An Economic Analysis of Investor Protection in Corporations With Concentrated Ownership

Bennedsen, Morten; Wolfenzon, Daniel Working Paper An Economic Analysis of Investor Protection in Corporations With Concentrated Ownership Working paper, No. 15-2000 Provided in Cooperation with: Department of Economics, Copenhagen Business School (CBS) Suggested Citation: Bennedsen, Morten; Wolfenzon, Daniel (2000) : An Economic Analysis of Investor Protection in Corporations With Concentrated Ownership, Working paper, No. 15-2000, Copenhagen Business School (CBS), Department of Economics, Frederiksberg, https://hdl.handle.net/10398/7551 This Version is available at: https://hdl.handle.net/10419/208437 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/ Institut for Nationaløkonomi Handelshøjskolen i København Working paper 15-2000 AN ECONOMIC ANALYSIS OF INVESTOR PROTECTION IN CORPORATIONS WITH CONCENTRATED OWNERSHIP Morten Bennedsen Daniel Wolfenzon Department of Economics - Copenhagen Business School Solbjerg Plads 3, DK-2000 Frederiksberg An Economic Analysis of Investor Protection in Corporations With Concentrated Ownership Morten Bennedsen ¤ & Daniel Wolfenzon. ¤¤ December 2000. Abstract We provide a theoretical analysis of the relationship between investor protection and the performance of corporations with concentrated ownership. We present an incomplete contracting model of a corporation with concentrated ownership and apply it to analyze two types of investor protection. First, we analyze the cost and bene¯ts of imposing super-majority requirements on certain important policy issues in the corporation. Second, we analyze why it can be in the interest of the corporation to impose restrictions on the free transferability of shares. ¤ Copenhagen Business School, CEBR and CIE. Corresponding address: Department of Economics, CBS, Solbjerg Plads 3, DK 2000 F. Email: [email protected]. Phone: ( +45 ) 38 15 26 07. Fax: ( +45 ) 38 15 25 76. ¤¤ Michigan Business School and Chicago Business School. 1 Introduction A central issue in the corporate governance literature during the last twenty years has been the connection between the degree of agency problems and the performance of corporations. The size of an agency problem is closely related to the ability of owners to protect their investment. In particular, this has been emphasized in the so called incomplete contracting literature (e.g. Hart 1995), which focuses on the consequences of agents not being able to write complete contracts on all possible future contingencies. Obviously, in a world of incomplete contracts it is important to understand how investors' share holdings can be protected either through a corporation's charter or through the legal system and how such investor protection a®ects the performance of a corporation. This is the topic of the present paper. A recent empirical literature has studied this issue in a global context ( see La Porta, Lopez-de-Silanes and Shleifer 1998 and La Porta, Lopez-de-Silanes, Shleifer and Vishny 1998). They have shown various important facts about ownership structures and protection of investors around the world. First, concentrated ownership is common all around the world and is dominating outside the Anglo-Saxian world. Second, there is evidence for the real agency problem in many ¯rms are between di®erent classes of shareholders and not between the management team and the group of owners as the traditional corporate governance literature has focused on. Third, the degree of protection of shareholders in general and minority shareholders in particular varies a lot across countries. Finally, the degree of shareholders protection has real implications for dividend policy and ownership structure. All these features ¯ts badly with the traditional model in corporate governance of a public traded ¯rm with dispersed and weak owners that are exploited by a powerful and self interested management team. Instead, it may seem more appropriate to analyze how di®erent classes of shareholders form and how some groups of shareholders seize control over the corporation and exploit other groups of shareholders ( Shleifer and Vishny 1997). In the present paper we begin to analyze the link between protection of 1 share holding and the performance of corporations with concentrated ownership. In particular, we are interested in analyzing how protection of minority shareholders can a®ect the e±ciency and the distribution of rent in a corporation. Obviously, investor protection can be delivered in a large number of ways. To structure the analysis we have chosen to focus on two topics: Imposing super-majority requirements on central policy issues in the in the corporation and allowing for free transferability of shares in a corporation. We have picked these two topics because they seem to be very important not only according to the global facts mentioned above, but also in the corporate law literature ( see Clark 1986, O'Neal ??, or Easterbrook and Fischel 1991). To our knowledge, this paper is the ¯rst formal economic analysis of these issues. In Section 2 we set up an incomplete contracting model of a corporation with concentrated ownership. It is a simple model where the owners of a corporation hire a self interested manager to run the ¯rm. The manager can be chosen among the owners or be an outside manager with no ownership stake in the ¯rm. The owners can ¯re the manager if a majority ( which size can be stipulated in the corporate charter or by corporate law) wishes to do so. Associated with the manager's actions is a distribution of private bene¯ts to the manager and the owners. Di®erent actions are supported by di®erent groups of owners. Thus, our model endogenize the formation of various classes of owners. The manager's need to be backed by a majority of the owners gives rise to a con°ict between the majority and the minority shareholders and the outcome of this con°ict is a®ected by how shareholders are protected. In Section 3 and 4 we apply the model to analyze the cost and bene¯ts of providing protection to minority shareholders through changing the size of the majority necessary to ¯re the manager. Allowing groups of minority shareholders a veto right to ¯re the manager, naturally limits the amount exploitation these minority shareholders can be exposed to. Legal scholars have long argued that there is a trade-o® between protecting minority shareholders' investment and the °exibility the management need to run the ¯rm 2 e±ciently. For instance, Easterbrook and Fischel notice that \Drafters of the organizing documents of a closely held corporation cannot avoid a trade-o®. On the one hand, they must provide some protection to minority investors to ensure that they receive an adequate return on the minority shareholder's investment if the venture succeeds. On the other hand, they cannot give the minority too many rights, for the minority might exercise their rights in opportunistic fashion to divert returns." (Easterbrook and Fischel 1991, p.238.). In Section 3 we show that imposing super-majority requirements improves e±ciency when the manager can take non-contractible actions and there are complete information about the actions taken by the manager. The intuition is that with a super-majority requirement, the manager must have support from more shareholders than under a simple majority rule. This limits the manager's opportunities of pursuing projects that are not in the interest of all the owners. We also argue that none of the owners should object to such an super-majority in the certainty case. We then, in Section 4, introduce uncertainty about the value of the corporation which give rise to a trade-o® between protection of minority shareholders and the likelihood of costly deadlocks, de¯ned as situations where owners decide to replace the manager. We show that uncertainty can increase the payo® to the majority shareholders in the absence of a super-majority rule. Hence, providing veto rights to a group of minority shareholders may be resisted by the management and the existing majority shareholders both because it may decrease e±ciency and because it decreases the rent theses agents can obtain from the ¯rm. In short, we establish the trade-o® described in the legal literature, but only in the case of uncertainty. Section 5 analyzes the consequences of restricting shareholders right to resale their shares. From a ¯rst glance it could be argued that allowing exploited minority shareholders to opt out of the corporation limits the amount these shareholders can be exploited and, thus, increases e±ciency. However, this argument is °awed, because the balance of power in the corporation, i.e. the distribution of majority and minority shareholders, is endogenous. 3 For instance, we show, that allowing shareholders to sell cash °ow without selling votes alters the balance of power in the corporation, such that the new group of majority shareholders has a tendency to concentrate votes but not cash °ows. This decrease e±ciency in the corporation through increasing the amount of share holding that can be exploited. This argument explains why most close corporations have rules restricting the transferability of shares. Clark (1986), referring to close corporations, observes: \Shareholders ::: will usually want to restrict the transferability of their shares. ::: Sometimes the continuing shareholders will want the exiting shareholder to sell to the corporation, rather than to any of themselves, in order to preserve the existing balance of power' ( Clark ( 1986) p. 763, emphasis added). Conclusions are drawn in Section 6 and all proofs are delegated to the appendix. 2 The Model An entrepreneur ( also denoted the initial owner or the founder) seeks ¯nance to set up a ¯rm that at a future date yields a potential cash °ow of size r. She sells cash °ow rights, c,andvotes,v, to a number of outside investors. The timing of the model is as follows, Date 1 Firm established at cost °<1. Founder sells ownership stakes fv i ;c i g, i2I=f1; :::; Ig,whereIis the set of new owners. De¯ne v=fv i g i 2 I and c=fc i g i 2 I . Date 2 A manager, m, is hired. The manager can be one of the owners or an outside manager with no ownership stake in the ¯rm. De¯ne I ¡ m = Infmg ( =Iif the manager is not an owner) and I m =I ¡ m [fmg as the set of owners and management. Having a manager is necessary to create any value in the ¯rm. The manager picks a non-contractible action a2A. Associated with this action is a vector of private bene¯ts, fb ( a) i g I m r, to the manager and each of the owners. There is a private e®ort cost for the manager of choosing action aequal to ( P i 2 I m b ( a) i ) 2 r. 4 Privatebene¯tsarereceivedbytheagentsatdate3ifandonlyifthe manager is still present in the corporation. Date 2 1/2 The manager can be replaced with an alternative manager at any point after date 2. The alternative manager cannot do anything except from canceling the action chosen by the previous management. It costs kr, 0·k<1, to replace the manager and the decision has to be backed by a majority, which size is stipulated in the corporate charter, of the owners. Date 3 If the manager is not replaced, then the ex post value of the ¯rm, given action a,is ( 1¡P i 2 I m b ( a) i )r. The ex post value is paid out in dividend to all owners. In addition, the owners and the manager receive their private bene¯t, b ( a) i r; i 2I m . If the manager is replaced, the ex post value of the ¯rm is ( 1¡k)r which is paid out in dividend to the owners. Assumption 1. Assume Ais so large that any non-negative distribution of private bene¯ts is feasible, i.e. the manager chooses b2R # I m + . Assumption 1 implies we can suppress the action, a, and instead assume the manager chooses a distribution of non-contractible private bene¯ts. De- ¯ne the aggregate level of diversion as ¹ b´P i 2 I m b i . How is the manager selected? We can distinguish between at least three types of ¯rms: ( a) Some ¯rms will need a professional manager with some speci¯c skills the investors do not possess, i.e. these ¯rms hire an outside manager. ( b) In many ¯rms the founder will keep on operating the ¯rm after having sold the bulk of the ¯rm to outside owners. ( c) In other ¯rms, the new owners will go together and pick a manager among them self. The focus in the present analysis is on how investor protection a®ects e±ciency in 5 corporations and not on how management is elected. 1 We therefore simply assume that the manager is in place at date 2. There are many quali¯ed agents who are able to manage the ¯rm implying that the reservation wage is competed down to zero. If the manager is ¯red she receives also zero utility from running the ¯rm, but she keeps any ownership stake she possesses. 3 Investor protection when ¯rm value is certain In this section, we characterize the equilibrium of the model when the ¯rm value, r, is certain and known to all agents. We are interested in the consequences of having di®erent majority requirements on the amount of diversion in the model, on the distribution of private rent among owners and manager, on e±ciency and on the probability of having a dead-lock, de¯ned as a situation where the manager is replaced. Let ºbe the amount of votes necessary to replace the manager. For instance º= 50 pct. is a simple majority rule and º= 10 pct. means that any group of shareholders that possess at least 10 pct. of the outstanding votes can ¯re the management. For any set A2I m ,denotec ( A)=P i 2 A c ( A)and v(A)=P i 2 A v(A) as the amount of cash °ow (respective votes) that group Apossesses. De¯ne 5(v;º)asthefamilyofstrong coalitions of owners, i.e. the family of sets of owners which support is su±cient to keep the manager in place, i.e. P i 2 A [f m g v i >1¡ºpct. 8A25 ( v;º):A strong coalition is thus an element of 5 ( v;º). Furthermore, let 4 ( v;º)´fA25 ( v;º): 1 This is analyzed in our related work on close corporations ( see Bennedsen and Wolfenzon 1998 ) . The model presented here can be thought of as an incomplete contracting version of our previous model of a close corporation. The incomplete contracting framework is more suitable to analyze the topic of investor protection and in addition it avoids some of the assumptions of our previous model: ¯rst, the action taken by a single manager is a non-contractible action who cannot be in°uenced by anyone. Second, there is no board in the model, only owners and a manager, hence, the particular procedure to select the board ( voting rules, number of board members, etc. etc. ) is not an issue. Finally, there is not imposed any exogenous distribution rule of the diverted cash °ow among the shareholders. 6 Á i (¯ i ;±)2ffire; keepgis owner i's vote on replacement of the manager. The correct equilibrium to use is a perfect Baysian equilibrium, de¯ned as De¯nition 2 ( Equilibrium ) . ff¯(r);±(r)g;f¹ i (¯ i ;±)g i 2 I ¡ m ;fÁ i (¯ i ;±)g i 2 I ¡ m gis an equilibrium if and only if 1) f¯(r);±(r)gmaximizes the manager's expected utility given ff¹ i ( ¯ i ;±)g i 2 I ¡ m ;fÁ i ( ¯ i ;±)g i 2 I ¡ m g. 2 ) Á i ( ¯ i ;±)maximizes owner i's expected utility given ff¯ i ;±g;¹ i ( ¯ i ;±);fÁ j g j 2 I ¡ m nf i g g. 3 ) ¹ i ( ¯ i ;±)is updated according to Bayes rule for all i. We analyze the model of the previous section under some simplifying assumptions. Assumption 2. 1 ) Thereisnodeadweightlossofdiversion. 2 ) The manager is an outside wealth constrained manager. 3 ) Ownership is distributed according to a one-share-one-vote assumption. Part 1) simpli¯es exposition. Notice, there are still two kinds of e±ciency cost left in the model, namely the rent left to the wealth constrained manager and the replacement cost when the manager is ¯red. Part 2) reduces notation in the following. Part 3) makes life easier and can be motivated by the optimality of one-share-one-vote in the case where there is no ¯ring cost and an outside manager. We say that the equilibrium is a separating equilibrium if it satis¯es De¯nition 2, all agents strategies are pure and if either ± ( r)6=± ( r)or ¯ i ( r)6=¯ i ( r)forsomei2I. If all agents strategies are pure and ± ( r)=± ( r) and ¯ i ( r)=¯ i ( r) for all i2I, then the equilibrium de¯ned in De¯nition 2 is denoted a pooling equilibrium. Furthermore, for analytical convenience, we solve for a symmetric equilibrium, where all owners in a given class are 13 treated equal, i.e. all majority owners (respective all minority owners) receive thesameamountofprivatebene¯ts. Lemma 1. The following constraints are necessary conditions for a symmetric pooling equilibrium: (1) M(¯;±)24(v;º); (2) ¯ i =08i2InM(¯;±); ( 3) ¯ i +±c i ¸ ( 1¡k)E 1 r ( ¯ i ;±)c i 8i2M ( ¯;±); ( 4) X i 2 I ¯ i +±· ( 1¡k)r; ( 5) ¯ m ( r)=r¡X i 2 I ¯ i ¡±¸0: Theorem 2. 1 ) r· 2 ¡ ( 1 ¡ k ) c ( M ¤ ) ( 1 ¡ k ) c ( M ¤ ) ris a necessary and su±cient condition for the existence of a pooling equilibrium. 2 ) Necessary conditions for the existence of a separating equilibrium without dead-locks are, ( a)P i 2 I ¯ i ( r)+± ( r)=P i 2 I ¯ i ( r)+± ( r); ( b)r< 2 ¡ ( 1 ¡ k ) c ( M ¤ ) ( 1 ¡ k ) c ( M ¤ ) r: The theorem shows that the set of separating equilibria without deadlocks is small and a prober subset of the set of pooling equilibria. The separating equilibria without deadlocks are supported by the owners always believe that the state is good whenever the manager does not take the equilibrium action and the manager in equilibrium is indi®erent between the two actions. Thus, we do not want to put to much emphasize on these equilibria. We proceed by characterizing the set of symmetric pooling equilibria. Lemma 1 tells us that such equilibria does not have deadlocks. The best symmetric equilibrium for the majority owners are the one where condition ( 5) in Lemma 2 binds, i.e. where there is zero rent left to the manager in the bad state of the world. 14 Corollary 1. The majority owners' prefered equilibrium is given by, ±=0; ¯ i =08i2InM ¤ ; ¯ i =minf1 c(M ¤ )r;(1¡k)rg8i2M ¤ ; ¯ m (r)=½maxf0; (1 ¡(1 ¡k)c(M ¤ ))rgif r=r; maxfr¡r;(1¡(1 ¡k)c(M ¤ ))rgif r=r; ¹ i ( ¯ i ;±)=1 28i2I: The best equilibria for the manager is the ones where the majority owner is indi®erent between ¯ring the manager or not, i.e. where condition ( 3) in Lemma 2 binds. Corollary 2. The manager's prefered equilibrium is given by, ±=0; ¯ i =08i2InM ¤ ; ¯ i = ( 1¡k)E 1 r ( ¯ i ;±)8i2M ¤ ; ¯ m ( r)=r¡ ( 1¡k)c ( M ¤ )E 1 r ( ¯ i ;±): ¹ i (¯ i ;±)=1 28i2I: We have drawn these solutions in Figure 2. The horizontal axis measures the di®erence in the value of the ¯rm in the two states of the world, which re°ects the degree of uncertainty in this model. The vertical axis shows the per share unit amount of rent to each majority owner. Since dividends are zero in the absence of any dead weight loss of diversion, ¯ i measures thereturnpersharetomajorityowneri. 2 The area between the solid line and the dashed line constitute the set of symmetric pooling equilibria in the model. 2 Notice, this is where the one-share-one-vote assumption simpli¯es the exposition. Alternatively, we could have de¯ned symmetric treatment of majority owners as the same amount of private bene¯t per unit of cash °ow. 15 Dead-locks Best eq. for maj. owners Best eq. for manager. A A’ B C C’ D’ Dr _ r _ ((1/c(M*))r_ (1-k)r_ (1/c(M*)(1-k))r_ ((2-c(M*)(1-k))/c(M*)(1-k))r_ _i, i in M* Figure 2: Amount of private bene¯t for majority owners in the best and worst symmetric pooling equilibrium. The solid line pictures the equilibrium prefered by the majority owners. The amount of rent to each owner depends on the amount of uncertainty. If ris less than Cthe manager pays out (1 ¡k)rin each state of the world implying that the majority owners together receive v(M ¤ )(1 ¡k)r.The manager herself receives r¡v ( M ¤ ) ( 1¡k)rin the good state and r¡v ( M ¤ ) ( 1¡ k)rin the bad state. In this case, the majority owners' return is as if there was only a good state in the world. Thus, the manager pays all the cost of having private information, she would be strictly better o® if the state of theworldwasobservable,sinceshewouldthenbeabletopaylessprivate bene¯t out to the majority owners in the bad state of the world. Uncertainty improves e±ciency in this equilibrium, since it reduces the rent the manager can extract to herself in the bad state of the world. At point C,r¡c ( M ¤ ) ( 1¡k)r= 0. Thus the wealth constrained manager 16 cannot pay more out in the bad state of the world. At this point the maximum rent per share in a symmetric equilibrium is achieved. For r2 ( C; D) the manager pays rout to the majority owners in both states of the world. This leaves the manager with more rent in the good state of the ¯rm. At point Downer i's expected value of the ¯rm is su±ciently high in equilibrium, such that she expects to bene¯t from ¯ring the manager. Hence, if r>Da symmetric pooling equilibrium is not sustainable anymore. Instead one of two types of dead-lock occurs: either there exists a mixed strategy equilibrium where the manager is ¯red with some positive probability; or, there is no equilibrium at all. In the ¯rst case the ¯rm's value decreases because the expected ¯ring cost is strictly positive. In the second case we have the decision vacuum often described in the legal literature ( see for example Easterbrook and Fischel 1991). 3 ThedashedlineinFigure2representsthemanager'spreferedequilibrium. In this equilibrium the majority owner's per share private bene¯t is kept down to where she is indi®erent between ¯ring the manager or not. Again, this equilibrium is sustainable up to point D, where the manager pays out all the ¯rm'svalueinthebadstateoftheworld.Theexpectedamountofrentleft to the manager is equal to the rent attained by the manager if the state of the world was observable and equal to 1 2 r+ 1 2 r. Uncertainty, therefore, does not improve nor decrease e±ciency in this equilibrium. In sum, if there is a limited amount of uncertainty, i.e. r<D, uncertainty as such is not bad for e±ciency reason, because it may force the manager with private information to pay out more dividend in the bad state of the world. However, if there is signi¯cant uncertainty, i.e. r>D,itgiveriseto costly dead-locks in the ¯rm. In this case uncertainty can decrease e±ciency. Figure 2 provides an interesting insight into what happens when the investor protection increases through requirements of super majority by increasing º. This is illustrated by the arrows in the ¯gure. An increase in º increases the amount of cash °ow internalized by any group of owners with 3 In the next iteration of the paper we intent to provide a characterization of the mixed strategy equilibria. 17 the least cash °ow property and this has two e®ects on the set of equilibria. First, it lowers the maximum rent per share a majority owner receives. This happens because the maximum rent is attained when the manager pays out all the ¯rm's rent in the bad state of the world and this value is not affected by the voting rule. However, since there are now more shares included in the majority, each share receives less rent. This e®ect is represented by the shift in the solid line from point Ato point A 0 . When there is little uncertainty, i.e. when r<C 0 , then increasing the majority requirements increases e±ciency without lowering any owner's rent even in the best equilibria for the owners. Therefore, in the limit when the uncertainty disappears, we get the same insight as in the previous section, namely that requiring supermajorities over management replacement increases welfare and increases the return to a group of previous minority shareholders' return, without lowering the return to any other group of shareholders. It is worth emphasizing, however, that the movement from Cto C 0 implies that the maximum level for each owner is attained at a lower level of uncertainty. The second e®ect is the reduction in the set of equilibria without deadlocks. This is represented by the shift from point Dto D 0 . A pooling equilibrium requires that the majority owners are over-compensated in the bad state such that the equilibrium compensation is larger than they expect to receive by ¯ring the manager. Hence, in the bad state, the manager uses some of the rent she exploits from the minority shareholder and distribute this to the majority shareholders. When there is an increase in the size of the cash °ow hold by any set of owners possessing the minimum cash °ow property, there is less share holding left to exploit and, therefore, less rent to distribute among a larger group of majority shareholders. Thus, the resource constraint in the bad state is more binding implying that dead-locks occur for a lower level of uncertainty. In these cases, an increase in the majority requirement lowers welfare, since we move from an equilibrium without dead-locks to a situation where either the manager is ¯red with a certain probability or there exists no equilibrium. From the founder's perspective, the bene¯t of increasing ºdepends on 18 which equilibrium the ¯rm ends up in. If there is little uncertainty about the ¯rm value, there is no cost of imposing a super-majority from the founder's perspective. However, the bene¯t may also be limited if the agents end up in the prefered equilibrium for the owners in the case of a simple majority. When there is signi¯cant uncertainty there is an increased cost through the increased likelihood of a costly dead-lock. If we compare these e®ects to the situation without uncertainty, it is worth emphasizing that increasing ºimproves welfare for sure in the absence of uncertainty, but that there may be a tradeo® between the increased likelihood of a costly deadlock and the decreased amount of diversion in the case with uncertainty. Furthermore, an increase in the majority requirement is against the manager's interest, because she has less opportunity of diverting cash °ow to her self. More surprisingly it may often also be against the interest of the existing majority owners, partly because it increases their chances of incurring costly dead-locks cost, partly because it decreases the bene¯t they mayhaveextractedfromthemanagerevenintheabsenceofdead-locks. 5 Transferability of shares Close corporations are characterized by having concentrated ownership and that owners frequently choose to restrict the transferability of shares. Legal scholars argue that restricting the transferability of shares can be a e®ective way to preserve the balance of power in a corporation ( seethequotefrom Clark ( 1986) in the introduction). In our model, free transferability of shares is costly. The reason is that, by trading shares to improve the balance of power in their favor, shareholders endupwithamajoritycoalitionthatconcentratesvotesbutnotcash°ows. From Theorem 1 we know that this reduces e±ciency in the corporation. Therefore, it is in the interest of the initial owner to restrict the transferability of shares. We provide an example that illustrates this point. For simplicity we assume that the replacement cost is zero and the manager is an wealth constrained outside manager. 19 Consider the following ownership structure: Votes CashFlow I 1 40% 40% I 2 35% 35% I 3 25% 25% From Theorem 1, the manager chooses shareholders 2 and 3 as the majority coalition and diverts 40 pct. of the cash °ow in the ¯rm. Hence, for the initial owner, the sum of the dead-weight cost and the cost of leaving rent to the future manager is 0:4. Now, if shares are tradable before the manager chooses his action, shareholder 1 can sell ( or even give away for free) one fourth of her shares to an outside investor. The new ownership structure becomes: Votes CashFlow I 1 30 % 30 % I 2 35 % 35 % I 3 25 % 25 % I 4 10 % 10 % Notice that the balance of power in the corporation has been altered and that now the majority coalition is formed by shareholders 1 and 3 with a cash °ow share of 55 pct. The manager now diverts 45 pct. of the resources in the corporation. Shareholder 1 is strictly better o®, since she has changed status from being an exploited minority shareholder to be an exploiting majority shareholder. Hence, free transferability allows the owners to dispose cash °ows. By doing this they become more attractive partners to participate in the majority coalition. The sum of the dead-weight cost and the cost of leavingrenttothefuturemanagerisnowincreasedto0:45. Hence, the ability of shareholder 1 to sell her cash °ow is bad for the initial owner. Therefore, as legal scholars suggest, a restriction on free transferability is in the interest of the initial owner since it preserves the balance of power in the ¯rm. Obviously, we need to consider the equilibrium behavior of the three shareholders, but we conjecture that this will only make matters worse. In the case where there are no restrictions on how cash °ow can be sold, it is 20 not hard to construct an example where the only equilibrium is one where all owners sell all their cash °ow implying that the manager diverts everything. In a more realistic case where cash °ow only can be sold bundled to votes according to a one-share-one-vote rule, the lower bound of the cash °ow possessed by any majority coalition is 50 pct. 6 Conclusion The distribution of ownership determines the allocation of power in a corporation, i.e. determines how di®erent classes of owners form. Furthermore, concentration of ownership creates a con°ict between controlling owners and management on one side and non-controlling owners on the other side. In the presence of this con°ict we have studied how various forms of investor protection a®ect the performance of a corporation with concentrated ownership and the return di®erent classes of owners receive on their investment. 21 Appendix Proof of Theorem 1 Proof. First we prove the following necessary conditions for ff b;d g ; f Á i g i 2 I ¡ m g is a subgame perfect equilibrium: Lemma 2. 1 ) M ( b;d ) 2 S ( v ;º ) , i.e. the manager is not ¯red ex-post. 2 ) b i =0 8 i 2 I n M ( b;d ) : 3 ) i 2 M ( b;d ) ) b i =max f 0 ; ( 1 ¡ k ¡ d ) c i g 4 ) M ( b;d ) ) = Arg min A 2 S ( v ;º ) c ( A ) ´ M ¤ , i.e. the selected majority has the minimum cash °ow property. Proof. Part 1 ) The maximum utility the manager can attain by being ¯red is c m . By choosing b m = k and d =1 ¡ k the manager is not ¯red, since M ( b;d ) = I ¡ m , and the manager's utility is k + c m ( 1 ¡ k ) ¸ c m . Part 2 ) Assume not, i.e. there exists an i s.t. b i + c i d< ( 1 ¡ k ) c i and b i > 0. By choosing b , the manager is not replaced, since it is a solution. Consider action b 0 ;d 0 given by b 0 j = b j 8 j 2 I ¡ m nf i g , b 0 i = 0 and b 0 m = b m + b i . Notice d 0 = d and M ( b 0 ;d 0 ) = M ( b;d ) , hence the manager is not replaced when choosing b 0 . Furthermore, the manager is strictly better o®. A contradiction. Part 3 ) .If d> 1 ¡ k then c i d> ( 1 ¡ k ) c i implying that i 2 M ( b;d ) even if b i =0. Thus, by the same argument as in Part 2, b i > 0 is never a solution. Assume d< 1 ¡ k and b i > ( 1 ¡ k ¡ d ) c m for some i 2 M ( b;d ) . Then the manager can deviate by choosing ( b 0 ;d 0 ) where b 0 j = b j 8 j 2 I ¡ m nf i g , b 0 i = ( 1 ¡ k ¡ d ) c m and b 0 m = b m + b i ¡ b 0 i >b m . Part 4 ) . Assume not, i.e. c ( M ( b;d ) ) >c ( M ¤ ) . Consider action ( b 0 ;d 0 ) given by d 0 = d , b 0 i = max f 0 ; ( 1 ¡ k ¡ d ) c i g8 i 2 M ¤ , b 0 i =0 8 i 2 I ¡ m n M ¤ ,and b 0 m = b m ¡ P i 2 I ¡ m b 0 i + P i 2 I ¡ m b i = b m + ( 1 ¡ k ¡ d ) ( c ( M ( b;d ) ¡ c ( M ¤ ) ) >b m where we have used that ( b;d ) satis¯es Part 3 ) . Hence, the manager is strictly better o® by deviating, a contradiction. Using Lemma 1, we set up the manager's problem as, max f b;d g ( b m ¡ 1 2¹ b 2 + c m d ) r s.t. ( 1 ) b i =0 8 i 2 I ¡ m n M ¤ ( 2 ) b i =max f 0 ; ( 1 ¡ k ¡ d ) c i g8 i 2 M ( b;d ) ( 3 ) M ( b;d ) = M ¤ ( 4 ) 0 · d =1 ¡ ¹ b 22