Professional Documents
Culture Documents
Game Theory 19.09.2021
Game Theory 19.09.2021
Why “pure”
It is deterministic set. No probabilities are attached or
randomisation is done.
The normal form
1. The finite set of players, N={ 1,2, ,n}
2. A collection of set of pure strategies, (𝑆1 ,
𝑆2 , … … … … 𝑆𝑛 )
3. A set of payoff function
{𝑢1 , 𝑢2 , … … … … 𝑢𝑁 )
each assuming a payoff value to each combination of
chosen strategies.
Examples
Prisoner's dilemma
Two prisoners. They are being question by the police in
different rooms. Each can confess or not
Prisoner B
Prisoner A Not Confess
Confess
Not -1,-1 -9,0
Confess
Confess 0,-9 -6,-6
Rock-Paper- scissor
Player B
Rock Paper Scissor
T -1, 1 1, -1
TailT
2
Stag Hare
1
2, 2 0, 1
A
Stag
Player A
B 1, 0 0, 0
Hare
0, 0 1, 1
B
ovie Opera 2, 1 0, 0
Movie 0, 0 1, 2
D 1, 3 2, 2
D
D: dominant pig
S: submissive pig
Big John and Little John
Big John and Little John eat coconuts, which dangle from a
lofty branch of a palm tree. Their favourite coconut palm
produces only one fruit per tree.
To get the coconut, at least one of them must climb the tree
and knock the coconut loose so that it falls to the ground.
Energy measurements show that a coconut is worth 10 Kc
(kilocalories) of energy, the cost of running up the tree,
shaking the coconut loose, and running back down to the
ground costs 2 Kc for Big John, but is negligible for Little
John, who is much smaller.
Moreover, if both individuals climb the tree, shake the
coconut loose, then climb down the tree and eat the
coconut, Big John gets 7 Kc and Little John gets only 3 Kc,
because Big John hogs most of it; if only Big John climbs the
tree, while Little John waits on the ground for the coconut
to fall, Big John gets 6 Kc and Little John gets 4 Kc (Little
John eats some before Big John gets back down from the
tree); if only Little John climbs the tree, Big John gets 9 Kc
and Little John gets 1 Kc (most of the food is gone by the
time Little John gets there).
Three person Games
Suppose each person chooses a number between 1. 2, 3. The payoff
is 4 times minimum number announced minus his own number.
B B B
A 1
1
3,3,3
2
3,2,3
3
3,1,3 1
1
3,3,2
2
3,2,2
3
3,1,2 1
1
3,3,1
2
3,2,1
3
3,1,1
A A
2 2,3,3 2,2,3 2,1,3 2 2,3,2 6,6,6 6,5,6 2 2,3,1 6,6,5 6,5,5
3 1,3,3 1,2,3 1,1,3 3 1,3,2 5, 6,6 5,5,6 3 1,3,1 5, 6,5 9,9,9
A P -100, -1000,
-100
5, -5, 5 P -100, -1000, -5 5, -10,
-10
C chooses P C chooses Y
Solution Concepts
Assumptions and set up”
1. Players are rational: A rational player is one who chooses
his action 𝑠𝑖 ∈ 𝑆𝑖 to maximise his payoff consistent with his
belief about what is going on in the game.
2. Players are intelligent: Knows every thing about the game:
the actions, outcomes,
3. Common knowledge
4. Self-enforcement:
In non-cooperative game theory a solution must be self
enforcing. The player should not have any incentive for
unilateral deviation.
Dominance in Pure Strategy
What is a dominated strategies:
◦ Let 𝑠 = (𝑠1 , 𝑠2 , … … … … 𝑠𝑁 )be a strategy profile
◦ s-i := (s1, …, si-1, si+1, … sn)
◦ s’=(s-i,s’i) := (s1, …, si-1 ,s’i, si+1, … sn)
Strictly dominated strategy:
◦ An strategy sj* Sj is strictly dominated if there exists any other strategy sj’
such that for all strategy profiles s S:
uj(s-j, sj’) > uj(s-j, sj*) for all s-j S-j
Of course, it is not rational to play strictly dominated strategies
Examples
Prisoner B
Not Confess
Confess
Prisoner A
Not -1,-1 -9,0
Confess
Confess 0,-9 -6,-6
Iterated Elimination of
Strictly Dominated pure Strategies
Since strictly dominated strategies will never be played,
one can eliminate them from the game
This can be done iteratively
If this converges to a single strategy profile, the result is
unique
This can be regarded as the result of the game, because it
is the only rational outcome
In any case, if by iterated elimination of dominated
strategies there is only one strategy left for each
player, the game is called a dominance-solvable game.
Player B
L C R
U 4,3 5,1 6,2
Player A
M 2,1 8,4 3,6
D 3,0 9,6 2,8
Does Player 1 have a dominated strategy?
Hint: Compare b and d.
Iterated Elimination of Dominated Strategies-Stage 1
Player 2
x y z
Now x and y are dominated and eliminated. Next for player 1 ‘d’is the strategy he will
choose when Player 2 plays z.
Weakly Dominated Strategies
Instead of strict domination, we can also go for weak
domination:
◦ An strategy sj* Sj is weakly dominated if there exists a strategy
sj’ such that for all strategy profiles s S:
uj(s-j, sj’) ≥ uj(s-j, sj*)
and for at least one profile a A:
uj(s-j, sj’) > uj(s-j, sj*).
When there is a weakly dominant strategy, if the
player is rational and cautious, then he will play the
dominant strategy
In this game, player 1 (firm) has a strictly dominant
strategy: “hire.” Player 2 has only a weakly
dominated strategy. If players are rational, and in
addition Player 2 is cautious, then Player 1 hires
and Player 2 shirks.
Each one of two tea shops charges its own price for
a cup, either Rs2, Rs4, or Rs.5. The cost of
obtaining and serving the tea can be neglected. It is
expected that 6000 cups per month are drunk in a
shop by tourists, who choose one of the two shops
randomly, and 4000 cups per month are drunk by
natives who go to the bar with the lowest price, and
split evenly in case both shops offer the same price.
What prices would the shops select?
The game, as all games considered so far, is
symmetric. Let me illustrate in one instance how to
compute the payoffs. If shop A charges Rs. 2 and
shop B charges Rs4, then all natives will choose A.
Therefore tea shop A will serve 4000 cups to the
natives, and 3000 cups to tourists, serving 7000
cups in total, making 7000 *Rs. 2 = Rs.14000.
Teashop B will only serve 3000 cups to tourists,
making 3000 *Rs. 4 =Rs.12000
The payoff matrix, with values in thousands of Rs, is
For each tea shop, move “4” strictly dominates move “2”, therefore we could eliminate
both moves “2” to
Proposition:
After iterated elimination of dominated strategies, players
will select 1.
Solution
2
Suppose all players choose, the maximum number 100. Then ∗ 100 =
3
66.66667 ≈ 67.
Round 1: All numbers in the interval [68 ,100] are weakly dominated by 67.
Firstly, the numbers in this interval cannot be two third of average of numbers
chosen between 68 and 100. Thus, if any play wins by choosing a number in the
interval [68 ,100], then he will do so, by choosing 67 itself. Thus all numbers in
this interval are weakly dominated by 67, and eliminated.
Round 2: Using the above logic we can eliminate the numbers above
2 2 2 2
( ∗ 100) = ( )2 ∗ 100 = ∗ 67 = 44.444 ≈ 45
3 3 3 3
Thus, after two rounds of elimination, we are left with numbers in the
interval [1, 45]
Round 3: again by same logic we can eliminate all the numbers
2 3 2
above ( ) ∗ 100 ≈ ∗ 45 = 30.
3 3
Thus after round 3, we are left with integers in the interval [1,
30]
2 4 2
Round 4: We can eliminate numbers above ( ) ∗ 100 ≈ ∗
3 3
30 = 20.
Following this logic, all players will ultimately choose 1.
Advertising game
Suppose two competing brands can choose one of the three marketing
campaigns Low(L), Medium (M) and High (H),
Player B
L M H
L 6,6 2,8 0,4
Player A
M 8,2 4,4 1,3
H 4,0 3, 1 2,2
Nash equilibrium
A Nash Equilibrium occurs when each player chooses a
strategy that gives him/her the highest payoff, given the
strategy chosen by the other player(s) in the game.
("rational self-interest")
Alternatively, a Nash Equilibrium occurs when each player
chooses his best responses.
More generally, at Nash equilibrium if no player wants to
unilaterally deviate to an alternative strategy
A Nash equilibrium corresponds to a steady state.
Definition:
The pure strategy profile 𝑠 ∗ = (𝑠1∗ ,𝑠2∗ , … … … … … … 𝑠𝑛∗ ) ∈ 𝑆 is a strict Nash
equilibrium if 𝑠𝑖∗ is a best response to 𝑠−𝑖∗
for all i ∈ N, that is ui(𝑠𝑖∗ , 𝑠−𝑖
∗
) > ui(𝑠𝑖′ 𝑠−𝑖
∗
) for
all 𝑠𝑖′ 𝑆𝑖 and all i ∈ N.
What is a best response?
The strategy 𝑠𝑖 𝑆𝑖 is player i’s best response to his opponent’s
strategies 𝑠−𝑖 𝑆−𝑖 if
ui(𝑠𝑖 , 𝑠−𝑖 ) > ui(𝑠𝑖′ 𝑠−𝑖 ) for all 𝑠𝑖′ 𝑆𝑖
If 𝑠𝑖 𝑆𝑖 is a strictly dominated strategy for player i, then it cannot be a
best response to any 𝑠−𝑖 𝑆−𝑖 .
Corollary
A strictly Dominant strategy, that survives IESDS must be included in the
set of best response strategies
Best Response Function of a Player
we define the function B byi
any action in Bi(s− ) is at least as good for player i as every other action of
i
player i
when the other players’ actions are given by s−i. We call B the best
i
response
function of player i.
The function B is set-valued: it associates a set of actions with any list of the
i
other players’ actions. Every member of the set B (s−i) is a best response of
i
player i to s−i:
if each of the other players adheres to s i then player i can do
−
Step 1: For every strategy of player 2( for each column) find the statefy of player 1 that gives him the
highest payoff( Best response)
Strategies of Player 2 Best response of player 1
L D
C M
R M
Step 2:
Strategies of Player 1 Best response of player 2
U R
M C
D C
The Nash equilibrium is found where the best response correspondence of each player will intersect. In
this case (M,C) is Pure Strategy Nash Equilibrium of this game.
Strict and Non strict PSNE
A A
2 2,3,3 2,2,3 2,1,3 2 2,3,2 6,6,6 6,5,6 2 2,3,1 6,6,5 6,5,5
3 1,3,3 1,2,3 1,1,3 3 1,3,2 5, 6,6 5,5,6 3 1,3,1 5, 6,5 9,9,9
A P -100, -1000,
-100
5, -5, 5 P -100, -1000, -5 5, -10,
-10
C chooses P C chooses Y
Synergistic Relationship Game
e∗2
𝑒1
e1∗
e1∗ = e∗2 = c
First Price sealed bid Auction
A first-price auction differs from a second-price auction only in that
the winner
pays the price she bids, not the second highest bid. Precisely, a first-
price sealed bid auction (with perfect information) is defined as
follows.
Players The n bidders, where n ≥ 2.
Actions The set of actions of each player is the set of possible bids
(nonnegative numbers).
Preferences The payoff of any player i is vi − bi if either bi is higher
than every
other bid, or bi is at least as high as every other bid and the number
of every other player who bids bi is greater than i. Otherwise player i’s
payoff is 0.
Proposition
One Nash equilibrium of a first-price sealed-bid auction is (b1, . . . ,
bn) = (v2, v2, v3, . . . , vn), in which player 1’s bid is player 2’s valuation
v2 and every other player’s bid is her own valuation.The outcome
of this equilibrium is that player 1 obtains the object at the price
v2.
Proof: If player 1 raises her bid above v2, she still wins, but pays a higher
price and hence obtains a lower payoff. (𝑣1 − 𝑣2 − 𝜖). If player 1 lowers her
bid then she loses, and obtains the payoff of 0. Player 1 has no incentive for
unilateral deviation.
If any other player changes her bid to any price at most equal to 𝑣2 the
outcome does not change. If she raises her bid above v2 she wins, but
obtains a negative payoff. Has no incentive for unilateral deviation.
Contributing to a public good
Each of n people chooses whether or not to contribute a fixed amount toward the
provision of a public good.
The good is provided if and only if at least k people contribute, where 2 ≤ k ≤ n;
If it is not provided, contributions are not refunded.
Each person ranks outcomes from best to worst as follows:
(i) any outcome in which the good is provided and she does not contribute,
(ii) any outcome in which the good is provided and she contributes,
(iii) any outcome in which the good is not provided and she does not contribute,
(iv) any outcome in which the good is not provided and she contributes.
Formulate this situation as a strategic game and find its Nash equilibria.
Is there a Nash equilibrium in which more than k people contribute?
One in which k people contribute?
One in which fewer than k people contribute?
There are n players, each of which has two strategies: Contribute (C) or
don't contribute (D). The payoffs for player i are as follows, given that 𝛼
other players contribute.
if n≥ 𝛼 ≥ k, player i is better off not contributing (the good will be provided
without her contribution).
If 𝛼= k − 1 people contribute, player i is better off contributing (the
good will only be provided with her contribution which she values
higher than the good not being provided).
If 0 ≤ 𝛼 ≤ 𝑘 − 2 people contribute, player i is better off not contributing
(the good won't be provided anyway).
Clearly, player i best response/optimal strategy does depend what the other
players are doing, so there is no dominant strategy/dominated strategy.
Nash Equilibrium Solution:
First, is there a NE where more than k people contribute?
Answer: no, since this means that from the point of view of
some player i, k others already contribute, so the best
response for this player is to play D.
Second, is there a NE where exactly k people contribute?
Yes: from a contributing player i's point of view 𝛼 = 𝑘 − 1
people play C so he should play C as well. This is true for all
players who play C. For all players who play D, 𝛼 = 𝑘 so
playing D is optimal as well. Note that we cannot say who
contributes. Each possible combination of contributing
players/defecting players is a NE.
Finally, are there NE were strictly less than k people
contribute?
Yes, there is one NE where no one contributes: if nobody else
contributes 𝛼 = 0 and it doesn't pay to contribute for player i
either. So D for all players is a NE.
Four people must drive from A to B at the same time. Two routes are available, one via X and
one via Y. (The roads from A to X, and from Y to B are both short and narrow;
in each case, one car takes 6 minutes, and each additional car increases the travel
time per car by 3 minutes. (If two cars drive from A to X, for example, each car takes
9 minutes.) The roads from A to Y, and from X to B are long and wide;
on A to Y one car takes 20 minutes, and each additional car increases the travel time per car
by 1 minute;
on X to B one car takes 20 minutes, and each additional car increases the travel time per car
by 0.9 minutes.
Formulate this situation as a strategic game and find the Nash equilibria. (If all four people take
one of the routes, can any of them do better by taking the other route? What if three take one
route and one takes the other route, or if two take each route?)
Solution
A strategic game that models this situation is:
Players The four people.
Actions The set of actions of each person is {X,Y} (the
route via X and the route
via Y).
Preferences Each player’s payoff is the negative of her
travel time.
Nash equilibrium
In a Nash equilibrium two people take each route. (In any other case, a person taking
the more popular route is better off switching to the other route.
For any such action profile, each person’s travel time is either 29.9 or 30 minutes
.(depending on the route they take).
If a person taking the route via X switches to the route via Y her travel time becomes
12 + 21.8 = 33.8 minutes; if a person
taking the route via Y switches to the route via X her travel time becomes 22+12 =
34 minutes. For any other allocation of people to routes, at least one person can
decrease her travel time by switching routes. Thus the set of Nash equilibria is the
set of action profiles in which two people take the route via X and two people take
the route via Y.
A new road is built
Now suppose that a relatively short, wide road is built from X to Y, giving each
person four options for travel from A to B: A–X–B, A–Y–B, A–X–Y–B, and A–Y–
X–B.
Assume that a person who takes A–X–Y–B travels the A–X portion at the same time as someone
who takes A–X–B, and the Y–B portion at the same time as someone who takes A–Y–B. (Think of
there being constant flows of traffic.)
On the road between X and Y, one car takes 7 minutes and each additional car increases
the travel time per car by 1 minute. Find the Nash equilibria in this new situation.
Compare each person’s travel time with her travel time in the equilibrium before the road from X to
Y was built.
Solution
Suppose minjej =𝑒
Let player 𝑖 chooses 𝑒𝑖 = 𝑒 ∗ > 𝑒
The payoff 𝑢𝑖 = 2𝑒 − 𝑒 ∗
The player 𝑖 can improve his payoff if he moves towards
𝑒 and he will have incentive to deviate 𝑒.
So any player with effort level 𝑒 ∗ > 𝑒, will have
incentive to deviate to 𝑒.
A model of electoral competition.
First proposed by Hotelling (1929) and popularized by Downs (1957).
Setup:
◃ Parties/candidates compete by choosing a policy on the line
segment [0, 1]. The party with most votes wins; if there is a
draw, each party has a 50% chance of winning.
◃ Parties only care about winning, and will commit to the
platforms they have chosen.
◃ Each voter has a favorite policy on [0, 1]; her utility decreases
as the winner’s position is further away from her favourite policy
Voter’s Preference regarding one dimensional
Policy variable:
Single peaked preferences: Single-peaked – each voter has an ideal point and
the further away from the ideal point the lower their utility. Each voter’s
distaste for any position is given by the distance between that position and
her favorite position. In particular, for any value of k, a voter whose favorite
position is x∗ is indifferent between the positions x∗ − k and x∗ + k.
The payoff of a voter whose favorite position is x∗, as a function of the winning position, x.
Single Peaked preference: examples
The allocation of votes between three candidates, with positions x1, x2, and x3.
Consider: Two Party case
Action: A party will choose a value for policy variable x.
Best response for a party:
Fix the position 𝑥2 of candidate 2 and consider the best position for candidate 1. First
suppose that 𝑥2 < m, the median value of policy variable based on voter’s preferences.
candidate 1 takes a position to the left of 𝑥2 then candidate 2 attracts the votes of
all citizens whose favourite positions are to the right of (𝑥1 +𝑥2 )/2 , a set that includes the
50% of citizens whose favorite positions are to the right of m, and more.
Thus candidate 2 wins, and candidate 1 loses.
If candidate 1 takes a position to the right of 𝑥2 then she wins so long as the dividing line
between her supporters and those of candidate 2 is less than m. If she is so far to the right
that this dividing line lies to the right of m then she loses.
Player 1 prefers to win than to lose, and is indifferent between all the
outcomes in which she wins, so her set of best responses to 𝑥2 is the
set of positions that causes the midpoint (𝑥1 +𝑥2 )/2 of the line segment
from 𝑥2 to 𝑥1 to be less than m. (If this midpoint is equal to m then the
candidates tie.) The condition (𝑥1 +𝑥2 )/2 < m is equivalent to 𝑥1 < 2𝑚 −
𝑥2.
𝑥1 + 𝑥2 = 2𝑚
If we superimpose the two best response functions, we
see that the game has a unique Nash equilibrium, in
which both candidates choose the position m, the voters’
median favorite position.
The edges of the shaded area, which correspond to pairs
of positions that result in ties, are excluded from the best
response functions. The outcome is that the election is a
tie.
This is the so-called ”median voter theorem
The Median Voter Theorem only strictly holds if there is a single issue.
If there are two or more issues that parties take stands on, but only one
election, there is no guarantee that the median voter's preference will win
on any issue.
Moreover, even with single-peaked preferences, multiple voting
dimensions make it possible for voting cycles to arise. Chaos
Suppose there is a continuum of voters, with favorite
policies uniformly distributed on [0, 1], and the number
of parties is 3 (L, C, R). Do we still have the equilibrium
that all parties choose m?
No. One of the parties can move slightly to the left or the
right of the median voter position, and win the election.
Would the three parties positioning at 0.45, 0.55, 0.6 be a
NE?
Yes. L wins already; C and R cannot win by moving
anywhere.
Voting Game
Two candidates, A and B, compete in an election where there are n number
of voters and n is even. Of the n voters, n/2 support candidate A and n/2
support candidate B. Each citizen decides whether to vote, at a cost, for the
candidate she supports, or to abstain. A citizen who abstains receives the
payoff of 2 if the candidate she supports wins, 1 if this candidate ties for first
place, and 0 if this candidate loses. A citizen who votes receives the payoffs
2−c, 1−c, and−c in these three cases, respectively, where 0< c <1. Find the
set of all pure-strategy Nash equilibria.
Solution
There are four possible cases:
Case 1: All voters vote. In this case both candidates will be
tied. Payment for those who has voted is 1-c.
Possibility of deviation. If one voter does not vote, he gets
zero as his candidate will loose. Thus it is best for him to
vote.
Case 2: Not every one votes and candidates are tied.
In this case the person who has not given his vote gets 1.
However if he is casting his vote his payoff will go up to 2-
c>1 as his candidate will win. Thus this is not a Nash
equilibrium as the player has incentive to deviate.
Voting game continued
Case 3: Not everyone votes and one candidate wins by one
vote: The voter who does not cast his vote and supports
the loser will get zero. However he can improve his
condition deviating; if he gives his vote so that the result is a
tie and the voter gets 1-c >0. Thus, it is not a Nash
equilibrium.
Case 4: Not everyone votes and one candidate wins by
more than one vote: In this case, a voter who votes
for the wining candidate would probably deviate to not
voting, since 2-c < 2.
Thus, the only Nash equilibrium is the one where everyone
votes.
Oligopoly models and Game Theory
The word Oligopoly is derived from two Greek words –
‘Oligi’ meaning ‘few’ and ‘Polein’ meaning ‘to sell’.
The Oligopoly is a market structure with finite number of
sellers, where each seller’s decision will affect the profitability
of others.
A duopoly is a special form of oligopoly with two firms
operating in a market.
In game theoretic terminology, duopoly or Oligopoly can be
defined as the market structure where firms are strategically
interdependent.
In this case we will consider the situations where firms
simultaneously choose there decision variables.
Case1: Firms simultaneously choosing Quantity
levels( Cournot Model)
Two firms
• Maximising profit and Producing homogenous product.( We can also have models with
heterogeneous product)
• Neither firm knows the other’s output decision when making its production choice as
they move simultaneously
Each firm makes an assumption about the other’s decision
• firm 1 assumes firm 2’s output to be given number
• likewise for firm 2
• Each is fully described by its cost function.
Two firms labelled f = 1,2
Firm f produces output qf
So total output is:
• q = q1 + q2
Market price is given by:
• p = p (q)
Firm f has cost function Cf(·)
So profit for firm f is:
• p(q) qf –Cf(qf )
Each firm’s profit depends on the other firm’s output
• (because p depends on total q)
Firm 1’s problem is to choose q1 so as to maximise
1(q1; q2) := p (q1 + q2) q1 – C1 (q1)
Differentiate P1 to find FOC:
1(q1; q2)
————— = p q(q 1 + q2) q1 + p(q1 + q2) – C 1(q1)=0
q
q1
1(q1; q2)
————— =0
q 1
𝜕2 𝜋1 1 𝜕2 𝜋1 2 =0
• 12 d𝑞 + 2 1 𝑑𝑞
𝜕𝑞 𝜕𝑞 𝜕𝑞
1(q1; q2) = const Slope of Iso-profit curve of firm 1
𝑑𝑞 2 pq(q1 + q2) q1 + p(q1 + q2) – Cq1(q1)
=−
q0 1(q1; q2) = const 𝑑𝑞1 𝑝𝑞
1(q1given
Firm 1’s choice ; q2) =that
const
2 chooses output q0
q1
April 2018 99
R1(·) is the reaction function observed from profit-maximisation by firm 1
Gives firm’s reaction 1 to fixed output level of competitor:
• q1 = R1 (q2)
• Slope of reaction function of Firm 1:
2 2
𝑑𝑞2 𝜕2 𝜋1 Τ𝜕𝑞1 𝜕2 𝜋1 Τ𝜕𝑞1
• ቚ = − 2 Τ 2 1 = − <0 ( unambiguously for linear dd function)
𝑑𝑞1 𝑅1 𝜕 𝜋1 𝜕𝑞 𝜕𝑞 𝑝𝑞𝑞 𝑞1 +𝑝𝑞
Firm 2’s problem is also solved in the same way
We get q2 as a function of q1 :
• q2 = R2 (q1)
• Slope of reaction function of firm 2
𝑑𝑞2 𝜕2 𝜋2 Τ𝜕𝑞2 𝜕𝑞1
• ቚ =− 2 <0
𝑑𝑞1 𝑅2 𝜕2 𝜋2 Τ𝜕𝑞2
Treat the above as a pair of simultaneous equations
Solution is a pair of numbers (qC1 , qC2)
• So we have qC1 = R1(R2(qC1)) for firm 1
• and qC2 = R2(R1(qC2)) for firm 2
This gives the Cournot-Nash equilibrium outputs
Second order condition of Profit Maximization
and Stability of Reaction function diagram :
2 12
𝜕 𝜋1 Τ𝜕𝑞 <0
2 Τ 22
𝜕 𝜋2 𝜕𝑞 <0
2 Τ 12
𝜕 𝜋1 𝜕𝑞 𝜕 2 𝜋2 Τ𝜕𝑞1 𝜕𝑞 2
2 >0
𝜕 2 𝜋1 Τ𝜕𝑞 2 𝜕𝑞1 𝜕 2 𝜋2 Τ𝜕𝑞 2
From this we get
2
𝜕2 𝜋1 Τ𝜕𝑞1 𝜕2 𝜋2 Τ𝜕𝑞2 𝜕𝑞1
>
𝜕2 𝜋1 Τ𝜕𝑞2 𝜕𝑞1 𝜕2 𝜋2 Τ𝜕𝑞 22
𝑑𝑞2 𝑑𝑞2
ቚ > ቚ
𝑑𝑞1 𝑅1 𝑑𝑞1 𝑅2
Thus the reaction function of Firm 1 must be steeper than the reaction function of Firm 2.
Cournot-Nash equilibrium :
𝑞01
q2
R1(·)
b
qC2
R2(·)
𝑞02
a
q1
qC1
It also is the Nash equilibrium of a simple quantity game:
• players are the two firms
• moves are simultaneous
• strategies are actions – the choice of output levels
• functions give the best-response of each firm to the
other’s strategy (action)
A numerical Example
Take the case where the inverse demand function is:
p = b0 – bq
And the cost function for f is given by:
Cf(qf ) = C0f + cf qf
So profits for firm f are:
[b0 – bq ] qf – [C0f + cf qf ]
Suppose firm 1’s profits are
Then, rearranging, the iso-profit curve for firm 1 is:
b0 – c1 C01 +
q2 = ——— – q1 – ————
b b q1
Firm 1’s profits are given by
1(q1; q2) = [b0 – bq] q1 – [C01 + c1q1]
So, choose q1 so as to maximise this
Differentiating we get:
1(q1; q2)
————— = – 2bq1 + b0 – bq2 – c1
q 1
𝑀𝑅1 = 𝑀𝐶 1
FOC for an interior solution (q1 > 0) sets this equal to zero
Doing this and rearranging, we get the reaction function:
b0 – 𝑐 2 𝑞 1 (
𝑞2 = − Assuming an interior solution exists)
2b 2
Nash Equilibrium Solution
Solving the reaction functions we get
a + c2 – 2c1 a + 𝑐 1 – 2𝑐 2
𝑞𝑐1 = 𝑞𝑐2 =
3𝑏 3𝑏
Firms with homogenous marginal cost:
60
30 - .5q2
30
10
0 15 25 30
Output of seller 1
In similar fashion, we derive a best reply function for seller 2. It is given by
q2 = 30 - .5q1
q2
30 q2 = 30 - .5q1
0 60 q1
So we have a system with 2 equations and 2 unknowns (q1 and q2) :
q1 = 30 – .5q2 q2 = 30 – .5q1
30
Equilibrium
20 Seller 2’s BRF
0 20 30 60 q1
Bertrand Model
• Bertrand model
– Oligopolistic competition
– Firms compete - setting prices
• Demand function
D( pi ) if pi p j
Di ( pi , p j ) (1 / 2)[ D( pi )] if pi p j
0 if pi p j
• Now, no Nash equilibrium exists with two firms operating in the market.
• Clearly, any outcome where p1 < c1 or where p2 < c2 is not an equilibrium (at least one firm
will earn negative profits and can profitably deviate).
• Any outcome where min[p1,p2] > c1 is not an equilibrium; at least one firm could increase
their profit by lowering their price.
• p1 = p2 = c1 is not an equilibrium; firm 2 could profitably lower their price.
• p1 ≥ c1, p2 < c1 is not an equilibrium; firm 2 could increase their price and increase its profit.
• Thus, no equilibrium exists.
Bertrand Model with Heterogeneous Firms
Suppose Cost functions C(q) = cq for both firms.
Let demand function is qi = α – pi + bpj, where α > c, 0 < b < 2.
Firm 1 and 2 choose prices simultaneously.
So, now we have a well-behaved problem with continuous profit functions, and well-defined best response
functions.
Firm 1 solves: maxp1 (α – p1 + bp2)(p1 – c)
This gives FOC: α – 2p1 + bp2 + c = 0
So BR1: p1 = (α + bp2 + c)/2
p1
Equilibrium p1*=p2*=30
1= p1q1 = p1(90-2p1+p2)= 30(60)=1800
2= p2q2 = p2(90-2p2+p1)= 30(60)=1800
No Bertand Paradox with Differentiated Products
Tragedy of commons
Suppose there are n players say firms in the world each choosing how much
to produce. The production activity consumes some of the clean air that
surrounds our planets. There is a total amount of clean air equal to K and
any consumption of clean air comes out of this common resource. Each
player i chooses his own consumption of clean air for production, 𝑘𝑖 ≥ 0
and the amount of clean air left is therefore
𝐾 − σ𝑛𝑖=1 𝑘𝑖
The benefit of consuming an amount 𝑘𝑖 ≥ 0 gives player i a payoff equal to
ln(𝑘𝑖 ), and no other player benefits from i’s choice. The utility of player i is
𝑣𝑖 (𝑘𝑖 , 𝑘−𝑖 ) = ln 𝑘𝑖 + ln(𝐾 − σ𝑛𝑗=1 𝑘𝑗 )
Let the Nash equilibrium solution be (𝑘1∗ , 𝑘2∗ ………….. 𝑘𝑛∗ ) where 𝑘𝑖∗ =
∗
𝐵𝑅𝑖 (𝑘−𝑖 ) for all 𝑖 ∈ 𝑁
Maximising the objective function of each n players we
get their best response functions( aka reaction function)
𝜕𝑣𝑖(𝑘𝑖 ,𝑘−𝑖 ) 1 1
= − σ𝑛 =0
𝜕𝑘𝑖 𝑘𝑖 (𝐾− 𝑗=1 𝑘𝑗 )
(𝐾−σ𝑛
𝑖≠𝑗 𝑘𝑗 )
thus 𝑘𝑖∗ = 𝐵𝑅𝑖 𝑘−𝑖
∗
= for all n agents.
2
Under symmetry condition
𝑘1∗ = 𝑘2∗ …………..= 𝑘𝑛 ∗
= 𝑘∗
Thus 𝑘 ∗ = 𝐾/(𝑛 + 1)
This is the PSNE of the game
Social Planner
Suppose a social planner is maximising the joint profit to
determine the air consumption of all n agents.
V= σ𝑛𝑗=1(𝑙𝑛𝑘𝑗 ) + n ∗ ln(𝐾 − σ𝑛𝑗=1 𝑘𝑗 )
𝜕𝑉 1 𝑛
= − 𝑛 =0 for all i
𝜕𝑘𝑖 𝑘𝑖 (𝐾−σ𝑗=1 𝑘𝑗 )
Again using the symmetry condition we have
1 𝑛
′− ′ = 0
𝑘 𝐾−𝑛𝑘
Or
′ 𝐾
𝑘 = < 𝐾/(𝑛 + 1)
2𝑛
Final Offer Arbitration
Arbitration is a procedure in which a dispute is submitted,
by agreement of the parties, to one or more arbitrators
who make a binding decision on the dispute.
In Final-offer arbitration, in which the two parties make
offers to the arbitrator, and then the arbitrator picks one
of the offers to be the wage settlement.
How the arbitrator makes her decision?
Final Offer arbitration Continued.
Firstly, the game starts with the Firm and the union making
simultaneous offers, which will we denote respectively by wf and
wu. The arbitrator has a settlement in mind which we will call x,
and we assume that she simply chooses whichever offer is closer
to x, provided that wf < wu. If x < (wf + wu)/2 then she chooses
wf , and if x > (wf + wu)/2, then she chooses wu. This is expressed
in the figure below, where we consider x to be any value on a
real line, with wf and wu fixed.
The Model
Considering the terminology we just introduced, the arbiter knows her ideal
settlement, x, but the other parties do not . They assume that x is randomly
distributed by a cumulative probability distribution F(x), with corresponding
probability density function f(x).
𝑑𝐹(𝑥)
Note that =f x
𝑑𝑥
This is a property of cumulative distribution functions.
With these assumptions kept in mind, the parties want to know the
probability that their offer is chosen; these probabilities can be expressed as
These probabilities tell the firm and the union what chances they have of
their offer getting chosen. We can use these probabilities to determine
the expected wage settlement, denoted by E[settlement]. Thus we have
that
This equation is very important to both the firm and the union; the Firm
wants to minimize this value, but the union wants to maximize it. If the
offer(𝑤𝑓∗ ; 𝑤𝑢∗ ) is to be a Nash equilibrium of this game, then the following
must be
true:
max
This means that the pair (𝑤𝑓∗ ; 𝑤𝑢∗ ) ) must solve the first
order conditions for both of the optimization problems
above. These first order conditions:
The left hand sides are equal implies
Or
This means that the difference between the offers must equal the
value of the density function at the median of the arbitrator's
preferred settlement (x).
Now suppose that the arbiter's preferred settlement is normally
distributed with mean 𝜇 and variance 𝜎 2 . This means that we
have the density function