23 Implicitly,these models seem to assume larger settings in which agents face cash-in-advance constraints and suffer shocks to income or preferences,causing them to sell securities.These models assume symmetrically that there are also noise traders who have sudden urges to buy securities,though it is less clear what the source of this urge is exactly.Intuitively,these traders either sell securities at too low a price or buy securities at too high a price because they are uninformed and the prices at which securities are traded are not fully revealing. Gorton and Pennacchi(1990)observe that these noise traders should recognize their problem, namely,that they lose money when they trade securities with better-informed traders.Consequently,they should demand securities with the property that when they are traded it is not possible for insiders to benefit at the expense of less informed traders.Thus,a security is said to be "liquid"if uninformed traders can sell it(unexpectedly)without a loss to more informed traders.The higher the variance of the value of a security,the greater the potential losses to insiders when uninformed traders must sell.If securities could be valued independently of information known only to the informed traders,then these securities would be highly desirable for trading purposes.Gorton and Pennacchi (1990)argue that splitting the cash flows of an underlying portfolio to create debt and equity can create such"liquid" securities,namely the debt.If the debt is riskless,then there can be no information advantage that other agents could possess.Uninformed agents with unexpected needs to sell securities can sell these securities to satisfy their liquidity needs.Financial intermediaries are the natural entities to create such securities,as they hold diversified portfolios of assets.Consequently,their debt should be used for transactions purposes. Holmstrom and Tirole (1998)provide another rationale for intermediaries based on a third notion of"liquidity."They begin by deriving a demand for"liquidity"that emanates from firms rather than consumers.There are three dates in their model,0,1,and 2.At date 0 the entrepreneur running the firm raises outside financing.At date 1 there is a"liquidity shock"requiring the entrepreneur to invest more in the project if it is to obtain a return at date 2.After the realization of the liquidity shock,the decision to continue or not is made,followed by the entrepreneur's effort choice.If the project is continued,then an outcome is realized at date 2 and contract payments are made.Because there is a moral hazard problem in inducing the entrepreneur to expend effort,outside investors cannot be promised the full social value of the investment.The firm raises less financing than the first-best social optimum.If the firm can store the initial resources,then it faces a dilemma.It can reduce the amount it invests at date 0,to have an amount to hedge against a liquidity shock.Or,it can invest more at date 0,but then have less on hand if it needs more at date 1. Now,suppose there is no storage and no aggregate uncertainty.The only way to transfer value across time is to use claims issued by firms.In general equilibrium,some firms will need resources at
23 Implicitly, these models seem to assume larger settings in which agents face cash-in-advance constraints and suffer shocks to income or preferences, causing them to sell securities. These models assume symmetrically that there are also noise traders who have sudden urges to buy securities, though it is less clear what the source of this urge is exactly. Intuitively, these traders either sell securities at too low a price or buy securities at too high a price because they are uninformed and the prices at which securities are traded are not fully revealing. Gorton and Pennacchi (1990) observe that these noise traders should recognize their problem, namely, that they lose money when they trade securities with better-informed traders. Consequently, they should demand securities with the property that when they are traded it is not possible for insiders to benefit at the expense of less informed traders. Thus, a security is said to be “liquid” if uninformed traders can sell it (unexpectedly) without a loss to more informed traders. The higher the variance of the value of a security, the greater the potential losses to insiders when uninformed traders must sell. If securities could be valued independently of information known only to the informed traders, then these securities would be highly desirable for trading purposes. Gorton and Pennacchi (1990) argue that splitting the cash flows of an underlying portfolio to create debt and equity can create such “liquid” securities, namely the debt. If the debt is riskless, then there can be no information advantage that other agents could possess. Uninformed agents with unexpected needs to sell securities can sell these securities to satisfy their liquidity needs. Financial intermediaries are the natural entities to create such securities, as they hold diversified portfolios of assets. Consequently, their debt should be used for transactions purposes. Holmström and Tirole (1998) provide another rationale for intermediaries based on a third notion of “liquidity.” They begin by deriving a demand for “liquidity” that emanates from firms rather than consumers. There are three dates in their model, 0, 1, and 2. At date 0 the entrepreneur running the firm raises outside financing. At date 1 there is a “liquidity shock” requiring the entrepreneur to invest more in the project if it is to obtain a return at date 2. After the realization of the liquidity shock, the decision to continue or not is made, followed by the entrepreneur’s effort choice. If the project is continued, then an outcome is realized at date 2 and contract payments are made. Because there is a moral hazard problem in inducing the entrepreneur to expend effort, outside investors cannot be promised the full social value of the investment. The firm raises less financing than the first-best social optimum. If the firm can store the initial resources, then it faces a dilemma. It can reduce the amount it invests at date 0, to have an amount to hedge against a liquidity shock. Or, it can invest more at date 0, but then have less on hand if it needs more at date 1. Now, suppose there is no storage and no aggregate uncertainty. The only way to transfer value across time is to use claims issued by firms. In general equilibrium, some firms will need resources at
24 date 1 and some will not.A second-best arrangement would allow firms with large needs for resources at date I to utilize the market value of those firms with low needs at date 1.How would this actually work? A firm with a liquidity shock at date I cannot meet its needs by selling claims at date 1;it is too late to do that.Could the firm instead hedge against an adverse liquidity shock at date 1 by buying claims on other firms at date 0,and then selling those claims at date 1?There are two problems with this arrangement. First,if the moral hazard problem is severe enough,then a market in firm claims will not supply enough "liquidity."Second,there is an inefficient distribution of liquid assets(the claims that can be sold at date 1 by firms needing resources).Firms without adverse liquidity shocks end up holding claims at date 1 that they do not need. An intermediary can provide liquidity by issuing claims to investors at date 0 on its value at date 2.At date 0 it contracts with each firm to provide a line of credit at date 1.The maximum credit line is incentive compatible with the entrepreneur making an effort.Unlike claims in the financial market, which cannot be made contingent on a firm's liquidity shock,firms only draw on the credit line at date I to the extent that they need resources.If there is aggregate uncertainty,then this arrangement may not work,and there can be a role for a government bond market. Both Gorton and Pennacchi (1990)and Holmstrom and Tirole (1998)have intermediaries creating securities that have desirable state contingent payoffs.In Gorton and Pennacchi,the bank creates demand deposits whose value does not depend on the state of the world.This security is in demand because its value is not state contingent and,therefore,uninformed traders will not lose to better-informed traders who know the state of the world.In Holmstrom and Tirole,the intermediary creates a security, the credit line,which is valuable because it is state contingent;it is only drawn on when a firm needs resources at the interim date.Capital market securities issued by firms cannot replicate this state- contingent payoff. F.Banks as Commitment Mechanisms An important question concerns why illiquid bank assets are financed by demand deposits that allow consumers to arrive and demand liquidation of those illiquid assets.Calomiris and Kahn(1991) and Flannery(1994)link the fragility of bank capital structures to the role of banks.These authors begin with the assumption that banks are somewhat opaque institutions,more so than nonfinancial firms. Evidence for this opaqueness compared to nonfinancial firms can be found,for example,in Slovin, Sushka,and Polonchek(1992)and Morgan(2000).Calomiris and Kahn(1991)argue that bank demand 6 Because the entrepreneur pays a fee at date 0 for the credit line,he can borrow at a lower rate at date 1 than if he issued securities at date 1.This lower rate leads to greater effort for any amount of borrowing
24 date 1 and some will not. A second-best arrangement would allow firms with large needs for resources at date 1 to utilize the market value of those firms with low needs at date 1. How would this actually work? A firm with a liquidity shock at date 1 cannot meet its needs by selling claims at date 1; it is too late to do that. Could the firm instead hedge against an adverse liquidity shock at date 1 by buying claims on other firms at date 0, and then selling those claims at date 1? There are two problems with this arrangement. First, if the moral hazard problem is severe enough, then a market in firm claims will not supply enough “liquidity.” Second, there is an inefficient distribution of liquid assets (the claims that can be sold at date 1 by firms needing resources). Firms without adverse liquidity shocks end up holding claims at date 1 that they do not need. An intermediary can provide liquidity by issuing claims to investors at date 0 on its value at date 2. At date 0 it contracts with each firm to provide a line of credit at date 1. The maximum credit line is incentive compatible with the entrepreneur making an effort.6 Unlike claims in the financial market, which cannot be made contingent on a firm’s liquidity shock, firms only draw on the credit line at date 1 to the extent that they need resources. If there is aggregate uncertainty, then this arrangement may not work, and there can be a role for a government bond market. Both Gorton and Pennacchi (1990) and Holmström and Tirole (1998) have intermediaries creating securities that have desirable state contingent payoffs. In Gorton and Pennacchi, the bank creates demand deposits whose value does not depend on the state of the world. This security is in demand because its value is not state contingent and, therefore, uninformed traders will not lose to better-informed traders who know the state of the world. In Holmström and Tirole, the intermediary creates a security, the credit line, which is valuable because it is state contingent; it is only drawn on when a firm needs resources at the interim date. Capital market securities issued by firms cannot replicate this statecontingent payoff. F. Banks as Commitment Mechanisms An important question concerns why illiquid bank assets are financed by demand deposits that allow consumers to arrive and demand liquidation of those illiquid assets. Calomiris and Kahn (1991) and Flannery (1994) link the fragility of bank capital structures to the role of banks. These authors begin with the assumption that banks are somewhat opaque institutions, more so than nonfinancial firms. Evidence for this opaqueness compared to nonfinancial firms can be found, for example, in Slovin, Sushka, and Polonchek (1992) and Morgan (2000). Calomiris and Kahn (1991) argue that bank demand 6 Because the entrepreneur pays a fee at date 0 for the credit line, he can borrow at a lower rate at date 1 than if he issued securities at date 1. This lower rate leads to greater effort for any amount of borrowing
25 deposits include the right to withdraw at anytime at par along with a sequential service constraint in order to control the risk taking activities of bankers.If information about the banker's decisions must be produced at a cost,then individual depositors who expend resources to produce the information will get into line to withdraw at the bank first.Because the sequential service constraint is a first-come-first- served rule,it rewards those depositors in line first,and so information-producing depositors will recover more than other depositors.This argument was the first to suggest that banks'capital structures are deliberately made fragile so as to commit to not engaging in certain activities.From this viewpoint, fragility is a positive attribute of banks.Jean-Baptiste (1999)also argues that the instantly callable feature of demand deposits is necessary as a device to discipline bankers. Flannery (1994)makes a related argument.He argues that bank creditors cannot effectively control bank asset substitution because of the ease of flexibly altering the bank portfolio,but they can estimate a bank's riskiness at any point in time.To control bankers,short-term debt is used because changes in bank risk will be reflected in financing costs.Again,the basic point is that the capital structure of banks is designed to be fragile,so that it functions as a commitment mechanism.Flannery and Sorescu(1996)show empirically that bank debt prices do reflect bank risk Diamond and Rajan (2001)use this idea that fragility is a commitment device to construct a model of bank-like financial intermediation.In their model,entrepreneurs need to raise money from outside investors to finance their projects.The specific abilities of the entrepreneur are important for the project to generate high cash flows,that is,if the entrepreneur refuses to work,then the project is worth less when someone else runs it.Moreover,the entrepreneur cannot commit to stay with the project.A lender,however,can build a relationship by lending to entrepreneur and learning about the project.If this relationship lender "liquidates"the project by separating the entrepreneur from the project,then the project is worth less than it would be worth with the entrepreneur,but more than if it is run by someone other than the relationship lender. Because the entrepreneur cannot commit to stay with the asset,Diamond and Rajan say that the asset is"illiquid."This "illiquidity"makes it possible for the entrepreneur to hold up the relationship lender.Because potential relationship lenders anticipate this holdup problem,the amount that the entrepreneur can borrow is limited.Lenders also have problems because they may face a realized liquidity shock at an interim date.If a relationship lender needed cash at the interim date,the project would have to be sold to a non-relationship investor in whose hands it is worth even less.The prospect of such a shock makes relationship lending expensive,if not prohibitive. The consequences of this chain of illiquidity could be mitigated if the relationship lender could borrow against the full value of the loan when faced with a liquidity shock.But this requires that the relationship lender commit to not separate from the project in the future.Diamond and Rajan argue that a
25 deposits include the right to withdraw at anytime at par along with a sequential service constraint in order to control the risk taking activities of bankers. If information about the banker’s decisions must be produced at a cost, then individual depositors who expend resources to produce the information will get into line to withdraw at the bank first. Because the sequential service constraint is a first-come-firstserved rule, it rewards those depositors in line first, and so information-producing depositors will recover more than other depositors. This argument was the first to suggest that banks’ capital structures are deliberately made fragile so as to commit to not engaging in certain activities. From this viewpoint, fragility is a positive attribute of banks. Jean-Baptiste (1999) also argues that the instantly callable feature of demand deposits is necessary as a device to discipline bankers. Flannery (1994) makes a related argument. He argues that bank creditors cannot effectively control bank asset substitution because of the ease of flexibly altering the bank portfolio, but they can estimate a bank’s riskiness at any point in time. To control bankers, short-term debt is used because changes in bank risk will be reflected in financing costs. Again, the basic point is that the capital structure of banks is designed to be fragile, so that it functions as a commitment mechanism. Flannery and Sorescu (1996) show empirically that bank debt prices do reflect bank risk. Diamond and Rajan (2001) use this idea that fragility is a commitment device to construct a model of bank-like financial intermediation. In their model, entrepreneurs need to raise money from outside investors to finance their projects. The specific abilities of the entrepreneur are important for the project to generate high cash flows; that is, if the entrepreneur refuses to work, then the project is worth less when someone else runs it. Moreover, the entrepreneur cannot commit to stay with the project. A lender, however, can build a relationship by lending to entrepreneur and learning about the project. If this relationship lender “liquidates” the project by separating the entrepreneur from the project, then the project is worth less than it would be worth with the entrepreneur, but more than if it is run by someone other than the relationship lender. Because the entrepreneur cannot commit to stay with the asset, Diamond and Rajan say that the asset is “illiquid.” This “illiquidity” makes it possible for the entrepreneur to hold up the relationship lender. Because potential relationship lenders anticipate this holdup problem, the amount that the entrepreneur can borrow is limited. Lenders also have problems because they may face a realized liquidity shock at an interim date. If a relationship lender needed cash at the interim date, the project would have to be sold to a non-relationship investor in whose hands it is worth even less. The prospect of such a shock makes relationship lending expensive, if not prohibitive. The consequences of this chain of illiquidity could be mitigated if the relationship lender could borrow against the full value of the loan when faced with a liquidity shock. But this requires that the relationship lender commit to not separate from the project in the future. Diamond and Rajan argue that a
26 bank can achieve such a commitment by designing a fragile capital structure,as follows.If the relationship lender issues demand deposits that are subject to collective action problems among the depositors,then if the relationship lender threatens to withdraw from the project,depositors will run the bank and the relationship lender will receive no rents. As Diamond and Rajan note,this fragile structure is not first-best if banks face undiversifiable liquidity shocks.In this case,runs may occur because of high liquidity demand rather than because of bank moral hazard.Diamond and Rajan(2000)use this problem to motivate the existence and optimal level of bank equity capital.We return to this point in Section IlI below. G.Empirical Tests of Bank Existence Theories Theories of the existence of bank-like financial intermediaries link banks'activities on the asset side of their balance sheets with the unique liabilities that banks issue on the liability side of their balance sheets.Such a link is important for establishing what it is that banks do that cannot be replicated in capital markets.As we have seen,these arguments take two linked forms.First,the banks'balance sheet structure may ensure that the bank has incentive to act as delegated monitor or information producer. Second,by virtue of holding a diversified portfolio of loans,banks are in the best position to create riskless trading securities,namely,demand deposits. Two papers,in particular,construct empirical tests of hypotheses about links between the two sides of bank balance sheets.Berlin and Mester(1999)look for a link between bank market power in deposits markets and the types of loan contracts that the bank enters into with borrowers.?"Core deposits"are those deposits,demand deposits and savings deposits,which are mostly interest rate inelastic.To the extent that a bank has such core deposits,it can safely engage in long-term contracts with borrowers;in particular,it can smooth loan rates.Using a large sample of loans from the Federal Reserve's Survey of Terms of Bank Lending to Business,they find that banks that are more heavily funded through core deposits do provide borrowers with smoother loan rates in response to aggregate shocks. Kashyap,Rajan,and Stein(2001)empirically analyze the link between loan commitments and demand deposits.While demand deposits are liabilities and loan commitments are assets,the two securities both commit the bank to potentially meet demands for cash.That is,depositors may withdraw their deposits and borrowers may draw on their loan commitments.To be prepared for such contingencies,each of these security types requires the bank to hold liquidity.As long as the demands for 7 Hannan and Berger(1991)and Neumark and Sharpe(1992)provide evidence of bank monopoly power in the retail deposit market
26 bank can achieve such a commitment by designing a fragile capital structure, as follows. If the relationship lender issues demand deposits that are subject to collective action problems among the depositors, then if the relationship lender threatens to withdraw from the project, depositors will run the bank and the relationship lender will receive no rents. As Diamond and Rajan note, this fragile structure is not first-best if banks face undiversifiable liquidity shocks. In this case, runs may occur because of high liquidity demand rather than because of bank moral hazard. Diamond and Rajan (2000) use this problem to motivate the existence and optimal level of bank equity capital. We return to this point in Section III below. G. Empirical Tests of Bank Existence Theories Theories of the existence of bank-like financial intermediaries link banks’ activities on the asset side of their balance sheets with the unique liabilities that banks issue on the liability side of their balance sheets. Such a link is important for establishing what it is that banks do that cannot be replicated in capital markets. As we have seen, these arguments take two linked forms. First, the banks’ balance sheet structure may ensure that the bank has incentive to act as delegated monitor or information producer. Second, by virtue of holding a diversified portfolio of loans, banks are in the best position to create riskless trading securities, namely, demand deposits. Two papers, in particular, construct empirical tests of hypotheses about links between the two sides of bank balance sheets. Berlin and Mester (1999) look for a link between bank market power in deposits markets and the types of loan contracts that the bank enters into with borrowers.7 “Core deposits” are those deposits, demand deposits and savings deposits, which are mostly interest rate inelastic. To the extent that a bank has such core deposits, it can safely engage in long-term contracts with borrowers; in particular, it can smooth loan rates. Using a large sample of loans from the Federal Reserve’s Survey of Terms of Bank Lending to Business, they find that banks that are more heavily funded through core deposits do provide borrowers with smoother loan rates in response to aggregate shocks. Kashyap, Rajan, and Stein (2001) empirically analyze the link between loan commitments and demand deposits. While demand deposits are liabilities and loan commitments are assets, the two securities both commit the bank to potentially meet demands for cash. That is, depositors may withdraw their deposits and borrowers may draw on their loan commitments. To be prepared for such contingencies, each of these security types requires the bank to hold liquidity. As long as the demands for 7 Hannan and Berger (1991) and Neumark and Sharpe (1992) provide evidence of bank monopoly power in the retail deposit market
27 cash on loan commitments and on deposits are not perfectly correlated,there are economies of scale to holding cash against both types of contingencies.They find that banks make more loan commitments than other types of intermediaries and that,within the banking sector,banks with high ratios of transaction deposit to total deposits also have high ratios of loan commitments to total loans. The dramatic increase in loan sales constitutes a challenge,both theoretically and empirically,to arguments concerning bank existence.In a loan sale,the cash flows from a loan on a bank's balance sheet are sold to investors in the capital markets,through issuance of a new security (a secondary loan participation).This seems paradoxical:the borrowing firm could have issued a security directly to the same investor in the capital markets without going to the bank,and yet chose to borrow from a bank.The above arguments for the existence of financial intermediation imply that the bank loan should not be resold because if it can be resold there is no incentive for the bank to screen ex ante or monitor ex post Gorton and Pennacchi (1995)explore these issues empirically,testing for the presence of incentive- compatible arrangements that could explain loan sales.One of their main findings is that the bank keeps a portion of the cash flows that is consistent with maintaining incentives.The idea is that the bank faces the same incentives as it would have had the entire loan been kept on its balance sheet.There are now a number of papers on this subject,but the basic paradox of loan sales remains unexplained.Indeed,the paradox is somewhat deepened to the extent that banks can transfer the credit risk of their loans to third parties via credit default swaps.Market participants seem to rely on banks'incentives to maintain their reputations for monitoring,but the efficacy of this mechanism is largely unexplored. H.Bonds versus Loans If banks monitor borrowers in ways that cannot be accomplished by dispersed bondholders,or produce information that capital markets investors cannot produce,then how can bonds and loans coexist?Why don't loans dominate bonds?This poses the question of the existence of bank loans in a different light.A number of authors have addressed this issue,attempting to differentiate between bonds and loans,in terms of their characteristics,but also in such a way that firms will demand both. Detragiache (1994)presents a model in which firms use both bonds and loans.Bonds (or synonymously "public debt")cannot be renegotiated,while loans(synonymously "private debt")can be costlessly renegotiated.Loans are senior to bonds.Equity holders face an asset substitution problem at the initial date,and renegotiation with the bank or liquidation may occur at the interim date.In renegotiation only the senior lender,the bank,can forgive debt,so bank debt has a clear advantage.But 8 The idea that dispersed lenders cannot renegotiate effectively compared to a single lender,like a bank,is commonly assumed.Bolton and Scharfstein(1996)provide the theoretical foundation for this notion
27 cash on loan commitments and on deposits are not perfectly correlated, there are economies of scale to holding cash against both types of contingencies. They find that banks make more loan commitments than other types of intermediaries and that, within the banking sector, banks with high ratios of transaction deposit to total deposits also have high ratios of loan commitments to total loans. The dramatic increase in loan sales constitutes a challenge, both theoretically and empirically, to arguments concerning bank existence. In a loan sale, the cash flows from a loan on a bank’s balance sheet are sold to investors in the capital markets, through issuance of a new security (a secondary loan participation). This seems paradoxical: the borrowing firm could have issued a security directly to the same investor in the capital markets without going to the bank, and yet chose to borrow from a bank. The above arguments for the existence of financial intermediation imply that the bank loan should not be resold because if it can be resold there is no incentive for the bank to screen ex ante or monitor ex post. Gorton and Pennacchi (1995) explore these issues empirically, testing for the presence of incentivecompatible arrangements that could explain loan sales. One of their main findings is that the bank keeps a portion of the cash flows that is consistent with maintaining incentives. The idea is that the bank faces the same incentives as it would have had the entire loan been kept on its balance sheet. There are now a number of papers on this subject, but the basic paradox of loan sales remains unexplained. Indeed, the paradox is somewhat deepened to the extent that banks can transfer the credit risk of their loans to third parties via credit default swaps. Market participants seem to rely on banks’ incentives to maintain their reputations for monitoring, but the efficacy of this mechanism is largely unexplored. H. Bonds versus Loans If banks monitor borrowers in ways that cannot be accomplished by dispersed bondholders, or produce information that capital markets investors cannot produce, then how can bonds and loans coexist?8 Why don’t loans dominate bonds? This poses the question of the existence of bank loans in a different light. A number of authors have addressed this issue, attempting to differentiate between bonds and loans, in terms of their characteristics, but also in such a way that firms will demand both. Detragiache (1994) presents a model in which firms use both bonds and loans. Bonds (or synonymously “public debt”) cannot be renegotiated, while loans (synonymously “private debt”) can be costlessly renegotiated. Loans are senior to bonds. Equity holders face an asset substitution problem at the initial date, and renegotiation with the bank or liquidation may occur at the interim date. In renegotiation only the senior lender, the bank, can forgive debt, so bank debt has a clear advantage. But 8 The idea that dispersed lenders cannot renegotiate effectively compared to a single lender, like a bank, is commonly assumed. Bolton and Scharfstein (1996) provide the theoretical foundation for this notion