Industrial Organization 02: Oligopoly

Industrial Organization 02
Oligopoly
Marc Bourreau
Telecom ParisTech
Marc Bourreau (TPT) Lecture 02: Oligopoly 1 / 42

Outline
Oligopoly, definition and examples

Cournot model, where firms compete in quantities
Bertrand model, where firms compete in prices
Bertrand paradoxe
From Bertrand to Cournot: capacity constraints
Cournot competition with n firms
Comparison of market powers: monopoly, Cournot, and Bertrand

Introduction
Introduction
Definition of an oligopoly
A market or industry in which a small number of firms compete.
Most markets fit this description: telecommunications, software industry,

but also mineral water industry, etc.
In an oligopolistic market, a firm cannot ignore the behavior of competi-
tors...
... and the reaction of competitors to its own decisions.
The theory of oligopoly aims at studying these strategic interactions.

Introduction
An example
The oligopoly of mineral water producers in France:

Three firms (Danone, Nestlé, and Neptune) share around 80% of the mar-
ket.
Products are marketed under various brand names (Danone: Evian, Volvic,
...; Nestlé: Perrier, Contrex, ... ; Neptune (formerly Castel): Saint-Yorre,
Vichy, Cristalline, ...).
Emergence of the Castel group in 1993 following a decision of the European
Commission which authorized the acquisition of la Société des eaux minérales
du bassin de Vichy.
Differentiation between mineral waters (health, well-being), different types
of packaging, makes the price comparisons difficult.
Interdependancy of strategic decisions: if a firm increases the price of
mineral water, its competitors can chose to do the same or not to react,
hoping to capture a part of the customer base.

Cournot model
Cournot model
Cournot model (1838), developed by Augustin Cournot, a French engineer.

Two firms produce identical (homogenous) goods ("perfect substitutes")
and compete in quantities.
The price is determined in the way that the entire quantity of produced
goods is sold. Therefore one sets p = P(Q) where Q = q1 + q2 is the total
quantity produced.
The marginal cost of producing each unit of the good is assumed constant
and identical for the two firms: c
Then, the profit function of the firm i is
Πi = (P(Q) − c)qi .
For simplicity, we assume that the inverse demand function is P(Q) = 1−Q.

Cournot model
Cournot model
Each firm sets its quantity to maximize its profits, given the quantity set
by the other firm.
The first-order condition of the maximization problem is:
1 − (q1 + q2 ) − c − qi = 0.
We compute a symmetric equilibrium such that qi = q. Thus:
1−c
q? = .
3
and
(1 − c)2
Π? = .
9

Cournot model
Cournot model
The equilibrium profit is:

(1 − c)2
Π? = .
9
In Cournot competition, firms make profits!
However, the Cournot model seems somewhat unrealistic:

Few examples of markets where firms set quantities rather than prices;
We don’t know well how the price is set up on the market.
A model where firms set prices rather than quantities? → Bertrand model.

Bertrand paradox The model
Bertrand model
Bertrand model (1883), developed by Joseph Bertrand, a French engineer.

Two firms produce identical goods ("perfect substitutes") and compete in
prices.
The demand function is q = D(p).
The marginal cost of production is constant and identical for both firms: c
We assume (not indispensable) that the market is split evenly if firms offer
the same price. We thus have:
if pi < pj


 D pi
Di (pi , pj ) =  if pi = pj .

 1
2 D p i
if pi > pj

0


The profit function of firm i is:
Πi (pi , pj ) = (pi − ci )Di (pi , pj ).

Bertrand paradox Bertrand equilibrium
Bertrand equilibrium
We find the Nash equilibrium for this one-step game.
Bertrand paradox
The game has a unique Nash equilibrium such that firms set p∗1 = p∗2 = c. A the
equilibrium, we have Π∗1 = Π∗2 = 0 and W = W ∗ .
A strong result:
When the number of firms goes from one (monopoly) to two (duopoly),
the price decreases from the monopoly price to the competitive price and
stays at the same level as the number of firms increases further.
Two firms suffice to reach a perfectly competitive equilibrium.
Because this result is extreme and generally inconsistent with reality, it is
called the "Bertrand paradox".

An example
The price war between Intel and AMD, chip manufacturers, in 2006.
AMD ended the year with a loss and Intel’s benefits dropped by 42% in
2006.
Interview with Mario Rivas (CEO of AMD, Source:

l’Expansion) :
"This price war with Intel was ridiculous. I wish I was
able to match my prices to the market rate (...). The
problem was that, despite the 2 billion dollars of stocks
we had, Intel had 7 factories to run at full capacity. They
cut the price and we have no choice but to follow their
pricing".

Demonstration
If p∗1 > p∗2 > c:

Then firm 1 increases its profit by setting p∗1 = p∗2 − .
If p∗1 = p∗2 > c:

Then firm 1 increases its profit by setting p∗1 = p∗2 − , as for a small ,
D(p∗1 )(p∗1 − c)/2 < D(p∗1 − )(p∗1 − c − )
If p∗1 > p∗2 = c :

Then firm 2 increases its profit by setting p2 = p∗2 + .

Bertrand competition with different marginal costs
Let’s assume that c1 < c2 .

If costs are close enough: c1 < c2 < pm (c1 )
There is a unique Nash equilibrium with p∗1 = c2 − et p∗2 = c2 .
Only firm 1 makes profit: Π∗1 = (c2 − c1 )D(c2 ) et Π∗2 = 0.
If firm 1 is much more efficient than firm 2: c1 < pm (c1 ) < c2

There is a unique Nash equilibrium with p∗1 = pm (c1 ) and p∗2 = c2 .
Only firm 1 makes profit: Π∗1 = Πm
1
and Π∗2 = 0.

Bertrand paradox Solutions to Bertrand paradox
Solutions to Bertrand paradox
Four important solutions to "Bertrand paradox", corresponding to four impor-

tant assumptions on which the model relies :
1 Products are homogeneous
2 Short-run competition (a static analysis of a one-step game)
3 Firms do not have capacity constraints
4 Consumers are perfectly informed
Relaxing one of these assumptions solves the paradox:

1 Product differentiation
2 Dynamic competition (repeated interactions)
3 Firms are capacity-constrained
4 Imperfect information

Product differentiation
Let’s assume, for instance, a geographic differentiation.

Two icecream vendors, 1 and 2, are located at the two extremities of a
beach
If p1 = c, is p2 = c + > c possible?
Consumers close to vendor 2 can prefer buying a little more expensive
from 2 than moving to 1!
Horizontal and vertical differentiation theory: the Hotelling model...
In the presence of product differentiation...

A situation such as that pi > c can be an equilibrium.
→ See the lecture about differentiation.

Dynamic competition
Bertrand competition model assumes that firms compete only during a

single period.
In this situation, as we start from p1 = p2 > c, a firm has strong incentives
to undercut its price.
In a more dynamic framework, what can happen?
In a dynamic situation, firms are going to take account the effect of its price
cut on its rival’s behavior in future periods.
If there is a "punishment" (price war...), it will compare the short-term and
long-term gains.
In a repeated interaction framework...

A situation with pi > c can be an equilibrium.
→ See the lecture about collusion.

Capacity constraints
Bertrand model assumes that firms do not have capacity constraints.

If p1 = p2 = c, the two firms share the demand: D(c)/2
If firm 2 increases slightly its price, p2 = c + , we assumed that firm 1
meets the demand, that is D(c).
But firm 1 can be unable to meet the whole demand: capacity constraints
In that case, firm 2 can increase her price and still retain a part of the market
When firms face capacity constraints...


Imperfect information
With imperfect information...

The Diamond paradox

Consumers are not informed about prices
Cost to move from a store to another (search cost)
If p1 = p2 < pm then deviation is possible to p1 + /2

Price competition with capacity constraints A model
Price competition with capacity constraints
We are going to study a price competition model when firms have constraints
in production capacity.
We consider the following model:

Two firms compete in prices, but are capacity-constrained
The demand function is linear, D(p) = 1 − p
The inverse demand function is: p = P(q1 + q2 ) = 1 − (q1 + q2 )
Firm i cannot produce more than its production capacity qi . Thus, we have
qi ≤ qi
The unit cost of acquiring capacity qi is c0 ∈ [3/4, 1].
There is no production cost (c = 0).

Price competition with capacity constraints A model
We need to define a rationing rule.

The rationing rule determines "which consumers" are being served when
the firm cannot meet the entire demand.
We apply the "efficient" rationing rule: consumers with higher willingness to
pay are first served.
If p1 < p2 and q1 < D(p1 ) :
(
e 2 p2 = D p2 − q1 if D p2 ≥ q1
D .
0 if D p2 < q1
Other possible rule: the "proportional" rationing rule

Price competition with capacity constraints Results
Results
We obtain the following result:

The unique Nash equilibrium is such that the firms set the same price

p∗ = 1 − q1 + q2 .
Consequence: at the stage of defining their production capacities, firms have a

gross profit equal to:
g
h i
Πi = 1 − qi + qj qi ,
That is? The profit function in the Cournot model.

Proof: preliminaries
First, given that c0 ∈ [3/4, 1], what is the upper bound for qi ?
To answer the question, what is the maximum profit of firm i ?
→ It is the monopoly profit!
As D(p) = 1 − p and c = 0, the monopoly price is equal to... ? 1/2
Thus the monopoly profit is... ? 1/4 − c0 qi
Therefore qi ≤ 1/3 because we have 1/4 − c0 qi ≥ 0.

Proof of existence

Now, let’s demonstrate that p∗ = 1 − q1 + q2 is a Nash equilibrium.
Observation: p∗ > c because q1 + q2 < 2/3.
Can firm i lower her price?

No, she would not increase her profit as she is already at the maximum of her
production capacity.
→ "Bertrand competition" does not work because firms face capacity con-
straints.

Proof of existence
If the firm i sets a higher price p > p∗ ?

Firm j captures the entire demand... up to the maximum of its production
capacity.
A fraction of the demand is not satisfied and firm i takes it back: we call it
the residual demand
The residual demand is equal to 1 − p − qj
Thus firm i makes profit:

p 1 − p − qj = 1 − q − qj q.
It is the Cournot profit! It is concave in q and its derivative in q = qi is
1 − 2qi − qj ≥ 0 as qi ≤ 1/3,
Therefore, the optimal value of q is q = qi .

Proof of uniqueness

p∗ = 1 − q1 + q2 is the unique Nash equilibrium.

p1 = p2 = p > P q1 + q2 is not an equilibrium because at least one firm
does not reach its maximum production capacity. Therefore she can lower
her price.

p1 = p2 = p < P q1 + q2 is not an equilibrium. With pi = pi + , the firm
would sell the same capacity (its production capacity) at a higher price.
p1 < p2 not feasible, because firm 1 is encouraged to increase its price.

Price competition with capacity constraints More general result
More general result
We consider a two-stage game in which:

First, firms choose simultaneously their production capacity;
Firms observe the chosen capacities and choose simultaneously their prices
(price competition).
Kreps and Scheinkman (1983): capacity + price = quantities

If the demand function is concave and we use the efficient-rationing rule, then
the equilibrium f the two-stage game is equivalent to the equilibrium of Cournot
competition (competition in quantities).

Cournot or Bertrand competition?
Which one is the best-suited model, Bertrand or Cournot?

Informal rule (rule of thumb)
If the production capacity can be easily adjusted, Bertrand competition model
better represents the duopoly competition. Otherwise, if it is difficult to adjust
the production capacity, the Cournot model is more appropriate.
Examples of markets where production capacity is not easy to adjust? → markets for
physical goods (car manufacturers, airplanes, cement industry...).
Examples of markets where production capacity is easy to adjust: → markets for

services (bank, insurance...)...

Bertrand or Cournot? The example of the recorded

music industry
Let’s consider the recorded music industry.
We observe an evolution in the distribution of recorded music:

from the sales of CDs in physical stores,
to digital files on web platforms such as iTunes,
and now to streaming platforms.
What type of competition model (Cournot, Bertrand) represents better the old
distribution model? And the new one?
What are the consequences we can predict from this change in the competition
model ?

The effect of the digitization: Britannica vs Encarta
Britannica: a 200-year-old encyclopedia, 1600 euros the full collection.

Encarta: a product launched by Microsoft in 1992 (acquisition of Funk and
Wagnalls) at 49,95 dollars/euros
Response of Britannica to Microsoft’s entry?
Online encyclopedia at 2000 euros per year
Drop in sales of Britannica to 50% between 1990 and 1996
Online encyclopedia at 120 euros/an
CD for 200 euros, then for less than 100 euros.

Let’s assume a linear demand D(p) = 1 − p

n firms on the market compete in quantities (Cournot competition)
The cost fonction of firm i is Ci (qi ) = cqi
What is the equilibrium price?
What is the equilibrium profit?


The inverse demand function is P(Q) = 1 − Q, where
n
X
Q= qi .
i=1
The profit function of a firm i is

Πi = (P(Q) − c)qi .
The first order condition of the maximization problem is:
1 − Q − c − qi = 0.
We look for a symmetric equilibrium with qi = q. Thus
1−c
q= .
n+1
and
(1 − c)2
Π= .
(n + 1)2
The equilibrium profit is:

(1 − c)2
Π= .
(n + 1)2
The more numerous firms are, the lower Cournot profit is.
If the number of firms is large, firms hardly any profit.

Comparison in terms of market power
Assumptions:
Let’s assume there are n firms,
With the same marginal cost of production, c.
We define Li = (pi − c)/pi the Lerner index for firm i.
Comparison: monopoly, Bertrand, Cournot

In a monopoly, we have?
1
. Li =

In Bertrand competition, we have? → Li = 0.
In Cournot competition, we have?
αi
Li = , where αi is the market share of firm i


Proof
Let P(Q) represent the inverse demand function, with Q = q1 + q2 the total
quantity produced
Then, firm i profit function is

P qi + qj − c qi
The FOC is :
P qi + qj − c + qi P0 qi + qj = 0,
That is,

P−c qi P0 qi + qj qi P qi + qj Q
0
= − =−
P P Q P
qi Q
= − ,
Q PD0
ie.
αi
Li = .

An application: mergers
Merger in the French mobile telephony market
The French mobile telephony market had a client base of 71 million of

subscribers in March 2015 (according to ARCEP). (That is, a penetration
rate of 107%).
There are four principal operators: Orange (35,6%), SFR (28%), Bouygues
Télécom (14,2%) and Free (13,2%)
Mobile virtual network operators (La Poste, etc.) have low market shares
(8,7%).
There have been rumors of mergers between Bouygues Télécom and Free.
We are going to build a simple competition model to study the incentives
of Bouygues Télécom and Free to merge.

For simplicity, let’s assume that the mobile market presents 4 firms.
We thus ignore the competitive pressure coming from MVNOs.
We assume that these 4 firms produce homogeneous (identical) goods and
compete in quantities.
Let qi firm i’s quantity and Q = q1 + q2 + q3 + q4 the total quantity.
The demand function is D(p) = 50 − p.
Questions:
1 Determine the Nash equilibrium for a quantity competition game. Com-
pute the quantity and profit in equilibrium for each firm.
2 We assume that firms 3 and 4 merge (Bouygues and Free) and become a
new firm, BF. Compute again the equilibrium, the produced quantity and
equilibrium profit for firm BF, 1 and 2.
3 Are firms 3 and 4 better off merging? How do firms 1 and 2’s profits
evolve?

1 Determine the Nash equilibrium for a quantity competition game. Com-

pute the quantity and profit in equilibrium for each firm.
The quantity in equilibrium for a Cournot model with n firms is:
50
q∗ (n) = .
n+1
The Cournot profit with n firms is:
2500
Π∗i (n) = .
(n + 1)2
Therefore, for n = 4, we have: q∗ (n) = 50/5 = 10 and Π∗i = 2500/25 = 100

1 We assume that firms 3 and 4 merge (Bouygues and Free) and become a
new firm, BF. Compute again the equilibrium, the produced quantity and
equilibrium profit for firm BF, 1 and 2.
We move from an oligopoly with 4 firms to an oligopoly with 3 firms

Thus, for n = 3, we have: q∗ (n) = 50/4 and Π∗i = 2500/16 = 156.25

1 Are firms 3 and 4 better off merging? How do firms 1 and 2’s profits
evolve?
Bouygues and Free are not well advised to merge because 156.25 < 100 *
2: their global profit decreases!
Whereas, Orange and SFR are better off: 156.25 > 100 !

Merger in Cournot competition

Let’s consider a market with n > 1 firms
Marginal cost is constant and identical: c
Linear demand: P(Q) = a − bQ
In the equilibrium, we have:
1 (a − c)2
Π∗i (n) = .
b (n + 1)2
In case of a merger of k firms, n − k + 1 firms will be present on the market.

In that case, a merger is profitable if
Π∗i (n − k + 1) ≥ kΠ∗i (n)
Merger in Cournot competition

In a Cournot competition, a merger is profitable only if it involves more than
80% of firms in the market.
Merger in Bertrand
Let’s consider a market with n > 1 firms

Marginal costs are constant and identical: c
In the equilibrium, p∗1 = ... = p∗n = c and profits are 0.
If k < n firms merge, how the equilibrium is modified?
It is unchanged: the price stays equal to c and profits are 0.
Merger in Bertrand competition

In a Bertrand competition, a merger is profitable only if it involves all (100%)
of the firms in the market.

Conclusion on mergers
Merger decisions cannot be explained only by the incentives to reduce compe-

tition in the market.
Other dimensions to explain merger?

Synergies: cost cutting...
Becoming an industry leader (the competition model changes: Stackelberg
rather than Cournot)
Portfolio strategy: extending product range, economies of scale in sales
and marketing, incentives to bundle.

Take-aways
Take-aways
Bertrand competition between identical firms leads to marginal-cost pric-

ing ("Bertrand paradox").
The Bertrand paradox is not verified anymore if we relax one of the 4
principal assumptions...
If capacity constraints can be easily adjusted in short term, it is more likely
that firms compete in Bertrand. If production capacity remains fixed in
medium term, firms compete in Cournot.
When n firms compete in Cournot, their profit is inversely proportional to
the number of firms playing in the market.
A merger is not always profitable in Cournot competition; it needs to
involve at least 80% of the firms in the market. In Bertrand competition, it
requires 100% of the firms in the market. Mergers cannot be only explained
by an incentive to reduce competition.

Industrial Organization 02: Oligopoly

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Industrial Organization 02: Oligopoly

Uploaded by

Copyright:

Available Formats

Industrial Organization 02

Marc Bourreau (TPT) Lecture 02: Oligopoly 1 / 42

Oligopoly, definition and examples

Marc Bourreau (TPT) Lecture 02: Oligopoly 2 / 42

Most markets fit this description: telecommunications, software industry,

Marc Bourreau (TPT) Lecture 02: Oligopoly 3 / 42

The oligopoly of mineral water producers in France:

Marc Bourreau (TPT) Lecture 02: Oligopoly 4 / 42

Cournot model (1838), developed by Augustin Cournot, a French engineer.

Marc Bourreau (TPT) Lecture 02: Oligopoly 5 / 42

We compute a symmetric equilibrium such that qi = q. Thus:

Marc Bourreau (TPT) Lecture 02: Oligopoly 6 / 42

The equilibrium profit is:

In Cournot competition, firms make profits!

However, the Cournot model seems somewhat unrealistic:

Marc Bourreau (TPT) Lecture 02: Oligopoly 7 / 42

Bertrand model (1883), developed by Joseph Bertrand, a French engineer.

The profit function of firm i is:

Πi (pi , pj ) = (pi − ci )Di (pi , pj ).

Marc Bourreau (TPT) Lecture 02: Oligopoly 8 / 42

We find the Nash equilibrium for this one-step game.

Marc Bourreau (TPT) Lecture 02: Oligopoly 9 / 42

Interview with Mario Rivas (CEO of AMD, Source:

Marc Bourreau (TPT) Lecture 02: Oligopoly 10 / 42

If p∗1 > p∗2 > c:

If p∗1 = p∗2 > c:

D(p∗1 )(p∗1 − c)/2 < D(p∗1 − )(p∗1 − c − )

If p∗1 > p∗2 = c :

Marc Bourreau (TPT) Lecture 02: Oligopoly 11 / 42

Bertrand competition with different marginal costs

Let’s assume that c1 < c2 .

If firm 1 is much more efficient than firm 2: c1 < pm (c1 ) < c2

Marc Bourreau (TPT) Lecture 02: Oligopoly 12 / 42

Solutions to Bertrand paradox

Four important solutions to "Bertrand paradox", corresponding to four impor-

Relaxing one of these assumptions solves the paradox:

Marc Bourreau (TPT) Lecture 02: Oligopoly 13 / 42

Let’s assume, for instance, a geographic differentiation.

In the presence of product differentiation...

→ See the lecture about differentiation.

Marc Bourreau (TPT) Lecture 02: Oligopoly 14 / 42

Bertrand competition model assumes that firms compete only during a

In a repeated interaction framework...

→ See the lecture about collusion.

Marc Bourreau (TPT) Lecture 02: Oligopoly 15 / 42

Bertrand model assumes that firms do not have capacity constraints.

When firms face capacity constraints...

Marc Bourreau (TPT) Lecture 02: Oligopoly 16 / 42

With imperfect information...

The Diamond paradox

Marc Bourreau (TPT) Lecture 02: Oligopoly 17 / 42

Price competition with capacity constraints

We consider the following model:

Marc Bourreau (TPT) Lecture 02: Oligopoly 18 / 42

Price competition with capacity constraints

We need to define a rationing rule.

Other possible rule: the "proportional" rationing rule

Marc Bourreau (TPT) Lecture 02: Oligopoly 19 / 42

We obtain the following result:

Price competition with capacity constraints

Consequence: at the stage of defining their production capacities, firms have a

That is? The profit function in the Cournot model.

Marc Bourreau (TPT) Lecture 02: Oligopoly 20 / 42

D(p∗1 )(p∗1 − c)/2 < D(p∗1 − )(p∗1 − c − )