maximum. The largest
row minima is 5.6, so the offense can ensure itself a gain of 5.6 yards by
playing the Short Pass, this is her maximin. The lowest of the column maxima
is 5.6, so the defense can be sure of holding the offense down to a gain of 5.6
yards by deploying its Pass Defense. This is the defense' s minimax.

Now from this two given strategy choices, we see the maxim in and the
minimax values are belonging to the same cell of the game table. Thus offense
maximinis the best response to the defense' s minimax strategy and vice versa,
which is basically the Nash equilibrium of the game. The equilibrium entails
the offense attempting a Short Pass while the defense defends against a Pass.
A total 5.6 yards will be gained by the offense (and given up by the defense).

Table 20.13: Minimax Method

Defense
Run Pass Blitz RowMin
Run 2 5 13 Min=2
Offence Short Pass 6 5.6 10.5 Min=5.6
Medium Pass 6 4.5 1 Min= 1
Long Pass 10 3 -2 Min=-2
Column Max Max=lO Max=B Max= 13

If the minimax method fails to fmd an equilibrium in some zero sum game,
then we conclude that the game has no Nash equilibrium is pure strategies (we
will address the concept of mixed strategy equilibrium in the next section).
Moreover this minimax method cannot be applied to the non zero sum games.
As in those games the opponent's best is not necessarily the defense's worst.

20.7.2 Nash Parables

There are various other ways in which Nash equilibrium concept has been
motivated within game theory. These motivations are parables in the sense
that we will only offer a verbal description of each one. Some of these
motivations have been precisely worked out in mathematical models, some
are accepted as simple intuitive proposition. In either case these parables are
worth telling because Nash equilibrium is the most widely used solution
concept in game theory. The parables will reassure you the reasonableness of
the solution concept.

a) Play Prescription
One can think of a Nash equilibriums*as a prescription for play. If the
strategy vector is proposed to the players, then it is stable prescription in
the sense that no one has an incentive to play otherwise. By playing an
alternative strategy, a player would simply lower her payoffs, if she
thinks that others are going to follow their part of the prescription.
b) Pre-play Communication
How would the players in a game find their way to a Nash equilibrium?
One answer that has been proposed is that they could coordinate on a
Nash equilibrium by way of pre-play communication,· i.e., they could
coordinate by meeting before (the game is actually played) and discussed
24 their options. It is not credible for the players to agree on anything that is
 not a Nash equilibrium because atleast one player would be worse off Static Games of
against such agreement. Complete
Information

c) Rational Introspection
A related motivation is rational introspection: each player could ask
himself what she expects will be the outcome to a game. Some candidate
outcome will appear unreasonable in that there are players who could do
better than they are doing, i.e., there will be players not playing a best
response. The, only time no players appears to be making mistake is when
each is playing a best response, i.e., when we are at a Nash equilibrium.

d) Focal Point
Another motivation is the idea that a Nash equilibrium forms a focal
point for the players in a game, The intuitive idea of a focal point was
first advanced by Thomas Schelling in 1960 in his book: The Strategy of
Conflict. It refers to a strategy vector that stands out from the other
strategy vectors because of some distinguishing characteristics. Consider
the two Coordination game: (i) two players would have to write down
either heads or tails. They will be paid if their choices match. (ii) To
players have to meet in Canaught Place, New Delhi and have to choose a
time for their meeting. Again they are paid only when their chosen time
coincide. Schelling found that in a disproportionate number of cases,
students choose head in the case (i) and 12:00 pm in the case of (ii). A
. Nash equilibrium strategy vector is a focal point because it has the
distinguishing characteristics that each player plays a best response under
that strategy vector.

e) Trial and Error

If player started by playing a strategy vector that is not a Nash
equilibrium, somebody would discover that she could do better. If she
changes her strategy choice and we are still not in a Nash equilibrium,
somebody else might want to change her strategy. This process of trial
and error will go on till we reach Nash equilibrium and then nobody has
an incentive to deviate from that strategy choice.

20.7.3 Some Common Examples of Nash Equilibrium

Let us now examine the Nash equilibrium concept in the well known strategic
games (Figure 20.2 - 20.9), discussed in the Section 20.5.

1) Battle of Sexes (See Table-20.2)

Here the two players (Husband and Wife) have two strategy choices
Football (F)and Movie (M). This game has two Nash equilibrium (H, H)
and (M, M), i.e., either the couple go for Football together or Movie.

The best response to the player 1 i.e., Husband, is to play (F) against his
wife's choice of (F) and earns a payoff of 3. If the Husband deviates and
play (M) against his Wife choice of (F) he is worse off as he gets 0
(because 3 > 0). Similarly, playing F is the optimal strategy for Wife
(with a payoff 1) against her Husband's strategy of (F). If she deviates
and play (M) against her husband's choice of (F), she is worse off as
compared to the strategy choices (F, F), because 1>0.

Similarly, the best response to the Husband, is to play (M) against his
wife's choice of (M) and earns a payoff of 1. If the Husband deviates and
25 '
Non-Cooperative play (F) against his Wife choice of (M) he is worse off as he gets 0
Game Theory-I (because 1 > 0). Similarly, playing M is the optimal strategy for Wife
(with a payoff 3) against her Husband's strategy of (M). If she deviates
and play (F) against her husband's choice of (M), she is worse off as
compared to the strategy choices (M, M), because 3>0.

So, (H, H) and (M, M) are the two best response strategies of the couple
and are therefore Nash equilibrium.

2) Prisoner's Dilemma (See Table - 20.3)

Here (Confess, Confess) is the mutually best response strategy of the 2
convicts and therefore the unique Nash equilibrium of the Prisoners'
Dilemma game where each convict will be imprisoned for 6 years
(rrl (C, C) = -6, similarlyrrz(C, C) = -6). Given that one of them
stick to the strategy Confess, none of the convicts has an incentive to
deviate from the strategy Confess. Because if any of them, deviates she
will be worse off (e.g., (rrl (NC, C) = -9, similarlyrrz(C, NC) = -9).

If both the convicts play Non-Confession (NC, NC) the suspected crime
cannot be established' because of the absence of eye witness and therefore
they would get a nominal punishment of 1 year imprisonment
(rrl(NC,NC) = O,rrz(NC,NC) = 0). But if any of the players defects
from (NC, NC) and confesses his payoff would increased as he would be
set free for his honesty (asrrl (C, NC) = 0 andrrz(NC, C) = 0). So for
each of the player, there is an incentive to deviate from the strategy Not
Confess (NC). Moreover, if any of the player ditches from NC (given the
other stick to NC) it becomes really costly for the player who has not
deviated. Thus if the Convict 1 deviates from the (NC) and plays C,
whereas the Convict 2 plays NC, it is Convict 2 who will be' punished for
9 years of imprisonment and the Convict 1 will be set free (i.e.,
rrz (C, NC) = -9, rrl (NC, NC) = 0,). Similarly, you may check the case
when the Convict 2 deviates from (NC) and plays Confess (C), where as
the Convict 1 plays (NC). In that case the Convicts 1's payoff will be 9
years of imprisonment and the Convict 2 will be set free (rrl (C, NC) =
-9), rrz(NC, C) = 0)).

Hence the strategy (NC, NC) cannot be sustained as an equilibrium

strategy. Both the convicts face a credible threat that given he participates
(i.e., plays NC), his rivals may defects and push him to 9 long years of
imprisonment. This mutual threat pushes them, to play safe by choosing
(Confess, Confess)i.e., (C, C). This strategy (C, C) is the mutually best
response strategy or Nash equilibriumwith a payoff = (-6, -6), although it
is not socially Pareto optimal. For, both of them would have been better
.off by playing (NC, NC).

3) Hog and Dove or Chicken Game (See Table-20.6 and Table-20.S)

So far the Chicken game is concerned we have mentioned in the earlier
section that it is mutually beneficial for the players to play different
strategies. Because the loss of swerving is so trivial compared to the
crash that occurs if nobody swerves, the reasonable strategy would seem
to be to swerve before a crash is likely. Yet, knowing this, if one believes
one's opponent to be reasonable, one may well decide not to swerve at
all, in the belief that he will be reasonable and decide to swerve, leaving
the other player the winner. This unstable situation can be formalized by
26 saying there is more than one Nash equilibrium, which is a pair of
 Static Games of
strategies for which neither player gains by changing his own strategy Complete
while the other stays the same. In this case, the pure strategy equilibria Information
are the two situations wherein one player swerves while the other does
not i.e., (St, Sw) and (Sw, St). See Table - 20.6.

For any matrix with the payoffs in Figure 20.8, where W > T > L > X, the
Hog Dove game has multiple Nash equilibrium i.e., (H, D) and (D, H).
Consider the strategy (H, D). Given that Player I plays H, it is optimal
for the Player 2 to play D because his payoff from playing D (i.e., L)
exceeds his payoff from playing H (L > X). Similarly, given Player 2
plays D, it is optimal for the Player 1 to play H because his payoff from
playing H (i.e., W) exceeds his payoff from playing H (which is X) since
(W > X). Therefore the strategy (H, D) is mutually the best response
strategy or the Nash equilibrium. You may also check the strategy (D, H),
by the similar reasoning it is the mutually best response strategy or the
N ash equilibrium.

You may also check why the strategies like (H,H) or (D, D) are not the
Nash equilibrium. In such cases, given the other player's strategy, each of
the players has an incentive to deviate.

4) Coordination Game
In game theory, coordination games are a class of games with multiple
. pure strategy Nash equilibria in which players choose the same or
corresponding strategies. Consider the ,Minimum-Effort Coordination
game given in Figure 20.9. In this game it is optimal for both the players'
to play the same strategy i.e., putting low effort (1,1) or high effort (2, 2)
as the best response strategy at the equilibrium. So, (1,1) or (2, 2) are the
Nash equilibrium strategy.

Notice that the low effort outcome (1, 1) is an equilibrium, since a costly
increase in effort by only one person will not raise the amount produced.
Given Player 1 plays to put low effort(l), it is optimal for the player 2, to
put low effort, because (1- c) > (1- 2c), with 0 < c <1. Similarly, given
Player 2 plays to put low effort, it is optimal for the player 1 to put low
effort because (1- c) > (1- 2c). As long as c is less than 1, payoffs are
maximized when all players choose the highest possible effort. So, high
effort outcome (2, 2) is also mutually best response strategy and therefore
a Nash equilibrium. In this case you may check that a reduction in effort
by only one person will lower the minimum by more than the cost
savings c.

Note that any common effort level constitutes a Nash equilibrium, since a
costly unilateral increase in effort will not raise the minimum and a
unilateral decrease will reduce the minimum by more than the cost when
c < 1. However, this argument does not depend on the number of players,
so noncritical changes in c and n will not alter the set of Nash equilibria,
despite the reasonable expectation that efforts should be high for
sufficiently low effort costs and low numbers of participants. See Table-
20.9.

20.7.4 Relation between the Solution Concepts

The general relation between IEDS and Nash equilibrium can be summarized
as follows:
27
Non-Cooperative For any game in which there is an outcome to IEDS. It must be the case that
Game Theory-I the outcome is a Nash equilibrium. The iterated elimination of strictly
dominated strategies eliminates all but the one strategy which is the Nash
equilibrium of the game. In a n player Normal form of a game, jf the strategy
is a Nash equilibrium then they survived the iterated elimination of strictly
dominated strategies. However not every Nash equilibrium can be obtained as
the outcome to IEDS.
. '

, Consider a 2 player 3 strategy game. Suppose the strategy (a2, b3) is the IEDS
solution. In order to establish that this pair is also a Nash equilibrium, we need
to show that a2 has a higher payoff than a1 or a3 when played against b3,
Likewise b3 has a higher payoff than b2 or b1 against player l' s strategy a2,
Suppose a1 is the strategy of the player 1 that is eliminated in the last round of
elimination and likewise b1 is the last round of elimination. We can now
proceed by the following three step proof:

Step 1: Regardless of whether player 2 plays b30r b'', a2 has a payoff as high
as ai, In particular against b 3, a 2 is at least as good a response as ai,

Step 2: Similarly.b ' is as good aresponse for player 2 as bi against a2,

How is a2 's payoff compare with those of a3? Since 3 got eliminated in the
very first round it must have been dominated by either/ 1. If it is dominated by
2 we are done. (Why?) If it is dominated by 1, then we know that it does
worse than 1, against player 2's strategy of 3, But in turn 1 has no higher a
payoff than 2 against that same strategy 3, We have therefore concluded:
Step 3: For player 12 is the best response against+, Likewise+ is the best
response against 2,
Table 20.14
Player 2
1 2 3
Player 1 1
2
3

The reverse implication is not necessarily true; there may be N ash equilibria
that are not the solution to IEDS, This may be easy to see in any game, such
as Battle of Sexes, in which there is a Nash equilibrium but the game is not
dominance solvable. Another situation where Nash does not imply IEDS can
arise when there are multiple Nash equilibria; some of these Nash equilibria
would be eliminated by iterated elimination of dominant strategies.

As regards the relation between dominant strategy solution and Nash

equilibrium, exactly the same results hold. First note that every dominant
strategy solution is an I~DS outcome, if there is dominant strategy every other
strategy is immediately eliminated. But we just learnt that every IEDS
outcome is a Nash equilibrium, It follows therefore that every dominant
strategy solution is a Nash equilibrium. The converse need not holds" The
coordination game has a dominant strategy solution however is also a Nash
equilibrium.

28
 Static Games of
Check Your' Progress 3 Complete
Information
1) What is minimax strategy?
.......................................................................................
......................................................................................
......................................................................................
......................................................................................
......................................................................................

2) Find solution to the following game for pure strategy.

L R
T (3,1) (2,-20)
M (5,0) (-10,1)
B (-10,0) (24,2)

......................................................................................
......................................................................................
......................................................................................
. .

......................................................................................
3) How do you fmd a relation between elimination of dominated strategies
(IEDS) and Nash equilibrium .
......................................................................................
......................................................................................
......................................................................................
......................................................................................

20.8 MIXED STRATEGY NASH EQUILIBRIUM

This section will present more general class of strategy called mixed strategies
that players can profitably employ. Players or agents could follow a strategy
of randomizing over the set of available actions according to some probability
distribution. Such a strategy is called Mixed strategy.

20.8.1 Definition and Example

Consider the example of Battle of Sexes game (given in Figure 20.5.2). You
may recall that you have two pure strategy Nash equilibrium (f, f) and (m, m).
Actually those are not the only choices. For instance the players can toss a
coin. If it is head, they can go to the Football match and if it is tail they can go
to the Movie instead. Hence there are actually three strategies to choose from:

i) (F, F)
ii) (M, M) and
iii) A coin toss i.e., a strategy with equal likelihood to going to Football or
Movie. .
29
r

Non-Cooperative Note that the coin toss has clearly enlarged the set of initial choices as strategy
Game Theory-I (iii). is evidently distinct from (i) and (ii). Strategies (i) and (ii) as well as
every other strategy discussed so far are called Pure Strategies. A strategy like
(iii) is called the Mixed Strategy. In fact strategy (iii) is just one of the many
mixed strategies.

Definition: Suppose a player has M pure strategiess", S2, , SM. A

mixed strategy for the player is the probability distribution over his pure
strategies; i.e., it is a probability vector (pI, p2, , pM) with pk ~ 0, k =
1,2, , M and Lf1=1 pk =1. Every pure strategy is in a trivial sense also a
mixed strategy in which pI = 1and pk = 0, for all other pure strategies.

The obvious question now is how a mixed strategy payoff will be evaluated?
It follows that the right way to evaluate the uncertainty inherent in .a mixed
strategy is to take its expected payoff. Expected payoffs are computed in two
steps:
Step 1: Weight the payoff to each pure strategy by the probability with which
that strategy is played.
Step 2: Add up the weighted payoffs.

Let us recall the Battle Sexes game as shown in Table-20.14. Let aand f3
denote the profile of mixed strategies of the Husband and Wife. Which mean
to say that, Husband is playing the F and M with probabilities a and(1 - a)
and the Wife is playing the F and M with probabilitiesf3 and (1 - f3).Let
TCH (., .) and TCw (. , .) be the payoff functions of the Husband and Wife,
respectively'. Thus, TCH(F, F) denotes the payoff of Husband when both the
Husband and Wife are going for Football. Similarly, TCH(M, F) denotes the
payoff of Husband when Husband going for Football and Wife going for
Movie. Similarly, you can interpret TCw(F, M) and TCw(F, F).
The Battle of Sexes game has 4 possible outcomes (F, F), (F, M); (M, F), (M,
M).

The outcome (F, F) occurs with a probability (a, f3)

The outcome (F, M) occurs with a probability {(a. (1 - f3)}
The outcome (M, F) occurs with a probability{(l - a). f3)}
The outcome (M, M') occurs with a probability {(1 - a). (1 - f3)}

We will now calculate the. expected payoffs of Husbands..L, f3), for playing,
the.pure strategies (F) and '(M) under the assumption that the Wife is playing
the Mixed Strategy. Thus, CH(F, f3) = Expected payoff of Husband for
playing the pure strategies (F), under the assumption that the Wife is playing
the Mixed Strategy f3, (1 - f3). Thus,
cH(F,f3) = f3 TCH(F, F) + (1- f3) TCH(F, M)=f3. 3 + (1 - f3). 0 = 3f3 .. (20.9)
CH(M, f3)=Expected payoff of Husband for playing the pure strategies (M),
under the assumption that the Wife is playing the Mixed Strategy f3, (1 - f3).
Thus,
CH(M, f3) = f3 TCH(M, F) + (1 - f3) TCH(M, M)= f3.0 + (1"'" f3).1 = (1- f3)
. (20.10)
Similarly, we will now calculate the expected payoffs of the Wife
cw(F, a), for playing the pure strategies (F) and (M) under the assumption
30 that the Husband is playing the Mixed Strategy with probabilities a, (1 - a).
Thus, Ew(F, a) = Expected payoff of Wife for playing the pure strategies (F), Static Games of
under the assumption that the Husband is playing the Mixed -Strategy Complete
Information
a, (1- a).
That is, Ew(a, F) = a 1fw(F, F) + (1 - a) 1fH(M, F) = a .1 + +(1 - a). 0=
a (20.11)
Ew(M, a) = Expected payoff of Wife for playing the pure strategies (M),
under the assumption that the Husband is playing the Mixed Strategy a, (1-
a). So,
Ew(a, M) = a 1fw(F, M) + (1 - a) 1fH(M, M)= a.O + +(1 - a). 3 =
3(1 - a) (20.12)
Table 20.14: Battle of Sexes under Mixed Strategy
Wife
Football (F) Movie(M)
p (1- P)
Football 1fH(F, F), 1fw(F, F) 1fH(F, M), 1fw(F, M)
Husband (F) , = (3, 1) = (0, 0)
(a)
Movie 1fH(M, F), 1fw(M, F) 1fH(M, M), 1fw(M, M)
(M)' = (0, 0) = (1, 3)
(1~ a)

Under the Mixed Strategy the expected payoffs of Husband from F and M
should be equal (if the expected payoffs are not equal then there can't be
mixed strategy equilibrium).
Therefore equating equation (i) and (ii) for Husband. Thus,
EH(F,f3) = EH(M,f3)
1 3
~ 3f3 = (1- f3) ~ f3 = 4~ (1- f3) = 4:·········· (20.13)
Repeating the same for the Wife, by equating (iii) and (iv) we get:
EW(a, F) = Ew(a,M)
3 1
~ a = 3(1 - a~ ~ a = "4 ~ (1 - a) = "4 (20.14)

Therefore, the Mixed Strategy Nash equilibrium is:

{a, (1- a),f3, (1- f3)} = {G,D, G,D}

Thus by extending the defmition of Nash equilibrium to allow for mixed'
strategy, we simply require that each player's mixed strategy be a best
response to other player's mixed strategy. Thus a, (1- a) is the best response
of the Husband against the Wife's mixed strategy f3, (1 - f3) and vice versa.
Since any pure strategy can be represented as the mixed strategy that puts zero
probability on all of the player's pure strategies, this extended definition
subsumes the earlier one.
In a two player normal form of game G = {5v 52' Uv U2}, the mixed strategies
( a", f3*) are a Nash equilibrium if each player's mixed strategy is a best
response to the other player's mixed strategy, so that, a* must satisfy:
1fl (a*, f3*) ~ 1fl (a, f3*) for every probability distribution over 51 and f3* must
satisfy: 1f2(a*,f3*) ~ 1f2(a*,f3), for every probability distribution over 52'
31
r

Non-Cooperative Plugging the values of a, (1- a), and {3,(1- {3),we can derive the
Game Theory-I numerical values of the Expected payoffs of the Husband and Wife from
equations (20.9), (20.10), (20.11) and (20.12).
From (i) expected payoff of the Husband from the pure strategy F:cH(F, {3) =
3{3= 3). = ~
4 4

From (ii) expected payoff of the Husband from the pure strategy
M:CH(M,{3) = (I - {3) = ~
From (iii) expected payoff of the Wife from the pure strategy F:cw(a, F) =
3
a=-
4

From (iv) expected payoff of the Wife from the pure strategy M:cw(a, M) =
3(1- a) = ~
4

The above values of expected payoffs of the Husband and Wife are derived
for their pure strategies of F and M when the opponent is playing the mixed
strategies.
Now let us consider that both Husband and Wife are playing the mixed
strategies. The expected payoff of the Husband from the mixed strategies is
CH(a,{3) = a.cH(F,{3)+ (1- a)cH(M,{3)
= a{{3 TrH(F, F) + (1 - {3)TrH(F, M)} + (1 - a){ {3TrH(M, F) +
(1 - {3)TrH(M, M)} (20.15)

Expected payoff of the Wife from the mixed strategies is:

cw(a, {3) = {3cw(a, F) + (1 - {3)cw(a, M)
= {3{ a Trw(F, F) + (1- a) TrH(M, F)} + (1- {3){ a Trw(F, M) + (1-
a) TrH(M,M)} ... (20.16)
13 33 3
=-- + --=-
44 44 4

Thus in .the equilibrium the expected payoffs of both the Husband and Wife
when both of them are playing a mixed strategy are equalized. Moreover this
value is equal to the payoff of the Husband or Wife under pure strategy when
the opponent is playing mixed strategy at the equilibrium (a*,{3*).==>

CH(a,{3) = cw(a,{3) = cH(F,{3) = cH(M,{3) = cw(a,M) = cw(a',F)

It should be noted in the finite player and finite strategies there are games
with no Nash equilibrium in pure strategies but there will always be such an
equilibrium in mixed strategies.
There are 3 possibilities, viz.,
1) If cH(F,{3) > cH(M,{3) ==> Husband's unique best response is to play
pure strategy F.

2)1 f CH(M, {3) > CH(F, {3) ==> Husband's unique best response is to play
pure strategy M.
3) If cH(F,{3) = cH(M,{3) ==> All mixed strategies of the Husband yields
the same expected payoff and hence all are the best responses. Same way
you can do it for the Wife.
32
 /
For the ~esp~nse function of the Husband: Static Games of
Complete
1 .
1) For fJ = -4 ==>cH(F,fJ) = cH(M,fJ) ==>Exp. Payoff of Husband from all Information

mixed strategy givesher the same payoff. Thus every mixed strategy is
his best response.
2) If fJ < ~4 ==>cH(F,fJ) < cH(M,fJ) ==>Exp. Payoff of Husband from M
exceeds his expected payoff from all mixed strategy that assign positive
probability to F.==>Unique best response for Husband is M.
1
3) If fJ > - ==>cH(F,fJ) > cH(M,fJ) ==>Expected payoff of Husband from
4
F exceeds his expected payoff from all mixed strategy that assign positive
probability to M. That impliesunique best response for Husband is F.
So, the Husband's best response function is:
1
0, and fJ < 4"
1
a: 05: a 5: l,andfJ = 4"
1
1, fJ > 4"
Similarly, Wife's best response is:

. { 0, and a < 4"

3
Bw(a) = 3
fJ: 0 5: fJ 5: 1, and a = 4"
3
1, a>- 4
This game has three mixed strategy Nash equilibrium (as evident from three
points of intersections of the Player 1 and Player 2's Best Response
functions): (a, fJ) = (0, 0), (~,4 ~), (1, 1). See Figure 20.1.
4

Proposition: Every strategic game in which each player has finitely many
strategies has a mixed strategy Nash equilibrium.

fJ
/1\
(1,1)
1

1
4 ~------------------~-------
a
-
3
(0,0) 1
4

Fig. 20.1: Mixed Strategy of Battle of Sexes Game

33
Non-Cooperative Recall in the mixed strategy equilibrium we have assumed CH(F, P) =
Game Theory-I
CH(M, p)and cw(a, M) = cw(a, F)in the equilibrium.
Proof: Let us assume statement 1 and 2 hold.
1) The Husband is playing the mixed strategy with probabilities a and (1 -
a)for F and M respectively (0< a<I).
2) But the expected payoff of the husband from F and M (when Wife is
playing the mixed strategy) are not equal i.e., cH(F, P) =1= CH(M, P).
Let us assume cH(F,P) > cH(M,P)
Now from 20.15, we know, the expected payoff of.the Husband when both of
them are playing the mixed strategies is CH(a,p) = a,cH(F,P)+ (1-
a)cH(M,P)
Differentiating the above equation with respect to a (the probability with
which husband plays the F):8e~~,p)=1.cH(F,P)(-1),cH(M,P) =
CH(F, P) - cH(M, P) >0 (because of the assumptionj=e 8e~~,p) >0 ===>
Expected payoff of the Husband from the mixed strategy is increasing in c, As
increases the expected payoff of the Husband from the mixed strategy i.e.,
CH(a,p) increases. So, it is optimal for the husband to put a = 1.
Therefore, the husband can maximize his expected payoff CH(a, P)by playing
F with probability 1 and Movie with probability O.
Hence the contradiction to the statement 1 and in equilibrium CH(F, P) =
CH(M,P)·
Similarly, we can show,
8eH(a,p)
if CH()F,p < CH(M,P ) ===> 8a
1. cH(F, P)( -l),cH(M, P)= cH(F, P) - CH(M, P) < 0 (because of the
assumptionj=e 8eH(a,p) < 0 ===>
8a
Expected payoff of the Husband from the mixed strategy is decreasing in u,
===> As a increases CH(a, P) falls. So it is optimal for the husband to put a = O.
Thereforethe husband can maximize his expected payoff CH(a, P)by playing F
with probability 0 and M with probability 1.
Hence the contradiction. So in equilibrium.sj, (F, P) = CH(M, p)(proved)
Example: The following matrix is a partially specified payoff matrix of the
two player strategic game is given below. Both the players are playing mixed
strategy. Figures in the bracket against each strategy are showing the
probabilities of their mixed strategy. This example is illustrating a method of
checking whether a mixed strategy profile is a mixed strategy equilibrium.

Player 2
(a,p) L(P = 0) C (.!.) R (~)
3 3
3
Player 1 T(a =-) (.,2) (3,3) (1, 1)
4

M (0) (., .) (0, .) (2, .)

B (.!.) (.,4) (5, 1) (0,7)
4

34
r

Static Games of
So we have to check the indicated pair of mixed strategy {(~, 0, ~), (0, ~,~)} is
4 4 3 3 Complete'
a Mixed Strategy Nash equilibrium. This also suffices by the properties of Information
Mixed Strategy equilibrium. Note that Husband is playing T and B with
positive weights and Wife is playing C and R with positive weights. Consider
the player 1's expected payoff from 3 pure strategies (T, M, B) provided
player 2 plays mixed strategy: '
125
1) c1(T,[J) = O. (.) +-3
333
+ 1.-=-
2) Cl (M, [J) = O. (.) + -31 .0 + -23. .2 = -
4
3

3) Cl (B, [J) = O. (.) + -13 .5+ 2

- .0=
3
5
-
3

Expected payoff of the Husband when both Husband and Wife are playing
mixed strategy:

Cl (a, [J) = ~C1(T,[J) + O. cl (M, [J) + ~C1(B, [J) = (~. i)+ (0. ~) + ~·i i
=

Thus, in Mixed Strategy Nash equilibrium: C1(a*,[J*) = c1(T,[J*) =

Cl (B, [J*) = ~
3

Similarly, C1(a*,[J*) >

·33
c1(M,[J) -sbecause ~ >~. This is because player l's
mixed strategy assigns positive probabilities to T and B and zero probability
toM.
You may repeat the same exercise for the Player 2.

Check Your Progress 4

1) What is a mixed strategy?

.........................................................................................

........................................................................................

2) Consider the example of Matching of Pennies game (in Table- 20.1).

i) Show that there is no pure strategy Nash equilibrium.
ii) Show that if the Player 1 plays Head with probability p and the Player
2 plays Head for sure, then player 1's expected payoff is (2p -1).
iii) What is the Player 2's expected payoff?
.........................................................................................
........................................................................................

........................................................................................

3) Consider a Hawk-Dove (or Chicken) Game (t = Tough, c= Concede)

Player 2
Tough (t) Concede (c)
Player 1 Tough (t) (1, -1) (10,0)
Concede (c) (0, 10) (5,5)
35
Non-Cooperative
Consider the Mixed Strategy in which Player 1 plays t with probability+,
Game Theory-I . 11
while Player 2 plays t for sure. Then what is the player 1's expected
payoff?

4) Show that there is no Mixed Strategy Nash equilibrium in the following

game:

Player 2
Player 1 Left (L) Middle (M) Right (R)
Up (U) (1,0) (1,2) (0, 1)
Down (D) (0,3) (0, 1) (2,0)

5) Find the mixed strategy Nash equilibrium of the following game:

Player 2
L R
Player! T (2, 1) (0,2)
B (1,2) (3,0)

20.9 LET US SUM UP

A strategic form of game is described by the list of players the strategies
available to each player and the payoffs to any strategy combination, one
strategy for each player. A dominant strategy gives higher payoffs than every
other strategy regardless of what the other player's do. A dominant strategy
solution to a game exists when every player has a dominant strategy.
No rational player will play a dominated strategy but would rather play one of
his undominated strategies. A rational player would not expect his opponent to
play a dominated strategy either. Elimination of dominated strategies can lead
36 to a chain of reaction that successively narrows down how a group of rational
players will act. If there is eventually a unique prediction, it is called the IEDS Static Games of
solution. Complete
Information
Nash equilibrium is the most popular solution concept with wide range of
applications. A strategy combination is a Nash equilibrium if each player's
.strategy choice is a best response against her opponent's choices in that
combination. Any IEDS solution has to be a Nash equilibrium however there
may be Nash equilibrium that are not IEDS solutions. Same thing is true of a
dominant strategy solution, it must be a Nash equilibrium but the re~erse is
not true.
A mixed strategy is a probability distribution over a player's pure strategies;
not every pure strategy needs to be included in every mixed strategy. The
payoff to a mixed strategy is computed as expected payoff to its component of
pure strategies. A mixed strategy can dominate a pure strategy even if the
latter is undominated by every other pure strategy. There are games with no
Nash equilibrium in pure strategies but there will always be such an
equilibrium in mixed' strategies.

20.10 KEY WORDS

Common Knowledge: A fact is common knowledge if all players know it,
and know that they all know it, and so on. The structure of the game is often,
assumed to b~ common knowledge among the players.
Dominating Strategy: A strategy dominates another strategy of a player if it
always gives a better payoff to that player, regardless of what the other players'
are doing. It weakly dominates the other strategy if it is always at least as
good.
Extensive Game: An extensive game (or extensive form game) describes
with a tree how a game is played. It depicts the order in which players make
moves and the information each player has at each decision point.
Game Theory: Game theory is the formal study of decision-making where
several players must make choices that potentially affect the interests of the
other players.
Game: A game is a formal description of a strategic situation.
Mixed Strategy: A mixed strategy of a player in a strategic game is a
probability distribution over the player's action. As a special case, a mixed
strategy can be the deterministic choice of one of the given pure strategies.
Nash Equilibrium: A N ash equilibrium, also called strategic equilibrium, is
a list of strategies, one for each player, which has the property that no player
can unilaterally change strategy and get a better payoff.
Payoff: A payoff is a number, also called utility, that reflects the desirability
of an outcome to a player, for whatever reason. When the outcome is random,
payoffs are usually weighted with their probabilities. The expected payoff
incorporates the player's attitude towards risk.
Player: A player is an agent who makes decisions in a game.
Rationality: A player is said to be rational if she seeks to play in a manner
which maximizes her own payoff. It is often assumed that the rationality of all
players is common knowledge.
Shapley Value: States in which costs and rewards are distributed among a
number of participants in order of the importance of that each plays in the
cooperative activity. 37
r

Non-Cooperative Core: Set of feasible allocations that cannot be improved upon by a subset (a
Game Theory-I coalition) of the economy's consumers
Strategic Form: A game in strategic form, also called normal form, is a
compact representation of a game in which players simultaneously choose
their strategies. The resulting payoffs are presented in a table with a cell for
each strategy combination.
Strategy: In a game in strategic form, a strategy is one of the given possible
actions of a player. In an extensive game, a strategy is a complete pian of
choices, one for each decision point of the player. .
Zero-Sum Game: A game is said to be zero-sum if for any outcome, the sum
of the payoffs to all players is zero. In a two-player zero-sum game, one
player's gain is the other player's loss, so their interests are diametrically
opposed.

20.11 SOME USEFUL BOOKS

'1) Avinash K. Dixit(2015), Susan Skeath andDavid H. Reiley, Games of
Strategy, W. W. Norton and Company; 4 edition.
2) Drew Fudenberg and Jean Tirole(1991), Game Theory, MIT Press.
3) Drew Fudenberg and David Levine (1998), The Theory of Learning in
Games, MIT Press, ..
4) Robert Gibbons (1992), A Primer in Game Theory, Harvester
Wheatsheaf, .
5) Martin J. Osborne and Ariel Rubinstein (1994), A Course in Game
Theory, MIT Press.
6) Michael Maschler, Eilon Solan and Shmuel Zamir (1992), Game Theory,
Cambridge University Press.

20.12 CHECK YOUR PROGRESS

ANSWERSIHINTS
Check Your Progress 1
1) A game in which players are able to make enforceable contracts.
2) Games in which players make moves at different times or in turn.
3) If all of the players know it (it is mutual knowledge) and all of the
players know that all other players know it and all other players know
that all other players know that all other players know it, and so on.
4) A strategy consisting of possible moves and a probability distribution
(collection of weights) which corresponds to how frequently each move
is to be played.
5) A set of strategies, one for each player, such that no player has incentive
to unilaterally change her action.
6) In a game where players are unable to make enforceable contracts
outside of those specifically modeled in the game.
7) Payoffs are numbers which represent the motivations of players. These
may represent profit, quantity, "utility", or other continuous measures
(cardinal payoffs), or may simply rank the desirability of outcomes
38 (ordinal payoffs).
r
I
I
I
I

8) A specific move or action that a player will follow in every possible Static Games of
I attainable situation in a game. Complete
i Information
I 9) A strategy is weakly dominant if, regardless of what any other players
I

I
do, the strategy earns a player a payoff at least as high as any other
I strategy, and, the strategy earns a strictly higher payoff for some profile
I
I of other players' strategies.
I' 10)

.. Player 2
L R
Player! L -1,1 1, -1
R 1, -1 -1,1

11)
I
I •
Player 2
H T
Player! H 1, 1 1, 1
T 1, 1 1, 1

12)

Player 2
H. T
Player! H 1, -1 -1,1
T -1,1 1, -1

13)

Player 2
H T
Player! H 1, 1 1, 1
T 1, 1 1, 1

Check Your Progress 2

1) A strategy is strictly dominated if there is some other strategy which offers
a higher utility, regardless of the strategies of the other players.

2) A dominant strategy equilibrium is reached when each player chooses

their own dominant strategy.

3) Player 1 has two strategies and player 2 has three. S'l= lup, down} and
S2={left, middle, right}. For player 1, neither up nor down is strictly
dominated. Up is better than down if 2 plays left (since 1>0), but down is
better than up if 2 plays right (since 2>0). For player 2, however, right is
strictly dominated by middle (since 2>1 and 1>0), so player 2 being
rational will not play right.

Thus if player 1 knows that player 2 is rational then player 1 can eliminate
right from player 2's strategy space. So, if player 1 knows that player 2 is
rational then player 1 can play the game as if it was the game depicted below.

39
Non-Cooperative
Player 2
Game Theory-I
left middle
Player 1 up 1,0 1,2
down 0,3 0,1
In the figure above, down is strictly dominated by up for player 1 and so if
player 1 is rational (and player 1 knows that player 2 is rational, so that the
second game applies) then player 1 will not play down. Consequently, if
player 2 knows that player 1 is rational and player 2 knows that player 1
knows that player 2 is rational (so that player 2 knows that the second game
applies) then player 2 can eliminate down from player l's strategy space,
leaving the game looking like below.

Player 2
left middle
player 1 Up 1,0 1,2

And now left is strictly dominated by middle for player 2, leaving (up,
middle) as the outcome of the game. This is process is called the iterated
elimination of strictly dominated strategies.

Check Your Progress 3

1) Minimizing the possible loss for a worst case (maximum loss) scenario.
2) The row player can get a value of 4 (if the other player plays R) or 5 (if the
other player plays L), so: vraw = 4. The column player can get 1 (if the
other player plays T), 1 (if M) or 4 (if B). Hence: v col = 1.
3) The iterated elimination of strictly dominated strategies eliminates all but
the one strategy which is the Nash equilibrium of the game.

Check Your Progress 4

1) A mixed strategy is one in which a player plays his available pure
strategies with certain probabilities
2) Solve yourself after reading carefully solution concepts through Nash
equilibrium.

3) Hint: When player 2 plays t for sure, player l' s expected payoff: ~(-1)
. 11
9 9
+ (1--).0
11 = --
11 When player 2 plays c for sure, then player l's
9 9 100
expected payoff: 11(10)+ (1-11).5= 11

4) Do yourself after reading Section 20.8

5) Do yourself reading after reading Section 20.8

20.13 EXERCISES
1) An employee (player 1) who works for a boss (player 2) can either work
(Q) or shirkrf'), while his boss can either monitor the employee (M) or
ignore him (I). Like most employee-boss relationships, if the employee is
working then the boss prefers not to monitor, but if the boss is not
monitoring then the employee prefers to shirk. The game is represented in
the following matrix:
40
 Static Games of
Player 2 Complete
Information
M I
Player! W 1, 1 1,2
S 0,2 2,1

a) Draw the best response functions of each player.

b) Find the N ash equilibrium of this game.
Answer: (a)
Let p be the probability that player 1 chooses Wand q the probability that
player 2 chooses M. It follows that Vi (W, q) > Vi (Si q) if and only if 1 > 2(1

_ q), or q > ~, and V2(P, M) > V2(P,I) if and only if p + 2(1 - p) > 2p + (1 -
2
1
p), or p <-. It follows that for player 1,
2

p= if q<-1

.
° 1
2
1
pE[O,I] if q =-
2
1
p= 1 if q>-
2

And for player 2

q= 1 ·f p<-1
1
2

qE[O,l] if p =~
2
if P>21
q=
° 1

Answer: (b)
Best response corresponds to Nash equilibrium (p, q) = G, D
2) Use the rninirnax method to get the mixed strategy of the game.
Solve the following game:
B

A Q ~J
Solution
It is a 2 x 2 matrix and no saddle point exists. So the solution is

!J !
2

v= a11a22 - a21a12 = 20- 3

(a11 +a22)-(a12 +a21) 9-4
. 41-
 Non-Cooperative v = 17/5
Game Theory-I
SA = (xl, x2) = (1/5, 4/5)
SB = (yI, y2) = (3/5, 2/5)
3) Consider the following game.
a b c d
w 2,0 0,5 1, ° 0,4

y
x 4,1
2,1
2,1
5,0
0,2
0,0
1,
0,3
°
Z 0,0 1, ° 4,1 0,0
a) Compute the set of mixed strategies.
b) Compute the set of all Nash equilibria.
Ans. (a)
w is dominated by a mixed strategy putting 2/3 on x and 1/3 on z.
a is dominated by mixed strategy putting 3/5 on c and 2/5 on d.
b is dominated by mixed strategy putting 3/5 on c and 2/5 on d.
y is dominated by a mixed strategy putting Y2 on x and Y2 on z
d is dominated by c.
x is dominated by z.
The set of rationalizable strategies is {z} x {c}
b) The only Nash equilbrium is (z,c) because there is only 1 mixed
strategy for each player
4) Let PI and P2 be the prices of a homogeneous product produced by firms
1 and 2 respectively. Each firm chooses their price Pi E [0, 00) . There is
no restriction on price and the firms choose their price simultaneously. Di
(Pi, pj) is the demand function of the ithfirm,

F(p) '" ···if P, < P,

1
- F(p.) ... ···if P,
2 I I
= P, J

° ······ifpj> Pj

Both the firms have a constant marginal cost, c. The payoff function of the ith
firm is given by 7lj (Pi' Pj) = (Pi - c) D, (Pi' Pj) •

Show that the Nash equilibrium of the game is (PI*=C, P2*=C).

Ans. Let (Pi*, pj*)be another Nash equilibrium of the game. Now there could
be following mutually exhaustive possibilities:

i) Case I Pi* > P j * > c ~ In this case the entire market is served by the Jh
firm. If the ith firm reduces its price slightly such that its price is below
the Jh firm but above the marginal cost, then it will get the entire market.
Therefore, Pi* is not the best response of the ith firm and hence such a
price configuration cannot be the N ash equilibrium.

42
r

•
. ii) Case 11 P, * > Pi* > c -+ Here the t firm is not behaving optimally (the Static Games of
Complete
argument is the same as case I). Information

iii) Case IIIPj* = Pi* > c -+ Given the equilibrium price of one firm the
other is not behaving optimally. Reducing price by ; (where ; is
negligibly small positive quantity) one firm can increase its profit given
the other firm is charging its equilibrium price.
iv) Case IV p/ > Pi* = C -+ Firm i is not doing its best. It can increase its
profit by charging a price, which is greater than c but slightly, lower than
the pricepj * .

v) Case V Pi * > p/= c -+ Here firm j is not behaving optimally (the

argument is same as case
Thus, we see there is no other combination of price in which both the
firms are doing their best. Therefore, no other price configuration other
than (Pl*=C, P2*~C) which is a Nash equilibrium.

5) Solve the following normal form game using the method of iterated
elimination of strictly dominated strategies.

Player 2
Left Middle Right
Player 1 Up (1,0) (1,2) (0, 1)
Down (0,3) (0, 1) (2,0)

Ans. The strategy "Right" of player 2 is strictly dominated by the strategy

Middle". Therefore, we can eliminate it from the table. The reduced table is
given below:

Player 2
Left Middle
Player 1 Up (1,0) (1,2)
Down (0,3) (0, 1)

Notice that in the reduced table for player 1, the strategy "Down" is
strictly dominated by the strategy "Up". Therefore, we can remove it
from the table. The further reduced table is as the following:

Player 2

Left Middle
Player 1
Up (1,0) (1,2)

Now, there is a single strategy left for player! and for player 2 the
strategy "Left" is strictly dominated by "Middle". Therefore, the
equilibrium is (Up, Middle)

6) Can we solve the following normal form game using the mini-max
method? If possible then solve it.

43
Non-Cooperative
Game Theory-I
Player 2
Left Middle Right
Up 5, 1 3,3 1,5
Player 1 Straight 6,0 2,4 1, 5
Down 1,5 0,6 0,6

Ans. It is possible to solve the game with minimax method because it is a

constant sum game (the constant being 6)
7) Consider the following game of 2 players. Suppose Player 1 has 3
strategies U, M, D, which are played with probabilities (0.2, 0.3 and 0.5)
respectively. Player 2 has 4 strategies L, Mi, M2 and R. Player 2 plays L
and M2 with probabilities 0.4 and 0.6. Show that player l's expected
payoff is 2.32. What about Player 2' s expected payoff?

Player 2
L Ml M2 R

Player 1
U
M
1,
2,4
° 4,2
2,0
2,4
2,2
3,1
2,1
D 4,2 1,4 2,0 3,1

8) Show that there is no Mixed Strategy Nash equilibrium in the Prisoner's

Dilemma Game.
9) Show that apart from the two pure strategy-equilibrium, there is also a
mixed strategy equilibrium in a chicken game (as given in Table-20.6).
10) Find out the unique Nash equilibrium in pure strategy in the following
game. Show that there is no Mixed Strategy Nash equilibrium in the
following game:

Player 2
Left (L) Centre (C) Right (R)
Top (T) (0,4) (4,0) (5,3)
Player I, Middle (M) (4,0) (0,4) (5,3)
Bottom (B) (3,5) (3,5) (6,6)

44
UNIT 21 STATIC GAMES WITH
COMPLETE INFORMATION:
APPLICATIONS
Structure
21.0 Objectives
21.1 Introduction
21.2 Game Theoretic Applications in Common Property Resources
21.2.1 Prisoners' Dilemma Resource Conservation Game
21.2.2 Chicken Game
21.2.2.1 Problem of the CPR Provision - Two Person Chicken Game
21.2.2.2· Chicken Game: Heterogeneous Group and Collective Action
21.2.3 Coordination Game in Fishing

21.3 Best Response Function

21.3.1 Synergistic Relationship

21.4 Quantity Competition and Price Competition

21.4.1 Cournot Competition
21.4.1.1 Comparative Statics
21.4.1.2 Coumot Competition vs Collusion
'21.4.1.3· Numerical Example
21.4.2 Bertrand Price Competition

21.5 War of Attrition

21.6 Hotteling's Location Game
21.7 Let Us Sum Up
21.8 Key Words
21.9 Some Useful Books
21.10 Check Your Progress AnswerslHints
21.11 Exercises

21.0 OBJECTIVES
After going through this unit; you will be able to:
• understand the nature of strategic interactions among the agents in using
Common Property Resources;
• appreciate application of three different categories of normal form games .
viz., Prisoners' Dilemma Game, Chicken Game and Coordination Game
(already introduced in the Unit 20) in common property resource (CPR)
Management;
• derive best response functions for computation of Nash equilibrium in the
case when payoff functions are continuous; and
• learn derivation of N ash equilibrium in the case of Coumot Duopoly
(quantity competition), Bertrand Duopoly (Price competition), War of
Attrition Game and Hotelling's Linear City Model.
45
Non-Cooperative
Game Theory-I 21.1 INTRODUCTION
Game theory is a multi-person decision theory that analyses situations in with
payoffs to agents. It involves the analysis of conflict, cooperation, and (tacit)
communication and extends applications into several fields, such as
economics, politics, law, biology, and computer science. Quite a number of
examples under discrete payoff structure and continuous payoff structure are
presented here to analyse Nash equilibrium in pure strategy. These
applications help learn theoretical insights, to understand how to formalize
game theoretic problem from an informal description of multi-person decision
making.

21.2 GAME THEORETIC APPLICATIONS IN

COMMON PROPERTY RESOURCES
Common property resources are often explained in terms of the physical
attributes of the goods or resources. They share two characteristics: (a) the
difficulty of excluding individuals from benefiting from the resource and (b)
the subtract ability of the benefits consumed by one individual from those
available to others. Such characteristics give rise to problems of (i)
appropriation and (ii) provision. While the first relates to extracting more than
the socially optimal level, the second problem refers to the difficulty of
investing for the "adequate" provision of the CPR or in activities related to its
improvement or maintenance. Both these problems can be considered to
generate negative consequences for users.

Commonly used games for depicting CPR problems

Three games, viz., (a) the prisoner's. dilemma game (b) the chicken game and
(c) the assurance game are used for handling CPR problems. The prisoner's
dilemma game has been widely used to represent the appropriation problems
encountered in CPRs.
When limits for resource use cannot be established, all users of the CPR will
want to use as much as possible so as to maximize their own profit, resulting
in over-appropriation of the resource. In the case of natural resources such
over-appropriation can lead to overexploitation.
In the case of resource provision, the problem arises with .free riders
benefiting from the resource (as a public good) provided without having
contributed to its provision. Take forexample, investment for the maintenance
or improvement of the CPR. This problem has been modeled, depending on
the provision technology, with the prisoner's dilemma or with the assurance
game. The former having no-one contributing to the resource provision and .',
the latter having players contributing if, and only if, the others are to .
contribute.
Lastly, the chicken game has been used to illustrate assignment problems
when there are different resource locations with different richness.

21.2.1 Prisoners' Dilemma Resource Conservation Game

Consider a village pasture which is common property. Let two local villagers
graze their cattle there. Note that villagers can graze more cattle with the same
effort cost.

46
 Table 21.1: Prisoners' Dilemma Game Static Games with
Complete Information:
Player 2 Application

Participate (p) Defect (d)

Player 1 Participate (s, s) (v, r)
(p)
Defect (d) (r, v) (t, t)

But beyond a certain level grazing, the pasture can no more be sustainable. If
the villagers desire to maintain its sustainability, the number of cattle grazing
there must be restraint so that the extraction of the resources does not exceed
the regeneration capacity of the pasture. When such a participation occurs in
the conservation of the commons, it results in payoff (s, s)given in Table
21.1.

However, if any of the participants them defects from participation (given the
other player participates) then defector's payoff increases and the one who is
participating gets nothing. That is,(p, d) = (v, r) or,(d, p) = (r, v). So
there is always an incentive among the villagers to free-ride or overgraze
(defect).
Thus, mutual defection or free riding (d, d) becomes the dominant strategy
equilibrium. The strategy, defection, for each of the villager dominates the
strategy participation, irrespective of the other players' choice of action (or
participation is dominated by the strategy defection for each). This sort of
outcome is described by Mancur Olson (1965) as - individual rationality
produces collective irrationality.
Here the cooperation is not sustainable as each of the villagers faces a credible
threat that given he participates, his rivals may defect which will push him to
zero payoff. Such a mutual threat pushes them to play safe by choosing
defect (d, d).The outcome is unique pure strategy Nash equilibrium (d, d),
with payoff =(t, t). This strategy (d, d) is the mutually best response strategy
or Nash equilibrium although it is not social / Pareto optimal and both the
villagers over exploit the pasture.
Basically, lack of institutional control may lead the individuals to over exploit
the natural resource base. So long as there are multiple users competing for
the resource, and pay-off structure follows inequality, r > s > t > v,
extraction by one creates negative externalities on others causing Hardin's
tragedy of the commons.

21.2.2 Chicken Game (or, Hawk-Dove Game)

A chicken game refers to a set up that typically describes two players heading
toward each other. If the players continue on the same path, they bump into
each other; if one swerves out of the way and other doesn't, the swerver
"loses" and is labeled the chicken, while the second, implicitly braver player
wins'. Hawk-Dove game also is based on the description of the same scenario
in which players compete for resources by being either aggressive or passive.
Both Chicken and Hawk-Dove are anti-coordination games, in the sense that
these can be thought of as the opposite of a coordination game. The
underlying idea here is that players use a shared resource. In coordination
games, sharing the resource creates a benefit for all as the resource is non-
rivalrous, and the shared usage creates positive externalities. In anti-
coordination games the resource is rivalrous but non-excludable and sharing
comes at a cost (or negative externality).
47
--,-~-

Non-Cooperative 21.2.2.1 Problem of the CPR Provision - Two Person Chicken Game
Game Theory-!
Consider a situation when everybody agrees something has to be done, but the
problem is who will actually do it?
There may be many situations where collective problem is solved through
unilateral actions. The situation arises, when collective goods can be obtained
at sufficiently low costs in relation to its benefits. This may create an incentive
to the bigger stake holders to undertake the responsibility of procuring it.
Given the total gains associated with the production of collective goods or
avoidance of collective bad is pretty large in relation to individual's costs
share in avoiding it, unilateral contribution may be the Nash equilibrium.
Consider two small scale marine fishermen A and B, who stands threatened
by the invasion of a fleet of foreign trawlers in their traditional fishing zone.
Their income is presently 10 units. But to maintain the level of present catches
they need to have inshore water legally protected against the encroachments
of foreign boats. They can secure the legal protection from the government by
making a lobbying effort (which is a public good), which costs altogether, say
Rs.4.
Table 21.2: Fishermen Game

Fisherman 2
Cooperation (C) Defection (D)
Cooperation (C) (x, x ) = (8, 8) (a, b) = (6, 10)
Fisherman 1 (NE)
'Defection (D) (b, a) = (10, 6) (y, y) = (2, 2)
(NE)

If they cooperate by sharing the costs, their payoffs are: (C, C) = (x =

10 - 2 = 8,x = 10- 2 = 8). But if only one cooperates and the other defects,
then the payoffs are: (C, D) = (b = 10 - 4, a = 10) = (6,10) or, (D, C) =
(a = 10,b = 6).

The lobbying is so rewarding that it pays a single agent to bear the entire
costs. The agents without lobbying effort (D, D) will face a disastrous
situation in their payoff (y = 2, Y = 2). So there are two-pure strategy Nash
. equilibrium (C,D)or (D, C). Politically active agent (or sucker) will
unilaterally bear the entire costs of procuring the collective goods.

In this kind of game the players have no dominant strategy (unlike the PD
game). As we have seen in PD game, each player prefers others to rake the
responsibility of procuring all goods, while he refrains from moving. However
contrary to the PD games, in this kind of game, each player is willing to take
full responsibility for that action if the other refrains to do anything. The
consequence of nobody doing the action is so disastrous that either of the
players would undertake it if the other refuses to do it.
Other Applications of Chicken Game are:
• Maintenance and surveillance of irrigation system
• Maintenance and surveillance of pastures
• Maintenance and surveillance of grazing fields
• Maintenance and surveillance of hunting grounds
48
r

In some fishing communmes the problem of conflict of appropriation of Static Games with
public goods is solved by reorganizing the claim of the first entrants into the Complete Information:
Application
fishing grounds. There may be some conflicting situations when ill case of
catching of lobsters or bottom dwelling species, the number of fishermen
exceeds the number of locations, where enough lobsters can be caught to
justify the effort involved.

21.2.2.2 Chicken Game: Heterogeneous Group and Collective

Action
Olson (1965) predicts that richer party will necessarily contribute because he
has greater stake. To appreciate such a result, consider a case of two herders.
The rich one has a herd of prized goats, whose furs are used for making
upscale Kashmeri sweaters to be exported and sold in some expensive salons
of Paris. The other poor herder lias a herd of humble goat used for making pet
meat. The payoff matrix of the game is shown in Table 21.3. This game has
,
two pure strategy equilibrium (C, D) or (D, C) (leading to an asymmetric
I. payoffs), and one mixed strategy equilibrium, where all outcomes have
I positive probability of occurrence. (C, D)is the focal point equilibrium, where
it is likely that the rich herder will play C in equilibrium.
Table 21.3: Game with Asymmetric Payoffs

Poor Herder
Cooperation (C) Defection (D)
Poor Cooperation (C) (x, x) = (15, 3) (13,5) (NE)
Herder
Defection (D) (17,0) (NE) (2,0)

If we assume that the rich is more powerful in terms of economic and political
power, it is more likely that class has abetter ability to take prior moves than
the poorer or weaker party. If this is true, then the appropriate tool for analysis
of decision problem involving rich and poor is not the normal form or
strategic of game but the extensive or sequential form of chicken game (which
will be introduced in the Unit 22). The solution concept in that case is
subgame perfect Nash equilibrium.

21.2.3 Coordination Game in Fishing

. In a coordination game, players benefit from working together. So the
solutions to some of problems arise only on cooperation. In this class of
games, there is no incentive for either party to cheat since it will result in a
worse outcome than that from cooperating.
Assurance Game also provides insights to solution of. problems through
cooperation. It goes by a generic name "Stag Hunt" game due to its following
formulation: Two hunters can either jointly hunt a stag (an adult deer) or
individually hunt a rabbit. Hunting a stag is challenging and requires mutual
cooperation. If either of the hunters hunts a stag alone, the chance of success
is minimal. Thus, hunting a stag is most beneficial although requires a lot of
trust among the participants.
In tragedy of the commons, therefore, we need to look for solutions through
coordination game instead of Prisoners' Dilemma type of game. For, the
incentive to conserve depends on whether others are willing to conserve.
To see that suppose that there are two fishermen (A and B), independently
deciding whether to put one or two boat at the sea for catching fish. Given the'
resource constraint, each of them taking two trawler boats are not sustainable: 49
Non-Cooperative and will thus result in low payoff (2, 2) as given in Table 21.4. So a
Game Theory-I cooperative behaviour is reflected by the strategy of taking (one Boat),
whereas free riding is reflected when taken (two Boats).
There are two pure strategies N ash 'equilibrium: good equilibrium - (1 Boat, 1
Boat); bad equilibrium - (2 Boat, 2 Boat). If both A and B can convey
properly their intention regarding cooperation there is a chance of having
good equilibrium - (1 Boat, 1 Boat), where both will have fairly good payoff
of (8, 8). However there is a chance of having a bad equilibrium as well. Note
that in this game a genuine trust problem persists regarding other player's
intended action. A may fear that B will abstain from cooperation not because
B prefers free-riding but because B's expectation about A;sbehaviour is
pessimistic and vice versa for B vis-a-vis on A.
Table 21.4: Example of Coordination Game
Fisherman B
1 Boat 2 Boat
Fisherman A 1 Boat (8,8) (NE) (1,6)

2 Boat (6, 1) (2,2) (NE)

Check Your Progress 1

1) What is a 'Common Resource'?

2) Write a short note on 'Common Property Resources'.

3) Why the players choose strategies that are not Pareto optimal in common
property of grazing field.

4) What is focal point equilibrium?

......................................................................................
50
5) Why Chicken game is called anti-coordination game? Static Games with
Complete Information:
Application

21.3 BEST RESPONSE FUNCTION

If each player has a fewer actions, we can find the Nash equilibrium by
examining each action profile and see if it satisfies the condition of Nash
equilibrium. In more complicated games where the payoff functions are
continuous, we obtain the Nash equilibrium through the best response
junction. In a two player game, the action profile si (i = 1,2) is a Nash
equilibrium of a strategic game if and only if each player's action is a best
response to the other 'player's action. So, s; = b1 (sD ands~ = b2 (s;).
Thus each player has a unique best response to every action of the other
player. (s;, sDis the Nash equilibrium if player 1's action s{ is her best
response to player 2's action s~ and player 2's action s~ is the best response
to player l' s action s;.
21.3.1 Synergistic Relationship
Note that in a synergistic relationship if two individuals devote more effort,
both are 'better off. To see that, consider two individuals (i = 1,2) involved in
synergistic relationship of marriage. When they devote more effort they are
better off. So their set of actions is their effort levels given that of the other.
As each individual i has infmitely many actions,the best response functions of
each individual have to be ·constructed. For that we take their preferences as
reflected by their payoff functions.
Payoff function of the ith individual is:
tt, = ai(c + aj - aa, ai = Non negative number = Effort level of the
individual i, c > 0, constant.
Trj = aj(c + a, - aj), aj = Non negative number = Effort level of the
individuali, c > O.

First order condition (FOC) :::~

1
= (c + aj - ai) - ai = 0 (21.1) and

on'
_1
oaj
= (c - aj + aa - aj = 0 (21.2)

Solving for a, from the equation 21.1 we geta, = c:aj

= bi(aj)i. e.,best
response function of the ith individual. Note that it is a linear function with
positive slope (i) in the ajai plane. S~larly, the best response function ofthe
jth player (from 21.2) isaj = c:a i
= bj(ad. Again, it a linear function with
positive slope (2) in the ajai plane (see, Figure 21.1). It may be seen that the
best response function b, (ad is steeper than b, (aj), and both are positive.

51
 a2rr. a2rr·
Non-Cooperative
Game Theory-I
The second order condition (SaC): aa.;
!
= -2 ai < 0 andaa.;
1
= -2 aj <0
a rr· = 1 > O. So the effort levels a.anda, are strategic complements.
2
andaa.a~.
! 1

Solving the two best response functions (equation 21.1 and 21.2) we get:
c+a'
b.l ( a.J ) = a.l = --land
2

The solution or the Nash equilibrium of this example has obtained the best
response functions such that, at = aj = c .

a~
l

1
-c
2

1
-c a~
2 J

Fig.: 21.1: Synergistic Relationships

Check Your Progress 2

1) Defme Best Response function .
.......................................................................................
......................................................................................
.......................................................................................
......................................................................................
......................................................................................
2) What is strategic complement and strategic substitute?
......................................................................................
......................................................................................
......................................................................................
......................................................................................
......................................................................................
52
r
3) Discuss the meaning of a synergistic relationship. Static Games with
Complete Information:
Application

21.4 QUANTITY COMPETITION AND PRICE

COMPETITION
In this section we will be presenting the case of duopoly fIrms simultaneously
competing with each other through quantity or price. If the strategic
interaction is through quantity the firms are under Coumot competition. If it is
through price, then the firms are under Bertrand competition.

21.4.1 Cournot Competition

Firms compete with each other by simultaneously choosing the level of output
of a homogenous product believing that the output levels of their rival firms
remain unchanged. Thus, output is the strategic variable.

With entry and exit' of the firms are restricted, let us consider a duopoly
market structure with two symmetric firms, summarized by their demand
function and identical cost function.

Market demand function: p(Q) = a - bQ = a - b(q1 + q2)' .... (21.3)

where, Q = q1 + q2, a, b > O.

. I Q) {a - bQ, forQ < a/b d '. . I d

Precise y, p( = 0, for Q > alb, an negative output IS ru e out.

Cost function: Ci(qi) = ciqi, i = 1,2 (21.4)

Each firm's strategy space can be defmed as: Si = [0, oc), output is
continuously divisible and negative output is not feasible.

Thus the payoff function of the 2 firms: 1f1(qV q2), 1f2(qV q2)
There is interdependence in each firm's Payoff functions. Payoff functions are
simply the profit functions and are realized by a combination of strategies
chosen by competing firms. If qi increases then 1fi(') also increases till a
certain point. However, if q, increases, then the profit of the rival firm 1fl.)
falls. So if both firms increase their output, aggregate profit may as well fall.

Equilibrium in the Coumot duopoly is defined by N ash equilibrium. The

nature of interdependence is a sort of Prisoner's Dilemma game wherefq], q;)
is said to be Coumot Nash equilibrium if and only if

1f1 (q;, q;) 2:: 1f1 (qv q;) (21.5a)

1f2(q;,q;) 2::1f2(q;lIq2) (21.5b)

In order to derive the Nash equilibrium output (q;, q;) we have to derive the
best response functions (reaction function) of both the firms,

Consider the profit function ofthe firm 1: 1f1(qV q2) = p(Q). q1 - c1q1

53
Non-Cooperative Thus, 1fl(qVqZ) = ql(a - bq, - bqz) - clql'
Game Theory-I
Maximizing the profit function of the firm 1 with respect to.qv we get the

FOC: arrl(qvq2) =(a - bql - bqz) - bql - Cl =0 , or,(a - Zbq, - bqz) -

aql
Cl = 0
(a-Cl) q2
Therefore, ql = 2b - 2" Rl(qz) (21.6)

R 1( q z) ~ Reaction function of the firm 1or, the best responsefunction of the

_ 1. Rl(qz)is a linear function decreasing in (qz), slope = :ql~
firm -~
z 0
and the intercept = (a-Cl) > O.
Zb

Similarly, we have to derive the reaction function or the best response

function of the firm 2.
Consider the profit function of the firm 2: 1fz(qv qz) = p(Q). qz - czqz
Then, 1fz(qv qz) == ([z(a - bql - bqz) - czqz·
Maximizing the profit function of the firm 2 with respect toqz we get

FOe: arr2(ql,q2) (a - bql - bqz) - hqz - Cz = 0 , or,(a - bql - 2bqz) -

aq2
Cz =0
C
Therefore.
l' , qZ = (a-
Zb 2) - ql
Z = R Z(q 1) (21 .7)

Rz (ql) is the reaction function of the firm 2 or, the best response function of
the firm 2. RZ(ql)is a linear function, decreasing in (ql), with slope aq2 = ~
aql
C q2 q l
- ~ 0 and the intercept = (a- 2) > O. (Given that aa = -.!. or, aa = - 2).
Z Zb ql Z q2
Thus, both the reaction functions are downward sloping straight lines. If both
are plotted on the (qz, ql) plane, RZ(ql) is steeper than Rl (qz) ~
a 1 IR1(qZ) = -.!.and aq1 IRz(ql) = -2. If the rival's output increases, it is
q
a q2 Z a q2
optimal for the incumbent firm to reduce the output. The intuition is quite
simple, i.e., if the ith firm increjses its output (for a given level of output of
the rival), price will fall. In order to maintain the high price the rival firm will
have to drop its output. Taking the second order cross partial derivative of the
two firms' profit function we get a2;1(~Vq2) -i < O. Thus, qlandqz are
q2 ql .
strategic substitutes of each other. As the firm 2 expands its outputq-,
marginal profit ~rrl~:~q2) of the firm 1 falls. Same is true for firm 2. In such.
case the reaction functions are always negatively sloped. However, if the
second order cross partial derivative is positive, then ql andqz are strategic
complements of each other. In that case reaction functions are positively
sloped.
Given the reaction functions of firms 1 and 2, we have a unique intersection
which defines the Cournot Nash equilibrium. Solving the reaction functions
(equation 21.6 and 21.7) corresponds to the intersection of the functions.
Thus,
a-ZCl + C2
*
ql = 3b ' qz
*
= a +Cl3b- C2 .
.•.......... (21.8)

~ Nash Equilibrium ~ Equilibrium output of firm 1 and 2. See Figure

(21.2).
54
 Static Games with
Complete Information: '
Application

d-q
b

Fig. 21.2: Nash Equilibrium Under Cournot Competition

The equilibrium industry output:

Q = ( *+ *) = a'-,Zct + C2 + a +Cl -C2 = Za -Cl -C2

.. '" ... (21.9)
,ql qz 3b 3b 3b
;

Equilibrium price: p* = a - bQ = a - b Za -;~ -C2 = a+ c~ + C2 •••••••• (21.10)

Suppose the firm 1 is more efficient than firm 2 wtthc; <

Cz • Then the output of the low cost efficient firms 1 is greater than the output
of the high cost inefficient firm 2.
.
Plugging the equilibrium values from equation 21.8, 21.9, 21.10 in the profit
function of the firm 1, we get the two equilibrium profits of the firm 1 and 2
respective as shown below:
1C*(~ *) = (a+ Ct + C2) (a-zct + C2) _ C (a-zct + C2 )
1 ql' qz 3 3b 1 3b

............ (21.11a)

............ (21. 11b) ..

21.4.1.1 Comparative Statics

Suppose the firm 1 becomes more efficient. That means Cl falls. Other things
remaining unchanged, the intercept term(a;:d falls. Equilibrium output of
firm 1 increases (from q; to q;*) and that of firm 2 falls (from q;' to q;'*). See
Figure 21.2. From equation 21.8 you may see the fall in
Cl will lead to fall in equilibrium price p", The equilibrium profit of the firm

I will increase and that of the firm to will fall (you may check olri~qi,qi) <0
Cl
011'* ( * *)
and 2 %q2 > 0). See Figure 21.3.
OCl

55
r
Non-Cooperative
Game Theory-I

Fig. 21.3: Comparative Static (in Cl) Under Cournot Competition

21.4.1.2 Cournot Competition vs. Collusion

Let us consider two symmetric firms (i.e., Cl = C2) under Coumot
competition. Then from the best response functions of the duopolist,·
Rl(q2) and R2(ql) from equation 21.4 and 21.5, the equilibrium output ofthe
firm 1 and 2 are
c- _ (a-c) c- _ (a-c) (21 12)
ql - 3b""' ·q2- 3b"" .
Given the firm 2' s output is 0, the firm l' s best response is to produce
.
monopo 1y output, i.e., R 1 (0) -- qlM q2 =0 I -
-
(a-c)_
Zb- . ·1 1
ql'M. sum ar y, R 2 (0) --

q2M ql I =0 -
- (a-c) -
Zb - qM, ( 2 1.1 3)

[Note: You may check the equilibrium monopoly output under the linear
demand function p(q) = a - bq, and constant marginal cost C(q) = cq. The
profit function ofthe monopolist is rrM = pq - cq = (a - bq)q - cq] .
M
FOC of the profit maximization of the monopolist: d1l = (a - 2bq - c) =
dq
o~ qM = (a-c)
2b
is the equilibrium monopoly output. The monopoly profit
(a-c)2
W ill b e=--.
4b

Suppose the two symmetric duopoly firms agree to collude together into a
single entity to take output decision jointly and thereby share the monopoly
profit equally. Then the colluded firms behave just like a monopolist. If the
marginal costs are constant, any division of monopoly output qM (any
division of profit) between the 2-frrms will give industry profit equal to the
monopoly profit. All possible division of monopoly output between the two
duopolist is shown by the line segment qtt q~. Any equal division of
monopoly output and profit corresponds to point M. (See Fig 21.3).
We have seen under symmetric condition,Coumot equilibrium output (from
equation 21.12)isqf* = qi* = qC = (a-c).
3b
Each firm would have a profit:
rr*c - rr*c - rr*c _ (a-c)2
1 - 2 - - 9b

56
 Hence Static Games with
M _ (a....•
c) > c _ (a-c) Complete Information:
q - 2b q - 3b ==> Application
Monopoly output is greater than Cournot output at the firm level.

Also, the monopoly profit = (a-C)\CCournot profit ofthe individual firm)

4b
(a-c)2 1 (a-c)2 (a-c)2 1
of -- ==> - . -- > -- ==> - Monopoly profic-Coumot profit of the
9b 2 4b 9b 2
individual firm.
So, if both the firm under duopoly forms a collusion and produce and share
the combined monopoly profit, then they will get higher profit under collusion
as compared to their duopoly profit. That implies both the firms have an
incentive to collude. If each of them restricts its output to half of monopoly
level of output %- (instead
M
of Cournot Nash equilibrium output) the profit of
each firm will be higher. In fact, any point inside the lens of two iso-profit
curves reflects Parato superior point compared to Coumot equilibrium.
Now the question is, if both the firms agree to 'collude, will the agreement
sustainable?
Under collusion, although both the firm decided to produce
M M' M
(L, L). However,each firm (given the other firm is producing CL) has an
. 2 2 , . 2 .
incentive to deviate according to its best response function or the reaction
M
function and realize even higher profit by producing qf (i= l , 2) instead of L,
2

h . qid = R ~(qf)
were 2 ,1,. J. = 1, 2 . P'oint M (qM
2' .qM) .
2 ISnot th e equui
ilibnum
. . IS
as It .
not profit maximizing for any of the firm, This is because it is not in the
reaction function (or, the best response function) of any of the firm. The
collusive agreement is not sustainable as it is not a Nash equilibrium. See
Figure 21.4.

ql

2b

qM
2

c
qM q2
2 2b b

q~
Fig. 21.4: Cournot vs. Collusion

57
r
Non-Cooperative Duopolists will be better off if they collude. But given the other firm
Game Theory-I producing the collusive output, each has an incentive to deviate. The firm
which respects the agreement will be worse off (compared to its Coumot
outcome) while the firm which ditches the partner firm by deviating will be
better off. So there is a credible threat for cooperation for each of the firm
because the other may ditch. As a result both the firms are worse off when
each responds to its own private incentives. Hence a Coumotduopolist faces a
situation of Prisoner's Dilemma.

21.4.1.3 Numerical Example

We are considering a duopoly market structure with two symmetric firms,
with the following market demand function and identical cost function.
Inverse demand function:
p(Q) = 130 - Q = 130 - 1. (ql + q2)' where, Q = ql + q2,
Constant marginal costs of each firm: c = 10 ~ TC1 = 10qv TC2 = 10q2
Profit function of the firm 1 and 2:
rr1(qv q2) =p(Q) ql - cql = {130 - (ql + q2)}ql - 10ql;
rr2(qv q2) =p(Q) q2 - cq2 = {130 - (ql + q2)}q2 - 10q2'
FOC of profit maximization of the firm 1

8rrl(ql. q2) = 130 - (ql + q2) - ql - 10 = 0 or 2ql = 120 - q2

8ql '
q
Reaction function ofthe firm lis R1(q2) = 120 - 2= 60 - ~q2 (a)
2
Similarly, FOC of profit maximization of the firm 2 is

8rr2(qlo q2) = 130 - (ql + q2) - q2 - 10 = Oor 2q2 = 120 - ql'

8q2
Reaction function of the firm 2 isR2(ql) = 120-Ql = 60 - ~ql (b)
. 2 . 2

Solving equation (a) and (b), Coumot Nash equilibrium output of each firm is
qf* = q~* = qC = 40.
Therefore, the Coumot industry level output: QC* = qf* + q~* = 40 + 40 =
80
Equilibrium Coumot Price=p(Qc*) = 130 - QC* = 130 - 80 = 50
Coumot firm level profit of the furml =
1lf*(qf*, qi*) = p(Qc*)qf* - cqf*= 40(130 -80-10) = 40.40 = 1600.
Similarly, equilibrium profit ofthe firm 2=1l~*(qf*, qi*) = 40.40 = 1600
We get the industry level profit as 1600 + 1600 = 3200.
Now consider the two firms collude together by jointly producing the
monopoly level of output and share the profit equally.
Monopoly profit (or, collusive profit) of the industry=r'" = (130 - qM)qM -
10qM
Monopoly output (or, collusive output) of the -industry can be obtained by
maximizing rrMw.r.t. qM.TheFOC of the profIt maximization of the
drrM
monopolist=- = (130 - 2qM - 10) = 0 ~ 2qM = 120 ~ qM = 60.
dQ

58
 M M Static Games with
So the collusive output of the each fIrm(L ,L) = (30,30). Complete Information:
2 2
Application
Equilibrium price: pM = 130 - qM = 130 - 60 = 70.
Equilibrium profit under collusion-er" = (130 - qM)qM - 10qM =
M M
60(70 - 10) = 60.60 = 3600. Given the firms share the output (L, L), the
2 2
M M
individual firm share of the profit: (~,~) = (1800,1800). Thus the firms
2 2
. nM nM
are better off by colluding because they get higher profit share: (-,-) =
2 2
(1800,1800) > (nf*, ni*) = (1600,1600). That means, each firm has an
incentive to collude.
However, each firm has an incentive to. deviate given the partner firm is
M
committing to collusion by producing L.
2

I• Thus, the output of the firm 1 under defection (when firm 2 sticks to collusion
. qM qM 1

I by producmg 2:= 30 = Ri(q2 = 2) = Ri(30) = 60 - 'iq2 = 60-

15 = 45 (from equation (a).
I Similarly, the output' of the firm 2 under defection (when firm 1 sticks to
. :. qM qM. 1
collusion by producmg - = 30 = R2(qi = -)2 = R2(30) =
,
1 . 2
60 - 15 = 45 (from .equation (b).
60 - -q2 =
2

Profit of firm 1 under defection (when firm 2 sticks to collusion)=. tt 1(qi

45,q2 = 30) ~ (130:"- 45 - 30)45 -10(45) = 45(55 -10) = 45(45) =
2025.
The profit of the firm 2 when firm 1 defects=n2(qi = 45, q2 = 30)

= (130 - 45 - 30)30 - 10(30) = 30(55 - 10) = 45(30) = 1350.

Since, 2025 > 1800, the firm which defects (given the partner firm commits to
cooperation) has an incentive to do so, because it becomes substantially better
off from the collusive outcome. However the firm which sticks to the
collusive commitment is really worse off. That is, 1350 < 1800 < 1600.The
firm is not only worse off that of than the collusive outcome, it looses further
as compared to the Cournot outcome. So given the other cooperate there is an
incentive for defection for each firm. There is a credible threat also because if
any of the firms defects the partner who is cooperating will worse off. This
credible threat actually drives the firm to play Coumot strategy.

21.4.2 Bertrand Price Competition

Let there be two firms producing a homogenous commodity. Each firm
chooses a price simultaneously and produces enough output to meet the
demand (no capacity constraint), given the prices chosen by other firms
(Pi> 0).
Payoff functions (or the profit functions) are: ni (Pi' P2) , n2(Pi, P2) .
Firms are symmetric in cost. Cost function=Ci(qi) = c. qio i = 1, 2, c =
Constant marginal cost.
Demand function faced by the firms=
Di(Pi ,P2) = a - P, 'V P :5 aand
Di(Pi , P2) = 0, 'V P > a and c < a.

59
r

NOD-Cooperative
Game Theory-I
Rationing Rule: Di(Pl ,P2 ) =

Here is the rationing rule.

1) If Pi < c (for either i= l , 2), Pj > c i. e., Pi < Pj' Firm i will capture
the entire market. But Firm i earns negative profit ttt, < 0). In this
situation firm i can raise the profit by increasing Pi till it is equal to c, i.e.,
Pi = C.

2) If Pi = candp, > c i. e., Pi < Pj ,then firm i is charging lower price than
firm j. The consumers will be purchasing from firm i only. So the entire
market will be captured by firm i and firm j will be out competed
i. e. , qj = O.Then firm j can slash the price till Pj = c in order to capture
the market share.
3) If Pi = Pj' then the market will be equally shared by the two firms.
4) If Pi > c, Pj > c, and Pi > Pj thenthe entire market will be captured by
firm j and firm i's market share will bezero. But the firm j is earning
positive profit since Pj > c. In that case firm i can reduce the price below
Pj in order to capture the entire market. This price undercutting will
continue till Pi = Pj = c.
Profit or payoff function of [rrm i under the rationing rule is
(a - Pi)(Pi - c), andif Pi < Pj,Pi ~ a
a-pi .
Tri(Pl, P2 ) = (-2-)(Pi - C), andif Pi = Pj,Pi ~ a
{
. 0 if Pi > Pj

In contrast to the Cournot game (where the profit function is continuous with
respect to the strategic variable i.e., quantity), there is a discontinuity feature
in the payoff functions under Bertrand price competition. The profit of a firm
can sharply drops to zero for charging a very little higher price. A slight
undercutting would enable it to grab 100% market share.

Nash equilibrium under Bertrand price competition (under homogenous

product industry) is a set of price (Pi *, Pj *) such that

The Nash equilibrium under Bertrand price competition (under homogenous

product industry with symmetric firms) is the pair of price strategy (Pi *, Pj *)
where, both the firms will charge competitive price and therefore make zero
profit i.e., Pi * = P! * = (c, c).In this situation the reciprocal undercutting
reduces the price to marginal cost. This is called Bertrand Paradox.
1) Two firms are sufficient for competitive outcome.
2) Competition between 2 firms result in complete dissipation of profits.
Consider a situation when
1) If Pi * > Pj *> c. Firm i makes zero profit but it can do better by charging
a price Pj * - E, where, where Pj * - E > c. It then captures the entire
market and makes a positive profit {(Pj * - E - c). D(p/ - E)}. It
further implies firm j can retaliate by undercutting its price below
60 (p j * - E) and make positive profit.
 Static Games with
2) If Pi * = Pj * = P * > c. Both the firms are making a profit {~ *-
Complete Information:
E - c). D (p * - E) }. If any of the firm charges a little lower price Application
(p * - E) it can capture the entire market and makes a positive profit
«p * - E - c). D(p * - E), which is higher than before. But that
induces the rival firm to retaliate for further price cutting.
3) If Pi * > Pj * = c. Firm j is making zero profit even through it serves the
entire market. Firm i's market share is zero but can capture half the
market by undercutting its price to Pj * = c.
Hence under homogenous product industry with symmetric firms under price
competition th reciprocal undercutting reduces the price to marginal cost (c) at
the equilibrium.
Following diagram (Figure 21.5 a, b, c) is showing the firm i's profit for
different price strategies of the firm i and j.
(b) ©
(a) , f1
C-<p.J <pM p. >pM
J

pM,

Fig. 21.5 (a, b, c): Profit Function of the ith Firm

Now we have to derive the best response function for the firm i and j, i.e., we
have to derive those strategies which give the firm i (and firm j) maximum
payoffs.
1) If Pj < c, ~ Firm i's profit is negative if Pi :::;Pj and zero if Pi > Pj'
So any price Pi > Pj is the best response to Pj· ~ Bi(pj) = {Pi: Pi >
Pj}
Thus in Figure 21.5 (a), the dark segment of the firm i's profit function
shows the relevant profit, whenp, < c. When Pi = Pj < c, both the
form will earn negative profit, as the market' will be equally shared
between the two frrms. So ith firm profit is a discrete point shown in the
• . I
negative aXIS.
2) If Pj = C, then analysis is similar to the previous case except that
Pi ~ Pt yields Tri = 0 and hence B, (pj) = {Pi: Pi ~ Pj}
3) If < Pj :::;Pm,then Firm i can set price Pi :::;P! and earn positive profit.
Tri(Pl pj)is increasing in Pi till Pi :::;Pj and Tri(Pi Pj) = 0 at Pi > Pj·
I I

So the best response function is Bi(pj ) = {Pi: c < Pi s Pj}

61
Non-Cooperative See Figure 21.5(b), the dark segment of the firm i's profit function shows
Game Theory-I that for a given Pi : if the firm i chargesp, < c, it will capture the entire
market but earn negative profit. For any Pi : c < Pi < Pj , firm i will earn
positive profit and enjoy entire market. When Pi = Pj ,the market will
be shared between the two firm, so the firm i's profIt turned out to be a
discrete point, For any price Pi > Pj , firm i will loose the entire market
share and earn zero profit,
4) If Pj > Pm, then charging the monopoly price (Pm) is the unique best
response of firm ii. e., Bi(pj ) = Pm'
See Figure 21.5( c), the dark segment of the firm i 's profit function shows that
for a given Pj > Pm, for any price less than Pj .firm i will enjoy the entire
market share and is increasing in Pi' For Pi = Pj' firm i's profIt drop to a
discrete point and becomes zero for any price Pi > Pj'
Thus firm i's best response function is
{Pi: Pi > Pj}, andif Pj < C

Bi(pj) = {Pi: Pi ;:::Pj }, andif Pj = C

{
{Pi: C < Pi ~ Pj}, if C < Pj ~ Pm
{Pm}, if Pm < v,
Similarly, the firm j' s best response function is
{p{ Pj > Pi}, andifp, <C
Bj(Pi)= .{Pj:Pj ;:::pj},an~ifPi =C
{
{Pj. C < Pj ~ pd, Lf C < Pi ~ Pm
{Pm}, if Pm < Pi
Thus the best response function of the Bertrand game is qualitatively different
from Cournot game. Because of the discontinuity feature, for some actions of
the rival j, firm ihas no best response and for some action it has multiple best
response (see Figure 21.5). The point of intersection of the two best response
functions shows that Pi * = Pj * = (c, c) as the Nash equilibrium.

pM

c ----- c
" "II " "II
""
" 45° I
I
""
" 45°
I
I
0
" C
" I
pM Pj
0 C pM
Pj

Fig. 21.6: Best Response Functions of the Bertrand Competitors

62
r
Static Games with
Check Your Progress 3
Complete Information:
Application
1) What do you mean by tacit collusion?
.....................................................................................
......................................................................................
......................................................................................
.
.....................................................................................
.

......................................................................................
2) What does Bertrand paradox refer to?
.....................................................................................
......................................................................................
.....................................................................................
.....................................................................................
.....................................................................................
3) In a Coumot duopoly when does a Nash equilibrium occur?
.....................................................................................
.....................................................................................
., .
.....................................................................................
.....................................................................................
4) Consider the case of Coumot Duopoly (n = 2) when the market demand
is p = 10 - Q and the firms are symmetric in costs with C(qa =
1. qi, where i = 1,2.' Show that the Cournot equilibrium quantity
(qi' qa = (3,3), equilibrium price p = 4 and profit at the firm
level = 9.
.....................................................................................
......................................................................................
.....................................................................................
.....................................................................................
.....................................................................................
.....................................................................................
5) Consider the case of Coumot Duopoly (n = 2) when the market demand
is p = 10 - Q and the firms are symmetric in costs with C(qi) =
1. qi' where i = 1,2. Show that if both firms colludes and shares the
profit equally, then the equilibrium output of each firm (under
collusion)= (2.25, 2.25), Price = 5.5, Each firm's share of profit = 10.125.
.....................................................................................
.....................................................................................
. ..~ .
.....................................................................................
.....................................................................................
....................................................................................... 63
r

Non-Cooperative 6) Show that by cheating on the Cartel (Collusion) a firm increases its own
Game Theory-I profits but necessarily lowers its riva1' s profit by an even larger amount.

..
..................................................................................... .

21.5 WAR OF ATTRITION

Wars of attrition often arise in business besides in other settings. The game
arises when two (or more) firms compete with each 'other, each one losing
money. but hoping that the competitor will eventually give up and exit the
industry. When playing the game, each firm must decide whether to cut its
losses and exit, or alternatively, tough it out in the hope that the competitor
will soon exit.

A Simple Example
For our understanding the War of Attrition game, let us consider the following
simple example:

Suppose two companies, A and B, must decide each month whether to spend
Rs.1000 lakh. If in the first month one company spends the Rs.1000 lakh and
the other does not, the game is over - the first company becomes a monopolist
worth Rs.10000 lakh, and the second company must exit from the race. If
neither invests Rs.lOOO lakh in the first month, the game is also over, with
neither company losing nor making money. However, if both companies
spend Rs.l 000 lakh in the first month, neither one wins anything.

Now move to the second month, where again each company has to decide
whether to spend Rs.1000 lakh. If both companies again spend Rs.1000 lakh,
the game moves to the third month, and so on. If, at the start of some month,
one of the companies spends Rs.lOOO lakh and the other does not, the first
company wins the Rs.l0000 lakh prize. But of course many months (and
much money) could go by before this happens.

Suppose you are Company A, and one of your classmates is Company B.

What should you do in this situation? Think carefully about the following
questions:
1) Is it rational to spend Rs.IOOO lakh in the first round of this game? Why
or why not?
How would you decide whether or not to spend the Rs.1000 lakh?
2) Would it be rational to start playing the game with a plan to exit if, after
three or four rounds, your opponent has not yet exited? Why or why not? '
3) Is there anything that y,ou could say to your opponent, or that your
opponent could say to you, that would affect your decision to start
playing the game? Is there anything you or your opponent could say that
would affect your decision to continue playing if, after two or three
rounds, neither of you has dropped out?
64
 r

Again, think carefully about these questions before proceeding further. Static Games with
Complete Information:
Initial conceptualization of the gamehas come from strategies related to the Application
conflict between two animals fighting over a prey. Each one chooses the time
at which it intends to give up. Fighting involves efforts. So, it is costly.
Naturally, each animal prefers the fight as short as possible. If both animals
intend to give up at the same time, then each has an equal chance of obtaining
the prey. Let the fighting (or waiting time) is continuous which starts at 0 and
continues infinitely. Also, let there be two players iandj. Each player's set of
action is the set of possible concession times. Let Vi be the ith individual's
value ofthe object of dispute (Vi> 0).
1) If the player i concedes first at the timer., his payoff is (- td, as he
spent ti units of time and does not obtain the object.
2) On the other hand, if the other player concedes first at the timer., player i
gets the object. So ith player's payoff: (Vi - tj).

3) If both the players concede at the same time ti = tjthenplayer i's payoff
1
= -2 (Vi - td·

Thus the payoff's of the ith player:

andif - t:t < t:J
1
2" (Vi - td, and if ti = tj

(Vi - tj), if ti > tj

For N ash equilibrium we need to derive the best response functions.

Intution: If tjis small, i.e., player j' s intended concession time is early enough
then it is optimal for player i to wait for the player j to concede. In that case
player Cs optimal concession time will be later than tj.
If tjis large, i.e., player j's intended concession time is long then it is optimal
for player i to concede immediately (if tj > vd.
Following diagram (21.6 a, b, c) is shows the player Cs payoff in three cases:
1. tj < Vi; 2. tj = vi;3. tj > Vi.
(a) ©
(b)

•
Ok-------_e------
It. = Vj t,
I J
I
I
I

~. Fig. 21.7 (a, b, c): Player i's Payoff Function

l 65
 Non-Cooperative Best response function of the player i (for Vi > vi»
Game Theory-I

{ti: ti > tj}, andif tj < Vi.

Bi(tj) = {
{ti: ti = ° or ti > t!}, and if tj = Vi
0, if tj > Vi
Best response function ofthe player j (for Vi > Vj):
{t/ tj > tJ, and if ti < Vj
Bj(ti) = { { t/ tj = ° or t, > t!}, and if ti = Vj
0, If t, > Vj

(b)
(a) t·J

tj B/tj)

Bj(tj)
Vj

vj
--------- ~
'" . I
'" " I
'" I
.'
I
M •
I
lIo..
,

Fig. 21.8 (a, b) : Player i andj's Best Response Functions (Bi(tj), B/ti).

Superimposing the player i and j's bestresponse functions, we see there are
two areas of intersection: 1. Vertical axis ~ Vi 2. Horizontal axis at and to the
right of Vj.

Nash equilibrium is (t/, t/): 1. t, = 0, tj ~ Vi,

2. tj = 0, ti ~ Vj

That is, player i concedes immediately and player j concedes at time Vi or

later and Player j concedes immediately and player i concedes at time Vj or
later.

Check Your Progress 4

1) Discuss the similarities between chicken game and war of attrition game .
.......................................................................................
.......................................................................................
.......................................................................................
.......................................................................................
••• ••••• 4 ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

.. , .
. 66 .......................................................................................
r
I

2) What is the easiest way of finding the Nah equilibrium in the war of Static Games with
attrition game? Complete Information:
Application

3) Write the properties of War of Attrition.

21~6 HOTELLING'S LOCATION GAME

Hotelling's linear city model has introduced the notions of locational
equilibrium in a duopoly in which two firms have to choose their location
taking into consideration consumers' distribution and transportation costs.

Initially the rn'odel was developed as a game in which firms first chose a
location and after a selling price for their products. In order to set their
business in the best location to maximize profits, the firms will have to
evaluate three key variables: competitors' location, customers' distribution
and transportation costs. We can consider static model where firms choose
their location and prices simultaneously.

The basic idea is that two ice cream vendors are on a beach that stretches the
0-1 interval. Customers are uniformly distributed along that interval. The
vendors simultaneously select a position. Customers go to the closest vendor
and split themselves evenly if the vendors choose an identical position. Each
vendor wants to maximize its number of customers.

The Nash equilibrium is for both vendors to select the median location (0.5);
doing this guarantees each vendor half the business, but deviating to any other
point generates strictly less.

We reframe the question as two candidates placing themselves along an

ideological spectrum, with citizens voting for whichever one is closest. This is
seen in the following:

Consider the case of electoral competition where there are two political parties
1 and 2, viz., (n = 2) contesting for votes. Suppose there exists a linear city [0,
1], where voters are uniformly distributed over the interval [0,1] with a unit
density i.e., each points represents one voter. It is assumed that the political
parties have no political conviction but wants to maximize their vote share by'
strategically locating themselves across the linear city. On the other hand,
voters have the political conviction and therefore the voters try to locate
themselves nearer to their political ideology. What would be the Nash
equilibrium voting configuration?

67
r
Non-Cooperative
Case 1: X1 < X2 *- -1
Game Theory-I 2

0
• •
Xl X2
•-1
2
1
Fig. 21.9 (a)

Party 1 and 2 are located at a distance x1andx2, (X1 < X2 *- ;-) distance
from the origin (0 at the extreme left). All the voters to the left of X1 will will
all vote to Party 1. However the voters to the right of Xv (X1 < X2 ) will be
splitted between Party 1 and 2. Thus, the vote share of Party 1: X1 +
~ (X2 -
2
X1) = ~(
2
X1 + x2). In order to understand whether this
configuration is a Nash equilibrium, we have to justify whether it is a
mutually best response strategy.
Suppose the Party 1 deviates from X1 distance to the right of Xv to say, X1"
where X1' = X1 + E, where E > O. .

Now the vote share of the Party 1 is: X1'~ (X2 - X1 ')=~(X1' + X2)=~( X1 +
. 2 2 2
E+X2) = ~( X1 + X2) + ~ E. SO the Party 1 gains by relocating itself to the
2 2
right. So the Party 1 has an incentive to deviate. Hence this is not the Nash
equilibrium strategy.

Case 2: X1 = X2 *- ~2

o • 1
2
• 1
Figure 21.9 (b)

In this configuration both the parties are located together but not in the centre.
Suppose P is the point where both are located. Each party will have equal vote
share, i.e.,;- x1 +;- (1 - X1) = ;-.

Now if any party gains from deviation we can claim that this is not a Nash
equilibrium. Suppose the Party 1 deviates from X1 distance to the right of
Xv to say X1', where X1' = X1 + E, where E > O. So the Party l' s vote
share =(1-
.
X1') + ~(
2
X1' - X1)'

Let the initial distance of both the parties was less than the centre of the linear
city i. e., (1 - X1') > ~~
2 22.
X1' - X1) >~. So, if any of the
(1 - X1') + ~(
parties deviates (to the right in this example) there are gains from deviation.
Such a configuration is also not the N ash equilibrium.
1
Case 3: X1 = X2 = -
2

•
o 1
Fig. 21.9 (c)

Here, each of the Party is getting exactly half vote share.

We will see that as any of the parties deviates (to its left or right), it looses
some vote share. So deviation by either party won't be gainful.

68
 Suppose the Party 1 deviates to its left at Xl', where Xl' = Xl - C. Static Games with
Complete Information: .
SO the Party 1's vote share after deviation: Xl' + ~(Xl - Xl) [( Xl - Application
1 1 1 1
c) + z( Xl - Xl + c)]=[(Xl - c)+zc] = (Xl - zC) < z" See that the
deviation by either party is not gainful. Hence Xl = X2 = .!. is the Nash
2
equilibrium. So both the parties would prefer to locate themselves at the
centre in order to maximize their vote share.
Check Your Progress 5
1) Suppose consumers are located uniformly along the line segment from 0
to 1. If firm A and B are located somewhere within the line segment,
discuss the Nash equilibrium you will get.

2) If two candidates choose their positions before the election in the

Hotelling's linear city formulation, why would they tend to run to the
centre as the election progresses?

3) In linear city model of Hotelling two fIrms share - transport costs with
their customers.

21.7 LET US SUM UP

In this unit we extend some of the basic ideas in game theory, particularly,
static games with complete information. We begin by discussing problems
encountered in common property resources and brought in prisoners' dilemma
game, assurance game and chicken game to find solutions. Then we will turn
to models of duopolistic competition. We first consider the choice of strategic
variable when duopolists compete in the sale of homogenous products - in
t particular, implications of choosing quantities instead of prices. The War of
Attrition game is seen to draw inferences on what it means to play this game
I· rationally.
69
Non-Cooperative Towards the concluding part of the unit we have included Hotelling's linear
Game Theory-I city model to examine the locational equilibrium in a duopoly in which two
firms have i(l choose their location taking into consideration consumers'
distribution 'and transportation costs. A basic conclusion of the model is that
of firms choosing to be in the extremes i.e.; they choose maximum
differentiation.

21.8 KEY WORDS

Assurance Game: A situation where there is a conflict between safety and
cooperation.
Capacity Constraint: A firm is unable to increase its output production due
to the increase in demand at least in the short run. The supply of the firm
remains insensitive to price.

Cartel: An association of manufacturers or suppliers with the purpose of

maintaining prices at a high level and restricting competition.

Chicken Game (or, Hawk-dove game, or snowdrift game): A situation where

each person prefers not to yieldto the other, but the outcome where neither
person yields is the worst for both.
Common Property Resource: Natural resources owned and managed
collectively by a community or society rather than by individuals.

Coordination Game: A situation where two persons ought to choose the

same strategy.

Homogenous Product: Product that cannot be distinguished from competing

products from different suppliers.

Monopoly: Market structure characterized by a single seller, selling a unique

product in the market.
Strategic Complement: Decisions of two or more players are called strategic
complements if they mutually reinforce one another.

Strategic Substitute: Decisions of two or more players mutually offsetting

one another.

21.9 SOME USEFUL BOOKS

1) Avinash K. Dixit(20l5), Susan Skeathand David H. Reiley, Games of
Strategy, W. W. Norton and Company; 4th edition.
2) Drew Fudenberg and Jean Tirole(1991), Game Theory, MIT Press..
3) Drew Fudenberg and David Levine (1998), The Theory Of Learning in
Games, MIT Press,
4) Robert Gibbons (1992), A Primer in Game Theory, Harvester
Wheatsheaf.
5) Martin J. Osbome and Ariel Rubinstein (1994), A Course in Game
Theory, MIT Press.
6) Michael Maschler, Eilon Solan, and Shmuel Zamir (1992), Game Theory,
Cambridge University Press.

70
 Static Games with .
21.10 CHECK YOUR PROGRESS Complete Information:
ANSWERSIHINTS Application

Check Your Progress 1

1) A common resource is a resource, such as water or pasture, that provides
users with tangible benefits. A major concern with common resources is
overuse, especially when there are poor social-management systems in
place to protect the core resource. Common resources that are not owned
by anyone are called open-access resources.
2) Most resources are congestible (rivalrous). This means that if some of the
resource is used for one purpose the same chunk cannot be used for
another purpose: one use precludes another use.Consider congestible
resources that have a zero price, i.e., everyone can take as much of the
resource as is available without having to pay for it. Such resources are
called common-property resources.
3) Each of the villagers faces a credible threat that if he participates, his
rivals may defects and push him to zero payoff.
4) A solution that people tend to use because it seems natural, special, or
relevant to them in the absence of communication.
5) Coordination games are linked to the economic concept of positive
externalities. That is, benefit reaped from being in the adopting the same
strategy as that of other agents. However, under negative externalities
choosing the same action creates a cost rather than a benefit. This class of
game, Chicken game falls in that, is anti-coordination game.

Check Your Progress 2

1) Strategy that produces the most favourable outcome for a player, taking
other players' strategies as given.
2) Decisions of two or more players are called strategic complements if
they mutually reinforce one another, and they are called strategic
substitutes if they mutually offset one another.
3) If two individuals devote more effort, both are better off.

Check Your Progress 3

1) No formal agreements on market sharing.
2) The possibility that all that is necessary for prices to be the same as in
competitive markets is for there to be two firms competing in a market.
3) When the firms' reaction curves intersect.
4) Do yourself as per the clue includedin the question.
5) Do yourself as per the clue included in the question.
6) Do yourself as per the clue included in the question.

Check Your Progress 4

1) The War of Attrition is a game something like Chicken stretched out over
time, where both players start with Continue, and the game ends when the
first one picks Swerve. Until the game ends, both earn a negative amount
per period, and when one exits, he earns zero and the other player earns a
reward for outlasting him.
71
Non-Cooperative 2) By getting the best response functions and plotting them.
Game Theory-I
3) Wars once started can last for a long time with no definite end; even
winner may lose money; random termination; reduce cost of fighting
induces lengthening of the war.

Check Your Progress 5

1) A variety of Nash equilibria are possible depending on the assumption
about the cost of travel for consumers.
2) Due to the fact that the candidates locate in the centre in the Nash
equilibrium of the game.
3) Linear

21.11 EXERCISES
1) Write a short note on the CPR management dilemma.
2) Derive the Nash equilibrium under the Cournot competition in the case of
3 symmetric firms with constant marginal costs (c). The market demand
function is p(Q) = a - bQ = a - b(ql + qz + Q3)' Derive the
equilibrium price and profit at the firm level as well as industry level.
3) Consider a Bertrand duopoly market structure. Defme Bertrand Paradox.
4) . Derive the best response function of the War of Attrition game. Show
graphically and explain the Nash equilibrium.
5) Discuss the Nash equilibrium under Hotelling's location game in the
context of electoral competition with two political parties.
6) Consider an example of the Tragedy of Commons. Show how the
phenomena of over fishing can be represented as a Prisoners' Dilemma.
You may set up the game with two players, each of which can undertake
low or high fishing activity.
Solution:
The case of over fishing should be set up in a manner similar to this:

India
Low
High fishing fishing
activity activity
V.S High fishing activity 1,1 3,0
Low fishing activity 0,3 2,2

The sustainable fishing catch is higher when both nations undertake low
fishing activity.
However, there is then an incentive for both to increase fishing. In fact,
high fishing is a
dominant strategy for both players. We therefore, end up with the worst
outcome.
7) There is competition that conveys the following information. Prize worth
Rs.l, cost c to fight per period. Each player chooses probability p .of
exiting per period. If both exit, there is a split of the/ prize. Best response
requires %p = p- c + (1 - p) %p. Value of exit = value of staying in.
72
r
Solve the game to get how long the the war would last, expected profits Static Games with
Complete Information:
for competitors.
Application
Solution:
Get = 1 - "';1 - 2c. Thus, C = (1/2)c. That means for c = 0.1, war will
last five years on average.
Expected profits areY2p. As profits are increasingin c, the effect on
competitor outweighs own effect.

M ar k et deeman d iIS given

. b y p (Q) = {1400 - Q if Q I
<
.
140
ot rwrse
8) There are two firms, each with unit costs = Rs.20. Firms can choose any
quantity.
i) Define the reaction functions of the firms;
ii) Find the Coumot equilibrium;
iii) Compare the Cournot equilibrium to the perfectly competitive
outcome and to the monopoly outcome.
iv) One possible strategy for each firm is to produce half of the
monopolistquantity. Would the resulting outcome be better for both
firms (Pareto dominant)? Explain why this is not the equilibrium
outcome of the Coumot game.
Solution:
i) In a Coumot duopoly the reaction function of Firm A identifies its
optimal response to any quantity produced by Firm B. In the presence of
.private firms, the optimal quantity is the one that maximizes JIA, Firm
A's profit, where
JIA = P(Q)qA - cqA
= (140 - qA - qB)qA - 20qA
The first order condition for profit maximization is:

anA =140-2q -q -20=0

aqA A B

_ 120-qB
qA ( qB) - 2

Since the game is symmetric (firms have identical cost functions), the
reaction function of firm B is:

A ( ) _ 120-qA
qB qA - 2

ii) Coumot equilibrium is identified by the quantities that are mutually best
responses for both firms; so, they are obtained by the solution of the
following two-equation system:

The equilibrium quantities are

73
r
Non-Cooperative
Game Theory-I
Q* = 80and the equilibrium price is
P(Q*) = 140 - 80 = 60
and firms' profits are:

n; = n: = 60q: - 20q: = 1600

iii) The competitive equilibrium outcome is characterized by P(Q) = c = 20.

So total quantity should be:
P(Q) = 140 - Q = 20, or, Q = 120
In such a case firms' profits are zero.
The quantity produced by a monopoly is obtained by the usual first order
condition:
The price under monopoly is P(qM) = 140 - 60 = 80 and profits are
JIM = 80 x 60 - 20 x 60 = 3600
So, profits under monopoly are higher than the sum of firms' profits
under Coumot competition; i.e., JIM > flA > JIB.
iv) If the each firm agreed to produce half of the monopolist quantity
qM = 40, their profits would be llM = 1800, larger than the Coumot
2 2
profits. So, a Pareto improvement with respect to Coumot equilibrium
would be possible. However, this cannot be an equilibrium since firms'
strategies are not mutually consistent best response; that is

ch C;) q; for all i

=1= = A,B

9) See the following problem:

Market demand is given by
lOO-Q if Q < 100
P(Q) = { .
o otherwise

Suppose that two firms both have average variable cost c = $50. Assuming
that fIrms complete in prices, then:
1) Defme the reaction functions of the firms;
2) Find the Bertrand equilibrium;
3) Would your answer change if there were three firms? Why?
Solution
1) The construction of the reaction function in a competition a la Bertrand
proceeds in the following manner.
Since firms compete in prices, we· need to use the direct demand
function where Q is a function ofP. From the inverse demand:
100- P if P < 100
Q(P) = { .
o otherwise

There are two firms, firm 1 and 2. Consider now the effect of the price
choice of any firm I on its own profits for any given price chosen by
firm j with i, j = 1, 2 and i '" j
74
I-
I a) Pi > P, : Il, =0 Static Games With
Complete Information:
In this case, firm 1 sells no output and therefore gets zero profits. Application

This is where firm i shares the total profits in the industry. The total
profits of the industry are here determined in the following way.
Price minus average cost gives the profit per unit of output (Pi - 50)
and this is multiplied by the total output (lOO-Pj).
c) Pi < Pj : lli = [(Pi - 50) (100 - Pj)]
Here firm i gets the whole profits of the industry.
In order to firm the reaction function of firm I we can distinguish 3
different cases. .
a) Pi (Pj) = 50 if P, ~ 50
In such a case, any price Pi < Pj will make negative profits, so [If m i
will prefer to lose the race rather than beating j on price. More
precisely we can .say that [If m i will never set a price below c = 50.
This is usually portrayed as the strategy that if Pj < 50, Pi = 50.
b) Pi (Pj) = Pj - 1::' if 50 < Pj ~ 75 = pM
Where pM is the monopolistic price. If Pj > 50, there is potential for
'positive 'profIts, as long as P, < 100, at which point quantity
demanded falls to zero, so that profits would be zero too. The first
issue is, should firm i match the price of [If m j, or attempt to
undereut its rival? Intuition suggests that the [If m should undercut
its rival rather than match Pj, This can be shown formally by
comparing the profits for [If m i if it matches Pj, lit, with the
profits, n,", it gets if undercuts Pj by a small amount I:: so that Pi =
Pj - 1::.

IIt = ~ [ (p j -: 50) (100 - Pj) ]

ll~ = [ (p j - E- 500) (100 - Pj + I::) ]

II~ = 2llt - E(150 - 2Pj - 2E)

This means:
A
As I:: -) 0, that is as the Pi, gets closer and closer to Pij, llt -) 2lli : if
firm i sets its price just below Pij' its profit will be roughly double the
profit from matching Pj. However, the fact that [If m i will do better by
just undercutting firm j than it does by matching firm j does not mean that
just undercutting is always the best strategy. When the [If m has the whole
market, its optimal price is the monopoly price. Therefore, when Pij is
greater than the monopoly price, firm j should set the monopoly price. In
this example, the monopoly price pM = 75. Following this discussion:
c)' Pi (Pj) = 75 ifPj > pM
The reaction function of any firm is then the following:

75
Non-Cooperative 50 if PJ-< 50
Game Theory-I
P (P)J
1
= P,J - E if 50 < P,J-< 75

\ 75 if Pj > 75

2) The Nash equilibrium is where the two reaction functions interest, which
is where PI = P2 = 50. In such a case the best response of the two firms
are mutually consistent. This is actually the only point in which this is so.
(Please check this result by your own Can you find another Nash
equilibrium?). This is the same as the competitive solution. Price is set
equal to average (= marginal) cost and zero profits are made.

3) No because an increase in competition does not change the price set by

the firms that cannot be lower than marginal cost. This is the essence of
the so-called Bertrand paradox: two firms are enough to achieve the
competitive outcome.

76
r

UNIT 22 DYNAMIC GAMES WITH

COMPLETE INFORMATION

Structure
22.0 Objectives
22.1 Introduction
22.2 Extensive- form Representation of Dynamic Games
22.2.1 Strategies in Extensive-form
22.3 Dynamic Games of Complete and Perfect Information
22.3.1 Backward Induction
22.3.2 Strategies in Dynamic Games with Complete Information
22.3.3 Subgaines
22.3.4 Subgame-Perfect Nash Equilibrium
22.3.5 Backward Induction and Subgame Perfection
22.3.6 Application 1: Stackelberg Competition
22.3.7 Application 2: Sequential Bargaining
22,4 Dynamic Games of Imperfect Information
22.4.1 Imperfect Information and Backward Indu~tion
22.4.2 Subgames with Imperfect Information
22.4.3 Strategies with Imperfect Information
22.4.4 Finding SPNE with Imperfect Information

22.5 Repeated Games

22.5.1 Two-Stage Repeated Games
22.5.2 Finitely Repeated Games
22.5.3 Infinitely Repeated Games
22.5.4 Application 3: Collusion between Coumot Duopolists
22.6 Let Us Sum Up
22.7 Key Words
22.8 Some Useful Books
22.9 Answer or Hints to Check Your Progress/Hints
22.10 Exercises

22.0 OBJECTIVES
After going through this unit, you will be able to:
• understand the idea behind dynamic games;
• present the dynamic games in extensive form keeping in view complete,
perfect and imperfect information;
• use backward induction as a solution concept to arrive at subgame perfect
Nash equilibrium; and
• analyse repeated games.
77
Non-Cooperative
Game Theory-I 22.1 INTRODUCTION
We have seen in Unit 20 the presentation ofa normal form game. In the
discussion there players choose their strategies simultaneously, without
knowing the choices of the other players. So these are static games. In this
unit we take up dynamic games by allowing the players have the choice
moving sequentially. That is, players make moves at different times or in turn.
So, one player chooses her action before the others choose theirs. That means
players who move later in the game have additional information about the
actions of other players.

22.2 EXTENSIVE-FORM REPRESENTATION OF

DYNAMIC GAMES
In the extensive form representation of the game we get the following features:
• Decision points: The dots, or nodes, represent a decision that must be
made by some player. The name of the player is usually placed above or
next to the dot.
• Actions: The lines, or branches, represent different actions that a player
can take at each decision point. The label is the name of the action, like
"enter" or "exit", etc. '
• Payoffs: The numbers at the bottom of the tree represent the payoffs to the
players. The convention is that the payoffs are given in order of the timing
at which that player moved first. So if Player 1 moves first, Player 2
moves second, Player 3 moves third, and so on, the payoffs are given in
the order: Player 1, Player 2, Player 3...

Such a method depicts the tree form of the game. Thus, game trees are made
up of nodes and branches, which are used to represent the sequence of moves
and the available actions, respectively. To appreciate the idea, consider two
players, Black Smith and White Elena, who are playing a sequential game.
Black moves first and has the option of Up or Down. White then observes his
action. Regardless of what Black chooses, she then has the option of High or
Low. The tree form of the game is follows:

Fig. 22.1: Tree Form Presentation of Sequential Games

78
In the game tree (Figure 22.1), decisions of the players are depicted by square Dynamic Games with
nodes. Note that node a is the decision node where player Black chooses Complete
Information
between Up and Down. Since node a is the first node, it is also known as the
initial node. Nodes ob and c are the decision nodes at which Player White
chooses between High and Low. The triangle-shaped ending nodes on the
right are the terminal nodes, which also have the payoffs for each player
associated with each outcome listed beside them.

22.2.1 Strategies in Extensive-Form

In sequential games, it is important to clearly define what is meant by
strategy. Game theorists define a strategy as a complete contingent plan of 'I
I

actions. In other words, a strategy specifies what action a player will take at
each decision node. Consider the game given above between players Black
and White.
Player Black has two strategies available - Up and Down. Player White,
however, actually has four strategies available since there are two nodes to
consider - b and c- and two possible actions at each node - High and
Low. Thus, strategies available to player White summarised in Table 22.1.
Table 22.1: Strategies of the Player

Strategies of Player Black

Up Down
.Strategy 1 High High
;
Strategy 2 High Low
. Strategy 3 Low High
Strategies of
Player White Strategy 4 Low Low

Because actions always lead to reactions, an important aspect of strategy in

sequential games is that players must consider - and plan for - their
opponent's reactions. In the example above, if player Black wants to
maximise his payoff, he must consider how his rival player White will react if
he moves Up and how she will react if he moves Down.

Check Your Progress 1

1) Differentiate between static and dynamic game.

2) What is extensive form representation of a game?

79
Non-Cooperative 3) What is meant by strategy in an extensive form representation of a game?
Game Theory-I

........................................ ~ : .

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 0.0 ••••••••••••••

22.3 DYNAMIC GAMES OF COMPLETE AND

PERFECTINFOErndATION
We have seen above that in perfect information games, the history is common
knowledge. However, there is a class of games where players move without
necessarily knowing the past. Such games are categorized as imperfect
information games.

In the present unit our analysis of dynamic games is restricted to complete

information by which we mean that the term is used to describe a game in
which knowledge about other players is available to all participants. We also
discuss the class of games in which the player with the move knows the full
history of the play of the game thus far. Therefore, we call these games with
perfect information: For example, Chess is a game with perfect. information as
each player sees all of the pieces on the board at all times.

Note that perfect information is different from complete information. In

complete information game, the structure of the game and the payoff functions
of the players are commonly known. In games of perfect information, on the
other hand, every player observes other players' moves, but may lack some
information on others' payoffs, or on the structure of the game. Thus, a game
with perfect information mayor may not have complete information.
However, when a game is described as having perfect information, complete
information is often assumed, unless specified otherwise.

22.3.1 Backward Induction

Extensive game's with perfect information can be analyzed by backward
induction. A backward induction is the process of reasoning backwards in
time, from the end of a problem or situation, to determine a sequence of
optimal actions. It proceeds by first considering the last time a decision might
be made and choosing what to do in any situation at that time.

To understand the process of applying backward induction consider the

quality choice game where consumers' choice of buying or not depends on the
quality of services (high or low) (see,Figure 22.2).The service provider, player
I, makes the first move, choosing High or Low quality of service. Then the
customer, player 11, is informed about that choice. Player 11 can then decide
separately between buy and don't buy in each case. The resulting payoffs are
also given in the figure.

80
 Dynamic Games with
Complete
(2,2) Information

(1, 1)

Fig. 22.2: Quality Choice Game

Here, player II moves last. Since she knows the play will end after her move,
she can safely select the action which is best for her. If player I has chosen to
provide high quality service, then the customer prefers to buy, since her
resulting payoff of Z is larger than I when not buying. If the provider has
chosen Low, then the customer prefers not to purchase. The choices available
to player II are indicated by arrows in figure.

When the last move is decided, backward induction proceeds to the player
making the next-to-last moves (and the moves continue). In Figure 22.2,
player I makes the next-to-last move, which in this case is the first move in the
game. Being rational, he anticipates the subsequent choices by the customer.
He therefore realizes that his decision between High and Low is effectively
between the outcomes with payoffs(2,2) or (1,1). Clearly, his choice for High
results in a payoff of 2 for him. In case of a selection of Low, the outcome is
payoff 1. So the unique solution to the game, as determined by backward
induction, is that player I offers high quality service, and player II responds by
buying the service.

See that the concept of backwards induction corresponds to the assumption

that it is a common knowledge that each player will act rationally at each node
where she moves - even if her rationality would imply that such a node will
not be reached. More precisely at each node i the player is certain that all the
players will act rationally at all nodes j that follow nodei; and at each node i
the player is certain that at each node j that follows node i the player who
moves at j will be certain that all the players will act rationally at all no desk
that follow node j, ...ad infmitum.

22.3.2 Strategies in Dynamic Games with Complete

Information
In an extensive game with perfect information, backward induction usually
yields unique choices at the players' decision nodes. However, there is an
exception to it. For, when a player is indifferent between two or more moves
at a node the selection of the option is blurred.

The solution through backward induction specifies the way the game will be
played. Starting from the root of the tree, play proceeds along a path to an
outcome. Note that the analysis yields more than the choices along the path.
Because backward induc~on looks at every node in the tree, it specifies for 81
 Non-Cooperative every player a complete plan of what to do at every point in the game where
Game Theory-I the player can make a move. Note that a player's plan includes a node even
when that point may never arise in the course of play. Such a plan is called a
strategy of the player. For example, a strategy of player II in Figure 22.2 is
"buy if offered high-quality service, don't buy if offered low quality service."
Such a strategy for player II's is obtained by backward inductiop.

With strategies defined as complete plan of move by the player, one can
obtain the strategic form representation of the extensive games. Recall that the
strategic form game tabulates all strategies of the players. In the game tree,
any strategy combination results into an outcome of the game, which can be
determined by tracing out the path of play arising from the players adopting
the strategy combination. The payoffs to the players are then entered into the
corresponding cell in the strategic form. Figure 22.3 shows the strategic form
for quality choice example. The second column is player II's backward
induction strategy, where "buy if offered high-quality service, don't buy if
offered low-quality service" is abbreviated as H: buy, L: don't.
:I'

H:buy; H:buy; L: . H: don't;L:

L:buy don't H: don't; L:buy don't

High 2,2 2,2 1,0 1,0

Low 0,3 1,1 0,3 1,1

Fig. 22.3: Strategic Form of the Extensive Game in Fig. 22.2.

A game tree can therefore be analyzed In terms of the strategic form. It is not
difficult to see that backward induction defmes a Nash equilibrium. In Figure
22.3, it is the strategy combination (High; H: buy, L: don't).

The sequential games converted into normal form typically contain

redundancies. Figure 22.3, for example, has eight cells, but the game tree in
Figure 22.2 has only four outcomes. Every outcome appears twice, because
there are two strategies of player II which differ only in the move that is .not
reached after the move made by player I. All the combinations of moves by
player II must be distinguished as strategies. Otherwise, any tw~ of them may
produce to different outcomes, depending on the action of player I.
See that in Figure 22.3, cell with (Low; H: don't, L: don't) is an equilibrium.
Here the customer never buys. So Low is the best response of the service
provider such an expected behaviour of player II. Although H: don't is not an
optimal choice, player II has no occasion to make that move. Therefore,
player II is not better off by changing her strategy. Thus, while we have an
equilibrium, it a suboptimal move. We require additional tool of subgame for
reaching a perfect equilibrium.

22.3.3 Subgames
A subgame is a subset of any game that includes an initial node. To
understand subgames use the extensive form representation. In the following
game tree (Figure 22.4) there are six subgames other than the game itself, two
of them containing two subgames each.

82
 Dynamic Games with
Complete
Information

Fig. 22.4: Subgames in a Game Tree

To understand the concept of subgame, it will be useful remember the

following points:
• Any game is asubgame of itself. Subgames other than the original game
itself are called proper subgames.
• For games of perfect information, every node other than a terminal node
defmes a subgame.
• Any subgame is a game in its own right, satisfying all of our rules for
game trees.
• A strategy for the original game also defines a strategy for each of its
subgames, sometimes called a' continuation strategy

22..3.4 Subgame Perfect Equilibrium

A subgame-perfect equilibrium is a solution of the game where not only
overall, but also each subgame is a Nash equilibrium. To understand the
concept consider the following game: player 1 has to decide between going up
or down (UID), while player 2 has to decide between going left or right (LlR).
We represent .the game using the strategic form specifying all the possible
strategies for player 2:
_ go Right if player 1 goes Up, go Left otherwise;
_ go Left if player 1 goes Up, go Right otherwise;
_ go Right no matter what;
_ go Left no matter what.
The extensive form of the game is given in Figure 22.5 (game tree on the left).
For its strategic form see the presentation in the right part of the figure. Since
it is a sequential game, we describe all possible outcomes depending on player
2 decisions, as seen in the game matrix.
2,5 PLAYER 2
R,L L,R R,R L,L

PLAYER 1
5,2
0,0
U 2,2
- -5,2 --
2,5 2,5

PLAYER 1

D 0,0 3,1 3,1 0,0

3,1

Fig. 22.5: Extensive-Form Game Converted to Normal Form

83
Non-Cooperative The equilibrium of the game can be seen by considering both the parts of the
Game Theory-I figure. It may be seen that Up-Left is a Nash equilibrium. However, it's not a
perfect equilibrium. In order to fmd the subgame-perfect equilibrium, we need
to solve the game through backwards induction. Starting at the last move of
the game, we proceed to the second to last move, and so on. In the game
player 2 will choose Left if player 1 goes Up, and Right if player 1 goes
Down, since these are the moves that maximise his utility. Because there is
complete information, each player's payoffs are known. So, player 1 knows
these choices in advance, and will therefore choose to go Down, because the
payoff will be greater. Therefore, Down-Right is the subgameperfect
equilibrium.

22.3.5 Backward Induction and Subgame Perfection

In extensive-form games, we can have a Nash equilibrium profile of strategies
where player 2' s strategy is a best response to player l' s strategy, but where
she will not want to carry out her plan at some nodes of the game tree.
For example, consider the following game, given in both normal-form and
extensive- form.

Player 2
L R
Player 1 A 1,2 1,2
B 0,0 2,1
1

0,0 2,1
This game has two Nash equilibria: (A,L) and (B,R). If the players have
planned to select right move before the game starts, then the NE profile (A,L)
makes sense. If player 2 is convinced that player 1 will play A, then playing L
is a best response. In fact, she would do everything in her power to convince
player 1 that she will play L, so that he plays A.

While player 2 might "threaten" to play L before the game starts, if the game
proceeds and it is actually her turn to move, then L is not rational. The threat
is not credible, and player 1 should realize this and play B.
Thus, L might be a reasonable plan for player 2 if she thinks player 1 will play
A, but it is not rational to carryout the plan after player 1 plays B. It is not
sequentially rational for player 2 to play L.

If sequential rationality is common knowledge, then players will look ahead

and realize that participants of the game will not take actions down the tree
84 that are not rational at that point.
 Dynamic Games with
For games of perfect information, backward induction is the process of Complete
"looking ahead and working backwards" to solve a game based on common Information
knowledge of sequential rationality. The process is as given below:
In Figure 22.6 by looking ahead to player 2's decision node, we find that her
optimal choice is R. So we can convert her decision node into a terminal node
with payoffs (2,1).
1

0,0 2, 1
Fig.' 22.6: Optimal Choice in Extensive-Form Game

Next, examine the smaller games. In Figure 22.7, player 1 can either choose
A and reach the terminal node with payoffs (1,2), or B and reach the terminal
node with payoffs (2,1).
1

1,2

0,0 2, 1
Fig. 22.7: Extensive-Form Game with Optimal Choice BR

We find that player l's optimal choice is B. So, backward induction leads to
the strategy profile (B,R), with payoffs (2,1).
Note that every [mite game of perfect information can be solved using
backward induction. If each player has a strict preference over his possible
terminal node payoffs (no ties), then backward induction gives a unique
sequentially rational strategy profile. The resulting strategy profile is a Nash
equilibrium.
The following points may be useful in while dealing with the solution concept
subgame perfect equilibrium:
• A strategy profile for an extensive-form game is a subgame perfect Nash
equilibrium (SPNE) if it specifies a Nash equilibrium in each of its
subgames.
85
 Non-Cooperative • The idea behind SPNE is that even if a NE strategy profile dictates that
Game Theory-I
certain subgames are not reached, we require that what the players would
. do conditional on reaching those subgames should constitute a NE. The
"off-the-equilibrium-path" behaviour can be important, because it affects
the incentives of players to follow the equilibrium.
• Notice that every SPNE must also be a NE, because the full game is also a
subgame.
• For [mite games of perfect information, any back ward induction solution
is a SPNE and vice-versa. The advantage of SPNE is that it can be applied
to games of imperfect information too.

22.3.6 Application 1: Stackelberg Competition

With quantity competition, the timing of production .rnatters. In the
Stackelberg Game, firm 1 produces its output first. Then firm 2 observes
q1before deciding its ownoutput.

Let us adapt the Cournot model covered earlier. Marginal production cost is
.,. equal to 100, and market inverse demand is given by p = 1000 - q1 - qz .
(If the expressionis negative, we take the price to be zero.)

Firm l's strategy set is51 = {q1: q1 ;::: O}.

Firm 2' s strategy set is more complicated, reflecting the extensive-form nature
. of the game. Firm 2 can let q2 depend on q 1.

5z = {functions qz (q1) : qz (q1) ;::: 0 for eachq.},

A Nash equilibrium is a profile of strategies, (qi, q;(q1)).In any SPNE, the
subgame following any q1must be aNE. In other words, q; (q1) must be a
best-response to q1.

Notice the distinction: Under simultaneous-move Cournot competition, firm 2

chooses a quantity, which in the NE -is a best response to []fm 1's quantity qi,.
Under Stackelberg competition, firm 2 chooses a function describing firm 2's
choice at each information set, and in the SPNE this function is BRz(q1).
Firm 2 must bestrespond to every possible output from []fm 1, not just qi.

Now that we have found the strategy of firm 2 satisfying sequential

rationality, we can work backwards and findq]. Firm 1's payoff function is

U1(qv qz) = [1000 - q1 - qZ]q1 - 100q1·

In a SPNE, firm 1 chooses a best response to firm 2'sstrategy function, which
we can substitute into the payoff function:

U1(qVq;(q1)) = [1000- q1 - (450 - ~1)] q1 -100q1

= [550 - ~1] q1 - 100q1 = [450 - ~1] q1 -100 q1

Taking the derivative with respect to ql, setting it equal to zero, and solving
for q1yields450 - ql = 0

qi = 450.

The SPNE is( 450,450 _ ql ).

86 Z
 Dynamic Games with
Along the equilibrium path, we see that firm I chooses output of 450, and firm
Complete
2 chooses an output of 225. The corresponding price is 325. This is very Information
different from the Cournot quantities, (300, 300), and price, 400.
You can check that there is a first mover advantage. Firml receives a higher
payoff than firm 2, and a higher payoff than it would receive under
simultaneous play.
Although the game has one SPNE, there are many NE that violate sequential
rationality, for example:
q1 = 64 q2(q1) = 418ifq1 = 64 = 10000therwise.

22.3.7 Application 2: Sequential Bargaining

How will people bargain over a pie of a certain size? A "reasonable" or "fair"
division of it could be (1/2,1/2) there number of agents involve is two.
However, bargaining can involve bluffing and posturing, there is no certainty
that an agreement will be reached. A paper by Rubinstein (1982), however,
showed that a sequential bargaining would yield a unique subgame perfect
equilibrium. We consider this model.

1) The Model
Take two players, one and two, playing the game sequentially and
making offers about how to divide a pie of size one. Time runs from
t = 0,1,2, ..... At time 0, playerone can propose a split (xo,l -
. xo)(with Xo E [0, 1]), which player 2 canaccept or reject. If player 2
accepts, the. game ends and the pie is consumed. If player two rejects, the
game continues to time t = 1. She now proposes a split(yv 1 - Y1)'
Such a proposal of player two may be accepted or rejected by player
one. The sequence of the game goes to ad infinitum.

Let us assume that both players want a larger slice, and also that they
both dislike delay. Thus, if agreement to split the pie (x, 1 - x)is reached
at timet, the payoff for player one is 8fx and the payoff for player two
is 8~(1 - x), for some 81,82 E (0,1).

2) A Finite-Horizon Version
We consider a variant of the game where there is some [mite number of
offers Nthat can be made. If we solve for the subgame perfect
equilibrium through backward induction, we need to start from the final
offer.

For simplification assume N = 2. Then at date 1, player two will be able

to make a final- take-it-or-leave-it offer. Given that the game is about
to end, player one will accept any split. Therefore, player two can
offery = O.
Consider its implication for date zero. Player two expects that if she
rejects player one's offer, she can get the whole pie in the next period and
the total payoff is 82, Thus, to get her offer accepted, player one must
offer player two at least 82, It follows that player one will offer a split
(1 - 82, 82),and player two will accept.

See that in the N = 2 offer of the sequential bargaining game, the

unique SPE involves an immediate (1 - 82,82) split.
87
Non-Cooperative 3) Solving Rubinstein Model
Game Theory-I
We consider infinite time horizon to solve the bargaining game examined
above for two periods. The problem here is we cannot apply backward
induction. Fortunately Rubinstein (1982), showed that the solution has a
remarkably simple form.

Suppose player one makes an offer at a given date t. then player two.
acceptance of it depends on her belief about what she will get if she rejects.
This in turn depends on what sort of offer player one will accept in the next
period, and so on .. Even with such a time horizon we will show there is a
unique subgame perfect equilibrium in the sequential bargaining game.

When player one proposes, she suggests a split (x,l - x)withx = (1-
82) / (1 - 8182), Player two accepts any division giving her at least 1 - x.
When player two proposes, she suggests a split (y,l - y)with Y = 81 (1 -
82) / (1 - 8182), Player one accepts any division giving her at least y. Thus,
bargaining ends immediately with a split(x, 1 - x).

We start by examining that proposed equilibrium is actually an SPE. By a

classic dynamic programming argument of the one-step deviation principle,
we know that it suffices to check that no player can make a profitable
deviation from her equilibrium strategy in one single period.

Now consider a period when player one offers. Player one has no profitable
deviation. She cannot make an acceptable offer that will get her more thanx.
Because when the offer made by her is rejected she will get y = 81x the next
period, or 8f x in present terms, which is worse than x. Player two also has no
profitable deviation. If she accepts, she gets 1 - x. If she rejects, in the next
period she gets 1 - y, or in present terms 82(1 - x) = 82(1 - 81x). A
similar argument applies to periods when player two offers.

We now show that the equilibrium is unique. To do this, let El, 171denote the
lowest and highest payoffs that player one could conceivably get in any
subgame perfect equilibrium starting at a date where he gets to make an offer.

To begin, consider a date where player two makes an offer. Player one will
certainly accept any offer greater than 81171 and reject any offer lessthan 81171'
Thus, starting from a period in which she offers, player two cansecure at least
1 - 81171 by proposing a split (81 Vv 1 - 81171)' On the otherhand, she can
secure at most 1 - 81v1.

Now, consider a period when player one makes an offer. To get player two to
accept, he must offer her at least 82 (1 - 81Vl)tO get agreement.

Thus:

171 ~ 1 - 82(1 - 811\).

At the same time, player two will certainly accept if offered more than

82(1 - 81171)' Thus:

El ~ 1 - 82(1 - 81El)'

It follows that: El ~ 1~8; ~

l-U1U2
171' Since 171 ~ El by definition, we know that
in any subgame perfect equilibrium, player one receives Vl = (1-
82) / (1 - 8182), We can make the same argument for player two.
88
Few Comments on Rubinstein Model of Bargaining Dynamic Games with
Complete
1) It helps to be patient. Note that player one's payoff, (1 - 02) / (1 - Information
0102), is increasing in 01 and decreasing in 02. The reason is that if you
are more patient, you can afford to wait until you have the bargaining
power (i.e. get to make the offer).
2) The first player to make an offer has an advantage. With identical
di~count factors 0, the model predicts a splitC: 0 1: o)which is better for
°~
J

player one. However, as 1, this ftrst mover advantage goes away.

The limiting split is (112, 112).

3) There is no delay. Player two accepts player one's first offer.

4) The details of the model depend a lot on there being no immediate

counter-offers. With immediate counter-offers, it turns out that there are
many many equilibria.

Check Your Progress '2

1) Differentiate between games with complete information and perfect
information .

.............~ ' .
2) What is backward induction?

....................................................... ' .

3) Define a subgame.

4) Defme the concept of Subgame Perfect Equilibrium.

89
Non-Cooperative 5) Explain the relationship between Backward Induction and Subgame
Game Theory-I Perfection .
........................................................................................
........................................................................................
..........................................................................................
........................................................................................

........................................................................................
6) What are the three important components a price leader in the
Stackelberg model is assumed to know?
........................................................................................

........................................................................................
........................................................................................

........................................................................................
........................................................................................

7) Point out the main insight of Rubinstein' sbargaining model.

........................................................................................

........................................................................................
........................................................ ' .

........................................................................................

........................................................................................

8) Sequential Bargaining problem is solved using: i) subgame or ii)

backward induction .
.......................................................................................
. ... ... ... ... ... . ... ~ .
........................................................................................
........................................................................................

.........................................................................................
9) Use backward induction to determine the subgame perfect equilibrium of
the following game:
A,B
0,2

2, 1
A,B

1,0
........................................................................................
........................................................................................

........................................................................................
90
 Dynamic Games with
10) Use backward induction to determine the subgame perfect equilibrium of
Complete
the following game: Information
A,B,C
1, 1, 1
A,B,C
2,3,2

0,0,2

1,2,4

0,2,0
........................................................................................
........................................................................................
. .
........................................................................................
........................................................................................

........... , .
11) Use backward induction to determine the subgame perfect equilibrium of
the following game:

A,B
4,5
A,B
1,0

2,2
........................................................................................
........................................................................................

........................................................................................
........................................................................................
........................................................................................

22.4 DYNAMIC GAMES OF IMPERFECT

INFORMATION
Often players do not have full access to all the information which is relevant
to their choices. Such games with imperfect information model exactly which
information is available to the players when they make a move. 91
Non-Cooperative A dynamic game of complete but imperfect information can be represented in
Game Theory-I extensive form by using non-singleton information sets to indicate what each
player knows (and does not know) when he or she has to the move. Thus,
there are three analytically equivalent defmitions involving the rules, viz.,
In dynamic games of imperfect information,
1) at some stage of the. game some players do not know its entire
development before the current move;
2) players may move simultaneously; and
3) some players at some stages may move while not observed by the others.

These games can be represented in extensive (game tree) form by using non-
singleton information sets to indicate what each player knows (and does not
know) when she has to the move. Since each player does not know all earlier
moves, we must partition the set of player's decision nodes into subsets, called
information set. We connect the nodes are connected with a dashed line if the
moving player does not know from which node of the game tree he is making
the move (see Figure 22.8).Sometimes instead of dotted line we can use oval
format to represent information set.

(1,-1) A wins

(-1,1) a wins
(-1,1) a wins

Fig. 22.8: Extensive-form Game with Information Set

When information sets are all singleton (made of a single node) the game is
one of perfect information. In games of imperfect information however, some
information sets includes more than one node.
Example: Prisoner dilemma in extensive form.

B Bs co

cooperate defect axJ,.4": .•..••.

v~.~
,,:~
~ ~.-.~ J ~4

The game is symmetric: rows and columns can be exchanged in normal

form. So, first and second move can be exchanged in extensive form.

92
 Dynamic Games with
22.4.1 Imperfect Information and Backward Induction Complete
In games of imperfect information backward induction procedure has to be Information

considered when
• a player does not know which node he is moving from he cannot single
out the best action;
• preceding player into the game tree cannot anticipate what she will do;
• common knowledge is not able to give the usual information;
• a dominated strategy not available.
A special case: prisoner dilemma
3,3
Bs

coo e •I

<,;
A

.... ~:~ : •

BI~
2,2
In tree above given asa special case of prisoner dilemma, even if B does not
know if she is in as
or in BI she can observe that, whatever her starting node,
"defect" gives her more utility then "cooperate":
• 4 instead of 3 if A has done "cooperate";
• 2 instead of I if A has chosen "defect" This is because "cooperate" is a
dominated strategy.

22.4.2 Subgames with Imperfect Information

A subgame is a subset of the extensive form that satisfies the following
criteria: .
• begins at a node (singleton);
• includes all nodes following this initial node and no others;
• does not cut any information sets (if two nodes are part of the same
information set they belong to the same subgame).
The characteristics of subgames are:
• self-contained. extensive forms, meaningful trees on their own;
• subgames that start from nodes other than the initial node are called proper
subgames; and
• in a game of perfect information, every node initiates a subgame.
Figure 22.9(1) presents these characteristics.
Figure 22.9(2): Information Sets and Subgames

93
Non-Cooperative
3,3
Game Theory-I

I
A •
I .1,4
• 4,1

l~
B~

In the figure player A has one information set that is a singleton. Player B has
one information set with two nodes (none of which is a singleton). See that no
subgames exist for PD but the whole game.

,
j

I
I
I

I~~~~.
I \i

~~:"....,..
=--~------.•
1 •••••

c
Fig. 22.9(2): Information Sets and Subgames

In the figure above each player A, B, C has three moves: {V, M, D} ; {u, rn,
d} ; {D, Jl, ()}.
• At the first stage A has one information set (singleton).
• At the second stage B knows that A either chooses {D} or {V,M} that are
the two information sets of B.

• At the third stage C knows that B has chosen among {Dd,Dm} , {Du},
{Md,Mm,Vd,Vm}, [Mu.Uu} that are four C's information sets.
Therefore, the game has three subgames (two proper subgames): i) the
original game, ii) a subgame following D, iii) a subgame following the path
D-u.
Through the concept of information set game theory can face strategic
interactions that are partly sequential and partly simultaneous moves.

94
r

Dynamic Games with

22.4.3 Strategies with Imperfect Information Complete
To determine NE, it is necessary to know the strategies of each player i (i = 1, Information
2, ... , n). A definition of strategy (a complete plan of action) is needed for
games in extensive form. For any given player i a strategy specifies what that
player should do at any information set.

22.4.4 Finding SPNE with Imperfect Information

First solve each proper subgame and then go back to the whole game or
viceversa. It should always produce the same result.
Example: strategies in the tree-J

1.0

2,3

By definition
• A has four strategies: {UV, UE, DV, DE}
• B has 'two strategies: {u, d}
Although nv
and DE are sequences of moves never played by A, they
cannot be discarded.
Convertion of Extensive Games to Normal Form

B
u cl
1.0
uv 4,1 1,0
2,3

The matrix shows three NE: {UV,u}, {DV,d}, {DE,d}

... and back to the tree.
Example: strategies in the tree-2
Non-Cooperative
Game Theory-I

NE: {UV,u}, {DV,d}, {DE,d}

Are all those equilibria coherent with backward induction?
By backward induction if A has the opportunity to make her second move, she
will choose E (2)1). So the only strategy profile coherent with backward
induction is {DE,d} with payoff(3,4).
Again refer back to the tree -2.
You will find the NE: {UV,u}, {DV,d}, {DE,d}. The same result is more
quickly achievable through the concept of subgames.

There are two subgames:

• the first starts at the node where A moves for the second time;
• the second is the whole game {UV,d} and {DV,d} do not satisfy subgame
perfection. The only SPNE is {DE,d} with payoff(3,4).
Example: Game Representation with Three Players

96
 Dynamic Games with
Complete
Information

With three players we represent the game in normal form. See that the
representation of the game is done with two matrices as given below:

A plays U A plays D
C C
Left Right Left Right
u 3, 1, 1 3, 1, 1 B u 1,2,2 2,0,3
B 0,0,0 4,2,1
d 3, 1, 1 3, 1, 1 d

A: matrix player; B: row player; C: column player

The game has the following NE: {U,u,left}, {U,u,right}, {U,d,left}, {D,d,right}.
For instance {U,u,left} yields payoff (3,1,1) is one NE.
Consider the subgame where only players Band C play (See figure below).

7 22
.1
dill

C
u/:'-
B/' _1 ri8~ 0,3 left right
·C;,
\7" 010
B
u 2,2 013

d O,ft 1,1
2,1
It can be seen that "right" is a dominant strategy for player C and the matrix of
the subgame has the only NE {d,right}. That excludes {U,u,left} from being a
SPNE. The only SPNE of the original game is {D,d,right} leading to the
. payoff (4,2,1).

Check Your Progress 3

1) What the difference between perfect and imperfect information?
........................................................................................
........................................................................................
........................................................................................
........................................................................................
97
Non-Cooperative 2) What is an Information set?
Game Theory-I

3) In games of imperfect information when would you consider backward

induction procedure?

4) What is a subgame perfect Nash equilibrium?

.........................................................................................

5) A "credible" threat is a threat of an action that a player would be willing

to undertake if---------

6) A subgame perfect equilibrium is a Nash equilibrium that involves only--

22.5 REPEATED GAMES

In this section we analyse whether threats and promises about future can affect
the current behaviour when a game is played repeatedly. Also we will define
the concept of sub-game perfect Nash equilibrium for repeated games
98
22.5.1 Two-Stage Repeated Games Dynamic Games with
Complete
Information
In two-stage repeated game, players play the game once again after the
payoffs of the game is collected by them. Consider the Prisoners' Dilemma
given in normal form in the following table

Player 2 Player2
L2 M2 L2 M
2
Player 1 Ll (1, 1) (5,0) Player 1 Ll (2,2) (6, 1)
Ml (0,5) (4,4) Ml (1,6) (5,5)

Suppose the players play the game twice, and they observe the outcome of the
first game before playing it once again. We further assume that the payoff of
the entire game, taking two stages together is simply the sum of the payoffs
from the two stages (we assume there is no discounting, that is, the same
amount of payoff in any period gives the same amount of utility. Generally, if
we think in monetary terms we will prefer Rs. 100 today than Rs. 100
tomorrow. This is because there is a discount rate, which is embedded in our
inter-temporal choice decisions).

It is easy to see that the Nash equilibrium of the first stage of the game is (Ll,
L2) with payoff (1, I); we have added the equilibrium payoff of the first stage
to the payoffs of the game in the next stage. In the second stage also there is a
unique N ash equilibrium of the two-stage prisoners' dilemma.

22.5.2 Finitely Repeated Games

The argument stated above holds more generally. We generalise the concept
of repeated games to games with fmite number of repetitions. Let G = {AI,
A2... , An; UI,U2,..., un} denote a static game of complete information in which
players 1 through n simultaneously choose their strategies al to an respectively
from their strategy spaces SI through Sn. This game is repeated T times after
collecting and observing the outcomes of each game. The game G is called the
stage game of the repeated game.

Defmition: Given a stage game G, let G(T) denote the finitely repeated game
in which G is played T times, with the outcome of all the previous games
observed before the next play begins. The payoff for G(T) is simply the sum
of the payoffs from the T stage game.

Proposition: If a stage game G has a unique Nash equilibrium then, for any
finite T, the repeated game G(T) has a unique sub-game perfect Nash
equilibrium: the Nash equilibrium of G being played in every stage of the
game.

It is interesting to investigate if there is more than one Nash equilibrium in the

stage game itself. What will be the nature of the equilibrium when the game is
repeated? We consider a simple two period game, which is a simple extension
of the prisoners' dilemma such as there are two Nash equilibria in it. In
addition to the strategies Ll and Ml, we add another strategy RI at the
disposal of player 1. Similarly, we add the strategy R2 to the strategy space of
player 2. The game is described below in normal form. .

99
Non-Cooperative
Player 2
Game Theory-I
L2 M2 R2
Player Ll (1,1) (5,0) (0,0)
1
MI (0,5) (4,4) (0,0)
RI (0,0) (0,0) (3,3)

As a result of adding the two strategies and the distribution of payoffs now
there are two pure strategy Nash equilibrium, namely, (LI, L2) and (RI, R2).
Suppose the above stage game is played twice with the first stage outcome
being observed before the second stage begins.

Since the stage game has more than one Nash equilibrium, it is now possible
for the players to anticipate that different first stage outcomes will be followed
by different stage game equilibria in the second stage. Let us suppose for
example the players anticipate that (RI, R2) will be the second stage outcome
if the first stage outcome is (MI, M2), but (LI, L2) will be the second stage
outcome if anyone of the eight other first stage outcomes occur. Thus, the
game reduces to the following one shot game, where (3, 3) has been added to
the (MI, M2) cell and (1, 1) has been added to rest of the cells.

Player 2
L2 M2 R2
LI (2,2) (6,1) (1,1)
Player 1
MI (1,6) (7,7) (1,1)
RI (1,1) (1,1) (4,4)

There are three pure strategy Nash equilibrium of the above game (LI, L2),
(MI, M2) and (RI, R2). Let us denote the outcome of the repeated game as
[(w, x): (y, z)]; where (w, x) is the first stage outcome and (y, z) is the second
stage outcome. Therefore, the Nash equilibria namely, (LI, L2), (MI, M2)
and (RI, R2) can be achieved in the simplified one shot game if the outcomes
of the repeated game are [(LI, L2), (LI, L2)], [(MI, M2); (RI, R2)] and [(RI,
R2), (LI, L2)] respectively (if the first stage outcome is (LI, L2), the second
stage outcome has to be (LI, L2) according to the players' anticipation and so
on for each of the one shot game's Nash equilibria). The first and the last
Nash equilibria of the repeated game are sub-game perfect and simply
concatenate the Nash equilibrium outcome of the stage game. But the Nash
equilibrium (MI, M2) of the one shot game is possible if the outcome of the
repeated game has the sub-game perfect outcome [(MI, M2); (RI, R2]],
which means in the first stage, the players chose (MI, M2) which is not a
Nash equilibrium of the stage game. We can conclude that cooperation can be
achieved in the first stage of a sub-game perfect outcome of a repeated game.
Proposition: We extend this idea to a stage game being played T times, where
T is any finite number. If G = {AI, A2... , An; UI, U2,..., Un}is a static game of
complete information with multiple Nash equilibria, then there may be sub-
game perfect outcomes of the repeated game G(T) in which for any t< T, the
outcome in stage t is not a Nash equilibrium of the stage game G.
The main points to extract from the above example is that credible threats or
promises about the future behaviour can affect current behavior sub-game
perfection (as we described in the previous unit) may not be a definition
100 strong enough to embody credibility.
22.5.3 Infinitely Repeated Games Dynamic Games with
Complete
In the finite horizon case, the main focus was that credible threats or promises Information
about future behaviour can influence current behaviour and if there are
multiple Nash equilibria of the stage game, then there may be sub-game
perfect outcomes of the repeated game G (T) in which, for any t < T, the
outcome of a stage game is not a Nash equilibrium of G. Whereas in case of
infmitely repeated games a stronger result is true. Even if the stage game has a
unique Nash equilibrium, there could be sub-game perfect outcome of the
infmitely repeated game in which no stage's outcome is a Nash equilibrium of
the stage game G.

Player 2
L2 M2

Player I LI (I, I) (5,0)

MI (0,5) (4,4)

An infinitely repeated game is an extension of a fmitely repeated game, it

being played infinitely. Suppose the prisoners' dilemma game is to be
repeated infmitely and for each t, the outcome of the t-I preceding plays of the
stage game is observed before the tthstage begins. Simple summation of the
payoffs from this infinite sequence of stage games does not provide a useful
measure of a player's payoff in the infinitely repeated game. This is because
receiving a payoff 4 is better than receiving a payoff I in every stage but a
summation of the payoff I, repeated till infinity and that of 4 is same, which is
infmity. To tackle this problem, we introduce the concept of discount rate. As
we have argued earlier, Rs. 100 today is not the same as Rs. 100 tomorrow. If
the rate of interest is 'r', one can earn (100xr) one-year later in addition of the
principle Rs. 100. Therefore, Rs. 100 today is worth Rs. 100(1 +r) tomorrow.
To fmd the present value of a future income or future stream of income, we
must discount it to get the present value of the future income. To get the
present value of future income to be received t years later we multiply, it with
-C 1 )t' This fraction -c 1 ) is called the discount factor; it is generally denoted
l+r l+r
by O. We can apply this method of calculating the present value of an income
stream to calculate the present, value of the payoffs of an infinitely repeated
game.
Defmition: Given the discount factor 0, the present value of the infinite
sequence ofpayoffs 7r1,7r2,7r3:··isgiven by

7r1 + 07r2 + 027r3 + ... = ~CX) at-i. tt,

Lt=l
Let us consider the infmitely repeated prisoners' dilemma, where the payoff of
the players is the present value of the player's payoff from the stage games.
The discount factor is a
for both the players. We want to show that
cooperation, that is, (MI, M2) can occur in every stage of a sub-game-perfect
outcome of an infinitely repeated game, even though the only Nash
equilibrium in the stage game is cooperation, which is (LI, L2).
Suppose the ith player begins the infinitely repeated game by cooperation and
then cooperates in rest of the periods if and only if both the players have
cooperated in the previous stages of the game. Formally, the strategy of the ith
player is Play Mi in the first stage.

101
 Non-Cooperative In the ehstage if the outcome of all the preceding (t-l) has been (Ml, M2) then
Game Theory-I play Mi in the tthstage otherwise play u. (i = 1, 2) (This type of strategy is
called a trigger strategy). If both the player follows this trigger strategy, then
the outcome of the infinitely repeated game would be (Ml, M2) in every stage
of the game. Now we will prove that given some conditions on the value of 8,
the above trigger strategy is a N ash equilibrium of the infinitely repeated
game, and such an equilibrium is sub-game-perfect.

To show that the above described trigger strategy is a Nash equilibrium of the
game, we have to show that if the player i adopts the trigger strategy, the best
response of the player j is to adopt the same strategy. The present value of the
payoffs that will be generated, for the j'hplayer, if both of the players stick to
the trigger strategy is given by

234
V =4+48+48 +48 ...=--.
1-8

If the fhplayer deviates from the trigger strategy, that is, she plays Lj in the
"
first stage, which will eventually lead to non-cooperation from player i in the
second stage (Li) and consequently from player j (Lj) also, the discounted
payoff of the fh player is given by (the payoff of the first stage for the lh
th
player would be 5, as the i player in the first period following her trigger
strategy will play Ri, and for the remaining periods the payoff of the player t
2
would be 1)V' = 5 + la + la + la ••• = 5 + ~a
3
1-

Therefore, playing Mj, or following the trigger strategy is optimal for the lh
player given that the ith player .sticks to her trigger strategy if and only if
_4_ ~
i -a
5 + _a_ ~
r-a
a ~~.4
Therefore, the trigger strategy 0 is the N ash
equilibrium for both the players if and only if a ~ ~.
4

Now we are in a position to formally define an infinitely repeated game,

history, strategy and sub games of an infinitely repeated game.
Definition of Infinitely Repeated Game: Given a stage game G, let G (00, 8)
denote the infinitely repeated game in which G is repeated forever and all the
players have the same discounting factor denoted by 8. For each t (t is a
positive integer), the outcomes of the t-l previous plays are observed before
the tth stage begins. Each player's payoff in G (00, 8) is the present value of the
player's payoff from the infinite sequence ofthe stage game.

History of an Infinitely Repeated Game: In the finitely repeated game G(T) or

the infinitely repeated game G (00, 8), the history of play through stage t is the
record of the player's choices in the stage 1 through t.

oStrategy of an Infinitely Repeated Game: Notice while defming the repeated

game in section we used the notation G = {AI, A2 ... , An; UI,U2,,,o,un} while in
previous units we described a static game of complete information as G = {SI,
S2 ... , Sn; UI,U2,,,o,u-I: remember Si is in a static game of complete information.
implies strategy spaces whereas the Ai s in a dynamic or repeated game of
complete information implies the action space. The players might have chosen
(all, a2I, a3I,... anI) in stage 1 (suppose there are n players) and (al2' a22,a32,...
an2) and so on. For each player i and for each stage s, the action aisbelongs to
the Action space Ai.
Thus, in an infinitely repeated game there is a difference between strategy and
actions. In a fmitely repeated game G(T) or an infinitely repeated game G (00,
102 8), a player's strategy specifies the action the player will take in every stage of
r
the game, for each possible history throughout the game. So, strategy is a Dynamic Games with
complete plan of action. It is defined for each possible circumstance and for Complete
Information
every stage of an infinitely repeated game.

Sub-game of an Infmitely Repeated Game: In the finitely repeated game G(T),

a sub-game beginning at the stage t+ 1 is the repeated game in which G is
played T-t times, denoted by G(T-t). There are many sub-games that begin at
the stage t+ 1, one for each possible history of the game through stage t.
Whereas for an infinitely repeated game G (00, 8), each sub-game is identical
to the original game G (00, 8). As in the fmitely repeated game, there are as
many sub-games beginning at stage t+ 1 of G( 00, 8) as there are possible
histories of play. through stage t. In the case of two stage prisoners' dilemma
(with two strategies available to each player), there are four sub-games,
corresponding to the second stage game that will follow the four possible first
stage outcomes. Similarly, the extension of the two stage prisoners' dilemma
where each player had three strategies at their disposal has nine sub-games. A
sub-game is not only a piece of the game which starts at a point where the
history of the game thus far is common knowledge among the players but also
includes all the moves that follow this point in the original game.

Sub-game Perfect Nash Equilibrium of an Infinitely Repeated Game (Selton

1965): A Nash equilibrium is sub-game perfect if the players' strategies
constitute a Nash equilibrium in every sub-game of the infinitely repeated
game. Sub-game perfect Nash equilibrium is refmement of the concept of
Nash equilibrium, which means for a strategy profile to be sub-game perfect,
it must be a Nash equilibrium first and then it must satisfy an additional test,
that is, it must be a Nash equilibrium in every sub-game ofthe game.
Our objective of describing all the defmition above was to show that the
trigger strategy that we have already defined in the game of infinitely repeated
prisoners' dilemma is sub-game perfect. Therefore, we need to show that the
trigger strategy constitute a Nash equilibrium on every sub-game of that
infmitely repeated game. Recall that every sub-game of an infmitely repeated
game is identical to the game as a whole. In the trigger strategy, Nash
equilibrium of the infinitely repeated prisoners' dilemma sub-games could be
-grouped into two classes:

• subgames in which all the outcomes of the earlier stages have been (RI,
R2)
• sub-games in which the outcome of at least one stage differs from (RI,
R2).
If the players adopt the trigger strategy for the game as a whole, then (i) the
players' strategies in a sub-game in the fist case are again the trigger strategy,
which we have already shown to be the Nash equilibrium of the game as a
whole.
The players' strategies in a sub-game in the second case are simply to repeat
the stage game equilibrium (Lj, L2) forever, which is also Nash equilibrium of
the game as a whole. Therefore, the trigger strategy Nash equilibrium of the
infmitely repeated prisoners' dilemma is sub-game-perfect.

22.5.4 Application 3: Collusion between Cournot Duopolists

Friedman was first to show that cooperation could be achieved in an infinitely
repeated game by using trigger strategies that switch forever to the stage game
Nash equilibrium following any deviation. The original application was to
collusion in a Coumot oligopoly.
103
Non-Cooperative Take the case of Cournot oligopoly: if the aggregate quantity in the market is
Q = (q) + Q2)'
Game Theory-I
and the market clearing price is P = a - Q. Assuming Q < a,
and each firm has a marginal cost c, if the firms' choose their quantities
simultaneously, then the unique Nash equilibrium of the game is both firm
producing (a-c), which we call the Cournot quantity and denote it by q.. Since
3

the equilibrium aggregate quantity2 (a-c) , exceeds the monopoly quantity

3
(a-c)
qm = -2 -. Clearly, both the firms would be better off if each firm produced
(a-c)
half of qm, the monopoly quantity--.
, 4

We will consider an infinitely repeated game based on this Cournot stage

game when both the firms have the discount factor a. We will seek for a sub-
game. perfect Nash equilibrium in which both the firms collude and their
payoffs are more
, than Cournot payoff.
Let us consider the following trigger strategy:
Produce half of the monopoly output (qn/2) in the first period. Continue to
produce the same in the eh period if both the firms produced (qm/2) in the r-r"
period; otherwise produce the Coumot output q..

The profit to one firm when both the firms produce (qn/2) is denoted byrr;
i.e. rrm = qm Ca _ q . _ c) = (a-c)2
, 2' 2 m 8

[Check qm= (a - c)/2]

Whereas the profit accruing to each firm when both produce qc is denoted
. (a-c)2
byzr., So,rrc = qcCa - 2qc - c) = --
9

Finally, if firm 1 is going to produce (qu/2) this period, then the quantity that
maximises j's profit in this period, is obtained solving the following simple
. qm

maximisation problem: maxqj (a-~-c) qj. The solution of the problem can
be obtained from the first order condition of profit maximisation, that is,

(a - 2q , ..!. q)
...!!!. - C = O. So, qJ' = a---c
2
qm (a-c)
Or q. = a---c2.2 3(a-c) with
:::;;:-_.
J 2 . 2 'J 2 8
9
associated profIt (a-c)2. We will denote this profit by rrd(d stands for
, 64
deviation).
Therefore, it is a Nash equilibrium for both the firms to play the trigger
strategy, given earlier provided that,present value of payoff from the trigger
strategy 2: present value of the payoffs deviated in the first period. .

rrm rrm rrm rrm (1l:)

Or'2 -+a-+a2_+a3_+"'>rrd+arr
2 2 2 - c
+a2rr c + .. ·tooo=-=
1-8
rrm
2(1-8)

1 1 m 8
Or -'-rr
'1-8 2
>
-
rrd +-n
1-8 c

Substituting the values of !id' and z; into the above equation, we get if
6 2: :7'
then the inequality will hold and the tri~ger strategy will be sub-game
perfect.
Thus, we see that collusion in infinitely repeated games can fetch extra
payoffs to the firms.
104
 Dynamic Games with
Check Your Progress 4 . Complete
1) What is a strategy in a repeated game? Information

........................................................................................
........................................................................................
........................................................................................
........................................................................................
........................................................................................
2) What do mean by a stage game?
.........................................................................................
........................................................................................
........................................................................................
.........................................................................................
.......................... ' .
3) An infinitely repeated game promising A per period with continuing
cooperation, B for the one period gain from cheating, and C per period
for the non-cooperative solution (B > A> C) will have a subgame perfect
Nash equilibrium in trigger strategies providing-----
.........................................................................................
........................................................................................
........................................................................................
........................................................................................
........................................................................................
4) A cartel-like collusive solution can be a Nash equilibrium only in price-
setting games with -------
........................................................................................
........................................................................................
........................................................................................
.........................................................................................
5) a) What is the equilibrium of the following game if it is played as a
I

sequential game with Player A going first?

b) If this game is played as a sequential game with Player B going first,
the equilibrium will be:
. Player B
Left Right
, ~
-, 7 2
. Up -, "- -,
-, -, -,

Player A
1
,
.,
i'
8
"",
Down
" -, -, -, 3 " -,
-, -,
5

4
6
" -, -,
-,
105
Non-Cooperative
Game Theory-I

6) What is a behavioural strategy?

22.6 LET US SUM UP

In this unit we have discussed the dynamic games with complete information.
The dynamic game is a sequential game where the players move after
observing the decision taken by the rivals. The complete information is further
divided as perfect information and imperfect infomiation. Perfect information
refers to the fact that each player has the same information that would be
available at the end of the game whereas imperfect information only reduces
uncertainty but (unlike perfect information) does not eliminate it. We proceed
to discuss the solution concept in dynamic games with its representation in
extensive form and invoking backward induction. Such a technique has helped
refined the Nash equilibrium with multiple equlibria by elimination the
incredible threat. It is also seen that the decision trees with non-singleton
information sets as found in case of game with imperfect information. The
backward induction does not work in such cases. Therefore such games are
. solved with subgame perfect equilibrium. In the process we have learnt that
subgame is a subset of any game that includes an initial node (which has to be
independent from any information set) and all its successor node. With
identification of subgamesit is possible to have a subgame perfect equilibrium
that representsNash equilibrium if every subgame of the original gameis also
a Nash equilibrium. Using 'such solution concepts we have seen their
applications in Stackelberg Competition and sequential bargaining.

The unit discussed repeated games that analyses whether threats and promises
about future can affect the current behavior. Besides learning the presentation
of these games, we have been exposed to their solution mechanismsby
considering finite and infinite time horizons. In the process we have seen the
change that occurs in one shot games like prisoner's dilemma. The use
oftrigger strategy where a player initially cooperates but punishes the
opponent if observed a certain level of defection. Bilateral bargaining is seen
as an attempt to resolve a bargaining situation, i.e., to determine the particular
form of cooperation and the corresponding payoffs for two players involved.
Using the framework of infinitely repeated games we have seen the process of
collusion fetching extra payoff.

106
 Dynamic Games with
22.7 KEY WORDS Complete
Information
Complete Information: Knowledge about other players is available to. all
participants.
Incredible Threat: A threat made by a player in a Sequential Game which
would not be in the best interest for the player to carry out.
Backward Induction: Backward induction is a technique to solve a game of
perfect information. It first considersthe moves that are the last in the game,
and determines the best move for the playerin each case.
Common Knowledge: A fact which all players know.All. know that they all
know it, andso on.
Dominating Strategy: A strategy dominates another strategy of a player if it
always gives a better payoff tothat player, regardless of what the other players
are doing. .
Extensive Game: A game that describes with a tree how a game is
played. Game: A game is a formal description of a strategic situation.
Mixed Strategy: An active randomization, with given probabilities, .that
determines theplayer's decision.
Perfect Information: At any point in time in the game only one player makes
a move,and knows all the actions that have been made until then.
Rationality: A player seeks to play in a manner which maximizes her
ownpayoff. .
Subgame Perfect Equilibrium; An equilibrium in which all actions are Nash
equilibria for all subgames.
Perfect Information: Each information set consists of onlyone node.
Perfect Recall: Each player recalls exactly what he did in thepast.
Discount Factor: If the rate of return is r (which is generally positive), then
1
the ratio- - is called the discount factor. The discount factor in game theory
l+r
literature is generally denoted by a.
Finitely Repeated Games: When a game is repeated fmite number of times
.
it is called a finitely repeated game .
Infinitely Repeated Games: When a game is repeated infinite number of
times it is called a fmitely repeated game.
Present Value: A future asset evaluated at present time is called the present
.value of the asset. If t periods later an asset is worth A, then it's present value
now is given by the formula- A )t.
C l+r

Stage Game: In a repeated game the initial game, which is repeated in every
time period is called the stage game.
Stackelberg Competition: A strategic game in which the leader firm moves
first and then the follower firms move sequentially.

22.8 SOME USEFUL BOOKS

1) Fundenberg, Derw and Jean Tirole (1980), Game Theory, MIT Press.
2) Tirole Jean (1989), The Theory of Industrial Organisation, MIT Press,
Cambridge Massachusetts, London, England. 107
Non-Cooperative 3) Andreu Mas-Colell, Michael D. Whinston and Jerry R. Green (2005),
Game Theory-I Microeconomic Theory, Oxford University Press.
4) Robert Gibbons (1992), A Primer in Game Theory, Harvester Wheatsheaf
5) Martin J. Osbome (2003), An Introduction to Game Theory, Oxford
University Press.

22. 11 ANSWER OR HINTS TO CHECK YOUR

PROGRESSIHINTS
Check Your Progress 1
1) In a static or simultaneous move game players move one time
simultaneously. Any game with sequential moves falls under the category
of dynamic games. A key difference is that in a static game no new
information is revealed to any of the players during the game before they
make their play.
2) The extensive form is a way of describing a game using a game tree. It's
simply a diagram that shows that choices are made at different points in
time (corresponding to each node). The payoffs are represented at the end
of each branch.
3) A complete contingent plan of actions of players as they move sequentially.

Check Your Progress 2

1) In a game of complete information, the structure of the game and the
payoff functions of the players are commonly known but players may not
see all of the moves made by other players, there may also be a chance
element (as in most card games). Conversely, in games of perfect
information, every player observes other players' moves, but may lack
some information on others' payoffs, or on the structure of the game.
2) The process of reasoning backwards in time from the end of a problem to
determine a sequence of optimal actions.
3) A subset of any game that includes an initial node (which has to be
independent from any information set) and all its successor nodes.
4) A subgame perfect equilibrium is a refinement of a Nashequilibrium. A
strategy profile is a subgame perfect equilibrium if it represents a Nash
equilibrium of every subgame of the original game. .
5) If the players played any smaller game that consisted of only one part of
the larger game and arrived at a Nash equilibrium of that smaller game,
then they have a subgame perfect equilibrium of the larger game. A
common method for determining subgame perfect equilibria in the case
of a finite game is backward induction. However, backward induction
cannot be applied to games of imperfect or incomplete information
because this entails cutting through non-singleton information sets.
6) a) the market demand curve, b) its own cost function, c) its rival's
reaction function.
7) Under complete information, equilibrium strategies are determined by
the relative impatience of the bargaining parties. In equilibrium, the
proposer makes an offer so that the responder is indifferent between
accepting the offer and rejecting it, given the cost of waiting; and the
responder accepts the offer.
108
8) Backward induction. Dynamic Games with
Complete
9) {S ; t } with payoffs of (1,0) .. Player A's equilibrium strategy is S; B's Information
equilibrium strategy is "t if N." To characterise a subgame perfect
equilibrium, one must find the optimal strategy for a player, even if the
player is never called upon to use it. In this case, although player B
never has to select between "t" and "b," the fact that the player would
select "t" is what makes playing "S" an equilibrium for player A.
10) {N, N, N ; b ; d } with payoffs (2,3,2). Starting at the end, Player A
would select "N" in either of the last moves he may have to make. If
player B is asked to make a decision, he selects b (knowing that player A
will then select A). If player C is asked to make a decision, he selects d,
knowing that player A will then select N. Now considering the first
period, player A chooses N.
11) {S,S,N; n,n,s } with payoffs of (4,5). Start with the last decision and
work backwards to the root of the tree. At each step, be careful to
concentrate only on the payoffs of the player making the decision.

Check Your Progress 3

1) Under perfect 'information each player has the same information that
would be available at the end of the game. So each player knows or can
see other player's moves. Take for good example the game of chess,
where each player sees the other player's pieces on the board.lmperfect
information, on the other hand, appears when decisions have to be made
simultaneously, and players need to balance all possible outcomes when
making a decision. For example, in a card game each player's card are
hidden from the rest of the players.
2) In the extensive form, an information set is a set ofdecision nodes such
that: Every node in the set belongs to one player. When play reaches the
information set, the player with the move cannot differentiate between
nodes within the information set, i.e. ifthe information set contains more
than one node, the player to whom that set belongs does not know which
node in the set has been reached.
3) When (i) a player does not know which node he is moving from he
cannot single out the best action; (ii) preceding player into the game tree
cannot anticipate what she will do; (iii) common knowledge is not able
to give the usual information; and (iv) a dominated strategy not available.
4) A subgame perfect Nash equilibrium is an equilibrium such that players'
strategies constitute a Nash equilibrium in every subgame of the original
game.
5) It may be found by backward induction, an iterative process for solving
finite extensive form or sequential games. First, one determines the
optimal strategy of the player who makes the last move of the game.
Then, the optimal action of the next-to-last moving player is determined
taking the last player's action as given. The process continues in this way
backwards in time until all players' actions have been determined.
6) In a position to do so.
7) Credible threats.

Check Your Progress 4

1) Strategy is a complete plan of action. A player's strategy specifies the
action the player will take in every stage of the game. 109
Non-Cooperative 2) In a repeated game an extensive form game which consists in some
Game Theory-I number of repetitions of some base game called a stage game .
..i

3) i > CC - A)/CA - B).

4) Infinite replications.
5) a) Down, Right; b) Down, Left
6) Independent assignment of strategy in each move is called the
behavioural strategy. That is, you resort to independent coin toss every
time an information set is encountered

22.12 EXERCISES
1) The simultaneous move game is played twice with the outcome of the
first stage being observed before playing the second stage. There is no
discounting factor. Can the payoff (4,4) be achieved in the first stage in
a pure strategy sub-game perfect Nash equilibrium? If so, give strategies
that do so. If not then, why not?

Player 2
Left Center Right
Top 3,1 0,0 5,0
Middle 2,1 1,2 3,1
Player 1 Bottom 1,2 0,1 4,4

2) Suppose there are n firms in a Cournot oligopoly. Inverse demand

n

function is given by P (Q) = a - Q, where Q = Eqi . Consider the

i=l
infinitely repeated game based on this stage game. What is the lowest
value of a
such that the firms can use trigger strategies to sustain the
monopoly output level in a sub-game perfect Nash equilibrium? How
does the answer vary with n?
3) Explain with a suitable example that in an infinitely repeated game a
feasible payoff can be obtained as a sub-game perfect Nash equilibrium
of the game, for appropriate discounting factor.
4) Does the concept of sub-game change when we talk about infinitely
repeated games.
5) a
Suppose in a Cournot model the discount factor < 9/17. Now, clearly
the firms cannot support a quantity as low as half of the monopoly
a
quantity. (see Section 25.5). But for any value of it is sub-game perfect
Nash equilibrium to repeat the Cournot quantity forever. Therefore, the
most profitable quantity that trigger strategy can support is between half
of monopoly output and Cournot output. Consider the following trigger
strategy to compute that level of output.
Hint: Produce q* in the first period. In the tth period produce q* if
both firms have produced q* in each of the t-I periods. Otherwise
produce the Cournot quantity.
6) Consider the Bertrand Duopoly model with constant marginal cost. We
have seen in the previous unit that the unique Nash equilibrium of this
110
game is (Pl*= P2*=C). Consider the following trigger strategy:
 Dynamic Games with
Player i at period t=O, chooses pm (monopoly price) and at period t>O, will Complete
choose pi= pm if and only if outcome of all earlier stages have been (pm.pm); Information
otherwise choose c forever.
Ans. Let us consider that player 2 strictly follows the trigger strategy. If
player 1 also follows the strategy, the present value of her future payoff
rrm rrm rrm rrm (rr~) rrm
stream is V = -2 + a - 2 + a2 - 2 + a3 - 2 + ...to 00 = --I-a = --2(I-a)
But if she deviates from her strategy, the present value of her future payoff
°
stream is given by V' = tt'" + a(O + + 0 .. · tooo) = tt'", Therefore, there
is no point for him to deviate from the trigger strategy if

V > v' or, -(2 rrm

1-
m
a) > tt or,
a >-.1
2

Thus, the trigger strategy is a Nash equilibrium, but to show that it is sub
game perfect Nash equilibrium, you need to show that the trigger strategy
induces Nash equilibrium in every sub-game of the infinitely repeated
Bertrand duopoly.
There are two possible situations
Case I: No one deviates at any point. Then the sub-game will look: like the
game itself, (pm, pm) is chosen in all earlier stages and the strategy profile
induces (pm, pm) in the subsequent games, which we have shown to be
Nash equilibrium provided the restriction on a.
Case II: If something else happens in the game, then the strategy will
induce (c, c) at all subsequent games, which we have shown to be a Nash
equilibrium ill the previous unit.

111