Shareholder Protection and Outside Blockholders: Substitutes or Complements? by Sergey Stepanov∗

Should outside blockholders be more common in countries with weaker shareholder protection? I show that there can be a U-shape dependence of the outside ownership concentration on the quality of shareholder protection. This result is in line with the recent empirical evidence questioning the traditional law and finance view. In my model, a lower cost of private benefit extraction makes outside monitoring less desirable for an entrepreneur, hereby calling for a lower outside blockholder’s share. However, a low blockholder’s share may provoke collusion between the entrepreneur and the blockholder, which hampers raising funds from dispersed shareholders. This trade-off yields the described U-shape relationship. (JEL: G 32, K 22)

1

Introduction

Most of the empirical law-and-finance research documents substantial differences in the patterns of ownership and control of public companies around the world. Barca and Becht (2001), La Porta, Lopez-de-Silanes, and Shleifer (1999), among others, find that in most countries, with the exception of the U.S. and the UK, control over companies is often concentrated in the hands of few large owners. A number of studies argue that legal protection of shareholders is a major explanatory factor for the cross-country differences in the ownership and control distribution (La Porta, Lopez-de-Silanes, and Shleifer (1999), La Porta et al (1998), Himmelberg, Hubbard, and Love (2004)).1 These papers document a negative relationship between the strength of shareholder protection and ownership concentration. A few theoretical models, e.g., Himmelberg, Hubbard, and Love (2004) and Shleifer and Wolfenzon (2002), rationalize this effect. Several recent papers reconsider the traditional law-and-finance view. Using a larger sample of companies that the previous studies did, Holderness (2009) finds that the U.S. ∗ New

Economic School, Moscow. I am very grateful to Mathias Dewatripont, Mike Burkart, Sergei Guriev, two anonymous referees and the editor for their comments and suggestions. I have also benefited from discussions with Marco Becht, Aleksandra Gregoric, Guido Friebel, Gianluca Papa and Cristina Vespro. I wish to thank as well seminar participants at ECARES and SITE. 1 Surveys by La Porta et al (2000b) and Denis and McConnell (2003) summarize this strand of research.

1

2 is roughly a median country in terms of companies’ ownership concentration. Controlling for firm-level characteristics, Holderness (2011) finds that legal shareholder protection is unrelated to ownership concentration. Spamann (2010) reexamines the data on the “antidirector rights index” across countries, which has been used as a primary measure of shareholders protection for more than a decade, and finds that the index values require substantial corrections. Furthermore, the author finds that when the corrected values are used, the empirical link between shareholder protection and ownership concentration disappears. Thus, the recent empirical findings suggest that the relationship between shareholder protection and ownership concentration is not straightforward. In this paper, I focus on the outside ownership concentration and present a theoretical model which obtains a Ushape dependence of the outside ownership concentration on the quality of shareholder protection. In a nutshell, the logic behind this result is as follows. An entrepreneur (manager) raises funds by selling equity shares in a competitive financial market. He chooses the ownership structure so as to minimize the inefficiencies, which he internalizes from the ex ante perspective. Costly monitoring by an outside blockholder is needed in order to reduce inefficient private benefit extraction by the manager. When shareholder protection deteriorates, private benefits become more valuable, and the manager prefers to reduce monitoring by decreasing the outside block size. However, when the law becomes too bad, the outside blockholder is tempted to collude with the manager for sharing private benefits. This reduces other investors’ willingness to provide funds. In order to eliminate the incentive to collude, the blockholder’s share has to be large enough. Further worsening of shareholder protection makes collusion more attractive, which requires to further raise the blockholder’s share in order to preserve the no-collusion incentive. Thus, the overall relationship between the law and outside ownership concentration turns out to be U-shape. Models by Himmelberg, Hubbard, and Love (2004) and Shleifer and Wolfenzon (2002) have one large controlling shareholder - the insider - and these papers yield a negative relationship between the quality of the law and insider ownership. Thus, they essentially study the effect of the law on insider ownership, without allowing for a presence of a large outside shareholder, whose interests could (partially) coincide with those of dispersed shareholders rather than with those of the insiders. In contrast to these studies, my model explicitly treats insiders and outside blockholders as distinct parties. Researchers have recognized both positive and negative sides of outside blockholders (see, e.g., Becht, Bolton and R¨oell (2002), section 5.1). Their frequently emphasized role is monitoring of managers (or, more generally, insiders, including controlling shareholders that are closely involved in management). By monitoring and exercising their control outside blockholders restrict insiders’ opportunism, being essentially a mechanism of corporate governance.2 However, there is a danger that such blockholders may use their 2 For

example, Lins (2003), using a sample of above 1400 firms from 18 emerging markets, finds that large non-management blockholders increase firm value. Maury and Pajuste (2005), in a sample of 136 Finnish listed non-financial firms, find that the presence of large non-family blockholders in a family firm is associated with a greater valuation, suggesting that non-family

3 control rights to expropriate other shareholders or collude with insiders. A widespread view on outside blockholders as a mechanism of corporate governance is that they are particularly important when legal shareholder protection is weak (Bergl¨of and Pajuste (2003), Bergl¨of and von Thadden (2000), Shleifer and Vishny (1997)). When legal protection is good, outside blockholders are not needed because investors are already well protected by law, and, since blocks are costly due to the lack of diversification or liquidity reasons, the optimal ownership structure should be dispersed. On the contrary, when legal protection is bad, a large outside shareholder is a very important (if not the only) instrument of restricting managerial opportunism. In line with this view, Burkart, Panunzi and Shleifer (2003) provide a model which yields a negative relationship between the quality of shareholder protection and the size of an outside blockholder. In contrast to the just described view, this paper derives a non-monotonic relationship between legal protection and outside ownership concentration. I look at the problem of corporate governance as a problem of the conflict of interest in the “triangle” of a manager, an outside blockholder and a continuum of dispersed shareholders. The model combines an optimal monitoring analysis a la Pagano and R¨oell (1998) with an analysis of the effect of collusion between the informed outside blockholder and the manager. I model legal shareholder protection through the cost for the manager of deriving private benefits at the expense of the shareholders. This cost is a deadweight loss, reflecting the difficulty of expropriation under a given quality of the law. An increase in this cost is associated with strengthening shareholder protection. My story goes as follows. The manager, who has no money, is trying to raise external funds by selling equity shares in a competitive market in order to implement a project. However, after the injection of funds has been made he has an incentive to derive private benefits instead of maximizing shareholder value. This creates the classical agency problem as in Jensen and Meckling (1976). In order to get financed the manager can reduce the agency problem by exposing himself to monitoring by a blockholder. When legal protection is very strong, the manager can resolve the agency problem by just keeping an equity stake large enough to create incentives not to derive private benefits. My focus, however, is on regimes with not so good legal protection, where blockholder monitoring is crucial to ensure financing. While monitoring is able to reduce managerial opportunism, a prospect of collusion between the blockholder and the manager reduces the willingness of the dispersed shareholders to provide funds. Since the investors break-even in the competitive financial market, from the ex ante perspective, the manager bears both the cost of monitoring3 and the cost of private benefit extraction. Hence, his aim is to choose the ownership structure that would minimize the combination of these costs subject to the investors’ participation constraint. While doing that the manager is concerned with two things. First, by setting the right blockholder’s share he wants to induce the optimal level of monitoring, i.e., the one that achieves the optimal balance between costly monitoring, which raises shareholder value, blockholders restrict private benefit extraction by families. 3 Through a reduced price of the block, since the blockholder has to be compensated for her monitoring effort.

4 and the inefficient private benefit extraction. Second, he wants to credibly commit to avoid collusion by choosing the sum of his own share and the blockholder’s share large enough such that the two parties jointly prefer profit maximization. These two goals, however, may be in conflict. In my model they are not when shareholder protection is relatively good, and they are when it is bad. When legal protection is relatively good (the cost of private benefit extraction is relatively high), there exists an ownership structure that both produces the optimal monitoring and is collusion proof. Like in Pagano and R¨oell (1998), the manager optimally trades off inefficiency of expropriation against the monitoring cost. When the law becomes worse, private benefits become more valuable for the manager. Hence, it becomes optimal to reduce monitoring in order to increase expected private benefit extraction. As a result, the optimal outside block decreases as the law deteriorates. When legal protection becomes sufficiently bad, i.e. private benefits become sufficiently valuable, collusion with the manager becomes very attractive for the blockholder and its threat starts driving the result. The blockholder’s share that would induce optimal monitoring under no threat of collusion violates the collusion-proofness constraint. Therefore, a larger share has to be allocated to the blockholder in order to kill the incentive to collude, which comes at the cost of excessive monitoring. As legal protection worsens further, collusion becomes more difficult to avoid and the blockholder’s share has to be increased in order to preserve the “no collusion” incentive. Thus, overall, I obtain a U-shape dependence of the outside blockholder’s share on the quality of shareholder protection. My model is closely related to two papers: Burkart and Panunzi (2006) and Burkart, Panunzi and Shleifer (2003). Like my model, these papers incorporate both the effect of blockholder monitoring and the effect of collusion between the blockholder and the manager. Burkart, Panunzi and Shleifer (2003) obtain a negative dependence of the outside blockholder’s share on the strength of shareholder protection. The initial owner of the firm seeks to maximize her welfare by selling a fraction of equity to the public and hiring a professional manager. As legal protection deteriorates, the manager is able to receive a higher rent from private benefit extraction. In order to keep this rent at zero, monitoring has to be more intensive when the quality of the law is poorer, which is achieved by the initial owner through keeping a larger equity stake.4 Burkart and Panunzi (2006), however, notice that in a more general model the link between shareholder protection and the blockholder’s stake can be both negative and positive, depending on the character of interdependence between the law and monitoring. While the finding of Burkart and Panunzi (2006) parallels this paper’s result in that strengthening legal protection does not necessarily lead to lower outside ownership concentration, there are important differences between the two papers. In particular, collusion does not qualitatively change the effect of the law in Burkart and Panunzi 4 The

paper also looks at the decision of a company founder to delegate management to a professional manager. The authors show that at sufficiently low levels of legal protection the founder prefers to manage the firm himself.

5 (2006)5 , while in this paper it does. Moreover, in Burkart and Panunzi (2006), while the blockholder’s share may or may not go up, monitoring always goes up as legal protection weakens. In this paper, on the contrary, monitoring may decrease when legal protection becomes weaker. The setups of Burkart, Panunzi and Shleifer (2003) and Burkart and Panunzi (2006), probably, better fit large, initially close firms, with no need for new investment, the owners of which want to hire a professional manager in order to improve the firm’s performance. In these papers, the initial owner maximizes her own welfare subject to either the manager’s participation constraint (Burkart, Panunzi and Shleifer (2003)) or the manager’s initiative constraint (Burkart and Panunzi (2006)). The optimal share the initial owners retains essentially balances the need to reduce managerial opportunism with the necessity to secure the manager’s participation or initiative.6 My analysis, on the contrary, better fits entrepreneurial companies, in which there is no question of hiring a new manager, but which seek external finance to realize their investment opportunities. There is no question of either managerial participation or initiative. Instead, the manager-entrepreneur wants to commit not to engage too much in inefficient private benefit extraction by exposing himself to monitoring. When private benefits become more attractive (shareholder protection weakens), he naturally prefers to preserve more of the private benefits and be monitored less, which calls for a lower blockholder’s share. However, the necessity to commit not to engage in collusion with the blockholder pulls the blockholder’s share in the opposite direction. It is worth noting that, despite having rather different setups and mechanisms that determine the link between the law and the ownership structure, both this paper and Burkart and Panunzi (2006) obtain that there is no simple relationship between the law and outside ownership concentration. This fact reinforces the validity of both papers’ claim that outside ownership concentration is not necessarily a substitute for legal protection. Another related paper that argues that improving shareholder protection may lead to a greater concentration of outside ownership is Stepanov (2010). In addition to shareholder monitoring, which may prevent managerial misbehavior ex ante, the paper introduces a possibility to sue the manager after his self-dealing is revealed. The paper shows that lowering the cost of litigation for an individual shareholder, which is naturally considered an increase in shareholder protection, may result in an ownership structure with a greater outside blockholder’s share and more monitoring, when courts are biased towards managers. This happens because, under manager-biased courts, ex ante monitoring is a more efficient mechanism of shareholder protection than ex post litigation, which implies that litigation should be used less relative to monitoring. How5 Collusion

does not qualitatively change the effect of legal protection in Burkart, Panunzi and Shleifer (2003) either. 6 The interaction between blockholder monitoring and managerial initiative was first examined in Burkart, Gromb and Panunzi (1997). Too much monitoring reduces the initiative. Because the initiative is valuable for shareholders, a blockholder’s share must not be too large in order to credibly avoid overmonitoring.

6 ever, when suing is easy for an individual shareholder, the shareholders cannot commit to the optimal use of litigation. A greater blockholder’s share is then needed in order to induce greater prevention of managerial misbehavior through ex ante monitoring so as to minimize occasions for suing the manager ex post. The paper is organized as follows. Section 2 sets up the model. In section 3 I solve the model. Section 4 analyses the effects of legal protection on the ownership structure. Section 5 discusses the results. In section 6 I analyze robustness of the results and compare them with those of the related papers. Section 7 concludes the paper. I have also created an online appendix accessible at http://www.nes.ru/dataupload/files/professors/online appendix 19Mar2012.pdf, which discusses several extensions of the model. 2

The model

All agents are risk-neutral. There is an entrepreneur (whom I will call the manager) who has a positive NPV project and necessary managerial skills but does not have funds. Implementing the project requires the investment outlay I which has to be raised from outside investors.7 I assume that due to high competition between potential investors the manager has full bargaining power at the financing stage, i.e. he will maximize his payoff subject to the investors’ participation constraint. I assume also that all funds are raised by selling equity. The only verifiable variable in the model will be the project’s return (security benefits), which will be binary: either 0 or some fixed value R, and the manager is assumed to have limited liability. So, I do not constrain the space of feasible contracts by considering only equity contracts, because any contract in such a framework can be interpreted as equity. Although other interpretations are equally possible (e.g., debt) the purpose of the model is to provide implications for the ownership structure; therefore, I will stick to the “equity” interpretation. The manager retains fraction αm of the shares for himself, sells part αb of the shares as a block to a large shareholder (the blockholder), and the remaining part αd – as dispersed equity.8 The funds can be used by the manager both for creating verifiable return and for deriving a non-verifiable private benefit. Lack of commitment to abstain from private benefit extraction will reduce the shareholders’ willingness to provide finance in the first place. However, the shareholders can limit managerial opportunism through monitoring. The role of the blockholder will be to solve a collective action problem in monitoring. The blockholder, however, may sometimes be tempted to collude with the manager for sharing the private benefit instead of pursuing shareholder interests. The following 7 Alternatively

we can assume that the manager initially is a sole owner of a firm which runs with zero profit. He receives zero private benefits and has no cash. The firm has a positive NPV project that requires investment I. 8 I allow for only one outside blockholder. In the online appendix I discuss several modifications with multiple outside blockholders and show that they yield the same relationship between shareholder protection and outside ownership concentration as the basic model does.

7 subsection sets up the game formally. 2.1 The game The game is illustrated in Figure 1. The manager’s payoff is denoted by M (Mc in the case of collusion), the blockholder’s payoff – by Sb (Sbc in the case of collusion), and the dispersed shareholders’ payoff – by Sd . The sequence of the events is as follows: t = 0. The manager decides on the ownership structure of the firm (αd , αb ) and share αm of the cash-flow rights he retains. The manager also sets the prices of the block and the dispersed equity and makes a take-it-or-leave-it offer to the prospective blockholder and dispersed investors. It is assumed that monitoring capital is not scarce. The manager can set different per-share prices for the block and for dispersed equity. The investors (both the prospective dispersed shareholders and the blockholder) decide whether to provide funds. I assume that investing more than I in the firm is unproductive. However, I allow the manager to raise any K ≥ I and simply pocket the excess funds. t = 1. The manager pockets K − I and has two options for using I: he can either create a verifiable return (we will call this option profit maximization), or derive nonverifiable private benefit B (private benefit extraction, self-dealing).9 Simultaneously with the manager’s action, the blockholder chooses her monitoring effort. Monitoring means the following. The blockholder chooses how much to invest in monitoring – the monitoring cost c, which is born solely by her. With probability p(c) she becomes “informed”. Being informed means that in case the manager has chosen private benefit extraction, the blockholder identifies the ways of managerial self-dealing and can force the manager to maximize profits instead. In case the manager has chosen profit maximization, the blockholder realizes that no self-dealing is taking place. With probability 1−p(c) the blockholder stays uninformed – even if the manager is self-dealing and the blockholder rationally expects that, she does not see how this self-dealing is taking place. I assume that in this case she is unable to stop self-dealing.10 Function p(·) is defined on [0, ∞), everywhere increasing, differentiable and strictly concave, p(0) = 0, p(∞) = 1, p′ (0) = ∞, p′ (∞) = 0. I assume that the dispersed shareholders cannot coordinate for monitoring, and since each of them is of measure zero no one of them will monitor.11 t = 2. If the manager has chosen profit maximization, the return is R with probability 1−q and 0 with probability q. The return is distributed among all shareholders (including 9 The

binary choice structure facilitates the exposition. If the manager could select any convex combination of profit maximization and self-dealing his choice would eventually be binary anyway due to the linearity of his maximization problem. 10 Alternatively, I could assume that monitoring simply reduces the level of self-dealing. As I show in the online appendix, the results of the model would remain the same. 11 If I had introduced a small fixed cost of monitoring, then “no monitoring” would be a dominant strategy for any dispersed shareholder.

8 the manager) according to their shares, and the manager obtains no private benefit. Let us denote the expected return in this case, (1 − q)R, by Π. Hence, a shareholder with share α receives αΠ in expected terms. I assume Π > I, that is, if the manager maximizes profits, the project’s NPV is positive. If the manager has chosen self-dealing, and the blockholder is uninformed, the manager obtains private benefit B and the return is zero with certainty. If the manager has chosen self-dealing, and the blockholder is informed, the blockholder has two options: 1. She can force the manager to switch to profit maximization (other outside shareholders would obviously support her in this case). Then the payoffs are exactly as if the manager had chosen to maximize profits.12 2. She can collude with the manager for sharing private benefit B.13 I assume that the parties bargain according to the generalized Nash bargaining solution. If the parties do not agree on sharing the private benefits (that is, do not collude), profit maximization follows. That is, the disagreement point utilities of the manager and the blockholder are αm Π and αb Π correspondingly. The surplus from collusion is Σ ≡ B − (1 − αd )Π. In the case of collusion, the manager’s and the blockholder’s payoffs are respectively Mc = αm Π + µΣ and Sbc = αb Π + (1 − µ)Σ, where subscript c stands for “collusion”, and µ ∈ [0, 1] reflects the bargaining power of the manager and is exogenous.14 I assume that if the parties are indifferent between collusion and profit maximization, they choose the latter. In case there is no financing at t = 0, all the parties get a zero payoff. I make the following crucial assumption: Assumption 1. B < Π. This assumption says that private benefit extraction is inefficient. I assume that the cost of extracting private benefits, Π − B, is a deadweight loss. Assumption 1 implies that the first-best solution is to maximize profits. The model does not allow for retrading shares in the market before the profits are realized. The implicit assumption is that the market is sufficiently transparent and/or trading involves sufficiently large transaction costs. Similarly to Pagano and R¨oell (1998) and Burkart, Gromb and Panunzi (1997), it can be shown that in a transparent market no party can gain by retrading, because any attempt to trade becomes immediately reflected in the stock price. While perfect transparency of trades is an extreme assumption, in reality significant trades of stock rarely remain unnoticed either due to disclosure rules or because they lead to abnormal trading volumes. Additionally, trading involves 12 I

show in the online appendix that the results remain qualitatively the same if we introduce a cost of switching to profit maximization, provided that the cost is not too large. 13 Thus, collusion is about sharing private benefits and occurs after monitoring. In Pagano and R¨oell (1998) collusion happens before monitoring and is about the level of monitoring. I discuss this difference in modeling collusion in detail in section 6. 14 Making µ a function of α and α would not change the results of the model in any way. m b As it will be clear in subsection 3.1, the equilibrium solution does not depend on µ.

9 Figure 1 The game. M + Sb = B Sd = 0 Collusion t=0 Manager chooses the ownership structure (ab, ad) and the prices Investors decide whether to buy the shares

t=1

p(c) Blockholder is informed t = 2

Manager chooses between profit maximization and private benefit extraction Blockholder chooses monitoring effort c

Profit maximization 1 – p(c) Blockholder is uninformed

M = amΠ Sb = abΠ Sd = adΠ M=B Sb = 0 Sd = 0

transaction costs. Thus, I believe that a more realistic model with some market intransparency and transaction costs would not produce results qualitatively different from this model. Another way to rule out retrading is to assume that the manager and the blockholder include the prohibition to trade their stock explicitly in the contract. From the ex ante perspective, the manager would like to commit that the optimal ownership structure remains unchanged, so such explicit restriction should be optimal for him ex ante. 2.2 Legal protection We assume that legal shareholder protection determines the value of B. Lower B means a higher cost of private benefit extraction. Hence, we associate lower B with better legal protection. The magnitude of B reflects the restraints that both the contents of the law and law enforcement put on expropriating shareholders. Better law makes finding selfdealing opportunities more difficult and forces a manager to search for more complicated, and thus more costly, ways of expropriation (e.g., via establishing complex and nontransparent structures with intermediary companies). Stronger legal protection also imposes higher expected penalties on wrongdoers (e.g., because of a higher probability of being caught and found guilty). While it is common to assume that the law determines the cost function of private benefit extraction (La Porta et al (2002), Shleifer and Wolfenzon (2002), Himmelberg, Hubbard, and Love (2004), Burkart and Panunzi (2006), Burkart, Panunzi and Shleifer (2003)) it is not obvious what the exact relationship between the law and the cost function should be. The only feature that seems unquestionable is that better law should increase (at least weakly) the cost of self-dealing for any given amount of self-

10 dealing. My model obviously satisfies this requirement. Since the manager’s choice consists of only two options, my model is equivalent to a model with a linear cost of self-dealing with the marginal cost being (1 − B/Π)x, where x ∈ [0, Π] is the amount of the return diverted. Such stealing technology would lead to a corner solution, which my model yields by construction. The law increases the slope of this function. La Porta et al (2002), Shleifer and Wolfenzon (2002) and Himmelberg, Hubbard, and Love (2004) share the latter feature as well. However, in those papers the cost is a convex function, which results in an interior solution for expropriation. I will discuss the robustness of my results with respect to the form of the cost function in section 6. In Burkart, Panunzi and Shleifer (2003) and for the most part of Burkart and Panunzi (2006) expropriation does not involve ex post inefficiency, i.e., there is no ex post cost of transferring wealth from the shareholders to the manager, and legal protection just sets an upper bound on the amount of the return that can be diverted. This implies that, regardless of the joint equity share of the controlling parties, they would always want to divert the maximum amount of the return, allowed by the law, unless it is exogenously assumed that the blockholder pursues the small shareholders’ interests and cannot (or does not want to) collude with the manager.15 In my model, the choice between expropriation and profit maximization depends on the equity stake of the controlling parties, which is arguably more realistic. Both better legal protection and a higher equity share of the manager-blockholder coalition increase the chance of the “no expropriation” outcome. The aim of my analysis is to examine how the legal parameter B affects the choice of the ownership structure (αd , αb ). Before proceeding with the solution I am going to make an assumption that ensures the need for blockholder monitoring. 2.3 Legal protection and the need for an outside blockholder It is easy to see that when legal protection is sufficiently strong, specifically when B < Π − I, the first-best (profit maximization) will be achieved in equilibrium. Since investors are perfectly competitive, the manager obtains the whole surplus from his relationship with investors. Hence, it is ex ante optimal for the manager to commit to profit maximization, provided that investors provide financing. When B < Π − I, any αm that satisfies B/Π < αm ≤ (Π − I)/Π makes the manager better off from maximizing profits than from self-dealing at t = 1 (because αm Π > B) and satisfies the investors’ participation constraint (1 − αm )Π ≥ I. The manager then could choose αm = (Π − I)/Π and would obtain then the whole NPV of the project, Π − I. Thus, indeed, in line with the widespread view, when the law protects shareholders well enough, blockholder monitor15 The

authors consider two scenarios: one in which the blockholder’s interests are assumed to be perfectly aligned with those of small shareholders (e.g., because private benefits are not transferable), and one in which there is no assumption of alignment. In the latter case, due to the ex-post efficiency of expropriation, the blockholder always colludes with the manager and they divert the maximum possible amount of the return. In Section 5.2 of Burkart and Panunzi (2006) the authors allow for partial alignment by introducing a convex cost of expropriation. In that case an interior solution for expropriation is possible.

11 ing is not needed and the manager can be disciplined through an equity compensation scheme. The focus of this paper, however, is on the situations where a blockholder is needed, i.e. when legal protection is not good enough. Therefore, I make the following crucial assumption: Assumption 2. B > Π − I. This assumption says that legal protection is not strong enough for the parties to write a contract on the firm’s return that would prevent expropriation and satisfy the investors’ participation constraint at the same time. In other words, Assumption 2 says that the agency problem is serious enough that the equity share the manager needs to keep in order to credibly commit not to expropriate the shareholders is too large to raise the required funds. Thus, under Assumption 2 the first-best is not achievable. In the terminology of Tirole (2001) it is called the “dearth of pledgeable income” problem. Although Assumption 2 rules out the alignment of the manager’s and the shareholders’ interests, it does not rule out the alignment of the interests of the managerblockholder coalition with those of the dispersed shareholders – for a given quality of the law, there is a threshold share of the manager-blockholder coalition above which the parties jointly prefer profit maximization. Thus, as we will see, the presence of the blockholder who monitors the manager and abstains from collusion with him is crucial for restricting managerial opportunism and ensuring financing by dispersed shareholders. Throughout the subsequent analysis, when I will speak about B, I will use terms “good” and “bad” legal shareholder protection (or law), having in mind that we are in the world restricted by Assumption 2. That is, the values of B that correspond to “good” legal protection will be the ones that are not much above Π − I. In other words, the law can be “good” in my model but not “too good”. 3 Solution I am looking for subgame perfect equilibria of the game, determined by: (a) the pair (αd , αb ), (b) the decisions of the dispersed shareholders and the blockholder whether to provide funds or not, (c) the manager’s choice between self-dealing and maximizing profits, (d) the choice of c by the blockholder, (e) if the manager has chosen self-dealing and the blockholder is informed, the decision of the manager and the blockholder whether to collude or not. There will be three relevant conditions/constraints in the analysis: the collusionproofness constraint, the optimal monitoring condition, and the investors’ participation constraint. I will derive them now when solving the game. It will turn out that there is never collusion in equilibrium, and the blockholder’s share is such that the optimal balance between the inefficient private benefit extraction and costly monitoring is achieved, given the collusion-proofness constraint.

12 3.1 Formal solution We solve the game by backward induction. Assume the manager has chosen private benefit extraction. At t = 2 the manager and the blockholder abstain from collusion whenever they jointly gain (weakly) more from profit maximization than from deriving private benefits, i.e., whenever Σ ≤ 0. Let us call it the collusion-proofness constraint (CP): (1 − αd )Π ≥ B or (CP)

αd ≤ 1 − B/Π

At t = 1 self-dealing is a strictly dominant strategy for the manager. It can be seen from the following reasoning. The necessary condition for financing requires that (1 − αm )Π ≥ I, otherwise the investors would never provide funds. This condition, combined with Assumption 2, implies that αm Π < B. It means that when the blockholder is uninformed, the manager is better off deriving B than maximizing profits. Notice that due to the assumptions about p(·), there will always be a positive probability that the blockholder is uninformed. If the blockholder becomes informed, either profit maximization or collusion follows at t = 2. If there is profit maximization, then the manager has not lost anything from having chosen to self-deal. If there is collusion, the manager gets Mc = αm Π + µΣ > αm Π. Let us now turn to the blockholder choice of monitoring at t = 1. Assume that at t = 2 collusion is not optimal for the coalition of the manager and the blockholder, that is (CP) holds. The blockholder maximizes her expected utility net of the monitoring cost, p(c)αb Π−c. Due to our assumptions about p(·) the solution c∗ (αb ) is always interior and is determined by the first order condition: (M)

p′ (c∗ ) =

1 αb Π

Let us call this condition the monitoring condition (M). Since p′′ (·) < 0, c∗ is increasing in αb , which is very natural – the higher the blockholder’s stake the more she is concerned with the equity value and the more she monitors. Hence, provided that (CP) holds, and assuming that the manager raises K ≥ I, the participation constraint (P) of the investors is: p(c∗ (αb ))(αb + αd )Π − c∗ (αb ) ≥ K or (P)

αb + αd ≥

c∗ (αb ) + K Πp(c∗ (αb ))

To sum up, if after period t = 0 the values of αd and αb are such that constraints (P) and (CP) hold then in the subgame perfect equilibrium of the subgame that starts at t = 1:

13 (a) the manager tries to self-deal, (b) the blockholder monitors with intensity c∗ (αb ), (c) if the blockholder gets informed, there is no collusion and the firm’s strategy is changed to profit maximization. Will the manager at t = 0 want to choose an ownership structure that guarantees the absence of collusion, i.e. such that (CP) holds? The answer is “yes”. Lemma 1 Collusion never happens in equilibrium of the entire game. The formal proof of the lemma can be found in the appendix. The intuition behind this lemma is as follows. Since the manager always extracts all the expected surplus from his relationships with the investors, and extracting private benefits is inefficient, he prefers to establish a commitment that would assure that all the investors get their money back and collusion does not happen. There is also the issue of the loss from monitoring. However, if the ownership structure is chosen so that collusion occurs, the blockholder will still monitor the manager in order to get a piece of the private benefits. As I show in the appendix, if such a structure is consistent with the blockholder providing at least I (when collusion is expected to occur, dispersed shareholders do not provide financing), there exists another ownership structure which precludes collusion and induces exactly the same monitoring effort. Thus, if it is feasible to satisfy jointly (CP), (P) and αb + αd ≤ 1, then in equilibrium the manager will do so. Otherwise no financing takes place at t = 0. The necessary and sufficient condition for such ownership structure to exist is that the ownership structure with αb = 1 satisfies (P) for K = I. Such allocation of ownership obviously ensures the absence of collusion. At the same time it induces monitoring that maximizes the net investors’ payoff – when all the equity belongs to one shareholder, she will maximize the shareholder value net of the monitoring cost. Thus, if αb = 1 does not satisfy (P), no other ownership structure does. Hence, I make the following assumption: Assumption 3. Πp(c∗ (1)) − c∗ (1) − I > 0 The manager’s payoff as of t = 0 is p(c∗ (αb ))αm Π+[1−p(c∗ (αb ))]B+K −I. Obviously (P) must be binding, as the manager’s payoff is increasing in K. Then, taking into account that αm = 1 − αd − αb , the manager’s program at t = 0 can be written as: max{V (αb ) = p(c∗ (αb ))Π + [1 − p(c∗ (αb ))]B − c∗ (αb ) − I} αb ,αd

(feasibility) (P) (M) (CP)

s.t.: αb + αd ≤ 1, αb ≥ 0, αd ≥ 0 p(c∗ (αb ))(αb + αd )Π − c∗ (αb ) ≥ I 1 c∗ (αb ) is determined by p′ (c∗ ) = αb Π B αd ≤ 1 − Π

14 Hence, the manager maximizes the aggregate ex ante welfare subject to (P), (M), (CP) and the feasibility constraints. In the above program and further in the text, with a slight abuse of notation, I call constraint p(c∗ (αb ))(αb +αd )Π−c∗ (αb ) ≥ I participation constraint (P) (it is just the original participation constraint when K = I). Assume that the unconstrained maximization of V (αb ) results in some αb = αb . Then, if there exists αd , such that (αd , αb ) satisfies all the constraints, the unconstrained maximization of V (αb ) solves the manager’s program with respect to αb . Assume that such αd can be found. Since V (αb ) depends on αb only through c∗ (αb ), we can maximize V with respect to c∗ and then derive the optimal αb . Since p(·) is a strictly concave function and Π > B, V (c∗ ) is strictly concave too, and, thanks to our assumptions about p(·), it has an interior maximum. The first order condition yields: p′ (c∗ ) =

1 Π−B

This is the condition for the ex ante optimal monitoring. The optimal block αb is the one that equalizes the blockholder’s choice of monitoring with the optimal monitoring, i.e. using (M) we can write: 1 1 B = , i.e. αb = 1 − αb Π Π−B Π

(OM)

Let us call it the optimal monitoring (OM) condition. Remember, that this solution is valid only if we can find αd ≤ 1 − αb , such that (CP), (P) and the feasibility constraints are satisfied. If such αd does not exist, αb will be inevitably larger than αb and there will be too much monitoring with respect to the optimal level. In this case, the solution will be at the intersection of (CP) and (P) (taken as equalities) because such ownership structure produces monitoring as close to the optimal level as possible, given (CP), i.e., αb will be determined by c∗ (αb ) + I B − α = 1 − b Πp(c∗ (αb )) Π Let us denote this value by αbc , where c indicates that the collusion-proofness constraint is binding. Thus, the equilibrium is: (a) At t = 0 the manager chooses [

αb∗

= max{1 − B/Π,

αbc },

αd∗

] c∗ (αb∗ ) + I B ∗ ∈ − αb , 1 − , Πp(c∗ (αb∗ )) Π

the investors provide at least I. (b) At t = 1 the manager tries to self-deal, the blockholder monitors with intensity c∗ (αb ),

15 (c) At t = 2 if the blockholder gets informed, there is no collusion, and the firm’s strategy is changed to profit maximization. Otherwise private benefit extraction occurs. Notice that whenever αb∗ = αb the ownership structure is not uniquely determined with respect to αd (except when αb = αbc ). The reason is that the manager is only concerned with setting the blockholder’s share optimally. Once it is set, the dispersed equity share can be varied (as was mentioned at the beginning, the manager can raise more than I in exchange for more equity and simply pocket the excess funds). Neither the shareholders’ nor the manager’s expected welfare changes, only the composition of the manager’s share and the dispersed equity share changes. At the end of section 4 I discuss how the model could be modified to yield a unique solution for αb∗ for any parameters values. The focus of the paper, however, is on the outside ownership concentration. Therefore, multiplicity of optimal αd∗ is not a problem for the purpose of my model. 3.2 Graphical interpretation A graphical interpretation of the problem is a convenient way to understand the solution. Figure 2 depicts all the constraints in space (αd , αb ). Lines (P) and (CP) are the corresponding constraints taken as equalities. The collusion-proofness constraint (CP) is just a vertical line at αd = 1 − B/Π, and the optimal monitoring line, (OM), is a horizontal line at αb = 1 − B/Π. The most difficult thing is to understand how (P) looks like. Lemma 2 establishes the properties of (P) that are relevant for my analysis.16 Lemma 2 Given Assumption 3, constraint (P), taken as an equality, has the following properties in space (αd , αb ): (a) it is downward sloping, (b) αb = 0 when αd = ∞, (c) it intersects the line αb + αd = 1, and there is only one intersection point in the north-eastern quadrant (i.e. satisfying αd ≥ 0, αb ≥ 0) The formal proof of the lemma is presented in the appendix. So, the set of the ownership structures that allows to attract finance without collusion is bounded by (P), (CP) and αb + αd = 1. If B is sufficiently small, then (CP) is far enough on the right and (OM) is sufficiently high. In this case, optimality requires to set αb = αb , and the set of the optimal allocations is segment GH. If B is too high (B ′ in the figure), then (CP) is too much on the left and (OM) is too low ((CP’) and (OM’) respectively), and the solution is unique – it is point F, at which αb∗ = αbc and αd∗ = 1 − B ′ /Π. 16 Though

in Figure 2 constraint (P) is depicted convex, it is ambiguous whether it is actually convex for any functional form of p(·), satisfying our assumptions on p(·). However, this is not important for our results. What is important is the properties of (P) established in Lemma 2.

16

ab

Figure 2 Graphical solution. (CP') (CP)

1

ab* = abc

F (P)

ab* = ab = 1 -

B P

G

(OM)

H

ad + ab = 1 (OM') E 1 ad 1-B/P 1-B'/P Note: αb is the solution when the collusion-proofness constraint is not binding; αbc is the solution when it is binding.

4 The effects of legal protection Now we are ready to examine how the parameter of the law B affects the choice of the ownership structure by the manager. Let us look at Figure 3. When B is low enough (CP) is rather far on the right, while (OM) is rather high. So, we are in the situation when point H is above point F. In this case, the goal of achieving the optimal monitoring does not conflict with the goal of avoiding collusion, because the manager can choose the optimal outside block size αb and then simply keep his share large enough to ensure no collusion. Hence, the collusion-proofness constraint (CP) is not binding. The solution is segment GH, a bold line in Figure 3. The equilibrium blockholder’s share αb∗ is the one that induces optimal monitoring: αb∗ = αb . As B goes up, (OM) moves down, and (CP) moves to the left (to become (OM’) and (CP’) in the figure). Hence, αb goes down, i.e. the optimal outside block decreases when the cost of private benefit extraction falls. The segment GH shrinks and becomes G’H’. Since higher B means weaker shareholder protection, this result is contrary to the widely-held view that weaker legal protection should be associated with higher outside ownership concentration. However, it is analogous to the one in Pagano and R¨oell (1998), who argued that a lower cost of private benefit extraction should be associated with a lower optimal outside blockholder’s share.

17 Figure 3 Effects of an increase in B when B is low. ab (CP'') (CP') (CP) 1

B­

(P) a*b = a b = 1 - B / P a*b = 1 - B′ / P a*b min = 1 - Bˆ / P

G

(OM)

H

(OM')

H'

G'

(OM'')

G''=H''=F'' F'

1-

F

B B′ Bˆ 11P P P

E 1

ad

b > B ′ > B. (CP’) and (OM’) corespond to B ′ ; (CP”) and (OM”) correspond to Note: Here B b α∗ B; b min is the lowest value that αb takes in equilibrium.

At some value of B points G, H and F coincide (G”=H”=F” in the figure), I will b For B = B b there is a unique equilibrium ownership structure, denote this value by B. this is the highest value of B when (CP) constraint is not yet binding for the choice of αb . The corresponding αb is denoted by αb∗ min . b the necessity to avoid collusion drives the choice of the ownership strucFor B > B ture. Look at Figure 4 now. As B goes up further the manager cannot anymore choose αb as the solution, because that would lead to collusion for any αd that satisfies (P). The manager still wants to ensure the monitoring intensity as close to the optimal one as possible. Therefore, he chooses the blockholder’s share as small as possible, which corresponds to the intersection of (P) and (CP). In Figure 4 the solution is at point F’ (the corresponding blockholder’s share is αbc ) and this point goes up along constraint (P) as B rises further. The outside block gets larger and the dispersed equity share becomes smaller. In fact, in my model, if the blockholder becomes informed she can be treated as an insider ex post. The collusion-proofness constraint then can be viewed as the constraint on insider ownership that requires the alignment of the insiders’ and outsiders’ interests. Under good law, the alignment is achieved even for a small joint insiders’ share and is not a concern – the only concern is the optimal monitoring. Under bad law, the

18

ab

Figure 4 Effects of an increase in B when B is high. (CP')

(CP)

1

(P)

B­ ab* = a

F'

c b

G=H=F

ab* = ab* min

(OM)

E

(OM')

Bˆ 1 ad P b (CP’) and (OM’) corespond to B ′ . Note: Here B ′ > B. 1-

B′ P

1-

alignment consideration becomes the driving force – insider ownership has to increase with a decrease in the quality of the law in order to keep the alignment. This reasoning parallels the one in Shleifer and Wolfenzon (2002) who argue that a lower quality of legal protection requires higher insider ownership to credibly abstain from expropriation. One may still wonder why the blockholder’s share has to be increased when the joint share goes up. Why not keep it optimal and just increase the manager’s share to preserve the alignment? The answer is simple – the participation constraint must hold. If only the manager’s share is increased while keeping (CP) binding, a too small stake will go to the outside shareholders, and monitoring induced by the optimal blockholder’s share will be insufficient for the investors to break even. So, I obtain a U-shape dependence of the outside ownership concentration on B b In other words, I obtain a Uwith the minimum value αb∗ min which is reached at B. shape dependence of the outside ownership concentration on the quality of shareholder protection. The result is illustrated in Figure 5. The above analysis can be summarized in a proposition. Proposition 1 Legal protection of shareholders has a non-monotonic effect on the optimal outside blockholder’s share. When legal protection is sufficiently weak (i.e., the cost of private benefit extraction is below a certain critical level), the optimal outside

19 blockholder’s share is decreasing in the quality of legal protection; when legal protection is sufficiently strong (i.e., the cost of private benefit extraction is above the critical level), the optimal outside blockholder’s share is increasing in the quality of legal protection. Figure 5 Legal protection and the size of the outside block. ab*

ab* min Optimal monitoring

Collusion-proofness

B

B

The model also yields a prediction about the effect of the law on the dispersed equity share for high values of B. In Figure 4 the ownership structure is determined uniquely and it is clear from the picture that an improvement in legal protection (a decrease in B) leads to an increase in the dispersed equity share. However, in Figure 3 the equilibrium dispersed equity share is not unique, and strengthening legal protection simply gives the manager a wider choice of the dispersed equity share (compare G’H’ and GH), which does not allow us to make a definitive prediction. Proposition 2 Improvements in legal protection (a decrease in the cost of private benefit extraction) raise the dispersed equity share when legal protection is sufficiently weak (i.e., below a certain critical level). The positive relationship between shareholder protection and ownership dispersion is consistent with the empirical findings by La Porta et al (1998) and the theoretical work by Shleifer and Wolfenzon (2002), arguing that firms in countries with stronger b the link shareholder protection have higher ownership dispersion. However, for B < B is indefinitive. If I had introduced an infinitesimal incentive effect for the manager from having a larger share even when αm Π < B, then the solution would likely always be at constraint (P), because the manager would want to keep as large share as possible in order to create incentives for himself (assuming that this would be ex ante efficient). In such a case, the model would yield an inverted U-shape relationship for the dispersed equity share: with a decrease in B the collusion-proofness constraint would move αd∗ up for high values of B (as it does now), but the optimal monitoring condition would move αd∗ down for low values of B. On the other hand, if instead of an incentive effect I had introduced very small liquidity or risk aversion considerations, they would call for as large αd∗ as possible, i.e., the solution would always be at (CP) constraint. Then the result of Proposition 2 would

20 hold for any quality of shareholder protection within the limits set by Assumption 1. 5 Discussion of the results One way to view the results of this paper is through the trade-off between the ex ante optimality for the initial owner (the manager) and the ex post optimality for the participants of the game (the blockholder and the manager). Ex ante the manager is concerned with two things. First, he wants to credibly induce the optimal level of monitoring, i.e., the one that achieves the optimal balance between costly monitoring, which ensures certain positive shareholder value, and the inefficient private benefit extraction. Second, he wants to credibly commit to refrain from collusion. On top of that, the manager needs to satisfy the investor’s participation constraint, i.e., he cannot retain a too large share of equity. Ex post, however, the monitoring intensity is determined by the blockholder’s share and the decision whether to collude or not is determined by the joint share of the manager and the blockholder. The ideal ownership structure is the one that induces the ex ante optimal decisions ex post. The problem is, however, that it is not always possible. Specifically, it is possible when legal protection is good and it is not when legal protection is bad. This is the source of a U-shape link between legal protection and the optimal blockholder’s share. When the law is good, collusion can be avoided even through relatively small ownership concentration. Hence, there exists an ownership structure such that the optimal monitoring is induced, collusion does not happen and the outside shareholders’ equity share is large enough to secure their participation. The ex ante optimality requires to increase managerial opportunism when legal protection weakens (but does not become too weak) and expropriation becomes more valuable (similarly to Pagano and R¨oell, 1998). The fact that collusion is unattractive for a wide range of ownership structures provides the manager with enough freedom to adjust the outside block optimally to changes in the law. When the law becomes too bad, the ex ante optimal behavior cannot be ensured by the choice of the ownership structure, because of the conflict between the provision of the optimal monitoring incentives and the incentive not to collude. This conflict arises because collusion becomes too attractive ex post. To credibly avoid it, the total ownership concentration (the joint share of the manager and the blockholder) has to be large enough. The investors’ participation constraint restrains the manager from achieving this task by simply choosing his own share large enough and setting the outside block size optimally. As a result, the outside block is inevitably too large to produce the ex ante optimal monitoring. To put it another way, in order to attract financing, the optimal monitoring has to be sacrificed for the commitment not to collude. As legal protection worsens further, the size of the “collusion pie” goes up and both the total ownership concentration and the blockholder’s share increase correspondingly. As discussed in the Introduction, Holderness (2011) and Spamann (2010), in contrast to the earlier law and finance empirical literature, find no relationship between shareholder protection and ownership concentration. My model provides an explanation for

21 such results: when the relationship is non-monotonic, it is well possible that no statistically significant relationship is found in a linear regression. Hence, this paper suggests that in empirical studies researcher should account for a potential non-monotonicity of the effect of the law on ownership concentration. 6 Robustness and comparison with the related papers The driving forces behind the non-monotonicity result of Proposition 1 are: (1) ex post inefficiency of self-dealing under relatively strong legal protection, (2) necessity to sustain ownership concentration in order to kill the incentives to collude under weak legal protection. The relationship between the law and the blockholder’s share in the “threatof-collusion” zone (i.e., under high enough B) is simply due to the alignment effect, which looks rather natural: in order to provide incentives to maximize shareholder value, the controlling coalition needs to be allocated a larger equity stake when the attractiveness of private benefits is higher. The result for the “no-threat-of-collusion” zone, however, may look more debatable. However, this result holds for a wider class of cost-of-expropriation functions than just a linear function (recall that the present model is essentially a model with a linear cost of stealing, with the marginal cost of stealing 1 − B/Π). The critical condition for the positive relationship between the law and outside ownership concentration in the “nothreat-of-collusion” zone is that for any manager’s share consistent with the investors’ participation (i.e., not too large) his incentives are not too sensitive to changes in the law. In this case, strengthening legal protection does not lead to a sufficient reduction in the expropriation, but still increases its deadweight loss for a given amount of the private benefits. Hence, the overall result is a raise in the deadweight loss. This effect needs to be offset by an increase in monitoring, which is provided through raising the blockholder’s share. In such a case, Proposition 1 is valid. The “low sensitivity of expropriation to law” condition trivially holds when the manager’s decision is the corner solution (steal everything), as in the present model. Such solution arises regardless of the shape of the cost-of-stealing function, provided that the marginal cost of expropriation is low for any share of stolen profits. Another trivial example is when the cost-of-stealing function is a piecewise linear function as depicted in Figure 6. In this example, stealing involves a low enough cost until a certain point, and becomes prohibitively costly after this point. Better law increases the slope of each piece without moving the breakpoint. Provided that the slope of the left piece does not become too steep, with this type of function the manager would steal up to the breakpoint regardless of the quality of the law. Clearly, if we slightly modify the cost function by “smoothening” the breakpoint, the conclusion would not change. My model is closely related to two recent papers that looked at the same problem but provided different results: Burkart and Panunzi (2006) and Burkart, Panunzi and Shleifer (2003). Burkart, Panunzi and Shleifer (2003) propose a model that delivers a negative relationship between the quality of the law and outside ownership concentration in professionally managed companies. In that paper the ownership structure is chosen by a company’s founder who sells a share of her company and hires a professional manager,

22

Cost of diversion

Figure 6 Piecewise linear cost of diversion.

P Amount of profit diverted

subject to the manager’s participation constraint (job acceptance). The founder retains a block and monitors the manager, i.e., she essentially becomes an outside blockholder. Private benefit extraction by the manager involves no cost at all and the law just limits the amount of self-dealing from above. In such a case, there is no ex post inefficiency of private benefit extraction. Thus, monitoring is ex post inefficient, but is needed to reduce the managerial rent from private benefit extraction. As the law worsens, the manager is able to steal more and, hence, more monitoring is needed. Since, in the model, the monitoring incentive is independent of the quality of the law, the outside blockholder’s share needs to be increased in order to raise her monitoring effort.17 Burkart and Panunzi (2006), in a setup similar to Burkart, Panunzi and Shleifer (2003), qualify the result of the latter paper and show that a positive relationship between shareholder protection and outside ownership concentration can be possible as well, depending on the character of interdependence between the law and monitoring. In particular, they show that when legal protection and monitoring are either complements or independent for the blockholder, the optimal size of the outside block always decreases when the law improves, while when they are substitutes, it may increase. In both mentioned papers, in equilibrium, monitoring unambiguously rises as the law worsens, because the role of monitoring is to reduce the manager’s rent. The reason why the blockholder’s share may go down in Burkart and Panunzi (2006) is that, when legal protection and monitoring are substitutes, reducing the quality of the law may provoke a too large increase in monitoring for a fixed blockholder’s stake. Then, in order to secure the manager’s initiative (which is equivalent to securing participation in Burkart, Panunzi and Shleifer (2003)), the blockholder’s share needs to be reduced. In my model, the role of monitoring is to reduce the inefficiency of private benefit 17 When

legal protection becomes too poor, the founder prefers to manage the firm himself instead of hiring a manager. In this case, the ownership structure becomes irrelevant.

23 extraction, which goes down as the law becomes laxer – hence the positive relationship between the quality of the law and monitoring in the no-collusion area. If we turn now to the role of collusion, it plays a very different role in this paper and the two discussed papers. The qualitative results in Burkart, Panunzi and Shleifer (2003) and Burkart and Panunzi (2006) do not change if the possibility of collusion is introduced. Due to the fact that ex post private benefit extraction does not involve a cost, no ownership concentration can make the manager-blockholder coalition abstain from collusion. Thus, collusion avoidance consideration is absent from their papers.18 In the present paper, however, such consideration plays a very important role and collusion completely reverses the effect of the law when legal protection becomes bad enough. Thus, there are important differences in the setups between my paper and theirs, which drive the difference in the results. I do not intend to say that one setup better fits the real world than the other one. Rather I think that their models and this one analyze different types of companies. The setups of Burkart, Panunzi and Shleifer (2003) and Burkart and Panunzi (2006), probably, better fit large, initially close firms, with no need for new investment, the owners of which want to hire a professional manager in order to raise performance. My analysis, on the contrary, better fits entrepreneurial companies, in which there is no question of hiring a new manager, but which seek external finance to realize their investment opportunities. There is no question of either managerial participation or initiative. Instead, the manager-entrepreneur wants to commit not to engage in inefficient private benefit extraction by exposing himself to monitoring. When private benefits become more attractive (shareholder protection weakens), he naturally prefers to preserve more of the private benefits and be monitored less, which calls for a lower blockholder’s share. However, the necessity to commit not to engage in collusion with the blockholder pulls the blockholder’s share in the opposite direction. At the same time, the fact that both the present paper and Burkart and Panunzi (2006), despite different setups, obtain that the blockholder’s share does not have to decrease with the quality of the law reinforces both papers’ claim that outside ownership concentration is not necessarily a substitute for legal protection. Pagano and R¨oell (1998) have an extension of their basic model, in which the manager and the outside blockholder can collude before monitoring takes place (subsection IV.2 of the paper). The two parties simply select the level of monitoring that maximizes their joint payoff. As a result, in their paper, collusion leads to undermonitoring. In contrast, in my paper the possibility of collusion leads to overmonitoring. While some pre-monitoring agreement between the manager and the blockholder is possible, I would argue that post-monitoring collusion is more likely to be sustainable. First, pre-monitoring collusion requires that the blockholder can commit to the agreed level of monitoring, which implies that the choice of monitoring is at least observable. While, to some extent this can be achieved (e.g., by placing a certain agreed number of the blockholder’s representatives in the board of directors), a large part of monitoring activ18 At

the end of Burkart and Panunzi (2006) the authors just briefly discuss what can happen if private benefit extraction is inefficient and thus higher ownership concentration reduces the incentive to expropriate dispersed shareholders.

24 ity can probably be hidden. Second, pre-monitoring collusion requires that the manager can commit to make a transfer to the blockholder in exchange for lower monitoring. If such a transfer is to be made after the choice of monitoring, the manager has a clear incentive to renege on his promise. At the same time, making a transfer before the choice of monitoring may be problematic because the manager has not yet received his private benefits out of which he could make a transfer. In my model, private benefits are already available when the two parties bargain. Also, when the blockholder is informed, she has evidence of managerial misbehavior in her hands. Implicitly, she could threaten the manager by promising to go to a court with this evidence if the manager does not give her a part of the private benefits. Notice that such threat would likely be credible, provided that a court could revert (some of) the manager’s self-dealing. 7 Conclusion The main point of this paper was to show that outside ownership concentration can both decrease and increase with the quality of legal shareholder protection. Under very good legal protection an outside blockholder is not needed at all. The focus of the paper, however, was on the situations when legal protection is not so good and blockholder monitoring is crucial to ensure financing. The paper looked at the link between the law and the ownership structure in a detailed way, treating differently the firm’s insider (the manager) and the outside blockholder. Legal protection affects both the incentives of the manager-blockholder coalition to expropriate outside shareholders and the monitoring incentives of the blockholder. Weaker shareholder protection makes private benefits more attractive (less costly) for a manager. On the one hand, this makes monitoring less desirable, hence, calling for a lower blockholder’s share. On the other hand, lower blockholder’s share implies a greater incentive to collude with the manager. Overall, this trade-off yields is a non-monotonic effect of shareholder protection on the outside ownership concentration. This results contrasts the traditional law and finance view, but is in line with the recent empirical evidence on the relationship between law and ownership concentration. Similarly to Burkart and Panunzi (2006) and Stepanov (2010), the model illustrates that the relationship between law and companies’ ownership structures is generally multifaceted and non-monotonic. The paper suggests that in future empirical work researchers should account for a potential non-monotonicity of the link between shareholder protection and ownership concentration. Appendix A.1 Proof of Lemma 1 Assume that the manager chooses (αd , αb ) such that (CP) does not hold. Then collusion will occur at t = 2. Consequently, the dispersed shareholders will provide no funds at t = 0 and financing can come only from the blockholder. Assume the blockholder provides funds K ≥ I at t = 0. The manager will obviously go for private benefit extraction. The surplus of the colluding parties is Σ ≡ B − (1 − αd )Π > 0. If the

25 blockholder becomes informed, her stake in collusion is Sbc = αb Π + (1 − µ)Σ < Π, the manager’s stake in collusion is Mc = αm Π + µΣ = B − Sbc , where subscript c stands for “collusion” and parameter µ reflects the bargaining power of the manager. The blockholder’s monitoring effort is then: cc = arg max{p(c)Sbc − c} and satisfies c the first order condition: 1 p′ (cc ) = Sbc I am going to show now that the manager can make himself better off by selling no equity to dispersed shareholders (αd = 0), which ensures the absence of collusion, and picking the blockholder’s share such that her monitoring effort under no prospect of collusion equals cc and her welfare is unchanged. To ensure that under no prospect of collusion the blockholder’s monitoring effort is cc the manager should set her share α bb such that α bb Π = Sbc , i.e., α bb = Sbc /Π. Since Sbc < Π, share α bb is below 1, i.e., it exists. The manager’s share is α bm ≡ 1 − α bb . The blockholder’s welfare is obviously unchanged since she obtains the same as in the collusion case (b αb Π = Sbc ) when monitoring is successful. Thus, she would provide the same amount K. The manager is better off because the probability that the blockholder is informed has not changed and, in the event when the blockholder is informed, he gets (1− α bb )Π ≡ Π−Sbc +K −I instead of B −Sbc +K −I before. An alternative way to show that the manager is better off, is to recall that ex ante the manager prefers to reduce ex post inefficiency. Given that the monitoring intensity is the same, but collusion leads to inefficient private benefit extraction, he prefers the solution without collusion. Hence, I have proved that the manager will always prefer to choose the ownership structure that ensures the absence of collusion at t = 2. Q.E.D.

A.2 Proof of Lemma 2 1. Let us prove that (P), taken as an equality, is downward sloping. It is easy to see if we rewrite it as [p(c∗ (αb ))αb Π − c∗ (αb )] + αd p(c∗ (αb ))Π − I = 0. The term in square brackets is the blockholder’s utility when she optimally chooses her monitoring effort. This term is obviously increasing in αb , since the blockholder gets more from an increase in her share even if she does not optimally adjust her monitoring effort. The function p(c∗ (αb )) is increasing in αb as well since c∗ (·) and p(·) are increasing functions. Therefore, if we want to keep (P) binding, an increase in αb must be compensated by an appropriate decrease in αd . 2. Since c∗ (0) = 0 and p(0) = 0, it is straightforward to see that αb = 0 when αd = ∞. 3. First, let us show that (P) can intersect the line αb + αd = 1 at most twice. When αb + αd = 1 (P) becomes: Πp(c∗ (αb )) − I = c∗ (αb ) In terms of c, the left-hand side is a concave function of c, while the right-hand side is a 45◦ line. Hence, in terms of c, the above equation can have two solutions at most. Since

26 c∗ (αb ) is a strictly increasing and continuous function, the equation has two solutions at most in terms of αb too. Furthermore, Assumption 3 says: Πp(c∗ (1)) − c∗ (1) − I > 0 It means that point (0, 1) lies above (P). Since αb = 0 when αd = ∞, and (P) is a continuous function, then (P) intersects αb + αd = 1 in the north-eastern quadrant. Furthermore, since (P) can intersect αb + αd = 1 at most twice, there is only one intersection point the north-eastern quadrant. Q.E.D. References Barca, F., and M. Becht (eds.) (2001) The Control of Corporate Europe. Oxford University Press, London. Becht, M., P. Bolton, and A. R¨oell (2002), “Corporate Governance and Control,” ECGI Finance Working Paper No. 02/2002. Bergl¨of, E., and A. Pajuste (2003), “Emerging Owners, Eclipsing Markets? Corporate Governance in Central and Eastern Europe,” in: P. K. Cornelius and B. Kogut (eds.), Corporate Governance and Capital Flows in a Global Economy, Oxford University Press, New York, pp. 267–302. Bergl¨of, E., and E.-L. von Thadden (2000), “The Changing Corporate Governance Paradigm: Implications for Developing and Transition Economies,” in: S. S. Cohen and G. Boyd (eds.), Corporate Governance and Globalization: Long Range Planning Issues, Edward Elgar Publishing, pp. 275–306. Burkart, M., and F. Panunzi (2006), “Agency Conflicts, Ownership Concentration and Legal Shareholder Protection,” The Journal of Financial Intermediation, 15, 1–31. Burkart, M., D. Gromb, and F. Panunzi (1997), “Large Shareholders, Monitoring, and the Value of the Firm,” The Quarterly Journal of Economics, 112, 693–728. Burkart, M., F. Panunzi, and A. Shleifer (2003), “Family Firms,” The Journal of Finance, 58, 2173–2207. Denis, D. K., and J. McConnell (2003), “International Corporate Governance,” The Journal of Financial and Quantitative Analysis, 38, 1–36. Himmelberg, C. P., G. Hubbard, and I. Love (2004), “Investor Protection, Ownership, and the Cost of Capital,” World Bank Policy Research Working Paper No. 2834. Jensen, M. C., and W. H. Meckling (1976), “Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure,” The Journal of Financial Economics, 3, 305–360. Holderness, C. G. (2009), “The Myth of Diffuse Ownership in the United States,” The Review of Financial Studies, 22, 1377–1408. Holderness, C. G. (2011), “Do Differences in Legal Protections Explain Differences in Ownership Concentration?” Working Paper, Boston College. La Porta, R., F. Lopez-de-Silanes, and A. Shleifer (1999), “Corporate Ownership around the World,” The Journal of Finance, 54, 471–517. La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny (1998), “Law and finance,” The Journal of Political Economy, 106, 1113–1155. La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny (2000a), “Agency Problems and Dividend Policies around the World,” The Journal of Finance, 55, 1–33.

27 La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny (2000b), “Investor Protection and Corporate Governance,” The Journal of Financial Economics, 58, 3–27. La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny (2002), “Investor Protection and Corporate Valuation,” The Journal of Finance, 57, 1147–1170. Lins, K. (2003), “Equity Ownership and Firm Value in Emerging Markets,” The Journal of Financial and Quantitative Analysis, 38, 159–184. Maury, B., and A. Pajuste (2005), “Multiple controlling shareholders and firm value,” The Journal of Banking and Finance, 29, 1813–1834. Pagano, M., and A. R¨oell (1998), “The Choice of Stock Ownership Structure: Agency Costs, Monitoring, and the Decision to Go Public,” The Quarterly Journal of Economics, 113, 187–225. Shleifer, A., and D. Wolfenzon (2002), “Investor Protection and Equity Markets,” The Journal of Financial Economics, 66, 3–27. Shleifer, A., and R. Vishny (1997), “A Survey of Corporate Governance,” The Journal of Finance, 52, 737–783. Spamann, H. (2010), “The “Antidirector Rights Index” Revisited,” The Review of Financial Studies, 23, 467–486. Stepanov, S. (2010), “Shareholder Access to Manager-biased Courts and the Monitoring/Litigation Trade-off,” The RAND Journal of Economics, 41, 270–300. Tirole, J. (2001), “Corporate Governance,” Econometrica, 69, 1–35. Sergey Stepanov New Economic School Nakhimovsky prospekt 47, office 1721 117418, Moscow Russia E-mail: [email protected]