Académique Documents
Professionnel Documents
Culture Documents
Solutions to Assignment 8
Advanced Microeconomics
1. To find the Subgame Perfect equilibrium of the game we need to solve the game by
backward induction. Consider the second period, the player who gets to make
the offer is making a take-it-or-leave-it offer in which he demands the entire pie
for himself: x2 = 1 if player A makes the offer and x2 = 0 if player B makes
the offer. The offer is accepted.
Consider now the end of period 1. Assume that the outcome of the randomiza-
tion device in the first period is β then it is player A that has to decide whether
to accept or reject the offer. Notice that if player A decides to reject the offer
his expected payoff is δ 23 . This implies that it is subgame perfect for player A
to accept any offer x1 ≥ δ 23 and reject all offers such that x1 < δ 32 .
Now if, in this case, player B makes an offer x1 < δ 23 player A will reject it
and player B’s payoff will be: δ 13 . If instead player B offers δ 23 to A she will
accept it and player B’s payoff will be 1 − δ 32 . Notice that 1 − δ 23 > δ 13 since
1 > δ. Hence, if the outcome of period 1’s randomization device is β player B’s
equilibrium offer is x1 = δ 23 and his payoff is 1 − x1 = 1 − δ 23 .
Assume now that the outcome of the randomization device in the first period
is α then it is player B that has to decide whether to accept or reject the offer.
Notice that if player B decides to reject the offer his expected payoff is δ 13 . This
implies that it is subgame perfect for player B to accept any offer such that
1 − x1 ≥ δ 13 and reject all offers such that 1 − x1 < δ 13 .
Now if, in this case, player A makes an offer such that 1 − x1 < δ 31 player B
will reject it and player A’s payoff will be: δ 23 . If instead player A makes an
offer such that 1 − x1 = δ 31 to B, he will accept it and player A’s payoff will be
1 − δ 13 . Notice that 1 − δ 13 > δ 32 since 1 > δ. Hence, if the outcome of period 1’s
1
randomization device is α player A’s equilibrium offer is such that 1 − x1 = δ 13
and her payoff is x1 = 1 − δ 13 .
The expected Subgame Perfect equilibrium payoff to player A is:
1 1 2 2 (1 + δ)
1−δ + δ =
3 3 3 3 3
1
If δ = 2
we get that the expected payoff to both player A and B is 21 .
Player 1: If t odd (i.e., it is his turn to make an offer), propose (1, 0).
If t even, accept the offer if 2 proposes (1, 0), reject otherwise.
Player 2: If t odd, accept any offer.
If t even, propose (1, 0).
It is obvious that given player 2’s strategy, player 1 has no incentive to deviate
at each node of the game. Now consider a deviation of player 2. If player 2
rejects 1’s offer and instead proposes other than (1, 0), then provided δ > 3/4,
player 1 will reject the offer knowing that he can get 1 next period (we use
the “one period only deviation” principle). Hence, player 2 has nothing to gain
from deviation. In contrast with the original Rubinstein game, player 2 cannot
offer (1 − ε, ε) which will be accepted by 1, where ε is an arbitrary small positive
number.
To show ( 34 , 14 ) is a SPE consider the following strategies:
2
Player 1: If t odd (i.e., it is his turn to make an offer), propose ( 43 , 14 ).
If t even, accept the offer if 2 proposes (m, 1 − m), m ≥ 3/4; reject other-
wise.
Player 2: If t odd, accept the offer if 1 proposes (m, 1 − m), m ≤ 3/4.
If t even, propose ( 34 , 41 ).
3. The unique mixed strategy Nash equilibrium of game A is (D, D) with payoff
(0, 0).
The three mixed strategy Nash equilibria of game B are (S, S) and (B, B) both
with payoff (4, 4), and then 21 , 12 with expected payoff (2, 2).
The Subgame perfect equilibrium strategies for the dynamic game are:
• Play D in period 1.
• If in period 1 the outcome of the game is (D, D) then play S in period 2.
3
• If in period 1 the outcome of the game is different from (D, D) then each
1
player randomize with probability 2
on S and B in the second period.
1 1
Notice that it is subgame perfect to punish deviations in period 2 since ,
2 2
is a Nash equilibrium of the period 2 subgame. Moreover the strategies above
are Subgame Perfect if and only if:
1−δ 1−δ
2 (4 + δ 4) ≥ (5 + δ 2)
1−δ 1 − δ2
or
1
δ≥ .
2
This Subgame Perfect equilibrium is such that the average discounted payoff to
each player is:
1−δ
(4 + δ 4) = 4.
1 − δ2
4. The set of feasible payoffs is the convex hull of (0, 0), (1, 2) and (2, 1). This is the
set
{(v1 , v2 ) | v1 ≤ 2v2 , v2 ≤ 2v1 , and v1 + v2 ≤ 3}
The highest symmetric payoff is 23 , 32 as the sum of the parties’ payoffs is lower
Notice that the average payoffs oscillate around the point 32 , 32 infinitely often
3
because whenever player i’s payoff is above 2
it is at most two periods before if
decreases below 23 . (Using δ ≥ 9
10
we can verify that (S, S), (B, B), (B, B) has
4
3
average payoff lower than 2
for player 1.)
The average payoff of player 1, as well as the payoff of player 2, converge. Hence,
they must converge to 23 , 32 . No player will deviate as he loses utility in both
5. In the case at hand: maxa gi (a) = 5, v i = 0, δ = 0.9 and the full dimensionality
assumption is satisfied.
The conditions that the punishment period length T and the payoff vi0 (v i <
vi0 < vi ) need to satisfy are:
(1 − δ) max gi (a) + δ(1 − δ T )v i + δ (T +1) vi0 < (1 − δ N )wij + δ N (vi0 + ε). (4)
a