18 Section IV below.Here,we simply note that the idea it attempts to capture is that consumers cannot coordinate to go to any securities market at the same time to trade;they are busy doing other things such as shopping,eating,sleeping,working,etc.Thus,Diamond and Dybvig's assumptions that demand deposits cannot be traded and that no other securities markets are open are not completely without foundation. This point is important because Jacklin (1987)and Haubrich and King (1990)argue that the existence of Diamond and Dybvig intermediaries requires the restriction that consumers only have nontraded demand deposits available to them.Jacklin (1987)begins by asking,why does a securities market fail in the Diamond and Dybvig model?In order to highlight the importance of trading restrictions and preferences for Diamond and Dybvig's result that intermediation is the best insurance arrangement,he proposes an alternative arrangement that uses traded securities.Suppose that there are firms in Diamond and Dybvig that own the two-period production technology.Each firm raises capital by issuing dividend-paying shares at date 0.Consumers buy the shares,entitling them to set the production policy and to set dividend policy about the amount paid out to share owners at date 1.The "dividends"on Jacklin's equity are set to smooth income in exactly the desired way;they are not just pass-throughs from the firms.Shareholders of record at date 0 receive the dividend at date 1 and then can sell the share in a share market at date 1.At date 1 consumers learn their preferred consumption streams. Early consumers will want sell their shares ex dividend to late consumers.Jacklin shows that the social optimum obtains with this share market in place.Thus,the bank cannot do any better. Jacklin goes on to show that the result that the intermediary cannot improve upon trading dividend-paying shares is not true in general.Recall that,in Diamond and Dybvig,some consumers find that they must consume early;it is all or nothing.If instead preferences are smooth,so that one type of consumer will learn that he has a stronger preference for earlier consumption than the other type,then it can happen that demand deposits dominate traded equity shares,but only under certain conditions. Furthermore,if demand deposits can be traded,then optimal risk sharing does not occur regardless of preferences.Finally,Jacklin argues that,if new assets can be introduced,individuals will deviate from either the demand deposit arrangement or from the economy with traded dividend-paying shares.These points lead Jacklin to conclude that the Diamond and Dybvig"demand-deposit"intermediary can only exist if trading restrictions limit consumers to the type of demand deposits that Diamond and Dybvig model.This highlights the importance of the sequential service constraint and its interpretation. Haubrich and King(1990)revisit in detail the issue of financial intermediation in settings where agents are subject to privately observable income shocks.Their main conclusions are similar to Jacklin (1987),namely,that "demand deposits uniquely provide insurance only if there are restrictions on financial side exchanges,which may be interpreted as exclusivity provisions or regulations on security
18 Section IV below. Here, we simply note that the idea it attempts to capture is that consumers cannot coordinate to go to any securities market at the same time to trade; they are busy doing other things such as shopping, eating, sleeping, working, etc. Thus, Diamond and Dybvig’s assumptions that demand deposits cannot be traded and that no other securities markets are open are not completely without foundation. This point is important because Jacklin (1987) and Haubrich and King (1990) argue that the existence of Diamond and Dybvig intermediaries requires the restriction that consumers only have nontraded demand deposits available to them. Jacklin (1987) begins by asking, why does a securities market fail in the Diamond and Dybvig model? In order to highlight the importance of trading restrictions and preferences for Diamond and Dybvig’s result that intermediation is the best insurance arrangement, he proposes an alternative arrangement that uses traded securities. Suppose that there are firms in Diamond and Dybvig that own the two-period production technology. Each firm raises capital by issuing dividend-paying shares at date 0. Consumers buy the shares, entitling them to set the production policy and to set dividend policy about the amount paid out to share owners at date 1. The “dividends” on Jacklin’s equity are set to smooth income in exactly the desired way; they are not just pass-throughs from the firms. Shareholders of record at date 0 receive the dividend at date 1 and then can sell the share in a share market at date 1. At date 1 consumers learn their preferred consumption streams. Early consumers will want sell their shares ex dividend to late consumers. Jacklin shows that the social optimum obtains with this share market in place. Thus, the bank cannot do any better. Jacklin goes on to show that the result that the intermediary cannot improve upon trading dividend-paying shares is not true in general. Recall that, in Diamond and Dybvig, some consumers find that they must consume early; it is all or nothing. If instead preferences are smooth, so that one type of consumer will learn that he has a stronger preference for earlier consumption than the other type, then it can happen that demand deposits dominate traded equity shares, but only under certain conditions. Furthermore, if demand deposits can be traded, then optimal risk sharing does not occur regardless of preferences. Finally, Jacklin argues that, if new assets can be introduced, individuals will deviate from either the demand deposit arrangement or from the economy with traded dividend-paying shares. These points lead Jacklin to conclude that the Diamond and Dybvig “demand-deposit” intermediary can only exist if trading restrictions limit consumers to the type of demand deposits that Diamond and Dybvig model. This highlights the importance of the sequential service constraint and its interpretation. Haubrich and King (1990) revisit in detail the issue of financial intermediation in settings where agents are subject to privately observable income shocks. Their main conclusions are similar to Jacklin (1987), namely, that “demand deposits uniquely provide insurance only if there are restrictions on financial side exchanges, which may be interpreted as exclusivity provisions or regulations on security
19 markets.If these restrictions cannot be implemented,then our environment does not rationalize banks" (p.362;emphasis in original).They also make the useful distinction between two separate issues.One is the fact that the available investment technology is illiquid in the sense that no return is earned if the two- period investment is ended early.The other is that risk averse consumers with privately observable income shocks have a demand for insurance.They argue that a securities market is as good as banks in providing liquidity.In their model,the bank's comparative advantage is in providing insurance against private income shocks rather than providing liquidity per se,but that advantage still depends on trading restrictions Hellwig (1994)and von Thadden (1998)examine how banks function when additional considerations are introduced into Diamond and Dybvig's structure.Hellwig shows that if market returns at the interim (early-consumption)date are subject to systematic "interest-rate shocks,"banks optimally do not provide insurance against such interest rate risk.Von Thadden shows that if depositors can join outside coalitions that engage in market activity,banks'ability to provide insurance is severely curtailed, and banks are more constrained as long-term investment opportunities are more reversible.Intuitively,ex ante insurance makes the return to holding deposits at the interim date deviate from returns available by directly investing,allowing arbitrage. Diamond(1997)responds to Jacklin (1987),Haubrich and King (1990),Hellwig(1994),and von Thadden (1998)in a model with both banks and a securities market in which (by assumption)only a limited subset of agents participate in the market.The main focus of the paper is on the interactions between bank provision of"liquidity"and the depth of the market.As more agents participate in the securities market,banks are less able to provide additional liquidity. Allen and Gale(1997)introduce a different smoothing role for financial intermediaries,namely, that they are unique in providing a mechanism for smoothing intertemporal intergenerational risks.Allen and Gale study the standard overlapping(risk averse)generations model with two assets,a risky asset in fixed supply and a safe asset that can be accumulated over time.>The risky asset lasts forever and pays out a random dividend each period.The safe asset consists of a storage technology.First,consider the market equilibrium in this economy.Perhaps counterintuitively,the safe asset is not a useful hedge against the uncertainty generated by the risky asset.Because the risky asset's returns are independently and identically distributed in any period,a representative young agent solves the same decision problem at any date.Old agents supply the risky asset inelastically,so the equilibrium price of the risky asset is 5 Freeman(19)and Qi(199)introduce Diamond and Dybvig banks into an overlapping generations model,but do not consider intertemporal smoothing of risk
19 markets. If these restrictions cannot be implemented, then our environment does not rationalize banks” (p. 362; emphasis in original). They also make the useful distinction between two separate issues. One is the fact that the available investment technology is illiquid in the sense that no return is earned if the twoperiod investment is ended early. The other is that risk averse consumers with privately observable income shocks have a demand for insurance. They argue that a securities market is as good as banks in providing liquidity. In their model, the bank’s comparative advantage is in providing insurance against private income shocks rather than providing liquidity per se, but that advantage still depends on trading restrictions. Hellwig (1994) and von Thadden (1998) examine how banks function when additional considerations are introduced into Diamond and Dybvig’s structure. Hellwig shows that if market returns at the interim (early-consumption) date are subject to systematic “interest-rate shocks,” banks optimally do not provide insurance against such interest rate risk. Von Thadden shows that if depositors can join outside coalitions that engage in market activity, banks’ ability to provide insurance is severely curtailed, and banks are more constrained as long-term investment opportunities are more reversible. Intuitively, ex ante insurance makes the return to holding deposits at the interim date deviate from returns available by directly investing, allowing arbitrage. Diamond (1997) responds to Jacklin (1987), Haubrich and King (1990), Hellwig (1994), and von Thadden (1998) in a model with both banks and a securities market in which (by assumption) only a limited subset of agents participate in the market. The main focus of the paper is on the interactions between bank provision of “liquidity” and the depth of the market. As more agents participate in the securities market, banks are less able to provide additional liquidity. Allen and Gale (1997) introduce a different smoothing role for financial intermediaries, namely, that they are unique in providing a mechanism for smoothing intertemporal intergenerational risks. Allen and Gale study the standard overlapping (risk averse) generations model with two assets, a risky asset in fixed supply and a safe asset that can be accumulated over time.5 The risky asset lasts forever and pays out a random dividend each period. The safe asset consists of a storage technology. First, consider the market equilibrium in this economy. Perhaps counterintuitively, the safe asset is not a useful hedge against the uncertainty generated by the risky asset. Because the risky asset’s returns are independently and identically distributed in any period, a representative young agent solves the same decision problem at any date. Old agents supply the risky asset inelastically, so the equilibrium price of the risky asset is 5 Freeman (1988) and Qi (1994) introduce Diamond and Dybvig banks into an overlapping generations model, but do not consider intertemporal smoothing of risk
20 constant and nonstochastic.Because the dividend is nonnegative,the safe asset is dominated and is not held in equilibrium. This market equilibrium is in contrast with the portfolio allocation that would occur for an infinitely lived agent facing the same investment opportunities.Such an individual can self-insure against low dividend periods by holding a buffer stock of precautionary savings in the form of the safe asset. Intuitively,when the dividend is high,the individual saves some of the dividend for a"rainy day"when the dividend is low.In the overlapping generations setting,a social planner can make a Pareto improvement by following the same type of rule. The market equilibrium in the overlapping generations model cannot achieve the allocation that the social planner could achieve because private agents cannot trade before they are born,while the social planner can,in effect,trade at all dates.In particular,the social planner trades ex ante,that is,before the realization of the path of dividends.A representative young agent,however,is born into a world where the dividend has just been realized.There is no willingness to implement insurance once the state is known.For example,suppose the dividend just realized is low.Then,the social planner would like to implement a transfer from the young to the old,to smooth their income.On the other hand,if the dividend just realized is high,then the social planner would like to transfer some of that to the current young.Some excess may be saved for the next period.These transfers insure that each generation receives the expected utility targeted by the social planner. It is well known that markets are incomplete in overlapping generations models,but the point made by Allen and Gale is that a long-lived financial intermediary may be the institutional mechanism to provide for this intertemporal smoothing.The intermediary would hold all the assets and offer a deposit contract to each generation.After accumulating large reserves,the intermediary offers (almost)all generations a constant return on deposits,independent of the actual dividend realizations.How such an institution would be set up initially,and how it would be maintained when some agents will have incentives to renege on the arrangement,are not clear.Allen and Gale loosely interpret the institution as corresponding to German universal banks. Consumption insurance that implements smooth patterns of intertemporal consumption plans is at the center of the model of consumer behavior of neoclassical economics.Another central notion concerns the use of"money"to facilitate exchange.The search models of money or models with cash-in-advance constraints attempt to explain why "money"exists.The notion of banks as consumption-smoothing institutions attempts to wed these two ideas.Bank liabilities are seen as claims that facilitate consumption smoothing.But,there is no notion of exchange in the model,no sense in which transactions are taking place where bank "money"is being used to facilitate the smoothing.Instead,agents are essentially isolated from each other;there is no trade with other agents where "money"buys goods
20 constant and nonstochastic. Because the dividend is nonnegative, the safe asset is dominated and is not held in equilibrium. This market equilibrium is in contrast with the portfolio allocation that would occur for an infinitely lived agent facing the same investment opportunities. Such an individual can self-insure against low dividend periods by holding a buffer stock of precautionary savings in the form of the safe asset. Intuitively, when the dividend is high, the individual saves some of the dividend for a “rainy day” when the dividend is low. In the overlapping generations setting, a social planner can make a Pareto improvement by following the same type of rule. The market equilibrium in the overlapping generations model cannot achieve the allocation that the social planner could achieve because private agents cannot trade before they are born, while the social planner can, in effect, trade at all dates. In particular, the social planner trades ex ante, that is, before the realization of the path of dividends. A representative young agent, however, is born into a world where the dividend has just been realized. There is no willingness to implement insurance once the state is known. For example, suppose the dividend just realized is low. Then, the social planner would like to implement a transfer from the young to the old, to smooth their income. On the other hand, if the dividend just realized is high, then the social planner would like to transfer some of that to the current young. Some excess may be saved for the next period. These transfers insure that each generation receives the expected utility targeted by the social planner. It is well known that markets are incomplete in overlapping generations models, but the point made by Allen and Gale is that a long-lived financial intermediary may be the institutional mechanism to provide for this intertemporal smoothing. The intermediary would hold all the assets and offer a deposit contract to each generation. After accumulating large reserves, the intermediary offers (almost) all generations a constant return on deposits, independent of the actual dividend realizations. How such an institution would be set up initially, and how it would be maintained when some agents will have incentives to renege on the arrangement, are not clear. Allen and Gale loosely interpret the institution as corresponding to German universal banks. Consumption insurance that implements smooth patterns of intertemporal consumption plans is at the center of the model of consumer behavior of neoclassical economics. Another central notion concerns the use of “money” to facilitate exchange. The search models of money or models with cash-in-advance constraints attempt to explain why “money” exists. The notion of banks as consumption-smoothing institutions attempts to wed these two ideas. Bank liabilities are seen as claims that facilitate consumption smoothing. But, there is no notion of exchange in the model, no sense in which transactions are taking place where bank “money” is being used to facilitate the smoothing. Instead, agents are essentially isolated from each other; there is no trade with other agents where “money” buys goods
21 Rather agents fear missing out on long-term investment opportunities because of possible shocks to their preferences.Agents trade only with the bank. E. Banks as Liquidity Providers Bank liabilities function as a medium of exchange.This basic observation leads to ideas and models concerning "liquidity"that are quite distinct and perhaps more natural than viewing bank liabilities as allowing consumption smoothing.A medium of exchange is a set of claims or securities that can be offered to other agents in exchange for goods.Such claims can dominate barter and may dominate government-supplied money.What are the advantages of privately-produced trading claims to be a medium of exchange?One class of these models considers settings where agents cannot contract and trade with each other due their inability to meet at a single location.Without"money"they must barter, and this is clearly inefficient.This generates a need for a payments system,essentially a trading center or bank that can produce and net claims.A second notion of liquidity is related to the information properties of claims that are privately produced as a medium of exchange.The focus is on reducing trading losses that agents who need to consume face when other traders with private information seek to use this information to make trading profits.Yet a third notion of liquidity uses a setting where moral hazard problems limit firms'ability to borrow to meet unexpected investment needs.Because moral hazard limits the effectiveness of transactions between firms with excess liquidity and firms that need liquidity,a bank that provides contingent liquidity to those that need it can dominate a decentralized market The first view of banks as liquidity providers concerns the role of banks in the payments system. Freeman (1996a,b)models an environment where agents are spatially separated and the timing of transactions is such that they cannot simultaneously trade at a central location.The problem in the model is that some agents,buyers,wish to consume goods from other agents,but have no goods that the buyers want to offer these sellers in exchange.Nor do buyers have any money,though later at another location they will be able to sell their goods in exchange for money.So,buyers issue i.o.u.'s-promises to pay at the central location next period with fiat money-to the sellers.Fiat money is used to settle the debts, but money and private debt coexist.Now,at the central clearing location it may happen that all creditors and debtors arrive simultaneously,in which case clearing occurs directly.If arrival is not simultaneous, however,settling can take place through a clearinghouse.The clearinghouse accepts money in payment of i.o.u.'s and pays off i.o.u.'s presented.However,if creditors arrive first,then the clearinghouse must have some means of paying them before the debtors arrive.A basic point of Freeman is develop the notion of the clearinghouse issuing its own i.o.u.'s,bank notes for example,that can circulate and be redeemed for fiat money later
21 Rather agents fear missing out on long-term investment opportunities because of possible shocks to their preferences. Agents trade only with the bank. E. Banks as Liquidity Providers Bank liabilities function as a medium of exchange. This basic observation leads to ideas and models concerning “liquidity” that are quite distinct and perhaps more natural than viewing bank liabilities as allowing consumption smoothing. A medium of exchange is a set of claims or securities that can be offered to other agents in exchange for goods. Such claims can dominate barter and may dominate government-supplied money. What are the advantages of privately-produced trading claims to be a medium of exchange? One class of these models considers settings where agents cannot contract and trade with each other due their inability to meet at a single location. Without “money” they must barter, and this is clearly inefficient. This generates a need for a payments system, essentially a trading center or bank that can produce and net claims. A second notion of liquidity is related to the information properties of claims that are privately produced as a medium of exchange. The focus is on reducing trading losses that agents who need to consume face when other traders with private information seek to use this information to make trading profits. Yet a third notion of liquidity uses a setting where moral hazard problems limit firms’ ability to borrow to meet unexpected investment needs. Because moral hazard limits the effectiveness of transactions between firms with excess liquidity and firms that need liquidity, a bank that provides contingent liquidity to those that need it can dominate a decentralized market. The first view of banks as liquidity providers concerns the role of banks in the payments system. Freeman (1996a,b) models an environment where agents are spatially separated and the timing of transactions is such that they cannot simultaneously trade at a central location. The problem in the model is that some agents, buyers, wish to consume goods from other agents, but have no goods that the buyers want to offer these sellers in exchange. Nor do buyers have any money, though later at another location they will be able to sell their goods in exchange for money. So, buyers issue i.o.u.’s – promises to pay at the central location next period with fiat money – to the sellers. Fiat money is used to settle the debts, but money and private debt coexist. Now, at the central clearing location it may happen that all creditors and debtors arrive simultaneously, in which case clearing occurs directly. If arrival is not simultaneous, however, settling can take place through a clearinghouse. The clearinghouse accepts money in payment of i.o.u.’s and pays off i.o.u.’s presented. However, if creditors arrive first, then the clearinghouse must have some means of paying them before the debtors arrive. A basic point of Freeman is develop the notion of the clearinghouse issuing its own i.o.u.’s, bank notes for example, that can circulate and be redeemed for fiat money later
22 Green(1997)builds on the Freeman model,arguing that a clearinghouse "netting by novation" can also achieve the efficiency gain in Freeman's model.In the same vein,McAndrews and Roberds (1999)model the efficiency gains from introducing banks that allow for centralized netting of claims.A bank can lend to firms via overdrafts.The firms are willing to accept payment in bank funds since the income funds can be used to repay the overdraft loan.Banks can provide"liquidity"to the extent that the payments they are requested to make are offsetting.Williamson(1992)also presents a model in which fiat money and private bank "money"coexist in equilibrium.Cavalcanti and Wallace(1999)study a random-matching model in which some agents,called banks,can produce information about the trading histories of other agents,called nonbanks.The equilibrium is one in which the banks issue and redeem private bank notes. In these models,banks issue private money to facilitate their role in clearing transactions.This is related to the historical experiences during which banks actually did issue their own private money, notably during the American Free Banking Era,1838-1863.During this period hundreds of different banks'monies circulated.Early economic historians and monetary theorists viewed the experience as a failure,arguing that it was marked by "wildcat banking"which justified a role for the government in the provision of a fiat currency.Following earlier work by Rockoff(1974,1975),Rolnick and Weber(1982, 1983,1984,1985)reexamine failure rates over the cross section of states with different banking regimes and conclude that the period was not marked by such episodes.Gorton (1996,1999)analyzes the prices of private bank notes and concludes that the market for banknotes worked well in pricing the risk of bank failure and in preventing wildcatting.Some experiences of Illinois,New York and Wisconsin during the Free Banking period are studied by Economopoulos(1988,1990). The pricing of free bank notes raises another issue concerning the production of liquidity.When offered a bank liability in exchange for goods,the seller of goods must recognize the risk that the bank can fail before the liability is honored.If some agents have private information about the likelihood of bank failure,they may be able to benefit from this when trading bank liabilities.An important property of a medium of exchange may well be that there is little or no such risk;that is,the value of the medium of exchange is independent of such considerations.But then it must be riskless in the sense that its value does not depend on the likelihood of the bank failing.This intuition is developed in the second view of liquidity,exemplified by Gorton and Pennacchi(1990). Gorton and Pennacchi (1990)begin with a common assumption of financial market models, namely the existence of"noise traders"or"liquidity traders."Kyle (1985)originally introduced these traders as a reduced-form modeling device,following Grossman and Stiglitz (1980).These models do not explicitly examine the motives of these noise traders;instead,they are posited to conveniently trade and lose money,making it profitable for other traders to undertake costly information production
22 Green (1997) builds on the Freeman model, arguing that a clearinghouse “netting by novation” can also achieve the efficiency gain in Freeman’s model. In the same vein, McAndrews and Roberds (1999) model the efficiency gains from introducing banks that allow for centralized netting of claims. A bank can lend to firms via overdrafts. The firms are willing to accept payment in bank funds since the income funds can be used to repay the overdraft loan. Banks can provide “liquidity” to the extent that the payments they are requested to make are offsetting. Williamson (1992) also presents a model in which fiat money and private bank “money” coexist in equilibrium. Cavalcanti and Wallace (1999) study a random-matching model in which some agents, called banks, can produce information about the trading histories of other agents, called nonbanks. The equilibrium is one in which the banks issue and redeem private bank notes. In these models, banks issue private money to facilitate their role in clearing transactions. This is related to the historical experiences during which banks actually did issue their own private money, notably during the American Free Banking Era, 1838-1863. During this period hundreds of different banks’ monies circulated. Early economic historians and monetary theorists viewed the experience as a failure, arguing that it was marked by “wildcat banking” which justified a role for the government in the provision of a fiat currency. Following earlier work by Rockoff (1974, 1975), Rolnick and Weber (1982, 1983, 1984, 1985) reexamine failure rates over the cross section of states with different banking regimes and conclude that the period was not marked by such episodes. Gorton (1996, 1999) analyzes the prices of private bank notes and concludes that the market for banknotes worked well in pricing the risk of bank failure and in preventing wildcatting. Some experiences of Illinois, New York and Wisconsin during the Free Banking period are studied by Economopoulos (1988, 1990). The pricing of free bank notes raises another issue concerning the production of liquidity. When offered a bank liability in exchange for goods, the seller of goods must recognize the risk that the bank can fail before the liability is honored. If some agents have private information about the likelihood of bank failure, they may be able to benefit from this when trading bank liabilities. An important property of a medium of exchange may well be that there is little or no such risk; that is, the value of the medium of exchange is independent of such considerations. But then it must be riskless in the sense that its value does not depend on the likelihood of the bank failing. This intuition is developed in the second view of liquidity, exemplified by Gorton and Pennacchi (1990). Gorton and Pennacchi (1990) begin with a common assumption of financial market models, namely the existence of “noise traders” or “liquidity traders.” Kyle (1985) originally introduced these traders as a reduced-form modeling device, following Grossman and Stiglitz (1980). These models do not explicitly examine the motives of these noise traders; instead, they are posited to conveniently trade and lose money, making it profitable for other traders to undertake costly information production