Ec432A Spring, 2022 Game Theory Review
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Ec432A
Spring, 2022
Game Theory Review
Key points:
1. Non-cooperative Game (Week 9’s notes)
• Comply-Cheat scenarios profits calculation:
= 0 to derive $ response function, which is a function of %
= , − &'() =
$ + % =
= 0 to derive % response function, which is a function of $
= , − &'() =
$ + % =
Using two equations to find two unknowns and .
• Credible Punishment calculation:
|
|
Firm 2 |
|
|
|
Comply |
Cheat |
Firm 1 |
Comply |
($3757, $3757) |
($2907, $4347) |
Cheat |
($4347, $2907) |
($3456, $3456) |
{Comply, Comply} is at least as good as your payoff in {Cheat, Cheat}
$3757 − α ≥ $3456
Remove the incentive to Cheat
$3757 − ≥ $4347 −
• Repeated Game calculation:
1 + + % + , + ⋯ =
-./012 = 3757 + 3757 + 3757% + ⋯ = 3757(1 + + % + ⋯ ) = > -3)'( = 4347 + 3456 + 3456% + ⋯ = 4347 + 3456(1 + + % + ⋯ ) = 4347 +
This is a subjective probability: what chance do you believe the game will continue?
2. Cooperative Game (Week 10’s notes)
• Finding two NE and the corresponding calculation
|
|
Firm 2
|
|
|
|
Comply |
Cheat |
Firm 1 |
Comply |
($3757, $3757) |
($3407, $3747) |
Cheat |
($3747, $3407) |
($3456, $3456) |
B -./012 C ≥ ( -3)'()
3757 + 3407(1 − ) ≥ 3747 + 3456(1 − )
≥ 83%
• Risk factor calculation:
Risk factor = 1 − , the probability resulting in the Cheat-Cheat NE.
This is also a subjective probability, what’s your belief on the other firm’s cooperation?
With two different equations
|
|
gov |
|
|
|
Tax (p) |
No tax (1-p) |
firm |
Merge (q) |
1000, 4
|
2000, 5 |
Separate (1-q) |
1200, 4.8 |
1800, 4.5 |
(/ ) > (4 ) => <
( ( ) < (5 ) => >
Each firm’s profit (minus taxation if getting tax)
Tax revenue (q_u * tax rate)
Firm production level q_u + q_d
The government must hold a belief that the firm has more than 3/13 chance to merge to choose the No-tax strategy. The firm must hold a belief that the government has more than 1/2 chance to choose the tax policy to merge. Note that this will also be observed by an outsider (like a local resident) to calculate the corresponding probability.
3. Incomplete Information
• Bayes’ Rule calculation:
Updated belief based on the new information. Consider the nod-and-smile gesture in the job interview example.
• Bayes Nash Equilibrium calculation: Steps of solving this type of game:
a. If You play strategy A for certain, find the other player’s best response in a format: (type 1 with probability : strategy , type 2 with probability 1 − : strategy Y)
b. Calculate your payoff under this best response: (6 )
c. Assuming the same best response made by the other player, calculate the payoff if you switch to strategy B: (7 )
d. Let (6 ) > (7 ) to find the probability .
e. Under this probability, BNE = (A, (type 1: strategy X, type 2: strategy Y))
f. If You play strategy B for certain, find the other player’s best response in a format: (type 1 with probability : strategy Y, type 2 with probability 1 − : strategy X). Note, these X and Y are made up strategies, could be the same as above.
g. Calculate your payoff under this new best response: (7 )
h. Assuming the same new best response made by the other player, calculate the payoff if you switch to strategy A: (6 )
i. Let (6 ) < (7 ) to find the probability .
j. Under this probability, BNE = (B, (type 1: strategy Y, type 2: strategy X))
k. Outside the probability range you found above, no BNE can be achieved.
• Separating or Pooling NE calculation: Steps of solving this type of game:
a. Write the format of the possible equilibria
Separating strategy = (type 1 with probability : strategy A, type 2 with probability 1 − : strategy B) or (B, A)
Pooling strategy = (type 1 with probability : strategy A, type 2 with probability 1 − : strategy A) or (B, B)
b. Check the separating equilibrium
If the other player plays (A, B), in each payoff matrix, you can find your optimal responses accordingly. Assume that strategy A is dominant, then use (6 ) > (7 ) to find the probability.
c. Under this probability range, BNE = (A, (A, B)). Note that it could also be (B, (A, B)), (A, (B, A)) or (B, (B, A)) based on the matrix tables you observe.
d. Check the pooling equilibrium
If the other player play (A, A), in each payoff matrix, you can find your optimal responses accordingly. Assume that strategy B is dominant, then use (6 ) < (7 ) to find the probability.
e. Under this probability range, BNE = (B, (A, A)). Note that it could also be (A, (A, A)), (B, (B, B)) or (A, (B, B)) based on the matrix tables you observe.
4. Applications
• Game Tree calculation:
Be sure to compute the hidden probability, such as both efforts ’ failure probability in the
case 3 example
• Cournot calculation:
Emphasize the response function definition:
$ = $ (% ) = 2 − 2
% = % ($ ) = −
Note that results would be different if marginal costs are different: ($ ) = $ $ and (% ) = % % :
Then, write down profit functions for each firm first, take the derivatives with respect to the corresponding firm’s output, set it to be 0 to derive the response function.
• Stackelberg calculation:
Because firm 2 enters into the market as a follower, it produces based on firm 1’s output level as its best response:
% = % ($ ) = −
Firm 1, on the other hand, can take into this activity into its consideration to maximize its profit in the 2nd stage:
$ = B − ($ + % )C $ − $
= K − L $ + M − NOP $ − $
= Q − − R $ − $
= − − $ − = 0
So, in Stackelberg competition:
-the leader has higher profits
-the follower has lower profits This is called a first mover advantage.
Note that if $ ≠ % , still follow these steps, but change to $ or % based on the equations you use. And and will be functions of , , $ and % .
• Bertrand calculation:
Firms determine the prices, $ and % , in a competitive market. The following analysis is important
a. If $ < : % > $
b. If $ > : % < $ and % > .
c. If $ = : % = $ =
2022-05-04