8 B.Banks as Delegated Monitors Diamond (1984)offered the first coherent explanation for the existence of financial intermediaries.?Diamond's intermediaries "monitor"borrowers.Since monitoring is costly,it is efficient to delegate the task to a specialized agent,the bank.The notion of monitoring borrowers has become an influential idea,which subsequent researchers have further developed. Not only do Diamond's intermediaries contain most of the important elements of a theory of intermediation,discussed above,but he also identifies and solves a fundamental problem at the root of intermediation theory.That problem concerns the fact that whatever problem the intermediary solves to add value with respect to borrowers would seem to imply that lenders to the intermediary would face the same problem with respect to their lending to the intermediary.In Diamond(1984),the intermediary "monitors"borrowers on behalf of investors who lend to the intermediary.But,then it would appear that the lenders to the intermediary have to"monitor"the intermediary itself.How is this problem,which has come to be known as the "monitoring the monitor"problem,solved?Diamond (1984)was the first to recognize and then solve this problem. In Diamond (1984)borrowers must be "monitored"because there is an ex post information asymmetry in that lenders do not know how much the firm has produced.Only the individual borrower observes the realized output of his project,so contracts cannot be made contingent on the output. Consequently,a lender is at a disadvantage because the borrower will not honor ex ante promises to pay unless there is an incentive to do so.The first possibility Diamond considers to solve this contracting problem is the possibility of relying on a contract that imposes nonpecuniary penalties on the borrower if his payment is not at least a certain minimum.This contract is costly because such penalties are imposed in equilibrium,reducing the utility of borrowers.If,instead,the lender had available an information production technology,then the information asymmetry could be overcome by application of this technology,at a cost.Perhaps this would be cheaper,and hence more efficient,than imposing nonpecuniary penalties.Diamond termed production of information about the borrower's realized output, at a cost,.“monitoring.” The notion of"monitoring"in Diamond(1984)appears inspired by Townsend(1979),but there is a critical difference.In Townsend the lender must bear a cost to determine whether the borrower has the resources to repay the loan or not,a decision made after the borrower's project output has been realized and after a payment has been offered to the lender.That is,in Townsend,the decision by a lender to monitor a borrower is made after the entrepreneur has made a payment to the lender;it is contingent on 2 For reasons of space,we do not survey the previous transaction cost-based literature.For surveys of this literature, see Benston(1976)and Baltensperger(1980)
8 B. Banks as Delegated Monitors Diamond (1984) offered the first coherent explanation for the existence of financial intermediaries.2 Diamond’s intermediaries “monitor” borrowers. Since monitoring is costly, it is efficient to delegate the task to a specialized agent, the bank. The notion of monitoring borrowers has become an influential idea, which subsequent researchers have further developed. Not only do Diamond’s intermediaries contain most of the important elements of a theory of intermediation, discussed above, but he also identifies and solves a fundamental problem at the root of intermediation theory. That problem concerns the fact that whatever problem the intermediary solves to add value with respect to borrowers would seem to imply that lenders to the intermediary would face the same problem with respect to their lending to the intermediary. In Diamond (1984), the intermediary “monitors” borrowers on behalf of investors who lend to the intermediary. But, then it would appear that the lenders to the intermediary have to “monitor” the intermediary itself. How is this problem, which has come to be known as the “monitoring the monitor” problem, solved? Diamond (1984) was the first to recognize and then solve this problem. In Diamond (1984) borrowers must be “monitored” because there is an ex post information asymmetry in that lenders do not know how much the firm has produced. Only the individual borrower observes the realized output of his project, so contracts cannot be made contingent on the output. Consequently, a lender is at a disadvantage because the borrower will not honor ex ante promises to pay unless there is an incentive to do so. The first possibility Diamond considers to solve this contracting problem is the possibility of relying on a contract that imposes nonpecuniary penalties on the borrower if his payment is not at least a certain minimum. This contract is costly because such penalties are imposed in equilibrium, reducing the utility of borrowers. If, instead, the lender had available an information production technology, then the information asymmetry could be overcome by application of this technology, at a cost. Perhaps this would be cheaper, and hence more efficient, than imposing nonpecuniary penalties. Diamond termed production of information about the borrower’s realized output, at a cost, “monitoring.” The notion of “monitoring” in Diamond (1984) appears inspired by Townsend (1979), but there is a critical difference. In Townsend the lender must bear a cost to determine whether the borrower has the resources to repay the loan or not, a decision made after the borrower’s project output has been realized and after a payment has been offered to the lender. That is, in Townsend, the decision by a lender to monitor a borrower is made after the entrepreneur has made a payment to the lender; it is contingent on 2 For reasons of space, we do not survey the previous transaction cost-based literature. For surveys of this literature, see Benston (1976) and Baltensperger (1980)
9 the amount of the payment.Hence,it is known as "costly state verification."In Diamond,however, monitoring is not state contingent and the cost must always be borne because,in Diamond,the monitoring cost must be incurred before the output realization of the borrower's project is known to anyone. This difference between Townsend and Diamond,with respect to monitoring,leads to another difference.In Townsend,the costly state verification problem motivates the form of the contract between a borrower and lender:it is a debt contract(since random monitoring is assumed away;see Boyd and Smith (1994)).In Diamond,the optimal contract between the borrower and the lender is a debt contract in the absence of monitoring,but once monitoring is introduced,the optimal contract is undetermined.It is feasible for the contract to be an equity contract,for example.On the one hand,this does not matter for Diamond's basic argument,but,on the other hand,it seems potentially important for understanding why agents trading in markets cannot replicate the function of the intermediary,as we discuss further below. The monitoring solution may dominate the contract that imposes nonpecuniary penalties,but it raises another problem.If a single borrower has many lenders,then each lender will have to bear the cost of monitoring,which in turn will lead to duplication of monitoring costs or free riding problems among individual lenders.This raises the prospect of a third solution.If the task of monitoring were delegated to a single agent,free riding and duplication of monitoring costs problems could potentially be eliminated.But if the lenders were to delegate the task of monitoring,then the same problem would still exist,but at one step removed.That is,the individual lenders would then face the task of monitoring the agent delegated to monitor the borrower(s).This is the problem of"monitoring the monitor."Diamond (1984)presents the first coherent theory of banking that solves the problem of monitoring the monitor To be more precise,the problem of"monitoring the monitor"is this:lenders to the intermediary can reduce monitoring costs if the costs of monitoring the intermediary are lower than the costs of lenders lending directly to borrowers and directly incurring the monitoring costs.Diamond's fundamental result is to show that as an intermediary grows large,it can commit to a payment to depositors that can only be honored if,in fact,the intermediary has monitored as it promised.If not,then the intermediary incurs nonpecuniary penalties,interpreted by Diamond as bankruptcy costs or loss of reputation. To see the argument,we follow Williamson's(1986)presentation of the Diamond result;unlike Diamond,it does not rely on precise contractual specification of nonpecuniary penalties,which is rarely seen in practice.Williamson's monitoring technology follows Townsend,so Diamond's result does not depend on the timing of monitoring(that is,whether it is state contingent or not).A brief outline of the essential part of the Williamson model is as follows.Borrowers need resources to invest in their projects. They invest K units of endowment at date 0 and receive Kw at date 1,where w is a random variable distributed according to the density f(w).As shown by Gale and Hellwig (1985),the optimal contract between the borrower and a lender is a debt contract.At date 1 borrower j has a realized return of wi per
9 the amount of the payment. Hence, it is known as “costly state verification.” In Diamond, however, monitoring is not state contingent and the cost must always be borne because, in Diamond, the monitoring cost must be incurred before the output realization of the borrower’s project is known to anyone. This difference between Townsend and Diamond, with respect to monitoring, leads to another difference. In Townsend, the costly state verification problem motivates the form of the contract between a borrower and lender: it is a debt contract (since random monitoring is assumed away; see Boyd and Smith (1994)). In Diamond, the optimal contract between the borrower and the lender is a debt contract in the absence of monitoring, but once monitoring is introduced, the optimal contract is undetermined. It is feasible for the contract to be an equity contract, for example. On the one hand, this does not matter for Diamond’s basic argument, but, on the other hand, it seems potentially important for understanding why agents trading in markets cannot replicate the function of the intermediary, as we discuss further below. The monitoring solution may dominate the contract that imposes nonpecuniary penalties, but it raises another problem. If a single borrower has many lenders, then each lender will have to bear the cost of monitoring, which in turn will lead to duplication of monitoring costs or free riding problems among individual lenders. This raises the prospect of a third solution. If the task of monitoring were delegated to a single agent, free riding and duplication of monitoring costs problems could potentially be eliminated. But if the lenders were to delegate the task of monitoring, then the same problem would still exist, but at one step removed. That is, the individual lenders would then face the task of monitoring the agent delegated to monitor the borrower(s). This is the problem of “monitoring the monitor.” Diamond (1984) presents the first coherent theory of banking that solves the problem of monitoring the monitor. To be more precise, the problem of “monitoring the monitor” is this: lenders to the intermediary can reduce monitoring costs if the costs of monitoring the intermediary are lower than the costs of lenders lending directly to borrowers and directly incurring the monitoring costs. Diamond’s fundamental result is to show that as an intermediary grows large, it can commit to a payment to depositors that can only be honored if, in fact, the intermediary has monitored as it promised. If not, then the intermediary incurs nonpecuniary penalties, interpreted by Diamond as bankruptcy costs or loss of reputation. To see the argument, we follow Williamson’s (1986) presentation of the Diamond result; unlike Diamond, it does not rely on precise contractual specification of nonpecuniary penalties, which is rarely seen in practice. Williamson’s monitoring technology follows Townsend, so Diamond’s result does not depend on the timing of monitoring (that is, whether it is state contingent or not). A brief outline of the essential part of the Williamson model is as follows. Borrowers need resources to invest in their projects. They invest K units of endowment at date 0 and receive w ~ K at date 1, where w ~ is a random variable distributed according to the density f(w). As shown by Gale and Hellwig (1985), the optimal contract between the borrower and a lender is a debt contract. At date 1 borrower j has a realized return of wj per
10 unit invested.Borrower j pays the lending intermediary a gross rate of return R in a state,wi,where there is no monitoring and R(wi)when there is monitoring.Define the set B=(wj:R(wi)<R and Be=(wi: R(wi)s R).Finally,let r denote the certain market return,required by risk neutral investors. When the intermediary has m borrowers,each investing K,then the total return to the intermediary (before compensating depositors)is: m=K∑minR(w,R}. j=1 By the strong law of large numbers: m(w)dw)dwv plim 1 Bc Consequently,since the intermediary's return must be at least the market return,r,if the following inequality holds: R(w)f(w;dw;+R∫f(wdwj-(次f(wjdw2r, Bc then,as the intermediary grows large,it can guarantee a certain return of r to its depositors. If the intermediary is finite sized,that is,it lends to a finite number of borrowers,then depositors must monitor the intermediary to ensure that the intermediary,in turn,is monitoring the borrowers.Since monitoring is costly,and given the certain market return that must be obtained,the depositors must be compensated for these monitoring costs by the intermediary.Compensating the depositors for monitoring costs incurred,lowers the profitability (utility)of the intermediary.However,the central result of Diamond (1984)applies here,namely,that the depositors need not monitor an infinitely large intermediary because such a firm can achieve r with probability one.In the limit,depositors do not need to monitor the intermediary.The"monitoring the monitor"problem is solved by diversification. One might object that,in practice,financial intermediaries are not infinitely diversified,and some credit risk is not diversifiable;also,it seems likely that a depositor finds it more difficult to monitor a large bank than to monitor a small bank.Krasa and Villamil (1992a,b)address these concerns.Suppose we modify Williamson(1986)by assuming that larger banks'returns are more costly to verify.If loan returns are stochastically independent of one another,Krasa and Villamil (1992a)apply the Large Deviation Principle to show that,so long as a depositor's cost of monitoring doesn't increase exponentially with bank size,the expected costs of monitoring a sufficiently large bank go to zero. Moreover,they show through examples that even relatively small banks(e.g.,32 loans)get enough gains from diversification to dominate direct lending.If some loan risk is systematic,the chance of bank failure is bounded away from zero as bank size grows(Krasa and Villamil,1992b).In this case,since the cost of
10 unit invested. Borrower j pays the lending intermediary a gross rate of return R in a state, wj, where there is no monitoring and R(wj) when there is monitoring. Define the set B={wj: R(wj) < R } and Bc ={wj: R(wj) ≤ R }. Finally, let r denote the certain market return, required by risk neutral investors. When the intermediary has m borrowers, each investing K, then the total return to the intermediary (before compensating depositors) is: π = ∑ = m j 1 m j K min{R(w ),R}. By the strong law of large numbers: π = ∫ ∫ + → ∞ B c B m j j j j R(w )dw f(w )dw mK 1 m plim . Consequently, since the intermediary’s return must be at least the market return, r, if the following inequality holds: ) f(w )dw r K c R(w )f(w )dw R f(w )dw ( B j j B c B ∫ ∫ j j j + j j − ∫ ≥ , then, as the intermediary grows large, it can guarantee a certain return of r to its depositors. If the intermediary is finite sized, that is, it lends to a finite number of borrowers, then depositors must monitor the intermediary to ensure that the intermediary, in turn, is monitoring the borrowers. Since monitoring is costly, and given the certain market return that must be obtained, the depositors must be compensated for these monitoring costs by the intermediary. Compensating the depositors for monitoring costs incurred, lowers the profitability (utility) of the intermediary. However, the central result of Diamond (1984) applies here, namely, that the depositors need not monitor an infinitely large intermediary because such a firm can achieve r with probability one. In the limit, depositors do not need to monitor the intermediary. The “monitoring the monitor” problem is solved by diversification. One might object that, in practice, financial intermediaries are not infinitely diversified, and some credit risk is not diversifiable; also, it seems likely that a depositor finds it more difficult to monitor a large bank than to monitor a small bank. Krasa and Villamil (1992a,b) address these concerns. Suppose we modify Williamson (1986) by assuming that larger banks’ returns are more costly to verify. If loan returns are stochastically independent of one another, Krasa and Villamil (1992a) apply the Large Deviation Principle to show that, so long as a depositor’s cost of monitoring doesn’t increase exponentially with bank size, the expected costs of monitoring a sufficiently large bank go to zero. Moreover, they show through examples that even relatively small banks (e.g., 32 loans) get enough gains from diversification to dominate direct lending. If some loan risk is systematic, the chance of bank failure is bounded away from zero as bank size grows (Krasa and Villamil, 1992b). In this case, since the cost of
11 monitoring banks that fail is increasing in bank size,there is a bank size past which the increase in monitoring costs dominates marginal benefits from additional diversification.Moreover,this optimal size diminishes as the systematic component of loan risk increases. Winton(1995a)addresses another issue,namely the role of bank capital.Suppose that the banker invests his own funds in the bank as"inside"equity capital.Being junior,such equity absorbs losses first, reducing the probability with which the bank defaults and depositors must monitor.Thus,bank capital is another mechanism for implementing delegated monitoring.Since the bankers'capital is fixed,it will be most helpful for smaller banks;also,the relative importance of capital versus diversification increases as more loan risk is systematic.3 Of course,Diamond (1984)does not explain all the characteristics of intermediaries.But,he elegantly explains the existence of intermediaries,in particular,as coalitions,of borrowers and lenders, which dominate the alternative of direct investment by investors in securities issued by firms.The securities market fails in the sense that intermediation,centralization of the task of monitoring,is a lower cost solution to the ex post information asymmetry between borrowers and lenders.Diversification is critical to intermediation providing a lower cost solution because diversification is critical to reducing the monitoring the monitor problem.The textbook idea that individual investors can diversify nonsystematic risk on their own does not take into account the role diversification plays in allowing an intermediary to be monitored costlessly (in the limit). Other papers that study banks as delegated monitors include Gorton and Haubrich(1987)and Seward(1990). C.Banks as Information Producers If information about investment opportunities is not free,then economic agents may find it worthwhile to produce such information.There will be an inefficient duplication of information production costs if multiple agents choose to produce the same information.Alternatively,a smaller number of agents could produce the information,becoming informed,and then sell the information to the uninformed agents.This,however,introduces the "reliability problem"originally identified by Hirshleifer(1971):it may be impossible for the information producer to credibly ensure that he has,in fact,produced the valuable information. A related problem concerns resale of the information.If an information producer could credibly produce valuable information,and then sell it to another agent,then there is no way to prevent the second 3 Winton(1995b)shows that further reductions in monitoring costs are possible if a class of"outside"equity holders is created,who are junior to depositors but senior to the banker
11 monitoring banks that fail is increasing in bank size, there is a bank size past which the increase in monitoring costs dominates marginal benefits from additional diversification. Moreover, this optimal size diminishes as the systematic component of loan risk increases. Winton (1995a) addresses another issue, namely the role of bank capital. Suppose that the banker invests his own funds in the bank as “inside” equity capital. Being junior, such equity absorbs losses first, reducing the probability with which the bank defaults and depositors must monitor. Thus, bank capital is another mechanism for implementing delegated monitoring. Since the bankers’ capital is fixed, it will be most helpful for smaller banks; also, the relative importance of capital versus diversification increases as more loan risk is systematic.3 Of course, Diamond (1984) does not explain all the characteristics of intermediaries. But, he elegantly explains the existence of intermediaries, in particular, as coalitions, of borrowers and lenders, which dominate the alternative of direct investment by investors in securities issued by firms. The securities market fails in the sense that intermediation, centralization of the task of monitoring, is a lower cost solution to the ex post information asymmetry between borrowers and lenders. Diversification is critical to intermediation providing a lower cost solution because diversification is critical to reducing the monitoring the monitor problem. The textbook idea that individual investors can diversify nonsystematic risk on their own does not take into account the role diversification plays in allowing an intermediary to be monitored costlessly (in the limit). Other papers that study banks as delegated monitors include Gorton and Haubrich (1987) and Seward (1990). C. Banks as Information Producers If information about investment opportunities is not free, then economic agents may find it worthwhile to produce such information. There will be an inefficient duplication of information production costs if multiple agents choose to produce the same information. Alternatively, a smaller number of agents could produce the information, becoming informed, and then sell the information to the uninformed agents. This, however, introduces the “reliability problem” originally identified by Hirshleifer (1971): it may be impossible for the information producer to credibly ensure that he has, in fact, produced the valuable information. A related problem concerns resale of the information. If an information producer could credibly produce valuable information, and then sell it to another agent, then there is no way to prevent the second 3 Winton (1995b) shows that further reductions in monitoring costs are possible if a class of “outside” equity holders is created, who are junior to depositors but senior to the banker
12 agent from selling it to a third agent,and so on.In other words,purchasers of the information can sell or share the information with others without necessarily diminishing its usefulness to themselves.This is known as the "appropriability problem."The returns to producing the information could not all be captured by the information producer,possibly making the production of information uneconomic(see Grossman and Stiglitz(1980)).The resale and appropriability problems in information production can motivate the existence of an intermediary. Leland and Pyle (1977)were the first to suggest that an intermediary could overcome the reliability problem.The intermediary can credibly produce information by investing its wealth in assets about which it claims to have produced valuable information.The starting point for Leland and Pyle (1977)is a single entrepreneur who has private information about an investment opportunity,but who has insufficient resources to undertake the investment.Since outside investors do not observe the entrepreneur's private information,there is an adverse selection problem.Leland and Pyle show that the entrepreneur's private information can be signaled by the fraction of equity in the project that the entrepreneur retrains,while he sells the remaining fraction to outside investors.At the end of their paper, Leland and Pyle suggest that financial intermediaries might efficiently solve the reliability and appropriability problems inherent in information production by issuing securities and using the proceeds to invest in a portfolio of securities about which the intermediary has become privately informed.After deriving his delegated monitoring model,Diamond(1984)also derives a Leland and Pyle model in which diversification lowers the intermediary's signaling costs compared to the entrepreneur's costs. Following Leland and Pyle,a number of papers,notably Campbell and Kracaw (1980),also argued that financial intermediaries might exist to produce information about potential investments, information that could not be efficiently produced in securities markets.Campbell and Kracaw(1980) show that appropriability and reliability problems can be eliminated if the information producer has a sufficient minimum amount of wealth to risk if he does not produce the information.To risk his own money requires that the intermediary actually invest on behalf of other agents.The paper,however,that most fully articulates the argument that coalitions of agents should form to produce information ex ante about potential investments is Boyd and Prescott(1986). The underlying problem faced by agents in Boyd and Prescott (1986)is an information asymmetry that occurs prior to contracting and investing,resulting in an adverse selection problem. Agents are of different types and this information is private to each agent.Each agent,however,is endowed with a technology to evaluate projects,that is,the technology can determine agent type.Ex ante information production can alleviate the adverse selection problem.This can be done in a market context, 4 See also Kihlstrom and Mathews(1990)and Duffie and Demarzo(1999)
12 agent from selling it to a third agent, and so on. In other words, purchasers of the information can sell or share the information with others without necessarily diminishing its usefulness to themselves. This is known as the “appropriability problem.” The returns to producing the information could not all be captured by the information producer, possibly making the production of information uneconomic (see Grossman and Stiglitz (1980)). The resale and appropriability problems in information production can motivate the existence of an intermediary. Leland and Pyle (1977) were the first to suggest that an intermediary could overcome the reliability problem. The intermediary can credibly produce information by investing its wealth in assets about which it claims to have produced valuable information. The starting point for Leland and Pyle (1977) is a single entrepreneur who has private information about an investment opportunity, but who has insufficient resources to undertake the investment. Since outside investors do not observe the entrepreneur’s private information, there is an adverse selection problem. Leland and Pyle show that the entrepreneur’s private information can be signaled by the fraction of equity in the project that the entrepreneur retrains, while he sells the remaining fraction to outside investors.4 At the end of their paper, Leland and Pyle suggest that financial intermediaries might efficiently solve the reliability and appropriability problems inherent in information production by issuing securities and using the proceeds to invest in a portfolio of securities about which the intermediary has become privately informed. After deriving his delegated monitoring model, Diamond (1984) also derives a Leland and Pyle model in which diversification lowers the intermediary’s signaling costs compared to the entrepreneur’s costs. Following Leland and Pyle, a number of papers, notably Campbell and Kracaw (1980), also argued that financial intermediaries might exist to produce information about potential investments, information that could not be efficiently produced in securities markets. Campbell and Kracaw (1980) show that appropriability and reliability problems can be eliminated if the information producer has a sufficient minimum amount of wealth to risk if he does not produce the information. To risk his own money requires that the intermediary actually invest on behalf of other agents. The paper, however, that most fully articulates the argument that coalitions of agents should form to produce information ex ante about potential investments is Boyd and Prescott (1986). The underlying problem faced by agents in Boyd and Prescott (1986) is an information asymmetry that occurs prior to contracting and investing, resulting in an adverse selection problem. Agents are of different types and this information is private to each agent. Each agent, however, is endowed with a technology to evaluate projects, that is, the technology can determine agent type. Ex ante information production can alleviate the adverse selection problem. This can be done in a market context, 4 See also Kihlstrom and Mathews (1990) and Duffie and Demarzo (1999)