Econ 703 Midterm suggested solution prepared by Ryo Okui and Rui Che

1
Note that the utility function is given by (13 pi )(pi 1) 1 i (pi , pi ) = 2 (13 pi )(pi 1) 0 if pi < pi if pi = pi if pi > pi

(1)

a.
The NE is (1, 1). Proof. As a rst step, we will see that (1, 1) is a NE. If player 2 plays 1, then player 1 cannot get a positive prot and playing 1 gives 0 to player 1. Therefore player 1 doesnt have an incentive to deviate. Exactly same argument shows that player 2 doesnt have an incentive to deviate either. Hence (1, 1) is a NE. Next we prove that there is no other NE. It is enough to consider the case p1 p2 Case 1: p1 p2 and p1 < 1 Player 1 obtain a negative prot. Therefore player 1 has an incentive to deviate to 0 which gives 0 prot to player 1. Case 2: p1 < p2 and 1 < p1 < 13 Player 2 gets 0. Therefore player 2 has an incentive to deviate to p1 which gives a positive prot to player 2. Case 3: p1 < p2 and p1 = 1 Player 1 gets 0. Therefore player 1 has an incentive to deviate to p1 + where is small enough so that player 1 gets a positive prot. Case 4: p1 p2 and p1 13 Player 1 has a non-positive prot. Therefore player 1 has incentive to deviate to 7 which gives strictly positive prot. Case 5: p1 = p2 and 1 < p1 13 Player 1 has incentive to deviate to p1 where > 0 satises that 1 (13 p1 + )(p1 1) > 2 (13 p1 )(p1 1). Such a always exists. If p1 7 the left hand side of the inequality is a decreasing function of and by taking the limit 0, that inequality is satised. If p1 > 7 then = p1 7 does work.

b.
The maximum payo that rm 1 can get at any NE is 0 Proof. playing 1 at the 1st stage and 1 at the 2nd stage after any history for both players constitutes a NE (actually a SPNE). And under that strategy prole, player 1 gets 0. Therefore the prot 0 is attainable. Suppose that there exists a NE which gives a positive prot to 1. Let (p1 , p2 , p1 , p2 ) be such a NE 1 1 2 2 where the superscripts indicate stage and the subscripts indicate player. 1

Note that under a NE, a stage game NE must be the 2nd stage outcome. therefore p2 (p1 , p1 ) = 1 1 2 = 1. This means player 1 cannot get a positive prot at the 2nd stage. Since this NE give a positive prot to player 1, we must have 1 < p1 p1 and p1 < 13. Then player 2 has an incentive to 1 2 1 deviate to the strategy to play p1 at the 1st stage and play 1 at the 2nd stage after any history for 1 some . As we saw in a., there exists a with which player 2 can get more prot than the original strategy prole. And since player 2 gets 0 at the 2nd stage by this deviation which is same as the prot at the 2nd stage under the original strategy prole, in total this deviation pays. Therefore player 2 has an incentive to deviate which contradicts (p1 , p2 , p1 , p2 ) is a NE. 1 1 2 2 p2 (p1 , p1 ) 2 1 2

c.
That is 18 for both of them. The total prot is 36. And the corresponding SPNE is that play 7 if that is the 1st stage or all history consists of 7. play 1 otherwise Proof. 36 is the highest possible total per stage prot by a simple algebra. So if it is attainable, then it must be the maximum per period payo in a SPNE. Now we show that the strategy prole above constitutes a SPNE for some large enough. Note that checking incentives to deviate for player 1 is enough. On path Under that strategy prole player 1s prot stream is (18, 18, 18, . . . ). Here note that we cannot have a most protable 1-shot deviation, however we can take a suprimum of prot from protable 1-shot deviations. It is very easy to show that the suprimum is 36 or (36, 0, 0, . . . ) in terms of prot 1 stream. So if 18 (1 )36 or 2 , player 1 doesnt have an incentive to deviate. O path In this game, the stage game NE is played at every stage, therefore there is no incentive to deviate. To sum up, there is no incentive to deviate at any subgame. Therefore that strategy prole constitutes a SPNE.

d.
As we saw in c., the minimum is
1 2

e.
Consider the following strategy prole. p1 = 2 1 p1 = 2 p2 = 1 p2 = 2 3 if 2 was played 1 otherwise 10 if (2, 3) was played 1 otherwise 7 if (2, 3, 10) was played 1 otherwise (2) (3) (4) (5)

where the superscripts indicate stage and the subscripts indicate player. This strategy prole constitutes a NE and player 1 gets (13 2)(2 1) + 0 = 11. Proof. First x the 1st stage strategy and consider deviations at the 2nd stage only. Player 1 cannot get a positive prot given player 2s strategy and gets 0 under this strategy. Therefore player 1 doesnt have incentive to deviate. Player 2 gets 36 which is the highest possible prot in a stage game. Therefore player 2 doesnt have an incentive to deviate. Then Consider deviations which alter the 1st stage outcome. If player 1 deviates at the 1st stage, then player 1 cannot get a positive prot since player 2 will play 1 in both stages. Therefore player 1 doesnt have an incentive to deviate. Player 2 can get a positive prot in the 1st stage by a deviation. The suprimum value of prot in the 1st stage from deviations is 11. However deviating in the 1st stage implies that player 1 will play 1 in the 2nd stage and player 2 can never get a positive prot in the 2nd stage. Therefore the suprimum value of prot from deviation is 11 which is less than the prot from the original strategy, 36. Therefore there is no incentive to deviate.

2
a.
The strategy sets for Alan and Bob are: SA = {(s1 , s2 , s3 , s4 )|s1 [0, 8]; A A A A A s2 A s3 A s4 A {(s1 , s2 , s3 , s4 )|s1 B B B B B 2 sB s3 B s4 B : : : : : : : [0, 8] [0, 4] {A, R}; [0, 8] [0, 4] [0, 2]; [0, 8] [0, 4] [0, 2] {A, R}} [0, 8] {A, R}; [0, 8] [0, 4]; [0, 8] [0, 4] [0, 2] {A, R}; [0, 8] [0, 4] [0, 2] [0, 1]} 3 (6) (7) (8) (9) (10) (11) (12) (13)

SB =

b.
Use backward induction: In the last period (t=4), Alan will accept any strictly positive oer to him. If hes oered zero, he should also accept it if we want to nd an SPNE (otherwise Bobs best response is not well-dened). Therefore Alans strategy at t=4 is accept any oer. Considering this, when we move back to Bob, hell oer zero to Alan and 1 to himself. Now go back to t=3. Alan knows that if the game goes to the last stage, Bob can get 1, therefore Bob will not accept anything less than 1 in this stage. Hell oer 1 to Bob and 1 to himself.. Bobs strategy at this period is accept any oer larger than or equal to 1 (Why larger than or equal to? Same reason as above.) and reject otherwise. Then go back to t=2... The SPNE that we get is (Notation: If its player i s turn to make the oer at period t, st represents the oer made to Alan i (how much Alan is going to get). s1 = 5 A A if s1 5 A s1 = B R otherwise A if s1 1 A s2 = A R otherwise s2 = 1 B s3 = 1 A A if s1 1 A s3 = B R otherwise s4 = A A s4 = 0 B

c.
Since backward induction gives us this unique strategy prole, the pure strategy SPNE is unique.

d.
There are many such NEs, so the answer is not unique. Below is just one of them: s1 = 8, s2 = R, s3 = 2, s4 = R A A A A s1 = A, s2 = 0, s3 = A, s4 = 0 B B B B

e.
There are many such NEs, so the answer is not unique. Below is just one of them: 2 s1 = 8, sA = A, s3 = 2, s4 = R A A A 1 = R, s2 = 0, s3 = R, s4 = 0 sB B B B