Notes on Monopoly Pricing

The Basic Monopoly Problem

• Recall the basic monopoly profit-maximization problem:

max pD(p) − C (D(p))

p

• Here, D(p) is the market demand the monopolist faces, C(q) is the cost

function

– Assume C(0) = 0; C 0 (q) > 0; C 00 (q) > 0

q2
∗ Example: C(q) = 2

• We assume in the basic setup that the monopolist can only choose a con-

stant per-unit price p for all consumers

– This causes inefficiently low quantity supplied, and inefficiently high

price

• If the monopolist can choose different prices to different consumers, how

does that change the outcome?

From Utility to Demand

• Let’s take a closer look at where the demand D(p) comes from

1
 • One convenient way is to say there are consumers with different “tastes”,

denoted by θ, for the monopolist’s product

• So, the consumer with taste θ gets the following utility from consuming q

units of the monopolist’s product:

V (q, θ) = θq

• Suppose, for convenience, that each consumer can eat at most 1 unit.

• This means, given p, all consumers with taste θ > p will buy 1 unit of the

good.

• Suppose the taste parameter θ is distributed over the interval [θ, θ̄], where

θ ≥ 0, according to the distribution F (·)

– Here, F (θ) tells you what fraction of consumers has taste below θ

– For convenience, let’s just say the total mass of consumers is 1, so

F (·) is a cdf, and f (·) is the pdf

• F (θ) = 0; F (θ̄) = 1

• By setting any price p = θ ∈ [θ, θ̄], the monopolist sells 1 − F (θ) units.

(why?)

• So D(p) = 1 − F (p)

Assumptions

• We make the following assumptions on the distribution F (·)

• F (θ) is continuous and differentiable for all θ ∈ [θ, θ̄]; (no atoms)

• f (θ) > 0 for all θ ∈ [θ, θ̄]

2
 f (θ)
• The hazard rate, 1−F (θ) , is increasing in θ.

– This is a widely used and well-known assumption on probability dis-

tributions.

– Many of the commonly known distributions satisfy this property.

General Pricing Schedule

• Now let’s consider a case where the monopolist does not have to choose a

constant per-unit price

• We allow for any payment schedule T (q)

• The constant per-unit p is called linear pricing: T (q) = pq

• So now, the consumer’s net payoff is U (q, θ) = θq − T (q)

• However, the consumer can decide not to buy anything, and get a payoff

of 0

• We also use a slightly different concept of cost of production here.

• Denote now the monopolist’s cost of producing q units for one consumer

as c(q)

• c(q) is not the total cost function, but rather the cost of producing a

product of q size meant for a single consumer

– Think 8 oz. coffee cup vs. 12 oz. coffee cup

– Alternatively you can think of q as measuring quality. The idea is to

abstract away from things like increasing returns to scale that adds

more complexity

• We still assume that c0 > 0, c00 > 0

3
 • With a non-linear pricing schedule T (q), the market demand D(p) isn’t

the monopolist’s constraint anymore

– The monopolist is no longer choosing a threshold level of consumer

to sell to

– Can design a pricing schedule T (q) where different consumers will

buy different quantities

– Still possible to set T (q) in such a way that excludes certain types

– But also possible to sell to all types and still make profits

• Crucial assumption: the monopolist only knows the distribution F (θ)

• Does not know the taste θ for any given consumer

• How should the monopolist make use of her information (knowledge of

F (θ))?

• How do we find the optimal (profit-maximizing) pricing schedule?

Revelation Principle

• The following result helps us find the optimal pricing schedule for the

monopolist:

– For any schedule T (·), which results in consumer with type θ max-

imizing his net payoffs by choosing quantity q and making pay-

ment T (q), there is an equivalent menu of quantity-payment pairs

(q(θ), T (θ)) that results in the same outcome.

• Think of this as the monopolist committing to sell quantity q(θ) and charge

T (θ), if a consumer goes and tells the monopolist that his taste parameter

is θ.

4
 • Here, q : [θ, θ̄] → [0, 1] and T : [θ, θ̄] → R

• We can then begin to formulate the monopolist’s problem:

wθ̄
max {T (θ) − c(q(θ))} f (θ)dθ
q(·),T (·)
θ

s. to θq(θ) − T (θ) ≥ θq(θ̂) − T (θ̂); ∀θ, ∀θ̂ (IC)

θq(θ) − T (θ) ≥ 0; ∀θ (IR)

Incentive-Compatibility Constraints

• Given a schedule of quantity-payment pairs (q(·), T (·)), the incentive-

compatibility constraints are equivalent to the following two conditions:

1. Monotonicity: q(·) is non-decreasing

rθ
2. Envelope Condition: For all θ ∈ [θ, θ̄], U (θ) = U (θ) + q(t)dt; where
θ
U (θ) := θq(θ) − T (θ)

Proof

• For any pair θ and θ̂,

U (θ) = θq(θ) − T (θ)

U (θ̂) = θ̂q(θ̂) − T (θ̂)

5
• Incentive compatibility means,

U (θ) ≥ θq(θ̂) − T (θ̂)

   
U (θ) − U (θ̂) ≥ θq(θ̂) − T (θ̂) − θ̂q(θ̂) − T (θ̂)
 
U (θ) − U (θ̂) ≥ θ − θ̂ q(θ̂)

• With roles reversed, we must also have

 
U (θ̂) − U (θ) ≥ θ̂ − θ q(θ)

• But this means

 
U (θ) − U (θ̂) ≤ θ − θ̂ q(θ)

• So,
   
θ − θ̂ q(θ̂) ≤ U (θ) − U (θ̂) ≤ θ − θ̂ q(θ)

U (θ) − U (θ̂)
q(θ̂) ≤ ≤ q(θ) ; assuming θ > θ̂
θ − θ̂

• There’s the monotonicity condition

• Now for any θ̂ > θ and any θ̂ < θ,

   
θ − θ̂ q(θ̂) ≤ U (θ) − U (θ̂) ≤ θ − θ̂ q(θ)

• So, taking the limit as θ̂ → θ, we get

U (θ) − U (θ̂)
U 0 (θ) = lim = q(θ)
θ̂→θ θ − θ̂

• So U 0 (θ) = q(θ), using the fundamental theorem of calculus, we get the

6
 envelope condition:
wθ
U (θ) = U (θ) + q(t)dt
θ

Payments

• Notice that

T (θ) = θq(θ) − U (θ)

wθ
T (θ) = θq(θ) − U (θ) − q(t)dt
θ

• This means that T (·) is uniquely pinned down, up to the constant U (θ),

by the choice of q(·)

rθ
• Also notice that given q(θ) ≥ 0, U (θ) = U (θ) + q(t)dt means satisfying
θ
the IR constraint for θ is enough to satisfy all IR constraints.

• What should be the monopolist’s choice for U (θ)?

Back to the PMP

• Using what we have learnt, we can rewrite the profit-maximization prob-

lem:
 
wθ̄  wθ 
max θq(θ) − c(q(θ)) − q(t)dt f (θ)dθ
q(·)  
θ θ

s. to q(·) non-decreasing (Monotonicity)

• The traditional way to solve this problem is to ignore the monotonicty

constraint, find the optimal q(·), and then check that monotonicity is

7
 satisfied.

• The maximization problem at hand is an optimal control problem.

• There is a neat short-cut that allows us to solve it without using optimal

control.

• Notice that the maximand is:

 
wθ̄  wθ 
θq(θ) − c(q(θ)) − q(t)dt dF (θ)
 
θ θ

• Using integration by parts, this can be rewritten as:

wθ̄  1 − F (θ)

θq(θ) − c(q(θ)) − q(θ) f (θ)dθ
f (θ)
θ

wθ̄  1 − F (θ)
 
= θ− q(θ) − c(q(θ)) f (θ)dθ
f (θ)
θ

• Notice now that we can just pointwise maximize the term inside the second

bracket, choosing q.

• The pointwise maximization problem is:

 
1 − F (θ)
max θ − q − c(q)
q f (θ)

1−F (θ)
• FOC: θ − f (θ) − c0 (q) = 0

• So the profit maximizing solution(assuming interior solution) is to choose

1−F (θ)
q(θ) such that marginal cost, c0 (q(θ)) = θ − f (θ)

f (θ)
• Remember the monotone hazard rate condition: 1−F (θ) is increasing in θ

1−F (θ)
• This means θ − f (θ) is strictly increasing in θ

8
 1 − F (θ)
c0 (q(θ)) = θ −
f (θ)

• The RHS is increasing for higher types, so the solution involves setting

higher marginal costs for higher θ

• Given c00 > 0, this means optimal q(θ) is increasing in θ. So monotonicity

is satisfied.

• Food for thought: think what the optimal q(θ) would be with constant

marginal cost.

• Is the monopolist’s quantity choice efficient (first-best) for any of the

types?

• Efficient allocation would maximize total surplus:

wθ̄
max {(θq(θ) − T (θ)) + (T (θ) − c(q(θ)))} f (θ)dθ
q(·)
θ

wθ̄
max {θq(θ) − c(q(θ))} f (θ)dθ
q(·)
θ

• Easy to see that this is achieved by setting c0 (q(θ)) = θ

1−F (θ)
• Monopolist chooses to set c0 (q(θ)) = θ − f (θ)

• Chooses less than the first-best quantity for all types θ < θ̄

• Chooses first-best quantity for θ̄ (no distortion at the top)