Title | Final Exam Solns 212 Fall 2020 |
---|---|
Course | Game Theory |
Institution | University of Pennsylvania |
Pages | 4 |
File Size | 217.9 KB |
File Type | |
Total Downloads | 95 |
Total Views | 162 |
Solutions to practice questions for the course...
Economics 212
Steven A. Matthews
Fall 2020
University of Pennsylvania
Final Exam Solutions
Open book, notes, and the Canvas course page. Do not communicate with anyone or consult the internet except Canvas. Indicate your reasoning. 2 hours, 90 points.
1. (20 pts) Consider the following normal form game. L
C
T
0; 2
0; 2
M
0; 1
1; 1
B
1; 0
0; 1
R
1 1 ;
0; 3
0; 0
(a) (4.5 pts) Redraw the matrix, but without payo¤s. Instead, in each box write “PO” if the corresponding strategy pro…le is Pareto optimal, and leave the box blank if it is not Pareto optimal.
(One-half point for each correctly
labeled box.) Soln:
(M ; C ) and (M ; R ) are e¢cient: any move away from one of them
will make at least one player worse o¤. No other pro…le is e¢cient: for any other pro…le, a move to either (M ; C ) or (M ; R) will make at least one player better o¤ and no player worse o¤. (b) (7 pts) Find all rationalizable strategy pro…les. Soln: T
is strictly dominated by any mixture of
it. In the resulting game,
L
M
and
B;
so we can delete
is strictly dominated by any mixture of
so we can now delete it. The resulting 2
C
and
R;
2 game has no strictly dominated
strategies. Hence, the rationalizable pro…les are
(M ; C ); (M ; R ); (B; C ); and (B; R ):
(c) (8.5 pts) Find all pure and mixed strategy Nash equilibria. Soln:
We know that a strategy pro…le is a NE i¤ it is a NE of the game
consisting of just the rationalizable strategies: C
R
M
1; 1
0; 3
B
0; 1
0; 0
The only pure strategy equilibrium is (M ; R): There is no equilibrium in which player 2 plays
C
For, if
M;
to
M
2 (C ) >
is
R;
0; player 1’s best reply would be
contrary to
2 (C ) >
0:
with positive probability. and player 2’s best reply
But there are mixed strategy equilibria of the form (p; R); where Any
p
2
[0; 1] is a best reply to
best reply to
i¤
p
Hence, (p; R) is a NE i¤
p
2
3p
R;
since
u1 (M ; R)
, 1
=
u1 (B; R ):
p
=
1 (M ):
But
R
is a
1 3:
p
1
[ 3 ; 1]:
1 = 2 ; play the in…nitely repeated game based on the stage game shown below. Let hC = ((C; C ); (C; C ); : : :)
2. (20 pts) Two players, each with the discount factor
denote the outcome in which (C; C ) is played forever. Find all values of
which the grim trigger strategy pro…le that has outcome
C
h
equilibrium outcome.
Soln:
C
D
C
10; 10
0; 6
D
11; 0
x; x
In any subgame that follows the play of a
D;
for
x
is a subgame perfect
the grim trigger pro…le calls
for (D; D ) to be played all the time, i.e., at any date after any history of that subgame. Thus, in such a subgame a player should always play his myopic best reply to
since he cannot a¤ect the future play of the other player.
D;
Hence,
(D; D ) must be an equilibrium of the stage game. This gives us one restriction: x
0:
Now, player 1 will not want to make a one-shot deviation from playing
C
if and
only if
10 1
11 +
, , x
1
20
9
11 + x
x;
1 = 2 : Similarly, player 2 will not want to make a one-shot deviation from playing C if and only if
using the fact that player 1’s discount factor is
10 1
Conclusion:
6+
, , x
1
20
6+x
14
x:
The grim trigger pro…le is a SPE if and only if
x
2
3. (25 pts) Consider an industry with the inverse demand function where
Q
is the sum of the outputs
q1
and
q2
[0; 9]: P (Q )
=
a
Q;
of the two …rms in the industry.
Each …rm can produce any nonnegative amount of the output at zero cost. In addition, before they make their output decisions, …rm 1 can advertise to increase demand: it chooses 1
= (a
q1
q 2 )q 1
1 = 48 a3 : Thus, given (a; q1 ; q2 ) their payo¤s are C (a) and 2 = (a q1 q 2 )q 2 :
at cost
a
C (a)
(a) (5 pts) Suppose …rm 1 chooses that …rm 2 sees both
q1
and
a
q1
at the same time as it chooses
before it chooses
q2 :
a;
and
Describe the nature of a
strategy for each …rm in this game. Soln:
A strategy for …rm 1 is a pair of numbers, (a; q1 ): A strategy for …rm
2 is a function possible (a; q1 ):
s2
:
R2+ ! R that speci…es an output 2
q2
=
s2 (a; q1 )
for each
(b) (20 pts) Find the subgame perfect equilibrium, and its outcome, of the game in (a). Denote the SPE as (a ; q1 ; s2): Use backwards induction to …nd it. As
Soln:
shown in the lecture slides, the sequentially rational strategy of …rm 2 is s
1 2 (a; q1 ) = 2 (a
q 1 ):
(1)
Firm 1’s best reply to s2 is found by solving the problem max(a a;q1
s2 (a; q1 ))q1
q1
C (a)
= max 1 aq1 2
a;q1
1 2 2 q1
1 3 48 a :
At the solution, the derivatives of the objective function with respect to and
q1
are equal to zero:
a
3 2 48 a = 0;
1 2 q1 1 2a
= 0:
q1
The solution of these two equations is (a ; q1 ) = (4; 2): So the SPE is (a ; q1; s2 ) = (4; 2; s2 ); where s2 is the function de…ned in (1). The SPE outcome is
(a
;q ;q ) 1 2
= (a
; q1 ; s2 (a ; q
1 )) = (4; 2; 1):
4. (25 pts) Two bidders participate in an all-pay common-value auction for an oil tract. Bidder
i
privately observes a signal s ~i : The signals s ~1 and s ~2 are uniformly
and independently distributed on [0; 1]: The value v ~ of the oil tract to whoever wins it is the product of the signals: v ~
If bidder is v ~
b:
bids
i
b
and wins (i.e.,
b
If she loses her payo¤ is
=s ~1 s ~2 :
is larger than the other bidder’s bid), her payo¤
b:
There is a strictly increasing di¤erentiable bid function is an equilibrium for each bidder
i
to bid
b (s i )
b
: [0; 1]
!R
such that it
when her signal is s ~i =
Derive
si :
this bid function. Soln: b(s)
1
= 3 s3 :
^ and bidder 2 uses the bid function bidder 1 bids b; ^ b( ): Then bidder 1 wins if b(~ s2 ) < b; or rather, if s ~2 < b1 (^ b) (since b( ) is strictly
Derivation.
Suppose s ~1 =
s;
increasing). Thus, the probability that bidder 1 wins is
Pr
s ~2 < b
1 ^ (b)
=
F
b
1 ^ (b)
=
b
1 ^ ( b ):
The expected value of the oil tract to bidder 1 conditional on s ~1 = winning, i.e., on s ~ < b1 (^b); is
s
and on
2
E
h s ~1 s ~2
j
s ~1
1 ^ (b)
=
s; s ~2 < b
=
s; s ~2 < b
i
b
1 ^ ( b)
=
s
b 1 ^ ^ b = b ( b )s
2
:
So bidder 1’s expected payo¤ is
Pr
^ b(~ s2 ) < b
E
h s ~1 s ~2
j
s ~1
3
1 ^ (b)
i
1 ^ (b)
2
^
b:
Let s ^ be the signal bidder
1
^ were to be her equilibrium would have to observe if b
bid, i.e., let s ^ be de…ned by b(^ s) the expression above for bidder terms of s ^
: s ^s
For b(
) to
s ^
2
= ^ b:
1’s
Then s ^
2 1 ss 2 ^
be the equilibrium bid function, s ^
0=
d ds ^
1
2 ^ 2 ss
b
1 (^
b): Substituting these into
expected pro…t gives us her expected pro…t in
b(^s) =
pro…t. The …rst-order condition s ^
=
=
b(^s): =
s must maximize this expected
s must satisfy is
b(^s) s^=s = [ss^ b(^s)]js^=s = s2 b0 (s):
Hence, b for all s
(s ) =
s
2
2 [0; 1]: Integrating both sides yields b (s )
If b(0) >
0;
b(0) =
a bidder with signal s
in which case she’d prefer to bid b(0)
0
= 0:
0
Z
= 0
s t
2
0
dt
would get the negative pro…t
to get
0
pro…t.
Thus, the equilibrium bid function is b(s)
4
1
= 3 s3 : 0
b(0),
This contradiction implies
1
= 3 s3 :...