Homework #5 Solutions

1. 1. As we are looking for symmetric BNE, suppose that in the equilibrium both players play
bidding strategy b that each type vj bids b(vj ) for some increasing differentiable function.
From player i’s point of view, if he bids bi , his expected payoff is (given that the other player
follows bj = b(vj ))

b−1
- (bi )
U (bi |vi ) = bi P r (bi < bj ) + {vi − b (vj )} dvj
0
b−1
- (bi )
= bi 1 − b−1 (bi ) + {vi − b (vj )} dvj
0

Taking FOC with respect to bi , we get

� �
−1 −1 1
1 − b (bi ) + bi � −1 + vi − b b−1 (bi ) =0
b (b (bi )) b� (b−1 (bi ))

This must be satisfied at bi = b(vi ). Thus, we have

b� (vi ) (1 − vi ) − b (vi ) + vi − b(vi ) = 0

In particular, we are looking for linear BNE: let b (vi ) = αvi + β. Using this, we have

(1 − 3α) vi + (α − 2β) = 0

1
So α = 3
and β = 16 . Thus, the symmetric linear BNE of this game is b(vi ) = 13 vi + 16 .
Comment: the professor made an announcement at the lecture that this game is NOT a
second price auction as the winner pays the other sibling’s bid to him/her (so if you lose, you
get what you did bid). In case you did not interpret the question in this way and reached
the conclusion that bidding one’s own value is BNE, I gave half points (5 points) assigned
for the problem.

2. 2. Writing as a Bayesian game:

N = {1, 2, · · · , n}, Ti = vi ∼ F , Ai = bi ∈ [0, ∞), payoff is vi − bi if win the auction and zero
otherwise. Types (individual values) are iid with pdf f and cdf F .
Computing the symmetric BNE in increasing differentiable strategies:
Suppose that in the symmetric BNE all players play bidding strategy b that bj = b(vj ). From
player i’s point of view, if he bids bi , his expected payoff is (given that all other players follow
bj = b(vj ))

(vi − bi ) P r [bi ≥ bj ∀j = i] = (vi − bi ) P r b−1 (bi ) ≥ vj ∀j = i

n−1
= (vi − bi ) F b−1 (bi )

1
 Taking FOC with respect to bi , we get

n−1
n−2 1
− F b−1 (bi ) + (vi − bi ) (n − 1) F b−1 (bi ) f b−1 (bi )
 

=0
b0 (b−1 (b i ))

Substituting bi = b(vi ) (as at the optimum, player i should also play bidding strategy b), we
get
1
− [F (vi )]n−1 + (vi − b(vi )) (n − 1) [F (vi )]n−2 f (vi ) 0 =0
b (vi )
We can rewrite this as
d 
b(vi )F (vi )n−1 = (n − 1)vi F (vi )n−2 f (vi )

dvi

Integrating both sides and since b(vi )F (vi )n−1 = 0 when vi = 0 (if value is zero you bid zero),
we get

Zvi
1
b(vi ) = (n − 1)tF (t)n−2 f (t)dt
F (vi )n−1
0

Using integration by parts, we get

R vi
0
F (v)n−1 dv
b(vi ) = vi −
F (vi )n−1

Comments: Do not forget to write this as a Bayesian game, as those are easy points at the
exam. I did not take points off for this pset grading, but if it were the exam if you missed
this part you could have lost 20-40% of the points for this problem.
3. (a) If θ < 0, the only equilibrium is to play Stag.
If θ = 0 there are equilibria where any number of players between 0 and k − 2 play Stag,
and where everyone plays stag.
For 0 < θ ≤ v, the equlibria are everyone plays Stag and everyone plays Rabbit. In
addition, if θ = v there are equilibria where any number of players between k and n
play Stag, and the rest play Rabbit.
If θ > v, then everyone plays Rabbit is the only equilibrium.
(b) There will be one monotone BNE. The value of Stag is decreasing in xi , and the value of
Rabbit is increasing in xi . For a given x∗ , there will be one x̂ that is indifferent between
Stag and Rabbit. This x̂ must equal x∗ for this to be a NE. x̂ is monotone increasing
x∗ , so there will be one BNE.
To find that BNE, consider some x∗ . If xi = x∗ , then he is indifferent to Stag and Rabbit.
He has expected value of x∗ when playing Rabbit. His expected value of playing Stag
is vP (xj < x∗ for k players). By the hint, the probability that k players have lower xj
than our x∗ is 1 − k/n, so our expected value simplifies to v(1 − k/n). Thus, for xi = x∗
to be indifferent, it must be x∗ = v(1 − k/n).

2
4. (a) In equilibrium, A picks p = 7000 after Peach and p = 2000 with probability 1/4,
p = 7000 with probability 3/4 after Lemon. After seeing 2000, Bob believes Lemon
with probability 1, and after 7000 Bob believes Lemon with probability 3/7 and Peach
with 4/7. These beliefs are derived from bayesian updating. After 2000 Bob accepts,
and after 7000 Bob accepts with probability 1/6. Bob is indifferent between accept and
reject after 7000, because he has expected utility 0. And Alice is indifferent between
2000 and 7000 after Lemon, because he has expected utility 2000 in either case.
To see consistency, consider θm = ((1−f)7000+f2000, (0.25)2000+0.757000; 1/6accept+
5/6reject, (1 − f)accept + freject), where the actions specified are for player A after
peach, player A after Lemon, Player B after seeing 7000 and after seeing 2000. Then
beliefs for player 2 are b(Lemon|2000) = 0.75/(0.75 + f) → 1 and b(Lemon|7000) =
0.75/(0.75 + (1 − f)) → 3/7.
(b) One equilibrium is A picks p = 5000 in either state. After any p ≤ 5000, B believes
that it is lemon with probability 1/2 and accepts. After any p > 5000, B believes that
the state is Lemon with probability 0.95 and rejects. For this to be consistent, we must
have σ m , β m that converge to σ and b. Define σ m to be:
Type Lemon plays
p = 5000 with probability 1 − 2f
uniform(0,5000) with probability f
5000+X, where x ∼ exponential distribution(1) with probability f
Type Peach plays
p = 5000 with probability 1 − 20f
uniform(0,5000) with probability f
5000 + Y, where X ∼exponential (1) with probability 19 f
β m (Lemon|p < 5000) = 1/2, and β m (Lemon|p > 5000) = 0.95, β m (Lemon|p = 5000) =
(1 − 2f)/(2 − 22f). With f → 0. In the limit, p = 5000 and b = limH→0 β m .

3
MIT OpenCourseWare
http://ocw.mit.edu

14.12 Economic Applications of Game Theory

Fall 2012

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.