Econ 1052: Problem Set 4 (With Solutions)

The solutions are detailed for learning purposes.

Problem 4: Designing an optimal selling mechanism

A monopolist is selling a divisible good to a buyer, and can sell any quantity q ∈ [0, 2]. The
buyer has private taste parameter v distributed according to CDF F : [0, 1] → [0, 1]. Assume
that F is strictly increasing, continuous, and regular, with associated density f .
If the buyer consumes quantity q and pays price p, their utility is

vq − p,

and the monopolist’s profit is

q2
p− .
2
The second term captures an increasing marginal cost of production. Observe that social
welfare is maximized by setting vi = q.
Your problem is to design an incentive-compatible and individually-rational selling mecha-
nism, consisting of a quantity function q : [0, 1] → [0, 2] and a pricing function p : [0, 1] → R.
The goal is to find an optimal mechanism, that is, one that maximizes the monopolist’s
expected profit.

1. Write down the buyer’s payoff when facing such a mechanism, which depends on their
true value v, their reported value v̂, and the functions p and q.

2. Using your last answer and the envelope theorem, write down an expression for the
utility U (v) of a buyer with type v, that depends on the lowest utility of a buyer with
type 0 and on the quantity function q.

3. If the mechanism maximizes the monopolist’s profits, we must have U (0) = 0. Explain
why.

1
 R1
4. Observe that the expected buyer utility is 0 U (s)f (s)ds. Substitute in your answers
from the previous parts, and use integration by parts and the Leibniz integral rule to
derive an expression for the expected buyer utility that includes the fraction 1−F (s)
f (s)
.

5. Observe that expected social welfare under the mechanism is

1
q(s)2
Z  
sq(s) − f (s)ds.
0 2

Observe also that the monopolist’s expected profit is equal to expected social welfare
minus expected buyer utility. Substitute in your answer from the last part, to derive
an expression for the monopolist’s expected profit that depends on the buyer’s virtual
value.

6. Write an equation that pins down the quantity provided to type v under the optimal
mechanism. Which types are provided with quantity 0? How does the optimal
mechanism’s quantity compare to the quantity that maximizes social welfare?

1. The buyer’s payoff when facing such a mechanism is

u(v, v̂) := v q(v̂) − p(v̂),

where v is the buyer’s true type and v̂ is their reported type.

2. Given a direct mechanism (q, p), a buyer of type v chooses a report v̂ to solve

max [v q(v̂) − p(v̂)] .

v̂∈[0,1]

Let v ∗ (v) denote an optimal report for type v, and define the indirect utility

U (v) := max [v q(v̂) − p(v̂)] = u v ∗ (v), v .

v̂∈[0,1]

Since
∂u(v̂, v)
= q(v̂),
∂v
we obtain, by the envelope theorem
Z v
q v ∗ (z) dz.


U (v) = U (0) +
0

2
3. We now show that for a profit-maximizing, incentive-compatible and individually
rational mechanism we must have U (0) = 0.
Clearly, if U (0) < 0, then individual rationality is violated for type 0, so such a
mechanism cannot be feasible.
Suppose instead that an optimal mechanism has U (0) > 0. Consider a deviation in
which the seller increases all payments by U (0):

p̃(v) := p(v) + U (0) for all v.

The allocation rule q is unchanged. The resulting utility of type v is

Ũ (v) = v q(v) − p̃(v) = v q(v) − p(v) − U (0) = U (v) − U (0).

Since q(v) ≥ 0 for all v, the envelope formula implies that U is nondecreasing, so
U (v) ≥ U (0) for all v. Hence Ũ (v) ≥ 0 for all v, and the new mechanism remains
individually rational. Incentive compatibility is preserved because subtracting the
same constant from each type’s payoff does not change any incentives to misreport.
However, the seller’s expected profit increases by U (0), contradicting optimality of
the original mechanism. Therefore, any profit-maximizing IC and IR mechanism must
satisfy U (0) = 0.

4. Using the envelope formula with U (0) = 0, we have

Z s
U (s) = q(z) dz.
0

The expected buyer utility is

Z 1 Z 1 Z s 
U (s)f (s) ds = q(z) dz f (s) ds.
0 0 0

Noting that dF (s) = f (s) ds, and applying integration by parts,

Z 1  1
Z 1
U (s) dF (s) = U (s)F (s) 0
− F (s) dU (s).
0 0

Since U (0) = 0 and F (0) = 0, and F (1) = 1, we get

 1
Z 1
U (s)F (s) 0 = U (1)F (1) = q(z) dz.
0

3
 Moreover, dU (s) = U ′ (s) ds = q(s) ds, so
Z 1 Z 1
F (s) dU (s) = F (s) q(s) ds.
0 0

Therefore,
Z 1 Z 1 Z 1 Z 1
U (s)f (s) ds = q(s) ds − F (s) q(s) ds = (1 − F (s)) q(s) ds.
0 0 0 0

Finally, writing this in terms of the density f ,

1 1
1 − F (s)
Z Z
U (s)f (s) ds = q(s) f (s) ds.
0 0 f (s)

5. The monopolist’s expected profit Π is equal to expected social welfare minus expected
buyer utility. Expected social welfare under the mechanism is
1
q(s)2
Z  
s q(s) − f (s) ds,
0 2

and from the previous part the expected buyer utility is

1
1 − F (s)
Z
q(s) f (s) ds.
0 f (s)

Hence
1 Z 1
q(s)2
 
1 − F (s)
Z
Π= s q(s) − f (s) ds − q(s) f (s) ds.
0 2 0 f (s)
Rearranging terms,
1
q(s)2
  
1 − F (s)
Z
Π= s− q(s) − f (s) ds.
0 f (s) 2

Define the buyer’s virtual value as

1 − F (s)
ϕ(s) := s − .
f (s)

Then we can write the monopolist’s expected profit as

1
q(s)2
Z  
Π= ϕ(s) q(s) − f (s) ds.
0 2

4
6. From the previous part, the monopolist’s expected profit can be written as
1
q(v)2
Z  
Π= ϕ(v) q(v) − f (v) dv,
0 2

where the virtual value is

1 − F (v)
ϕ(v) := v − .
f (v)
Since f (v) > 0 on [0, 1], and given an incentive-compatible allocation rule q, the
integrand
q(v)2
ϕ(v) q(v) −
2
is the contribution to profit from type v. Ignoring for the moment global incentive
constraints, for each type v the monopolist would like to choose q(v) to solve the static
problem
q2
 
max ϕ(v) q − .
q∈[0,2] 2
The objective is strictly concave in q, so the first-order condition characterizes the
maximizer whenever the interior solution lies in [0, 2]:

q2
 
∂
ϕ(v) q − = ϕ(v) − q = 0 =⇒ q = ϕ(v).
∂q 2

The constraint q ≥ 0 implies that if ϕ(v) < 0 the solution is instead q = 0. The upper
bound q ≤ 2 is never binding here, because v ∈ [0, 1] and

1 − F (v)
ϕ(v) = v − ≤ v ≤ 1 < 2.
f (v)

Therefore, pointwise profit maximization yields the quantity rule

 
∗ 1 − F (v)
q (v) = max{0, ϕ(v)} = max 0, v − .
f (v)

Under the regularity assumption, the virtual value ϕ(v) is nondecreasing in v, so q ∗ (v)
is also nondecreasing. Hence q ∗ satisfies the monotonicity requirement that underlies
incentive compatibility.
The types that receive zero quantity under the optimal mechanism are precisely those

5
for which the virtual value is nonpositive:

1 − F (v)
q ∗ (v) = 0 ⇐⇒ ϕ(v) ≤ 0 ⇐⇒ v− ≤ 0.
f (v)

If there is a unique cutoff type v0 ∈ [0, 1] such that ϕ(v0 ) = 0, then all types v < v0
receive q ∗ (v) = 0, while types v > v0 receive a strictly positive quantity.
For comparison, the quantity that maximizes social welfare for a given type v solves

q2
 
max v q − ,
q∈[0,2] 2

yielding the first-order condition v − q = 0 and hence

q SW (v) = v

(again, the upper bound 2 is not binding since v ≤ 1). Because

1 − F (v)
ϕ(v) = v − ≤v for all v,
f (v)

we have
q ∗ (v) = max{0, ϕ(v)} ≤ v = q SW (v) for all v,

with a strict inequality for all types with ϕ(v) > 0. Moreover, types with ϕ(v) ≤ 0 receive
q ∗ (v) = 0 under the optimal mechanism, even though the socially efficient quantity for
those types would be q SW (v) = v > 0. Thus, relative to the social-welfare–maximizing
allocation, the optimal selling mechanism distorts quantities downward, especially by
excluding low-virtual-value types from trade altogether.