Privatization as an agency problem: Auctions versus private negotiations

Journal of Banking & Finance 31 (2007) 2730–2750 www.elsevier.com/locate/jbf

Privatization as an agency problem: Auctions versus private negotiations Zsuzsanna Fluck a, Kose John b, S. Abraham Ravid a

c,*

Eli Broad College of Business, Michigan State University, 319 Eppley Center, United States b Stern School of Business, New York University, 44 W. 4th St., NY 10012, United States c Rutgers Business School, 180 University Ave., Newark, NJ 07102, United States Received 18 April 2006; accepted 1 December 2006 Available online 31 January 2007

Abstract This paper investigates the design of privatization mechanisms in emerging market economies characterized by political constraints that limit the set of viable privatization options. Our objective is to explain the striking diversity of mechanisms observed in practice and the frequent use of an apparently sub-optimal privatization mechanism: private negotiations. We develop a simple model in which privatization is to be carried out by a government agent, who plays favorites among bidders but is potentially disciplined by losing his private benefits of staying in office. If the political environment is such that the privatization agent himself aims at raising the fair value for the company, then privatization auctions and private negotiations are equally successful in raising public revenues. If, however, political considerations distort the agent’s incentives, it may be that a seemingly transparent auction will raise less revenue, than opaque private negotiations. We also show that information disclosure laws may have negative welfare implications: they may help the privatization agent to collude with some of the bidders to the disadvantage of non-colluding bidders.  2007 Elsevier B.V. All rights reserved. JEL classification: L33; G28; G34; K22; C72 Keywords: Privatization; Agency; Political constraints; Auctions; Private negotiations

*

Corresponding author. Tel.: +1 973 353 5540. E-mail address: [email protected] (S.A. Ravid).

0378-4266/$ - see front matter  2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jbankfin.2006.12.008

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2731

1. Introduction Privatization of state-owned enterprises has grown into an economic tidal wave during the 1990s. This has not left academics idle, and the literature on privatizations has mushroomed. Much less attention has been paid to privatization methods. Perhaps the most important puzzle, facing researchers in this literature is the striking diversity of privatization mechanisms. Frydman and Rapaczynski (1991) document the use of numerous and arguably sub-optimal mechanisms, particularly the overwhelming choice of private negotiations in Eastern Europe. Similarly, empirical papers, such as Dewenter and Malatesta (1997) or Steven et al. (1999) conclude that governments tend to privatize partially, under-price offerings even more than in the private market, sell companies at a pre-set price, or have special deals for different people, rather than just auction off the entire firm to the highest bidders. In their analysis of the Polish, Czech and Russian privatization processes, Boycko et al. (1994) argue that privatization mechanisms in these countries were selected on the basis of political constraints and were simplified to meet political feasibility even at the expense of economic principles. The importance of political considerations is also emphasized in Baldwin and Bhattacharyya’s (1991) study of the privatization of Conrail by the US government. They report that after a successful challenge to an auction, a public equity offering of Conrail netted the government $700 million more than the winning bid in that auction. In the context of bank privatizations, Bonin et al. (2005) show that the timing and choice of privatization methodology matters in transition economies. The critical role of political considerations in bank privatizations is also emphasized by Clarke et al. (2005), Megginson (2005), as well as Berger et al. (2005), Beck et al. (2005), Haber (2005) and Bonaccorsi di Patti and Hardy (2005) among others (see also Fluck et al., 1995). Our paper proposes a unique theory that explains how political constraints, which are ever present in privatizations, may shape different privatization mechanisms. Our main result thus explains the somewhat intriguing view expressed in the bank privatization literature that ‘‘share offerings produce lower performance gains than direct sales to concentrated strategic investors in weak institutional environments’’ (Clarke et al. (2005, p. 1925)). There is not much theoretical literature along these lines.1 In an article that is close in spirit, but different in focus, Perrotti (1995) suggests that, in the presence of moral hazard, a government may use partial privatization to signal that it is not going to expropriate the new owners. If this is not feasible, under-pricing may be optimal. Biais and Perotti (2002) also develop a model of different voting constituencies that leads to ‘‘Machiavellian Privatization’’. Political constraints, in a different sense, matter there. Our paper suggests that in the presence of agency problems within the privatization process, private negotiations may actually serve a purpose, even if the government can in fact auction the privatized firm to bidders. We also consider the effects of disclosure within a potentially corrupt system. 1 There is a related literature on the design of markets and IPO mechanisms. See for example, Berkovitch and Israel (1999), Boot and Thakor (1994), Diamond (1996), Ravid and Spiegel (1997), Titman and Subrahmanyam (1997). To exemplify the qualitative difference between IPO’s and privatizations, one can look at the ‘‘new popular notion’’ (as defined by a commentary by Margaret Coker in Business Week, December 13, 1999) in Russia which attempts to de-privatize companies that were sold to ‘‘undesirable’’, generally foreign, investors.

2732

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

The agent we have in mind is either the manager of a state-owned firm who is appointed to privatize the firm,2 or a bureaucrat from the state privatization agency. In either case the agent is in a position to potentially monopolize information about the company in question.3 Business and financial information is typically scarce in emerging market economies. Financial markets are nonexistent or at infancy. Many of the firms for sale have never been audited before and follow ancient accounting practices.4 Hence, a crucial piece of information may not be publicly available, and the official may have strong incentives to help his friends in this respect. In our model, the agent provides an informational edge to one of the bidders. This is obviously stylized, however, it is important to realize that our results go through even if what happens in reality is a meeting in a cafe´ where some important tip is passed on to a favorite bidder.5 The other important question is the objective function of the government or ‘‘the people’’. Clearly, selling companies below value can create scandals, and hence the government would like to raise adequate revenues. On the other hand, in emerging market economies there is public concern if entrepreneurs end up losing money in the privatization process.6 These goals translate into an incentive scheme that penalizes the agent for a positive value–price differential ex-post while not rewarding him for raising more money than the company is worth. Such incentives are not consistent with profit-sharing contracts that create incentives to overcharge for bad companies. Also, conflict-of-interest regulations naturally exclude explicit profit-sharing contracts between the Treasury and the bureaucrat in charge of privatization. Similar to Grossman and Hart (1988) or Harris and Raviv (1988) in a different context, we compare the outcomes of commonly used mechanisms. Auctions and private negotiations are the most widely employed privatization mechanisms in emerging market economies. The typical auctions are first-price sealed-bid auctions (‘‘The Revenge of the Nerds’’, 1994). Our paper thus investigates how much revenue privatization auctions and private negotiations yield in emerging market economies by explicitly modeling the political constraints discussed above. This paper provides new insights into collusion between a bidder and a seller’s agent. While collusion among bidders received a lot of attention in the literature (see McAfee and McMillan (1987, p. 724) or Rothkopf and Harstad (1994) for excellent surveys of this literature or Porter and Zona (1993) for a recent study of this type of collusion), collusion between a bidder and a seller’s representative is largely unexplored. 2

A commonly used privatization mechanism for small and medium-sized companies in Hungary is selfprivatization. The government appoints the manager of the state owned firm to privatize the company (Major, 1994). 3 For a different view of the advantage of developed markets in diffusing information as compared to emerging markets see Berkovitch and Israel (1999). 4 As J. Mark Mobius, portfolio manager of the Templeton Russia fund explains: ‘‘We invest in the face of very iffy data. The amount of information available leaves a lot to be desired’’. (Best Single Country Fund: Templeton Russia, Mutual Funds, March 1997, p. 48). 5 Besides information regarding the value of the company in question, the official may control information about the sequencing of future privatizations, future changes in the tax code, or other decisions at the planning stage that can materially affect the value of the company for the bidders. This information, though incremental, is very valuable and likely to be costly to produce for a market research firm. 6 To prevent severe losses by entrepreneurs, governments in these economies frequently grant bidders the right to sell the privatized company back to the government within a period of three to five years at the price it was purchased for, or, alternatively, offer compensation packages for unexpected losses incurred by entrepreneurs during the first few years of operation (Major, 1994).

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2733

We recognize that collusion among bidders may affect privatization auctions as well, however, to keep our model tractable we assume away this type of collusion here.7 We find that in addition to the negotiating skills of the agent, it is the degree of political constraints that determines which mechanism is more successful in raising funds. If the agent himself aims at raising fair value for the company, then privatization auctions and private negotiations are equally successful in raising public revenues. In this case the official will disclose all information publicly and the winning bid will approach value. If, however, political constraints distort the agent’s incentives, then one mechanism outperforms the other. In particular, if the distortion is moderate, then, surprisingly, private negotiations raise more value for a successful enterprise than privatization auctions. This is true even though the auction is fair and transparent, save for this little tip in the cafe´. The intuition is that even though a bureaucrat in charge of the privatization process may compromise both auctions and private negotiations, he or she has much more direct control over the outcome of private negotiations than over the outcome of auctions. Whereas in private negotiations the privatization agent can potentially attain his most preferred outcome, the winning bid in an auction may very well exceed or fall short of the agent’s target. If the agent’s target is sufficiently high, the expected winning bid in the privatization auction is likely to fall short of the agent’s target. Interestingly, our results hold regardless of whether information disclosure is mandated for the bidders in the economy. The rest of the paper is organized as follows. In Section 2 we describe the assumptions of the model. In Section 3 we discuss the benchmark case when the privatization agent voluntarily reveals information to all bidders. In Sections 4 and 5 we investigate the optimality of auctions versus private negotiations when disclosure rules are in effect and when they are not in effect. In Section 6 we discuss the regulatory implications of our model. We close the paper with some concluding remarks and recommendations for the design of privatization procedures. 2. The model 2.1. The basic setup We consider a privatization target that can be either a gem or a lemon with equal probabilities. The gem is valued at v, the lemon is valued at v. Two mechanisms are compared – a first-price sealed-bid auction and private negotiations. The comparison is in terms of the ability to raise a fair price for the company. There are two bidders. The bidders are riskneutral profit-maximizers. Each bidder is equally capable of operating the company. The bidders know the agent’s preferences and the distribution of company values. They may learn the value of the company they bid for if the agent reveals information to both or one of them. Any bidder may decide to purchase information from a research company at cost c. The privatization program is to be carried out by a government agent who derives utility from both pecuniary benefits, l, and private benefits from staying in office, W. The pecuniary benefits are the side-payments he receives if he colludes with a bidder. The non-pecuniary benefits, W capture the value of perks, power, and the agent’s ability to generate side-payments in the future. We normalize the agent’s salary to zero. The agent 7

For a real-life perspective on the subject see Cramton’s (1997) on FCC spectrum auctions.

2734

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

maximizes the sum of his monetary payoff and expected private benefits. Our specification of the agent’s preferences originates in the corporate control literature. Similar objective functions in the context of asset sale include Grossman and Hart (1988), Harris and Raviv (1988), Shleifer and Vishny (1989), Burkart et al. (1997). The agent’s monetary payoff, l, depends on the value–price differential, v–p, and on his bargaining power (negotiating skills), a. The higher the value–price differential the more the colluding parties can split as side-payment. Moreover, the agent can extract more from the bidder the better is his bargaining position. Note that the extreme case of no rent sharing rules out collusion. If the bidder can capture all the rents, the agent’s objective reduces to maximizing his expected private benefits. In this case the agent has no incentive to collude. If all rents accrue to the agent, then the bidders have no incentive to collude with the agent. As long as the official expects favors from the colluding bidder when he is dismissed (e.g. a high level appointment in the bidder’s firm), he is willing to split the rents with the bidder.8 Similarly, as long as the colluding bidder expects future favors from the privatization agent if the agent stays in office (e.g. information about companies to be privatized later), he is willing to share his rents with the agent. The agent’s expected private benefits depend on the likelihood that he remains in office. In the corporate finance context it is commonly assumed that the agent loses his private benefits of control with a certain probability. This would happen when he is dismissed due to decisions by the corporate board, proxy fights, and takeovers. Such models were developed by, among others, Harris and Raviv (1988), Berkovitch and Israel (1999), Stulz (1988), and Fluck (1999). In our context of privatization, the likelihood that an agent stays in office depends on his performance and on whether or not public attention is focused on privatization. If the public focuses on privatization issues and a company is sold at a price far below value then the agent is dismissed. Hence we model the probability, p, that the agent is dismissed as an increasing function of the value–price differential, v–p. That is, the larger is the discrepancy between the value of the company and the price the winner paid for it, the more likely it is that public attention should turn to privatization issues and the agent will be dismissed. The value of p is positive and increasing for v > p and 0 for v 6 p.9 Thus, the agent’s objective function takes the following form: maxð1  pðv  pÞÞ  W þ lððv  pÞ; aÞ;

ð1Þ

where p characterizes the economy, W the privatization agent and a the bargaining process. Notice that the privatization agent faces a trade-off when deciding how to price the firm: on one hand, he would like a low price to capture some of the buyer’s profits. On the other hand, he would like a high price to increase his chances of keeping his private benefits. We compare two mechanisms, a first-price sealed-bid (common-value) auction and private negotiations, with a minimum bid set at v in either case.10 Since one of our aims is to 8

In real life, bribes do not need to be in cash. Promises of future jobs or benefits when the official retires can be sufficient and commonly happen even in more law abiding countries – this may be sufficient to motivate the official to play favorites among bidders, and our results will qualitatively go through. 9 Note that the probability can be 100% as well – in that sense our setting is more general and realistic. 10 We assume that the number of bidders remains the same regardless of the sale mechanism. Clearly, if either mechanism will attract more bidders, it can skew the result in favor of that mechanism. Papers such as Bulow and Klemperer (1996) suggest that auctions can have a better outcome if more bidders appear. However, it can be that private negotiations will attract more bidders if they hope for a better outcome.

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2735

derive policy implications for the privatization process, we want to model auctions and private negotiations, in a manner that closely approximates the actual bidding process. In privatization auctions bidding is allowed only in currency units or multiples of currency units.11 Thus, in our model bidding is conducted in multiples of currency units with the smallest allowable bidding unit set at e. In other words, one can bid v, v + e, v + 2e, etc. Unlike open auctions that can always attract additional bidders (Hendricks et al., 1993; Bulow and Klemperer, 1996), potential bidders in privatization auctions of emerging market economies come from a tight group of insiders, competitor-firms, wealthy individuals, or investor groups. They are screened in advance to assure that they can raise the necessary funds to finance the deal. Some of these insiders have close ties with those in charge of the privatization process and tend to be frequent participants at other privatization auctions. Thus, we model privatization auctions and private negotiations with a fixed number of bidders, one of whom may have a continuing relationship with the privatization agent. We solve for a trembling hand perfect Nash equilibrium whenever several equilibria are possible. The timing of the auction is as follows. The privatization agent decides whether to reveal information to one or both parties or to none at all. We assume that if the agent decides to collude with one of the bidders, he will not cheat the colluding bidder. The colluding parties privately agree on the side-payment conditional on the value–price differential. If the agent’s bargaining power is a, then, for the good company, the bidder pays the agent a fraction of the expected value–price differential, where expectation is taken over the set of possible winning bids in the auction.12 As noted, at a cost of c anyone can find out exactly what the target company is worth. Our assumption that both the official and the research company know precisely the value of the firm, substantially simplifies the analysis without altering the nature of our results. With appropriate changes in the parameter restrictions, our results will carry through even if (1) the official’s information is noisy; (2) the information of the research company has lower precision than the privatization official’s; or (3) the information supplied by the research company is complementary to that of the privatization official. In the case of private negotiations the agent decides whether to reveal information to one or both parties or to none at all. If he decides to share information publicly, then he can only accept bids that approximate value. If he decides to reveal information privately, then he approaches one of the bidders to sell the company to the bidder in exchange of sharing the rents. Given a (the agent’s bargaining power) the parties negotiate a payment v + a(v  v) from the bidder to the agent. That is, each party’s share of the rents is proportional to his bargaining power. The payment for the good company is thus v + a(v  v), or the closest acceptable bid, and it is v for the bad company. Since the bidder is indifferent as to how much of his payment become public revenue and how much will

11

Much of the auction literature analyzes continuous auctions. However all real-life auctions are, almost by definition, discrete. In fact, in the largest auction to date in this country, the FCC spectrum auction, the FCC explicitly set discrete bid increments (the design also included discrete time intervals between bids). As it turned out, there were interesting behavioral consequences to the discrete bid intervals, as is the case in our paper (see Cramton, 1997). 12 Our result would be qualitatively the same, if the side-payment were an a fraction of the expected value–price differential with expectation taken over the set of possible winning bids in the auction conditional on the colluding bidder winning, or if the side-payments were to be paid ex-post.

2736

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

end up as a bribe, he lets the agent decide the price that will be publicly announced, and lets him pocket the rest of the payment. Naturally, the agent chooses the price that maximizes his objective function (1) subject to the constraints that l + p = v + a(v  v), l P 0, p P 0. Once the winner starts to operate the company, the value of the company is publicly revealed. If public attention is focused on privatization, and if there is a discrepancy between the value of the company and the price paid for it, then the official is dismissed. If the public attention is focused on issues other than privatization then the agent stays in office. It has been suggested that information disclosure laws generate more revenues by making the privatization process more transparent. Thus, we consider two scenarios in our model. In the first case, (Section 4) disclosure laws are in effect; whereas in the second one (Section 5) information disclosure is not mandated. When no disclosure laws are in effect, then the purchase of information remains private. When disclosure laws are in effect, parties who purchase information are expected to disclose their information. Naturally, disclosure rules do not bind the colluding bidder. Hence he appears to be uninformed from the point of view of the legal system. There is one additional assumption we require regarding the rate of detection for collusion. When bribery is a more serious offense than failing to reveal information, then, for collusion to be incentive compatible for the bidder, it must be the case that bribery is more difficult to discover. This assumption seems reasonable. As we noted, the collusion we have in mind is a casual and hard to detect conversation, that both parties wish to keep secret. This contrasts with business transactions such as sale of information by a research company that leave a paper trail. 3. The case of voluntary information revelation We first discuss the benchmark case in which the agent greatly values staying in one refuses to be bribed and voluntarily reveals information to all bidders. This s a scenario in which political constraints and sale mechanisms do not matter. Formally, W  ðpðv  pÞ  pðeÞÞ P v  p  e

ð2Þ

for every p P v. For every p, the right hand side shows the maximum bribe the official gets when he raises p and accepts a bribe rather than raising v  e and refusing side-payment. The left-hand side shows the difference between the expected private benefits of these strategies. Note that the condition is only sufficient – agents may inform both bidders even if the condition does not hold. Proposition 1. When the agent informs both bidders then the subsequent auction p ¼ v  e for the good company, and p = v for the bad company in any perfect equilibrium. The intuition is straightforward. If both bidders know that the company is a lemon, then neither bids more than v in equilibrium. If, however, they find out that the company is good, then they will not bid below v  e since they may lose the auction otherwise. If either bids v, then she may win or tie with the other bidder. In either case zero profit is

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2737

made. Either bidder can do at least as well by lowering her bid to v  e. Bidding v  e for the company yields an expected profit e/2 for both bidders in equilibrium.13 When there are no political constraints, then for the privatization agent maximizing utility is equivalent to minimizing the probability of dismissal. Proposition 2. Whenever (2) holds, the privatization agent voluntarily reveals information to both bidders in any perfect equilibrium. When (2) holds, then publicly revealing information maximizes the utility of the privatization agent (Proposition 1). Obviously, no strategy can improve upon this. As we shall see later, other strategies make the agent strictly worse off. Disclosure rules do not matter when (2) holds. Since both parties are receiving information from the agent, nobody purchases information, so there is nothing to disclose. Finally, whenever (2) holds, then an auction yields the same revenue as private negotiations. The agent who publicly reveals information extracts nearly full value: the bidders will be indifferent between buying at value and walking away. Proposition 3. Whenever (2) holds, the privatization agent raises v  e for the good company and v for the bad company through private negotiations. The following sections focus on approachable agents. An agent is called approachable if for a high enough bribe he is willing to compromise public revenues. When the agent is approachable, disclosure rules make a difference. 4. Privatization when disclosure laws are in effect We model disclosure rules in the spirit of Grossman and Hart (1980), that is, disclosure rules require that any information that a party has about the value of the company must be revealed to everybody. The disclosure we model is thus very much in the spirit of what insiders at publicly held companies are required to do when they take actions that substantively affect shareholder value14 (see Fig. 1). Disclosure rules have two implications in our model. They force the party who purchased information to reveal this information and thereby reveal their identity. Secret information transactions on the other hand, by nature, are not known to anyone but the parties involved. Since disclosure rules have no impact on the colluding parties who do not comply with them, they enable the colluding parties to fully realize their cost advantage over the non-colluding party. As a consequence, it does not pay for the non-colluding party to purchase information unless the cost of acquiring information is negligible. The outcome is an auction with asymmetrically informed bidders.15 13 One could ask the following question: if the official discloses the true value, why do we need an auction at all. As is shown in Propositions 1 and 3, in such case both auctions and private negotiations would extract nearly full value. 14 See for example, Bonin et al. (2005) for a discussion of the role of information disclosure in privatizations in transition economies. 15 Our paper is related to Wilson (1967), Milgrom and Weber (1982), Engelbrecht-Wiggans et al. (1983), Hendricks et al. (1993, 1994) on auctions of mineral rights with asymmetrically informed bidders.

2738

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750 The official reveals information to one, both or none of the bidders.

0

The bidders may (publicly) purchase information.

1

2

Auction or private negotiation takes place.

Ownership is transferred.

3

4

Fig. 1. The timing of the model when disclosure rules are in effect.

Proposition 4. When disclosure rules are in effect and when the privatization agent discloses information to one bidder only, then one of the following applies: (1) If c 6 e/4, then the non-colluding party always purchases information and the subsequent sealed-bid auction facilitates the transfer of the good company’s ownership at price p ¼ v  e. (2) If c > e/4, then no party purchases information. There is a unique perfect Nash equilibrium in mixed strategies in the auction in which the expected winning bid for the good company is significantly below the unconditional expected value. Proof. The proof is shown in an Appendix.

h

This proposition establishes our most interesting outcome – namely, that in corrupt regimes, seemingly open and transparent auctions are likely to bring disappointing outcomes, as experienced in bank privatizations (see Megginson, 2005; Clarke et al., 2005). When disclosure laws are in effect then, in equilibrium, the identity of the party who purchased information is disclosed and the identity of the colluding party is revealed in equilibrium since all parties have rational expectations. Therefore, as long as the information structure is symmetric and the cost of purchasing information is non-negligible, colluding parties can realize their cost advantage in obtaining information. Notice that if the information purchased by the uninformed bidder is of lower precision than the information obtained from the privatization official, the threshold for information purchase will be even lower than e/4. There is no pure strategy Nash equilibrium in the sealed-bid auction. Since the uninformed would never want to place a bid higher than unconditional expected value any bid above that value is a winning bid. 16 Such a bid is not a best response for the colluding bidder unless the uninformed bids unconditional expected value or if it is not an accept16 The uninformed bidder would be less concerned of overbidding if the mechanism were a second-price auction. In privatizations second-price auctions have rarely been used. In those instances when it was tried (for example in New Zealand), the public was outraged to learn how small the payment is going to be relative to the winning bid (see ‘‘The Revenge of the Nerds’’, 1994). Since it is difficult to control such public outcry, second-price auctions are generally considered politically infeasible.

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2739

able bid, the highest allowable bid below. In case the colluding bidder does bid this high, it is not a best response for the uninformed to make such a bid. However, if the informed bids less in the good state, then it is a best response for the uninformed to place a higher bid but never higher than unconditional expected value independently of the state. Staying out is not Nash equilibrium strategy either. There is a unique perfect Nash equilibrium in mixed strategies, however. In this equilibrium no bid exceeds the unconditional expected value with positive probability. The colluding party always bids v when the company is of low quality. When the company is of high quality, the informed mixes between his sure winning bid (the one just below the unconditional expected value) that may, however, be unnecessarily high, and lower bids that are potentially more profitable winning bids. The uninformed bids either cautiously or aggressively, each with high probability so as to minimize his potential realized losses and his opportunity losses. Collusion can be alleviated by setting information costs low or bidding increments high so that c 6 e/4. When this is the case, then scenario 1 arises. In this case collusion breaks down because the party that does not purchase information becomes suspect as a party to collusion, since a non-colluding party will always purchase information. Therefore, under condition c 6 e/4 collusion will become unprofitable and unsustainable. Obviously, private negotiations are not affected by disclosure rules. We illustrate the outcome of the auction and of private negotiations with a simple example for the case when c > e/4. Example 1. Let v = $1 million and v ¼ $6 million. The increment between subsequent bids is $1 million. The unconditional expected value of the company is $3.5 million. The government agent’s private benefit from staying in office is $4 million. The cost of information purchase is $750,000. The political constraints specified by the probability of dismissal are pðv  pÞ ¼ ðv  p  eÞ  0:08 if v  p 6 3 and pðv  pÞ ¼ ðv  p  e  0:2) otherwise. We present below the essence of the solution, a more detailed proof and description is available from the authors upon request. Private negotiations: The colluding bidder makes a payment v þ aðv  vÞ to the agent for the good company. The agent sets l and p to maximize (1) subject to l þ p ¼ v þ aðv  vÞ, l P 0, p P 0. The bidder’s profits are ð1  aÞðv  vÞ. When the value of the company is revealed, the agent faces a dismissal with probability pðv  pÞ. For example, for a = 2/3, the agent will negotiate a price of $4 million, and pocket a bribe of 1/3 of a million. His utility will be 4.01 million that is higher than the 4 million he can achieve by revealing information publicly. The bidder receives a profit of 5/3 of a million. The agent faces a dismissal with probability 0.08. The auction: After iteratively eliminating weakly dominated strategies, we find that the non-colluding party will never purchase information and will not bid more than expected value, $3.5 million. The colluding party will not bid above $3.5 million with positive probability for the good company and will not bid $1 million with positive probability for the bad company. In equilibrium, the colluding party bids 3 with probability 2/3 and 2 with probability 1/3 when the company is good. Furthermore, if the non-colluding party bids 1 with probability 3/5, and 3 with probability 2/5, then in the good state the colluding party is indifferent between bidding 2 and 3, prefers either of these bids to 4, and might as well bid 2 with probability 1/3 and 3 with probability 2/3. Similarly, if the colluding party bids 2 with probability 1/3 and 3 with probability 2/3 for the good company and 1 for the bad

2740

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

company, then the non-colluding party is indifferent between bidding 1 and 3 and might as well bid 1 with probability 3/5 and 3 with probability 2/5. The colluding party wins the good company with probability 11/15. The expected price for the good company is only $2.8 million, about 20% below the unconditional expected value and less than half of what the company is worth.17 This under-pricing happens even when the agent himself would prefer a price of $4 million and is able to negotiate this price in the absence of the auction. This would happen, for example, when a = 2/3. However, when the agent’s only option is an auction he prefers the expected winning bid of $2.8 million to the $5 million he can raise by foregoing his bribe and revealing information to both bidders. His utility from the latter is $4 million. When he reveals information to one bidder only, he raises $2 million with probability 1/5 and $3 million with probability 4/5 for the good company. He receives a bribe of 1/5 · a · $4 million + 4/5 · a · $3 million = 16/5 · a. Depending on a, he prefers revealing information to one bidder only. This happens for example when a = 2/3. In this case, the agent prefers revealing information to one bidder only, since by doing so, he expects the equivalent of 5.14 million in utility terms. So far we have assumed that when auction mechanism is used to privatize the company the official always reveals information to at least one party. Next we establish that the official will indeed, always designate at least one party as informed in equilibrium. As Proposition 5 below shows, not revealing information is a dominated strategy. Depending on his preferences, the agent either informs one bidder (see Example 1) or both (see the case of voluntary information revelation). Proposition 5. When disclosure rules are in effect and when a sealed-bid auction is used to sell the company, then in equilibrium the official will reveal information to at least one bidder. Proof. The proof is shown in Appendix. h Our main result: in an ex ante comparison private negotiations dominate auctions for a wide range of preferences and bargaining skills. Proposition 6. Suppose that c > e/4 and the official is somewhat disciplined by the threat of dismissal (i.e. p*, the price he would negotiate if he colludes exceeds the expected winning bid in the auction with asymmetrically informed bidders). Then, private negotiations raise more revenues for good companies, auctions raise more revenues for bad companies if either of the following holds: (a) Regardless of the mechanism used, the agent informs only one bidder; or (b) the agent reveals information to one bidder when conducting an auction and makes information public when conducting private negotiations. Proof. The proof is shown in Appendix. h Proposition 6 presents a striking result: even if the official is so corrupt that half of the company’s value disappears in the pocket of the colluding parties, private negotiations can

17

The expected price the colluding party pays for the good company when he wins is $2.7 million, so the colluding party pays less than the non-colluding party when winning the good company.

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2741

raise more value than the asymmetric information auction. This is illustrated in Example 1. Such logic may explain why one observes private negotiations in countries where corruption and collusion are widespread. If the official will distorts both outcomes, then a request by external monitors or the public to conduct an open auction may backfire, whereas private negotiations may increase social welfare and lead to better prices. Another comparison relevant for political considerations, is the ex-post comparison of the realized winning bid of the auction with the price attained through private negotiations. Corollary 6a. Suppose c > e/4 and p* > v + e. Then, when the agent reveals information to only one bidder there is at least one price in the auction that private negotiations can improve upon. Since any bid below expected value is sufficient to win the good company with positive probability in the auction with asymmetrically informed bidders, this corollary demonstrates that there is always a price in the auction private negotiations can improve upon ex-post, provided that p*, the price that maximizes the utility of the privatization official exceeds v. Since low bids are used with high probability (see Example 1) the domination can be quite strong. The reverse is not true: for a range of the agent’s preferences, all prices in the auction are dominated by private negotiations. The situation is somewhat different when no disclosure rule is in effect. We presented sufficient conditions for private negotiations to dominate auctions in nontransparent regimes. These conditions seem to be consistent with empirical findings in the bank privatization literature (see Clarke et al., 2005; Haber, 2005) as well as evidence cited in Megginson (2005), which culminates with his recommendation: governments may need to emphasize asset sales to foreign owners (Megginson, 2005, p. 1962). Our next section focuses on a setting in which the non-colluding party is not a transparent entity, and thus may have a ‘‘better shot’’ against a corrupt opponent. 5. Privatization when no disclosure laws are in effect When information disclosure is not mandated, it is potentially advantageous for the non-colluding party to purchase information (see Fig. 2). In this case the information structure is asymmetric: the colluding parties do not know whether they face an uninformed bidder in the auction. The lack of disclosure protects the The official reveals information to one both or none of the bidders.

0

The bidders may (privately) purchase information.

1

2

Auction or private negotiation takes place.

3

Ownership is transferred.

4

Fig. 2. The timing of the model when no disclosure rules are in effect.

2742

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

non-colluding bidder so that he can benefit from purchasing information with some probability. Having an informational advantage, a non-colluding bidder can compete away part of the rent the colluding parties could otherwise realize from their cost advantage. As a result more revenues can be raised from privatization when information disclosure is not mandated. Let EV* stand for the unconditional expected value of the target, if it is an allowable bid, and for the closest allowable bid from above if it is not an allowable bid. Then, Proposition 7. When no disclosure rule is in effect and when the privatization agent informs one bidder only, then, depending on the cost of information, c, one of the following applies: (1) If c 6 e/4, then the non-colluding party always purchases information and the subsequent auction facilitates the transfer of ownership for the good company at a price p ¼ v  e.     (2) If c > max vEV ; vEV2 e holds, then no party purchases information. There is unique 4 perfect Nash equilibrium in mixed strategies in the auction such that the expected winning bid for the good is significantly below the unconditional expected value.  company   vEV e ; (3) If e=4 < c < max vEV holds, then the non-colluding party will purchase 2 4 information with some probability. There is a unique perfect Nash equilibrium in the subsequent auction in which the non-colluding party plays mixed strategies conditional on whether or not he has purchased information. The expected price at which a good company is sold is higher when disclosure rules are not in effect than when they are in effect. Proof. The proof is shown in Appendix. h Note that the expected price at which the good company is sold is the same no matter whether disclosure rules are in effect. In scenario 1 the party that does not purchase information becomes suspect as a party to collusion, since a non-colluding party will always purchase information. Therefore, under condition c 6 e/4 collusion becomes unprofitable and unsustainable. In scenario 3 the expected price at which a good company is sold is higher when disclosure rules are not in effect because the non-colluding party will purchase information with some probability when his purchase cannot be detected by the colluding parties. When he learns that the company is good, the non-colluding party is willing to bid higher than when he is ignorant. He is willing to buy information as long as the cost of purchasing information can be recovered. After iteratively eliminating dominated strategies we find: (1) When uninformed, the non-colluding party (a) will not bid above expected value; (b) will bid v with positive probability for the good company. (2) No informed party will bid v with positive probability for the good company. (3) The colluding party will bid more aggressively than the non-colluding party. We illustrate the auction and of private negotiations in Example 2 below for the case when c > e/4. Example 2. Let v = $1 million and v ¼ $6 million and let the minimum increment be $1 million. Assume that the cost of information purchase is $750,000. The unconditional expected value of the company is $3.5 million. The political constraints and the outcome of the private negotiations are as in Example 1.

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2743

In equilibrium, the non-colluding party purchases information with probability 4/7 and if he learns the company is good, he bids either $3 million or $4 million with probability half each. If he learns the company is bad, he bids $1 million. With probability 3/7 he remains ignorant and bids no higher than $1 million. For the good company the colluding party bids $4 million half of the times, a bid high enough to win against an equally informed bidder and $2 million half of the times to take advantage of bidding against an uniformed bidder. For the bad company, the colluding bidder bids $1 million. Note that bidders bid more aggressively under no disclosure than under full disclosure. They both bid $4 million with positive probability under no disclosure, a bid neither placed with positive probability in the case of full information disclosure. Were the cost of information lower, they would bid even more aggressively. In contrast, if c were to exceed $2 million, the non-colluding party would never purchase information and would bid the same as in Example 1. The non-colluding party has no incentive to deviate from his equilibrium strategy even upon learning that the company is good. He is indifferent between bidding $3 million and $4 million for the good company both ex ante and ex-post provided the colluding party follows his equilibrium strategy. The non-colluding party would profit less by bidding higher than $4 or lower than $3 million after getting good information. Notice also that the non-colluding bidder does not bid the same as the colluding bidder even when they both know the value of the company. The reason is that even though the two bidders have the same information about the value of the company then, the non-colluding party is better informed than the colluding party: he knows when he is informed. The expected price for the good company in the auction is $3.4 million, about 20% higher than the $2.8 million price the auction raises under mandated disclosure. It still falls short of the $4 million the agent would raise through private negotiations and it is still below the unconditional mean. Again, when the agent’s only option is the auction he prefers the expected winning bid of $3.4 million to the $5 million he could raise by foregoing his bribe and revealing information publicly. His utility from the latter is $4 million. When he colludes, he raises $2 million with probability 3/14, $3 million with 1/7 and $4 million with 9/14 for the good company. His bribe is 3/14 · a · $4 + 1/7 · a · 3 million + 9/14 · a · 2 million = 18/7a. When for example a = 2/3, he prefers revealing information to one bidder only since by doing so, he expects 4.9 million in utility terms. So far we assumed that the agent reveals information to at least one bidder. The next proposition establishes that the privatization agent will disclose information to at least one party even when there is imperfect monitoring of information purchases. Depending on his preferences, he either colludes with one of the bidders (see Example 2) or reveals his information publicly (see the case of the voluntary information revelation).

Proposition 8. When no disclosure rules are in effect then the privatization agent always reveals information to at least one bidder. Proof. The proof is shown in Appendix.

h

Having established the equilibrium bids in the auction, it is straightforward to see that Corollary 6a would also hold for the case when information disclosure is not mandated. That is,

2744

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

Corollary 8a. When c > e/4 and when p*, the price that maximizes the utility of the privatization official, given a, exceeds v, then no matter what the preferences of the privatization agent are, there is at least one price in the auction which private negotiations can improve upon. It is important to note that the probabilities implied by the two disclosure scenarios are quite different. When information disclosure is not mandated, auctions produce higher revenues (see Proposition 7). Depending on the agent’s preferences, private negotiations may still raise more than auctions but for a smaller range of parameter values. Paradoxically, disclosure rules give rise to lower public revenues and higher bribes when market and political frictions are prevalent. 6. Implications and additional policy considerations There are several policy implications one can draw from our model. First, we have seen in scenario 1 in Propositions 4 and 7 that auctions designed with high bidding increments may give rise to higher public revenues and lower bribes. These results suggest that discrete intervals may play an important role in auctions. Intuitively, the higher the bidding increment, the more costly a wrong bid, and the more useful is information. Similarly, low cost information will reduce the unfair advantage of the colluding bidder and will make higher prices more likely in auctions. In our model if the cost of information is less than a quarter of the bidding increment, then the official cannot manipulate the auction by strategically releasing information to a favorite bidder. Furthermore, in this case no collusion between the official and a bidder is sustainable. In such an event, uninformed bidders will always purchase information, therefore, the fact that a bidder does not acquire information signals inside information and makes it easy to detect collusion or corruption. Hence, by setting the bidding increment appropriately high or by subsidizing market research, a government can significantly reduce corruption in the privatization process. The second, perhaps most striking implication of our model, is that the choice of privatization mechanisms should depend on the availability and dissemination of information and the severity of political constraints. Our model suggests that in economies or in industries where information is widely available, valuation is straightforward and political constraints do not matter, auctions and private negotiations are equally successful in raising revenues. In contrast, in economies or industries where information is scarce, valuation is highly uncertain and political constraints are present, private negotiations may dominate auctions. The crux of the issue is the recognition that a corrupt agent will be corrupt whether he is negotiating in private or organizing an auction. Whereas an official can both compromise auctions and private negotiations he or she has much less direct control of the outcome of an auction than that of private negotiations. This idea is consistent with the recommendations of both Clarke et al. (2005) and Megginson (2005) which amount to a suggestion of selling directly to private investors as opposed to share auctions. Our model calls attention to several other issues. First we note that the agent’s bargaining power a, affects the public revenues. The higher is a, the less likely is the official to voluntarily reveal information and the more likely he is to collude with one of the bidders regardless of the mechanism used. Thus, a higher a may actually translate into lower public revenues. On the other hand, even though a does not affect the outcome of the auction directly, (only through the official’s decision concerning the sharing of information), it directly influences the price attained through private negotiations. Given that the agent

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2745

colludes, the more bargaining power he has, or the better negotiating skills he possesses, the more public revenues private negotiations generate. These ideas lead to some interesting policy implications. For example, it is plausible to think that an entrenched privatization official who expects to be involved in many deals to come is more likely to have a substantial bargaining power as opposed to one who is involved in a single deal only. In particular, our theory suggests that the government can exert indirect control over the privatization process by nominating different officials to run each privatization program (as in Hungary for example) or by choosing officials who will be responsible for many transactions. Another important variable in our analysis is the non-pecuniary benefits, W. This variable represents the prestige of the privatization agent’s office or the agent’s ability to receive bribes in the future. Thus, agents with different potential tenure are likely to behave differently. Limiting tenure to a specific period, or limiting the number of deals the official can direct, will reduce W, and may increase the official’s willingness to accept bribes. On the other hand, by guaranteeing a period of employment, the government can push W upward, thus making the privatization agent less willing to accept bribes, at least in the early years of his career. Hence, there is a trade-off between W and a. The same measures that increase W, are likely to decrease a and vice versa. 7. Concluding remarks This paper has analyzed the privatization process within an agency framework. We have focused on the role of a privatization agent and demonstrated that a potential agency conflict may substantially affect the choice of privatization mechanism. An agency problem spanned by political constraints has interesting implications. Paradoxically, private negotiations may raise more value than auctions when the agent-in-charge highly values staying in office and uses his bargaining power to negotiate his target price. Disclosure rules that seem to render an auction more transparent may work in favor of corrupt bidders. Our research highlights how the effectiveness of the political process, the severity of the political constraints and the availability of information affect the choice of mechanism between privatization auctions and private negotiations. Moreover, our theory sheds some light on the widespread use of private negotiations in economies where political constraints are significant. There are several important directions for future research. One is the characterization of optimal contracts for government agents conducting privatizations of large-scale enterprises in countries with political constraints and agency problems. Such contracts might specify the length of tenure, the number of deals an agent is involved and the compensation/penalty structure. Other interesting problems awaiting future research include (i) the study of privatization mechanisms when the sale object has some private value to the buyer (for example, when the bank complements a bidder’s other business activities); or (ii) when the sale contract includes a buyback clause and there are moral hazard, effort provision, over-optimism and adverse selection problems in the post-sale operation of the enterprise. Acknowledgements We acknowledge helpful comments from Vicente Madrigal, Enrico Perrotti, Anthony Saunders, Lemma Senbet, Andrei Shleifer, Jayanthi Sunder, Robert Vishny, Ingo Walter,

2746

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

Luigi Zingales and participants in conference presentations at the Pacific Basin Finance Conference, The European Finance Association Meetings, The Annual Conference on Financial Economics, University of Maryland, at the American Economic Association meetings and at seminars at UC-Berkeley, Columbia, Haifa, NYU, Rutgers and Tel Aviv Universities, the City University of London, the University of Cyprus, ESSEC, Dartmouth, INSEAD UCLA, USC especially John Matsusaka and Jan Zabojnik, the University of Amsterdam, and the United Nations Research Council. Appendix Proof of Proposition 4 Step 1: Information purchase (c 6 e/4). Whenever the uninformed purchases information, then with probability 1/2 he learns that he is bidding for the good company. Since disclosure rules are in effect, his information acquisition is disclosed to the colluding parties. Consequently, both bidders bid v  e in perfect equilibrium (Proposition 1) and the non-colluding party makes a gross profit of e with probability 1/2. Consequently, it is only worthwhile for the uninformed to purchase information if c 6 e/4. Step 2: No information purchase (c > e/4). Nonexistence of pure strategy equilibrium. Suppose the uninformed bids bu and the informed bids v or bi conditional on the value of the company. If the informed bids more than E(v) for the good company then the uninformed bidder’s best response is v (any above v bid would bring an expected loss to the uninformed). However, if the uninformed bids v, then the informed’s best response is v + e for the good company. If the informed bids less than E(v) then the uninformed’s best response is to outbid him. However, for any bid by the uninformed, the informed’s best response to outbid him for the good company. Finally, if the informed bids E(v) for the good company, then the uninformed will stay away from this bid. Consequently, there exists no pure strategy equilibrium. Step 3: Characterization of the perfect equilibrium when c > e/4. The perfect equilibrium can be computed using iterated elimination of weakly dominated strategies. In a perfect equilibrium the only strategies that are played with positive probabilities are those that survive iterated elimination of weakly dominated strategies. Bidding v for the good company is strongly dominated for the colluding bidder. Bidding more than v for the bad company is weakly dominated for the colluding bidder. Bidding more than E(v) is weakly dominated for the non-colluding bidder. Taking the iteration one step further, bidding more than E(v + e) is also weakly dominated for the colluding bidder. If the colluding party bids E(v) or above with positive probability for the good company then the non-colluding party will never bid E(v) since bidding E(v) would yield him an expected loss. Step 4: Uniqueness when c > e/4. It follows from Step 3 that there exists ~bC 2 ðv; vÞ such that for every bC > ~ bC BRN (bC) = v where BRN is the best response correspondence for the non-colluding party. We select the smallest of these ~bC s and denote it by ^bC : We know from Steps 2 and 3 that whenever the non-colluding party bids between ðv; v  2 2Þ the colluding party’s best response is to outbid him. Similarly, whenever the colluding party’s strategy specifies an expected bid bC < ^ bC ; it is the non-colluding party’s best response to outbid him. Consequently, for any bC < ^ bC , the hyperplane bC = bN would separate C 1 N C 1 (BR ) and BR where (BR ) is the inverse correspondence of the colluding party’s best response correspondence. Hence, there exists no bC < ^bC ; such that

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2747

BRC(bN) = BRN(bC)1. Since from Step 3 we know that there exists no bN such that BRC(bN) = v, therefore, ^ bC is the unique solution to BRN(bC) = (BRC(bN))1. It follows from Step 2 that there exists no Nash equilibrium in which the colluding party would play a pure strategy, therefore, ^ bC must be the outcome of a mixed strategy. Since the colluding bidder would not be willing to play the same non-degenerate mixed strategies against different mixed strategies by the non-colluding bidder there is a unique bN that solves BRC ðbN Þ ¼ ^ bC . Consequently, there is a unique perfect equilibrium in mixed strategies. vþv Step 5: The expected price of the good firm is less than 2 . When E(v) is an acceptable bid, it trivially follows from Step 3 that no bidder bids higher than E(v) and that no bidder bids E(v) with probability 1 for the good company. Consequently, the expected price of the good firm is less than E(v). When E(v) is not an acceptable bid, then let EV* and EV* denote the closest allowable bid from above and below, respectively. Now we need to show that qI the probability the informed bids EV* is less than half. For the colluding bidder to bid EV* it must be the case that the non-colluding bidder bids EV* with some probability. For the non-colluding bidder to bid EV* it must be the case that EV þ ðv  EV Þ  ProbðEVN winsÞ P 0; 2 EV ProbðEVN winsÞ P : 2ðv  EV Þ EV Since vEV P1/2, Prob(EVN wins) P 1/4. However, for Prob(EVN wins) P 1/4, it must be  the case that the non-colluding bidder bids EV* for the good company with probability less than 1/2. But if it is the case, then the expected winning bid for the good company is less than E(v). Obviously, the colluding party may not bid as high as EV* with positive probability in which case the expected winning bid is even lower. h

Proof of Proposition 5 Step 1: c 6 e/4. The official is better off disclosing information to both bidders than not disclosing it to anyone. He gets no bribe either way, but in the former scenario he is more likely to stay in office. This so since an uninformed would never bid v  e with probability 1 regardless of what his beliefs are about the other bidder. The official’s expected gain from disclosing information to both bidders as opposed to none is bounded from below vv by ðpð 2 Þ  pðeÞÞ  W2 : When c 6 e/4, revealing information to one bidder dominates revealing information to both or none of the bidders. This is because when the official reveals information to one bidder then the non-colluding bidder will purchase information, the bids will tie at v  e for the good company and when the colluding bidder wins, the official also gets a side-payment. Step 2: c 6 e/4. Revealing information to one, both or none of the bidders may raise the same revenue when c > e/4. However, the official strictly prefers revealing information to one bidder only to his other options since this is the only way he can receive an additional side-payment. Consequently, not revealing information to any bidder is not part of any perfect equilibrium. h Proof of Proposition 6. The price the privatization official negotiates will exceed E(v). Since the expected winning bid in the auction is less then E(v) (see Proof of Proposition 4), the negotiated price will exceed any equilibrium winning bid in the auction. h

2748

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

Proof of Corollary 6a. It follows from the Proof of Proposition 4 that in the auction v  e will be a winning bid for the good company with positive probability in the perfect equilibrium when c > e/4. Whenever the price at which the utility of the agent is maximized exceeds v then private negotiations will attain at least v + e. It also follows from Proposition 4 that the highest winning bid for the good company in the auction is EV*. Consequently, if the price at which the utility of the privatization agent is maximized exceeds EV*, then the negotiated price will exceed any price attainable through auction. h Proof of Proposition 7. The proof of Steps 1, 2, 4 and 5 are identical to Steps 1, 2, 4 and 5 in Proposition 4 and are omitted. The proof of Step 3 is shown below. Step 3:   v  EV v  EV  e e=4 < c < max ; : 2 4 If the colluding party plays his equilibrium strategy derived in the Proof of Proposition 4, then it is worthwhile for the uninformed to purchase information and to outbid him as long as conditions of Proposition 7 hold. On the other hand, given that the non-colluding party purchases information, then the colluding party’s best response is to outbid him so that the resulting winning bid for the good company will be v  e (see also Proposition 1). If on the other hand, the non-colluding party bids v  e for the good company, then the colluding party will never purchase information as long as c > e/4. However, if the noncolluding party will never purchase any information then the colluding party will bid the same as in Proposition 4. Consequently, the non-colluding party will purchase information with some probability (less than 1) in any perfect Nash equilibrium and the winning bid will exceed those in Proposition 4. h Proof of Proposition 8 Step 1: c > e/4. The agent is at least as well or better off by disclosing information to both bidders than by not disclosing to anyone. When the official discloses information to both bidders, they always bid v  e for the good company (see Proposition 1) and they bid less otherwise. Step 2: c > e/4. Revealing information to one, both or none of the bidders may raise the same revenue when c > e/4. However, the official strictly prefers revealing information to one bidder only to his other options since this is the only way he can guarantee his side-payment. Consequently, not revealing information to any bidder is not part of any mixed strategy Nash equilibrium. h References Baldwin, C.Y., Bhattacharyya, S., 1991. Choosing the method of sale: A clinical study of conrail. Journal of Financial Economics 30, 69–98. Beck, R., Cull, L., Jerome, A., 2005. Bank privatization and performance: Empirical evidence from Nigeria. Journal of Banking and Finance 29 (8–9), 2355–2380. Berger, A.N., Clarke, G.R.G., Cull, L., Clapper, L., Udell, G., 2005. Corporate Governance and bank performance: A joint analysis of the static, selection and dynamic effects of domestic foreign and state ownership. Journal of Banking and Finance 29 (8–9), 2179–2222.

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

2749

Berkovitch, E., Israel, R., 1999. Optimal bankruptcy laws across different economic systems. Review of Financial Studies 12 (April), 347–377. Biais, B., Perotti, E., 2002. Machiavellian privatization. American Economic Review 92 (1), 240–258. Bonaccorsi di Patti, E., Hardy, D.C., 2005. Financial sector liberalization, bank privatization and efficiency: Evidence from Pakistan. Journal of Banking and Finance 29 (8–9), 2381–2406. Bonin, J.P., Hasan, I., Wachtel, P., 2005. Privatization matters: ‘‘Bank efficiency in transition countries’’. Journal of Banking and Finance 29 (8–9), 2155–2178. Boot, A., Thakor, A., 1994. Financial market design. Working Paper, University of Amsterdam. Boycko, M., Shleifer, A., Vishny, R.W., 1994. Voucher privatization. Journal of Financial Economics 35, 249–266. Bulow, J., Klemperer, P., 1996. Auctions versus negotiations. American Economic Review. Burkart, M., Gromb, D., Panunzi, F., 1997. Large shareholders, monitoring and the value of the firm. Quarterly Journal of Economics 112, 693–728. Clarke, G.R.G., Cull, R., Shirley, M.M., 2005. Bank privatization in developing counties: A summary of lessons and findings. Journal of Banking and Finance 29 (8–9), 1905–1930. Cramton, P., 1997. The FCC spectrum auctions: An early assessment. Journal of Economics and Management Strategy 6 (3), 431–495, Fall. Dewenter, K., Malatesta, P., 1997. Public offerings of state-owned and privately owned enterprises: An international comparison. Journal of Finance 52, 1659–1679. Diamond, D., 1996. Liquidity, banks and markets: Effects of financial development on banks and the maturity of financial claims, Working Paper, University of Chicago. Engelbrecht-Wiggans, R., Milgrom, P.R., Webber, R.J., 1983. Competitive bidding and proprietory information. Journal of Mathematical Economics 11, 161–169. Fluck, Z., 1999. The dynamics of the management-shareholder conflict. Review of Financial Studies 12 (April), 379–404. Fluck, Z., John, K., Ravid, S.A., 1995. Privatization in stages and the dynamics of ownership structure. PacificBasin Finance Journal 4, 277–296. Frydman, R., Rapaczynski, A., 1991. Markets and institutions in large-scale privatization. In: Corbo, V., Coricelli, F., Bossac, J. (Eds.), Reforming Central and Eastern European Economies. The World Bank, Washington, DC. Grossman, S., Hart, O., 1980. Disclosure laws and takeover bids. Journal of Finance 35, 323–334. Grossman, S., Hart, O., 1988. One share–one vote and the market for corporate control. Journal of Financial Economics 20, 175–202. Haber, S., 2005. Mexico’s experiments with bank privatization and liberalization. Journal of Banking and Finance 29 (8–9), 2325–2354. Harris, M., Raviv, A., 1988. Corporate control contests and capital structure. Journal of Financial Economics 20, 55–86. Hendricks, K., Porter, R.H., Tan, G., 1993. Optimal selling strategies for oil and gas leases with an informed buyer. American Economic Review Papers and Proceedings 83, 234–239. Hendricks, K., Porter, R.H., Wilson, C.A., 1994. Auctions for oil and gas leases with an informed bidder and a random reservation price. Econometrica 62, 1415–1444. Major, I., 1994. The constraints on privatization in Hungary: Insufficient demand or inelastic supply? Most 4, 107–146. McAfee, R.P., McMillan, J., 1987. Auctions and bidding. Journal of Economic Literature (June), 699–738. Megginson, W.L., 2005. The economics of bank privatization. Journal of Banking and Finance 29 (8–9), 1931–1980. Milgrom, P., Weber, R.J., 1982. The value of information in a sealed-bid auction. Journal of Mathematical Economics 10, 105–114. Perrotti, E., 1995. Credible privatization. American Economic Review 85 (4), 847–859. Porter, R.H., Zona, J.D., 1993. Detection of bid rigging in procurement auctions. Journal of Political Economy 101, 518–538. Ravid, S. Abraham, Spiegel, M., 1997. Optimal financial contracts for a startup with unlimited operating discretion. Journal of Financial and Quantitative Analysis (September), 269–286. Rothkopf, M., Harstad, R., 1994. Modeling competitive bidding: A critical essay. Management Science 40, 364– 384. Shleifer, A., Vishny, R., 1989. Management entrenchment: The case of manager-specific investment. Journal of Financial Economics 25, 123–139.

2750

Z. Fluck et al. / Journal of Banking & Finance 31 (2007) 2730–2750

Steven, J., Megginson, W.L., Nash, R.C., Netter, J.M., 1999. Share issue privatization as financial means to political and economic ends. Journal of Financial Economics 53 (2), 217–253. Stulz, R., 1988. Managerial control of voting rights: Financing policies and the market for corporate control. Journal of Financial Economics 20, 25–54. ‘‘The Revenge of the Nerds’’, The Economist, July 23, 1994. Titman, S., Subrahmanyam, A., 1997. The going public decision and the development of financial markets. Journal of Finance (June), 1045–1082. Wilson, R., 1967. Competitive bidding with asymmetric information. Management Science 13, A816–A820.