European Journal of Law and Economics, 17: 5–19, 2004 c 2004 Kluwer Academic Publishers. Manufactured in The Netherlands.
Leniency Program: A New Tool in Competition Policy to Deter Cartel Activity in Procurement Auctions KARINE BRISSET LIONEL THOMAS CRESE, Universit´e de Franche Comt´e, Besanc¸on, France
[email protected] [email protected]
Abstract This paper discusses the impact of a leniency program on incentives within cartels. The objective of this program is to encourage a cartel member to confess and implicate his co-conspirators with hard evidence about their collusive agreement. We develop a simple model of cartel behavior under a first-price sealed-bid procurement auction and we show how an effective leniency program can prevent the internal coordination of cartel members. Keywords: cartels, leniency programs, competition policy, auctions JEL Classification: L4, D43, D82
1.
Introduction
Cartels in auctions and procurement auctions are prohibited by article 81 E.C. Two reasons can explain this prohibition. Firstly, in a procurement auction, collusion between firms constitutes a restriction of competition and therefore increases the equilibrium price at which the market is allocated. Therefore, because of the social costs of public funds, it is necessary to prevent any collusion between firms in public procurement auctions. Secondly, in practice, the cartel leader, i.e. the cartel’s designed winner, is not systematically the firm with the lowest cost, giving rise to a loss of efficiency and therefore to a loss of welfare. For all these reasons, cartels must be investigated and sanctioned by the competition authorities. However, the competition authorities have often faced serious difficulties in detecting and obtaining sufficient hard evidence to prosecute cartel members. Because they are illegal, cartels often organize clandestine meetings and communication, making detection very hard. The prosecution of secret cartels is therefore difficult and costly for the authorities. Reports by the French competition authority relate that successful prosecutions against cartels in public and private procurement auctions often result from their being reported or from indiscreet cartels. In other cases, prosecution has failed because of a lack of hard evidence or because of a flaw. It is rather difficult to estimate the efficiency of cartel law enforcement. However, in France, we have about 170000 annual public markets and only 15 prosecutions against cartels in private and public procurement auctions. This does not mean that there are few cartels in procurement auctions but rather that the risk of being investigated and prosecuted
6
BRISSET AND THOMAS
is low. So, even if the sanctions are high, it will be insufficient to prevent cartels if the risk of detection is perceived as low. This risk depends not only on the probability of an audit but also on the probability of a successful prosecution. If the authorities cannot increase the number of investigations, the only way to overcome this enforcement problem is to provide incentives to cartel members to elicit information on the existence of the cartel. In fact, one way to fight cartels is by creating private incentives to play one member against the others. In this objective, in 1978, the U.S. Department of Justice was the first to introduce an Amnesty program and granted amnesty to the whistle-blowing cartel member who revealed the cartel’s existence before the opening of an investigation. But this program was unsuccessful (one application per year) for two reasons. The first one is that amnesty was not automatic but depended on the discretion of the Department of Justice. And the second was that, before an investigation, the risk of detection is rather low. So, if a firm decides to join the cartel, it is because it judges that it does pay despite the risk. But prior to any investigation, there is no increased risk, so the firm judges that the amnesty program does not pay because it simply cancels out any positive gains, by exempting it from any fine. That is why in 1993, the USA revised the Amnesty program with the following conditions. It automatically grants Amnesty to the first firm that reports the illegal activity and completely cooperates before an investigation is under way. Another change is to make amnesty possible even after an investigation is underway. Since this revision, the number of applications has multiplied to more than 20 per year and led to dozens of convictions and fines totaling well over $ 1 billion. Several other legislatures also have leniency programs. In 1996, the European Commission introduced its own “leniency program” leading to significant reductions in or exemptions from fines with the following conditions: – A 75–100% reduction in fines to the first firm that cooperates fully with the authorities – A 50–75% reduction in fines to the first firm that cooperates fully with the authorities after an investigation is under way. – A 10–50% reduction in fines to all the other firms whose confession makes the agency’s proof easier. Five years after its implementation, the effectiveness of this program being mitigated, the European Commission decided this year to improve its policy. The objective of the new program is to improve the transparency and certainty of the conditions under which any reduction of fines will be granted. Thus, the Commission will grant immunity from any fine if the firm is the first to submit evidence, which, in the Commission’s view, may enable it to adopt a decision to carry out an investigation in the sense of Article 14(3) of Regulation no. 17 in connection with an alleged cartel affecting the Community; or if the firm is the first to submit evidence which in the Commission’s view may enable it to find a violation of Article 81 EC in connection with an alleged cartel affecting the Community. In addition, if a firm provides added information, it can benefit from a reduced fine as follows:
LENIENCY PROGRAM
7
– first undertaking: a reduction of 30–50% – second undertaking: a reduction of 20–30% – subsequent undertakings: a reduction of up to 20%. Canada and the U.K. also have their own leniency programs, which are based in many respects on the U.S. experience. Germany announced a program in May 2000 and France adopted its own leniency program in 2001.1 One can then wonder about the effectiveness of such a program on the prevention and the detection of cartel activity. At present, there are few theoretical studies on these programs. Their effects are analyzed within the framework of tacit repeated agreements. Thus, Motta and Polo (1999) are interested in the implementation of an optimal mechanism of detection within a dynamic framework. In their analysis, the leniency program applies after the beginning of an investigation by the authority. The authors determine the conditions under which a leniency program can be adopted when it is not possible to prevent collusion. The optimal program consists in a total exemption from fines for all firms that collaborate with the authority. Nevertheless, they underline the pro-collusive effect of such a program because it may prevent deviant behavior within the cartel. Spagnolo (2000a) analyses the effects of a leniency program on incentives to deviate from the cartel’s agreements. He shows that a program of moderate reduction in fines before any investigation is rather ineffective. On the other hand, it is possible to encourage firms to report the collusive agreement before an investigation by rewarding them. In particular, when the fines are pure transfers and it is not possible to prevent the formation of a cartel, it is efficient to pay a maximum reward to the firm that cooperates fully. Spagnolo (2000b) analyses the effects of a leniency program on the incentive to deviate from the collusive agreement in the context of several markets within a dynamic framework. He also highlights the pro-collusive effect of such a program when it grants exemption from fine for disclosure before an investigation. Ellis and Wilson (2001) develop a simple model of cartel behavior under the conditions of Bertrand competition with differentiated products and find that in some situations introducing leniency programs causes the cartel to cease to be viable. However, in other cases, the leniency program only serves to raise cartel profits. None of these papers considers the problems of coordination within the cartel related to the asymmetry of information between members. In procurement auctions, in general, firms do not know how much each of their fellow cartel members is willing to submit for the market being allocated. If it is admitted that the collusive surplus and its division can depend on the firms’ ads, the latter may have the incentive to manipulate the information on their cost to argue for a bigger share. However, this manipulation can involve inefficiencies within the cartel and can prevent the maximization of the collusive surplus. It is thus in the interest of the cartel to design an optimal mechanism to be used in deciding who receives the market and how the collusive surplus is distributed. As the incentives of the mechanism will depend on the rule of division, this will have to be a firm, whose cost is high but does not behave like a firm, which would have a low cost and reciprocally. The other
8
BRISSET AND THOMAS
major problem is to ensure the participation of the firms. In practice, within the framework of repeated agreements, the collusion can be based on reciprocity. Within a static framework, it is necessary to define a rule of the cartel’s surplus division such that the firm does not feel injured by this illicit agreement. This must result in a rule that ensures the satisfaction of the individual rationality constraint. Otherwise, it could indeed be advantageous for a firm not to take part in the collusion. The purpose of this paper is precisely to investigate this problem and examine how a leniency program can disturb the cartel’s optimal mechanism between firms in a firstprice sealed-bid procurement auction. In particular, we show that it is possible to prevent the cartel’s formation by rewarding firms that report hard evidence to the competition authority. The remainder of the paper is organized as follows. Section 2 presents the model and analyzes the formation of the cartel in the absence of a leniency program considering nevertheless that the cartel is confronted with the risk of detection by the competition authority. Section 3 introduces the leniency program such as it actually applies in Europe. Section 4 analyses how to improve the efficiency of these programs.
2.
The model
We consider that a public market about furniture is allocated by a first-price sealed-bid auction where n firms may either have a high cost (c = 1) with probability p or a low cost (c = 0) with probability (1 − p). All firms are risk-neutral. Only the firm knows its true cost. The buyer’s reserve price2 is common knowledge and is equal to r with 1 > r > 0. We consider that firms can form a collusive agreement and that the relevant cartel consists of all firms in the market. We assume that the competition authority can audit the cartel with an arbitrary probability α. In the event of a successful prosecution, which happens with probability µ, the procurement auction may be cancelled and a maximal fine A¯ 3 imposed on each firm. If firms decide to form a collusive agreement, it is necessary to define an optimal mechanism, which ensures the efficiency of the cartel, i.e. which selects one of the low cost firms as the cartel leader.
2.1.
The cartel mechanism
When they collude, firms organize a knockout4 before the procurement auction to select any member with a cost equal to c = 0 as the cartel leader. The mechanism is the following: If r is the bid tendered by the cartel at the procurement auction,5 every member with c = 0 receives an equal share of the cartel profit (r − 0). Firms with c = 1 do not receive any transfer because they cannot increase the cartel profit. In this context, we can show that this mechanism is incentive compatible. Indeed, a bidder with c = 0 has no incentive to report cˆ = 1: when he reports cˆ = 1, his cartel payoff is zero but when he reports cˆ = 0, his cartel payoff is strictly positive. Likewise, a bidder with c = 1 has no incentive to report cˆ = 0. Indeed, assume that the
9
LENIENCY PROGRAM
number of firms that report c = 0 is s and to simplify, consider a situation with no risk of prosecution: – If this firm is the designated winner (with a probability 1/s), it must pay the cartel a total transfer of (s−1)r and receives r from the buyer. s – If he is not the designated winner (with a probability (1 − 1/s)), it receives from the winner a transfer of rs . Comparing its payoff when it reports cˆ = 0 to its zero payoff when it tells the truth and reports c = 1, we can see that: 1 r 1 (s − 1)r r −1 π(ˆc = 0/c = 1) = 1 − + − +r −1 = A¯
(3)
It rises directly from the comparison between (1) and (2).
Thereafter, we will consider that the condition (3) is satisfied in order to highlight the impact of a leniency program when it is not possible to prevent the cartel’s formation. Under these conditions, we will first study the impact of such a program as defined by European law. 3.
Effects of the corporate leniency program on cartel behavior
The current European leniency program consists in granting various reductions in fines to cartel members who provide explicit and hard evidence of its existence. In particular, a firm that reports the collusive agreement and provides explicit evidence before the opening of an investigation can benefit from a total exemption from fines. Our objective here is to understand in which conditions this program can prevent the cartel’s formation or facilitate the reporting of a cartel. To do this, we consider that the rules of the leniency program are as follows: – any member who reports the agreement before any investigation benefits from an exemption from fines, – any member who provides explicit evidence of the existence of the cartel after the opening of an investigation but before the authorities have any formal evidence benefits from a reduction in fines and pays only β A¯ with 0 < β < 1. Taking these rules into account, we can state the following proposition: Proposition 2. Such a leniency program is unable to prevent cartel formation. More precisely, firms do not find it profitable to report the cartel before the opening of an investigation.
11
LENIENCY PROGRAM
In the event of an investigation, the program becomes an incentive provided that: n
(1− p ) (1 − µ)r n(1− p) β
(1− p ) r (1 − µ) n(1− p)
µ
Proof: By reporting the collusive agreement before an investigation, a firm will not pay a fine and will have zero profit. However taking condition (3) into account, we know that a low cost firm perceives an expected positive profit, before any procedure of investigation, higher than the positive expected profit than it would have in competition. Thus, it is not profitable to report the collusive agreement. As for high cost firms, they do not gain anything if they report; we can assume that they will not report. After the opening of an investigation, the incentive to report will also depend on the behavior of the other firms. High cost firms will not have any reason to report. On the other hand, low cost firms find themselves in a situation of prisoner’s dilemma where the game is as follows:
c = 0 firm R
N.R
¯ −β A¯ −β A;
¯ −β A¯ − A;
¯ − A¯ −β A;
(1− p ) ¯ (1 − µ)r n(1− p) − µ A;
c = 0 firm R N.R
n
n
(1− p ) ¯ (1 − µ)r n(1− p) − µ A
R: report, N.R: no report.
It is easy to verify that it is a strictly dominant strategy for a low cost firm to report collusion when β < µ − (1 − µ)r
(1 − p n ) 1 . n(1 − p) A¯
(1− p n ) 1 As β > 0, this implies also that A¯ > r (1 − µ) n(1− . p) µ
We can then state the following corollary: Corollary 3. When it is efficient, the leniency program after the opening of an investigation has a pro-collusive effect ex ante. Proof: n We know that when an investigation is underway (α = 1) and when β < µ − (1 − (1− p ) 1 µ)r n(1− , firms report the collusive agreement. So, the low cost firm’s expected payoff p) A¯
12
BRISSET AND THOMAS
before an investigation then becomes: Eπc (c = 0, β) = (1 − α)r
(1 − p n ) − αβ A¯ n(1 − p)
This profit is by assumption greater than Eπc (c = 0), the individual profit within the cartel in the absence of a leniency policy. Consequently, if the low cost firm’s individual rationality constraint would not be satisfied in the absence of a leniency policy, it could become so in the presence of such a program. Thus, we have just shown that the current leniency program presents inefficiencies. On the one hand, it is of no interest before an investigation; on the other hand, it supports the formation of a cartel ex ante when it encourages the reporting of a cartel in the event of an investigation. In what follows, we study how to improve the performance of a leniency program and show that it is even possible to prevent the cartel’s formation by rewarding firms that report the collusive agreement. 4.
How to improve the efficiency of a leniency program
Our conclusions on the pro-collusive effect of leniency programs are similar to those of Spagnolo (2000a, 2000b) and Motta and Polo (1999) even if the framework is different. Nevertheless, none of these authors is interested in the problem of the optimal coordination between members within the cartel. Apparently, without leniency program or in the presence of a simple program of reduction in fines, the internal mechanism is very easily conceivable within our auction framework. In particular, the mechanism does not grant any transfers to high cost firms, which does not prevent the latter revealing their true value and not reporting the cartel. On the other hand, if the leniency program becomes more attractive by granting a reward to members who report the collusive agreement, the stakes are no longer the same. In the United States there already exists, for federal markets, a system of premiums to reward any person, outside of the collusive agreement, who reports any fraud in these markets. This premium accounts for 30% of the fines paid by defrauders when they are detected and condemned. Our idea is to analyze how the application of such a system to cartel members can disturb the cartel and prevent the formation of the collusive agreement. We thus consider the following program: – Any cartel member providing explicit evidence of the existence of the cartel before an ¯ investigation by the authorities receives a monetary reward, λ A. – Any cartel member who provides explicit evidence of the existence of the cartel after the beginning of an investigation but before the authorities have formal evidence, benefits from a reduction in fine and pays only β A¯ with 0 < β < 1.
13
LENIENCY PROGRAM
Granting a reward for any explicit report will oblige the cartel to bribe high cost firms to induce them not to report collusion. Under these circumstances, what becomes of the cartel’s optimal mechanism? Let us suppose that the cartel grants a transfer T to any high cost firm. The cartel’s residual profit is then shared between all low cost firms. First, we study the subgame after the opening of an investigation.
4.1.
Effects of the leniency program after an investigation is underway
Let us consider at this point that the authorities have just begun an investigation but that they do not have reasonable evidence to prove the existence of a collusive agreement. Under these conditions, we analyze the effects of the leniency program on the behavior of both types of firms, knowing that the cartel bribes high cost firms. Suppose for the moment that the collusive agreement was formed in the preceding step and that it could imagine a coordination mechanism ensuring the participation of all firms. Suppose that nobody reports the agreement at this point, so the low cost firm’s interim payoff when it reveals the truth is: n−1
Eπ2 (c = 0) = (1 − µ)
j
Cn−1
j=0
(r − 0 − (n − 1 − j)T ) n−1− j (1 − p) j − µ A¯ p j +1
The first term defines its expected payoff taking into account the fact that there exists among the other members a random number j of low cost firms. The second term defines the fine to be paid in the event of detection by the authorities, which happens with probability µ. After simplification, we obtain:
p − pn 1 − pn Eπ2 (c = 0) = (1 − µ) (r − 0) −T n(1 − p) 1− p
− µ A¯
¯ If this firm reports the collusive agreement, it will pay: −β A. In the absence of any report, the high cost firm’s interim expected payoff is: Eπ2 (c = 1) = (1 − µ)T − µ A¯ Insofar as it receives a transfer from the cartel, it can now be prosecuted in the event of ¯ Then, the incentive to detection. So, if it reports the collusive agreement, it will pay: −β A. report (R) will depend on the game between firms with the same cost and between firms with different costs. The equilibrium of these games can be highlighted in the following diagrams:
14
BRISSET AND THOMAS Game between 2 low cost firms under the assumption that high cost firms do not report Firm 1 R
N.R
¯ −β A¯ −β A; ¯ − A; −β A¯
¯ − A¯ −β A;
Firm 2 R N.R
Eπ2 (c = 0); Eπ2 (c = 0)
Game between two high cost firms under the assumption that low cost firms do not report Firm 1 R
N.R
¯ −β A¯ −β A; ¯ −β A¯ − A;
¯ − A¯ −β A;
Firm 2 R N.R
Eπ2 (c = 1); Eπ2 (c = 1)
Game between two asymmetric firms Low cost firm R
N.R
¯ −β A¯ −β A; ¯ −β A¯ − A;
¯ − A¯ −β A;
High cost firm R N.R
Eπ2 (c = 0); Eπ2 (c = 1)
Under the assumption that −β A¯ > Eπ2 (c = 0) ≥ Eπ2 (c = 1), “report” in the event of an investigation is a strictly dominant strategy for all firms. In what follows, we will consider that this assumption is satisfied and thus that the leniency program is effective ex post. Our objective is then to analyze how an effective leniency program can disturb the cartel ex ante i.e. before an investigation. 4.2.
The effects of an effective leniency program
By granting a reward to cartel members who report the collusive agreement before an investigation, we can disturb the cartel at various levels: – That encourages high cost firms to assert payment within the cartel under the credible threat of reporting the collusive agreement in the event of non-satisfaction. – that can then prevent the formation of the cartel by making the participation non rational for low cost firms, – or that can result in not being able to define an optimal revelation mechanism within the cartel.
15
LENIENCY PROGRAM
In fact, the cartel must now overcome three types of constraint: – to make sure that the firms do not find it profitable to hide information on their cost: incentive constraint. – to ensure the voluntary participation of the firms: individual rationality constraint. – to make sure that no firm, whatever its type, finds it profitable to report the collusive agreement before an investigation: ex ante non-denunciation constraint. Incentive constraints. For a low cost firm, it is necessary that this firm does not have an incentive to announce cˆ = 1: Eπc (ˆc = 0/c = 0) ≥ Eπc (ˆc = 1/c = 0)
(4)
By disregarding the risk of detection, if the firm reveals its true cost, it obtains: n−1
(r − 0 − (n − 1 − j)T ) n−1− j (1 − p) j p j + 1 j=0 (r − 0)(1 − p n ) p − pn = −T n(1 − p) 1− p
Eπc (ˆc = 0/c = 0) =
j
Cn−1
While announcing cˆ = 1, it will have: Eπc (ˆc = 1/c = 0) = T The incentive constraint (4) then imposes: r ≥T n
(5)
For a high cost firm: it is necessary that this firm does not have an incentive to announce cˆ = 0: Eπc (ˆc = 1/c = 1) ≥ Eπc (ˆc = 0/c = 1)
(6)
If it reveals its true cost, it obtains: Eπc (ˆc = 1/c = 1) = T
(7)
Let us suppose now that this firm manipulates the information on its cost. To do this, let us consider that the number of firms which add c = 0 is s ≤ n. Two cases are then possible. – If the firm is not selected as the cartel leader, an event which happens with probability (1 − 1/s), it receives a transfer of r −0−(n−s)T . s
16
BRISSET AND THOMAS
– If it is selected as leader (probability 1/s), it must pay the other members a total transfer ) of (s−1)(r −0−(n−s)T + (n − s)T and receives r from the auctioneer. s Let us compare its expected payoff when it announces cˆ = 0 with the payoff when it announces c = 1: 1 r − 0 − (n − s)T Eπc (ˆc = 0/c = 1) = 1 − s s 1 (s − 1)(r − 0 − (n − s)T ) + − − (n − s)T + (r − 1) s s r − 1 − (n − s)T = 0. – For high cost firms, the condition is as follows: Eπc (c = 1) ≥ λ A¯
(11)
It is also noticed that the low cost firm’s incentive constraint imposes that: Eπc (c = 0) ≥ Eπc (c = 1) Consequently, the ex ante non-denunciation constraint is summarized with (11), that is: T ≥
(λ + αβ) ¯ A 1−α
(12)
It is noticed that the satisfaction of the constraint (12) involves the satisfaction of the constraint (9). Studying these constraints, we can then state the following lemma: Lemma 4. Firms can perfectly collude if and only if constraints (5), (8) and (12) are simultaneously satisfied. In what follows, with a numerical example, we show that the presence of such a leniency program disturbs the formation of the collusive agreement whereas this is not the case in the absence of leniency programs or in the presence of the corporate European leniency program. 4.3.
Numerical application
We define our variables in order to ensure the formation of the collusive agreement in the absence of a leniency program, thus satisfying the condition (3): r (1 − αµ)(1 − p n ) − n(1 − p) p n−1 > A¯ αµ n(1 − p) In addition, we choose β in such a way that the leniency program encourages firms to report after the beginning of an investigation: 1−µ p − pn 1−µ 1 − pn 0 < β < min 1; µ − −T T; µ − r n(1 − p) 1− p A¯ A¯
18
BRISSET AND THOMAS
Lastly, we decide that the non-denunciation constraint is binding to restrict the weight of the constraint: (λ + αβ) ¯ T˜ = A 1−α We consider that r = 0.75, p = 0.25, α = 0.2, µ = 0.3, λ = 0.1, A¯ = 1, β = 0.1, n = 5. So, we have: T˜ = 0.15. The individual rationality constraint (8) is satisfied since we have: 0.515 > T˜ . On the other hand, the incentive constraint is binding with: nr = 0.15 = T˜ . Insofar as T˜ and nr do not depend on p, we can then state the following proposition: Proposition 5. Taking into account the numerical data above, the leniency program with rewards in the event of a report before an investigation prevents the formation of the collusive agreement when the number of firms is higher than 5. Proof: Above, we showed that the incentive constraint (5) is binding for n = 5. As r is decreasing in n, we deduce that for n > 5, the condition (5) is no longer satisfied. n Consequently, there is no collusive mechanism, which satisfies all constraints. Therefore, the formation of the collusive agreement is prevented. 5.
Conclusion
Our objective in this paper has been to suggest some economic incentives to improve the efficiency of leniency policy on cartel behavior. We have highlighted that the rules of the actual European program present some inefficiencies. In particular, providing full exemption from fine for any member who reports the cartel before any investigation is not an incentive. On the other hand, we have shown how the leniency program could become an effective tool for the prevention of anti-competitive behavior by rewarding firms that provide hard evidence of the cartel’s existence. The leniency policy can then create private incentives to play inefficient firms against efficient ones. Acknowledgments The authors would like to thank two anonymous referees for useful criticisms and remarks. Any errors and shortcomings are our own.
Notes 1. The principles seem more ambiguous than those of the European law. The article (loi N.R.E no. 2001-420, Titre II, ch. 2, Art. 73-3) does not specify anything as to the percentages of exemption granted according to the time when the report takes place (before or after the beginning of an investigation), which leaves a significant discretionary power to the competition authority. 2. The reserve price is the maximal price at which the buyer accepts market allocation. 3. This upper boundary can be derived from a firm’s limited liability.
LENIENCY PROGRAM
19
4. A knockout is a private auction held by the cartel to determine which firm gets the market and how much it pays the other members. 5. In order to maximise the collusive surplus, the cartel leader must submit the maximal possible price, i.e. the reserve price. 6. We can easily show that, in this situation, each firm uses a mixed strategy with distribution G n (b) = (1 − p( br )n−1 )/(1 − p) on support [b, r ] with G n (b) = 0.
References Ellis, C. & Wilson, W. (2001). “What Doesn’t Kill us Makes us Stronger: An Analysis of Corporate Leniency Policy.” Mimeo, University of Oregon. Journal Officiel de la R´epublique Fran¸caise, no. 113, 16 mai 2001, 7776, Imprimerie nationale, Paris. McAfee, R. P. & McMillan, J. (1992). “Bidding Rings,” American Economic Review. 82, 579–595. Motta, M. & Polo, M. (1999). “Leniency Programs and Cartel Prosecution.” Forthcoming in International Journal of Industrial Organization. Spagnolo, G. (2000a). Optimal Leniency Programs. Mimeo, Stockholm School of Economics. Spagnolo, G. (2000b). “Self-Defeating Antitrust Laws: How Leniency Programs Solve Bertrand’s Paradox and Enforce Collusion in Auctions.” Mimeo, Stockholm School of Economics. U.S Department of Justice (1993). “Corporate Leniency Policy.” Antitrust Division. Available at http//www.usdoj. gov/atr/public/guidelines/lencorp.ht
