Industrial Organization 03: Collusion

Industrial Organization 03
Collusion
Marc Bourreau
Telecom ParisTech
Marc Bourreau (TPT) Lecture 03: Collusion 1 / 40

Outline
1 Definition
2 Why cartels are forming?
3 Why do we not observe only cartels?
Sanctions against cartels in Europe and in the US
Instability of cartels due to the strategic interactions
4 A model of collusion
5 Factors favoring collusion: examples.

Collusion Definition
Introduction
Adam Smith (The Wealth of Nations, Book I, 1776)

People of the same trade seldom meet together, even for
merriment and diversion, but the conversation ends in
a conspiracy against the public, or some contrivance to
raise prices.

Definition
Definition of collusion
We talk about collusion (or cartels) when firms on a market agree to realize
profits superior to the "normal" profits they should get in a situation of compe-
tition.
→ "Normal" profits are those of the non-cooperative Nash equilibrium (Bertrand,

Cournot...).
Collusion can be:

Explicit: the firms agree explicitly on prices, but also on quantities, pro-
duction capacities, investments in R&D, etc.
Tacit: behavior that allows firms to realise profits superior to the "normal"
profits, without an explicit agreement.

Why Cartels are Forming?
Firms in a market are tempted to collude to raise their market power (their
profits)...
We suppose that in the case of competition, firms "maximize their profit". Why
then forming a cartel makes it possible for firms to raise their profits?
Because a situation of competition is comparable to the prisoner dilemma:

A firm decides of its strategy (by setting prices, quantities, investments...)
to maximize its profits...
... but it doesn’t take into account the effect of its decision on the other
firms.
In a cartel, the firms take into account how the decisions of each of one
them impacts the profits of the others.

Different Types of Cartels
Price fixing
Quantity fixing (rare)
Geographic markets repartition
etc. (other ideas?)
But...
We can have parallel behaviors without collusion
Some concerted practices are authorized under some conditions (ex: Re-
search Joint Ventures)

Why Do We Not Observe Only Cartels ? The Sanctions against Cartels
Why Do We Not Observe Only Cartels ?
Why all firms are not colluding?
→ Because collusion is forbidden (competition policy). But this is not sufficient

to discourage cartel formation!
→ Because firms in a cartel have incentives to "cheat", to "deviate" from the

cooperative equilibrium.
Break unilaterally the collusion agreement to set, for example, a lower
price... to capture a larger share of demand.
If each firm expects that the others will cheat, collusion is not possible.
Punishment mechanisms are required to dissuade the "cheaters ".

Cartels and Public Policy
Cartels are socially inefficient:

Loss of social welfare when the price is higher than marginal cost: "dead-
weight loss" (welfare criterion).
Transfer from consumers to producers (consumers surplus criterion).
Yet, in the 19th century, cartels were common and legal in the US (steal, sugar,
petroleum...).
In France, since 1810, the article 419 of the Code Pénal forbids price-fixing
agreements "prices above and below of what would have been set by the free and
natural competition"...
... but in practice, the Courts punish few cases ("good" cartels).

At the end of the 19th century, beginning of the 20th, the American Congress
voted two laws to fight against the high prices set by the "trusts":
The Sherman Antitrust Act in 1890
The Federal Trade Commission Act in 1914
These laws forbid explicit agreements which aim at reducing competition in-
tensity.
Sherman Act (1890):

Section 1. Every contract, combination in the form of trust or otherwise, in
restraint of trade or commerce... is declared to be illegal. Section 2. Every
person who shall monopolize, or attempt to monopolize, or combine or
conspire with any person or persons, to monopolize any part of the trade or
commerce... shall be deemed guilty of a felony...

In Europe, Article 85 of the Rome Treaty (1958):

The following shall be prohibited...: (a) directly or indirectly fix purchase or
selling prices... (b) limit or control production... (c) share markets or
sources of supply...
Now Article 101 of the Treaty on the Functioning of the European Union.
The laws which prohibit the cartels are entitled:

Antitrust Law in the US
Competition Policy in Europe.

Examples of financial sanctions for collusion (source: OCDE 2000):
Case Sanction
Worldwide cartel of graphite 300 millions $ and sentences in
electrodes the US ; 218 millions euros in Eu-
rope
Anti-competitive agreement re- 378 millions French francs
lated to the TGV
Worldwide cartel of vitamins 1 billion $ and sentences in the
US ; 855 millions euros in Europe
Worldwide cartel of lysine et cit- 200 millions $ and sentences in
ric acid the US

Legislations implementation:
The interdiction is applied more or less strictly according to the countries
(US more strictly) and according to the time (stricter nowadays)
The interpretation of what constitutes an anti-competitive agreement may
also vary from one country to another (ambiguities).
R&D agreements
Professional association, information exchanges...
Leniency programs:
A company which participated in a cartel which would inform on the cartel,
may benefit from a partial or complete sanction reduction.
Introduced in the US in 1993, in Europe in 2002, in France in 2001.

Why Do We Not Observe Only Cartels ? Cartels Instability
The Stability of Cartels
Should the collusion be tacit or explicit, it requires repeated interactions and a

punishment mecanism:
Setting prices superior to the competitive prices (for example) is held by an
agreement (tacit or explicit) that in the case of deviation from the collusive
equilibrium, the cheater would face a "punishment".
The "punishment": a temporary price war, or any other actions aiming at
reducing the deviant firm’s profits.
The "punishment" should be sufficiently costly compared to the deviation
profits.
It is even more efficient if it is not costly for the firms implementing it.
Any "deviation" should be detected.

Why Do We Not Observe Only Cartels ? Cartels Instability
An Example of Cartel
The diamants cartel:
DeBeers Group, founded in 1870, dominates the worldwide diamant mar-
ket.
Mostly controls trade rather than production: Central Selling Organization
(CSO).
More than 80% of the worldwide production transit through the CSO.
Benefits for producers to go through CSO:

Evaluation, publicity, price stability.
Fear of reprisal in case of "deviation".
A deviation:
Zaire (Mobutu) tried to deviate (in 1981).
Two months later, a huge quantity of unidentified diamants flooded the
market: price fall by almost 40%.

A model of collusion
A Model of Collusion
Two firms sell identical goods (perfect substitutes)

Same marginal cost of production, c
→ Same assumptions as in the Bertrand competition model
But we consider a repeated game with infinite horizon ("supergame")

At each time t = 0, 1, ..., ∞, the two firms set prices simultaneously and
conditionally to the prices previously set
δ denotes the discount factor (value today of one euro, which will be
obtained only in the next period)
pm denotes the monopole price, πm denotes the monopoly price

The Discount Factor
δ is the discount factor (value today of one euro, which will be obtained only
in the next period). In general, δ < 1.
For example, an investor could invest 1 euro today to get 1+r euro at the next
period (r: return rate per peiod). We would get:
1
δ= .
1+r
When we apply the discount factor to future profits, we talk about discounted
value of future profits.
→ For example, to evaluate a project, we usually calculate the "Net Present

Value" (NPV) = discounted income - discounted costs

The Collusion Strategy
Let’s consider the following strategy (also called "trigger strategy"):
Collusion period:
A firm sets the price p = pm (monopoly price) if the other firm has set pm in the
previous periods.
Punishment period:
A firm sets p = c after a deviation (i.e. if the other firm has set a price p < pm in
the previous period) and for all following periods.
How do we determine if the tacit collusion agreement is viable?
Comparison of the discounted value of future profits in the case of collusion

with the discounted value of future profit in the case of deviation.

Discounted Profits in Case of Collusion
Every period, both firms set the monopoly price. Therefore, every period, both
firms share the monopoly profit.
The discounted value of future profits is:
1 m 1 1
Π= π + δ πm + δ2 πm + ...
2 2 2
1 m
Π= π 1 + δ + δ2 + ... ,
2
1 m 1

Π= π .
2 1−δ

Discounted Profits in Case of Deviation
We look for the incentive to deviate from the anti-competitive agreement (i.e.
the net profit in case of deviation).
What would be "the best" deviation strategy (price change)?
Best deviation
At any period, the best deviation consists in setting a price equal to pm − .
Therefore, the deviant firm wins (at best)?
Π
e = πm + δ × 0 + δ2 × 0 + ... = πm

Collusion Sustainability
It is "not" in the interest of any firm to deviate if its profit in case of deviation is
lower than its profit in case of collusion, that is, if:
Π
e < Π,
or
1 m 1

πm < π ,
2 1−δ
therefore
Result: stability of collusion

Collusion is sustainable (an equilibrium) if and only if
1
δ> ≡ δ.
2

Collusion Sustainability
The collusion is sustainable if the firms value future profits high enough.
The parameter δ (limit discount factor) indicates to what extent the collusion is
feasible:
If a factor tend to decrease δ, we say that it facilitates collusion.
If a factor tend to increase δ, we say that it makes collusion more difficult.
Remark: there are other equilibria to this repeated game (e.g., the Bertrand
equilibrium of the one-period game is an equilibrium).

Collusion if the Number of Periods is Finite?
Equilibrium in a price competition game repeated T < ∞ times

There is only one (subgame perfect) Nash equilibrium such as p? = c.
Why?
Backward induction reasoning.
The last period (Tth) corresponds to the one-period Bertrand game, so the
unique equilibrium of this period is such that pT = c
At period T−1, firms expect that at the last period the Bertrand equilibrium
will prevail
They are in the same situation as in the last period and set pT−1 = c
etc.

Factors which Promote Collusion
Market Structure and Collusion

Does a more concentrated market structure (a lower number of firms) facilitate
collusion?
Let’s consider an oligopoly with n identical firms
The marginal cost is assumed constant and equal to c
Firms compete à la Bertrand
The discounted value of future profits in case of collusion is
1 M 1

Π= π ,
n 1−δ
What is the deviation profit? It is always Π

e = πm
The limit discount factor is
1

δ> 1− ≡ δ.
n

The threshold discount factor,
1

δ≡ 1− ,
n
increases with n, and therefore we obtain the following result:
Relation between market structure and collusion

Lower is the number of firms, the easier is collusion.
For example, two firms with a discount factor of 0.6 could hold a situation of
tacit collusion, but not three firms.

This result has been confirmed in practice.
In the cartel cases dealt with by the Department of Justice (DOJ) in the US
between 1990 and 2003:
77% were related to anti-competitive agreements between 6 firms or less
13% only related to anti-competitive agreements between 50 firms or more
Out of the 111 cartels detected and sanctioned by the European Commis-
sion between 1969 and 2009 (Combe and Monnier, 2012):
On average 7.7 participants and median of 5
50% of cartels comprised less than 5 participants and 75% less than 10.
Cartels often collapse upon entry.

Entry Barriers and Collusion
Do entry barriers facilitate or make collusion more difficult?
Collusion is more difficult if there are weak entry barriers:

Entry of competitors (e.g., "hit and run" entry) decreases collusion profits.
The possibility of competitors entry in the future weakens the "threat" of
"punishment" in case of deviation.
Let’s assume the following model:

Every period, there is a probability of entry µ from a competitor which sets
the competitive price p = c
With the probability 1 − µ, there is no entry and the two firms may try to
collude
A "hit and run" entry only last one period (the possibility of entry is
therefore determined every period).

In the case of collusion, both firms receive:

1 M 1 1
Π= π + 1 − µ δ πM + 1 − µ δ2 πM + ...,
2 2 2
The collusion is therefore sustainable if:

1 M 1 M δ

π + 1−µ π > πM ,
2 2 1−δ
with
1
δ> ≡ δ.
2−µ

The threshold discount factor δ increases with the probability of entry µ.
Therefore, we obtain the following result:
Relation between entry barriers and collusion

A higher probability of entry decreases the collusion possibility.
→ A higher probability of entry (weaker entry barriers) decreases future collu-

sion profits and therefore decreases the "cost" of a deviation (punishment).

Frequency of Interactions and Collusion

Do competitive interactions facilitate or make collusion more difficult?
Let’s take the basic model:

Two firms sell identical goods (perfect substitutes)
Same marginal cost of production, c
We assume that firms compete every T periods: period 1, T + 1, 2T + 1, etc.
The collusion is sustainable if
1 M
π 1 + δT + δ2T + ... > πM ,
2
which is equivalent to
1
δ> ≡ δ.
21/T
Relation between interaction frequency and collusion

More frequent interactions facilitate collusion.

Multi-Market Contacts and Collusion
Definition
When firms compete in several different markets.
Two firms compete in two independent (but identical) markets: market 1

and market 2.
Firms interact more frequently in market 1 (every period) than in the
market 2 (every 2 periods).
They collude in both markets and in case of a deviation, the "punishment"
applies also on both markets.
What is the threshold discount factor in both markets?
√
Interaction frequencies: if 1/2 < δ < 0, 71 = (1/ 2), collusion is sustainable
in market 1, but not in market 2

Collusion is sustainable on both markets if the deviation "profits" are superior

to the deviation "costs",
1 M 1 M πM πM
π + π ≤ δ + δ2 + ... + δ2 + δ4 + ... ,
2 2 2 2
or
4δ2 + δ − 2 ≥ 0,
with δ ≥ 0, 593.
For example, if δ ≥ 0, 6, collusion is not sustainable in market 2, but could be

implemented on both markets.
Therefore, multi-market interactions enable collusion in markets where, due to

their characteristics, it is a priori not sustainable.

Example of the air sector:

Air companies compete on numerous air routes, but not all.
For instance, in 1998, in the US on the 1000 first routes, the 3 main compa-
nies had a "contact" on 40% of these routes.
Multi-market contacts and collusion ?

It has been shown that this variable (the average multi-market contact
rate) had a positive impact on a route price (after controlling for other
price explaining factors).

Secret Price Cuts
Green and Porter model (1984):

A market with two firms
Each firm observe its price and its sale, but not the price and the sale of its
competitor
With a probability α, demand disappears: it equals 0 ("exogeneous shock")
If its demand is zero, a firm is unable to know if it is a lack of luck (demand
shock) or because its competitor has cheated.
In this context, perfect collusion (monopoly price even after a demand shock)
is not possible.

Secret Price Cuts
However, collusion is still feasible following this strategy:

Collusion phase: first set the monopoly price and maintain it as long as
the firm maintains its market share.
Punishment phase: when a firm demand is null, it enters a price war for a
limited number of periods (T), after which it comes back to the monopoly
price pM
Trade off on the choice of T:

The war price should be sufficiently long to dissuade "cheaters"
But a price war could have been triggered by a bad shock of demand and
not by a "deviation"
Firms have an incentive to limit the time of the price war

Secret Price Cuts
Conclusion of Green and Porter model

If price decreases are difficult to observe, periodic war price could be required
to sustain an anti-competitive agreement.
Remark: although there are periods of price wars at the equilibrium, no firm
cheats at the equilibrium.

If Demand Fluctuations are Unobserved
What would happen if firms could observe the demand fluctuations?
Let’s suppose that the demand is cyclical (ups and downs).

Demand shocks are independent from one period to another
In case of collusion, does the profit depend on the demand level?
No, due to the independence hypothesis
In case of deviation, does the profit depend on the demand level?
Yes, deviation more profitable if the demand is high.
Conclusion: It could be necessary to decrease the prices when the demand is

high. → Prices evolve in a "counter-cyclical" manner

"If You Find Cheaper Elsewhere..."
Some clauses could also facilitate collusion: for example, clauses such as "if
you find cheaper elsewhere, we reimburse the difference..." → allows to detect
deviations.
The "most favored customer clause"

What is the impact on deviation incentives?
If the firms decreases its price to capture the demand in the short-term, it
should reimburse the customers who had paid the monopoly price
→ This type of clauses highly reduce incitations to deviate and make cartels
more stable.

Other Factors
Other factors:
Cost asymmetries make collusion more difficult for two reasons:
Difficult to agree on a common pricing strategy.
It is more difficult to "discipline" the most efficient firms.
Growing markets: collusion is easier to sustain, as the profit of deviation
is weaker compare to the cost (as the market will keep growing).

Take-Aways
Take-Aways (1)
There are two types of anti-competitive agreements: explicit and tacit.

Cartel formation allows to raise the firms profits, as the firms take into
account how the decision of each of them impacts the profits of the others.
The possibility of collusion is realistic when we take the time dimension
into account (repeated interactions between firms).
Firms cannot always collude as they have the possibility to cheat, which
will lead to a punishing price war (cf. arguments from Stigler 1964, some-
times it may be more profitable to be outside a cartel!).
Collusion possibilities are facilitated by: the low number of firms, the
high valuation of discounted future profits, symmetry of costs, absence of
demand uncertainty.

Take-Aways
Take-Aways (2)
Summary table: conditions for cartel formation (stability):
Market characteristics Firms characteristics

– Low number of firms – Homogeneous firms in terms of
– Weak price elasticity of demand production costs
– Intermediate products – Homogeneous firms in terms of
– Low demand fluctuations and products
high fixed costs
– Entry barriers

Industrial Organization 03: Collusion

Uploaded by

Copyright:

Available Formats

Industrial Organization 03: Collusion

Uploaded by

Document Information

Original Description:

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 03: Collusion

Uploaded by

Copyright:

Available Formats

Industrial Organization 03

Marc Bourreau (TPT) Lecture 03: Collusion 1 / 40

Marc Bourreau (TPT) Lecture 03: Collusion 2 / 40

Adam Smith (The Wealth of Nations, Book I, 1776)

Marc Bourreau (TPT) Lecture 03: Collusion 3 / 40

→ "Normal" profits are those of the non-cooperative Nash equilibrium (Bertrand,

Collusion can be:

Marc Bourreau (TPT) Lecture 03: Collusion 4 / 40

Why Cartels are Forming?

Because a situation of competition is comparable to the prisoner dilemma:

Marc Bourreau (TPT) Lecture 03: Collusion 5 / 40

Different Types of Cartels

Marc Bourreau (TPT) Lecture 03: Collusion 6 / 40

Why Do We Not Observe Only Cartels ?

Why all firms are not colluding?

→ Because collusion is forbidden (competition policy). But this is not sufficient

→ Because firms in a cartel have incentives to "cheat", to "deviate" from the

Marc Bourreau (TPT) Lecture 03: Collusion 7 / 40

Cartels and Public Policy

Cartels are socially inefficient:

Marc Bourreau (TPT) Lecture 03: Collusion 8 / 40

Cartels and Public Policy

Sherman Act (1890):

Marc Bourreau (TPT) Lecture 03: Collusion 9 / 40

Cartels and Public Policy

In Europe, Article 85 of the Rome Treaty (1958):

The laws which prohibit the cartels are entitled:

Marc Bourreau (TPT) Lecture 03: Collusion 10 / 40

Cartels and Public Policy

Examples of financial sanctions for collusion (source: OCDE 2000):

Marc Bourreau (TPT) Lecture 03: Collusion 11 / 40

Cartels and Public Policy

Marc Bourreau (TPT) Lecture 03: Collusion 12 / 40

The Stability of Cartels

Should the collusion be tacit or explicit, it requires repeated interactions and a

Marc Bourreau (TPT) Lecture 03: Collusion 13 / 40

Benefits for producers to go through CSO:

Marc Bourreau (TPT) Lecture 03: Collusion 14 / 40

Two firms sell identical goods (perfect substitutes)

But we consider a repeated game with infinite horizon ("supergame")

Marc Bourreau (TPT) Lecture 03: Collusion 15 / 40

The Discount Factor

→ For example, to evaluate a project, we usually calculate the "Net Present

Marc Bourreau (TPT) Lecture 03: Collusion 16 / 40

The Collusion Strategy

Let’s consider the following strategy (also called "trigger strategy"):

How do we determine if the tacit collusion agreement is viable?

Comparison of the discounted value of future profits in the case of collusion

Marc Bourreau (TPT) Lecture 03: Collusion 17 / 40

Discounted Profits in Case of Collusion

The discounted value of future profits is:

Marc Bourreau (TPT) Lecture 03: Collusion 18 / 40

Discounted Profits in Case of Deviation

What would be "the best" deviation strategy (price change)?

Therefore, the deviant firm wins (at best)?

Marc Bourreau (TPT) Lecture 03: Collusion 19 / 40

Result: stability of collusion

Marc Bourreau (TPT) Lecture 03: Collusion 20 / 40