ide 6r
TRANSCRIPT
-
8/3/2019 IDE 6r
1/9
CHAPTER 6: OLIGOPOLY
1. Cover first single-period and multi-period games
2. Analyze standard oligopoly models: Cournot, Bertrand, and Stackelberg
3. The evidence
GAME THEORY: Formal models to analyze conflict (or competition) and cooperation (or collusion). It
is the study of how interdependent decision makers make choices.
Strategic Games: Describes interactions between mutually aware (and usually rational) players.
Why study game theory?
A deeper understanding of Oligopoly may conclude that the S-C-P model may not be a good guide for
public policy. For example, in the Bertrand price competition model, two firms competing is sufficient to
yield a perfectly competitive outcome of P = MC.
Although, game theory also suggests that cooperative behavior is easier to facilitate with more
concentration.
There are two broad types of games we will be looking at:
1. Static: Simultaneous Move Games
2. Dynamic: Sequential Move Games
SIMULTANEOUS MOVE GAMES
Use Game table or Payoff table or Matrix table. May be called the normal form, strategic form, ormatrix form of the game. We will primarily examine the following:
Prisoners Dilemma
Simultaneous, One-Move
Repeated
i) Finite Repetition
ii) Infinite Repetition
SEQUENTIAL MOVE GAMES
Use Game Trees. Tree is called the extensive form or game tree form
-
8/3/2019 IDE 6r
2/9
SIMULTANEOUS MOVE GAMES
The starting point for solving simultaneous-move games for the equilibrium is to look for dominant
strategies.
Dominant Strategy: Strategy that gives a player higher payoffs against every opposing strategy.
In a dominant strategy equilibrium, each player chooses an action that is the best response against any
action the other might take.
A more general situation for an equilibrium is a Nash Equilibrium. In a Nash Equilibrium each player is
doing the best it can, given the strategies of the other players. (To change is to be worse off)
Games with dominant strategies yield a Nash Equilibrium. However, a Nash Equilibrium can emerge
without dominant strategies.
(Note: Nash is a nobel prize winner in economics and is the main character in the movie Its a Beautiful
Mind)
PRISONERS DILEMMA
This is the classic game in game theory. It is a simultaneous-move, one-shot game with imperfect
information (do not know what the other player will do)
The characteristics and outcomes of the PD are:
1. Each player has two strategies
Cooperate or collude
Defect, cheat, or compete
2. Each player has a dominant strategy (defect from cooperation)
3. Equilibrium for both players is worse than cooperative outcome
4. Conflict between individual incentives & joint incentives
Importance:
I. Wide ranging applicability
II. Equilibrium outcome is a bad outcome for individual players
Issues:
In Adam Smith, perfect competition world, the dominant strategy is pursuing self-interest
(greed), which yields the best outcome for society
Players can achieve a better outcome for themselves by committing conspiracies againstsociety
It takes only two players to achieve the (desirable) competitive outcome
Note:
Repeated PD games may yield a different outcome than one-shot PD games. A reputation or signals may
be established to encourage or facilitate the cooperative outcome. For example the reputation or signal
might be to:
Punish defection
Reward cooperation
-
8/3/2019 IDE 6r
3/9
Consider the examples below
The first case is the classic PD game, where both players choose their dominant strategy Confess
Mumbles Action
Confess Dont Confess
Big Boys Confess 6 years, 6 years 1 year, 10 yearsAction Dont Confess 10 years, 1 year 3 years, 3 years
If someone can signal an alternative outcome for confessing, the outcome of the game can be altered.
Mumbles Action
Confess Dont Confess
Big Boys Confess death, death death, 10 years
Action Dont Confess 10 years, death 3 years, 3 years
For the signal to be credible, a reputation would need to be established which requires Repeated PD
Games.
PRISONERS DILEMMA REPEATED GAMES
Series of simple one-period simultaneous-move games
Because players can react to other players past actions, repeated games allow for equilibrium outcomes
that would not emerge in one-shot games. (Reputation Effects) There are two cases:
Finite:
In a finite game, it is not rational for any player to deviate from the dominant strategy in any period.
This can be seen by working backward. In the last period, with no future to worry about, it is rational to
play the dominant strategy. If it is known that everyone will play the dominant strategy in the lastperiod, then it is rational to play it in the next-to-the-last period. Repeating this logic, rational players
should play the dominant strategy in each period.
Chain-store paradox: Paradox is that despite the conclusion that predatory pricing is irrational to deter
entry in finite PD games, many firms are commonly perceived as engaging in predatory pricing. The
paradox is resolved by introducing uncertainty. If potential entrants lack certainty, then price cutting by
incumbents may alter entrants expectations about the incumbents future pricing strategy.
Infinite:
In infinite repeated games, dynamic pricing rivalry emerges and reputations can be established.
Cooperative (monopoly) pricing may be sustainable as an equilibrium, even though firms are makingdecisions non-cooperatively (w/o explicit collusion).
High market concentration facilitates cooperative pricing (cheaters are more easily detected)
A tit-for-tat strategy, in which a firm is prepared to match whatever strategy a competitor makes, is one
of the more compelling strategies that may lead to cooperative pricing.
-
8/3/2019 IDE 6r
4/9
SEQUENTIAL GAMES
Sequence of decisions made over time.
These games utilize a game tree and are solved using the fold-back method or backward induction.
Consider the game in the handout.
Beta
Do not Expand Small Expand
Do not Expand $18, $18 15, 20 9, 18Alpha Small 20, 15 16, 16 8, 12
Large 18, 9 12, 8 0, 0
In the simultaneous-move game the Nash Equilibrium is
SMALL, SMALL.
Now suppose that the game is sequential and Alpha chooses first. Seek the subgame perfect Nash
Equilibrium. Subgames are a game within a game.
Game tree for capacity game.
,
Do Not Expand 18, 18Do Not Expand Beta Small 15, 20
Large 9, 18
Do Not Expand 20, 15
Alpha Small Beta Small 16, 16
Large 8, 12
Do Not Expand 18, 9
Large Beta Small 12, 8
Large 0, 0
Alpha chooses first, but decides on the basis of Betas expected choices.
If Alpha goes Beta maximizes payoff going Alpha payoff
Do Not Expand Small (Beta payoff $20) $15
Small Small (Beta payoff $16) $16
Large Do Not Expand (Beta payoff $9) $18
Given Betas choices, Alpha will maximize payoff by moving first and going LARGE. Beta does not
expand.
-
8/3/2019 IDE 6r
5/9
Standard Models of Oligopoly Theory
(Single-period models of non-cooperative Oligopoly)
Single Period Static Models:
C: Cournot Duopoly Model (1838)
B: Bertrand Model (1883)
S: Stackelberg Model (1934)
Common Assumptions:
1. Number of firms is given & does not change
2. Produce a homogenous product (relaxed in B)
3. Firms have the same (constant) MC
Differences:
C, S: choice variable is quantity
B: choice variable is price
C, B: firms act simultaneously
S: interaction is sequential
Cournot Duopoly (2-firm) Model
Each firm chooses quantity given the other firm maintains their current output. [Output maintenance
assumption]
Market Demand: P = 30 ( q1 + q2 )
where q1 and q2 are perfect substitutes
Cost: AC = MC = $6
Monopoly Solution: Produce where MR = MC
MR = 30 2 Q = $6 = MC
Q = 12 P = $18 Profit = TR TC = 216 72 = 144
Competitive Solution: P = AC = MC = $6
Q = 24 P = $6 Profit = 0
Cournot Duopoly Solution:
Firm 1 faces the (inverse) demand curve (with choice variable q1)
P = ( 30 q2 ) q1
[Note typo in text on page 161, eq. (6.2)]
The demand curve is a residual demand (the market demand - q2 ). This firm will want to maximize
profit where MR1 = MC
[MR1 is the residual marginal revenue curve]
MR1 = (d R1 / d q1 ) = ( 30 q2 ) 2 q1 and MC = 6
-
8/3/2019 IDE 6r
6/9
Solve MR1 = MC for q1 taking q2 as given. This will yield Firm 1s reaction function or best-response
function.
( 30 q2 ) 2 q1 = 6
q1 = 12 ( q2 / 2 )
By symmetry, firm 2s reaction function is
q2 = 12 ( q1 / 2 )
In equilibrium, each firm makes -max decision, anticipating a -max decision of all competitors.
Called the Cournot-Nash Equilibrium. It is a Nash equilibrium since each is doing the best given the
others choice.
The equilibrium must satisfy all reactions functions simultaneously.
q2 = 12 ( q1 / 2 ) = 12 [6 (q2 /4)]
Solving for q2 yields
q2 = q1 = 8 and P = $14
andProfit = $64 for each firm
Monopoly Duopoly
Perfect Competition
(Social Optimum)
Price $18 $14 $6
Industry Output 12 16 24
Industry Profit $144 $128 0
Generalization: As the number of firms increases, quantity competition equilibrium approaches the
perfectly competitive outcome. This supports the S-C-P model (see Table 6.2)
-
8/3/2019 IDE 6r
7/9
Interpretation
The Cournot adjustment process shown above leads to the equilibrium where the two reaction functions
intersect. This suggests a sequential or dynamic process w/o learning.
A better interpretation is to consider this a Nash equilibrium game as a simultaneous output choice,
single-period model where all decision makers are rational (optimize) and strategic (consider rivalsactions).
But How do firms form their beliefs?
Problems with Cournot Model:
1. Single-period, static model is not realistic
2. The Cournot-Nash equilibrium is not a dominant strategy equilibrium. The output maintenance
assumption leads to this outcome.
3. The output maintenance assumption is nave. With learning and experience the players could
choose:
Monopoly outcome: (q1 = q2 = 6) and industry profit is maximized, or
Stackelberg outcome: Firm 1 commits to q1 = 12 and maximizes individual profit at $72 >
$64. Firm 2 passively reacts with q2 = 6 and profit = $36 < $64.
4. No basis for the underlying beliefs of the players
5. These criticisms have led to multiperiod models
q2
q1
Firm 1 Reaction Function
Firm 2 Reaction Function
-
8/3/2019 IDE 6r
8/9
BERTRAND MODEL OF PRICE COMPETITION
(firms set prices rather than output)
Features:
1. Duopolists sell homogeneous good at identical and constant MC
2. Consumers are astute and always purchase at the lower price
3. Low price firm gains the entire market. High price firm sells nothing
4. Each firm believes its rivals price is fixed
Consider the Prisoners Dilemma as a Price War
Firm 2
High Price (HP) Low Price (LP)
Firm High Price (HP) 10, 10 5, 12
1 Low Price (LP) 12, 5 7, 7
Now add the Bertrand assumption that low price always takes the entire market
Firm 2High Price (HP) Lower Price
Firm (HP) 10, 10 - FC, 15
1 Lower Price 15, - FC 0, 0
Each firm will want to undercut the rival to avoid losing all sales. In the limit each firm will price at MC
with profit of zero.
The Bertrand model has the extreme result that price competition by as few as two firms, yields the
perfectly competitive outcome. Lower price strategy always dominates, resulting in MC pricing.
Caveat:
1. Firms need unlimited capacity (there is no equilibrium with capacity constraints no one can
capture the entire market)
2. Must have homogeneous product
3. A differentiated product with soften the price competition
STACKELBERG MODEL
(sequential leader-follower model)
Go from a one-period simultaneous-move model to a two-stage sequential game (still considered static).Firms set output (still quantity competition like Cournot) but one firm acts as the leader and chooses
before the others (first-mover). The others choose according to their Cournot reaction function.
If Firm 1 chooses first, they would select the best point for them on Firm 2s reaction function.
-
8/3/2019 IDE 6r
9/9
Compared to the Cournot solution, the Stackelberg solution is for the leader to increase output and the
follower to produce less. Total market output increases and market price falls. Firm 2 passively follows
with a smaller output and smaller profit while the leader gains at its expense. However, the output is
below the competitive output and price is above competitive price.
Strategic and credible commitment by Firm 1 gives it an advantage when other firms follow passively.
q2
q1
Firm 1 Reaction Function
Firm 2 Reaction Function
12
Firm 1 selects q1
= 12, Firm 2 selects q2
= 6