You are on page 1of 28


An Analysis of Business Issues

Howard Davies
and Pun-Lee Lam
Published by FT Prentice Hall

Chapter 13:

Game Theory and Its Application

in Managerial Economics

■ On completion of this chapter you should:
– understand the place of game theory in Economics
– be able to represent and solve simple games
– apply game theory to the issue of collusion
– model Cournot, Bertrand and von Stackelberg
– be able to take a game-theoretic approach to entry
– appreciate the limits of game theory

A ‘Paradigm Shift’ Bringing Industrial
Economics and Managerial Economics
■ The ‘old’ Industrial Economics used the
Structure-Conduct-Performance paradigm
■ dependent variable is the performance/
profitability of a sector
– performance determined by structure; with high entry barriers, high
concentration and high product differentiation profits will be high
– cross- sector multiple regressions the dominant empirical technique
■ behaviour of individual firms (conduct) largely
A ‘Paradigm Shift’ Bringing Industrial
Economics and Managerial Economics
■ Industrial Organization now dominated by
game theory
– focus is on what happens within an oligopolistic
industry, not on differences across industries
– decisions taken by individual firms have become
the centrepiece of analysis
– case studies of firms’ conduct have become the
dominant empirical method

Basic Concepts in Game Theory
■ Warning! Game theory is difficult and can involve highly complex chains of
■ A game is any situation involving interdependence amongst ‘players’
■ Many different types of game:
– co-operative versus non-co-operative (which is the main focus)
– zero-sum, non-zero-sum
– simultaneous or sequential
– one-off versus repeated
• repeated a known number of times, infinite number of times or an unknown but
finite number of times
– continuous versus discrete pay-offs
– complete versus asymmetric information
– Prisoners’ Dilemma, assurance games, chicken games, evolutionary games

Representing and Solving Games
A Simultaneous Game in Strategic Form:
The Payoff Matrix

Company A’s Actions

High Price Low Price
Company B’s High Price 100A,100B 120A, -20B
Actions Low Price -20A,120B 50A,50B

Representing and Solving Games
■ The same game in sequential form
High Price
100A, 100B

Company A
High Price
120A, -20B
Low Price
Company B

High Price
-20A, 120B
Low Price

Company A

50A, 50B
Low Price

Representing and Solving Games
■ Use rollback or backward induction to solve sequential games
– if Company B has set a high price then A chooses a low price: the
‘B high/A high’ branch can be pruned
– if B set a low price then A will choose a low price: the ‘B low/ A
high’ branch can be pruned
– B is therefore choosing between a high price (-20) and a low price
(50): it chooses low price
– A will also choose a low price
■ For simultaneous games search for dominant strategies (ones
which will be preferred whatever the rival does)
– A prefers a low price if B sets a high price and a low price if B sets
a low price: low price is a dominant strategy
– low price is also dominant for B

This Simple Example Illustrates a
Number of Key Ideas
■ Nash Equilibrium - ‘a set of strategies such that each
is best for each player, given that the others are
playing their own equilibrium strategies’
■ Rollback and Dominant strategies
■ The Prisoners’ Dilemma
– an important class of game where players can choose
between co-operating or cheating/defecting
– the best outcome for both is co-operation but the ‘natural’
result is cheating
– collusion is a key example

How to Find Nash Equilibria?
■ 1.Note a distinction between pure strategies - the players choose
one or other ‘move’ with certainty - and mixed strategies - the
players choose ‘high price with a 60% probability and low price
with a 40% probability’. Mixed strategies are too complex to deal
with here: see Dixit and Sneath (1999) for a non-technical
■ For pure strategies
– look for dominant strategies
– if dominant strategies are not to be found, look for dominated
strategies and delete them
– for zero sum games use the minimax criterion: pick the strategy for
each player for which the worst outcome is the least worst
– cell-by-cell inspection: examine every cell and for each one ask
(does either player wish to move from this cell, given what the other
has done?)

For example

N o D om in an t Strategies: E lim in ate D om in ated S trategies

C o m pany A ’s A ctio ns
H ig h P rice M ed iu m P rice L o w P rice
C o m pa ny B ’s H ig h P rice 100A,100B 120A, 65B 60A, 65B
A ctio ns M ed iu m P rice 65 A ,120B 80A,80B 60A, 55B
L o w P rice 65A,60B 55A , 60B 50A,50B

How Many Nash Equilibria?
■ There may be no Nash equilibria in
pure strategies
■ There may be more than one Nash

■ The Prisoner’s Dilemma game, leading to low
price/low price illustrates a problem for firms trying to
collude. But managers will recognise the problem and
try to find ways round it. How can the dilemma be
– Repetition
– Punishments and rewards
– Leadership

Repetition and the Prisoners’ Dilemma
■ If the game is repeated an infinite number of times there are
very clearly advantages in co-operating and players can
observe each others’ behaviour, hence co-operation may be
■ BUT the logic of this depends on the number of rounds of the
game being infinite. If the number of rounds is finite, in the last
round there is no longer any incentive to co-operate. Hence
cheating will take place in the last round and therefore in the
round before….and so on. THE DILEMMA RE-APPEARS
■ COMMONSENSE suggests that solutions are found to this
problem, and a good deal of evidence supports that view,but the
basic insight remains.

Contingent Strategies in Repeated Games
■ Contingent strategies are strategies whereby the actions taken in
repeated games depend upon actions taken by rivals in the last
– grim strategy: co-operate until the rival defects and then defect forever; this is
‘too unforgiving’ - one mistake and the prospect of collusion is gone forever
– tit-for-tat: co-operate when the rival co-operated in the last round, defect if they
defected; this strategy makes permanent cheating unprofitable and out-
performs most others in simulations and experiments
• in terms of the original game, defecting in every round under tit-for-tat leads to profit
of 120,50,50,50,50, whereas not defecting leads to 100,100,100,100. Choice of
defect is only rational at a very high discount rate

Contingent Strategies in Repeated Games
■ If the game is not repeated an infinite number of
times, but there is a probability ‘p’ of another round
the analysis can follow the same logic but adjust the
discount rate so that $1 accruing two periods in the
future is discounted by p/(1+r)2 instead of 1/(1+r)2 .
■ If ‘p’ is relatively low, cheating becomes relatively
more profitable, because the future gains from co-
operation become less.

Penalties and Rewards to Support
■ Introducing penalties and rewards changes the pay-
off structure
■ Firms may ‘punish’ themselves in order to set up
structures which assist collusion
– e.g.’I will match any lower price set by my competitor’ or
looks like a highly competitive move, but it produces a pay-
off structure in which collusion is the outcome
■ Penalties and rewards might come from other sources
- consumers or the law might punish collusion - trade
associations might punish those who defect from the
Leadership to Support Collusion

L e a d e r s h ip a s a S o lu t io n

C o m p a n y A ’s A c t io n s
H ig h P r ic e L o w P r ic e
C o m p a n y B ’s H ig h P r ic e 1 0 0A,3 0 B0 1 2 0A, 1 2B0
A c t io n s L o w P r ic e - 2 0A,2 8 B0 5 0A,5 0B

Leadership to Support Collusion
■ If one firm has much larger pay-offs than the other it
may suit it to charge the higher price even if the rival
charges a lower price - see the example
■ Furthermore, the large firm may increase overall
profits by making side-payments to rivals
■ Saudi Arabia in the oil market?

Cournot and Bertrand Competition

■ From historical curiosities to ‘analytical workhorses’

■ Cournot competition
– two firms, identical products, firms choose output levels
– each firm’s profit-maximising output depends on the other
firm’s output; hence each firm has a reaction function
– as each firm will operate on its reaction function the point
where they cross is the Cournot Nash equilibrium
■ Bertrand competition
– two firms, identical products, firms choose price levels
– price is forced down to marginal cost

Von Stackelberg equilibrium
■ A ‘Leader’ and a ‘Follower’
■ The Leader chooses a price and the Follower then
chooses the price that suits it best, given what the
Leader has done
■ A sequential game and the solution is found by
working backwards. The Leader maximises his profit
(which depends on what the Follower does in
response to what the Leader does) by taking into
account what the Follower will do

Entry Deterrence
■ A major area of application for game
■ “Firms may deter entry by threatening
– in particular firms may build excess
capacity in order to deter entry by
indicating that they will cut price and
increase output if entry takes place
Entry Deterrence
■ But is the threat credible?
■ If it is more profitable for the incumbent to allow entry
and share the market the threat is not credible
■ However, if the incumbent can make commitments
which effectively force him to fight the entrant. For
instance, if the incumbent installs excess capacity so
that his cost per unit rises if market share is given up
to an entrant, he is forced to fight: see Figure 13.7

Airbus and Boeing: No
Government Intervention
B oeing and A irbus Produce a Super-Jum bo:
N o G overnm ent Intervention

A irbus
D evelo p Super- Stay O ut
Ju m bo
B o eing D evelo p Super- -100A,-100B 0A, 500B
Ju m bo
Stay O ut 500A,0B 0A,0B

Airbus and Boeing: No
Government Intervention
B oeing and A irbus Produce a Super-Jum bo:
B oeing isSubsidized by the U S G overnm ent

A irbus
D evelop Super- Stay O ut
Boeing D evelop Super- -100A, 10B 0A, 610B
Jum bo
Stay O ut 500A,0B 0A,0B

How Useful Is Game Theory?
■ A powerful tool, BUT
– outcomes are very sensitive to the protocols
– there may be many equilibria
– when using it to model real life cases there may be many different
options - “ a model may be devised to fit almost any fact” Saloner
– analysis works backwards - instead of theory to hypotheses to
empirical data - empirical data to theory
– sometimes players’ commonsense tells them what to do despite
multiple equilibria
– the requirement that firms do as expected (in rollback, for instance)
– where do the protocols come from, how do they change? GE and
Westinghouse found new protocols which helped them to collude -
why then and not before?

How Useful Is Game Theory?
■ An overall judgment?
■ Some very useful insights
– Nash equilibrium concept
– collusion; the price matching result
– entry deterrence; the importance of credibility
– Cournot and Bertrand provide determinate solutions for
■ The degree of complexity involved may limit its
usefulness as a predictive tool
■ The degree of rationality which has to be assumed on
the part of players is uncomfortable