Você está na página 1de 6

Problem Set 4: Sketch of Solutions

Information Economics (Ec 515) · George Georgiadis

Due in class or by e-mail to quel@bu.edu at 12:30, Monday, December 8

Problem 1. Screening
s2
A monopolist can produce a good in different qualities. The cost of producing a unit of quality s is 2. Consumers
buy at most one unit and have utility function
(
qs if they consume one unit of quality s
u(s|q ) =
0 if they do not consume

The monopolist decides on the quality (or qualities) it is going to produce and price. Consumers observe qualities
and prices and decide which quality to buy if at all.

1. Characterize the first-best solution.

2. Suppose that the seller cannot observe q, and suppose that


(
qH with probability 1 b
q=
qL with probability b

with q H > q L > 0. Characterize the second-best solution and consumers’ informational rent.

3. Suppose now that q is uniformly distributed on the interval [0, 1]. Characterize the second-best optimal
quality-pricing schedule.

Solution for Problem 1:

Part 1: Characterization of First-Best Solution Here we assume that the seller is perfectly informed about the
buyer’s characteristics. The seller can treat each type of buyer separately and offer her a type-specific contract
( Ti , si ) for each type qi . The seller solves the problem

max Ti s2i
Ti ,si

so that
qi si Ti 0.

Hence, differentiating with respect to si , yields

qi 2si = 0

1
which gives the first-best quality schedule
1
s(q ) = q.
2

Part 2: Two Types Using the revelation principle the seller’s problem is given by
h i h i
max b TL s2L + (1 b) TH s2H
Ti ,si

subject to

qL sL TL 0 (1)
qH sH TH 0 (2)
qL sL TL qL sH TH (3)
qH sH TH qH sL TL . (4)

Using the same procedure as in section 2.1.3, we can eliminate (ICL) and (IRH) and observe that (ICH) and (IRL)
will bind at the optimum. Thus, the reduced unconstrained problem is

max b(q L s L s2L ) + (1 b)[q H s H Dqs L s2H ].


s H ,s L

where Dq ⌘ q H q L . Differentiating with respect to s L and s H and accounting for the non-negativity of s, we
obtain

1 1 b
sL = max qL Dq, 0
2 2b
1
sH = qH
2

so that the quality for the low type is inefficiently low, but efficient for the high type. The transfers are given by

TL = max {q L s L , 0}
TH = qH sH s L Dq.

The informational rents are

uL = 0
uH = s L Dq 0.

There are no rents left to the low type, but the high type earns an informational rent as long as the low type is not
excluded from the market. There may be a corner solution, where the monopolist chooses to exclude the low type,
if the informational rent the monopolist has to pay becomes too large.

Part 3: Continuum of Types The seller’s problem can be written as follows:


ˆ 1 h i
max T (q ) s(q )2 f (q )dq
s(q ),T (q ) 0

2
subject to

qs(q ) T (q ) 0 (5)
qs(q ) T (q ) qs(qb) T (qb) for all qb 2 [0, 1]. (6)

Following the implementation and optimization problems described in section 2.3.3.1 and 2.3.3.2 we can rewrite
the seller’s problem as
ˆ 1 nh i o
max qs(q ) s ( q )2 f ( q ) s ( q ) [1 F (q )] dq.
s(q ) 0

The first order condition yields


[q 2s(q )] f (q ) [1 F (q )] = 0.

Rearranging this equation we obtain 


1 1 F (q )
s(q ) = q
2 f (q )
which, using the fact that q ⇠ U [0, 1], simplifies to

1
s(q ) = 0 for all q <
2
1 1
= q for all q ,
2 2

since quality cannot be negative. The pricing schedule is given by


ˆ q
T (q ) = qs(q ) s( x )dx.
0

Since s(q ) = 0 for q < 12 , T (q ) = 0 for all q < 12 . For all q 1


2, we have
ˆ q
T (q ) = qs(q ) s( x )dx
1
2
q
1 1
ˆ
= q (q ) (x )dx
2 1
2
2
q q2 1 q 1
= q2 + +
2 2 8 2 4
q2 1
= .
2 8
1
In this equilibrium, the monopolist sells only to consumers whose q is above 2, and thus there are efficiency
losses associated with this equilibrium relative to the perfect discrimination case. Servicing low-value consumers
forces the monopolist to either charge lower prices to the high types or to raise quality. Since quality is costly, the
monopolist finds it profitable to shut off part of the market in order to raise the profits from the other segment of
the market.

Problem 2. Screening #2

John has a grape farm and sells grapes to a liquor company. The relationship between John and the liquor company
is mutually exclusive: there is no outside options for both. The liquor company produces wines, and its profit

3
function is

p ( x, a) = 2a2 (a + 3 x )2 ,

where x is the amount of grapes supplied by John, and a is the type of the liquor company, which is not known
to John. We assume that Pr { a = 1} = Pr { a = 4} = 0.5. John offers a contract, which is represented by a function
T : R ! R where T ( x ) is the money paid by the liquor company when John supplies x amount of grapes. T can
be negative; John subsidizes the liquor company. We assume that there is no cost to produce grapes. Thus, John
wants to maximize the expectation of T ( x ).

1. Suppose that a is known to John. What is the optimal contract?

2. Suppose that a is not known to John. What is the optimal contract when John wants to supply grapes to both
types?

3. Suppose that a is not known. What is the optimal contract when John wants to accommodate the high type?

4. Compare John’s profits in parts 2 and 3. Discuss.

Solution for Problem 2:

Part 1: The profit maximization problem is

max Ta
x a ,Ta

subject to p ( x a , a) Ta 0.

Clearly, the constraint must be binding. Therefore, the first-best solution become ( a + 3, 2a2 ).

Part 2: When a is known only to John, the problem becomes

max 0.5T1 + 0.5T4


x1 ,T1 ,x4 ,T4
s.t. p ( x1 , 1) T1 0 ( IR1 )
p ( x4 , 4) T4 0 ( IR4 )
p ( x1 , 1) T1 p ( x4 , 1) T4 ( IC1 )
p ( x4 , 4) T4 p ( x1 , 4) T1 ( IC4 )

Following the standard argument, one can show that (IR 1) and (IC 4) are binding. One can also show that as long
as x4 x1 , (IR 4) and (IC 1) can be ignored. Then the problem can be reformulated as

max 0.5T1 + 0.5T4


x1 ,T1 ,x4 ,T4
s.t. p ( x1 , 1) T1 = 0 ( IR1 )
p ( x1 , 1) T1 p ( x4 , 1) T4 ( IC1 )

4
From the first two constraints, we have

T1 = 2 ( x1 4)2
T4 = ( x1 7)2 ( x4 7)2 ( x1 4)2 + 2.

Then, by plugging this into the objective function and ignoring the last constraint, one can find that x1 = 7 and
x4 = 7, which satisfies the last constraint. The corresponding Ts are T1 = 7 and T4 = 29.

Part 3: John will just maximize the profit of the high type and collect it through the transfer. Hence he sets

T = 32 = max p ( x, 4) = 32 (x 7)2 at x = 7.
x

Note that one has to verify whether the low type has no incentive to accept this offer. Indeed, if he accepts it, then
he gets p (7, 1) 32 = 39.

Part 4: John’s expected payoff in Part 2 is 11. On the other hand, his profit in Part 3 is 16. Therefore, John is strictly
better off selling his fruits only to the high type.

Problem 3. Auctions

A decision where to locate a hazardous waste dump is taken through an auction between n towns in a given
country. Call di town i’s disutility from taking on the dump. Assume the di ’s are uniformly and independently
distributed on [0, 1]. Call Ti the transfer the town requests from taking the hazardous waste dump. The lowest
bidder gets the dump and receives its requested transfer, which is paid equally from the other towns. Compute
the symmetric equilibrium of this auction. Is this an efficient allocation mechanism? Discuss.

Solution for Problem 3

Let b i (·) be the bid function of town i and bi = b i (di ) be the actual bid for disutility di . Denote
✓ ◆
G (di ) = Pr min d j  di .
j 6 =i

First, we appeal to the revelation principle such that the towns only report their type dˆi . That is, when submitting
its bid, town i solves the following problem
h i 1
max 1 G (dˆi ) ( b i (dˆi ) di ) G (dˆi ) E[b i |d i  dˆi ]
dˆi N 1

where
b i = min b j
j 6 =i

and
d i = min d j .
j 6 =i

We focus on a symmetric equilibrium so b i (di ) = b(di ). Furthermore, we assume that the bidding function is
increasing in di . Using the inverse bidding function b 1 (bi ) = dˆi for the reported types we can rewrite this problem
as h ⇣ ⌘i
1 1 1 1
max 1 G b ( bi ) ( bi di ) G( b (bi )) E[b i |d i b (bi )]
bi N 1

5
, 1 (b
h i 1
ˆ b i)
1
max 1 G( b (bi )) (bi di ) b( x ) g( x )dx
bi N 1 0

where g is the probability density function of d i . Taking the derivative with respect to bi we obtain the first order
condition
1 1 bi di 1 1 1 1
1 G( b (bi )) g( b (bi )) 1 (b
= b( b (bi )) g( b (bi )) 1 (b
b0 ( b i )) N 1 b0 ( b i ))

Substituting for the inverse bidding function we obtain

1
[1 G (di )] b0 (di ) g ( di ) [ b ( di ) di ] = b ( di ) g ( di )
N 1

where we also used that in equilibrium bi = b(di ). We now use the distributional assumptions to simplify this
2
differential equation. Since di ⇠ U [0, 1], we have 1 G ( d i ) = (1 di ) N 1 and g(di ) = ( N 1) (1 di ) N . Using
these expressions we find that after some simplification

2 2
(1 di ) N 1 0
b ( di ) N (1 di ) N b ( di ) = (N 1) (1 di ) N di

This differential equation can be solved by multiplying both sides by (1 di ) which gives us
h i
1 1
(1 di ) N b 0 ( di ) N (1 di ) N b i ( di ) = ( N 1) (1 di ) N di .

Integrating both sides we have

1
∂ ⇣ ⌘ 1
ˆ ˆ
N 1
(1 x ) b i ( x ) dx = (N 1) (1 x) N xdx.
di ∂x di

Integrating the right hand side by parts and after some algebra we obtain

N 1
(1 di ) N b ( di ) = (1 di ) N ( Ndi + 1) .
N ( N + 1)

Rearranging this equation we find


N 1
b ( di ) = ( Ndi + 1) .
N ( N + 1)
All that remains to check is that the bidding function is increasing which indeed it is for N 2

N 1
b 0 ( di ) = > 0.
N+1
N 1
Notice that the town with disutility of 1 will request N to take the dump whereas the firm with disutility of 0
N 1
will only request N ( N +1)
. The bids are increasing in the disutility di . This implies that the town with the lowest
disutility ends up with the dump since it demands the lowest compensation for accepting the dump. Finally, since
the bids are just transfers between the towns the mechanism is efficient.