|# / Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium IVIember Libraries http://www.archive.org/details/agencycostsinproOOacem DEWEY HB31 .M415 \/2^. 9s-i @ working paper department of economics A GENCY COSTS IN THE PROCESS OF DEVELOPMENT Daron Acemoglu Fabrizio Zilibotti 96-8 Feb. 1996 massachusetts institute of technology 50 memorial drive Cambridge, mass. 02139 J; AGENCY COSTS IN THE PROCESS OF DEVELOPMENT Daron Acemoglu Fabrizio Zilibotti 96-8 Feb. 1996 MASSAWU?'" m^ 13 1996 96-8 Agency Costs in the Process of Development Dciron Acemoglu Fabrizio Zilibotti* Massachusetts Institute Universitat of Technology Pompeu Fabra February 16, 1996 Abstract We analyze an economy where production is subject to moral hazard. The degree of the incentive (agency) costs introduced by the presence of moral hazard naturally depends on the information structure in the economy; cheaper is it is to induce correct incentives in a society which posesses better ex post information. The degree of ex post information projects and entrepreneurs in the economy; the depends on the number of more projects, the better the information. This implies that at the early stages of development, the range and the amount of information are limited and agency costs are high. Since the information created by a project is an externality on others, the decentralized economy is constrained inefficient; in particular, it does not of projects 'experiment' enough. The analysis of the role of information also opens the tigation of the development of financial institutions. We way to an inves- contrast the infor- mation aggregation role of stock markets and information production role of banks. Because the amount of available information increases with development, our model predicts the pattern of financial development observed in practice; banks first and stock markets later. JEL Classification: D82, E44, G20. KeyAvords: Agency Costs, Development, Information, Financial Institutions, Social Experimentation. *We thank Andres Almazan, Alberto Bisin, Russ Cooper, Peter Diamond, John Hassler, Torsten Jaume Ventura and seminar particUPF, Wharton for helpful comments and Persson, Sevi Rodriguez-Mora, Mfiria Saez-Marti, Ilya Segal, ipants at Copenhagen, Harvard-MIT, IIES-Stockhohn, Benjamin Levin for excellent research assistance. International Institute of Economic the Ministry of Education of Spain Zilibotti acknowledges the hospitality of the and financial support from Studies, University of Stockholm, (DGICYT PB93-0388). Introduction 1 The efficient allocation of resources who are not the reqmres that many tasks be delegated to agents residual claimants of the returns they generate. It full natural that agency relations play an important role in development [e.g. some Third World [e.g. many is therefore accounts of economic Mydral, 1968, North, 1990] and that high agency costs faced by societies North, 1990, p. have been argued to prevent their economic development 59]. The change in the factory system at the time of the British Industrial Revolution [see Mokjn:, 1991, for discussion], or the emergence of hierarchical organizations among the agency relations that process of development. is and professional management [see Chandler, 1977] are seem to have played an important Perhaps the most important example of agency relations in the credit market. Entrepreneurs be given the right incentives. It is borrowing funds for their activities need to well-known that financial intermediation was limited at the early stages of development, and economic growth and the growth of intermediated fimds went hand in hand over the past three centuries 1969, Kennedy, 1987, role in the King and Levine, 1994]. For instance. [e.g. Goldsmith, Goldsmith (1987) shows that in most pre-modern societies financial arrangements were extremely informal, and the same pattern arises from Townsend's (1995) study of Indian villages. These observations suggest that societies at the early stages of development were unable to have wide-ranging agency relations ties to raise and relied predominantly on family or village funds and ensure enforcement. This situation contrasts with the more developed and complex credit relations that we observe today. A number of other features related to the evolution of incentive contracts agency relations are also relevant to our investigation. stages and First, wliile at the early most firms were owner managed, the majority of large firms today have man- agement separated from ownership. This suggests that the 'high powered' incentives CEOs of the owners have been replaced by the weaker incentives of current day Berle and Means, 1932, Jensen and Meckling, 1976]. Moreover, even when is restricted to professional Murphy (1990) managers only, the show that the pay-performance [e.g. attention same pattern emerges. Jensen and sensitivity for CEOs has decreased substantially since the 1930s^ Second, while in less developed external financing re^ Unable to explain this pattern using any existing theory, Jensen and Murphy suggest that this due to political constraints. However, an impUcation of this explanation is that non-monetary methods of control should be used more often and thus current day CEOs should be replaced more frequently for poor performance. In contrast, Haddlock and Liuner (1994) find that the likeUhood is lies almost exclusively on banks and other direct lending relations [Goldsmith, 1987; Pry, 1995], stock and bond markets play an increasingly important role in developed economies. Since banks tjrpically screen and monitor the projects they finance [Dia- mond, 1984], their diminishing role suggests that the informational requirements of agency relations and thus the form of incentive contracts have been changing over time. This paper has three related objectives. ple explanation for Our main costs thesis why agency The first is costs are high at the early stages of development^. that the information structure of an is and formalize a sim- to offer economy determines agency and that over the process of development the information structure changes endogenously. According to North (1990, p. 57), the problem nication mechanism is required'. This 'communication mechanism' economy develops, the become cheaper to A weak in form a commu- poor societies and as an become more flows of information enforce. is 'to know when punishment provide the information necessary to to is efiicient direct prediction of our theory is and incentives the observed pat- tern of evolution of incentive contracts: at the early stages of development, large pimishments are necessary to induce the right incentives; but as the commimication mechanism develops, better risk-sharing (insurance) can be off^ered to entrepreneins and other agents, and incentive contracts can become second objective is to show that the less 'high-powered'. analysis of agency costs Our and information has important impUcations for the development of financial institutions. Our third objective to assess the efficiency of the evolution of agency relations is and financial institutions. As an example to borrow of the paper's money for main idea, consider the case of foreign trade. Given the to the behavior of the entreprenevir — e.g. and vmcertainty of risk has he picked a good trade, effort, is he steaUng part of the money, was the ship? — , amoimt an agent who wants it is relative he putting the weather or carelessness that sunk high agency costs are to be expected. In fact, in pre-modern economies, of dismissal has not changed since 1930s. •^We should note that the main argmnent of this paper is about shadow rather than actual agency costs. The actual agency costs in the villages studied by Townsend may be low because no one engages in entrepreneurship nor invests in risky projects. What is very high is the cost that an agent would have to incur if he decided to borrow money to become entrepreneiu, i.e. the agency cost at the margin. Similarly, some aspects of incentives in the complex contemporary organizations may be quite distorted, but absent the relatively efficient information flows of modern would be much more serious and perhaps such complex organizations would society, distortions not exist. while long-distance trade was an important activity, investors bore a large amount of risk and the risk-premium was very high [see a hypothetical situation in which there are Braudel, 1979]. In contrast, consider many other entrepreneurs borrowing funds for similar foreign trades. In this alternative scenario, investors can reduce the agency costs by using the information they obtain from other entrepreneurs regarding the unavoidable uncertainty of this trade, performance evaluation nmnber in the jargon of the relative they can filter out the literature, common shock. This is the At the early stages of development, hmited savings constrain story of our paper. the i.e. of projects (or entrepreneurs) as well as the information that can be used to write incentive contracts. Limited information in turn leads to high agency costs. and As the their increases, more entrepreneurs are active performance reveals a substantial amount of information to the which can be used we economy capital stock of the will also society, in devising the right incentives for each entrepreneur. Moreover, argue that the relative scarcity of information at the early stages of development favors banking over stock markets, and that economic development can be associated with a shift from bank finance to stock and bond markets, as observed in practice in the coiirse of financial development. Our model has three key featm'es: subject to moral hazard; amount (ii) (i) Production requires entrepreneurial Different projects have correlated returns; of savings determines the number effort (iii) of projects that can be imdertaken. The As a result, savings determine the amovmt of information which can be used in devising appropriate incentive schemes for entreprenein-s, with accumulation. Expressed differently, in an economy with moral hazard, the compensation of agents depends on idiosyncratic and ence their performance, and this lack of As the economy becomes richer full and agency costs decrease common shocks which influ- insiuance introduces high agency costs. and undertakes more projects, the compensation of agents can be conditioned on the success of other projects, therefore the variability introduced due to common diction of the model. turnover of shocks can be largely avoided. Gibbons and Murphy (1990) relation find that the compensation and CEOs depend significantly on the performance of other firms in the same industry and conclude that there evaluations In line with this pre- among top is executives. support for the presence of relative performance Haddlock and Lumer (1994) find even a stronger between these variables using data from the 1930s when the U.S. companies were much less diversified to one industry. than today, thus could more easily be classified to belong Since each project's performance reveals information relevant for others, infor- mation has pubUc good features. It is then natural to question the extent to which the market achieves an efficient allocation of resources. To answer this question we contrast the choice of a social plarmer subject to the relevant informational constraints to the decentralized equihbriiim. We show that the social planner would always choose to produce more information than the decentralized equiUbriimi by 'experimenting'. Further, this constrained inefiiciency result complex to the formation of In our is a pubhc good because stock prices of different and reveal the performance of each project and thus dertaken within the auspices of a bank may be two all is not necessarily publicly observed, and as a better equipped to deal with the free-rider problems at the early stages of development. to emphasize to be robust In contrast, detailed information regarding projects im- the relevant information. consequence banks shown financial coalitions. economy information firms are pubhcly observed is The comparison of banks to stock markets leads us distinct fimctions related to information: the first is information aggregation; the aggregation of available information, efiicient in this function. and stock markets are more The second is information production which will depend en- dogenously on the financial incentives provided by different arrangements. Precisely because the stock market is free-rider effects, therefore, more it is disadvantage of stock markets when there is efficient at aggregating information, it creates not always good at producing information. more dramatic at the early stages of This development information to be aggregated, and more need to produce addi- is less tional information. This accords well with the emphasis of a number of economic historians such as Cottrell (1992), Tilly (1992) and Kennedy (1987) who emphasize the role of banks in gathering information over the development process. Therefore, at the early stages of development, as it is observed in practice [Goldsmith, 1987], banks and other non-market institutions are the main channel of financial interme- As development diation. proceeds, however, more information can be aggregated and stock markets emerge. The fact that since the more information reduces agency work of Holmstrom information structure lation to the papers is a costs has been known at least (1979). However, to our knowledge, endogenizing the new step. The main mechanism we propose has some re- on tournaments and yard-stick competition [Holmstrom, 1982, Green and Stokey, 1983, Lazear and Rosen, 1981, Shleifer, 1985] which also argue that conditioning on the performance of other agents improves incentives, but these papers treat the niimber of projects and thus the information structure as given. Prom a large different perspective, number Diamond (1984) also discusses the advantages of a Our paper of projects in a moral hazard setting. also related to is the literature on rational expectations equihbria, see inter aha Green (1977), Gross- man (1979) and Kyle (1989). (1980), where information the information that is is The closest link a public good and relevant in their context market traders, whereas the perhaps to Grossman and is thus imderprovided. is is However, the private information of stock role of information in our of agency contracts. Ovir paper also shares a Stiglitz model common ground is to reduce the costs with the literature on the financial development and growth. Here especially. Greenwood and Jovanovic (1990), Bencivenga how and Smith (1991) and Greenwood and Smith (1993) discuss financial intermediation interacts with growth. in a related spirit discuss the interaction between Acemoglu and Zilibotti (1995) risk-diversification through finan- Newman (1993,1995) and Aghion and may also be used to endogenize agency arrangements and growth. Banerjee and cial Bolton (1993) propose a mechanism which costs. In these papers, the distribution of income is endogenous and on the form of loan contracts; as the economy develops agents become limited liabiUty constraints a pattern opposite to what contracts should The plan become is less serious. it impacts richer, thus However, this mechanism predicts observed in practice; as an economy develops, agency become more 'high-powered'. of the paper is as follows. The next section lays out the basic model and characterizes the equilibriiun with stock markets. Section 3 characterizes the social plarmer's choice and demonstrates that the decentralized equilibriiun inefficient. An Appendix financial development observed contains the proofs of the how our in practice. Section 5 main Propositions and Lemmas. The model 2 Set 2.1 2.1.1 We constrained Section 4 analyzes equilibrium with banking and demonstrates model predicts the pattern of concludes. is up of the model Timing of Events and Preferences consider a two-period economy where a large nxmiber (M) only derive utility from second-period consiunption. Each agent of identical agents is risk-averse with given by utility = EoU{co,ci,ei) where second period consumption and is Ci Agents who decide to become entrepreneurs agents, ei = = EoU{ci,ei) Eo\ogc{ -v{ei), will have to exert economy side of the consists of We entrepreneurial labor. will refer to these as 'capital-intensive' (y) sectors, respectively. two ship spend the end of this period, Each agent who hmnan second period of their consmne the proceedings Output is is / The in the labor-intensive sector dinring the is no consumption the workers no longer work; they simply of their investments from the in the labor-intensive sector of this labor input and rest saved and invested in the capital-intensive lives, X where capital (at rtm the capital-intensive sector firms. At the become workers. They work whole income life decide entrepreneur- they receive their salaries and consume this amount. in the first period, their sector in the first period of his period of their fives and receive a wage income. Since there sector. In the first the 'labor-intensive' (x) and the period of their lives acquiring the necessary first in the second period, they of the agents first 'capital intensive' (y) sector. economy is given by: = Al, A is aggregate labor productivity. Since all factor markets are assimied to be competitive, the entire output will accrue to the workers. particular, since all first period A{M — letters the N), where w and s income is saved, we have of agents who w= s =A and In W=S= denote respectively wages and savings and capital case denote aggregate variables. N, which will be number 0. for all other The sectors. has a choice of whether to become an entrepreneur. Agents and and > and the second sector uses savings and uses unskilled labor as the imique input, cost) eflFort 0, v'{.) 0. The production no = Also v{0) ei is effort. oiir key endogenous variable, decide to become entrepreneurs at time is 0. Production in the capital-intensive sector requires a project and an entrepreneur to transform the savings of workers into output. denoted by 'large' that -^ C/ = [0,A'']; nimiber (that is ^More constant). presicely, is each project is The set of available projects is in our calculations we will let A^ —> oo and M Each project can only be run by one entrepreneur we assume that if (M) represented by an integer and A^ more than one entrepreneur run the same — > ^, is a oo such thus N is project, they all also the the number amount open of The projects. level of production in project j depends on of capital (kj), but there are decreasing returns at the project level [so that in equilibrium not productive only [see discussion all the funds are invested in only one project]. Also, a firm employs an amoimt of savings larger than some if it below on The production this feature]. ezk^ where ^ G effort of the as; >D <D if kj if kj {0, 1, ^} is a stochastic variable D critical level, function for an entrepreneur with a project in the capital-intensive sector can be written Vj is whose reaUzation depends on the entrepreneur and on the state of nature. To simplify matters, level of we assume that each entrepreneur decides between high and low effort, and the utility cost of high effort = is v{.) by no other agent Whether the e. in this entrepreneiir has exerted high effort economy. The underlying state is can be Good with probability p and Bad with probabihty 2.1.2 We observed also unobservable and it — p. Uncertainty. assume that the set of projects three subsets, such that these subsets is, [/{o}) U = respectively, Each project has an Pr(j e 1 is - 1 U{oy | U U (where | \= N) can be partitioned into U Urj^ where the cardinality of — 7r)A'', [/{i} |= (tt — 6)N, Urj^ U{i} U[oy \= (1 U , | \ each of |= 8N. identical probability of belonging to each subset, thus Vj' - TT, Pr(j G t/{i}) = TT- 6, Common captures idiosyncratic uncertainty in our model. as the return of the projects in these subsets will t/m) = Piij e 6. G U, This feature shocks are also present depend on the imderlying state of nature. In particular, - Vj e [/{o} ^ Oj = 0. ^ r. 9i =0 -V;e[/,,}^Kj_^ { . 6j vi e U^e} 9j 6i make zero returns. = =6 =1 iff e^ = low and state is Bad; otherwise. = low = high iff Cj iff Cj and and state is Bad; state is Good; otherwise. This ensures that there are no property rights over the projects so that the Even if there were property rights and the right to run savers will always be paid the full returns. a project could be traded, these rights would have a price of zero until the point when therefore none of our key results would be aiFected. N= N, Therefore, high effort increases the expected return of a projects in two ways; it reduces the probabiUty of a failure in bad times, and we assvune that this latter effect implies that 8 is = is Table 1 and 'small', l/N, so the ex-ante probability infinitesimal. increases the probability good times. For technical reasons which of a very high return in soon, it become will clear f/r^ be a singleton. This let each project to belong to Urj^ for simimarizes the conditional probabiUties of the different realizations for each project. Table 1. Conditional probabiUties of realizations. Production Underlying State Effort Good High (l-TT) (prob.=p) Low (l-TT) Bad High (l-TT) Low 1 (prob.=l With and - p) ezk^ 1 N TT stock markets, the realization of 9 for every firm will be publicly observed this information wiU be used to form the posterior public improve the efficiency of contracts that induce high costs. Intuitively, is Bad was exerted). effort a stronger punishment of bad outcomes (a failure signals that effort than when the state is Good The reason signal extraction in belief regarding the A better inference of the miderlying state underlying state of natin-e. of nature level Zk^ is of nature will and thus reduce agency called for when the state was not exerted with high probability) (a failure is uninformative about whether or not to introduce the very high realization 6 a simple way — with 6 = 0, when all is effort to enable entreprenemrs exert effort, no information about the iinderlying state would be revealed. The presence of very high return implies that exert effort, in the when all projects are open, and when all this entrepreneurs Good state we would observe one project with return 6Z whereas the very high return would not be observed in the Bad state. Therefore, when all projects are open and rim by high effort, the imderlying state would be revealed and the common shock can be infinitesimal of retm-n is convenient as it filtered out perfectly. The assumption that 6 ensures that the effect of 6 on the aggregate rate and on the incentive compatibility condition of the entrepreneurs negligible. The important economic rim, signal extraction will is point is be harder, thus, as that the when only 8 be a few projects are number of projects inference about the underlying state of nature will improve. will increases, the It Table has to be noted that there are actually two important features embedded in The 1. first is the one discussed in the above paragraph: as the nimiber of projects increases, the information about the underlying state of nature improves. The second low effort feature which will play a crucial role in sections 3 and 4 that high and is have different consequences regarding information revelation. This second aspect will be discussed in detail later. The Stock Market 2.2 At the begirming of every period, a market which we and functions without any offer to costs of transaction. call the stock market opens Each entrepreneur makes a contract the market which determines the pa3anent associated with one share (sale price normalized to $1) of his business in each possible publicly observable state of nature. Thus, the contract for entrepreneur j is a mapping from the space of publicly observable events, denoted by E(n), into real numbers; Pj of the world a G 2(n) is : reahzed, each consumer receives an E(n) -^ 5R. After a state amount (dividend) Pj{cr) per share. Since the savings used in the capital-intensive sector fully depreciate after use, the capital value of each share after the dividend is zero, therefore Pj{(r) the post-realization price of a $1 share inclusive of the dividend. it is We also is assume that possible to trade shares in the stock market after the realization of the state of nature and before the dividend payment, and as a result, all agents observe the share price of each 'firm' (entreprenevir) and, via the price, they infer the realization of^. The set of observable events is conditioned upon the n = is publicly observable. Contracts, P, are conditional ^, because as discussed above, n will fraction of open projects, determine the amount of information that upon the following events: (i) what the information revealed about the whether the project is imderlying state Although the payment of each share could be conditioned upon is. successful or not; (ii) the performance of other specific projects, this will not have any information content above and beyond the conditioning upon the public assessment of the underlying state. Therefore, when condition the payments and dividends on a of two elements; the successful When first and the second this will cause we convenient, instead of the overall state of nature a, summary measure aj which is a vector denotes whether the project in question, project is will j, is the pubhc behef about the imderlying state of nature. no confusion we will drop the subscript j. The ex-post price of each share in each state u will be: (1) ' kj< D. if we Finally, will assume that all contracts are signed at the same point in time when agents decide their profession, thus an entrepreneur offers a contract to the market and the workers promise to [this invest a certain amount once they does not introduce any problems since there is receive their wages no vmcertainty regarding wages]. This assumption will make sure that entreprenemrs compete a la Bertrand and thus obtain no rents above their reservation assumed that entrepreneurs choose The 2.3 utility. Also, throughout the analysis it is their effort level after all contracts are signed^. decentralized equilibrium The Equilibrium Concept 2.3.1 The concept we use for the decentraUzed equihbrium will is an adaptation of the notion of Perfect Bayesian Equihbrium to an economy with competing principals. We require that (i) all be obtained by Bayes' beliefs rule, (ii) all entrepreneurs max- imize their returns given their contracts and choose the best contract for themselves and (iii) all workers behave optimally based on their beliefs taking actions as given. In particular, this projects and investment other agents' means that a worker can take the levels of other agents have implicitly assumed that shareholders are Uable for the project's project gives a zero return, the price turns negative set of his return. losses. When the and shareholders must pay to the entreprenetir the agreed wage out of their personal income [note that if the consimiption of the entrepreneur in the case of a failure were zero, nobody would choose to become entrepreneur since log(O) An open and hence the information revealed by these actions as given, and consider a deviation that maximizes ^We all = — oo]. model which would give exactly the same predictions with more notations is one where entrepreneurs can buy insurances against personal failures. Denote the insurance payment that the entrepreneur j receives in state a by ij {a) and the total compensation of the entreprenemby ujj{a). Since there are many projects, the insmrance provision can function without any residual alternative risk, thus we have ^^g£(n) to simi to zero, thus V entreprenem- {since this tj(a) is = 0. Also in each state, the total insurance transactions have — 0, Vct e S(n). Savers observe all insurance contracts of the ij {o^) crucial for incentives). r Pj{&) = \ The ex post price of each share ej(a)Zfcf-(wJ'(a)-i,( ^i^ 'f ^i if kj It is would then be: we not ^^ < D. easy to check that the two models give identical results. In the rest of the paper introduce the insurance market in order to keep the notation simpler. 10 will Analysis: Preliminary Results 2.3.2 We first characterize an equilibrium in which a number N < N projects are open and all entreprenevirs choose to exert high effort. We will then determine the equilibrium number N, and, of projects finally, prove that imder some parameter restrictions, the unique equilibriiun will entail aU entrepreneurs exerting high effort. Since all use the convention that if j > that aggregate risk induced by sampling randomness studying the model for the range in which N of We f, then project j will open after /. N suppose throughout the analysis that the number of open projects size D]. we projects are ex ante symmetric, without loss of any generality, A is is will will also so is 'large', neghgible [that is we are minimiun project large relative to This assumption implies that by a law of large number argvmient, the subset the N projects which are open will consist and a proportion of projects belonging to U{oy (approximately) of a proportion — (1 tt) belonging to U{iy. Now tt the maximization problem of a representative saver can be written as max J2 Vi^j) log Yl Pji^j)^j ^^>> j,aj where Sj is \ j =^ ^J J the amovmt that savings invested in project (entrepreneur) Now, consider a situation in which P(a-), to the market, and this induces Given high all effort in a; e will K, was Good (the high rate of return 6Z is high rate of return and 6Z let ujq is in each of the be conditioned. {cDq, (jJi,u!o, it is (lui); and ri{aj) N contract, effort, projects - and [it is details are aU projects, there are two pieces of information wages paid to the entrepreneiu when and successful same each entreprenevir to exert high whether or not the high return reward to an entrepreneur by offer the equiUbria must have this property upon which each entrepreneur's reward project. Second, j, a-j. aU entrepreneurs investors decide to invest an equal amount, straightforward to show that (2) j=l denotes the probability associated with the state omitted]. S s-t- 9Z is First, the return of his observed. Let us denote the ^i}- In particular, let publicly known observed) and Coq and a)i be the that the underlying state when he is luisuccessful and ui denote the corresponding wages (tJo) in case the not observed. Once entreprenemrial rewards are defined, the reaUzations of the ex-post price (or dividends) of each share are determined according to equation (1). 11 Prom (1) V(n,K) and = where we have (2), the utility of each saver can be written in a simple form + +pn log ^"l f— \m — nj - (ttZK^ - wui (1 - ttui case, the entrepreneurial 6Z 1"^ reward is a; € and contribution to and N is Bad, probability u 6 1 — p, or and Finally, there are M — N savers who open pn. is In this is effectively leave the determination of this may not be observed because the imderlying Good, but the is was not open, probability p{\ Since all nN — n). entrepreneurs exert high In this effort, in projects that are successful and pay — 7r)iV which are not successful and pay negative dividends. positive dividends N we because the underlying state {ujq,uji} is paid. (1 open, probability savers' utility out of the analysis [formally, both underlying states of nature there are of the + ^^ oo that makes this the exact solution to our model]. project that would be very successful case a reward is Note that since there {Coq^Cji}. Alternatively, the high rate of return 6 state Tr)u;o] (3) publicly observed is only one project with the high rate of return dZ, we have ^ -^ - ^, and n = ^ had already been defined above. Let us The high retiirn 6Z is only observed when the tmderlying state Therefore, the overall probabihty that project's contract {I 7r)i:;o) Good, probability p, and the project with the high return 6 n. - m= set explain this equation. is (l-p)]log (-^^)[ttZK'^ - [p(l-n) : projects, thus have invested an equal amount in each will we need to divide the revenue by logarithms and dividing the numerator and the denominator by M— iV. Taking N gives the utility of the representative saver (worker) as (3). We next write the participation and incentive compatibility constraints that need to hold for a representative entrepreneur to be willing to choose this profession and to exert high effort. (1 — pn)[7rloga;i + (1 — 7r)loga;o] (1 The first constraint is +pn[7rloga;i - p)7r(logu;i - for participation. + (1 log Wo) The - 7r)loga;o] > e > V{n,K) e left-hand side (4) (5) is the entrepreneur's expected utiUty and since the entrepreneur can decide to become a worker this has to be no smaller than V{n, K). Note that therefore (4) requires the utiUty of value of a saver. all agents are free to become savers, an entrepreneur to be greater than the maximized This makes the problem somehow non-standard in that 12 it is no longer a simple constrained maximization but a fixed-point problem. constraint expected is The second for incentive compatibility. It requires that the entrepreneur utility from high effort than from low In other words, effort. has higher it requires the loss of utiUty from lower rewards in the case of low effort to be less than the cost of high effort. and u{a) are determined by maximizing In equiUbrium, P(o-) (3) subject to the participation constraint (4) and the incentive constraint (5) with respect to ^1, cjo, (^0) and uJi. Proposition together with all other proofs 1 summarizes the results in the Appendix]. is Proposition 1 In an equilibrium where 1. The constraints 2. Each entrepreneur and (4) G = exp entrepreneurs choose high effort: all (5) hold with equality. receives the following rewards: UJl = = Cjq u>i where Tj-f-c;: > 1, B= ttG + GuJq = Buq Savers obtain the safe return -^^^^{ttZK^ 4. The expected utility obtained by all — (1 3. — agents tt) > with partial derivatives Vn{n, K) > for the safe return 1, and E = exp{e} > 1. Bojn). is equal to -+^ As the expression [the proof of this Proposition Q and Vxin, (^) K) > - (7) 1 0. shows, Bujq can be interpreted as the aver- age (per project) cost of entrepreneurship borne by the savers. Furthermore, from equation (7) we can observe the entrepreneur is that not getting receives a salary that {^j full is the cost incurred due to the fact that insm:ance; instead, with probability (1 depends on the outcome of the project and hence he 13 —pn), he is bearing some As n risk. increases this probability Finally, notice that savers are fully insinred by effort all feature as costs it entrepreneurs, there is and the associated due to the costs will go down. fact that conditional no aggregate uncertainty. This is on high a very useful enables us to completely isolate the impaxit of information on agency by removing the interactions between aggregate uncertainty and agency The Number 2.3.3 we Next, of Projects: characterize the choice of entrepreneurs exert high Assumption 1 costs. K and n still assuming that in equilibrivmi all effort. ^-^m > E (^) — 1 This condition implies that decreasing returns to capital in each project are sufficiently strong (low that, if possible, to /3), compared to the effort cost. therefore guarantees It open a new project and pay the compensation to one more entrepreneur wiU always be more profitable than to expand the scale of production in existing firms. In the absence of this would be affected, assumption none of our qualitative results but restricting attention to this set of parameter values simplifies the exposition. Recall now that aggregate savings, S, is equal to the wage income of the previous period. Then: Proposition 2 Suppose Assumption 1 holds. Then, in an equilibrium with high effort: (ii)If^>§=^S = A{M-N), Therefore, the maximimi mmiber kj^j^, of projects which andN = N. is consistent with the tech- nological constraints will be opened in a high effort equiUbrium, of savings increases more projects will be imdertaken. Once all and as the stock the projects are open, expansion wiU follow in the form of higher investment in each project. consequence the determines is level of savings how much As a determines the nrnnber of projects which in turn information becomes pubficly available and thus to induce the right incentives. Since the level of savings 14 is how costly it a one-to-one function of the level of labor productivity, A, respect to A (or interchangeably Proposition 3 Let u denote we will do oiir comparative static analysis with with respect to S)^. the entrepreneurial (^^) fine the agency costs by the ratio, [recall wage with equation contractible effort. (4)]- De- Then, agency costs are a decreasing function of the aggregate stock of savings. At the early stages when S is n low, also low is and thus agency In richer economies, the nvunber of projects which can be financed better information our analysis. and thus agency costs are lower. This is larger, there is one of the key results of demonstrates that, as often claimed informally, It is costs are high. e.g. North (1990), development goes hand in hand with the reduction of incentive (agency) costs and the reduction in these costs at the later stages is due to an improvement in the information structure of the economy. Moreover, as the economy develops, n, the probability that the underlying state underlying state Good, is we have is cjq discovered increases and because = (Di, when the and together with development, the agent will bear less risk and the incentive contracts wiU become less and less 'highpowered'. This High 2.4 is in all - as long as the cost of effort entrepreneurs choose high Assumption 2 We in the introduction. Effort as Equilibrium We now prove that equilibrium Une with the evidence discussed E ( J^) < f ^^ + also introduce is not too high, in the decentralized effort. 1. some additional notation: p denotes the probabiUty of discovering ex post that the underlying state has been Good conditional on the actual state being Good. ^Note that while the model forward. is static, to extend these results to a growing In Acemoglu and Zilibotti (1996a), assume that production we economy is straight- and on the labor consider an overlapping generation model, in the capital-intensive sector exerts a positive externaUty productivity of the labor-intensive sector, such that At — BY^, and At is the state variable of the model. In this case, the model exhibits standard neoclassical dynamics with convergence to a N grows and agency and the stock of savings increase. Therefore, all results can be easily re-interpreted the context of a growing economy (an earUer version with the details is also available on request stationary steady-state. Along the transitional path towards the steady-state, costs fall as At in from the authors). 15 - V{n, K\p) denotes the indirect entrepreneurs exerting high - b{n, K\p) and b'^{n, utility of each worker conditional on p, with all effort. K\p) respectively denote the average retiurn of one project net of the salary paid to the entrepreneur conditional on high effort and low effort. Let us now estabhsh that the return to a representative saver in the case of high effort for all projects is increasing in Lemma 1 Then^p' > Lemma Consider an allocation in which V{n,K\p') p, all entrepreneurs choose high > V{n,K\p) andb{n,K\p') > 2 Suppose Assumption 2 holds, then h{n,K\p) With low The the amount of available information. effort, effort. b{n,K\p). > U^{n,K\p), Vp. the cost of effort and the related agency costs are not incurred. rate of return of the project informational issues the - is lower. first effect is Lemma 2 estabUshes that - leaving aside outweighed by the second, and savers receive a higher retvirn from a high effort project than from a low effort project. Proposition 4 Suppose Assumptions ized in Propositions 1 Given Assumption and 2 2, is 1 and 2 hold. Then, the allocation character- the unique equilibrium. high effort has higher return and this induces each en- treprenem: to offer a contract that promises high effort. To see the intuition, sup- pose that an entrepreneiu offered a contract that would make him choose low effort, because the rate of return from this project would be lower than the alternatives, each saver wovild prefer to invest in other projects. Therefore, the entrepreneur with low effort would not be able to raise enough funds. We conclude this section with two remarks about the robustness of omx specification. First, in our formulation agency costs decrease as the nimiber of firms grows, because the probability of observing the nvraiber of firms. The same mechanism, fully reveahng signal, 6, is increasing in the that the inference of the imderlying state improves with the number of firms, would work more generally as long as the number of possible realizations of the underlying state were of the ber of projects. number Our specification with just of projects captures this as the num- two underlying states and a very large mechanism 16 same order in a parsimonious way. Second, our results depend on some form of technological non-convexity. be opened at all stages of development, then there However, the assimiption of minimvmi costs. function (that arising is, from the given exactly the ioi k < D, y = 0) is same retmrn would be no dynamics inessential as there already a non-convexity is one entrepreneur who needs to be it D would would complicate the analysis because the equilibrium size of projects in this case would depend on the discontinuity at agency as the savers. Thus, removing this assumption not alter our qualitative results, but The in requirement in the production size fact that each project requires the projects can If all number of projects. enables us to keep the project size constant irrespective of the nvunber of projects are open. Constrained Inefficiency 3 Constrained Efficiency and Experimentation 3.1 We have now established that decentrahzed equiUbrium induces may to choose high effort. However, this different information 1; if than high effort. a project that has exerted low information revelation This is the second feature we Although the exact a special featine of our model, is which is embedded link it is between than high amount instance, consider a is but it more complex matrix return but reveal more information. it We can has a lower has the potential of revealing socially useful considerably actions would require high effort, and of socially revealed by different production techniques. effort information. This featmre effort an example of a more think of choosing low effort in this setting as experimentation because direct return in Table discover that the state general issue; the trade-off between the private return and the useful information entrepreneurs not be optimal since low effort produces effort is successful, of nature has definitely been Good. all more general than our formahzation: of actions and a subset The and payoffs than Table of these 1: for some would have lower private actions that reveal more information would play the same role as low effort in our example. We now analyze the choice of a social plarmer representative agent subject to the who maximizes same informational the welfare of a constraints as the decentral- ized economy. Namely, the social planner will directly observe neither the underlying state nor the effort choices of the entrepreneurs. Under this informational constraint, the planner will choose and announce the set of contracts that will be offered to the 17 who agents decide to become entrepreneurs^. DN Proposition 5 3 Sh < (i) S < If effort in the N— -^ Sh, the social planner would induce no effort in r projects > S > open projects (Hi) If N — r projects, remaining DN (a) If such that; S > DN, where < < r N and N= and high -^. Sh, then the social planner would induce high effort in all (r = 0). then N all and r projects are open = 0. This proposition states that the social planner would choose to experiment at earlier stages of development Low effort yields if low effort is that the underlying state of nature is chosen and the project it 0, the decentralized equilibrium However note that even in this range. feature that the amount we discover increases p, the probability that the society discovers the state of nature. Since for > successful, is Good. Therefore, investing in low effort projects is eqmvalent to social experimentation because allocation has r (or labor productivity) is low. a lower return, thus everything else being equal, choosing low effort However, is costly. when the stock of savings S< is Sh, the constrained efficient constrained Pareto inefficient in the social planner's choice we have the of information available to the society on which incentive contracts can be conditioned increases along the process of development, thus agency costs are decreasing in A and S. Impossibility of Experimentation in Equilibrium 3.2 The still previous subsection demonstrated that the constrained efficient allocation in- volves some degree of experimentation. However, the stock market such an allocation fact that the stock is we know from not sustainable. Yet this section 2 that with is market set-up forces each entrepreneur to act alone and thus provides no means for internaUzing the informational externality. ition is partly due to the A possible intu- that financial coalitions can form in order to internalize this externality [see Boyd and Prescott, 1987 who suggest financial coahtions as a solution for a different type of externahty]; for instance, a group of entrepreneurs can get together and offer shares as an investment fund. Then, ^Note also that we axe finitesimal. This is still is may be conjectured that some positive maintaining the assumption that iV equivalent to assuming that the level of savings small, then the social planner there it is S is large or that would again prefer not to experiment. For instance, obviously no point in experimenting. 18 n is not in- N were when N = not too small. If \, amount of experimentation can be sustained an equilibrium. as We will show that this conjecture is not correct. We model the formation of financial coalitions any number of projects and these projects. There exist a finite to maximize make aries will zero a set of contracts j G Ji] and compete a < profit'''. > u^ More number I : a if thus in equilibrium we assume specifically, to each of the entrepreneurs — > ^'^ and P* : E(n) -^ a and will get in state u^jicr) 3?, i deals with [that actually With do this set-up some amount no all any generaUty that who gets in state is all In the risks. intermediaries risk. socially desirable. stated and proved in Proposition 6; here we can be far, it equilibriimi in the range of saving levels {S is If each — P*(cr) — Z)jeJi '^]{'^) the idiosyncratic of experimentation tion impossible. for is an intermediary rims a large number loss of they bear no to the existing literature. offers again a mapping and the equilibrium concept we have used so exists i each $1 invested. Thus we have Z)jej. 9j{a)Z'Kkj can diversify we suppose without this, therefore shown that there it is amoimt that entreprenein j as the in state a. If of projects, then in equilibrimn, rest of this section, level for intermedi- with P*(cr) as the amount the investor he accepts the contract. The amount the profit of the intermediary it it all assumed that intermediary and an investment package (fund) to the market which Ti{n) put $1 of financial intermediaries la Bertrand, from the observable states to a pajnnent u^j fi- a dividend stream that combines the returns of offer all profits entrepreneurs through A financial intermediary can costlessly nancial intermediaries (or investment funds). 'run' among < Sh) where This result will be formally briefly discuss the intuition First, the presence of free-entry and relate makes experimenta- an intermediary engages in experimentation, it also needs to run some other projects with higher returns to cover the costs of experimentation. However, another financial intermediary can then bid away the profitable projects who are cross-subsidizing the low effort entrepreneurs and leave only the loss-making experimentation project(s) to the intermediary. Therefore, even with complex financial coalitions, competition first and public availability of information make sure that any intermediary engaging in costly information creation will not be able to enjoy the monopoly power required to cover allocation without experimentation cannot ^For ofF-the equilibrium path behavior, owned by all its among the we assume that M agents. 19 on the other projects. Next, an be an equilibriiun either since a the agents in the Scime generation and distributed equally losses if all financial they make financial intermediaries are jointly profits or losses, these wiU be intermediary can do better than this allocation (through some degree of experimen- and attract a tation) non-existence result Stiglitz (1976). is In large portion of the funds. is Grossman and related to Grossman and arises when Grossman and no information Stightz, in our equilibrium in the original game [see is stealing high eff'ort effort projects in and Stiglitz, and and revealed and no equilibrium is in a sense natural exist in general. Also, in Rothschild poohng equilibrium by same consequences free-riding our model. However, it stealing as a financial intermediary has to be noted that again in Rothschild the non-existence problem arises without coahtions and also that nonis much more common than moral hazard models as ours. We wiU now show that as in Rothschild and Stiglitz 's insm-ance model a natural refinement of omi original equilibrium concept. Reactive Equilibrium, restore a unique equilibrium tions 1 and 2, make is sufficient to which coincides with equilibrium outcome of Proposi- where aU entrepreneurs exert high effort. Intuitively, previous equilibriimi concept, to disturb equilibrium to and on the information created by the low existence problems in adverse selection models as theirs in market traders Bernheim, Peleg and Whinston, 1987 for a def- this hcis the projects and Rothschild and equivalent to looking for a coahtion-proof and such an equiUbrimn does not profitable types This exists. economy the non-existence prob- Stightz's (1976) paper, entry can always destroy a away is financial coaUtions are introduced. This since introducing financial coalitions inition] Stiglitz (1980) StigUtz, the information of stock costs of gathering information, thus lem only no equilibrium This implies that no trader wants to incur the revealed by the market price. exists. In contrast to So, it was according to oiu sufficient for positive profits given the set of existing financial contracts. an entrant Whereas the Reactive Equifibrium imposes the additional requirement that the entrant should not be subject to yet another round of entry which would make her incur negative profits^. Before providing the formal definition, we introduce the following notation. denotes a set of contracts offered to savers by a financial coafition. A game U.{C\C U C C") is given in Acemoglu and Zihwhich financial intermediaries are sequentiaUy off'ering contracts or are withdrawing the contracts that they have previously oflFered. An equiUbriimi is reached when no intermediary wants to withdraw or make a fiu-ther offer. We show that for any cost of withdrawing contracts e > 0, there is a imique equifibrium which is the same as the Reactive EqmUbriiun. It is significant to note that Wilson's (1979) equilibriimi concept, which is in spirit very different than Reactive Eqmfibrium, was motivated by intermediaries withdrawing their contracts from the market but, in om- game, this equiUbrivun only exists when e = 0. ® theoretic justification for this equifibrium concept botti (1996b). Briefly, consider a game in 20 denotes the profits of the intermediary that offers C when also C is offered in the market. Then: A Definition 1 (i) {Ci} vector of contracts {Cj} is feasible; all is a Reactive Equihbrium agents are optimizing conditional on {Cj}, > derived by Bayes' rule, andVj, Tl{Cj\{Ci}) (ii) VC : n{C'\{Ci}liC') C U C") < The > 0, 3C" : all beliefs are 0. U{C"\{Ci}UC'[JC") > 0, andU{C'\{Ci}U 0. part of the definition first iff is common with our previous definition of equihb- rium; aU agents need to optimize and based on this, no intermediary makes negative profits. We also need to impose that these contracts are feasible, intermediary promises to undertake project second part j, it contracts are optimal only against all is if a financial enough funds to do raises the 'reactive' equilibriiun restriction. is that It so. The requires that the existing deviations which themselves would not turn unprofitable. Proposition 6 (i) V5 < Sh no (Perfect Bayesian) Equilibrium exists. and 2 where (a) V5, the allocation characterized by Propositions 1 all entrepreneurs exert effort is the unique Reactive Equilibrium. The second that there is part of this proposition is and fimctioning costlessly. always constrained inefficient because produces too 4 important for two reasons. First, it shows a well-defined and unique equilibrium in our economy even with complex coalitions forming have is little Second, the unique equilibrium it we leads to no experimentation, thus information. Financial Institutions and Information. We have so far established that (i) the amoimt of information increases with develop- ment, hence agency costs are lower in developed economies and (ii) the equihbrium of our economy, especially at the early stages of development, produces too little information. This second feature brings the question of may arise to deal strates that when with this inefficiency. access to information The is what financial institutions result in the previous subsection demon- not restricted, coalition formation will not 21 prevent the inefficiency. In this context we argue that banks can be thought of will on access to information and/or producing information that as placing restrictions cannot be easily transmitted. Before we start is it we have useful to recall that also suggested that stock markets have good information revelation and aggregation properties. That success of a project is observed publicly through the performance of its is, the share on the stock market and this impUes that the information contained in this performance regarding the underlying state this informational transmitted to is all the agents in the economy. Yet, advantage of stock markets creates a free-rider problem. wants to bear the cost of experimenting [by choosing the low private return to reveal information that useful to the whole society. is of stock markets in information aggregation mation production. In the rest of the As a result, will agent activity] the advantage makes them disadvantageous paper we No show that the market at infor- failure just discussed can explain the emergence of a speciahzed financial institution, which wiU We refer to as 'banks'. reasons we wiU emphasize two roles of banks [and treat each separately banks are not as make them will efficient at though they are in no way for expositional excliisive]. First, information aggregation as stock markets but this will well-suited to information production at the early stages by avoiding the Second, banks have alternative ways of obtaining information, free-rider problem. in particiilar as we emphasized by Diamond (1984), they have the capacity to monitor individual effort choices. The comparison empirically of stock markets we observe and banks is of considerable interest because that banks and other direct lending institutions played an important role in developing economies and stock and bond markets only emerged much later in the development process. For instance, in an important historical study, Goldsmith (1987) analyzes the financial structures often pre-modern societies and finds that institutions banking was developed in a number of them while stock market type were not observed at all except in Holland and there only due to special circumstances. However, since the nineteenth century stock markets have played an increasingly important role in the financing of Western economies. Despite to date has offered this new and existing ventures in many well-known historical sequence, economic theory no explanation^. ^An exception is Greenwood and Smith (1993) who discuss debt versus equity and conclude that equity mcirkets will never arise in equihbrium without government intervention. Another important contribution is Greenwood and Jovanovic (1990) who 22 also motivate the existence of Banks 4.1 as 'Experimenters'. Equilibrium With Beuiks Only 4.1.1 we In this subsection, drawn in this section outline the is first role of banks. The important distinction between institutions that lend to a group of borrowers through a bilateral relation and institutional arrangements whereby each borrower comes into contact with the whole market. Banking That sive bilateral relation. entrepreneurs and provides tion that the bank. the funds to these entrepreneurs, and the informa- produced by these entrepreneurs is Intuitively, would observe its if be modelled as an exclu- each bank enters into a relation with a number of is, all will is not observed by anyone other than a company enters the stock market, the whole economy performance. In contrast, if it obtains all its finances from a bank, We the economy will only have limited information on this company. that the fimctioning of banks Ce{Nb) that is is costly; if a bank runs iVj, projects, there incurred in terms of final output in every state. increasing and weakly convex. One also Ce(.) possible justification for these costs is assume is a cost positive, comes from the fact that banks often gather information by making most of their investment in a particular industry and thus are not well-diversified. Also, banks in practice incur administrative costs which would be part of Ce(.). economy when the only equilibrixun of this We now characterize the financial intermediation possibility is banking. Proposition 7 Let Ce{Nb) (i)3 bank is Se{Ce) such that active choosing low f = S< then V Ce < nZV^ such Se{Ce), there is that: a unique equilibrium in which one carries out experimentation in the form of f > 1 entrepreneurs effort. S > (a) If and if = CeN b, Se{Ce), then we have a unique equilibrium without experimentation, 0. Corollary 1 There Ce(.) such that in number I > 1 (i) exists and an open set of (ii) strictly increasing in Proposition 7, and convex functions instead of a unique bank, we have a of banks. by arguing that they run small scale experiments to extract information about the state of natiu-e. However, since each intermediary finances a continuum of projects but only experiments on a countable subset of them, experimentation is a costless activity in their model, and there is no comparison of the information production roles of diflFerent financial financial intermediaries institutions. 23 Since banks do not automatically reveal the information that is produced by the projects they are financing, they are subject to less severe free-rider problems. In bank particular, a economy in this receive outside information the information that Note since and there on an effectively is model there derlying state of nature, if is isolated island; no possibiUty of other banks produces since this information it in our is it does not free-riding on not pubhcly observed. is no source of micertainty other than the un- the aggregate performance of the bank were observed, the relevant information woiild again be perfectly revealed, thus the restrictions on information transmission introduced by banking could not work. However, transmission would not occur easily if, as it seems plausible, other unobservable we have ables affect the performance of the bank. Further, left this vari- out of the analysis the possibility that banks trade in information. Although interesting, this extension would not alter the quaUtative result as long as the costs of banking, Ce{.), are positive. Financial Development: Banks versus Stock Mcirkets 4.1.2 When will intermediation be carried out through banks rather than stock markets? In answering this question, The linear cost functions. marginal cost is we restrict the analysis to the case in generalization of this result to the case of increasing straightforward. At this stage, a most convenient. To write the utiUty of a saver aggregate saving level S < ND, we Then we have the (7). but with C = Cn which banks have diagrammatic analysis in the stock K=D substitute will be market economy with and n = -^ in equation utihty of a representative saver given by V{S) = log (('S') tZD"C can be thought as the average rate of M+% return on aggregate savings. Instead of the decentrahzed equilibrimn, where the underlying for the The 1 now state is publicly observable; consider a hypothetical we then have the same average rate of return on aggregate savings with difference is due to the fact that the true state (^ = (^gg = we would have C = Cce = expression TvZD^-'^ M+^ becomes known with probabiUty rather than n. Finally, in another hypothetical world in which effort contractible, economy ^^"^^^ ^^'^ is perfectly entreprenemrs will only be m+£\e-i\ paid their reservation return in the form of a constant salary. Now let us turn to banking. When all 24 projects are financed through banks. the tmderlying state perimentation. On is inferred more precisely since the banks will carry out ex- the other hand, banking comes at the cost of Cg per project. = Therefore, the average rate of retm:n on savings would be given by C <zZD^-'^-Ce% where \{i-pp') (1-PP-, M+% p' is the probabiUty that the Good state depends on the optimal extent of experimentation and n < p' < = revealed, p' when than less projects can be opened. Figure la plots the inverse of these four rates of return A'^ (plotting the inverse simplifies the diagrams). As given [but of course no experimentation Ss{Ce) SliCe) is ND SUCm) intuitive reason Cos^ why starts at the these curves start at the not matter relative to imderlying state At Figure lb. same point but are zero, the utihty of consimiption is minus lies Bank not observed, it is above infinity (^'^^ ND as 'monitors'. retiurn with contractible everywhere above C~^ [the same point ^~^ also starts at the this]. this point, all projects are is other savers as Ss{Cm) below the other curves, implying that the rate of effort is highest. all possible with stock markets] Figure la. Banks as 'experim^enters'. (^~^ lies max- before, the equiUbriimi will imize the utility of a representative saver taking the decision of is I is (be is that when and the cost of same savings effort does point, but because the imtil aggregate savings reach ND. nm with high effort and the imderlying state of natture revealed almost surely, thus the two curves meet again. Next to understand the shape of Q^ (thicker curve), suppose that then the (inverse of the) rate of return Constrained Pareto Optimum is banks fimction at no cost (i.e. given by the dashed curve which of this economy. 25 is Cg = 0), also the This line starts at the same point as the other ciirves, then case can use it remains below more information than the decentralized economy], and then the point where the bank finds optimal to stop experimentation, it the cost of banking Ce increases, Figure is up (^"^ shifts it finally, at meets (^~^ As . Inspection of this multiplicatively. suS&cient to establish the following proposition [except for the multiplicity aspect which is discussed below]. Proposition 8 3 Ce,SL{Ce), Ss{Ce) such S G If (i) [because the agency contracts in this (^~^ {SL{Ce),Ss{Ce)), then that for Ce > (^be < Ce; Cnn o,nd in equilibrium all funds are intermediated by banks. (a) If S > Ss{Ce), then < (he Cnn, o,nd in the absence of the stock market, there exist multiple equilibria; investment funds in one with banking and one with stock market intermediation. With investment funds, the only (Reactive) equilibrium has all savings intermediated through the stock market. Returning to Figure levels such that banking is less efficient is (say, no information when S > Ss{Ce) banking coaUtions among entrepreneurs can be If through financial intermediaries), stock market an equilibrium with banks. if we do not in the predictions of our state of nature observed in the development experience of many countries. on banks and Goldsmith, 1987]. The intuition for two types of institutions duce information but are is why there is higher) less less devel- is this historical switch between the worth discussing. As noted above, banks actively pro- inefficient at and since there experimentation diminishes. becomes typically bilateral direct lending relations [see information aggregation. In more developed economies the amoimt of information that needs to be aggregated N is Developed economies use institutions such as stock and bond markets quite widely whereas rely and a project that uses the stock market. model are consistent with the pattern which oped economies exclusively be the no firm enters the stock market, then economy regarding the underlying for will allow costless coahtions, there also exists Intuitively, if agency costs would be very high The not too large, there will be a range of saving the only eqmlibrium. Instead, tinique equilibrimn. However, is is than the stock market. formed costlessly there Ce la, if is is larger (that is, information available at no cost, the need for active As an economy grows rich, information production important and information aggregation, the activity at which stock markets have a comparative advantage, gains importance, hence the switch from 26 banks to stock markets^". In support of this thesis, stock market of Holland 'Hhe exchange's price we can quote Goldsmith on the list... was one of the most important sources of information for the period's international trade." and Thomas, 1973 for statements to the same argue in this paper, even the first effect]. [p. 217, see also North This suggests that, as we proper stock market of economic history appears to have acted as an important source of information. Therefore, our model explains the late emergence of stock markets by the different comparative advantages of various financial institutions in producing infor- mation. Naturally, the prediction that once the stock market disappear is unrealistic, but it is is introduced banks due to the simplicity of om- framework banks only perform 'experimentation' activity, as well as in which our assumption that all projects are homogenous. Banks 4.2 The as Monitors other role of banks is to reduce transaction and agency costs via monitoring. Assimie that each project can be interim monitored at the some constant cost, which we denote as Cm- Monitoring enables the bank to observe whether the entrepreneur exerts effort or not. Therefore, monitoring forces high effort. For simplicity, assimie that these banks necessarily monitor tions in projects. all More we also realistic specifica- which banks can decide to randomly monitor some projects, or to ex-post monitor only unsuccessful projects [as in Diamond, 1984] would give the same re- sults. Since each bank deals with a large number insurance to the entrepreneur and pay him a 3] so as to meet his participation constraint. of projects, flat wage it can provide perfect [O as defined in Proposition Thus we can write the return ^°Note that the model also predicts that at the very primitive stages of development fi-om (for S < SL{Ce)), the stock market performs better than banking. This is because stock markets are supposed to fxmction at no admistrative cost which is an unrealistic assmnption but makes the rest of oiu" mechanism easier to appreciate. Goldsmith (1987)'s analysis of pre- modern financial systems shows that at the very early stages, there was no intermediation, and it was only after a number of necessary economic conditions were met that banks started to play an important role in financial intermediation. Therefore, the historical pattern suggests a sequence of a period of almost no intermediation, then banking and then finally a period of stock market intermediation. In Acemoglu and Zihbotti (1996a), we show that when we allow for an alternative technology without division of labor and no need of financial contracts this technology will be chosen at the very early stages of development. Therefore combining the results of these two papers, we have predictions which are in line with the styhzed fact about financial development. 27 = banking in this case as ( C^ is the same as (^~J- (~^ over some range. But preferred, thus = "^gijiT^r^f and in Figure lb, Cm Therefore, as long as Cbm than a also monitoring if critical is This time for Cm > for is less • 0, it shifts Cm = up value Cm, the curve the curve 0, multipUcatively. (^ will be below very cheap, stock markets will never be Cm needs to be greater than another critical value Cm and in this case Proposition 8 will apply to monitoring banks exactly as to banks as experimenters. The sequence of financial institutions that arise in equihbriimi will again be in line with the historical pattern. and is only necessary do not work. when Intuitively, heavy reUance on direct monitoring more information is revealed, and it to rely on the stock market to aggregate this information rather than monitoring should play a societies is in line It is also less is optimal make heavy interesting to observe that the conclusion that important role in modern society than in more ancient with the view that information has become more decentralized and privacy has acquired a value 5 costly other methods of providing incentives to entrepreneurs In rich economies use of direct monitoring. is it did not have before. Concluding Comments Agency costs feature importantly in many accoimts of the process of development. High risk-premia, distorted incentives, corruption, limitations on the division of labor can all be related to the high agency costs faced by less developed economies. Why are agency costs high? We suggest an answer to this question based on the idea that the hmited range of projects xmdertaken in less developed economies leads to relatively less information for the society to for agents. be used in devising the right incentives This argument explains a number of styUzed facts about development of financial institutions and evolution of incentive contracts. The mechanism proposed in this paper opens a nimaber of avenues search. If indeed information reduces agency costs and is for future re- increasing in the amoimt of savings, similar arguments can be applied to imderstand the emergence of or- ganizations that rely on Fm-thermore, on financial more complex we have omitted divisions of labor and deeper in this paper the importance of arrangements and agency costs. The hierarchies. income distribution interaction between these two as- pects has not yet been investigated but potentially important in imderstanding both the development of financial institutions and the incentive problems faced by societies today. 28 many APPENDIX: Proofs Proof of Proposition 1. 1. Suppose that holds with strict inequaUty for the entrepreneur running project v. function could be increased by reducing the entrepreneur's salary in all states. Next consider the case where (5) holds with strict inequality. Then a meanpreserving contraction in cjj and Uq would still satisfy the incentive compatibility constraint but would increase the utility of the entrepreneur, thus cOi and a>Q can be reduced without violating the participation constraint and hence the objective function would be increased [note that due to log utility, we can always reduce these (4) The objective salaries 2. Let B D without hitting a boundary]. and G be as in the Proposition, uji Then by this Proposition. problem can be written = Gloq follows from (5) and Part 1 of substituting from (5) into (3) and (4), the maximization as: V{n, K) = m_ax_ - pn) logiir ZK'^ - (1 Bloq) + UlQ ,i'l jliJO pn log [(ttZX^ - - TTWi (1 - 7r)tDo)l + log L J " ( (A.l) ) \m — n) subject to: (1 The — pn)[7rlogG-|-loga;] conditions that (Do +pn[7rlog(Di = cji = the F.O.C's with respect to cjo, -7r)loga;o] wo and = max log wo Cj\ (we omit the IttZK^ J The tives part follows from the previous point. straightforward. Proof of Proposition can be opened. We will first (A. 2) 2. First part. maximize Then, the objective — (A.3) = V{n,K) + e (A.4) — nj \m ^1 f for V{n, K) and u}q which is given D 3. first is = V{n^K) as: (A.3) and (A.4) uniquely determine the solution by expressions (6) and (7) in the text. Substituting wq into (A.3) proves details). - Bujq] + log L log(Ba;o)-(l-pn)[logB-7rlogG] 4. e as: and the participation constraint (A.2) 3. - -Bwq (part 1 of the Proposition) are immediate from function can be written (eq. (A.l)) y(n, K) + (1 (7) The calculation of the partial deriva- D Suppose A{M — N) < ND, so not all projects with respect to n and subject to the resource K 29 {nK < ^) and constraint the equilibrium value of the minimum n and S. Since y„ By constraint must be binding. {K > D). Then, we size constraint > 0, V^- > will determine (see Proposition 1) the resource using the resource constraint to eliminate K, we can then rewrite (7) as: (l-pn) V{n) = Lz (-|^ log This expression constraint that j to be is + g first /x > condition for is /i order condition \(l-pn) m + n E (*) the Lagrangean multiplier of the to be strictly positive is size) -/i = (A.6) -1 minimum size constraint. A sufficient that: \(l-pn) Q is (A.5) is: \(l-pn) ^ 1-/3 > n which -1 (J^^ E (#)"--' plog(j.) ^(*) n where m + n fE - /?) log(n) - log maximized with respect to n subject to the (minimum n < j^. The 1-0 (1 ^(*) -1 \(l-3m) Q m + n E (*) (A.7) -1 equivalent to: 1-/3 m> n /3 m" pn) -1 (A.8) < E (A.9) Finally, \ (i-pn) Vn € (0, 1) Therefore, Assumption 1 (resource constraint) and 5= -j^DM and iV = , n H^ is sufficient to ensure that A{M — N) = S -^, and the first K (savings equal ND = D. Next, since = S wage income), we obtain that part of the Proposition is proved. A{M — N) > ND, the first part implies that the economy will open as part. If projects as possible, thus n 1. Then, from the fact that VK{'i-,K) > 0. it follows = S. Finally, a maximum of agents that the resource constraint will be binding, so — become entrepreneurs, can hence N) S. This completes the proof. Second = many N NK A{M = W= Proof of Proposition 3. First, consider the case in which effort were contractible. Then the salary of the entrepreneur would be independent of n and thus of the amount of savings, which implies V{n, K) = loginZRl^ -Q)+ log n m—n (A.10) and logu) = V(n,K) + e 30 (A.ll) where (A.ll) is the relevant participation constraint. Hence: l From equation (6) we know (A.12) + E-2Tn—n that: n bA(^-p") B^, = Then agency ^^^-"^ , „-" ^ZK^ B\(l-P") n l + ^(^) costs are given by: BuQ Tin— 71 is decreasing with n, as Proof of Lemma it 1. Since the (A.14) m Q UJ which (A.13) can be N(l-pn) verified by D differentiation. probabihty of discovering the Good state is p, the program to be solved becomes: max y(n,ii:) == [p(l - p) + +pplog (1 -p)] log {-kZK^ f— — nj \m ^^ {-kZK^ f— — nj \m ^^ - TTcDi (1 - TxuJi - (1 - tx)ujq) + 7r)wo) (A.15) subject to the constraints (1 -p)[7rloga;i - -7r)logwo] +pp[7rloga;i (1 (1 -p)7r(loga;i The first u = Buq - - (1 -7r)loga;o] log wo) > -e > V{n,K) (A.16) (A. 17) e wi = Guq and whatever the probability of discovering the underlying state, p. Following the order conditions of the problem, as in Proposition steps above, we can (1), yield solve: V[n,K\p) = log(7rZir^) b{n,K\p) = - log \^ + E Q \i^-PP) [^ (A.18) {kZKP) { pn) i(^+^(*r (A.19) -1 Straightforward differentiation shows that both expressions are increasing in p everywhere. D Proof of Lemima 2, We entrepreneurs choose low characterize the effort. To minimum cost of entrepreneurship when some when some projects are with start with, note that 31 effort, the economy can be subject to aggregate uncertainty since in the Bad state, of the low effort projects fail. In this proof, we will ignore the cost induced by the introduction of uninsurable risk, since this risk would make low effort less desirable and thus make our argument true a fortiori. Ignoring the uninsurable risk implies that the entrepreneurs who are induced to exert no effort are offered a flat wage, which we denote low all by a;'^ The On participation constraint for the no effort entrepreneur implies log(a;'^) V{n,K). the other hand, the participation and incentive constraints for the rest of the en- trepreneurs (4) and (5) together with the log = ( ^ + e + V{n,K). The two first part of Proposition give log{Bu)o) 1 = conditions together yield: Buo J' = E (*) (A.20) (i-pp) Then: {nZK^ - h{n,K\p)-b''{n,K\p) Bujo) - pnZK^ - BujQ B \ -kZK^ (1-p) The RHS of (A.23) N, i.e. ^^xH ^^ ^* i Proposition 1]. T^ZK^ / minimum at N = N [since the only term which depends on maximum at A'^ = iV see the expression for Bujq in the proof of reaches a ^^^ BuQ (A.21) \(i-P^) B E (*) \(i-PP) — At the minimum, the RHS of (A.21) becomes: -nZK^ (1-pp) 1 + -^ ^^(Jr) ^ M-N~^\ ^ M-N (A.22) \G^, and some simple algebra using the definition of E, G and B shows that the term within D is always positive as long as Assumption 2 is satisfied. brackets Proof of Proposition 4. Consider, first, an allocation such that Nd < % agents are entrepreneurs. Then there is an opportimity for a Nd+V*^ agent to become an entrepreneur and increase his expected utility since there are enough funds. Therefore, any equilibrium = -^ entrepreneurs. Consider now an allocation where r > of allocation must have N N entrepreneurs raise funds with a fiat wage contract at uie. Savers will naturally anticipate that they will exert no effort. By Lemma 2, the contribution of each low effort project to the savers' portfolio is lower than that guaranteed by high effort projects. So, each low effort project will raise less funds than and will not be feasible. Therefore in the D N equilibrium we cannot have r > 0. Finally, consider the case in which entrepreneurs offer contracts that satisfy the incentive compatibility constraint and thus savers anticipate high Now, if a iV + P* agent decides to become entrepreneur, he will not be able enough funds with either a high effort or low effort contract. Further none of the effort. to raise 32 entrepreneurs can improve on this allocation, therefore the allocation with D contract is an equilibrium and there cannot be any other. Proof of Proposition induced for all 5. = projects (r N high effort Consider as a benchmark the allocation in which effort is Now suppose that for a small number of entrepreneurs r 0). replace the incentive compatible contract with a flat wage w'^. Let p, p' be respectively the probabilities of discovering the Good state before and after the contract replacements. we To be we have p precise, = § and = p' ^+ (l - ^) [1 - (1 - ^ So, p' Tr)^. p + [l-(l-7rr](l-|) + ^(l-7rr. This increase in probability of discovering the underlying state from p to a change in the expected return equal (A^ where the {N — r) first - r) term -r [b{n, K\p') - the eflFect of the increase in the information which the remaining is b{n, K\p)] b{n, incentive contracts can be conditioned upon, from inducing no effort in r firms. (A.23) N [b{n,K\p') - b{n,K\p)] where the first term is cause p' will to: positive r \h{n,K\p') and the second is U^{n, K\p') and the second term can be written - - K\p) is (A.23) the loss return as: - negative by b'^{n,K\p') Lemmas 1 and (A.24) 2, respectively. We now proceed in two steps. First, we show that when a non-negligible fraction of projects are not open, i.e. when h = N — N has the same order of magnitude as A'' [that is 4- > 0], is always desirable. Second, we analyze the case in which h is small with respect to A'' so that n —> 1~ [that is ^ !^ 0]. In the first case, since then experimentation (infinitesimal) N is by assmnption 'large', then p' "^^ p -\- 7r(l — n) [note that since (1 - n) = j^, the assumption that h is not infinitesimal with respect to is crucial for this approximation] N Then, since b(n^K\p') (A.24) is — b{ji,K\p) > 0, there exists r > 0, small relative to TV, such that strictly positive. namely the limiting case in which n —+ 1~. Now, the previous argument does not apply because 6(n, K\p') — b{n, K\p) becomes infinitesimal. Next, consider the case of 'small' In this case, we need /i, to write the explicit expression for the benefit of experimentation, namely: N [6(n,ir|p') - b(n,K\p)\ = ^-kZK^N 1 < 1 + B \(1-PP') ^^(*) 1 + "^^[w] [\-VP) (A.25) which, after rearranging, gives: -zk^^^e[-8^) 1 (*)"""') + ;=fes (i (i-p) -(#) + ;#.^ (*)""'") " 1/iV (A.26) 33 can now take the limit of this expression as A'' —> oo and n —> 1~. Note that to calculate the limit of the second term (which is of the type ^) we use L'Hopital Rule. The We resulting expression is: lim N—oo.n—*1- E(^) (l-P) 1 + E (^) (i-p) - b{n, K\p)] = N M-N ttZK^- A^ [6(n, K\p') ,p\og N (^) [/I - - i^Yih + r)] = Q[h - (1 (1 - nYih + r)] M-N (A.27) which is a positive function of h [where the expression Next, consider the cost. We Q is defined for future use] have: r \h{n, K\p') lim N->oo x),n—!" - b^\n, K\p') = 1- 1rn ZK^ rH ,1-p 1 The ^<M + lih^^ (A.28) ( J^j net value of experimentation - which depends on r as well as on /i - is Q[h-{l-'Ky{h + r)]-rH and, for given h, experimentation positive. For r = 0, this (A.29) is is beneficial iff (A.29) 3r € N'^ such that this expression , equal to zero and for r 7^ it can be written fc>r(r).r-'^/«+'l-'')' 1 Equation (A. 23) - (1 is as: (A.30) - -kY gives: Sh^N min Tir) reN+ (A.31) the critical savings level such that such that V5 < Sh experimenting is optimal. Sh < ND, we immediately obtain that S > Sh experimenting has higher cost than benefit, therefore, in this range r = as claimed. This completes the proof. D where Sh is Finally, since Proof of Proposition 6. Part (i). First, we will show that there will exist a contract that makes positive profits if the allocation has positive amoimt of experimentation [Result (1)]. Then, we will prove that for S < Sh, there exists a contract that will make positive profit when the allocation in question has no experimentation [Result 2] These two results will prove that there exists no equilibrium with our previous equilibrium concept. Then we will move to prove the existence of a tmique Reactive Equilibrium. Result 1. Experimentation means that a financial intermediary that markets a set of projects U* is inducing a subset of this u** C U* to exert no effort - that is, r > 1 . entrepreneurs are offered a flat wage a;'®. Let us also denote the probability of discovering that the underlying state was Good by p as before. By Lemma 2, the return on these — r projects b{n,K\p). r projects, b'-^{n,K\p), is lower than the return on the other N 34 Now, another intermediary can attract all entrepreneurs in the set U*\u** by offering them a slightly higher reward in all states, and market this security to savers at rate of return b{n,K\p) — e, who will prefer this new security for e small enough; as long as u** ^ 0, this deviant intermediary would make positive profits. Thus no allocation with D experimentation can be an equilibrium. Result 2. Consider an allocation where a subset of projects U* <ZU = [0, N] is open Now another intermediary can enter, attract all the all entrepreneurs exert effort. entrepreneurs in the set U* by offering all but one of them a slightly higher reward in all and and the remaining one a states, the first flat wage (cj'^ + e), where lj^^ is part of Proposition 5 this portfolio gives higher utility to coalition can make defined by (A.22). By edl savers and thus the D positive profits. Part (ii). We will first show that the high effort allocation is an equilibrium [Result and then that no others exist [Result 4]- We focus on S < Sh, which is the region where an equilibrium did not exist. For the other cases an equilibrium exists with our previous equilibrium concept and is the high effort equilibrium. Let p = n he the information available in the absence of experimentation, and p' > p the information available when some experimentation is carried out (see the proof of Proposition 5 for details). Also, remember from Proposition 1 that savers bear no risk, thus we can work with average rates of returns without worrying about variability of returns. Result 3 Consider the candidate equilibrium allocation where a number L of intermediaries offer identical incentive compatible contracts of the type described by Proposition 1. Now, if we call this set of contracts, [Result 2] above shows that 3C'\U{C'\{Ci}L)C') > 0. However, as shown by [Result 1], for such there also exists C" such that a second round of entry is profitable, i.e. U{C"\{Ci} U U C") > 0. Then, to show that the candidate is a Reactive Equilibrium we only have to show that U{C'[{Ci} U C" U C") < 0. As it is characterized in the proof of [Result 1], C" steals all projects that had high effort, and leaves projects that have no effort with the incumbent intermediary that offered 3] d C C is 2, . The average h{n,K\p') Lemma 2, > when there is experimentation to a small nimiber of projects provides savers with a return up to 6(n, K\p'). But, by Lemma rate of return 6(n, K\p'). Therefore, must have C C" h'-^{n^K\p'). Furthermore, in order to attract savers away from {Ci\,C' from high effort that is b{n,K\p), where, again by offered at least the return b{n^K[p) > b''^{n,K[p). Now we claim (postponing by few lines the proof) that b^^{n,K\p') < b^^{n,K\p). This implies that the intermediary which had entered at the first stage offering is left only with the no effort projects and therefore making a loss, because it is offering savers a return b{n,K\p). This establishes that all entrepreneurs exerting effort is an equilibrium. C To finish the b'^^{n,K\p). J' = To proof of Result 5, we need to show see this, observe that: b^^{n,K[p') that, as claimed above, 6'^(n, K\p') = {pirZK^ — cv'-^) is < decreasing with p /Xpn) = i+^(g^j -VJi-^n) ^ is increasing with p. ;;r=;r Result 4- We now show that no other equilibrium exists. Consider an allocation in which some projects choose low effort (set ui /g = 0). If we can show that no such since: , E[-^) allocation can be an equilibrium, we will have established the uniqueness of the high by the above argument (i.e. that b{n,K\p) > b''^{n,K\p')), the low effort projects must be run by an intermediary who also runs a number of high effort projects - let us call the set of these projects with high effort Uh- But as long as Uh 7^ 01 there exists another intermediary who can offer C" and attract all these projects in Ufi- And this intermediary can offer a rate of return as high as b{n,K\p') because effort equilibrium. First, the experimentation of the first intermediary 35 who offered C still reveals the underlying Thus this new intermediary with C" can make positive profits. Therefore, to show that this candidate allocation is not an equilibrium it is sufficient to would no^ make negative profits, that is n(C|{Ci}U show that for any further entry C", does not do cross-subsidization C" U C") > 0. However, this is true by definition since of projects; the worst that can happen is that it loses u^ because C" offers a better return (say closer to the maximum rate fe(n,K|/j')), but C" never makes negative profits. state with probability p' . C C D < txZD^ ensures that the return to each project remains positive after paying the banking cost. Second, free entry im- Proof of Proposition 7. First, the condition that Ce implement the Constrained Social Optimum subject to payment of the 'administrative' cost of Ce per project. Note that no entrant can free-ride on the experimentation carried out by the incumbent bank because the replies that if a unique bank is sulting information does not adopted if and only if active, become it will public. Finally, the proof that experimentation will is below some threshold (5e) follows the proof which needs to be subtracted from the return of Ce the stock of savings of Proposition 5 except for the cost each project. The corresponding expression Se{Ce) = is: mm = r (A.32) ^_^^_^y where ,(i-p) rr { _B_Y'-P> _N_ 1-P This completes the proof of the Proposition. G'^ JV D liFirst note that as long as Ce(l) < nZK^, there exists an equilibrium with banking. This follows immediately from the fact that the allocation in which each project is run by one bank dominates no production. Second, note that in all equilibria, banks have to make zero profits; otherwise, another bank could enter and run exactly the same projects and pay e more to the savers in all states. Now we will construct an example where Ce(.) is increasing and convex and where there are a number of banks Proof of Corollary larger than 1 carrjdng out active experimentation. Let the savings level he S< Sh, then consider the following cost function for banks: ^ where Nb < -^ and > 0. Also Ce < nZD^ and Ce > Ce- Let Ce —> oo, then we will have at least two banks in this economy. Next, since at least two banks will have Nb projects and -j^ > 0, experimentation is profitable by the argument of the proof of Proposition 5. This establishes that for Cf{.) increasing and convex, we have an equilibrium with many banks and each bank (with the possible exception of one that may be too small) runs active experimentation. Next, suppose a sequence of increasing convex functions, {Cg(.)}fc — C|°(.), and a sequence of economies {£''^}fc such that economy k has the > 36 banking cost function C^. Then it immediately follows that 3k* such that \/k > k*, the equilibrium of economy E'^ has a number Z*^ > 1 of banks where each bank (but possibly one) runs experimentation with r*^ > 0. References [1 [2: [3] Acemoglu, D., and Zilibotti, F., (1995). 'Was Prometheus Unbound by Chance? Risk, Diversification and Growth.' MIT Working Paper, No. 95-12. Acemoglu, D., and Zilibotti, F., (1996a). Extent of Information?' in progress. 'Is Acemoglu, D., and 'Dynamic Financial Competition and the Zilibotti, F., (1996b). the Division of Labor Limited by the Reactive Equilibrium' in progress. [4] [5] [6] Aghion, P. and Bolton, P. (1993);'A Trickle ment.' Mimeo, London School of Economics. Down Theory of Growth and Develop- Banerjee, A. and Newman, A. (1993). 'Occupational Choice and the Process of Development' Journal of Political Economy, 101, pp. 274-298. Banerjee, A. and Newman, A. (1995). 'Migration, Integration and Development' MIT Mimeo [7] [8] [9] Bencivenga, V. and Smith, B. (1991). 'Financial intermediation and endogenous growth'. Review of Economic Studies, 58, pp. 195-209 Berle A. A. and G. C. Means (1932); The New York, McMillan Publishing Modem Corporation and Private Property Bernheim, D., Peleg, B., and Whinston, M. (1987); 'Coalition-proof Nash Equihbrium' Journal of Economic Theory J., and Prescott, E. (1986). 'Financial IntermediaryEconomic Theory, 38, pp. 211-32. [10] Boyd, [11] Braudel, F. (1979); Civilization and Capitalism: The Wheels of II, Harper and Row, New York. [12] Cottrell, (1992) in Cassis (editor) Coalitions.' Journal of Commerce Volume Finance and Financiers in European History: 1880- 1890. [13] Diamond, D. (1984). 'Financial Intermediation and Delegated Monitoring.' Review of Economic Studies, 51, pp. 393-414. [14] Pry, M. Edition. [15] (1995). Money, Interest, The Johns Hopkins and Banking in Economic Development. Second University Press. Gibbons, R., and Murphy, K. (1990). 'Relative Performance Evaluation For Chief Executive Officers' Industrial and Labor Relations Review, 43, pp. 30S-51S. 37 MIT LIBRARIES 3 9080 00903 4718 [16] [17] Goldsmith, R. (1987). Premodem Study, Cambridge University Green, J. A Financial Systems: Historical Comparative Press, Cambridge. (1977);'The Non-existence of Informational Equilibria' Review of Economic pp 451-463. Studies 44, [18] [19] [20] Green, J., and Stokey, N. (1983). 'A Comparison of Tournaments and Contests.' Journal of Political Economy, 91, pp. 349-64. and the Greenwood, J., and Jovanovic, B. (1990). 'Financial development, growth distribution of income.' Journal of Political Economy, 98, pp. 1076-1107. Greenwood, J., and Smith, B. (1993). 'Financial markets in development and the development of financial markets.' Mimeo. [21] Grossman, S., and Stiglitz, J. (1980). 'On the Impossibility of Informationally Efficient Markets.' American Economic Review, 70, pp. 393-408. [22] Haddlock, Turnover' [23] Holmstrom, , [24] C, and Lumer, G. (1994). 'Ownership, Compensation and Executive MIT Mimeo. B., (1979). 'Moral Hazard and Observability' Bell Journal of Economics 10, pp. 74-91 Holmstrom, B. (1982). 'Moral Hazard in Teams' Bell Journal of Economics , 13, pp. 323-340. [25] M. C. and W. H. Meckling (1976);'Theory of the Firm: Managerial Behavior, Agency Costs and Capital Structure' Journal of Financial Economics Vol 3, pp 305Jensen, 360. [26] Jensen, M Incentives' [27] Kennedy, C. and K. J. Murphy (1990); 'Performance Pay and Top-Management Journal of Political Economy Vol 98, pp 225-264 W. {1987); Industrial Structure, Capital Markets Press. and the Origins of British Economic Decline Cambridge University [28] [29] Kyle, A. (1989); 'Informed Speculation with Imperfect Competition' Review Economic Studies, 56, pp 317-356. Mokyr, J. Revolution: (1991). 'Editor's Introduction' in J. An Economic Mokyr (editor) British Industrial Perspective Westview. An New [30] Mydral, G. (1968); The Asian Drama: York, the 20th Century Fund. [31] Lazaer, E. and Rosen, S. (1981); 'Rank-Order Tournaments as Optimal Labor Contracts.' Journal of Political Economy, 89, pp. 841-64. [32] North, D. (1981). Structure and Change in Economic History. New York. [33] North, D, (1990); Institutions, Institutional Change and Economic Performance, Cambridge University Press, Cambridge. 38 Inquiry into the Poverty of Nations. W. W. Norton & Co., [34] [35] [36] Riley, J. (1979). 'Informational Equilibrium' Rosenberg, N., and Bridzell, L. Books. How the West Grew Rich. New York Basic M. and Stiglitz, J. (1976). 'Equilibrium in Competitive Insurance MarEssay on the Economics of Imperfect Information.' Quarterly Journal of Economics, 20, pp. 629-49. Rothschild, kets: [37] Jr. (1985). Econometrica 47, pp 331-359. An Shleifer, A. (1985). 'A Theory of Yardstick Competition.' Rand Journal of economics, 16, 319-27. Cassis op. [38] Tilly, (1992) in [39] Townsend, R. (1995); 'Financial Systems nal of Economics Vol 110, 1011-1046. [40] cit. in Northern Thai Villages' Quarterly Jour- Wilson, C. (1979). 'A Model of Insurance Markets Journal of Economic Theory 16, 167-201. 39 With Incomplete Information' Date Due 3 OCT. 1 IfeSt 2 3 1988 1 Qm^ Lib-26-67