Transcript What are the Incentives to Collude?

Collusion and Cartels
• What is collusion?
– An attempt to suppress competition
• What is a cartel?
– A group of firms who have agreed explicitly to
coordinate their activities to raise market price
or decrease market output.
Why doesn't everyone collude?
• Illegal.
– In the US, collusive agreements cannot be
enforced by legal contracts
– International cartels do exist, however.
• Hard to come to an agreement.
• Strong incentives to cheat -- collusion may
not be sustainable.
Antitrust Perspective on
Collusion
• Inefficiencies arise from collusion because price is
greater than marginal cost and thus there is
deadweight loss.
– But cartels can be worse than monopolies because they
may not achieve the economies of scale that
monopolies do.
• Sherman Act forbids any “contract, combination,
or conspiracy in restraint of trade”.
• FTC Act bans all “unfair methods of competition”.
Three Types of Collusion
• Tacit Coordination.
• Facilitating Practices.
• Explicit Conspiracy.
Tacit Coordination
• Spontaneous cooperation resulting
from strongly perceived
interdependence.
• For example, following a rival’s price
change.
• Difficult to achieve with lots of firms.
• Hard to find/prove/correct.
Facilitating Practices
• “Most Favored Customer” clauses or
price matching policies.
– Reduces sellers’ incentives to cut prices.
• Long-term customer contracts or exit
fees.
– Insulates sellers from new competition.
• Advance announcement of price
changes.
Facilitating Practices, con’t
• May have positive benefits along with
anti-competitive effects.
• Not necessarily illegal.
• Evaluated under a “rule-of-reason”.
– Compare benefits to harm -- only illegal
if there is a net harm.
Explicit Conspiracy
• Price fixing agreement.
• Formal cartel.
• Per se illegal.
What are the Incentives to Collude?
• Start with a simple model of a Bertrand Duopoly.
• Without collusion, p1*= p2*= c and thus i = 0 .
• Industry quantity is set at the perfectly competitive level.
• For a monopolist, price is set where MR = MC:
•P = a-bQ so MR = a -2bQ.
•Set c = a-2bQ and solve for monopoly quantity and price.
•QM = (a-c)/2b and PM = (a+c)/2
• So M = (PM - c)*QM = (a+c - 2c)/2 * (a-c)/2b = (a-c)2/4b
• If each firm produces 1/2 QM, each gets (a-c)2/8b
Collusion in the Prisoner’s Dilemma Framework
Collude
Collude
Defect
(a-c)2/8b,
(a-c)2/8b
Defect
0,
0
But what about the two empty cells?
To fill in the two empty cells:
If one firm sets price at the monopoly level, what price will the
cheater set?
• pi* (the BR) = pM- .
•Then the cheater gets all the demand and earns a profit only
slightly less than what a monopolist would get:
C = (PM--c)*(QM +)  ((a+c)/2 -c)*(a-c)/2b = (a-c)2/4b.
•Profit of non-cheater is 0.
Collusion in the Prisoner’s Dilemma Framework
Collude
Collude
Defect
(a-c)2/8b,
0,
(a-c)2/8b
Defect
(a-c)2/4b,
(a-c)2/4b
0,
0
0
Solving Repeated Prisoner’s Dilemma Games
• For infinite horizon repeated PD, “Grim Trigger” strategy
can facilitate collusion:
– Cooperate as long as other player cooperates, but once
he defects, defect forever.
– His defection “triggers” the punishment.
– “Grim” because punishment lasts forever.
• To check if there is a symmetric equilibrium with trigger
strategies:
– Make sure that cooperating is better than defecting if
other player has cooperated.
– Make sure that “punishment” is a credible threat, that
you will actually go through with it.
When Are Trigger Strategies are NE?
• Assume other player(s) also using a trigger strategy.
• If no one has defected, all others will cooperate this period.
• If you follow the trigger strategy, i.e. cooperate, you get
M/2 this period and you get M /2 each period in the
future.
– An infinite stream of payments of M can be written as
1/(1-)* M /2.
• If you defect, you get M this period but in all future
periods you get 0.
– Total earnings thus are M.
• Thus following the strategy is optimal if:
1/(1-)* M /2 > M, or  > 1/2.
Factors that Affect the Success of Collusion
• Ability to get “monopoly profits” in collusive phase.
– Elasticity of demand.
– Ability to restrict entry into the market.
• Ability and cost to reach collusive agreement.
– How many firms.
– Similarities between firms (size, costs, etc.).
• Amount of interaction between firms.
– The shorter the “periods” the higher the discount rate
will be.
• Ability to detect cheating
– Stability of market demand.
– Transparency of firms’ actions.
Modeling Collusion
• Infinitely repeated, non-cooperative game (i.e. can write
enforceable contracts).
• Can be either price or quantity competition.
• Look for Nash equilibrium strategies that involve profits in
excess of those predicted in a one-stage game.
• Three general classes of models have been developed:
 Imperfect monitoring: Can't directly observe players’
actions.
 Learning: Players don't know everything about market and
other players, but learn over time.
 Cyclical: Business cycles affect the difficulty of colluding.
Imperfect Monitoring Models
• Key element: can't directly observe players’
actions.
– Hidden actions: prices negotiated individually with
buyers.
– Demand has random component: can’t tell if low
demand is due to market or competitors.
• With imperfect monitoring, punishment phases -- often
seen as price wars -- are triggered by unexpected demand
shocks, but the length of the punishment phase does not
depend on the size of the shock.
Learning Models
• Key element: Players don't know everything about
market and other players, but learn over time.
– Industry wide demand may not be known initially (new
market).
– Firm’s cost may not be known initially.
• Firm’s change prices/quantities as they learn, which
punishment phases.
• With learning, price wars are separated by stable collusive
periods, severe wars are triggered by demand shocks and
the length of the war depends on the severity of the shock.
Cyclical Models
• Key element:Business cycles affect the difficulty
of colluding.
– Cheating today produces certain profits whereas future
conditions (and thus profits) are not certain.
• When times are good, incentives to cheat increase if
shocks are randomly distributed.
• If a good shock is more likely to be followed by another
good shock (the shocks are related), then the temptation to
cheat increases when times are bad.
Empirical Evidence on Collusion
• Slade, 1990, uses evidence of price wars, cartel
breakdowns, and other phenomenon to provide
insight into the nature of collusion
• Looks at three “cases studies”
– JEC Cartel.
– Retail gas data.
– Nickel prices.
Detecting Collusion
• Two Stage Approach to Detecting and
Proving Collusion:
– Identify markets that are susceptible to
collusion
– Determine whether firms are pricing collusively
• Why two stages?
– Focus resources where collusion is most likely
– Evaluate ambiguous conduct
Detecting Collusion, Con’t
• Historically, what factors have made
industries susceptible to collusion?
– Fraas and Greer find markets with small
number of firms and uncomplicated setting
conducive to collusion. Also trade associations
and common sales agencies facilitate collusion.
– Hay and Kelley find similar results and
additionally that product homogeneity increases
collusion.
Screening for Collusion
• Screening = Trying to identify industries
where collusion is occurring or can occur.
– Andrew Dick finds that common screening
variables miss between 50-80% of industries
where cartels are likely to have raised price.
Proving there is Collusion
• Difficult to prove without a “smoking gun”.
– Price-fixing agreement (on paper or tape).
– Agreement to divide territory (on paper/tape).
• Identical prices are predicted in most noncollusive models.
– Unilateral, independent identical prices is
known as “conscious parallelism” and is legal.
• To prove collusion, need evidence that firms
were acting contrary to self-interest, if not
for collusive agreement.