Chapter 13 Game Theory1

 Static Games
 Dynamic Games

An Overview of Game Theory

• Game theory is a set of tools used by economists and many others to analyze players'
strategic decision making.

• A game is a situation in which players (participants) make strategic decisions that take
into account each other's actions and responses. A game is an interaction between players
(such as individuals or firms) in which players use strategies.

• We concentrate on how oligopolistic firms behave within a game.

Oligopoly: a small group of firms in a market with substantial barriers to entry. In such
markets, the firms know that their actions significantly affect each other's profit, so
their actions depend on how they think their rivals will act.

Example: A normal-form game that specifies the players, their strategies, and the payoffs
for each combination of strategies.

Firm B

Advertise Don't advertise

10 (profit of A), 5 15, 0

Advertise
(profit of B)
Firm A
Don’t advertise 6, 8 10, 2

Definitions:

1
The teaching notes should only be used by students in Dr. Haozhen Zhang’s class of ECON2020.

1
• The payoffs of a game are the players' valuation of the outcome of the game (e.g., profits
for firms, utilities for individuals).
• The rules of the game determine the timing of players' moves and the actions players can
make at each move.
• An action is a move that a player makes at a specified stage of a game.
• A strategy: A rule or plan of action for playing a game. It is a plan that specifies the action
that a player will make based on the information available at each move and for any
possible contingency. For example, A will advertise (not advertise) if B advertises.
• Optimal strategy: Strategy that maximizes a player's expected payoff.
• Dominant strategy Strategy that is optimal no matter what an opponent does.
Advertising is a dominant strategy for Firm A. The same is true for Firm B: No matter what
firm A does, Firm B does best by advertising.
• Strategic interdependence occurs when a player's optimal strategy depends on the actions
of others.

Assumptions:
• All players are interested in maximizing their payoffs
• All players have common knowledge about the rules of the game
• Each player's payoff depends on actions taken by all players
• Complete information (payoff function is common knowledge among all players).
However, it is not perfect information (player knows the full history of the game up to the
point he is about to move). Because the move is simultaneous, neither firm knows what
action its rival will take when it makes its own decision. We will relax this assumption
when we examine sequential games.

We will examine both static and dynamic games.

2
13.1 Static Games
• In a static game, each player acts simultaneously, only once and has complete
information about the payoff functions but imperfect information about rivals' moves.

• Examples: employer negotiations with a potential new employee, Farm Boy' choice
of locations; Competition between Air Canada and Westjet on the Toronto-
Montreal route

• Consider a normal-form static game of complete information which specifies the players,
their strategies, and the payoffs for each combination of strategies.

Example: Quantity-Setting Game

• Quantities, q, are in thousands of passengers per quarter; profits are in millions of dollars
per quarter

Copyright Person Canada

Predicting a Game's Outcome

3
• We can precisely predict the outcome of any game in which every player has a dominant
strategy.
• Dominant strategy: A strategy that produces a higher payoff than any other strategy for
every possible combination of its rivals' strategies. That is, I'm doing the best I can no
matter what you do. You're doing the best you can, no matter what I do.
• Rational players will always choose the dominate strategy and avoid strategies that are
dominated by other strategies.
• Airline Game:
• If United chooses high-output (qu=64), American's high-output strategy maximizes
its profits
• If United chooses low-output (qu=48),, American's high-output strategy still
maximizes its profits.
• For American, high-output is a dominant strategy.
• The strategies of United are the same as those of United.
• The high-output strategy is dominant for American and for United. This is a dominant
strategy equilibrium. But note that not all games have dominant strategies.

Copyright Person Canada

 Players choose strategies that don't maximize joint profits. Called a prisoners' dilemma
game, all players have dominant strategies that lead to a profit that is less than if they
cooperated.

Nash Equilibrium

4
• The best response is a strategy that maximizes a player's payoff, given its beliefs about its
rivals' strategies.
• A set of strategies is a Nash equilibrium if, when all players use these strategies, no player
can obtain a higher payoff by choosing a different strategy. That is, each player is doing
the best he/she can, and no player has an incentive to deviate from a Nash equilibrium.
• Nash Equilibrium: I'm doing the best I can given what you are doing. You're doing the best
you can, given what I am doing.
• Every game has at least one Nash equilibrium, and every dominant strategy equilibrium is
a Nash equilibrium, too. In our class, we only examine the pure strategy Nash equilibrium
in which each player chooses a single action.
• In a game without dominant strategies, calculate best responses to determine Nash
equilibrium

Example qa=qb=64 is the NE

5
Questions: 1. Is there a dominant strategy for American or United?
• Answer: NO. (Dominant strategy: A strategy that produces a higher payoff than any other
strategy for every possible combination of its rivals' strategies.)

2. What are N.Es?

A two-step procedure.
1. determine each firm's best response to any given strategy of the other firm.
i.e.
 For American
o If U chooses qu=96, A's best response is qa=48. Strategy (qu=96, qa=48)
o If U chooses qu=64, A's best response is qa=64. Strategy (qu=64, qa=64)
o If U chooses qu=48, A's best response is qa=64. Strategy (qu=48, qa=64)
 For United
o If A chooses qa=96, U's best response is qu=48. Strategy (qu=48, qa=96)
o If A chooses qa=64, U's best response is qu=64. Strategy (qu=64, qa=64)
o If A chooses qa=48, U's best response is qu=64. Strategy (qu=64, qa=48)

2. Are there any pairs of strategies (a cell in the profit table) that are the best responses for
both firms?
We can see the only when (qu=64, qa=64), neither firms want to deviate from its strategy.

Multiple Equilibria

6
 • Many oligopoly games have more than one Nash equilibrium

1. determine each firm's best response to any given strategy of the other firm.
 the best response of firm 2

If firm 1 chooses "Do not enter", firm 2 choose "enter"

If firm 1 chooses "enter", firm 2 choose "do not enter"

 the best response of firm 1

If firm 2 chooses "do not enter", firm 1 choose "enter"

If firm 2 chooses "enter", firm 1 choose "do not enter"

2. Are there any pairs of strategies (a cell in the profit table) that are the best responses for
both firms?

Yes, (firm 1 chooses "enter", firm 2 choose "do not enter") and (firm 1 chooses "do not
enter", firm 2 choose "enter")

In both Nash equilibriums, neither firms want to deviate from the strategies.

It is difficult to see how the firms choose strategies unless they collude and can enforce their
agreement.

Note: in this example, firms do not have dominant strategies. In these cases, we have to
calculate best responses to determine Nash equilibrium
More examples:
1. Failure to Maximize Joint Payoffs

7
In the Nash equilibrium of the first advertising game above, firms do not maximize joint profits.
In the second, they do.

1. determine each firm's best response to any given strategy of the other firm.
 the best response of firm 2

If firm 1 chooses "Do not Advertise", firm 2 choose "advertise"

If firm 1 chooses "advertise", firm 2 choose "advertise"

 the best response of firm 1

If firm 2 chooses "Do not Advertise", firm 1 choose "advertise"

If firm 2 chooses "advertise", firm 1 choose "advertise"

2. Are there any pairs of strategies (a cell in the profit table) that are the best responses for
both firms?

Yes, (firm 1 chooses "advertise", firm 2 choose "advertise")

In the Nash equilibrium of the first advertising game, firms do not maximize joint profits.

Another way to find a NE for this game: "advertise" is a dominant strategy for both firms, so
NE is (firm 1 chooses "advertise", firm 2 choose "advertise").
Every dominant strategy equilibrium is a Nash equilibrium. However, some games do not have
dominant strategies. In these cases, calculate best responses to determine Nash equilibrium

Second game:

8
1. Determine each firm's best response to any given strategy of the other firm.
 the best response of firm 2

If firm 1 chooses "Do not Advertise", firm 2 choose "advertise"

If firm 1 chooses "advertise", firm 2 choose "advertise"

 the best response of firm 1

If firm 2 chooses "Do not Advertise", firm 1 choose "advertise"

If firm 2 chooses "advertise", firm 1 choose "advertise"

2. Are there any pairs of strategies (a cell in the profit table) that are the best responses for
both firms?

Yes, (firm 1 chooses "advertise", firm 2 choose "advertise")

In the Nash equilibrium of the second advertising game, firms maximize joint profits.

Another way to find a NE for this game: "advertise" is a dominant strategy for both firms, so
NE is (firm 1 chooses "advertise", firm 2 choose "advertise").

13.2 Dynamic Games

9
In dynamic games:
Assumptions:
• players move either sequentially or repeatedly
• players have complete information about payoff functions
• at each move, players have perfect information about previous moves of all
players

Sequential Games
In a Stackelberg Airlines game, American is the leader firm that chooses its output level
first. Given American's choice, United, the follower, picks an output level. A normal-form
game is as follows:

We use an extensive-form diagram or a game tree to show

• the number of players (N)
• the sequence of their moves
• the actions they can take at each move
• the payoff function over all possible strategies.

10
This Stackelberg game tree shows
• decision nodes(boxes with players): indicates which player's turn it is
• branches (lines extending out or the boxes): indicates all possible actions available
• subgames: subsequent decisions available given previous actions

Subgame Perfect Nash Equilibrium

• To predict the outcome of the Stackelberg game, we use a strong version of Nash
equilibrium.
• A set of strategies forms a subgame perfect Nash equilibrium if the players' strategies are
a Nash equilibrium in every subgame. The subgame perfect Nash equilibrium is also a
Nash Equilibrium.
• This game has four subgames; three subgames at the second stage where United
makes a decision and an additional subgame at the time of the first-stage decision.
• We can solve for the subgame perfect Nash equilibrium using backward induction.

Backward Induction in Airline Game

• Backward induction is where we determine:
• the best response by the last player to move
• the best response for the player who made the next-to-last move
• repeat the process until we reach the beginning of the game

11
• Airline Game
• If American chooses 48, United selects 64 to maximize its profit at 5.1, then
American's profit=3.8
• If American chooses 64, United selects 64 to maximize its profit at 4.1, American's
profit=4.1
• If American chooses 96, United selects 48 to maximize its profit at 2.3, American's
profit=4.6
• Thus, American chooses 96 in the first stage.
• Then American chooses 96 and United selects 48 are the action that firms choose in the
SPNE.

Dynamic entry game.

Copyright Person Canada

Backward induction:
1. Determine the potential rival's response in the second stage to each possible action taken
by the Incumbent in the first stage.
 If the Incumbent does not invest, its rival enters because 4>0, and the Incumbent's
profit = 4
 If the Incumbent invests, its rival does not enter because 0>-1, and the
Incumbent's profit = 8
2. Determine the Incumbent's decision, given its potential rival's response.
 The Incumbent will invest because 8>4, Incumbent chooses "investment" and
entrant chooses "do not enter" is the SPNE

12
Note that if there is no threat of entry, the Incumbent will not pay for the investment. This is
an example that competition could increase productivity.
Repeated game:
In a repeated game, a firm can influence its rival's behavior by signaling and threatening to
punish.
For example:

Example. Repeated Airline Game

Copyright Person Canada

One period N.E. is (qa=64, qu=64), which does not maximize joint profits.

One airline firm could use a low-quantity strategy for a couple of periods to signal to the other
firm its desire that the two firms cooperate and produce that low quantity in the future.

The airline can threaten to punish a rival for not restricting output.

American could announce or somewhat indicate to United that it will follow the following
strategy:
• American will produce a smaller quantity each period as long as United does the same. So
both will have profits of 4.6
• If United produces the larger quantity in period t, American will produce the larger
quantity in period t + 1 and all subsequent periods. So both will have profits of 4.1.

If United believes that American will follow this strategy, United knows that

13
 1. If it produces lower quantity, it will make $4.6 million each period;
2. if it produces larger quantity at t, it will make a higher profit of $5.1 million at t but
a lower profit of $4.1 million in each following period (t+1, t+2...).

United's best policy or action: if it values future profits nearly as much as current profits, it
produce the lower quantity in each period because one-period gain of deviating from the
collusive output level will not compensate for the reduced profits in future periods.

Note:
Cooperation is more likely in a game that will continue forever or one that will end at an
uncertain time. If firms repeatedly play the same game indefinitely, they should find it
easier to collude.

For a firm's announced strategy to be a credible threat, rivals must believe that the firm's
strategy is rational in the sense that it is in the firm's best interest to use it.

For example, if the firms know that they are going to play the game for T periods. They
will both cheat in the last period. As a result, the last period is like a single-period game.

14