cartels and tacit collusion - crest | center for … · 2016-06-17 · 5 cartels and competition...
TRANSCRIPT
CARTELS AND TACIT COLLUSIONAdvanced Industrial Organization 1
THIBAUD VERGÉ
CREST-LEI
ENSAE 3A / Master APE (2009-2010)
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 1 / 47
Introduction
Outline
1 Introduction
2 Basic Theoretical Model
3 Factors Facilitating Collusion
4 Informational issues
5 Cartels and Competition Policy
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 2 / 47
Introduction
Cartels and Competition Policy
Fighting cartels is at the core of competition policy.
Biggest European Cartel Cases2008 Car Glass 1384 me Saint-Gobain (896 me), Pilkington (370 me)2009 Gaz 1106 me E.O.N (553 me), GDF-Suez (553 me)2007 Elevators 992 me ThyssenKrupp (480 me)2001 Vitamins 790 me Hoffmann-LaRoche (462 me)2007 Switchgear 750 me Siemens (396 me)
In France2009 Temporary Work 94 me Manpower (42 me), Adecco (32 me)2008 Steel 575 me ArcelorMittal (309 me), KDI (169 me)2005 Mobiles 534 me Orange (256 me), SFR (220 me)
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 3 / 47
Introduction
What is Collusion?
Explicit or Tacit CollusionCollusion ≡ collusive agreement that should be outlawed
Collusion ≈ high prices, prices above the “competitive” level
Explicit Collusion: firms act through an organized cartel
Tacit Collusion: firms maintain high prices in a non-cooperativeway
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 4 / 47
Introduction
Reading List
MOTTA, M. (2004), Competition Policy: Theory and Practice,Cambridge (Chapter 4).
TIROLE, J. (1988), Theory of Industrial Organization, MIT Press(Chapter 6).
GREEN, E. and R. PORTER (1984), “Noncooperative CollusionUnder Imperfect Competition”, Econometrica, 52, pp. 87-100.
JULLIEN, B. and REY, P. (2007), “Resale Price Maintenance andCollusion”, Rand Journal of Economics, 38, pp. 983-1001.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 5 / 47
Introduction
The Main Ingredients of Collusion
Prisoners’ DilemmaP2
C NC
P1C (2,2) (−1,3)
NC (3,−1) (0,0)
Unique Nash Equilibrium: No cooperation (strictly dominantstrategies), zero payoffs.
Cooperation is Pareto dominant
When is cooperation possible?
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 6 / 47
Introduction
The Main Ingredients of Collusion
Suppose the game is repeated over time
Players could agree to cooperate (and obtain higher payoffs)
However, at any stage, a player would have a strong incentive todeviate
In order for cooperation to be sustainable, the rival player needs tobe able to detect the deviation . . .
. . . and to enforce the “agreed” (credible) punishment
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 7 / 47
Basic Theoretical Model
Outline
1 Introduction
2 Basic Theoretical Model
3 Factors Facilitating Collusion
4 Informational issues
5 Cartels and Competition Policy
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 8 / 47
Basic Theoretical Model General Framework
(Tacit) Collusion in a General Framework
The important variables1 πC the per-firm collusive profit
2 πD the profit of the deviating firm
3 πNC the per-firm non-cooperative profit (i.e. the equilibrium payoffin the Nash-equilibrium of the static game)
4 δ the discount factor
Idea: Collusion is a SPNE ifThe short term gain of deviation is lower than the long term loss(punishment).
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 9 / 47
Basic Theoretical Model Finite Number of Repetitions
Unique Subgame Perfect Nash Equilibrium
Backward Induction: Start with the last periodLast stage⇐⇒ Static GameUnique Nash Equilibrium, without cooperation.
Penultimate PeriodLast period profits are not affected by the historyPenultimate stage⇐⇒ Static GameUnique Nash Equilibrium, without cooperation.
Unique SPNENE of the static game repeated over time. No cooperation.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 10 / 47
Basic Theoretical Model Infinitely Repeated Game
Multiple Subgame Perfect Nash-Equilibria
General ResultThe NE of the static game is always a SPNE of the infinitelyrepeated game
No cooperation is always a SPNE.
FormallyIn an infinitely repeated game, strategies are complex
History: ht =(s1,0, s2,0; . . . ; s1,t−1, s2,t−1
)Choice at t depends on ht
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 11 / 47
Basic Theoretical Model Infinitely Repeated Game
Cooperation (collusion) as a SPNE?
QuestionCan collusion be an equilibrium?
Are there other equilibria?
Focus on particular equilibria with trigger strategiesIf nobody has deviated in the past, each player chooses tocooperate at date t
If somebody deviates from the collusive path at date t − 1, thenfrom t onwards players revert to the non-cooperation solution
Trigger strategies or reversal to Nash
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 12 / 47
Basic Theoretical Model Infinitely Repeated Game
Existence condition for a collusive equilibrium
Short-term gain (at t = 0)
Deviation: πD
Collusion: πC
Gain: πD − πC
Long Term Loss
Deviation:∑+∞
t=1 δtπNC = δ
1−δπNC
Collusion:∑+∞
t=1 δtπC = δ
1−δπC
Loss: δ1−δ
(πC − πNC)
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 13 / 47
Basic Theoretical Model Infinitely Repeated Game
Tradeoff Today / Future: Summary
Collusion as an equilibrium iff loss > gain
Loss: δ1−δ
(πC − πNC)
Gain: πD − πC
δ1−δ
(πC − πNC) > πD − πC ⇐⇒ δ > δ̃ ≡ πD−πC
πD−πNC
Conclusion
If firms are patient enough(δ > δ̃
)there exists a subgame perfect
Nash equilibrium (with trigger strategies) with collusion.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 14 / 47
Basic Theoretical Model Collusion under Price Competition
Price Competition with Perfect Substitutes
Bertrand Competition between n firms
πNC = 0
πC = ΠM
n
πD = ΠM (−ε)
Discount Factor Threshold
δ̃ =ΠM − ΠM
nΠM − 0
=n − 1
n= 1− 1
n
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 15 / 47
Factors Facilitating Collusion
Outline
1 Introduction
2 Basic Theoretical Model
3 Factors Facilitating Collusion
4 Informational issues
5 Cartels and Competition Policy
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 16 / 47
Factors Facilitating Collusion
Factors Facilitating Collusion
1 Concentration2 Entry3 Cross-ownership4 Regularity and frequency of orders5 Evolution of demand6 Symmetry7 Multi-market contacts8 Excess Capacities9 Price transparency / Exchange of information
10 Antitrust enforcement11 Leniency programs
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 17 / 47
Factors Facilitating Collusion Structural Factors
Concentration and Entry
ConcentrationMany identical firms under price competition
The gains from collusion are smaller the larger the number of firms
However, non-cooperation and deviation profits do not depend on the number offirms
Collusion is easier to sustain when the number of players is smaller
EntryThe easier the entry into an industry, the more difficult to sustain collusiveprices
If the entrant behaves aggressively, the incumbent firms (cartel members) willhave to react by lowering prices
If the entrant is willing to join the cartel, then collusion is more difficult as thenumber of firms increases
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 18 / 47
Factors Facilitating Collusion Structural Factors
Cross-ownership
Symmetric case with 2 firmsTwo identical firms competing in prices, each owning a share α ofthe rival firmπNC = 0πC = (1+α)ΠM
2
πD = ΠM + 0
Therefore: δ̃ =ΠM− (1+α)ΠM
2ΠM−0 = 1−α
2
ConclusionCollusion is easier when firms own shares of their rivals
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 19 / 47
Factors Facilitating Collusion Structural Factors
Evolution of DemandRotemberg and Saloner, American Economic Review, 1986
Stochastic DemandAt each period, demand can be either high (dH) or low (dL)
Demand shocks are i.i.d.Equal probabilitiesWe denote by:
ΠMH and ΠM
L the corresponding monopoly (collusive) profits
ΠM =ΠM
L +ΠMH
2 the expected monopoly profit
Collusion and deviation expected profits (state s = H, L at t)
Collusion: ΠMs
2 + (δ + . . .+ δt + . . .) ΠM
2 = ΠMs
2 + δΠM
2(1−δ)
Deviation: ΠMs + (δ + . . .+ δt + . . .)0 = ΠM
s
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 20 / 47
Factors Facilitating Collusion Structural Factors
Evolution of DemandRotemberg and Saloner, American Economic Review, 1986
Collusion is sustainable whenever:
ΠMs2
+δΠM
2(1− δ)> ΠM
s ⇐⇒δΠM
1− δ> ΠM
s ⇐⇒ δ > δ̃s ≡ΠM
s
ΠM + ΠMs
ConclusionNote that ΠM
H > ΠML ⇐⇒ δ̃H > δ̃L
Collusion is more difficult to sustain when demand is high
Short term gains are higher when demand is high
Long term (expected) losses are the same in the two states
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 21 / 47
Factors Facilitating Collusion Structural Factors
Growing and Declining Industries
Evolving industriesDemand is certain but changes over time
Monopoly profit in period t is given by θt ΠM
ConclusionWe then have: δ̃(θ) = 1
2θ
Collusion is easier to sustain in growing industries (θ > 1) than indeclining industries (θ < 1)
At t = 0, short term gains are identical but long term losses arehigher when future demand is expected to be higher
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 22 / 47
Informational issues
Outline
1 Introduction
2 Basic Theoretical Model
3 Factors Facilitating Collusion
4 Informational issues
5 Cartels and Competition Policy
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 23 / 47
Informational issues Temporary Price Wars
Identifying Deviations to Sustain CollusionGreen and Porter, Econometrica, 1984
Stigler (JPE, 1964): Collusion would break down because ofsecret price cuts
Green and Porter: If actual prices are not observable, collusion ismore difficult to sustain, but could still arise in equilibrium.
With secret price cuts, a firm does not know if a low demand isdue to a price cut (deviation from the collusive path) or a negativedemand shock (on the collusive path)
Idea: trigger a price war (reversal to Nash) but for a limitednumber of periods
A period of low prices (price war) does not necessarily meanthat there is no collusion
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 24 / 47
Informational issues Temporary Price Wars
Green and Porter, Econometrica, 1984A simplified version
n identical firms selling a homogenous product.Price competition.Uncertain demand shocks (i.i.d over time): demand can be eitherlow (D(p) = 0 for any price p) with probability α or high(D(p) > 0).Firms never observe the actual state of demand.Firms do not observe prices set by rival firms.
→ A firm facing a zero demand does not know whether it is:1 Because the demand was low.2 Because some rival(s) undercut its price.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 25 / 47
Informational issues Temporary Price Wars
Temporary Price Wars
Temporary Price Wars in EquilibriumEach firm sets the collusive price pm at the outset of the game.It continues to do so as long as all firms have positive demands.When at least one firm observes zero demand (which is assumedto be common knowledge), the industry enters the “punishmentphase”: each firm sets price equal to marginal costs for the next Tperiods.Collusive phases restarts at the end of the “price war”.
Some notationsV + represents the present discounted value of a firm in eachperiod which belongs to the collusive phase.V− represents the present discounted value of a firm at the startof the punishment phase.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 26 / 47
Informational issues Temporary Price Wars
Temporary Price Wars
Relationships between V + and V−
The following two equations holds:
V + = (1− α)
(π (pm)
n+ δV +
)+ αδV− and V− = δT V +
This yields:
V + =(1− α)
π(pm)n
1− (1− α) δ − αδT +1 and V− = δT V +
Collusion is possible as long as
V + ≥ (1− α)(π
(pm)
+ δV−)
+ αδV−
i.e. (after some computation . . .)
[δn(1− α)− (n − 1)] + δT +1 (αn − 1) ≥ 0 (1)
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 27 / 47
Informational issues Temporary Price Wars
When is collusion feasible?
Three possible cases
1 If α < α1 ≡ 1− n−1nδ
(which may be impossible if δ is too small or n is too large),then the left term is positive while the right term is negative. There may thereforeexist a values of T that satisfy (1).
2 If α1 ≥ α < α2 ≡ 1n , both terms are negative and equation (1) never holds.
3 If α ≤ α2, the left term is negative while the right term is positive. But:
n − 1− δn(1− α) = (n − 1)(1− δ) + δ(αn − 1) > δ(αn − 1) > δT +1(αn − 1),
and thus equation (1) cannot hold.
Collusion is feasible whenever:
α < α1 and T ≤ T min ≡ (ln δ)−1 ln(δn(1− α)− (n − 1)
δ(1− αn)
)(2)
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 28 / 47
Informational issues Temporary Price Wars
Some Conclusions
Temporary price wars may help sustaining collusion whendeviations are not easily observable.
Collusion is not feasible if the probability to be in the bad state istoo high.
The “punishment period” needs to be longer when the probabilityof the being in the bad state increases (i.e. ∂T min/∂α > 0).
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 29 / 47
Informational issues Some other informational problems
Information Exchange
Past (or Present) Prices and QuantitiesExchanging information about past (or present)
(Individual) prices and/or quantities
Aggregate demand
helps collusion.
French mobile phone cartel (e(534m) fines for Orange, SFR and BouyguesTelecom)
Future PricesPrivate (“cheap talk”) or public (credible) announcements about future prices help tosustain collusion
Airline Tariff Publishers (ATP) in the US
British Airways (fuel surcharges)
Collusion in multi-unit auctions
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 30 / 47
Informational issues Some other informational problems
Other Practices that Facilitate Collusion
Most-Favored Nation ClausesAlso called price matching clauses
Way of exchanging information through consumers)
Note that price comparison engines (Kelkoo, . . . ) or online agencies co-ownedby rivals (Expedia, Opodo, Orbitz, . . . ) can also serve the same purpose.
Resale Price MaintenanceJullien and Rey (RAND Journal of Economics, 2007)
Resale price maintenance prevents retailers from reacting to local demandshocks.
Helps to sustain collusion among producers since deviations are more easilydetected.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 31 / 47
Informational issues Vertical Relations
Resale Price Maintenance and CollusionJullien and Rey, Rand Journal of Economics, 2007
2 differentiated producers, selling through different retailers.Price competition between retailers.
Uncertain local demand shocks (i.i.d over time and acrossproducts):
Di(pi ,pj
)= d + εi − pi + σpj , with εi ∼ U ([−∆,∆]) .
The exact terms of the contracts between a manufacturer and itsretailer is never observed by the rival manufacturer.
Retail prices are observed at the end of each period.Contracts consists of two-part tariffs (of the form Ti(q) = wiq + Ai )and (only allowed) an imposed retail price.
A new retailer at each period (⇔ myopic retailers + no long termcontracts). This implies that retailers are in essence passive.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 32 / 47
Informational issues Vertical Relations
Equilibrium of the Static Game
Timing of the stage game1 Producers make take-it-or-leave-it offers.2 Retailer Ri observes εi and chooses its retail price pi (unless RPM
is used).3 Demands and profits are realized.
Retailer Ri ’s pricing decision
pi = arg maxp (pi − wi)(
d + εi − pi + σpej
)=
d + εi + wi + pej
2=
d + wi + pej
2+εi
2= pe
i +εi
2
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 33 / 47
Informational issues Vertical Relations
Equilibrium of the Static Game
Retailer Ri ’s expected profit
ERi (wi) = E((
pei +
εi
2− wi
)(d + εi −
(pe
i +εi
2
)+ σpe
j
))(3)
Manufacturer Mi ’s expected profit
EMi = E(
wi
(d + εi −
(pe
i +εi
2
)+ σpe
j
))+ ERi
= E((
pei +
εi
2
)(d + εi −
(pe
i +εi
2
)+ σpe
j
))= pe
i
(d − pe
i + σpej
)︸ ︷︷ ︸
≡π(
pei ,p
ej
)+
14
E(ε2
i
)︸ ︷︷ ︸≡v(∆)= ∆2
12
= π(
pei ,p
ej
)+ v (∆) .
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 34 / 47
Informational issues Vertical Relations
Equilibrium of the Static Game
Equilibrium of the Static GameExpected equilibrium prices:
pei = arg maxpπ
(p,pe
j
)=
d + σpej
2=⇒ wN
1 = wN2 = 0.
“Competitive” retail prices: pN1 = pN
2 = pN ≡ d2−σ
Manufacturer’s expected profits: EMN1 = EMN
2 = Π(pN)+ v (∆).
RPM in the Static Game
If Mi imposes a retail price pi , its expected profit is only π(
pi ,pej
).
It thus looses the benefit of retail prices that adjust to localdemand shocks (i.e., looses v(∆)).RPM is never used in the static game.
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 35 / 47
Informational issues Vertical Relations
Scope for Collusion
DefinitionCollusive prices:(
pM1 ,p
M2
)= arg max(p1,p2) (π (p1,p2) + π (p2,p1) + 2v (∆)) .
Recall: Π (p) = π(p,p).
Assumptions
k ≤ Π(
pN)
+ v (∆) ≤ Π(
pM)
First part: k is a per-period fixed cost paid by the manufacturers.The assumption ensures that selling is indeed profitable.Second inequality: There is scope for collusion (under RPM).
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 36 / 47
Informational issues Vertical Relations
Collusion without RPM
Denote by ΠF (resp. ΠF ) the maximal (resp. minimal) averageper-period profit that can be obtained in a fully symmetricequilibrium, and by sF (resp. sF ) the corresponding strategy.
Most profitable strategy: the strategy sF is of the form “chargean expected price equal to pF as long as past retail prices belongto[pF − ∆
2 ,pF + ∆
2
], and sF otherwise.
Necessary Condition 1: Perfect Detection
maxp
π(
p,pF)− Π
(pF)≤ δ
1− δ
(Π(
pF)
+ v (∆)− ΠF
)(PD)
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 37 / 47
Informational issues Vertical Relations
Collusion without RPM
Necessary Condition 2: Small DeviationsSmall deviations are less easily detected.
Indeed, if p ∈[pF −∆,pF + ∆
], the deviation may be undetected.
Therefore, the following condition must also hold for anyp ∈
[pF −∆,pF + ∆
]:
π(
p,pF)− Π
(pF)≤ δ
1− δ
∣∣p − pF∣∣
∆
(Π(
pF)
+ v (∆)− ΠF
)(SD)
Collusion without RPMMost profitable collusive equilibrium is such that pF maximizesΠ(pF)+ v(∆) subject to (PD) and (SD).
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 38 / 47
Informational issues Vertical Relations
Collusion with RPM
When RPM is used on the collusive path, all deviations areautomatically detected.
Necessary and sufficient condition
maxp
π(
p,pRPM)
+ v (∆)− Π(
pRPM)≤ δ
1− δ
(Π(
pRPM)− ΠRPM
)(RPM)
Collusion with RPMMost profitable collusive equilibrium is such that pRPM maximizesΠ(pRPM) subject to (RPM).
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 39 / 47
Informational issues Vertical Relations
Comparing without / with RPM
Constraints (PD) and (RPM)
maxp
π(
p,pF)
+ v(∆) +δ
1− δΠF ≤
11− δ
(Π(
pF)
+ v (∆))
(PD)
maxp
π(
p,pRPM)
+ v(∆) +δ
1− δΠRPM ≤
11− δ
Π(
pRPM)
(RPM)
RPMFacilitates collusion:
ΠRPM < ΠF , i.e., harsher punishments.Perfect detection for all deviations (no equivalent to (SD)).
But yields lower expected profits for identical average prices:additional term v(∆) in (PD).
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 40 / 47
Informational issues Vertical Relations
Welfare Effect and Conclusions
There exist values of the parameters (especially forintermediate values of the discount factor), for which RPM wouldbe used by firms to facilitate collusion.
Retail prices are then higher on average.
But they no longer react to local shocks.
With retail cost shocks, rigid prices are bad for consumers and totalwelfare.But for local demand shocks, consumers prefer rigid prices.
Overall, ambiguous effect with local demand shocks. But, inthe linear demand example, RPM can only be bad for welfareif σ ≤ 2
3 .
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 41 / 47
Cartels and Competition Policy
Outline
1 Introduction
2 Basic Theoretical Model
3 Factors Facilitating Collusion
4 Informational issues
5 Cartels and Competition Policy
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 42 / 47
Cartels and Competition Policy Detecting Collusion
Detecting Collusion
Need for hard evidence (minutes, memo, . . . )
What is a collusive price?
Uncertainty about price or demand
Problems using historical price data (i.e. price evolution)
“Price parallelism” can be explained by common shocksUnless “extreme parallelism” (e.g. Dyestuffs)
Facilitating practices can only be stopped but not used to findfirms guilty
Tacit (rk: requires coordination) or explicit collusion?
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 43 / 47
Cartels and Competition Policy Deterring Collusion
Deterring Collusion (Ex ante policies)
Heavy punishment when detectedExpected fines (fine x probability of audit), damages to “injured”parties, prison sentences
Black list of facilitating practices
Exchange of information about individual prices and/or quantitiesBest price clausesMinority shareholdings. . .
Auction design
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 44 / 47
Cartels and Competition Policy Leniency Programs
Leniency programs
As used against organized crime in order to obtain hard evidence
Leniency ≡ immunity from fines and/or prisonBreaking trust within the cartel (preventing cartel formation but alsodetecting cartel more easily)Saving resources for antitrust authorities
Ambiguous theoretical effects on deterrenceReduces the gains from collusionReinforces the possibility of punishment
Optimal leniency programs
See Spagnolo (mimeo, 2000), Motta and Polo (IJIO, 2003), . . .Automatic full immunity (from fines but also damages, possibly evenrewards) even after an investigation has started
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 45 / 47
Cartels and Competition Policy Leniency Programs
Leniency programs in the U.S.
First introduced in 1978 but redesigned in 1993
Automatic immunity from fines and prison sentences
Provide hard evidence before an investigation has begunFirst arrived only and conditional on leaving the cartel
Discretionary leniency for evidence provided once theinvestigation has been launched
Number of applicants rose from 1 a year to 2 a month
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 46 / 47
Cartels and Competition Policy Leniency Programs
Leniency program in the EU
First introduced in 1996 (inefficient) but redesigned in 2002
Complete immunity from finesProvide hard evidence even after investigation has begunFirst arrived only but discretionary partial immunity for second andthird arrived if it helps convict firmsNot for the ring-leader
Recent case (November 2006): Producers and traders ofsynthetic rubber fined a total of e519m
Bayer applied for leniency in December 2002 (was granted fullimmunity from a e204m fine)Once the investigation had started, Dow applied for leniency (wasgranted partial immunity - 40% - from a e107m fine)Eni (e272m), Shell (e160m), Unipetrol (e17.6m) and Trade-Stomil(e3.8m) were also involved
THIBAUD VERGÉ ( CREST-LEI ) Collusion Advanced IO 1 47 / 47