Micro II - Problem Set 7 - Suggested Solution

Problem 1: MWG 8.D.4. In this problem we analyze a bargaining game where each
players strategy is their request di . If the sum of requests exceeds 100 both players get
nothing, while if d1 + d2 100 each gets what they asked for. We will again use Player 1
as a reference.
a) Just like in the previous problem, we cannot have strictly dominated strategies
because for any d1 there exists a d2 > 100 d1 that will guarantee an outcome of
zero.
b) All di > 100 are weakly dominated because they guarantee the zero payoff, whereas
d 100 may lead to a positive one.
c) All cases where:
d1 = x,

d2 = 100 x,

where x [0, 100]

are Nash equilibria. So are all cases where d1 , d2 > 100 because there are no profitable deviations.

Problem 2: MWG 8.D.5. For simplicity of notation, we will represent the broadwalk
as the real segment [0, 1]. The strategy of each vendor, therefore, is a location xi [0, 1].
Vendor 1 is our reference.
a) Given x2 , the best response is to place x1 an infinitesimal amount  to the side with
more customers. That is:

x2 +  if x2 < 0.5
x1 = R1 (x2 ) =

x2  if x2 > 0.5

x
if x2 = 0.5
2

It is clear from the definition of the response function and Figure 1, that the only
NE is (x1 , x2 ) = (0.5, 0.5).
b) We can prove this by showing that a profitable deviation for at least one of the
vendors always exists:
1

 All three vendors at the same spot, each selling to 1/3 of the market cannot
be an equilibrium, because a vendor who moves  in the direction of the longer
portion of the market is guaranteed at least 1/2  > 1/3.
Two vendors in one spot and the third in another one cannot be an equilibrium
either. The vendor who is alone can move towards the two others, preserving
the market between him and the closer side while taking customers from the
other two. This will go on until they are practically all 3 in the same place
again. We have shown above that this cannot be an equilibrium.
Each vendor in a different location cannot be an equilibrium either, because the
ones that are on the sides can increase their market share by moving towards
the middle vendor. This will again lead to the same situation as in the first
case, which is not an equilibrium.
These three cases exhaust all possibilities and we have shown that none of them can
be an equilibrium, therefore, there is no equilibrium in this game.

Problem 3: MWG 8.D.8. We are asked to prove that we cannot have (non-degenerate)
mixed strategy Nash equilibria when we have convex strategy set Si and strictly quasiconvex utility function ui (). We can prove this by contradiction. Assume that {i , i }
is a non-degenerate mixed strategy Nash equilibrium, that is, strategies s1i and s2i are
both given positive probabilities. For this to be the case, s1i and s2i must both be best
responses to i at the same time. If we take a convex combination si of these two
strategies, by quasiconvexity of ui ():
ui (s1 , i ) = ui [s1i + (1 )s2i , i ] > ui (s1i , i ) = ui (s2i , i )
2

we see that it contradicts the definition of best response. Hence we have proven that a
mixed strategy cannot be a best response under these assumptions.
Problem 4: Modified Battle of the Sexes Game. This game is like the original
Battle of the Sexes, only the woman is trying to avoid the man.1
a) The idea is to realize that each player has four possible utility levels: both attending
their favorite event and achieving their goal with the other player (meeting or avoiding), being able to realize either one of these two or the other, and not being able
to realize any. By assumption, the social event is of secondary importance. Hence,
M
the following payoff matrix captures all of these features:

S
W

0, 3 2, 0

3, 1 1, 2

b) The best responses are already marked on the matrix and we can see they do not
intersect. This is sufficient to claim that there is no pure strategy NE.
c) Assume the woman plays S with probability p, while the man plays it with probability q. For the man to be mixing strategies, it must must hold that:
uM (S) = 3p + 1(1 p)
uM (B) = 0p + 2(1 p)
Equality requires that p = 1/4. By the same token, for the woman to be mixing
strategies:
uW (S) = 0q + 2(1 q)
uW (B) = 3q + 1(1 q)
For these expected payoffs to be equal we need q = 1/4. Figure 2 shows the response
functions in mixed strategies. The non-existence of pure strategy equilibria and the
only mixed strategy Nash equilibrium can be clearly seen.
Problem 5: MWG 8.E.1. Each player can condition its strategy on his own type.
These strategies are:

A|S, A|W

A|S, N |W

N |S, A|W

N |S, N |W

For this reason it has also a Matching Pennies game flavor.

Table 1: Payoff matrix for Problem 1.

AN
NA

M
4

M
4

w+s M
2 , 4

AN

2s , M
2

w2 , 3M
4

NN

0, M

w+s
2

w+s
4

w+s
4

M
2

NA

w+s M
4 , 4

M s M s
4 , 4
M w M
4 , 2

0, M
2

s
2

3M
4
M
2

s
4

w+s
w
4 ,2

4s , M4s

NN
M, 0
M
2 ,0

M w M w
4 , 4

M
2 ,0

0, M
2

0, 0

With four strategies for each player there will be 16 possible outcomes. We need to
compute the expected payoff for each player in each of them. If both play AA (shorthand
for attack regardless of the type) then each player expects:

 



1 1
1
1 1
1
M
w+s
(s) + (M s) +
(w) + (w) =

2 2
2
2 2
2
4
2
The square brackets are probabilities the player is strong or weak respectively, and for
each of these outcomes we have the possibility of meeting a strong or weak opponent
(round brackets).

1 1
(s) +
2 2

If P1 plays AA while P2 plays AN then P1s expected payoff is:

2
2 2
2
2
4

while P2s payoff is:

2 2
2
2 2
2
4
2
The payoff matrix is shown in Table 1.

The Bayesian-Nash equilibria will depend on

the parameters of the game:

M > w > s and w > M/2 > s :

{AA, N N } {N N, AA}

w > M > s and s > M/2 :

{AN, AN } {AA, N N } {N N, AA}

w > M > s and M/2 > s :

{AA, AN } {AN, AA} {AN, AN }

Problem 6:
(a) Given quasi-linearity of preferences in the private good, the Pareto efficient amount
of the public good is determined by maximization of the sum of the two utilities, or
max 10 g + (1 + 2 ) g g 2 ,
g

giving g = (1 + 2 1)/2.
(b) In a Bayesian Nash equilibrium, each players solves


1
2
gi (i ) = arg max E 5 gi + i (gi + gj (j )) (gi + gj (j )) ,
gi
2
where gj () is the equilibrium strategy of player j and the expectation is taken over
the possible realizations of j . The FOC gives
gi (i ) = i 1 E[gj (j )].
As a result,
E [gi (i )] = E(i ) 1 E[gj (j )]
= 2.5 E[gj (j )].
As a result, the ex ante expected level of the public good is given by
E [g1 (1 )] + E[g2 (2 )] = 2.5.
The equilibrium is not unique in that E[gj (j )] can be anywhere in [0, 2]. Given the
realizations of 1 and 2 , the actually provided amount of the public good is given
by
g1 (1 ) + g2 (2 ) = 1 + 2 2 E [g1 (1 )] E[g2 (2 )]
= 1 + 2 4.5.
5

Note that
(1 + 2 1)/2 T 1 + 2 4.5
1 + 2 1 T 21 + 22 9
8 T 1 + 2 ,
and so the actual equilibrium amount of the public good falls short of the PE amount
unless 1 = 2 = 4, when the two amounts coincide.