American Economic Association

Financial Contracting Author(s): Oliver Hart Source: Journal of Economic Literature, Vol. 39, No. 4 (Dec., 2001), pp. 1079-1100 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/2698520 Accessed: 26/03/2009 14:50

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aea.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic Literature.

http://www.jstor.org Journal of Economic Literature Vol. XXXIX (December 2001) pp. 1079-1100

Financial Contracting

OLIVER HART'

1. Introduction that is, the ratio of the market value of its debt to the market value of its FINANCIAL CONTRACTINGmight be equity?2 described as the theory of what kinds of deals are made between finan- Questions like these have been the ciers and those who need financing. Let focus of much of the very large corpo- me motivate the subject matter of this rate finance literature that has devel- article with the following questions: oped over the last forty years, and they have also been studied in the more (A) Suppose an entrepreneur has an idea recent financial contracting literature. but no money and an investor has My plan is to summarize some of the money but no idea. There are gains older literature (section 2) and then -from trade, but will they be realized? move on to some more recent thinking If the idea (project) gets off the (sections 3 and 4). Section 2 will be de- ground, how will it be financed? liberately brief and will not do justice (B) We see companies around the world to the older literature. Fortunately, with a wide variety of financial struc- there are excellent surveys by Milton tures. Almost all companies have Harris and Artur Raviv (1991), Andrei owners (i.e., shareholders or equity Shleifer and Robert Vishny (1997), and holders). Some have other claimants, Luigi Zingales (2000) that the reader e.g., creditors, preferred sharehold- can consult to supplement what I have ers, etc. Why? Does this matter, for to say. (The latter two papers also example, for corporate efficiency or have insightful things to say about the investment behavior? What deter- financial contracting literature.) mines a company's debt-equity ratio, 1 HarvardUniversity and London School of Eco- 2. Established Views nomics. This article is a revised version of the of Financial Structure Nancy L. Schwartz Lecture delivered at North- western University in June 2000. I would like to The modern corporate finance litera- thank Philippe Aghion, Patrick Bolton, Bengt Holmstrom, John Moore, Andrei Shleifer, and ture starts with the famous Modigliani Jean Tirole for helpful comments; Fritz Foley for and Miller (MM) theorem (Franco excellent research assistance; and Ehud Kalai and Mort Kamien for inviting me to give the lecture. I 2 Post-war, the value of long-term debt of large would also like to acknowledge research support U.S. corporations has been about half the value of from the National Science Foundation through the equity. See Franklin Allen and Douglas Gale National Bureau of Economic Research. (2000). 1079 1080 Journal of Economic Literature, Vol. XXXIX (December 2001)

Modigliani and Merton Miller 1958). cently) used to illustrate MM with one This striking irrelevance result can be of Yogi Berra's famous (mis-)sayings: paraphrased as follows: "You better cut the pizza in four pieces because Modigliani-Miller (MM): In an ideal world, I'm not hungry enough to eat where there are no taxes, or incentive or in- six."5 Apart from the crumbs, this formation problems, the way a project or firm seems to sum up the proposition pretty is financed doesn't matter. well. MM, although an enormously impor- A simple (too simple) way to under- tant benchmark, does not seem to de- stand this result is the following. A proj- scribe the world very well. To give one ect can be represented by a stream of example of a problem, if MM were uncertain, future cash flows or (net) empirically accurate, we might expect revenues. Each future revenue is firms to use no debt or large amounts equivalent to some amount of cash to- of debt, or firms' debt-equity ratios day; the exact amount is obtained by to be pretty much random. However, applying a suitable discount factor (if Raghuram Rajan and Zingales (1995) the future revenue is uncertain, we find that similar, systematic factors de- might apply a higher discount factor). termine the debt-equity ratio of firms Now add all the cash equivalents to- in different countries. In fact, I think gether to obtain the total value of the that it would be fair to say that, since its project-its present value, V, say. conception, MM has not been seen as a Suppose the project costs an initial very good description of reality; thus, amount C. Then the project is worth much of the research agenda in corpo- undertaking if and only if V > C, that is, rate finance over the last forty years has if and only if it contributes positive net been concerned with trying to find value. Now we get to MM.3 The finan- "what's missing in MM." ciers of the project-who put up the Researchers have focused on two C-have to get their C back. They can principal missing ingredients: taxes and get it back in a variety of ways: they incentive problems (or asymmetric in- could be given a share s of future reve- formation). In both cases the idea is nues, where sV = C. Or they could get that, because of some "imperfections," some debt (riskless or risky) that has a V is not fixed and financial structure present value equal to C. But, however can affect its magnitude. they get it back, they must get C, and simple arithmetic tells us that the en- 2.1 Taxes trepreneur who sets up the project will get the remainder V - C. That is, from The simplest tax story is the follow- the entrepreneur's point of view (and ing. In many countries, the tax authori- from the financiers') the method of ties favor debt relative to equity: in par- financing doesn't matter.4 ticular, interest payments to creditors Merton Miller (who, sadly, died re- are shielded from the corporate income tax while dividends to shareholders are 3 Actually the result that the project should be undertaken if and only if V > C can also be not. As a result, it is efficient for a firm thought of as being part of MM. to pay out most of its profits in the 4 This informal justification of MM can easily be form of interest-this reduces its tax made rigorous for the case where the entrepre- neur and investors are risk neutral. If the parties bill and thus increases the total amount are risk averse, however, a more subtle, "home- available for shareholders and creditors made leverage" argument is required. See Joseph Stiglitz (1974). 5 See Berra (nd). Hart: Financial Contracting 1081 taken together. (Of course, this increase entrepreneur) who initially owns 100 in firm value is at the expense of society percent of a firm. This manager will since the treasury receives less tax choose not to consume perks since each revenue,) dollar of perks costs more than a dollar This simple tax story is too simple: it in market value (and as owner he bears suggests that we should see much the full cost). Now suppose the man- higher debt-equity ratios than we actu- ager needs to raise capital to expand the ally do. For this reason, it has been firm. One way to do this is to issue eq- elaborated on in various ways.6 But ex- uity to outside investors. However, this tensions of the theory, however in- will dilute the manager's stake-he will genious, do not seem to be adequate to now own less than 100 percent of the explain the data: for example, Rajan and firm. As a result, he will consume perks, Zingales (1995) find that, while taxes in- since the cost of these is borne at least fluence debt-equity ratios, other factors partially by others. As noted, this is in- are important too. efficient since total value (firm value In- fact, in the last few years the plus the value of perks) will fall. literature has focused on a different Alternatively, suppose the manager departure from MM: incentives. borrows to raise capital. At least for small levels of debt, this does not dilute 2.2 Incentive (Agency) Problems the manager's stake. The reason is that The most famous incentive paper in the the debt must be paid back for sure (it corporate finance li'terature is Michael is riskless), which means that, in a mar- Jensen and William Meckling (1976). ginal sense, the manager still owns 100 Jensen and Meckling argue that the value percent of the firm (his payoff is V - D, of the firm or project V is not fixed, as where D is the value of the debt). As a MM assume: rather it depends on the result, he bears the full cost of perks actions of management, specifically and will not take them. their consumption of "non-pecuniary So far it looks as if borrowing is an benefits" (perks). Perks refer to things efficient way to raise capital. However, like fancy offices, private jets, the easy Jensen and Meckling argue that borrow- life, etc. These benefits are attractive ing becomes costly when debt levels are to management but are of no interest large. The debt then becomes risky, since to shareholders-in fact they reduce there is a chance that it won't be repaid. firm value. Moreover, it is reason- At this point, the manager will have an able to assume that they are ineffi- incentive to gamble with the firm's as- cient in the sense that one dollar of sets, e.g., to engage in excessively risky perks reduces firm value by more than investments. The reason is that if an a dollar.7 excessively risky project succeeds, the Jensen and Meckling use these ideas firm's profits are high and the benefici- to develop a trade-off between debt and ary is the firm's owner-that is, the equity finance. Consider a manager (or manager himself (recall that he has 100 percent of the equity); whereas if the 6 See, e.g., Miller (1977). project fails, the firm's profits are low 7 This assumption makes sense since managers can typically consume perks only in quite narrow and the losers are the firm's creditors ways; that is, if unconstrained, they might prefer since the firm is bankrupt. to spend an extra dollar on their children's educa- According to Jensen and Meckling, tion rather than a fancy office, but the former would look suspicious whereas the latter can be the optimal debt-equity ratio or capital defended (to shareholders and society). structure for the firm is determined at 1082 Journal of Economic Literature, Vol. XXXIX (December 2001) the point where the marginal benefit of ence for' borrowing rather than issuing keeping the manager from taking perks shares disappears.8 is offset by the marginal cost of causing In other words, a question unan- risky behavior. swered by Jensen and Meckling's analy- The effects that Jensen and Meckling sis is: why use financial structure rather emphasize are clearly important. How- than an incentive scheme to solve what ever, their analysishas a theoretical short- is really just a standard agency problem? coming. The incentive problem that Before we move on, it is worth men- Jensen and Meckling focus on is what tioning another strand of the agency economists call an agency problem, i.e., literature that focuses on private in- a (potential) conflict of interest between formation possessed by managers rather an agent who takes an action (in this than managerial actions. (This part of case, the manager choosing the level of the literature corresponds to the ad- perks) and a principal who bears the con- verse selection version of the moral haz- sequences of that action (other share- ard problem studied by Jensen and holders or creditors). There is a large Meckling.) A leading example of this lit- economics literature on agency prob- erature is Stewart Myers and Nicholas lems, but the main finding from that Majluf (1984). Like Jensen and Meck- literature is that the best way to deal ling, Myers and Majluf consider a man- with them is to put the agent on an ager who needs capital to expand the optimal incentive scheme. firm. Myers and Majluf ignore perks, An illustration may be helpful. Sup- but suppose that the manager has bet- pose you (the principal) hire me (the ter information about the profitability agent) to sell silverware; my job is to of the existing firm, i.e., assets in place, drive around the suburbs, knock on than investors. In particular, imagine people's doors, and try to interest them that the manager knows that these are in knives and forks. You may be worried worth a lot, whereas investors do not. that I will sit in my car listening to rap Then, if the manager acts on behalf of music and not selling your product. One current shareholders (e.g., because he solution is to pay me a fixed amount per holds equity in the firm himself), he set of silverware I sell (a piece-rate) will not want to raise capital by issuing rather than a fixed wage per hour. (Or new shares. The reason is that the new you might use a combination of the shares will be sold at a discount relative two.) to their true value, which dilutes the Applying this logic to the present value of the current shareholders' stake. context leads to the conclusion that the Instead the manager will raise capital manager's salary should be geared to by issuing (riskless) debt. Riskless debt firm performance, that is, the manager will not sell at a discount-the firm will should be put on an incentive scheme, simply pay the market interest rate on I =f (V), where V is firm market value. it. Hence no dilution will take place. But this can be done independently of the firm's financial structure, that is, in- 8 This point is elaborated on in Philip Dybvig dependently of whether the manager is and Jaime Zender (1991). Paying the manager ac- cording to total market value V has the drawback a shareholder. (In the silverware exam- that the manager may have an incentive to invest ple, I did not have to become a share- in unprofitable projects in order to raise V. This holder of the silverware firm to work problem can be overcome by deducting the capital raised from V before assessing the manager's hard.) Moreover, given an optimal in- salary, i.e., paying the manager according to value centive scheme, the manager's prefer- net of investment cost. Hart: Financial Contracting 1083

Thus Myers and Majluf provide another 3. Financial Contracting Literature: reason why MM fails: if managers have Decision and Control Rights superior information, they will want to sell new securities whose return is in- We have seen that incentive (agency) sensitive to this information (riskless problems alone do not yield a very sat- debt being the most insensitive security isfactory theory of financial structure. of all). The recent financial contracting litera- Myers-Majluf are surely right that ture (developed in the last fifteen years private information is an important de- or so) adds a new ingredient to the terminant of financial structure, and the stew: decision (control) rights.10 effect that they identify appears to be This literature takes as its starting empirically significant. However, their point the idea that the relationship analysis suffers from the same theoreti- between an entrepreneur (or manager) cal weakness as Jensen-Meckling. Fi- and investors is dynamic rather than nancial structure matters only because static. As the relationship develops over managers are (implicitly) on a particular time, eventualities arise that could not kind of incentive scheme. Specifically, easily have been foreseen or planned Myers and Majluf assume that man- for in an initial deal or contract be- agers act on behalf of current share- tween the parties. For example, how holders, e.g., because they hold equity many people in 1980 could have antici- themselves. But things don't have to be pated the fall of the Soviet Union or the this way. Suppose managers were paid a rise of the internet in the 1990s? In an fraction of the firm's total market value ideal world, a contract between a com- V. Then managers wouldn't worry about puter manufacturer (IBM, say) and a selling new shares at a discount, since software producer (Microsoft), written any loss in current shareholder value is in 1980, would have included a contin- offset by a gain in new shareholder gency about what would happen if the value and managers are paid on the internet took off-or for that matter basis of the sum of the two. With this would have had a clause guarding incentive scheme, managers are happy against Microsoft becoming the domi- to expand by issuing new equity, and nant supplier of operating systems. In financial structure no longer matters.9 practice, writing such a contingent con- tract would have been impossible: the future was simply too unclear. 9There is in fact a strict advantage to putting managers on an incentive scheme that rewards Economists (and lawyers) use the term them according to total shareholder value, rather "incomplete" to refer to a contract that than current shareholder value. As Myers and does not lay out all the future contin- Majluf show, the latter scheme may cause manag- ers to turn down some profitable new projects, be- gencies. A key question that arises with cause the dilution effect on current shareholder respect to an incomplete contract is: how value will be so great that they prefer not to in- vest. This inefficiency is avoided if managers are 10This recent literature should be contrasted rewarded according to total shareholder value. with an earlier literature based on costly state veri- John Persons (1994) argues that an incentive fication; see Robert Townsend (1978) and Gale scheme where managers are paid a fraction of the and Martin Hellwig (1985). In this earlier litera- firm's total market value is not "renegotiation- ture, an optimal contract between an entrepreneur proof": the board of directors (acting on behalf of and investor was analyzed under the assumption current shareholders) will always revise it. How- that a firm's profitability is private information, ever, Persons does not explain why the board acts but that this information can be made public at a on behalf of current shareholders or why the cost. This earlier literature did not stress contrac- board is given the power to revise the managerial tual incompleteness (as defined below) or focus on incentive scheme. the role of decision (control) rights. 1084 Journal of Economic Literature, Vol. XXXIX (December 2001) are future decisions taken? That is, sets or push the firm into bankruptcy. given that an incomplete contract is si- Moreover, if the firm enters bank- lent about future eventualities, and given ruptcy, then creditors often acquire that important decisions must be taken some of the powers of owners. in response to these eventualities, how A rough summary is that shareholders will this be done? What decision-mak- have decision rights as long as the firm is ing process will be used? solvent, while creditors acquire decision It might be helpful to give some ex- rights in default states. amples. Consider a firm that has a long- It is worth emphasizing the differ- term supplier. Advances in technology ence between this perspective and that might make it sensible for the firm in- described in section 2. According to stead to buy its inputs on the internet. MM, the firm's cash flows are fixed and Who makes the decision to switch? equity and debt are characterized by Or take a biotech firm that is engaged the nature of their claims on these cash in trying to find a cure for diabetes. flows: debt has a fixed claim while The firm has been pursuing a particular equity gets the residual. In Jensen and direction, but new research suggests Meckling, the same is true except that that a different approach might be now the allocation of cash flow claims better. Who decides whether the firm can affect firm value through manage- should change strategy? rial incentives. In neither case do votes Other examples concern whether a or decision rights matter. In contrast, firm should undertake a new invest- in the financial contracting literature, ment, whether the firm's CEO should decision rights or votes are key, even be replaced, or whether the firm should though, of course, as we shall see, cash be closed down. flow rights matter a lot too. The financial contracting literature It is also worth noting that there is an takes the view that, although the con- important distinction between the kinds tracting parties cannot specify what of decisions we are talking about here and decisions should be made as a function the managerial actions we discu-ssed in of (impossible) hard-to-anticipate-and- the context of Jensen-Meckling. Mana- describe future contingencies, they can gerial actions, e.g., the level of perks or choose a decision-making process in ad- effort, are usually assumed to be non- vance. And one way they do this is transferable (or hard to transfer): only through their choice of financial struc- the manager can choose them. In con- ture. Take equity. One feature of most trast, decision rights are (more easily) equity is that it comes with votes. That is, transferable: e.g., the decision about equity-holders collectively have the right whether to replace the CEO, say, can to choose the board of directors, which be taken by one party (shareholders) or in turn has the (legal or formal) right by another party (creditors). Hence, a to make key decisions in the firm- key design question is: how should deci- specifically, the kinds of decisions sion rights be allocated in the initial described above. contract/deal between the parties? To In contrast, take debt. Creditors do this we now turn. not have the right to choose the board 3.1 The Allocation of Decision of directors or to take decisions in the Rights firm directly. However, they have other The financial contracting literature has rights. If a creditor is not repaid, she tended to focus on small entrepre- can seize or foreclose on the firm's as- neurial firms-rather than a publicly Hart: Financial Contracting 1085 traded company or corporation-and sume that the project yields cash flows we will do this too for the moment. To in the amount of $V and private bene- make things very simple, consider a sin- fits in the amount of $B. Private bene- gle entrepreneur, a single investor, and fits are similar to the non-pecuniary a single project. The question is, how benefits discussed in Jensen-Meckling; should the right to make future deci- although private benefits may represent sions be allocated between the entre- things like psychic value, we suppose preneur and the investor? Who should that they have a cash equivalent, i.e., they have the right to replace the CEO or can be measured in dollars. The inves- terminate the project? tor is interested only in cash flows, while In order to answer this question, we the entrepreneuris interested in both cash obviously need a theory of why the flows and private benefits. These differ- allocation of decision-making authority ent interests create a potential conflict matters. Various possibilities have been between the entrepreneur and investor. advanced. One approach is based on the Since private benefits (like decision idea that decision rights are important rights) are very important in what follows, for influencing asset- or relationship- it may help to illustrate them in the cur- specific investments. Suppose individ- rent context. Consider an entrepreneur ual i is considering whether to invest who has developed an idea for a project. resources in learning about how to The entrepreneur is likely to get some make the project more profitable. If he personal satisfaction from working on the controls the project, and has a good idea, project, or from the project succeeding, he can implement this idea without in- that is over and above any cash flows terference from anyone else. This gives received. Also, if the project succeeds, the him a strong incentive to have an idea. entrepreneur's reputation is enhanced On the other hand, if someone else con- and he will do better in future deals. trols the project, i will have to get per- Personal satisfaction and reputational mission from this other person and may enhancement are both examples of pri- have to share the fruits of his idea with vate benefits since they are enjoyed by them; this will dilute his incentives. the entrepreneur but not the investor. The above approach has been used in Some private benefits are less in- the theory of the firmll but has been nocuous. Someone who controls a proj- employed less in the financial contract- ect can decide who will work on the ing literature. Instead, in this latter lit- project; the controller may choose to erature, researchers have focused on how appoint relatives or friends to key posi- the allocation of control rights affects tions even though they are incompetent the trade-off between cash flows and ("patronage"). The controller may also private benefits once the relationship is be able to divert money from the proj- underway. ect, e.g., he can set up other firms that The best-known paper adopting this he has an ownership interest in, and approach is Philippe Aghion and Patrick choose the terms of trade between the Bolton (1992).12 Aghion and Bolton as- project and these firms to suck cash out of the project. Patronage and diversion 11 See Sanford Grossman and Hart (1986), Hart are also examples of private benefits. and John Moore (1990), and Hart (1995). As noted, the existence of private 12 For related work, see Bolton and David benefits introduces a potential conflict Scharfstein (1990), Douglas W. Diamond (1991), Hart and Moore (1994) and Hart and Moore of interest between the entrepreneur (1998). and the investor. How is this conflict 1086 Journal of Economic Literature, Vol. XXXIX (December 2001) resolved? The answer is that this de- This outcome will be achieved if the in- pends to a large extent on who has vestor makes the decision since she the right to make decisions once the puts no weight on private benefits, but relationship is underway. not if the entrepreneur does (given his To understand this, consider a simple stake of 10 percent, he gains only $20 case where the entrepreneur is allocated from avoiding losses, but loses his full a fraction 0 of the project cash flows and private benefit of $100). the investor receives the remaining (1 - On the other hand, suppose that the 0). Suppose that the project is set up at losses from continuation are $80 rather date 0 and all decisions are taken and than $200. Now it is efficient to con- benefits earned at date 1. The date 1 tinue the project, and this time effi- objective functions of the entrepreneur ciency will be achieved if the entre- and investor are then as follows: preneur has decision-making authority, but not if the investor does (since the Entrepreneur: Max B + OV, investor is concerned only with loss Investor: Max (1 - O)V_ Max V avoidance) 13

It is also useful to write down the ob- 13We have not considered renegotiation. Sup- jective of a planner who is concerned pose the losses from continuation are $200. We with social (or Pareto) efficiency. In a saw that if the entrepreneur has control at date 1, he will keep the firm going even though this is first-best world where lump sum distri- inefficient. However, one thing the investor could butions are possible the planner would do is to offer the entrepreneur a payment in re- maximize the sum of the entrepreneur turn for closing the firm down. The entrepreneur requires at least $80 to make this worthwhile and and investor's payoffs (since both are the investor is prepared to offer up to $180-so measured in money), i.e., social surplus, presumably something in this range will be agreed B +V. upon. Similarly, if the losses from continuation are $80, and the investor has control, the entrepre- Social Planner: Max B + V. neur-if he has the money-could pay the inves- tor an amount between $72 and $92 to persuade her not to close the firm down. It is clear that these three objective In fact, in a world of perfect renegotiation, the functions are generally distinct. This famous Coase theorem tells us that the allocation suggests that it will indeed matter of decision rights doesn't affect the date 1 out- come at all: the parties will always arrive at the whether the entrepreneur or the inves- efficient outcome through bargaining. However, in tor makes ex post decisions. (The planner the present context, there is an important impedi- is a mere construct and so will not have ment to renegotiation: the fact that the entrepre- neur is wealth-constrained. (Presumably this is decision-making authority!) why the entrepreneur approached the investor in For example, suppose the only deci- the first place. If he was not wealth-constrained, sion to be made concerns whether the he could have financed the project himself.) Thus, while it may be relatively easy for the investor to project should be terminated or contin- bribe the entrepreneur to make a concession when ued (at date 1). Assume that E's private the entrepreneur has control, it is harder for the benefit from continuation is $100, but entrepreneur to bribe the investor to make a con- cession when the investor has control. In fact we that $200 in resources can be saved if have implicitly assumed that the entrepreneur has the project is terminated now rather no wealth, so that the only item of value he can than later. Also assume 0 = .1. offer to give up is his fraction 0 of V; this may not be enough to achieve efficiency. Note that he can't From a social surplus or efficiency give up B directly because B is a nontransferable perspective, the project should be ter- private benefit. minated (the $200 loss exceeds the Since renegotiation complicates the basic story, without changing the fundamental message that $100 private benefit and social surplus the allocation of control matters, I will ignore it in is represented by the sum of these). what follows. Hart: Financial Contracting 1087

Consider the issue of contract design lead to the destruction of significant at date 0. The parties have two instru- private benefits since the investor puts ments at their disposal: the allocation of all the weight on cash flows. Note that cash flow rights, represented by 0, and this contract has a simple interpretation: the allocation of control rights. (For the entrepreneur is a paid employee- simplicity, we have assumed that the he has no formal authority and gets a parties share cash flows in a linear man- flat wage (actually zero!). ner, but nothing significant depends on The question is, where between these this-the investor could hold convert- two extremes does the optimal contract ible, preferred stock, for example.) For lie? simplicity assume that the entrepreneur There is one case where there is a makes a take-it-or-leave-it offer to the simple answer. Suppose that, whatever investor at the contract-signing stage. decision is taken at date 1, the project Suppose also that both parties are risk yields a cash flow that is at least C (dis- neutral. Then the entrepreneur will counted back to date 0). Then the in- choose the contract to maximize his vestor can be given riskless debt with expected payoff subject to the inves- value C and the entrepreneur can be al- tor breaking even, i.e., recovering her located all the equity, i.e., he is the re- investment cost C (on average). sidual income claimant and has all the A simplification can be made. Since decision rights. This contract is feasible the investor's gross expected return is because the investor breaks even and fixed at C, an optimal contract will also optimal because there is no inefficiency: maximize the sum of the entrepreneur the entrepreneur maximizes B + V.14 and investor's payoffs, i.e., (expected) Unfortunately, in a world of uncer- social surplus, B + V, subject to the tainty, it is unlikely that the project investor breaking even. It follows that, cash flows will be large enough to sup- given two contracts, both of which have port riskless debt of value C given any the investor breaking even, the one decision. In order to understand what is that generates greater expected social optimal then, imagine that the parties surplus is superior. can anticipate-and contract on-certain It is useful to consider two polar con- events at date 1 (they are verifiable).'5 tracts. At one extreme, suppose the en- An example of an event might be a situ- trepreneur has all the cash flow rights ation where the firm has low earnings (0 = 1) and all the decision rights. Then and its product is not selling; in another the entrepreneur's objective function event the opposite may be true-the and the social planner's coincide, which firm has high earnings and its product means that an efficient outcome is guar- is selling. anteed. Unfortunately, the investor gets none of her money back! Thus, this con- 14 Note the importance of the condition that the tract is not feasible. project yields at least C whatever decision is taken. For riskless debt to be optimal, it is not At the other extreme, suppose that enough to suppose that the project can always the investor has all the cash flow rights generate C ex post if some decision is taken; such (0 = 0) and all the decision-making a decision might involve project termination, say, and the destruction of significant private benefits. rights. This contract maximizes the in- In this case, it may be better to allocate decision vestor's payoff and so the investor will rights on an event-contingent basis, as described more than break even-or at least, if she below. 15 An event is a subset of the set of all possible doesn't, the project can never go ahead states of the world (i.e., all possible contingen- at all. Unfortunately, this contract may cies). 1088 Journal of Economic Literature, Vol. XXXIX (December 2001) The advantage of allocating cash flow second set, where the cut-off is chosen and control rights to one party or the so that the investor breaks even. other will typically differ across these How good a job does the Aghion- events. For example, in one event it Bolton model do in explaining the fea- may be the case that a ruthless strategy tures of real-world financial contracts? of value (cash flow) maximization leads An interesting recent paper by Steven to an approximately efficient outcome N. Kaplan and Per Stromberg (2001) because private benefits aren't very argues that a good place to look is the important. Recall the example where venture capital sector (see also Paul closing the firm down saved $L and Gompers 1997).16 Venture capitalists wasted a private benefit of $100. If L = are private providers of equity capital 200, then indeed value maximization for young growth-oriented firms (high- generates an efficient outcome. tech start-ups). Although venture capi- On the other hand, in other events, talists often represent several large rich private benefits may be relatively more individuals or institutions, they corre- important, and value maximization may spond quite well to the single investor cause a significant loss of social surplus. of the Aghion-Bolton model. Similarly, This would be the case in the same the founder or founders of a start-up example if L = 80. company can be represented without Aghion and Bolton show that the too much of a stretch by a single entre- investor should have control-and preneur. The distinguishing feature of cash flow rights-in the first kind of venture capital deals is that the major event, and the entrepreneur should participants have a close relationship have control-and also possibly cash and are few in number. flow rights-in the second kind of Kaplan-Stromberg study 213 venture event. The reason is that giving the capital (VC) investments in 119 port- investor control and cash flow rights folio companies (firms) by fourteen VC in the first kind of event generates an partnerships. Most of these firms are in approximately efficient (social surplus the information technology and soft- maximizing) outcome and makes it eas- ware sectors, with a smaller number ier to satisfy the investor's break-even being in telecommunications. Kaplan- constraint; while giving the entrepre- Stromberg's main findings (from our neur control in the second kind of event point of view) are the following: prevents inefficiency and hence is de- sirable as long as it is consistent with (1) VC financings allow the parties to al- satisfying the investor's break-even con- locate separately cash flow rights, vot- straint. (Giving the entrepreneur cash ing rights, board rights, liquidation flow rights in the second kind of event rights, and other control rights. may be useful to bring the entrepre- (2) Cash flow rights, voting rights, con- neur's objective function in line with trol rights, and future financings are the social objective function.) frequently contingent on observable A very rough summary of the Aghion- measures of financial and nonfinan- Bolton model is thus the following. If cial performance. For instance, the we rank events from those where ruth- VCs may obtain voting control or less value maximization is least ineffi- board control from the entrepreneur cient to those where it is most inefficient, if the firm's EBIT-earnings before then the investor should have control in 16 For related work on biotechnology alliances, the first set and the entrepreneur in the see Joshua Lerner and Robert Merges (1998). Hart: Financial Contracting 1089

interest and taxes-falls below a pre- likely that the project cash flows can specified level or if the firm's net support something like riskless debt, in worth falls below a threshold. Also, which case an efficient outcome can be the entrepreneur may obtain more achieved by giving all the control rights cash flow rights if the firm receives and residual income rights to the entre- approval of a product by the Food preneur. Under these conditions there and Drug Administration (FDA) or is is no gain-and there can be a consider- granted a patent. able loss-from allocating control rights (3) If the firm performs poorly, the VCs to the investor. obtain full control. As firm perfor- Interestingly, there is one striking mance improves, the entrepreneur finding of the Kaplan-Stromberg study retains/obtains more control rights. that, although consistent with the Aghion- If the firm performs very well, the Bolton model, does not necessarily fol- VCs retain their cash flow rights, but low from it. This is that control rights relinquish most of their control and and cash flow rights shift to the VC if liquidation rights. The entrepre- the firm does poorly (number 3 in the neur's cash flow rights also increase above list). This makes perfect sense if with firm performance. we can identify poor performance with (4) VCs have less control in late rounds an event where ruthless value maximi- of financing (i.e., when the project is zation leads to an approximatelyefficient close to completion). outcome, e.g., because private benefits aren't very important relative to cash At a broad level, these findings fit flows.17 And indeed, this is quite plausi- very well with the Aghion-Bolton ble, in the sense that in bad events it model. First, as that model emphasizes, may be efficient that the project be ter- cash flow rights and control (decision) minated or the entrepreneur removed as rights are independent instruments, and CEO, and this is exactly what a ruthless indeed they are used independently: value maximizer would do. someone may be allocated significant However, things do not have to be cash flow rights without significant con- this way. It could be that ruthless value trol rights and vice versa. (To put it maximization leads to an efficient out- another way, there can be a substantial come in good events. For example, deviation from one share-one vote.) imagine that, if a start-up is very suc- Second, as the Aghion-Bolton model cessful, the founding entrepreneur is no predicts, to the extent that different longer the best person to run it, e.g., events can be identified, the allocation because his creativity gets in the way of of cash flow rights and control rights the professional approach to manage- will depend on these; here the events ment that is now desirable. If the losses correspond to performance as measured from keeping the entrepreneur on are by such things as earnings, net worth, high enough, then it is efficient to re- or product functionality (FDA or patent place him. However, the entrepreneur approval). Third, the fact that VCs have may resist replacement given his private fewer control rights in late financings benefit. Under these conditions, the can be understood as follows. In late only way to obtain an efficient outcome financings, a firm requires less cash is to put control in the hands of the VC. relative to future profitability, i.e., the investment cost C is, in effect, lower. 17 For a set of conditions guaranteeing this, see As we have seen, this makes it more Hart and Moore (1998). 1090 journal of Economic Literature, Vol. XXXIX (December 2001) In other words, this is a case where outside investor is costly (that is, the in- the VC should have control if the firm vestor has to expend time or resources performs well, since it is in good events to exercise control). Before we get into that cash flows are important relative to the details of the analysis, it is worth private benefits. emphasizing that neither the agency As noted, Kaplan-Stromberg do not approach of section 2 nor the control find this effect in the data, but the rights model of section 3 bears directly question is why? Possibly the answer is on this question. The agency approach, that the Aghion-Bolton model ignores as we have already argued, is really a an important variable: effort. That is, in theory of optimal incentive schemes reality, private benefits B and cash rather than capital structure; while the flows V are a function of ex ante effort control rights model helps to explain as well as ex post decisions. An entre- the optimal allocation of control be- preneur may have little incentive to tween insiders and outsiders, but not work hard to ensure that a good event why, given a particular level of control occurs, i.e., V is high, if his reward is to by insiders (in this case, zero), outsiders be replaced by a ruthless investor.18 In hold heterogeneous claims, i.e., some are other words, ex ante effort considera- shareholders while others are creditors. tions may explain why, empirically, con- In fact, one's first thought would be trol shifts to the investor in bad rather that diversity is bad since it creates than good events. conflicts of interest between different investors. Moreover, it is not clear 4. Costly Intervention and the Diversity why management should be affected by of Outside Claims diversity: why does it matter to them that in good states of the world share- In section 3, we discussed how con- holders have control, while in bad trol should be divided between an in- states creditors have control (given that sider (the entrepreneur) and an out- management never has control)? sider (the investor). However, in many One approach to the diversity issue is in countries like the large companies based on the existence of collective ac- United States, the United Kingdom, or tion problems. Imagine a large company as the Japan, insiders, represented by that has many (relatively small) share- of or do board directors management, holders. Then each shareholder faces the not have (voting) control in any state following well-known free-rider problem: rests with of the world. Rather control if the shareholder does something to investors. dispersed outside Moreover, improve the quality of management, those outsiders hold diverse claims: then the benefits will be enjoyed by all some are and others are shareholders shareholders. Unless the shareholder is creditors. altruistic, she will ignore this beneficial In this we discuss what section may impact on other shareholders and so for the of out- be responsible diversity will under-invest in the activity of moni- side claims. We will argue that diversity toring or improving management. For understood as part of an optimal can be example, an individual shareholder will when intervention an mechanism by not devote time and resources to per- 18 The entrepreneur may also have an incentive suading other shareholders to vote to to manipulate the accounts ex post, to make a replace an incompetent board of direc- good event look like a bad one, if he is likely to be replaced in a good event (e.g., he could throw tors. As a result, the management of a away money). company with many shareholders will Hart: Financial Contracting 1091 be under little pressure to perform well. lems is not entirely satisfactory for two (The threat of a hostile takeover bid can reasons. First, the existence of these overcome the shareholder passivity prob- collective action problems is assumed, lem to some extent, but, for all sorts of not derived: in particular, it is supposed reasons, is unlikely to eliminate it.) that shareholders face these problems In contrast, individual creditors can while creditors don't. However, things in principle obtain the full benefits of don't have to be this way. One could their actions for themselves and thus do imagine a company that sets itself up not face the same kind of free-rider so that each shareholder has the right problem (at least outside bankruptcy). to liquidate a fraction of the company's Suppose a creditor's debt is not repaid. assets unilaterally-in fact we see just Then she can seize some of the firm's such an arrangement with open-end mu- assets if her debt is secured; while if tual funds. Equally, one could imagine her debt is unsecured she can obtain a that creditors are required to act by a judgment against the firm and have a majority vote to seize assets or push a sheriff sell off some of the firm's assets. firm into bankruptcy. If most compa- She does not require other creditors to nies choose not to operate this way, we act. In fact, it is better for her if they do not, need to explain this; we shouldn't just since there are then more assets to seize! take it as given. So creditors impose discipline on Second, most collective action mod- management in a way that shareholders els of the trade-off between debt and do not. Specifically, whereas a manager equity assume that shareholders are who faces a large number of small share- completely passive. However, this view holders is unlikely to be penalized sig- is hard to square with the fact that com- nificantly if he fails to deliver high panies routinely pay out cash to share- profit or pay out large dividends, a man- holders in the form of dividends and ager who faces a large number of small share repurchases. If managers face no creditors knows that he must repay his pressure from shareholders, one would debts or he will be in trouble: his assets expect them to retain all their earnings: will be seized or he will be forced into wouldn't they always be able to find bankruptcy (which is assumed to be something better to do with a dollar unpleasant for him). However, there is than to pay it out to shareholders? a trade-off: too much discipline can be There is a third problem with most bad. While some debt is good in order collective action models of debt that is to force management to reduce slack, also worth mentioning. In these models too much debt is bad because it can it is typically the case that debt matters lead to the bankruptcy and liquidation only if a firm is close to bankruptcy. of good companies, and can prevent The reason is that, if not, then the firm management from financing profitable can pay off its current debts by borrow- new projects. Various papers have ex- ing against future income, i.e., current plored this trade-off and have used it to debt levels do not constrain manage- derive the optimal debt-equity ratio for ment. However, the idea that debt mat- a company.19 ters only if a firm is in extreme financial The view that financial diversity oc- distress does not seem very plausible. curs because of collective action prob- In recent years, Erik Berglof and Ernst-Ludwig von Thadden (1994) and 19 Representative contributions are Grossman and Hart (1982), Jensen (1986), Myers (1977), Mathias Dewatripont and Jean Tirole Rene Stulz (1990), and Hart and Moore (1995). (1994) have explored an alternative 1092 Journal of Economic Literature, Vol. XXXIX (December 2001) approach to diversity that proceeds more ity by buttressing the firm's financial from first principles. The basic idea be- position. (The model below focuses on hind these papers is that diversity is protection against future calamities good not because of the existence of rather than empire-building.)20 collective action problems, but rather Suppose initially that the firm has no because diversity changes incentives. In debt, i.e., all investors are shareholders. particular, suppose that a company has Assume also that these shareholders have a single investor (or group of homo- control rights, do not face collective geneous investors)-say, a shareholder action problems, and so could intervene with 100 percent control rights. This to force the manager to disgorge some shareholder has the right and the ability of the "free cash flow." However, inter- to intervene at any time; but assume, in vention is costly, e.g., it requires the contrast to what has gone before, that expenditure of time or resources. Then intervention is costly. Then this investor the manager will pay out just enough may choose not to act because the costs cash-d, say-to stop the shareholders of intervention exceed the benefits. In from intervening, i.e., such that the in- contrast, if the company has several in- tervention cost equals the inefficiency vestors with heterogeneous claims, it is generated by reinvesting earnings rather likely that for at least one investor the than paying them out. benefits of intervention exceed the Now assume instead that the firm costs. If this investor also has the ability owes some money to short-term credi- to intervene, management will be under tors that exceeds d-call the amount p. pressure. The conclusion is that hetero- Suppose that creditors do not have any geneous claimants can put more pres- powers that shareholders do not, i.e., sure on management than homogeneous their cost of intervention is just the claimants when intervention is costly. same. The manager could announce that It will be useful to present a very he will pay the creditors less than what simple model-in the form of a nu- he owes them-say d. However, credi- merical example-that illustrates this tors are unlikely to accept this-they approach. The model is based on (pre- will choose to intervene. Why? The rea- liminary) joint work with Moore and son is that if they agree to accept d draws on the ideas of Jeffrey Zwiebel rather than p, then the residual amount (1996), as well as those of Berglof and p - d will at best be postponed and von Thadden (1994) and Dewatripont possibly even canceled (this is the and Tirole (1994). nature of a debt claim). In either case, Let me begin with a verbal descrip- if we allow for discounting and uncer- tion of the model. Consider a firm that tainty, creditors won't get much of the has some current earnings, and is also residual. In contrast, if they intervene expected to be profitable in the future. The manager or board of directors of 20 Retaining some cash to protect against future calamities may be at least partly in the interest of the firm will have to decide how much shareholders, of course. However, to the extent of the current earnings to pay out to in- that the manager obtains a private benefit from vestors. It is plausible that the manager continuation, he may retain excessive cash. The model focuses on this excessive element of cash has his own (selfish) reasons for not retention. Excessive cash retention was a promi- paying out as much as the sharehold- nent feature of Kirk Kerkorian'sbattle against the ers would like, e.g., he might want to Chrysler board in the 1990s; for another example, see the discussion of Japanese companies in the engage in empire-building activities or Financial Times of October 17, 2000 ("Takeover protect against a possible future calam- Specialist Tries a Different Tactic," p. 10). Hart: Financial Contracting 1093

t=O 1 2 3

Company set up Earningsy, = 50 Liquidityshock k EarningsY2 = 90 at cost 56 kuniform (if firm survives) on [0,200] Figure 1. now, they may be able to get all of their calamity (environmental, legal, etc.), the p (assuming that the firm's cash flow firm will have to come up with $k to plus asset value exceeds p). To this end, survive. Here k is a random variable they will even be prepared to destroy with a known distribution as of date 0; value, e.g., liquidate productive assets the realization of k becomes known (to or cut off funds for good investment the manager and investors) at date 2. projects. The point is that creditors do Finally, if it survives the liquidity shock, not care about the firm's future profit- the firm earns y2 at date 3, which is ability given that the beneficiaries of fu- paid out to investors. ture profitability are shareholders rather For simplicity, we will assume C = 56, than them. yi = 50, k is uniformly distributed on To put it very simply, shareholders [0,200], y2 = 90, and the intervention cost are soft because they are the residual F = 18. Also, investors are risk neutral, income claimants while creditors are and the market interest rate is zero. tough because they are not. The time-line is illustrated in figure Now to the details of the model (or ex- 1. To simplify matters, we suppose that ample). The model is a slightly more com- the manager's only interest is to maxi- plicated version of the Aghion-Bolton mize his chances of surviving to date 3, model of section 3, with the one impor- i.e., of overcoming the liquidity shock. tant difference being that intervention (In others words, he gets a fixed private is costly. There are four dates. At date 0 benefit from running the firm between the firm is set up at cost C. This amount dates 2 and 3 and he maximizes the must be raised from outside investors expected value of this benefit; he's since the manager has no funds of his uninterested in cash.) Given that the own. At date 1 earnings of yi are real- manager is uninterested in cash, it will ized. The' firm's manager (or board of never be efficient for him to hold any directors) then has a choice about how income claims, i.e., all equity (in the much of yl to pay out and how much to sense of residual income rights) will be retain (retentions are placed in the held by outside investors. firm's bank account(s), which pay the Denote by el the amount the man- going rate of interest). At this stage a ager pays out to investors at date 1 and controlling outside investor can inter- by i = y1 - el the amount he retains. It's vene to undo the manager's pay-out useful to start with two polar cases. decision-but this costs F. (F can be in- 4.1 Investor Optimum terpreted as the cost of learning where the firm'-sbank accounts are.) We sup- From the point of view of investors pose that the manager's motive for (who hold all the income claims), the retaining funds is that the firm will be firm should survive at date 2 if and only hit by a liquidity shock at date 2. What if k < 90. The reason is that this is the this means is that, because of s'ome rule that maximizes present value at 1094 Journal of Economic Literature, Vol. XXXIX (December 2001) date 2: the firm is worth saving only if it Given a zero pay-out, the firm will costs less to save than it is worth (90). survive if and only if k < 140 and, in this Moving backwards in time, we see case, investors will receive 140 - k as that this outcome can be achieved as a dividend. Thus the firm's (present) long as the firm pays out all its earnings value at date 0 will be at date 1. The point is that, at date 2, 140 the manager can borrow up to 90 VMO= 1/200 1(140 - k)dk= 49.21 against date 3 earnings; and this will en- 0 sure that he can survive liquidity shock (MO stands for manager optimum.) Note k if and only if k < 90. that VMO < 56, i.e., the manager opti- In other words, from the point of mum is not feasible: if the manager has view of investors, it is best to leave no total control of the firm and investors slack in the firm at date 2. Under these (rationally) expect the manager to retain conditions, the firm's (present) value at all the date 1 earnings, they will not date 0 is given by finance the firm in the first place. 90 4.3 Shareholder Control VIO= 50 + 1/200 (90 - k)dk= 70.25. 0 Obviously the manager wants to find (IO stands for investor optimum.) The a way to persuade investors to finance first term represents the date 1 earnings the firm. In order to do this, he must that are paid out, while the second term cede some control to outside investors. represents the expected going concern One possibility is to allocate all the value from date 1 onwards: note that the control rights to a single shareholder (or firm borrows k at date 2 whenever k < a group of homogeneous shareholders), 90, and pays 90 - k out as a dividend to so that this shareholder can intervene at investors at date 3. date 1. In fact, in the Aghion-Bolton Observe that Vio > 56, so that, if the model of section 3, this was the way to manager can commit to the investor op- put maximumpressure on the insider (the timum, the firm will be set up at date 0 entrepreneur) and to maximize the re- (there is enough value to compensate turn to outsiders. However, in that model the investors). we ignored any costs of intervening. In the present context, there is an 4.2 Manager Optimum intervention cost of 18. Suppose there Once the firm has been set up, the is a single shareholder with all the con- manager has a quite different goal from trol rights. At date 1, the manager pays that of the investors: he wants the firm out an amount el. At that stage, the to survive to date 3. This means that he 21We are making an implicit assumption here. wants to retain as much earnings at date If k > 140, the manager cannot save the firm and 1 as possible. Suppose he retains i. a question arises as to what happens to the date 1 earnings of 50. We take the view that the manager Then he can add this to the 90 he can engages in a partial rescue, specifically, he keeps borrow against date 3 earnings and sur- the firm going for a fraction X of the period be- vive any liquidity shock k such that k < tween dates 2 and 3, where 90 X + 50 = kk. That is, there are constant returns to scale with respect i + 90. Obviously the manager wants i to to time for X < 1, so that the cash injection re- be as big as possible, i.e., i = 50. That quired to overcome the liquidity shock is propor- is, if the manager is unconstrained (e.g., tional to the length of the period (k > 1 is as- sumed to be infeasible). Given this assumption, he has full control), he'll never pay out the date 1 earnings are totally dissipated, which is anything at date 1. why they don't appear in the formu a for VMO. Hart: Financial Contracting 1095 shareholder decides whether to inter- always be structured with the above fea- vene. Intervention means that she can tures and we are simply trying to show choose to pay out more funds if she likes that there is a way to get the firm (obviously she'll choose to pay out financed at date 0. everything), but she has to incur the Note also that the assumption that intervention cost 18. the creditor can be reimbursed does not The manager will pay out just enough introduce an asymmetry between the to make the shareholder indifferent be- creditor and the shareholder: the share- tween intervention and not. It is easy to holder is also in effect reimbursed for see that the equilibrium value of ei is intervening since, as the residual in- 10, i.e., i = 40. Given this the firm will come claimant, she owns everything. To be saved at date 2 if and only if k < 130 put it another way, if the shareholder and its (present) value at date 0 will be reimbursed herself formally, then she 130 would simply be transferring funds from one pocket to another. Vs = 10 + 1/200 J (130 - k)dk= 52.25. 0 The timing is now as follows. First, the Note that Vio - Vs = 18, which is the manager makes a payment to the intervention cost. In other words, it is creditor-call the amount p. Second, the creditor indeed the case that, with ei = 10, the decides whether or not to in- shareholder is just indifferent between tervene. Third, the manager makes a intervening and not. payment to the shareholder-call the Unfortunately,Vs < 56. In other words, amount d. Fourth, the shareholder total shareholder control is not enough decides whether to intervene. to get the firm financed at date 0! I claim that the equilibrium is for the manager to pay 20 to the creditor and 4.4 Shareholder and Creditor Control nothing to the shareholder and for nei- ther party to intervene. The reason is There is a way to get the project the following. If the manager pays p < financed: it is to include a short- 20 to the creditor, the creditor will term creditor as well as a controlling choose to intervene since she is entitled shareholder. to collect 20 - p + 18 (what she is owed Suppose there is a creditor who is plus her reimbursement for interven- owed 20 at date 1. If the creditor is not tion) and the firm has 50 - p in retained fully paid, she can choose to intervene earnings. Since the manager loses 38 in at a cost of 18 (just like the share- funds altogether if the creditor inter- holder). Assume that, if the creditor venes, it is better for the manager to intervenes, she can seize retained earn- pay the creditor the 20 she is owed. ings and pay herself the remaining Given that 20 is paid out to the credi- amount she is owed and be reimbursed tor, and 30 is retained, there is no need for her intervention costs, i.e., if she is to pay the shareholder anything. The owed x, she can seize x + 18. reason is that the firm is now saved if Finally, suppose that, if the creditor and only if k < 120 and so is worth decides not to intervene, her remaining 120 debts are canceled, i.e., she is entitled - = to nothing further at date 3. 1/200 (120 k)dk 36 0 Some of these assumptions are strong, but for our purposes this does at date 1. If the shareholder intervenes, not really matter. The debt claim can she can increase the firm's value to: 1096 Journal of Economic Literature, Vol. XXXIX (December 2001)

90 the right to intervene in default states. 30 + 1/200 J (90 - k)dk= 50.25 A second comment concerns renego- 0 tiation. One argument that can be lev- by seizing the remaining 30 in retained eled against the beneficial role we have earnings. But the gain from intervention found for short-term debt and equity is equals 14.25, which is less than the that we have ignored the possibility of intervention cost 18; i.e., shareholder renegotiation between the shareholder intervention will not occur. and the creditor. Suppose the manager So, under short-term debt and equity, does not repay the creditor the full 20 the firm will be worth she is owed, e.g., instead the manager 120 pays only 10. It is easy to see that it is not in the collective interest of the = Vsc 20 + 1/200 J(120 - k)dk= 56 shareholder and the creditor for the 0 creditor to intervene since the slack in at date 0, where the first-term represents the system (given by Vio - Vs) is only debt repayment and the second term 18, the cost of intervention. In fact, if represents the value of date 3 dividends. the creditor intervenes, she is entitled Since Vsc equals the date 0 investment to seize 28 and will gain 10 in net cost C, it follows that the firm can now terms, while the shareholder's return be financed! will be reduced from Some comments on the model are in 130 102 order. First, it is worth rehearsing the f - to f - for the of A 1/200 (130 k)dk 1/200 (102 k)dk, intuition benefits diversity. 0 0 single shareholder with full control rights is not tough enough on manage- i.e., by 16.24. (The shareholder and ment. The reason is that intervention is creditor's incremental payoffs sum to costly and, although intervening per- less than zero because the creditor seizes mits the seizure of retained earnings only 28 rather than the 50 she would that would otherwise b-e used for unpro- seize if the shareholder and creditor ductive purposes (from the point of could coordinate.) view of the shareholder), the gross rate Since the shareholder's loss from in- of return on these earnings is positive tervention exceeds the creditor's gain, (in low k states they will be paid out one might expect the shareholder to as a dividend at date 3). So the gains bribe the creditor not to intervene; i.e., from seizing the retained earnings are pay off her remaining debt. (One way to not that high. In contrast, a short-term do this is through a debt-equity swap.) creditor has a very different objective Of course, anticipating this, the man- function: any funds left in the firm ac- ager has no incentive to pay the credi- crue to the shareholder, not to her (i.e., tor the full 20 in the first place and the their gross rate of return is zero), and disciplinary role of debt evaporates. so her incentive to intervene is much Although this argument has some greater. force, it is far from decisive. What is re- Note that it is the combination of ally involved is a matter of timing. It is cash flow, rights and control rights true that, if the shareholder has a that is vital here: it is important both chance to bribe the creditor after the that the creditor has a claim that is manager has decided how much to pay capped above (which makes her in the creditor, then this reduces the man- effect impatient), and that she has ager's incentive to pay the creditor. Hart: Financial Contracting 1097

However, it is just, as plausible that the zero and use the proceeds to provide shareholder must decide whether to a large financial cushion at date 2. bribe the creditor before the manager Of course, arbitrarily large issues of makes his decision (i.e., the share- new equity are not in the interest of holder's move in the game comes first). the initial shareholder and so it makes In this case the shareholder won't bribe sense for the initial contract between the creditor, knowing that, if she does the manager and investors to limit not, the manager will prefer to pay the these. We have implicitly supposed that creditor the full twenty rather than face no new equity issues are allowed at date intervention. 1 (without shareholder permission), but In other words, under the timing a similar logic will work if a limited where the shareholder moves first, the number of new shares can be issued, model is intact. Note that there is no particularly if the time horizon extends inefficiency on the equilibrium path: to greater than four dates: the manager any funds the manager pays out to the then faces a trade-off between issuing creditor reduce slack and so the share- shares today to create more slack and holder is happy to see these funds being issuing them in the future when an- paid out. other liquidity shock may hit the firm. Third, one might ask whether the Note that a limit on new equity is quite manager could avoid the reduction in realistic-in practice, companies are usu- slack at date 1 by borrowing against fu- ally authorized to issue a certain num- ture earnings. It is indeed possible for ber of new shares, but once this limit has the manager to borrow against future been reached, the company must get earnings-since the firm is not close to permission from existing shareholders bankruptcy-but this only makes mat- to issue more (in our model, permission ters worse for him. To raise 20 at date 1 will not be given). the manager must promise d > 20 at A fourth comment concerns the costs date 3 since the firm is not certain to and benefits of debt. In the model the survive the date 2 liquidity shock. But benefit of debt is that it is a way for the this means that at date 2 the condition manager to commit to pay out free cash for the firm to survive becomes 140 - d flow, which enables the manager to get > k, since the manager arrives at date 2 the firm financed; while the cost of having already mortgaged d. As a result debt is that the manager has less slack the manager is less likely to survive to guard against liquidity shocks, which than if he doesn't borrow, where the reduces his private benefit. However, in condition for survival is 120 > k. extensions of the model there would be What the manager really wants to do other costs of debt. For example, if the is to issue new equity rather than bor- debt level at date 1 exceeded 50, then row. If the manager issues new equity the manager would have to pay this by of value 20, he can pay his date 1 debt borrowing against date 3 earnings. without incurring any future obliga- However, this introduces debt overhang tions: the condition for date 2 survival at date 2 (in the sense of Myers 1977): becomes k < 140 (given that he has 50 if d is owed at date 3, the condition for in retained earnings). In fact, the man- the firm to survive the liquidity shock ager can go further. Even if there is no becomes 90 - d > k, which means that short-term debt, he could issue huge the firm may fail to survive even when amounts of new equity at date 1-to the survival generates value for investors point where the value of initial equity is (survival generates value for investors 1098 Journal of Economic Literature, Vol. XXXIX (December 2001) whenever 90 > k). If the debt level be- which, as we have pointed out, many comes even higher, the manager may models of the debt-equity trade-off are be forced to liquidate part of the firm at not.22 This is clearly true if the amount date 1, i.e., sell off some key assets, owed at date 1, p, is less than 10, since which may again be inefficient. we saw that the manager will have to Finally, if the debt level becomes pay the shareholder 10 - p to stop her huge, the firm may declare bankruptcy, from intervening. At the optimum p = and-depending on the bankruptcy pro- 10 and the dividend is zero. However, if cedure-the firm may be turned over to we allow for uncertainty in yi, debt re- the creditor. However, this creates a payments and dividends can both occur problem: the creditor will become the at an optimum. Suppose yi can be high residual income claimant and will act or low. There is a class of cases such "soft" instead of "tough." So another that when yi is low p is paid to the cost of debt is that, if it becomes too creditor and there is no dividend and large, the disciplinary role of debt- when yi is high p is paid to the creditor depending as it does on the existence and on top of this a dividend is paid to of multiple claimants with conflicting the shareholder. interests-is lost. Whether this theory of dividends is This last observation has two interest- adequate to explain the data is another ing implications. First, it shows that in matter. The theory suggests that the this model moderate debt levels matter. payment of dividends should be quite If the debt level is very small, it does irregular and the amounts far from con- not affect what the manager pays out at stant. There is, of course, a large em- date 1 (this is true if p < 10); on the pirical literature that finds the opposite: other hand, if the debt level is very dividends are regular and smooth. How- large, the disciplinary role of debt is ever, the recent evidence suggests that lost. As noted earlier, this conclusion things are changing. Also, the idea that contrasts with that of most collective ac- dividends are the result of pressure tion models of debt in the literature, from shareholders receives support from where debt matters only when a firm is a recent cross-country comparison by close to bankruptcy. Second, for debt to Rafael La Porta et al. (2000). have a role it is essential that the firm cannot declare bankruptcy too easily. 5. Conclusions Specifically, a "no-fault" procedure that Let me conclude briefly. I have allows any firm to declare bankruptcy discussed how economists' views of and carry out an automatic debt-equity firms' financial structure decisions have swap will be counterproductive since evolved from treating firms' profit- the manager can avoid the disciplinary ability as given, to acknowledging that role of debt. This may provide some managerial actions affect profitability, justification for the idea that a firm that to recognizing that firm value depends wants to declare bankruptcy should on the allocation of decision or control have to convince a disinterested party, a rights. I have tried to show that the de- judge, say, that it cannot pay its debts. cision or control rights approach is use- As a last comment, we should note ful, even though it is at an early stage of that the above model is consistent with development, and that this approach the payment of dividends (or repur- has some empirical content: it can chase of shares-we have not distin- guished between the two)-something 22 Zwiebel (1996) is an exception. Hart: Financial Contracting 1099 throw light on the structure of venture Grossman, Sanford and Oliver Hart. 1982. "Corpo- rate Financial Structure and Managerial Incen- capital contracts and the reasons for the tives" in The Economics of Information and diversity of claims. Uncertainty. John J. McCall, ed. Chicago: U. I have been quite selective in the Chicago Press, pp. 107-40. . 1986. "The Costs and Benefits of Owner- topics I have covered. For instance, I ship: A Theory of Vertical and Lateral Integra- have not discussed research on why tion,"J. Polit. Econ. 94, pp. 691-719. companies in most countries other than Harris, Milton and Artur Raviv. 1991. "The Theory of Capital Structure,"J. Finance, 46, pp. 299- the United States, the United Kingdom, 355. and Japan often have controlling share- Hart, Oliver. 1995. Firms, Contracts, and Finan- holders and exhibit deviations from one cial Structure. Oxford: Oxford U. Press. Hart, Oliver and John Moore. 1990. "Property share-one vote. There has been some Rights and the Nature of the Firm," J. Polit. very interesting recent empirical and Econ. 98, pp. 1119-58. theoretical work on this topic,23 but a . 1994. "A Theory of Debt Based on the Inalienability of Human Capital," Quart. J. discussion of this will have to wait for Econ. 109, pp. 841-79. another paper. . 1995. "Debt and Seniority: An Analysis of the Role of Hard Claims in Constraining Management," Amer. Econ. Rev. 85, pp. 567- REFERENCES 85. Aghion, Philippe and Patrick Bolton. 1992. "An . 1998. "Default and Renegotiation: A Dy- 'Incomplete Contracts' Approach to Financial namic Model of Debt," Quart. J. Econ. 113, pp. Contracting,"Rev. Econ. Stud. 59, pp. 473-94. 1-41. Allen, Franklin and Douglas Gale. 2000. Compar- Jensen, Michael. 1986. "Agency Costs of Free ing Financial Systems. Cambridge, MA: MIT Cash Flow, Corporate Finance and Takeovers," Press. Amer. Econ. Rev. 76, pp. 323-29. Bebchuk, Lucian. 1999. "A Rent-Protection The- Jensen, Michael and William Meckling. 1976. ory of Corporate Ownership and Control," "Theory of the Firm: Managerial Behavior, working paper 7203, NBER. Agency Costs, and Ownership Structure," J. Berglof, Erik and Ernst-Ludwig von Thadden. Finan. Econ. 3, pp. 305-60. 1994. "Short-Termversus Long-Term Interests: Kaplan, Steven N. and Per Stromberg. 2001. "Fi- Capital Structure with Multiple Investors," nancial Contracting Theory Meets the Real Quart. J. Econ. 109, pp. 1055-84. World: An Empirical Analysis of Venture Capi- Berra, Yogi. "Famous Yogi Berra Quotes," online: tal Contracts,"U. Chicago Grad. School Bus. http://www.yogiberraclassic.org/. La Porta, Rafael; Florencio Lopez-de-Silanes, An- Bolton, Patrick and David Scharfstein. 1990. "A drei Shleifer and Robert Vishny. 1998. "Law Theory of Predation Based on Agency Problems and Finance," J.Polit. Econ. 6, pp. 1113-55. in Financial Contracting,"Amer. Econ. Rev. 80, . 2000. "Agency Problems and Dividend pp. 94-106. Policies around the World," J. Finance 55, pp. Dewatripont, Mathias and Jean Tirole. 1994. "A 1-33. Theory of Debt and Equity: Diversity of Securi- Lerner, Joshua and Robert P. Merges. 1998. "The ties and Manager-Shareholder Congruence," Control of Technology Alliances: An Empirical Quart. J. Econ. 109, pp. 1027-54. Analysis of the Biotechnology Industry,"J. Ind. Diamond, Douglas W. 1991. "Debt Maturity Econ. 46, pp. 125-56. Structure and Liquidity Risk," Quart. J. Econ. Miller, Merton. 1977. "Debt and Taxes," J. Fi- 106, pp. 709-37. nance 32, pp. 261-75. Dybvig, Philip H. and Jaime F. Zender. 1991. Modigliani, Franco and Merton H. Miller. 1958. "Capital Structure and Dividend Irrelevance "The Cost of Capital, Corporation Finance and with Asymmetric Information," Rev. Finan. the Theory of Investment," Amer. Econ. Rev. Stud. 4, pp. 201-19. 48, pp. 261-97. Gale, Doug as and MartinHellwig. 1985. "Incentive- Myers, Stewart. 1977. "Determinants of Corporate Compatible Debt Contracts: The One-Period Borrowing,"J. Finan. Econ. 5, pp. 147-75. Problem," Rev. Econ. Stud. 52, pp. 647-69. Myers, Stewart and Nicholas Majluf. 1984. "Cor- Gompers, Paul A. 1997. "Ownership and Control porate Financing and Investment Decisions in Entrepreneurial Firms: An Examination of when Firms Have Information that Investors Convertible Securities in Venture Capital In- Do Not Have," J. Finan. Econ. 13, pp. 187- vestments," HarvardU. Grad. School Bus. 221. Persons, John C. 1994. "Renegotiation and the Im- possibility of Optimal Investment," Rev. Finan. 23 See the many Papers by La Porta et al. (e.g., Stud. 7, pp. 419-49. La Porta et al. 1998); and Lucian Bebchuk (1999). Rajan, Raghuram G. and Luigi Zingales. 1995. 1100 Journal of Economic Literature, Vol. XXXIX (December 2001)

"What Do We Know about Capital Structure? Optimal Financing Policies," J. Finan. Econ. 26, Some Evidence from International Data," J. pp. 3-27. Finance 50, pp. 1421-60. Townsend, Robert. 1978. "Optimal Contracts and Shleifer, Andrei and Robert W. Vishny. 1997. "A Competitive Markets with Costly State Verifica- Survey of Corporate Governance," J. Finance tion,"J. Econ. Theory 21, pp. 265-93. 52, pp. 737-83. Zingales, Luigi. 2000. "In Search of New Founda- Stiglitz, Joseph E. 1974. "On the Irrelevance of tions,"J. Finance 55, pp. 1623-53. Corporate Financial Policy," Amer. Econ. Rev. Zwiebel, Jeffrey. 1996. "Dynamic Capital Struc- 64, pp. 851-66. ture under Managerial Entrenchment," Amer. Stulz, Rene. 1990. "Managerial Discretion and Econ. Rev. 86, pp. 1197-215.