ECO290E: Game Theory

Lecture 3: Why and How is Nash Equilibrium Reached?

1 Lecture 3 2013 Winter

Review of Lecture 2

Each (normal-form) game consists of 1) players, 2)
strategies, and 3) payoffs.

Any two players game with finite strategies can be
expressed by a payoff (bi-) matrix.

In a Nash equilibrium, no one has an incentive to change
her strategy unilaterally.

A Nash equilibrium is not necessary Pareto efficient.

Ex) Prisoner’s Dilemma

There can be multiple Nash equilibria.

Ex) Coordination game

One equilibrium can be less efficient than (Pareto dominated
by) the other equilibrium.

Called “coordination failure.”

2 Lecture 3 2013 Winter

Four Reasons for NE
There are at least four reasons why we can expect Nash
equilibrium (NE) would realize.
1. By rational reasoning
2. Being an outstanding choice
3. A result of discussion
4. A limit of some adjustment process

⇒ Which factor serves as a main reason to achieve NE

depends on situations.

3 Lecture 3 2013 Winter

1. Rationality

Players can reach Nash equilibrium only by rational
reasoning in some games, e.g., Prisoners’ dilemma.

However, rationality alone is often insufficient to lead

to NE. (see Battle of the sexes, Chicken game, etc.)

A correct belief about players’ future strategies

combined with rationality is enough to achieve NE.

⇒ 2 - 4 help players to share a correct belief.

4 Lecture 3 2013 Winter

2. Focal Point

A correct belief may be shared by players only from
individual guess.

Class room experiments:

Choose a subway station in Tokyo and write down its name.

You will win if you can choose the most popular answer.

Most of the students are expected to write “xxx”.

Like this experiment, there may exist a Nash equilibrium

which stands out from the other equilibria by some
reason.

⇒ Focal Point (by Thomas Schelling)

5 Lecture 3 2013 Winter

3. Self-Enforcing Agreement

Without any prize or punishment, verbal promise
achieves NE while non equilibrium play cannot be
enforced.
⇒ NE = Self-Enforcing Agreement

Example: Prisoner’s Dilemma

Even if both players promise to choose “Silent,” it will not
be enforced since (S,S) is not a NE.

(C,C) is self-enforced (although there is no point of making
such a verbal promise…).

6 Lecture 3 2013 Winter

4. Repeated Play

Through repeated play of games, experience can
generate a common belief among players.

Example 1: Escalator

Either standing “right” or “left” can be a NE.

Example 2: Keyboard

“Qwerty” vs. “Dvorak”
⇒ History of adjustment processes determine which
equilibrium is realized: Economic history has an
important role.
⇒ “Path Dependence” (by Paul David)

7 Lecture 3 2013 Winter

Roles of Social Science

Explain the reason behind the frequently observed
phenomena.
⇒ NE serves as a powerful tool.

Predict what will happen in the future.

⇒ Although it is usually difficult to make a one-shot
prediction, NE may succeed to predict the stable situation
after some (long) history of adjustment processes.

8 Lecture 3 2013 Winter

What is Rationality?

A player is rational if she chooses the strategy
which maximizes her payoff (given other players’
strategies.)

Implicitly assume that we can describe her
preferences as payoff numbers.

The definition implies that a rational player

takes a dominant strategy whenever it is available.

never takes (strictly) dominated strategies.

If strategy s generates always higher payoff than s’,
then s’ is called (strictly) dominated by s.
9 Lecture 3 2013 Winter
Solving Game by Rationality

Player 2 Left Middle Right

Player 1
Up 0 2 1
1 1 0

Down 3 1 0
0 0 2

10 Lecture 3 2013 Winter

Iterated Elimination Process

What can players deduce if they completely understand (i.e.,
agree) that they are both rational players?

Neither of them will take strictly dominated strategies…

Step 1: Right is strictly dominated by Middle, so player 2 NEVER
takes Right.

Step 2: Given the belief that Player 2 never takes Right, Down is
strictly dominated by Up.Therefore, Player 1 will NOT take Down.

Step 3: Given the belief that Player 1 will not take Down, Left is
strictly dominated by Middle. Therefore, Player 1 will NOT take
Left.

Step 4: Only (Up, Middle) is survived after the iterated elimination
process!
⇒ This reasonable solution coincides with NE.

11 Lecture 3 2013 Winter

Limitation of Rationality

The process often produces a very imprecise
prediction about the play of the game.

Coordination, Battle of the sexes, Chicken, etc.

Nash equilibrium is a stronger solution concept than

iterated elimination of strictly dominated strategies, in
the sense that the players’ strategies in a Nash
equilibrium ALWAYS survive during the process, but
the converse is not true.
⇒ If the elimination process picks up a unique strategy profile
(e.g., previous example), then this MUST be a (unique) NE.

12 Lecture 3 2013 Winter

Is it Solved by Rationality?

Player 2 Left Middle Right

Player 1
Up 0 1 3
1 1 1

Down 3 1 0
0 2 0

13 Lecture 3 2013 Winter

Randomized Strategies

No strategy looks to be dominated…

If a player 2 randomizes L and R with 50% each, then

Such mixed (randomized) strategy yields 1.5 (as an expected
payoff) while M gives 1 irrespective of player 1’s strategy.

Therefore, M is eliminated by mixing L and R.

After eliminating M, we can further eliminate D (step 2) and L
(step 3), eventually picks up (U, R) as a unique outcome.

We essentially extend the concept of strategy, i.e.,

allowing randomizing over different (pure) strategies.

This type of strategy is called “mixed strategy”.

14 Lecture 3 2013 Winter

Spatial Competition Model

Players:Two ice cream shops

Strategies: Shop location along a beach (any integer
between 0 and 100)

Payoffs: Profits = The number of customers

Assumptions:

Customers are located uniformly on the beach.

Each customer goes to the nearest shop (and buys
exactly one ice dream).

If both shops choose the same location, each receives
half of the customers.
15 Lecture 3 2013 Winter
Nash Equilibrium

There is a unique NE in which both shops open at
the middle, i.e., s1 = s2 = 50.

Called “Principle of minimum differentiation”.

Why?
1. Choosing separate locations never becomes a NE.
2. Choosing the same locations other than the middle
point also fails to be a NE.
3. If both shops choose the middle, then no one has an
incentive to change the location.

16 Lecture 3 2013 Winter

Applications of Spatial Competition

Spatial competition (Hotelling) Model can explain

Two players competing in the same platform

Aiming for as much customers each could obtain

Results in taking very similar strategy one another…

Players Strategies Phenomena

Two Big Political Parties Target Policy Moderate Politics

Convenience Stores Location Two Stores Locate Nearby

TV Programs Broadcasting Time Same Topics on same time for

Different Channels
Manufacturers Taste or Appearance Similar Products (Coke, Audio
Players, Smartphone, etc)

17 Lecture 3 2013 Winter

Is There a Dominant Strategy?

Q: Is 50 a dominant strategy?

A: No!

If 50 is a dominant strategy, then payoff of choosing 50 must
be strictly larger than the payoff of choosing ANY other
strategy, for EVERY possible strategies of the other player.

Suppose other player’s strategy is 90. Then, choosing 50
yields payoff of 70 (= (50+90)/2), while choosing 80 will
achieve 85 (= (80+90)/2)

⇒ Choosing 50 is NOT always optimal.

18 Lecture 3 2013 Winter

Solution by Iterated Elimination

Step 1:A rational player never takes the edges, since 0
(100) is strictly dominated by 1 (99).

Step 2: 1 and 99 are never chosen if the players know
their rival is rational.

Step 3: 2 and 98 are never chosen if the players know
that their rival knows that you are rational.

Step 50: 49 and 51 are never chosen if the players
know that their rival knows that …
⇒ Both players choose 50 in the end!

19 Lecture 3 2013 Winter

Further Exercises

Find a real life example that can be described as a game
with multiple Nash equilibria in which focal point story
picks up a unique equilibrium.

List up real life stories of path dependence.

Consult an advanced textbook and study the idea of
“rationalizability” (each player only chooses a strategy
that can be a best response to opponents’ possible
strategies) and its connection to the iterated elimination
of strictly dominated strategies.

20 Lecture 3 2013 Winter