Two experts in monetary policy offer a unified framework for studying the role of money and liquid assets in the economy

*197*
*67*
*4MB*

*English*
*Pages 348
[382]*
*Year 2011*

- Author / Uploaded
- Ed Nosal
- Guillaume Rocheteau

*Table of contents : ContentsAcknowledgmentsIntroduction Models Where Money Is Useless Models Where Money Is Essential Beyond Monetary Exchange: Credit and Liquidity Tour of the BookChapter 1. The Basic Environment 1.1 Benchmark Model 1.2 Variants of the Benchmark Model 1.3 Further ReadingsChapter 2. Pure Credit Economies 2.1 Credit with Commitment 2.2 Credit Default 2.3 Credit with Public Record Keeping 2.4 Credit with Reputation 2.5 Further Readings AppendixChapter 3. The Role of Money 3.1 Money Is Memory 3.2 Decentralizing Allocations 3.3 Further ReadingsChapter 4. Money in Equilibrium 4.1 A Model of Divisible Money 4.2 Alternative Bargaining Solutions 4.3 Walrasian Price Taking 4.4 Competitive Price Posting 4.5 Further Readings AppendixChapter 5. Properties of Money 5.1 Divisibility of Money 5.2 Portability of Money 5.3 Recognizability of Money 5.4 Further Readings AppendixChapter 6. The Optimum Quantity of Money 6.1 Optimality of the Friedman Rule 6.2 Interest on Currency 6.3 Friedman Rule and the First Best 6.4 Necessity of the Friedman Rule 6.5 Feasibility of the Friedman Rule 6.6 Trading Frictions and the Friedman Rule 6.7 Distributional Effects of Monetary Policy 6.8 The Welfare Cost of Inflation 6.9 Further ReadingsChapter 7. Information, Monetary Policy, and the Inflation–Output Trade-Off 7.1 Stochastic Money Growth 7.2 Bargaining under Asymmetric Information 7.3 Equilibrium under Asymmetric Information 7.4 The Inflation and Output Trade-Off 7.5 An Alternative Information Structure 7.6 Further Readings AppendixChapter 8. Money and Credit 8.1 Dichotomy between Money and Credit 8.2 Costly Record Keeping 8.3 Strategic Complementarities and Payments 8.4 Credit and Reallocation of Liquidity 8.5 Short-Term and Long-Term Partnerships 8.6 Further Readings AppendixChapter 9. Money, Negotiable Debt, and Settlement 9.1 The Environment 9.2 Frictionless Settlement 9.3 Settlement and Liquidity 9.4 Settlement and Default Risk 9.5 Settlement and Monetary Policy 9.6 Further Readings AppendixChapter 10. Competing Media of Exchange 10.1 Money and Capital 10.2 Dual Currency Payment Systems 10.3 Money and Nominal Bonds 10.4 Recognizability and Rate-of-Return Dominance 10.5 Pairwise Trade and Rate-of-Return Dominance 10.6 Further ReadingsChapter 11. Liquidity, Monetary Policy, and Asset Prices 11.1 A Monetary Approach to Asset Prices 11.2 Monetary Policy and Asset Prices 11.3 Risk and Liquidity 11.4 The Liquidity Structure of Assets’ Yields 11.5 Endogenous Recognizability, Information, and Liquidity 11.6 Further Readings AppendixChapter 12. Liquidity and Trading Frictions 12.1 The Environment 12.2 Equilibrium 12.3 Trading Frictions and Asset Prices 12.4 Intermediation Fees and Bid–Ask Spreads 12.5 Trading Delays 12.6 Further ReadingsBibliographyIndex*

Money, Payments, and Liquidity

Money, Payments, and Liquidity

Ed Nosal and Guillaume Rocheteau

The MIT Press Cambridge, Massachusetts London, England

© 2011 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. For information about special quantity discounts, please email [email protected] mit.edu This book was set in Palatino by Newgen. Printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Nosal, Ed. Money, payments, and liquidity / Ed Nosal and Guillaume Rocheteau. p. cm. Includes bibliographical references and index. ISBN 978-0-262-01628-5 (hardcover : alk. paper) 1. Liquidity (Economics) 2. Monetary policy. 3. Money. I. Rocheteau, Guillaume. II. Title. HG178.N68 2012 339.5 3—dc22 2011007836 10 9 8 7 6 5 4 3 2 1

to our parents à nos parents

Contents

Acknowledgments Introduction 1

1

The Basic Environment 1.1 1.2 1.3

2

xi

13

Benchmark Model 14 Variants of the Benchmark Model Further Readings 19

Pure Credit Economies

18

21

2.1 Credit with Commitment 22 2.2 Credit Default 26 2.3 Credit with Public Record Keeping 2.4 Credit with Reputation 38 2.5 Further Readings 44 Appendix 45

3

The Role of Money 3.1 3.2 3.3

4

47

Money Is Memory 48 Decentralizing Allocations Further Readings 59

Money in Equilibrium 4.1

52

61

A Model of Divisible Money 62 4.1.1 Steady-State Equilibria 69 4.1.2 Nonstationary Equilibria 71 4.1.3 Sunspot Equilibria 74

31

viii

Contents

4.2

Alternative Bargaining Solutions 4.2.1 Bargaining Set 77 4.2.2 The Nash Solution 79 4.2.3 The Proportional Solution 4.3 Walrasian Price Taking 84 4.4 Competitive Price Posting 86 4.5 Further Readings 91 Appendix 93

5

Properties of Money

76

82

99

5.1

Divisibility of Money 101 5.1.1 Currency Shortage 102 5.1.2 Indivisible Money and Lotteries 5.1.3 Divisible Money 110 5.2 Portability of Money 111 5.3 Recognizability of Money 115 5.4 Further Readings 119 Appendix 121

6

The Optimum Quantity of Money 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9

7

107

127

Optimality of the Friedman Rule 129 Interest on Currency 133 Friedman Rule and the First Best 135 Necessity of the Friedman Rule 138 Feasibility of the Friedman Rule 142 Trading Frictions and the Friedman Rule Distributional Effects of Monetary Policy The Welfare Cost of Inﬂation 154 Further Readings 158

144 150

Information, Monetary Policy, and the Inﬂation–Output Trade-Off 161 7.1

7.2 7.3 7.4 7.5

Stochastic Money Growth 162 7.1.1 Symmetrically Uninformed Agents 164 7.1.2 Symmetrically Informed Agents 166 Bargaining under Asymmetric Information 167 Equilibrium under Asymmetric Information 174 The Inﬂation and Output Trade-Off 177 An Alternative Information Structure 184

Contents

7.6 Further Readings Appendix 189

8

Money and Credit

ix

188

195

8.1 Dichotomy between Money and Credit 196 8.2 Costly Record Keeping 202 8.3 Strategic Complementarities and Payments 207 8.4 Credit and Reallocation of Liquidity 214 8.5 Short-Term and Long-Term Partnerships 219 8.6 Further Readings 225 Appendix 227

9

Money, Negotiable Debt, and Settlement

229

9.1 The Environment 230 9.2 Frictionless Settlement 233 9.3 Settlement and Liquidity 236 9.4 Settlement and Default Risk 242 9.5 Settlement and Monetary Policy 246 9.6 Further Readings 247 Appendix 248

10

Competing Media of Exchange

251

10.1 Money and Capital 252 10.1.1 Linear Storage Technology 253 10.1.2 Concave Storage Technology 257 10.1.3 Capital and Inﬂation 260 10.2 Dual Currency Payment Systems 262 10.2.1 Indeterminacy of the Exchange Rate 263 10.2.2 Cash-in-Advance with a Twist in a Two-Country Model 265 10.3 Money and Nominal Bonds 270 10.3.1 The Rate-of-Return Dominance Puzzle 271 10.3.2 Money and Illiquid Bonds 273 10.4 Recognizability and Rate-of-Return Dominance 274 10.5 Pairwise Trade and Rate-of-Return Dominance 280 10.6 Further Readings 282

11

Liquidity, Monetary Policy, and Asset Prices 11.1 A Monetary Approach to Asset Prices 286 11.2 Monetary Policy and Asset Prices 291

285

x

Contents

11.3 Risk and Liquidity 295 11.4 The Liquidity Structure of Assets’ Yields 300 11.5 Endogenous Recognizability, Information, and Liquidity 11.5.1 Equilibrium 306 11.5.2 Equilibria with Recognizable Assets 308 11.5.3 Equilibria with Unrecognizable Assets 309 11.5.4 Multiple Monetary Equilibria 310 11.6 Further Readings 313 Appendix 314

12

Liquidity and Trading Frictions 12.1 12.2 12.3 12.4 12.5 12.6

317

The Environment 318 Equilibrium 320 Trading Frictions and Asset Prices 326 Intermediation Fees and Bid–Ask Spreads Trading Delays 332 Further Readings 337

Bibliography Index 357

339

329

305

Acknowledgments

The starting point of this book is a Federal Reserve Bank of Cleveland policy discussion paper. We began work on the policy discussion paper in 2004, and our objective was to provide a concise overview of the literature on the economics of payments using a uniﬁed framework. During this phase of the project we beneﬁted from the great working environment and resources provided by the Research Department at the Federal Reserve Bank of Cleveland. We want to acknowledge the unwavering support provided by the research director, Mark Sniderman, and by our colleagues Dave Altig, Mike Bryan, Bruce Champ, and Peter Rupert. The content of the book has been shaped by our collaborations with many researchers in the ﬁeld of monetary theory and payments: David Andolfatto, Boragan Aruoba, Aleksander Berentsen, Ricardo Cavalcanti, Ben Craig, Ricardo Lagos, Yiting Li, Sebastien Lotz, Peter Rupert, Shouyong Shi, Christopher Waller, Neil Wallace, Pierre-Olivier Weill, and Randall Wright. Several sections or chapters in the book come directly from our own work with these coauthors. For instance, the chapter on money under alternative trading mechanisms comes from some work with Boragan Aruoba, Christopher Waller, and Randall Wright. The sections on the divisibility and recognizability of money are derived from some work with Aleksander Berentsen and Yiting Li. The section on money and capital comes from some work with Ricardo Lagos. The model on liquidity in over-the-counter markets is also a joint work with Ricardo Lagos based on a paper by Darrell Dufﬁe, Nicolae Garleanu, and Lasse Pedersen. In writing this book, we stand on the shoulders of many scholars. In particular, the basic research agenda for the ﬁeld of monetary theory has been largely shaped by Neil Wallace. The search-theoretic approach to monetary economics was pioneered by Nobu Kiyotaki and Randall

xii

Acknowledgments

Wright, and the framework that we use throughout the book was developed by Ricardo Lagos and Randall Wright. This framework itself beneﬁted from the earlier work of Shouyoug Shi, Alberto Trejos, and Randall Wright. We would like to thank Steve Williamson for his comments on the 2005 policy discussion paper that led to this book, David Andolfatto, for his insightful comments on the ﬁrst half of the book, Neil Wallace who provided thoughtful comments for the introduction, Aleksander Berentsen who used this book to teach monetary theory at the universities of Basel and Zurich, and Stan Rabinovich who provided detailed comments for the entire book. We also thank graduate students at the Institute for Advanced Studies, Vienna, National University of Singapore, the Singapore Management University, and the University of California, Irvine, especially Ryan Baranowski, Giovanni Sibal, and Cathy Zhang. Finally, we have beneﬁted from the support of the MIT Press.

Introduction

Economics is all about gains from trade. But if gains from trade are to be realized, people must exchange one object for another. How does this exchange materialize? Exchange can be easy. If John has apples but likes strawberries more, and his classmate Paul has strawberries but likes apples more, then John and Paul can directly exchange strawberries for apples when they meet. There is a double coincidence of wants: John has what Paul wants and Paul has what John wants. In ﬁgure I.1 we represent the endowments and preferences of John and Paul. The × beside the names are their endowments and the arrows indicate the goods they would like to consume. Unfortunately, life is typically not that easy. So let’s complicate things just a little by adding another classmate, George, and a third commodity, tangerines, to the picture. George has tangerines and likes apples more but hates strawberries, Paul has strawberries and now likes tangerines more but hates apples, and John has apples and still likes strawberries more but hates tangerines. The preferences and endowments for John, Paul and George are now depicted in ﬁgure I.2. How do John, Paul, and George trade? If they can all meet at the same time and place, then exchange is just as easy as above. John gives apples to George, George gives tangerines to Paul, and Paul gives strawberries to John. But what if they can only meet in pairs? For concreteness, one Apples

John Figure I.1 Double coincidence of wants

Strawberries

Paul

2

Introduction

Apples

John

Strawberries

Paul

George

Tangerines

Figure I.2 Lack-of-double-coincidence-of-wants problem

can think that at different times, two classmates are randomly chosen and are brought together at a meeting in Las Vegas. Why Vegas? Since what happens in Vegas, stays in Vegas, what these classmates do in their meetings remains private information. If they can commit, then exchange is still easy. John can commit to give apples to George, George can commit to give tangerines to Paul, and Paul can commit to give strawberries to John. Sooner or later, all of the desirable exchanges will take place via pairwise meetings. But commitment seems rather strong; it is not a characteristic found in abundance in human interaction. But what if they are unable to commit? Then the trading arrangement described above—give your partner in a meeting your good if he desires it more than you—will not work. For example, if John gets strawberries from Paul, then he has no incentive to give apples to George, provided that John likes apples a little, because he gets nothing in return from George. Hence John might as well consume his own apples. So when they meet in pairs, we have the famous double-coincidence problem. Two coincidences are required if trade is to take place, so there is a problem if classmate A really likes what classmate B is holding, but classmate B does not like what classmate A is holding. If they are only willing to trade the good they have for the good they desire most, then the outcome is autarky. Autarky, however, need

Introduction

3

not be the outcome. For example, Paul could accept John’s apples in exchange for strawberries, even though Paul doesn’t like apples. After this transaction, when Paul meets George, Paul can trade his apples for George’s tangerines. In this example, apples are used as medium of exchange, meaning that the apple is accepted in trade by Paul not to be consumed, but to be traded later on for some other good, tangerines. This medium of exchange is useful or essential in the following sense. If there is no medium of exchange, then the outcome, autarky, is worse than the allocation that can be obtained with a medium of exchange. Typically people have to pay for the goods they acquire. And depending on the situation, the payment instruments—or media of exchange—can be commodities, real assets, and/or ﬁat money. What ends up serving as the economy’s payments instruments depends on many factors, such as the cost of storing apples compared to the cost of storing strawberries or tangerines, or how easy it is to recognize the quality of apples relative to that of strawberries or tangerines. What media of exchange will emerge in an economy? What are the factors that determine what will and will not be media of exchange? These are questions that this book addresses. But whatever the payment instruments are, the only kind that we study in this book are ones that are essential in the sense described above. Models Where Money Is Useless The most obvious and ubiquitous payment instrument is money. Although much ink has been spilled on the topic of money, some observers, such as Banerjee and Maskin (1996), believe that “money has always been something of an embarrassment to economic theory.’’ One reason for this unsatisfactory situation is that the “wrong’’ model is used to study money. The benchmark model in economics is that of Arrow and Debreu (1954) and Debreu (1959). The environment is frictionless: markets are complete and people can commit to all future actions. At the beginning of time, a market opens up and individuals choose the goods they want to buy and sell over all future contingencies. The only constraint an individual faces is a budget constraint. As the future unfolds, people make or accept delivery of goods as promised at the beginning of time. In the standard Arrow–Debreu environment, a competitive equilibrium is Pareto optimal. Pareto optimality necessarily implies that money cannot play an essential role in the economy. This observation also applies to the workhorse model of modern macroeconomics, the

4

Introduction

neoclassical growth model developed by Cass (1965) and Koopmans (1965), and Kydland and Prescott (1982). Fiat money has been forced into these models so that monetary policy can be studied. Since money is not essential, it has to enter the picture in some ad hoc fashion. Real money balances can be assumed to be a productive good and can enter either utility functions (e.g., Patinkin 1965) or production functions (e.g., Fischer 1974). This assumption seems odd; Fiat money is an intrinsically useless object, but is being treated as a standard consumption or intermediate good. Equally puzzling is that the price level enters the utility or the production functions. Along similar lines, Niehans (1971, 1978) suggested a transactions role for money by introducing exogenous transaction costs, and assuming that money has the lowest of these costs. Another popular approach, initiated by Clower (1967), is based on the observation that in monetary economies, goods are not traded for other goods directly. Goods are traded for money. To capture this “stylized fact’’ of monetary economies, Clower (1967) and Lucas (1980) introduce a restriction that requires that consumption goods be purchased only with money, called a “cash-in-advance constraint.’’ The problem with this description is that money enters the economy as a constraint that reduces the welfare of the economy, and not as a mechanism that overcomes exchange problems and enlarges the set of allocations that are feasible. The most prominent framework for policy analysis nowadays, the New Keynesian model of monetary policy proposed by Woodford (2003), takes money completely out of the picture by focusing on “cashless economies.’’ In such economies, money only matters as a unit of account given that prices are set in this unit of account and can only be readjusted infrequently. Models Where Money Is Essential Following Wallace (1998, 2001, 2010), we believe a reasonable modeling goal in the study of money, or any payment instrument, is that it be essential. None of the approaches described above satisfy the so-called Wallace (1998) dictum: [T]he proposed dictum is that money should not be a primitive in monetary theory. It is easy to describe in the abstract how to construct models that satisfy this dictum: specify both the physical environment and the equilibrium concept

Introduction

5

of the model in a way that does not rely on the concept called money or force the modeler at the outset to specify which objects will play a special role in trade. The physical environment and the equilibrium concept may include features that make trade difﬁcult, more difﬁcult than in the S[tochastic Competitive] G[eneral] E[quilibrium] model—features such as trading posts that are pairwise in objects, asymmetric information, or pairwise meetings. The model may also include assets that differ in their physical characteristics. For example, some assets may be indivisible and others not, some may be ﬁat objects while others throw off a real dividend at each date, some may physically depreciate more than others, some may be more recognizable than others, and some may yield disutility because they give off a noxious odor. Given such a speciﬁcation, the model determines—but, in general, not uniquely because there may be multiple equilibria—the values of the different assets and their distinct roles, if any, in exchange.

There are a number of models of essential, or useful, money grounded in a competitive environment. The competitive equilibrium in the overlapping generations (OLG) model, developed by Samuelson (1958), does not need to be Pareto efﬁcient. In an OLG model people are born at different dates, live ﬁnite lives, and the economy continues forever. The structure of the OLG model implies that credit—namely borrowing and lending—is not incentive feasible. If the (nonmonetary) competitive equilibrium is not Pareto efﬁcient, then the introduction of ﬁat money results in a Pareto improvement. Money is essential because it allows agents to engage in Pareto-improving (intertemporal) trades. The OLG model was the standard model for monetary economics for well over a decade. It is the environment that Lucas (1972) used to revolutionize macroeconomics. The authoritative statement and accomplishments of this framework can be found in Wallace (1980). As in the OLG model, money can play a useful role in Townsend’s (1980) turnpike model. The model has inﬁnitely lived agents moving along an endless linear “highway,’’ or turnpike, from one location to the next. Agents receive endowments in alternating periods, which creates a need for intertemporal trade. But agents with different endowment processes move in opposite directions along the turnpike. So agents of different types meet at most once, which makes credit arrangements infeasible. Just as in the OLG model, the introduction of ﬁat money leads to an allocation that is preferred by all agents in the economy. Ostroy (1973), Starr (1972), and Ostroy and Starr (1974, 1990) focused on the transactional role of money in an otherwise standard general equilibrium model. The exchange process, by which agents move from their initial endowments to a ﬁnal allocation, is modeled by (many)

6

Introduction

rounds of bilateral trade. In a round of bilateral trading, the value of goods that agent 1 wants from agent 2 may exceed the value of goods that agent 2 wants from agent 1. Because of this lack of double coincidence of wants, it can take many rounds of trade before all agents are able to move from the initial endowment to their ﬁnal (equilibrium) allocation. If money is introduced, then additional quantities of goods can be bought and sold, implying there will be fewer rounds of bilateral trading. If trading is costly, money is useful. A competitive environment, however, is not the most natural one to think about issues relating to money. For example, there is a strategic aspect to money: I accept an intrinsically useless object in trade because I rationally think others will accept it. As well, the mechanics of exchange—how people meet and exchange goods—is not formalized in a competitive environment. So it is difﬁcult to think about double-coincidence problems in such an environment. A natural way to capture strategic and double-coincidence issues is in a model of bilateral meetings. Jones (1976) was the ﬁrst to model the double-coincidence problem in a bilateral random meeting context. Diamond (1982) constructed a fully coherent equilibrium search model, but without money. In Diamond’s (1982) model a person cannot consume the good he produces, but goods produced by anyone else are perfect substitutes in consumption. Since there is never a double-coincidence problem, there are no impediments to exchange, once people have met. Diamond (1984) introduced money into his search model but it was accomplished by imposing a cash-in-advance constraint. In a series of papers, Kiyotaki and Wright (1989, 1991, 1993) added the doublecoincide problem identiﬁed by Jones (1976) into Diamond’s (1982, 1984) equilibrium search model. The big innovation in Kiyotaki and Wright (1989, 1991, 1993) was the introduction of heterogeneity over tastes and goods. The original Kiyotaki and Wright (1989) model focused almost exclusively on the emergence of commodity money as a medium of exchange. In a simple three-person environment, they designed the pattern of specialization, consumption, and production to create a lack of double coincidence of wants between agents. This heterogeneity is similar to that described in the John, Paul, and George example. They showed that certain goods will emerge as a medium of exchange depending on preferences, endowments, and beliefs. A somewhat stunning result was that in some equilibria, the good that serves as medium of exchange is the good with the highest storage cost, or the lowest rate of return. This

Introduction

7

ﬁnding was interpreted as a possible resolution for the long-standing rate-of-return dominance puzzle. The rate-of-return dominance puzzle, identiﬁed by Hicks (1935), is the lack of a compelling explanation for the observation that the rate of return on a medium of exchange is less than the ones of other assets in the economy. The puzzle is why people wouldn’t hold and use higher rate-of-return instruments as media of exchange. Kiyotaki and Wright (1991, 1993) extended the previous analysis to include an intrinsically useless object and demonstrated that this object can be valued in exchange and can raise society’s welfare. The equilibrium, however, is not unique. If, for example, people believe that money will be accepted as a means of payment in the future, then a monetary equilibrium prevails with ﬁat money as a universally accepted means of payments. Alternatively, if people believe that money will not be accepted as a means of payment in the future, then the equilibrium is characterized by barter only. The models described above are rather stark and simple. All objects are indivisible; agents can hold at most unit of output or one unit of ﬁat money, and in all meetings, objects trade one-for-one. One may reasonably ask, other than demonstrating that a medium of exchange can emerge, what can one learn from such a stylized environment. The answer is: A lot. Here are two examples. Kiyotaki, Matsui, and Matsuyama (1993) adopted a two-country, twocurrency version of the Kiyotaki–Wright model where they investigated the conditions under which a currency would emerge as an international currency, meaning a currency that is accepted as medium of exchange in both countries. This question was virtually impossible to address in reduced-form monetary models. Their answer was both intuitive and insightful. They found that the status of international currency depends on both fundamentals, such as the sizes of the countries and their degree of integration, as well as (self-fulﬁlling) beliefs and conventions. Williamson and Wright (1994) formalized the old idea developed by Jevons (1875) that recognizability is a key property for a good or commodity to be used as money. They considered an environment where there is a double coincidence of wants in all meetings, as in Diamond (1982, 1984), but goods can be produced in different qualities, and agents have some private information about the quality of their goods. They showed that (fully recognizable) ﬁat money can play a useful role even if there is no double-coincidence problem. By introducing a good of recognizable quality, ﬁat money, consumption goods of unknown

8

Introduction

quality become less acceptable, and hence agents have less incentive to produce them. Although the Kiyotaki–Wright model provides useful insights, it is unable to satisfactorily address some important and interesting questions in monetary theory. For example: How is the exchange value of money determined? The Kiyotaki–Wright model essentially evades this question since, by assumption, people can hold at most one unit of indivisible money, and one unit of money trades for one unit of output. To answer this and other interesting policy questions, these extreme assumptions have to be relaxed in a number of directions. The ﬁrst step to generalize the model environment, undertaken by Shi (1995) and Trejos and Wright (1995), was to endogenize the value of money. This was accomplished, despite the restriction that people hold at most one unit of indivisible money, by making output divisible. With divisible output, the quantity of goods that is traded for one unit of money is determined by bargaining between the two parties; hence one can speak sensibly about the value of money. Osborne and Rubinstein (1990) provide a systematic treatment of markets with bilateral trade and bargaining. The models of Shi (1995) and Trejos and Wright (1995) impose a pricing (or trading) mechanism in bilateral meetings between agents. Although the mechanism typically has axiomatic or strategic foundations, it is chosen arbitrarily and might not lead to allocations with good properties from society’s point of view. An alternative approach, proposed by Kocherlakota (1998) and strongly endorsed by Wallace (2010), is that of mechanism design. In a mechanism design approach, a planner chooses the trading mechanism among all incentive-feasible mechanisms. The mechanism that the planner chooses satisﬁes some desirable property; for example, it maximizes social welfare. A mechanism design approach can be helpful for establishing the essentiality of money. Recall that money is essential if, given the speciﬁcation of the environment, there is no other way to achieve (desirable) allocations. Regardless of how the value of money is determined—whether using strategic or axiomatic approach or a mechanism design approach—it is possible to examine how a change in the aggregate stock of money affects the value of money and output. However, because agents are restricted to hold at most one unit of money, the more interesting policy question of how a continuous change in the money supply affects inﬂation and output cannot be addressed. Initial progress on the modeling of money growth was made by simply relaxing the unit

Introduction

9

upper bound constraint on money holdings, within the context of a Shi–Trejos–Wright-type environment. Zhu (2003, 2005) provided existence results when money holdings are richer than {0, 1} for both indivisible and divisible money. Camera and Corbae (1999) and Molico (2006) provided numerical based solutions for these richer money holdings environments. Green and Zhou (1998) and Zhou (1999) assumed price posting by sellers and indivisible goods which made the model a bit more tractable. All these papers, however, demonstrate that departing from the unit money upper bound assumption signiﬁcantly complicates the analysis. The complications arise from the fact that the equilibrium is, in part, characterized by a distribution of money holdings that is determined jointly with terms of trade in bilateral matches. And characterizing these equilibrium objects jointly is not easy (at least, analytically). An alternative and clever approach to deal with unrestricted money holdings and divisible money is to change the economic environment in a way that implies that, in equilibrium, the money holdings of all agents of the same type are identical just before they are bilaterally matched. Since the distribution of money holdings is degenerate, the model becomes analytically tractable. Shi (1997), taking the lead from Lucas (1990), assumed that households are composed of a continuum of members—buyers and sellers—who pool their money holdings. This large-household structure implies that risks associated with the random matching process for individual buyers and sellers can be completely diversiﬁed away at the household level. Lagos and Wright (2005) instead introduced competitive markets that operate periodically and quasi-linear preferences. The competitive markets allow agents to adjust their money holdings following random-matching shocks. Since quasi-linear preferences eliminate wealth effects, all agents will make the same choices in the competitive market, except for the choice of the “quasi-linear good.’’ The Lagos– Wright environment can accommodate different pricing mechanisms in the decentralized exchange market (Rocheteau and Wright 2005), such as bargaining, price posting, and Walrasian pricing. Moreover the existence of periodic competitive markets allows for the reintroduction of Arrow–Debreu-type general equilibrium apparatus, such as state-contingent commodities (Rocheteau et al. 2008). Because of its ﬂexibility, the Lagos–Wright model has already generated a large body of applications and extensions (see Williamson and Wright 2010a, b). We use this model throughout the book.

10

Introduction

Beyond Monetary Exchange: Credit and Liquidity In this book we are interested in understanding how gains from trade can best be exploited in economic environments characterized by different sets of frictions. We do not require that trade be mediated by money since different frictions may dictate the use of different payment instruments. One of the key challenges in monetary theory is to provide an explanation for the coexistence of money and credit. To address this issue, we allow agents to use bilateral credit arrangements, or IOUs, to facilitate trade, as in Diamond (1987, 1990) and Shi (1996). One reason why coexistence is a challenge is that the frictions that are needed to make money essential typically make credit infeasible, and environments where credit is feasible are ones where money is typically not essential. By constructing environments where money and credit coexist, we are able to study interactions between monetary policy and the use of credit. We also study the notion of settlement, the process by which an obligation created by a credit relationship is ultimately extinguished. We examine how frictions in the settlement process affect the role of the monetary authority. Over time ﬁnancial innovations such as securitization have made the distinction between monetary and nonmonetary assets somewhat fuzzy. Individuals and ﬁrms have access to checkable equity and bond mutual funds, they can get home and car equity loans that are effectively consumption loans collateralized by assets, and they can use government bonds as collateral in many instances. So at least indirectly people use all sorts of assets to facilitate trade. In some extensions of our basic model we allow people to use assets other than ﬁat money or credit to facilitate exchange: assets such as capital, land, and government debt. Again, whether agents use these assets to conduct their transactions depends on the properties of assets, such as divisibility and recognizability, and on the frictions they face. In practice, monetary policy is conducted through open-market operations, where the monetary authority trades ﬁat money for bonds. A fully coherent model of monetary policy should include these two assets, and explain how they coexist even though bonds pay interest but money doesn’t. Or, put another way, a coherent model of monetary policy should address the rate-of-return dominance puzzle. We provide explanations for this puzzle based on the physical properties of the interest bearing asset, such as recognizability, and on conventions or self-fulﬁlling beliefs. Our approach to address the rate-of-return

Introduction

11

dominance puzzle can also be applied to other types of asset pricing anomalies. For instance, Lagos (2006) showed that a monetary model with bonds and equity can address the risk-free rate and equity premium puzzles. There is no universally accepted deﬁnition of liquidity. It is precisely because the concept of liquidity is somewhat vague that a model can be useful to clarify it. Clearly, when there are no frictions associated with trade, all assets (and goods) are equally liquid, as in the Arrow–Debreu model. However, if there are frictions associated with exchange, then some assets may be able to command greater amounts of goods in trade than other assets. Generally speaking, the liquidity of an asset has to do with the ease at which it can be used to ﬁnance a random spending opportunity. If it can only be sold on short notice at a discounted price or not at all, then the asset is said to be illiquid. One of our objectives will be to explain why different assets have different liquidity properties. When assets do have different liquidity properties, we investigate the implications that liquidity has for the distribution of asset returns, and for the relationship between asset prices and monetary policy. The notion of liquidity has a time, volume, and price dimension, and can be quantiﬁed by using measures related to the ease at which assets can be bought and sold. For example, liquidity can be measured by transaction costs, such as bid–ask spreads, trading delays, the time that it takes to buy or sell an asset, and trading volume. We will use the structure of our basic model, with decentralized trades and bilateral matches, to describe an over-the-counter asset market that can be used to think about these measures of liquidity. Tour of the Book The book is organized in twelve chapters. In the ﬁrst chapter we present the basic environment we will use throughout the book. Even though we introduce different twists along the way, the models we use all have some common ingredients: an alternating market structure with competitive and bilateral trades, and quasi-linear preferences, as in Lagos and Wright (2005). In chapters 2 to 5 we present benchmark economies with a single method of payment. In chapter 2 all trades are conducted with credit. We are interested in understanding whether an economy can achieve good allocations with credit for different sets of frictions. In chapter 3 we examine an economy whose frictions imply that credit arrangements are

12

Introduction

not incentive feasible, and we show that there is a role for ﬁat money. The role that ﬁat money fulﬁlls can be identiﬁed as a record-keeping technology. Chapter 4 uses the benchmark monetary economy to examine the allocations that emerge under different pricing mechanisms in the decentralized market. Chapter 5 studies how the properties of money, such as divisibility, portability, and recognizability, impinge upon its role as a medium of exchange and affects allocations. Chapters 6 and 7 are devoted to monetary policy. In chapter 6 we characterize the optimal rate of growth of money supply under various price mechanisms and frictions in the decentralized trade market. We carefully explain the circumstances under which the Friedman rule is optimal, achieves the ﬁrst-best allocation, and is feasible. In chapter 7 the money growth rate is assumed to be random, and we examine the relationship between inﬂation and output under different information structures. Chapters 8 and 9 look at economies where monetary exchange coexists with credit transactions. In chapter 8 we propose several environments where money and credit can coexist, and study how monetary policy affects the use of credit. In chapter 9 we introduce settlement frictions and investigate how these frictions affect the allocations, and if there is an optimal policy response. Chapters 10 and 11 consider the coexistence of money and other assets, such as another money, capital, and bonds. Chapter 10 studies monetary equilibria with multiple assets and focuses on the rate-of-return dominance puzzle. Chapter 11 investigates the implications for asset prices and monetary policy. Finally, chapter 12 adopts a continuous-time version of the model with intermediaries to describe the functioning of over-the-counter markets and to study how trading frictions affect asset markets, asset prices, and different measures of liquidity.

1

The Basic Environment

This book studies issues directly related to society’s need for media of exchange. Any such study requires a departure from the standard Arrow–Debreu model economy. In the Arrow–Debreu model, markets are frictionless and complete, agents can all get together at the beginning of time to buy and sell contracts, and they can commit to deliver or accept delivery of goods over all possible dates and contingencies. The basic structure of the Arrow–Debreu model implies that the economy can achieve a Pareto-efﬁcient allocation without needing objects like money or other ﬁnancial institutions. A good model of media of exchange should incorporate a number of key ingredients. We view the following as being necessary ingredients: 1. People cannot commit. If people can commit, then they can promise to repay their debts or make gifts, and there is no need for a medium of exchange. 2. The monitoring or record keeping of actions must be imperfect. As we will see later, a well-functioning record-keeping device can replicate the role played by a medium of exchange. 3. It must be costly for people to interact with one another. If people could costlessly get together to trade, then many trades could be arranged among groups of people without having to resort to a medium of exchange. A natural way to think about costly connections among people is that they meet in pairs. Moreover, if people meet in pairs, then monitoring what happens in these meetings may be difﬁcult. 4. There must be a problem of lack of double coincidence of wants. If there is not a double-coincide problem—meaning in every pairwise meeting each person wants what the other person has—then trades can be conducted through barter.

14

Chapter 1

5. The model must be dynamic. It would be difﬁcult to think about a number of (ﬁnancial) assets if the model was not dynamic. For example, who would be willing to accept ﬁat money, an intrinsically worthless piece of paper, in a static environment? Or, what is the meaning of debt, a promise to do something in the future, in a static model? It is, of course, possible to add to this list. For example, one may want to include imperfect recognizability as a key ingredient since it is useful in explaining the emergence of a uniform currency or the acceptability of an asset as a medium of exchange. We use the assumption of imperfect recognizability in many parts of the book, for example, to help explain the coexistence of money and higher rate of return assets. If assets are held, then they must be priced. It would be desirable to have these assets priced in competitive markets if only for convenience. So, although we require that bilateral trading relationships exist, we do not insist that all trades be conducted on a bilateral basis; i.e., some trades can be conducted on a competitive market. Finally, although it is not an absolute requirement, it would certainly be desirable if the model is analytically tractable. Tractability facilitates a better understanding of some issues or insights, and extending the model to address a large variety of topics related to money and payments. 1.1 Benchmark Model The benchmark model we use throughout the book will have the following characteristics. Time is discrete and continues forever. Each period is divided into two subperiods, called day and night, when different activities take place. During the day, trades occur in decentralized markets according to a timeconsuming bilateral matching process. We will label the day market DM, which can also stand for decentralized market. In the DM some agents can produce but do not want to consume, while other agents want to consume but cannot produce. For convenience, we label the former agents sellers and the latter buyers, which captures the agents’ roles in the DM. Our assumption on preferences—sellers have no desire to consume in the DM—and technologies—buyers are not able to produce in the DM—generates a double-coincidence problem in matches between buyers and sellers. The measures of buyers and sellers are equal, and are normalized to one.

The Basic Environment

15

A buyer meets a seller, and a seller meets a buyer, with probability σ . The parameter σ captures the extent of the trading frictions in the market. If σ = 1, the trading frictions are shut down (except for the pairwise meeting friction), and each agent can ﬁnd a trading partner with certainty. The parameter σ can be interpreted as capturing heterogeneity in terms of the goods that sellers produce and that buyers consume. We call the good that is produced and traded in the DM either the DM good or the search good, since trade requires a search activity. Exactly how production and trade are organized at night will depend on the issue that is under investigation. What can be said about the night market is that in general, it will be characterized by fewer frictions than those that plague the DM. We will label the night market CM, since this market will typically be a competitive market. At night, all agents can produce and consume. The good that is produced and consumed in the CM will be called either the CM good or the general good. Typically buyers will produce the general good in order to settle their debt or to readjust their asset holdings, and sellers will consume the general good in order to reduce their asset holdings. All goods, whether produced in the DM or in the CM, are nonstorable. So a search good cannot be carried into the CM, and a general good cannot be carried into the next DM. The perishability of consumption goods will prevent them from being used as means of payment. The preferences of the by t β t U b qt, xt , yt buyer and seller are given t s and t β U qt , xt , yt , respectively, where U b q, x, y and U s q, x, y are the buyer’s and seller’s period utility functions, q ∈ R+ is the quantity of the search good consumed and produced in the DM, x ∈ R+ is the quantity of the general good consumed in the CM, and y ∈ R+ is the amount of work undertaken in the CM. All agents discount between the night and the next day at rate r = β −1 − 1, where β ∈ (0, 1) is the discount factor. Although it is not crucial, we assume that the period utility functions are separable across subperiods: U b q, x, y = u q + U x, y and U s q, x, y = −c q + U x, y . More important, for tractability we will require that an utility agent’s function in the CM is linear in their hours of work, i.e., U x, y = v (x)−y.

16

Chapter 1

As we argue later, linearity plays a role in eliminating wealth effects and facilitates the determination of the terms of trade in the DM. The production technologies in the DM and CM are both linear in labor, where one unit of labor produces one unit of output. Therefore c q is the seller’s disutility (or cost) of labor in the DM and y is the agent’s disutility of labor in the CM. If v (x) is strictly concave, then we would typically get the choice of consumption in the CM satisfying x = x∗ , where v (x∗ ) = 1. For most of the book, and without loss of generality, we will simply assume that v (x) = x, so there is no gain from producing the general good for oneself. The timing of events and the preferences of agents are described in ﬁgure 1.1. In summary, the speciﬁcation of the period utility functions for buyers and sellers are U b (q, x, y) =u(q) + x − y,

(1.1)

U s (q, x, y) = − c(q) + x − y,

(1.2)

respectively. We assume that u (q) > 0, u (q) < 0, u(0) = c(0) = c (0) = 0, u (0) = +∞, c (q) > 0, c (q) > 0, and c(¯q) = u(¯q) for some q¯ > 0. We assume that the utility for the buyer in the DM is bounded below, which matters when there is negotiation between a buyer and a seller in the DM, so that utilities are not unbounded in the case of disagreements. Without loss of generality, we assume that u (0) = 0. An example of a (1−a) DM utility function for buyer is u q = q + b − b(1−a) , where b > 0 Discount factor across periods: β

DAY (DM)

NIGHT (CM)

σ bilateral matches between buyers and sellers.

Consumption/production of a general good

Buyer’s utility: u(q) Seller’s utility: −c(q)

Buyer’s utility: U(x, y) = x−y Seller’s utility: U(x, y) = x−y

Figure 1.1 Timing of events and preferences

The Basic Environment

17

but small. If a ∈ (0, 1), then b can be set equal to zero. This utility function is reminiscent of a constant relative risk aversion utility function, and approaches such a function as b goes to zero. Let q∗ denote the level of production and consumption of the search good that maximizes the match surplus between a buyer and seller, u(q)−c(q). It solves u (q∗ ) = c (q∗ ). Preferences in the DM are represented in ﬁgure 1.2. It can be seen graphically that q∗ maximizes the size of the gains from trade in the DM, i.e., the difference between u q and c q . The assumption that the utility functions for both buyers and sellers are linear in the general good is made for tractability purposes. In versions of the model where agents can hold assets, such as money or capital, a more general speciﬁcation for preferences would tend to generate a distribution of asset holdings when agents are subject to idiosyncratic shocks in the DM. The idiosyncratic shocks arise because of the randomness in the matching process in the DM. The heterogeneity in asset holdings is not eliminated by trading in the CM under a more general speciﬁcation of preferences due to wealth effects. In contrast, with (quasi-) linear utility, there are no wealth effects and agents, conditional on their type, will choose the same asset positions in the CM. The linearity of the CM utility function, U x, y , greatly c(q) u(q)

q* Figure 1.2 Preferences in a bilateral match

q

q

18

Chapter 1

simpliﬁes the determination of the terms of trade in the DM, which usually occurs through bargaining, and payment arrangements in bilateral matches. This linearity makes the continuation values in the bargaining problem linear. Note that the linear speciﬁcation for the utility over goods produced and consumed in the CM implies that there is no beneﬁt associated with producing the general good for one’s own consumption. The benchmark model can be reinterpreted as a representative household model, where the buyers are the households and the sellers are ﬁrms. Each ﬁrm has a technology that requires a discrete investment of k units of the general good in the CM to produce exactly one unit of a perfectly divisible intermediate good. The intermediate good is durable for one period, i.e., until the next CM. The ﬁrm can use the intermediate good in the subsequent period to produce the DM good and/or the CM good. The DM good is produced from the intermediate good according to a linear technology. The CM good is produced from the intermediate good according to the technology f (x), where f (0) = 0, f (0) = +∞ and f (1) = 0. The opportunity cost for the ﬁrm to produce the DM good is given by c(q) = f (1) − f (1 − q). Assume that −k + βf (1) = 0 so that a ﬁrm makes no proﬁts from producing only the general good. The proﬁts of the ﬁrms are transferred to households in a lump-sum fashion. 1.2 Variants of the Benchmark Model Generally speaking, we adopt some version of this benchmark model speciﬁcation throughout the chapters that follow. In all chapters there will be a DM, with bilateral matching of agents, and there will be a CM, where trades are more centralized and agents have linear utility. However, we will depart from some aspects of our benchmark model in order to focus on the problem at hand. For example, when we want to talk about capital formation, we will allow some goods to be storable; when we want agents to be able to borrow or lend before entering the DM, and after exiting the CM, we will introduce additional subperiods and match-speciﬁc heterogeneity. If we think that policy may affect the nature of the matching process, we will endogenize the extent of the search frictions; when it simpliﬁes the analysis, we will consider the case of ﬁnitely lived agents. When we do depart from the benchmark speciﬁcation, we will be very clear in explaining both how and why we are modifying the model.

The Basic Environment

19

1.3 Further Readings Jones (1976) examines a model with a double-coincidence problem, where agents meet in pairs and ﬁat money appears to be useful. The analysis, however, departs from rational expectations. Fully consistent models of bilateral exchange with trading frictions were introduced by Diamond (1982, 1984). Kiyotaki and Wright (1989, 1991, 1993) extend these models to incorporate a double-coincidence problem and a meaningful role for a medium of exchange. For related approaches, see also Oh (1989) and Iwai (1996). The basic model we consider adopts the environment of Lagos and Wright (2005). The version with ex ante heterogeneous buyers and sellers comes from Rocheteau and Wright (2005). Rocheteau, Rupert, and Wright (2007) examine an environment where agents’ CM utility functions are neither linear nor separable, but labor is indivisible and agents have access to lottery devices. Chiu and Molico (2010) do not use quasilinear preferences, but resort to numerical methods to solve the model. In most of the book we assume that the utility function in the centralized market is fully linear, as in Lagos and Rocheteau (2005). Shi (2006) explains the rationale that underlies the microfoundations of money, and why these foundations are necessary for monetary economics. Surveys and summaries of the literature are provided by Wallace (1998, 2000, 2010) and Williamson and Wright (2010a, b).

2

Pure Credit Economies

Consider an encounter between two individuals. One is hungry in the morning and wants to consume, but is only able to produce at night. Call him the buyer. The other can produce in the morning, but is only hungry at night. Call him the seller. If the buyer has nothing tangible to offer the seller in exchange for consumption goods, then the buyer and seller are unable to engage in a morning spot trade. In this event a simple solution would be for the buyer to promise to deliver some consumption goods in the future in exchange for some consumption goods now. Such a credit arrangement, however, may fail to materialize if the seller believes that after he produces for the buyer, the buyer will not repay his debt. In this chapter we are interested in characterizing the conditions under which bilateral credit is feasible, and the set of allocations that can be obtained in such credit economies. We are particularly interested in knowing if the best—socially desirable—allocations are among the feasible ones. We consider four related environments that can support credit arrangements but differ in terms of the amount of commitment, or trust, that agents possess, and on the punishments that can be imposed on a debtor who reneges on his obligation. We will start by considering the best of all possible worlds—similar to the standard Arrow–Debreu framework—where agents are always trustworthy. That is, agents have the ability to commit to repay their debts. In such an environment, there is nothing that prevents intertemporal gains from trade from being fully exploited: socially desirable allocations can always be achieved. In such a perfect world payment arrangements between agents are quite trivial. In our second environment, we assume that with positive probability, buyers are not able to produce when it is time to repay their

22

Chapter 2

debt. Different buyers may have different probabilities of default. If the buyer does not know any more than the seller—i.e., information is symmetric—then socially desirable allocations are still feasible. In this case the terms of trade reﬂect the possibility of default. If, however, buyers know their abilities, or probabilities, of repaying their debts and sellers don’t—i.e., information is asymmetric—then it becomes harder to achieve socially desirable allocations. In particular, if buyers are sufﬁciently different in terms of their probabilities to repay their debts, then the socially desirable allocations can no longer be obtained. In the ﬁnal two environments, we abandon the idea that agents can be trusted. If trade is to take place, trading arrangements must be self-enforcing. In the third environment, we assume there exists a technology—a public record-keeping device—that makes agents’ production levels publicly observable. This technology opens up the possibility of punishing someone who does not produce when he is supposed to. Whether or not socially desirable allocations can be achieved depends on how agents value future consumption, the size of the gains from trade, and the structure of the market. In the fourth environment, we assume that there does not exist a public record-keeping device, but, at times, agents are able to trade repeatedly among themselves. Repeated interactions allow for the possibility of trust building, where trust can be maintained by the punishment scheme of destroying a valuable partnership. We show that socially optimal allocations are feasible if it is hard to form a relationship— because, for example, the trading frictions are sufﬁciently severe—and if relationships are sufﬁciently stable. 2.1 Credit with Commitment The environments we will consider have the following characteristics: First, matches between buyers and sellers are formed during the day and are maintained at night. The fact that agents are matched for the entire period allows them to make promises—or negotiate debt contracts—during the day that can be settled at night. Second, there are no frictions—for example, no difﬁculties for debtors and creditors to ﬁnd one another—or no costs—for example, no administrative or enforcement costs—associated with settling debt at night: An agent can settle his debt by producing the general good at night, and transferring it to his creditor. Third, there are no tangible assets, such as money or

Pure Credit Economies

23

capital, that agents can use for trade purposes. We start with an economy where buyers can commit to repay their debts; then we consider environments where they cannot. We will describe the set of allocations that are feasible—such as the buyer’s consumption in a match cannot be greater than the seller’s production—and individually rational—meaning that trade is voluntary. We restrict the set of allocations to be symmetric across matches and constant over time. When a match is formed during the day in the decentralized market, DM, the buyer and seller must decide—either simultaneously or sequentially—whether to accept or reject the allocation (q, y), where q is the quantity of the search good produced by the seller for the buyer in DM, and y is the amount of the general good that the buyer promises to produce and deliver to the seller at night in the centralized market, CM. The buyer and seller will trade allocation q, y only if both of them accept it. We are agnostic in terms of how the allocation q, y is determined. For example, it might be the case that the allocation is an outcome from some bargaining protocol. Our objective is to describe all feasible and individually rational allocations that can be obtained through any trading mechanism. The sequence of events within a typical period is illustrated in ﬁgure 2.1. At the very beginning of the period, all agents are unmatched. During the DM, each agent ﬁnds a trading partner with probability σ . A buyer and seller who are in a match decide to accept or reject a proposed allocation (q, y). If either player rejects the proposal, then the match is dissolved; otherwise, the seller produces q units of the search or DM good for the buyer during the DM, and the buyer produces y units of the general or CM good for the seller at night in the CM. At the end of the period, all matches are destroyed. The expected lifetime utility of a buyer, evaluated at the beginning of the DM, is DAY (DM)

σ matches contract (q,y)

Matched sellers produce q

Figure 2.1 Timing of the representative period

NIGHT (CM)

Matched buyers produce y

Destruction of matches

24

V b = σ u(q) − y + βV b ,

Chapter 2

(2.1)

provided that both the buyer and seller accept allocation (q, y). According to (2.1), in the event that the buyer meets a seller, with probability σ , he consumes q units of the search good and produces y units of the general good. Since we focus on stationary allocations, time indexes are suppressed. The expected lifetime utility of a seller evaluated at the beginning of the DM is V s = σ −c(q) + y + βV s .

(2.2)

Equation (2.2) has an interpretation similar to (2.1), except for the fact that during the DM sellers produce (and buyers consume) the search good and in the CM sellers consume (and buyers produce) the general good. Since agents are able to commit, the only relevant constraints are buyers’ and sellers’ participation constraints, which are evaluated at the time that a match is formed. The participation constraints indicate whether agents are willing to participate in the trading arrangement q, y , i.e., whether they agree to the proposed contract. These constraints are u(q) − y + βV b ≥ βV b ,

(2.3)

−c(q) + y + βV s ≥ βV s .

(2.4)

According to (2.3), a buyer will accept allocation (q, y) if the lifetime utility associated with acceptance—the left side of (2.3)—exceeds the lifetime utility associated with rejection—the right side of (2.3)—or if his surplus from the trade, u(q) − y, is nonnegative. Condition (2.4) has a similar interpretation for the seller. Note that (2.3) and (2.4) consist only of single deviations to show the optimality of buyers’ and sellers’ strategies. After a deviation we assume that agents return to their proposed equilibrium strategies with their associated payoffs, given by the right sides of (2.3) and (2.4). These payoffs are identical to the βV b and βV s terms on the left sides of (2.3) and (2.4). From (2.3) and (2.4) the set of incentive feasible allocations, AC , is AC = (q, y) ∈ R2+ : c(q) ≤ y ≤ u(q) . (2.5) This set is represented graphically by the shaded area in ﬁgure 2.2. The gains from trade will be maximized if agents produce and consume q∗

Pure Credit Economies

25

y c(q) u(q)

q*

q

Figure 2.2 Incentive-feasible allocations under commitment

units of the search good u q∗ = c q∗ . From (2.5) it ∗ in the ∗DM, where is easy to check that q × [c(q ), u(q∗ )] ⊆ AC . When agents are able to commit, the intertemporal nature of the trades or any issues associated with search frictions are irrelevant for incentive feasibility; the efﬁcient level of production and consumption of the search good, q∗ , is incentive-feasible for any values of β and σ . The level of output for the general good, y, will determine how the gains from trade are split between the buyer and seller. Since any allocation in AC is incentive feasible, questions naturally arise regarding how the proposed allocation (q, y) will be chosen, and whether it will be efﬁcient. One way to address these questions is to impose an equilibrium concept or, equivalently, a trading mechanism on bilateral matches, and to characterize the outcome of this procedure. For example, we can assume that the allocation (q, y) is determined by the generalized Nash bargaining solution, where the buyer’s bargaining power is θ ∈ [0, 1]. If an agreement is reached, then the buyer’s lifetime utility is u q − y + βV b ; if they fail to agree, his lifetime utility is βV b . Similarly, if they reach an agreement, then the seller’s lifetime utility is y − c q + βV s ; if they fail, then his lifetime utility is βV s . The generalized Nash bargaining solution maximizes a weighted geometric mean

26

Chapter 2

of the buyer’s and seller’s surpluses from trade, u(q) − y and −c(q) + y, respectively, where the weights are given by the agents’ bargaining powers and a surplus is simply the difference between lifetime utility when there is agreement and lifetime utility when there is disagreement. The generalized Nash bargaining solution is given by the solution to θ 1−θ max u(q) − y y − c(q) q,y

(2.6)

subject to u(q) − y ≥ 0,

(2.7)

y − c(q) ≥ 0.

(2.8)

The solution to (2.6)–(2.8) is q = q∗ and y = (1 − θ)u(q∗ ) + θ c(q∗ ). See the appendix for details. The intuition that underlies the generalized Nash solution can be diagrammatically illustrated. The buyer’s surplus from a trade is Sb = u(q) − y, while the seller’s surplus is Ss = −c(q) + y. Hence the total surplus from a match is Sb + Ss = u(q) − c(q), and it is at its maximum when q = q∗ . All the pairs of surpluses (Sb , Ss ) that can be reached through bargaining, i.e., the pairs such that Sb + Ss ≤ u(q∗ ) − c(q∗ ), constitute the bargaining set, which is represented by the triangular area in ﬁgure 2.3. The Pareto frontier of the bargaining set is the pairs such that Sb + Ss = u(q∗ ) − c(q∗ ). The Nash solution is obtained graphically at the tangency point between a curve representing the Nash product, 1−θ θ , and the Pareto frontier of the bargaining set. u(q) − y −c(q) + y Note that the allocation is efﬁcient for any value of the buyer’s bargaining power θ ∈ [0, 1]. Furthermore, as one varies θ over [0, 1], the set of generalized Nash bargaining solutions varies over q∗ × [c(q∗ ), u(q∗ )]. Diagrammatically speaking, as θ increases, the solution moves down the Pareto frontier in ﬁgure 2.3. 2.2 Credit Default In the previous section, credit arrangements work remarkably well. In reality, however, credit arrangements may not function so smoothly. In particular, given the intertemporal nature of a debt contract, there is always a risk that something (bad) can happen between the time the contract is negotiated and the time it must be settled. For example, a buyer may be unable to or does not want to produce at the time of settlement; i.e., the buyer may default. For our ﬁrst pass at capturing the

Pure Credit Economies

27

Ss

u(q*) − c(q*)

Nash solution

Pareto frontier Nash product

u(q ) − c(q ) *

*

Sb

Figure 2.3 Nash bargaining

notion of default, we will assume that a buyer in a match can commit to produce in the CM if he is able to. But the buyer is subject to an exogenous, idiosyncratic productivity shock, which implies that he is able to produce in the CM with probability δ and is unable to produce with probability 1 − δ. Equivalently 1 − δ can be interpreted as the probability of an exogenous default. If buyers are homogeneous in terms of their default probabilities, then the expected lifetime utility of a buyer, evaluated at the beginning of the DM, is now given by V b = σ u(q) − δy + βV b ,

(2.9)

provided that both the buyer and seller accept allocation (q, y). This value function is similar to (2.1), except that y is replaced with δy, since there is a 1−δ probability the buyer will not produce in the CM. Similarly the expected lifetime utility of a seller evaluated at the beginning of the DM is now given by V s = σ −c(q) + δy + βV s .

(2.10)

28

Chapter 2

The set of incentive-feasible allocations is almost identical to (2.5), except that, like the value functions above, y is replaced by δy, i.e., the buyer’s promised CM output production is simply adjusted to compensate for the risk of default. This observation is valid as long as the production of the general good is unrestricted. In this case the risk of default has no effect on the set of incentive-feasible allocations. If, however, there is an upper bound on the quantity of goods buyers can produce in the CM, then for a sufﬁciently high probability of default, the set of feasible allocations will be reduced. Next we will see that default risk matters in the presence of heterogeneous buyers and private information. Assume that buyers are heterogeneous in terms of their probabilities of default. There is a measure πH of buyers with a high probability of repayment, δH , and a measure πL = 1−πH with a low probability of repayment, where δL < δH . Alternatively, and equivalently, we could assume that δ ∈ {δL , δH } is an idiosyncratic shock realized by the buyer at the beginning of the period, and these shocks are identically and independently distributed across periods, implying that buyers are ex ante identical. We denote the average probability of repayment by δ¯ = πH δH + πL δL . Assume that the probability of repayment is private information to the buyer. Upon being matched, the trading mechanism offers the buyer a menu of allocations {(qL , yL ), (qH , yH )}. The buyer either chooses an allocation from the menu or declines the offer. If an allocation is chosen, then the seller can either accept or reject it. Trade occurs if both agents accept an allocation. We consider menus of allocations that are stationary, symmetric across agents of a given type, and incentive compatible. By incentive compatibility, we mean that L-type buyers choose allocation (qL , yL ) over qH , yH and H-type buyers choose allocation (qH , yH ) over (qL , yL ). The value function for a buyer of type χ ∈ {L, H} evaluated at the beginning of the DM is Vχb = σ u(qχ ) − δχ yχ + βVχb ,

χ ∈ {L, H}.

(2.11)

This value function is analogous to (2.1), where the term δχ yχ takes into account the probability that the buyer repays his debt. The value function of a seller evaluated at the beginning of the DM is V s = σ E −c(qχ ) + δχ yχ + βV s .

(2.12)

Pure Credit Economies

29

The expectation is with respect to the type χ of buyer with whom the seller is randomly-matched, and (2.12) assumes that a χ -type buyers chooses allocation (qχ , yχ ). A menu of allocations is incentive-feasible if the following conditions are satisﬁed: u(qχ ) − δχ yχ ≥ 0, χ ∈ {L, H}, − c(qχ ) + yχ E δ| qχ , yχ ≥ 0,

(2.13) χ ∈ {L, H},

(2.14)

u(qL ) − δL yL ≥ u(qH ) − δL yH ,

(2.15)

u(qH ) − δH yH ≥ u(qL ) − δH yL .

(2.16)

The conditions (2.13) and (2.14) are the participation constraints for buyers and sellers, respectively. In (2.14), E δ| qχ , yχ represents the seller’s expected value of δ, given that the buyer chose allocation qχ , yχ . Both of these conditions indicate that each agent ﬁnds the proposed menu of allocations acceptable. Inequality (2.15) speciﬁes that an L-type buyer has no incentive to choose the allocation intended for H-type buyers. Similarly inequality (2.16) says that an H-type buyer (weakly) prefers allocation (qH , yH ) to allocation (qL , yL ). Let’s ﬁrst consider a pooling menu of allocations: these are allocations where (qH , yH ) = (qL , yL ) = (q, y). Note that for a pooling menu, the incentive-compatibility conditions (2.15) and (2.16) are automatically satisﬁed, and since the choice of the allocation in the ﬁrst stage of the game conveys no information about the buyer’s type, E δ| q, y = δ¯ as well, if condition (2.13) is satisﬁed for χ = H, then it is automatically satisﬁed for χ = L, since it is more costly for the H-type buyer to fulﬁll his obligation, meaning u q − δL y ≥ u q − δH y. Hence conditions (2.13) for χ = H and (2.14) deﬁne the set of incentive-feasible pooling allocations, AP , which is given by

c(q) u(q) AP = (q, y) ∈ R2+ : . ≤y≤ δH δ¯ The set of incentive-feasible pooling allocations, which is represented by the grey area in ﬁgure 2.4, shrinks as the ratio δH /δL increases. This is because there is a wedge between the expected cost of promising to repay one unit of the general good by the H-type buyer, δH , and the expected beneﬁt of such a promise for the seller, δH πH + δL πL . As δH /δL increases, the expected cost for the H-type buyer increases, relative to the

30

Chapter 2

δy c(q) u(q)

πH δH + πL δL u(q) δH

q

q* Figure 2.4 Incentive-feasible, pooling allocations under exogenous default

seller’s expected beneﬁt, and as a result trade opportunities diminish. The efﬁcient level of production and consumption of the search good, q∗ , can be implemented if

δH − δL ∗ c(q ) ≤ 1 − πL u(q∗ ). δH If δH = δL , then this condition is always satisﬁed. As δH /δL increases, the right side of the inequality decreases, which makes it less likely that the condition will hold. Consider next a separating menu of allocations. These are menus that have allocations characterized by qH , yH = qL , yL . The buyers’ incentive constraints, (2.15) and (2.16), can be rearranged to read δL (yL − yH ) ≤ u(qL ) − u(qH ) ≤ δH (yL − yH ),

(2.17)

while their participation constraints, (2.13) and (2.14), can be written as c(qχ ) ≤ δχ yχ ≤ u(qχ ),

χ ∈ {L, H}.

(2.18)

Pure Credit Economies

31

Since δH > δL , the incentive constraints (2.17) can only be valid if yL ≥ yH and qL ≥ qH , and because the allocation is a separating one, these inequalities are strict. As a result the low-probability repayment buyer consumes more, and produces more, than the high-probability repayment buyer in a separating menu. Hence a separating contract cannot implement an efﬁcient allocation since high- and low-type buyers trade different quantities in the DM. If there is a limit to the amount of general goods that an agent can produce in the CM—and this limit can be arbitrarily large—then from (2.18), as δL approaches zero, qL tends to zero. From (2.17) qH ≤ qL , which implies that qH tends to zero as well. Hence, in a menu characterized by separating allocations, trade will completely shut down if one of the buyer types defaults with probability one. 2.3 Credit with Public Record Keeping In the previous section a default by the buyer was an exogenous event. There was a risk that the buyer would be unable to produce and, hence, repay his debt. In this section we allow for the possibility of strategic default by relaxing the commitment assumption. By strategic default, we mean that the buyer chooses to default even though he has the ability to produce. In order to support trade in a credit economy when agents cannot commit, they must be punished if they do not deliver on their promises. The punishment that we impose is autarky: if an agent fails to deliver on a proposed allocation, then no one will trade with him in the future. Furthermore we will assume that the punishment is global, in the sense that all agents in the economy revert to autarky if at least one agent deviates from proposed play. The basic methodology that underlies the environment comes from the theory of repeated games. This literature teaches us that cooperative outcomes can be achieved by using threats of punishments that are credible. For such punishments to be feasible, players actions must be observable. Hence there is a need for a public record-keeping technol ogy. We can formally deﬁne a record as a list q(i), y(i) i∈[0,σ ] , where i represents a match and [0, σ ] denotes the set of all matches. This record is made available to everyone at the end of each CM. Note that the public record lists only quantities and not the names of the agents associated with the produced quantities. It is for this reason that any deviation from proposed play will result in a global punishment. If names were

32

Chapter 2

associated with quantities, then nonglobal, personalized punishments would be possible. It turns out, however, that very little is changed if individual punishments are possible. We discuss these issues at the end of this section. The chronology of events is as follows: At the beginning of the DM, a measure σ of buyers and sellers are randomly matched. In each match the allocation (q, y) is proposed, which agents simultaneously accept or reject. If the allocation is accepted, then the seller produces q units of the search good for the buyer. In the CM the buyer chooses to either produce y units of the general good for the seller or to renege on hispromise and produce nothing. At the end of the CM, a record q(i), y(i) i∈[0,σ ] of the DM and CM production levels for all matches is publicly observed. Based on this record, agents simultaneously decide whether to continue to trade in the subsequent period or to revert to autarky by playing the global punishment strategy. The global punishment strategy requires that all sellers refuse to extend credit to buyers in all future period matches. Given this punishment strategy, a particular seller in a match has no incentive to extend credit to a buyer since the buyer will not repay his debt. The buyer will renege because he cannot be (further) punished for this behavior, and according to the global punishment strategy, the buyer will not get credit in future matches. We restrict our attention to symmetric, stationary allocations (q, y) that are incentive feasible. Incentive feasibility implies not only that the buyer and the seller agree to allocation (q, y), as before, but also that the buyer is willing to repay his debt when it is his turn to produce. We assume that revert to autarky at the end of the CM whenever all agents q (i) , y (i) = q, y for some i ∈ [0, σ ], i.e., there is at least one trade that is different from the proposed one. Indeed having all agents revert to autarky is an equilibrium outcome in this situation. During the DM, matched buyers and sellers agree to implement allocation (q, y) if −c(q) + y + βV s ≥ 0,

(2.19)

u(q) − y + βV b ≥ 0.

(2.20)

Condition (2.19)—which is the seller’s participation constraint—says that a seller prefers allocation (q, y) plus the continuation value of participating in future DMs and CMs, βV s , to autarky at the time when the match is formed. The seller compares the payoff associated with acceptance to that of autarky because if the seller rejects the proposal, a (0, 0)

Pure Credit Economies

33

trade will be recorded and such a trade will trigger global autarky. Condition (2.20) has a similar interpretation to condition (2.19) but for the buyer; i.e., the buyer prefers suggested trade (q, y) plus the continuation value of participating in future trades to autarky. Note that the participation constraints (2.19) and (2.20) differ from the participation constraints when agents could commit—(2.3) and (2.4), respectively—since now agents go to autarky if they do not accept the proposed allocation (q, y), instead of just being unmatched for the current period. We next need to check that the buyer has an incentive to produce the general good since this production occurs after he consumes the search good in the DM. The buyer will have an incentive to produce the general good if −y + βV b ≥ 0.

(2.21)

The left side of inequality (2.21) is the sum of the buyer’s current and continuation payoffs if he repays his debt by producing y units of output for the seller; the right side is his continuation (autarkic) payoff of zero if he defaults. Note that the buyer’s participation constraint (2.20) is automatically satisﬁed if his incentive constraint (2.21) is satisﬁed. The value functions for the buyer and seller at the beginning of the period are by equations (2.1) respectively, still given and (2.2), b s i.e., V = σ u q − y / (1 − β) and V = σ −c q + y / (1 − β). These functions imply that the seller’s participation constraint (2.19) and the buyer’s incentive constraint (2.21) can be rewritten as − c(q) + y ≥ 0, σ u(q) − y ≥ y, r

(2.22) (2.23)

respectively, where r = β −1 − 1. Condition (2.22) simply says that the seller is willing to participate if he gets some surplus from trade. It is interesting that this participation condition does not depend on discount factors or matching probabilities. Condition (2.23) represents the incentive constraint for the buyer to repay his debt. The left side of (2.23) is the buyer’s expected payoff beginning next period, provided that he does not renege on his debt obligation this period; it is the discounted sum of expected surpluses from future trade. This expression depends on both the frequency of trades, σ , and the discount rate, r. The right side of (2.23) represents the buyer’s (lifetime) gain if he does not produce the

34

Chapter 2

general good for the seller this period. Not surprisingly, a necessary, but not sufﬁcient, condition for inequality (2.23) to hold is that the buyer’s surplus from the trade is positive, i.e., u(q) − y ≥ 0. Note that (2.22) and (2.23), along with (2.1) and (2.2), imply that V s ≥ 0 and V b ≥ 0, so agents are better off continuing to trade than being in autarky. The set of incentive-feasible allocations when agents cannot commit, but when public record keeping is available, APR , can be obtained directly from inequalities (2.22) and (2.23), i.e.,

σ APR = (q, y) ∈ R2+ : c(q) ≤ y ≤ u(q) . (2.24) r+σ This set, which is represented by the gray area in ﬁgure 2.5, is smaller than the set of incentive-feasible allocations when agents can commit, AC ; see ﬁgure 2.2. This is a consequence of the additional incentive constraint, (2.21), that must be imposed when buyers are unable to commit to repay their debts. The set APR expands as the frequency of trades, σ , increases or as agents become more patient, i.e., when r decreases. Note also that APR → AC when r → 0, since the cost of defaulting, which y c(q) u(q)

σ u(q) r+σ

q* Figure 2.5 Incentive-feasible allocations under public record keeping

q

Pure Credit Economies

35

is the expected discounted sum of future trade surpluses, becomes inﬁnite. The efﬁcient production and consumption level of the search good, q∗ , will be incentive feasible if c(q∗ ) ≤

σ u(q∗ ). r+σ

(2.25)

Suppose that inequality (2.25) holds for particular values of σ and r. If the probability of ﬁnding a future match, σ , is decreased, then the beneﬁt of avoiding autarky is reduced. If σ decreases sufﬁciently, then there will be no value for y that gives the buyer an incentive to repay his debt, and makes the seller willing to produce q∗ . In this situation the efﬁcient level of production and consumption of the search good, q∗ , is not incentive feasible. One can see this graphically, if the [σ/(r + σ )]u(q) curve in ﬁgure 2.5 intersects the c(q) curve at a value of q less than q∗ . Similarly, if buyers discount the future more heavily, i.e., if β decreases or if r increases, the buyer will have a greater incentive to renege on his debt since he cares more about his current payoff than future payoffs. For each level of search friction in the DM, σ ∈ (0, 1], there exists a ¯ ), such that if β ≥ β(σ ¯ ), then an threshold for the discount factor, β(σ ¯ ) is a efﬁcient allocation (q∗ , y) is incentive feasible. This threshold β(σ decreasing function of σ , which means that the efﬁcient level of production and consumption of the search good, q∗ , is easier to sustain ¯ ), then when there are lower frictions in the DM. If, however, β < β(σ the incentive-feasible allocations will be characterized by an inefﬁciently low level of the search good, i.e., q < q∗ . Two of the assumptions regarding punishments can be relaxed. First, we have assumed that if an agent in a match does not accept the proposed offer in the DM, then the economy will forever revert to autarky starting in the next period. This is reﬂected by the zero payoff on the right sides of (2.19) and (2.21). This assumption is harmless in the sense that if agents were not punished for rejecting the proposed offer, then they would still accept all of the equilibrium offers that are supported by the autarky punishment. Formally, we could replace the two participations constraints (2.19) and (2.20) with (2.3) and (2.4). Second, we have assumed that if an agent defects from proposed play, then the economy will revert to global autarky forever. Such an assumption is necessary when an agent who defects from equilibrium play cannot be identiﬁedby other agents in the economy. If, however, a record is now the list q(i), y(i), b(i), s(i) i∈[0,σ ] , where b(i) ∈ [0, 1] is

36

Chapter 2

the identity of the buyer in match i and s(i) is the identity of the seller in match i, then it is possible to support credit arrangements through individual punishments. That is, all of the credit arrangements above can be sustained without having to revert to global autarky in the event of a defection from a proposed allocation. So far we have been silent on how the allocation (q, y) is determined. One can think of the allocation as being the outcome of a bargaining game between a buyer and a seller. In chapter 2.1 we determined this outcome by using the axiomatic Nash solution. Here we will choose a noncooperative game form, where the buyer proposes an allocation to the seller, which the seller can either accept or reject. If the offer is accepted, then the allocation associated with it is publicly recorded and the seller produces the DM output associated with the allocation. If the buyer defaults on his obligation to produce y units of general goods in the CM, then the economy will revert to autarky. If the offer is rejected, then no trade takes place, and the buyer and the seller split apart. Without loss of generality, we assume that there is no punishment if the seller rejects the offer. The buyer’s problem is max u(q) − y , (2.26) q,y

subject to and

− c(q) + y ≥ 0,

− y + βV b ≥ 0.

(2.27) (2.28)

The buyer’s incentive-compatibility constraint (2.28) must hold; otherwise, the buyer would renege on his promise to repay y units of general goods and the seller, who understands the buyer’s incentive, would reject such an offer. The solution to (2.26)–(2.28) is such that q = q∗ and y = c(q∗ ) if c(q∗ ) ≤ βV b ; otherwise, q = c−1 (βV b ) and y = βV b . Substituting this solution into (2.1) and (2.2), it is easy to check that V s = 0 and σ u(q) − c(q) b V = . 1−β This simple noncooperative an efﬁcient allocation game implements if and only if c(q∗ ) ≤ βσ u(q∗ ) − c(q∗ ) /(1 − β) = σ u(q∗ ) − c(q∗ ) /r. This condition is diagrammatically depicted in the left panel of ﬁgure 2.6. This condition is equivalent to (2.25), so whenever the ﬁrst-best allocation is incentive-feasible, it can be implemented by a simple mechanism where the buyer makes a take-it-or-leave-it offer to the seller. If c(q∗ ) > σ u(q∗ ) − c(q∗ ) /r, then the quantity traded in a credit

q*

σ [u(q)−c(q)] r c(q)

q

Figure 2.6 Credit equilibrium under take-it-or-leave-if offers by buyers.

Efficient credit equilibrium

βV b =

q*

βV b =

q

σ [u(q)−c(q)] r

Inefficient credit equilibrium

qe

c(q)

Pure Credit Economies 37

38

Chapter 2

equilibrium is the largest solution to c q = σ u q / (r + σ ); see the solution qe in the right panel of ﬁgure 2.6. In this situation the quantity traded increases with σ and decreases with r. It should also be noted that autarky, i.e., q, y = (0, 0), is always an equilibrium. If agents anticipate that they will not be able to trade in the future because credit is not available, then they will not trade in the current period. If they trade in the current period, then buyers would renege on their repayment in the CM because they cannot be punished any worse than autarky. 2.4 Credit with Reputation The public nature of the record-keeping technology is a rather strong assumption. In this section we illustrate that a much weaker recordkeeping technology—private memory—can still be powerful in terms of sustaining credit arrangements when buyers and sellers have repeated interactions. It is well known that cooperation can be sustained when agents repeatedly interact with one another. With repeated interactions, agents are able to develop reputations for behaving appropriately. We assume that agents who are in a trade match during the DM can form a long-term partnership that can be maintained beyond the current period. That is, agents can continue their trade match or partnership into the next period if they so desire. We allow for both the creation and destruction of a partnership. At the end of each period, an existing partnership is exogenously destroyed with probability λ ∈ (0, 1). One can justify this exogenous destruction by supposing that the buyers and/or the sellers are hit by a relocation shock and, as a result, permanently lose contact with one another. Agents can also choose to terminate a partnership at will. For example, the seller may choose to dissolve the partnership by looking for alternative trading partners if the buyer does not deliver on his promise to produce the general good. This sort of termination is important because it provides the seller with a punishment vehicle—i.e., the destruction of the asset value of an enduring match or partnership—that is required to make a partnership viable in the ﬁrst place. Notice that walking away from a partnership when the buyer does not deliver on his promise to produce the general good is a subgame perfect equilibrium since both players make their decisions simultaneously. That is, since agents make their decisions simultaneously, walking away from a partnership is a best response for an agent to the same strategy by the other agent.

Pure Credit Economies

DAY

A fraction σ of unmatched agents find a match.

Matched sellers produce q

39

NIGHT

Matched buyers produce y

A fraction λ of matches are destroyed

Figure 2.7 Timing of the representative period

The chronology of events is illustrated in ﬁgure 2.7. At the beginning of the DM, unmatched buyers and sellers participate in a random matching process. With probability σ , a buyer (seller) is matched with a seller (buyer). Buyers and sellers whose match was not destroyed at the end of the previous period simultaneously and independently decide whether to look for the previously matched partner or look for a new partner. If two old partners search for each other, then they ﬁnd one another with probability one, and the partnership is maintained. If either one of them looks for a new partner, the match is terminated. In each match an allocation (q, y) is proposed, which agents can either accept or reject. If both agents accept the offer, the seller produces q units of the search good for the buyer in the DM. In the CM, the buyer chooses whether or not to honor his implicit obligation by producing y units of the general good for the seller. At the end of the CM, either the partnership is exogenously destroyed or it can continue into the next period. Partnerships can be formed and maintained only during the random matching process at the beginning of the DM. We will characterize the set of symmetric stationary equilibrium allocations for this economy. Let et denote the total measure of partnerships during the DM in period t, after the matching phase is completed. Assuming that buyers do not renege on their promises, the law of motion for et is et+1 = (1 − λ)et + σ [1 − (1 − λ)et ] .

(2.29)

According to (2.29), if there are et partnerships in period t, a fraction (1 − λ) of them will be maintained into period t + 1. Among the 1 − (1 − λ)et agents who are unmatched at the beginning of t + 1, a fraction σ ﬁnd new partners. In the steady state, et+1 = et = e¯ , which, from (2.29), implies that

40

e¯ =

Chapter 2

σ . σ + λ(1 − σ )

(2.30)

The number of matches increases with the matching probability σ and decreases with the destruction probability λ. Let Veb be the value function for a buyer who is in a partnership at the beginning of a period and Vub the value function for a buyer who is not, where e stands for employed in a partnership and u for unmatched. Then, assuming that the buyer does not renege and neither party voluntarily terminates the partnership, Veb = u(q) − y + λβVub + (1 − λ)βVeb ,

(2.31)

Vub = σ Veb + (1 − σ )βVub .

(2.32)

According to (2.31) the buyer receives q units of search goods in the DM and produces y units of general good in the CM. The partnership is exogenously destroyed with probability λ, in which case the buyer goes to the random-matching process at the beginning of the next period to ﬁnd a new partner. According to (2.32) an unmatched buyer ﬁnds a seller with probability σ . If the buyer does not get matched, then, with probability 1 − σ , he starts the next period unmatched. The closed-form solutions for Veb and Vub , whose derivation can be found in the appendix, are given by [1 − (1 − σ )β] u(q) − y b Ve = , (2.33) (1 − β) [1 − (1 − σ )(1 − λ)β] σ u(q) − y b Vu = . (2.34) (1 − β) [1 − (1 − σ )(1 − λ)β] Let Ves be the value function for a seller who is in a partnership at the beginning of the period and Vus the value function for a seller who is not. Then Ves = −c(q) + y + λβVus + (1 − λ)βVes ,

(2.35)

Vus = σ Ves + (1 − σ )βVus .

(2.36)

According to (2.35) the seller produces q units of the search good during the DM and consumes y units of the general good in the CM. With probability λ the partnership is destroyed, in which case the seller enters

Pure Credit Economies

41

the random-matching process at the beginning the next period. According to (2.36), with probability σ , the seller is matched with a buyer who likes his search good. The closed-form solutions for Ves and Vus are given by [1 − (1 − σ )β] −c(q) + y s Ve = , (2.37) (1 − β) [1 − (1 − σ )(1 − λ)β] σ −c(q) + y s Vu = . (2.38) (1 − β) [1 − (1 − σ )(1 − λ)β] Allocation (q, y) can be implemented as an equilibrium outcome if three sets of conditions are satisﬁed. First, agents who enter the DM unmatched, and subsequently become matched, will accept the proposed allocation (q, y) if the following participation constraints hold: Ves ≥ βVus ,

(2.39)

Veb ≥ βVub .

(2.40)

If the seller and buyer accept the allocation (q, y), then their expected payoffs are given by the left sides of (2.39) and (2.40), respectively, and if they reject, the continuation payoffs are given by the right sides. Second, at the beginning of a period, matched sellers and buyers who do not receive a relocation shock will agree to continue their partnership if Ves ≥ Vus ,

(2.41)

Veb ≥ Vub .

(2.42)

If the seller and the buyer choose to continue the partnership, their payoffs are given by the left sides of (2.41) and (2.42), respectively; if either or both choose to dissolve the partnership, the expected payoffs are given by the right sides. Clearly, conditions (2.41) and (2.42) imply that (2.39) and (2.40) hold, respectively. Note that from (2.33) and (2.34), and (2.37) and (2.38), the surpluses that a buyer and a seller receive are given by, respectively, (1 − σ ) u(q) − y b b Ve − Vu = , (2.43) 1 − (1 − σ )(1 − λ)β

42

Ves − Vus

Chapter 2

(1 − σ ) −c(q) + y = . 1 − (1 − σ )(1 − λ)β

(2.44)

From these surpluses,we can deduce that (2.41) and (2.42) are satisﬁed if u q − y ≥ 0 and −c y + y ≥ 0. Third, a buyer in a partnership must be willing to produce the general good for the seller in the CM. This requires that −y + λβVub + (1 − λ)βVeb ≥ βVub .

(2.45)

If the buyer produces y units of the general good, then his expected payoff is given by the left side of (2.45). If, however, he deviates and does not produce, then the partnership will be dissolved at the beginning of the subsequent period, and the buyer will start the next DM seeking a new match; the utility associated with this outcome is given by the right side of (2.45). This constraint can be re-expressed as y ≤ (1 − λ)β Veb − Vub or, using (2.43), y ≤ β (1 − λ) (1 − σ ) u q . The set of incentive-feasible allocations that can be supported by reputations is given by AR = (q, y) : c(q) ≤ y ≤ β(1 − λ)(1 − σ )u(q) .

(2.46)

This set is represented by the gray area in ﬁgure 2.8. The buyer’s incentive-compatibility condition (2.45) generates an endogenous borrowing constraint, y ≤ β(1 − λ)(1 − σ )u(q). This borrowing constraint indicates that the maximum amount the buyer can promise to repay in the CM depends on his patience, β, the stability of the match, λ, and market frictions, σ . The buyer is able to credibly promise to repay more, the more patient he is, i.e., the higher is β; the more stable is the match, i.e., the lower is λ; the greater is the matching friction, i.e., the lower is σ ; and the higher is his consumption the next DM, i.e., the higher is q. Note that the set of incentive-feasible allocations, AR , will be empty if either all matches are destroyed at the end of a period, if λ = 1, and/or if agents can ﬁnd partners in the DM with certainty, if σ = 1. The existence of credit relationships relies on the threat of termination, but such a threat only has bite if matches are not easily destroyed and if it is difﬁcult to create a new trade match. The efﬁcient production and consumption level of the search good, ∗ ∗ R q , is implementable if and only if q , y ∈ A , c(q∗ ) ≤ β(1 − λ)(1 − σ )u(q∗ ).

(2.47)

Pure Credit Economies

43

y c(q) u(q)

β(1 − λ)(1− σ)u(q)

q*

q

Figure 2.8 Incentive-feasible allocations with reputation

Agents are able to trade the quantity q∗ through long-term partnerships if the average duration of a long-term partnership is high, i.e., if λ is low, if the matching frictions are severe, i.e., if σ is low, and if agents are patient, i.e., if β is close to one. Figure 2.8 characterizes a situation where the efﬁcient production and consumption level of the search good is implementable. Diagrammatically speaking, the β(1 − λ)(1 − σ )u(q) curve must intersect the c q curve at a q > q∗ in ﬁgure 2.8. The model environment in this section provides an example where social welfare is not monotonic in the underlying search friction σ . Social welfare at the steady state, W , is deﬁned to be the measure of trade matches, e¯ given by (2.30), times the surplus of each match, i.e., W = {σ/ [σ + λ(1 − σ )]} u(q) − c(q) . Let W ∗ denote the maximum social welfare over the set of incentive-feasible allocations,

σ ∗ R W (σ ) = max u(q) − c(q) : q, y ∈ A . σ + λ(1 − σ ) To see that W ∗ (σ ) is not a monotonic function of σ , note that the measure of matches, e, is increasing in σ while the set of incentive-feasible q’s

44

Chapter 2

shrinks as σ increases. If σ = 0 or σ = 1, then W ∗ = 0 since at σ = 0 the number of matches is zero and at σ = 1 the set of implementable allocation is AR = {(0, 0)}; if, however, σ ∈ (0, 1), then W ∗ (σ ) > 0. Finally, we show that if the planner were free to choose the extent of the search frictions, σ , in the DM, then the optimal σ would have the repayment constraint, c(q) ≤ β(1 − λ)(1 − σ )u(q), bind and q < q∗ . To see this, deﬁneσ ∗ asthe solution to c(q∗ ) = β(1 − λ)(1 − σ ∗ )u(q∗ ). Clearly, for all σ ∈ 0, σ ∗ , the repayment constraint is not violated at q = q∗ ; hence a planner would never choose σ < σ ∗ . As σ increases above σ ∗ , the number of matches increases, but the repayment constraint starts to bind. A small increase of σ above σ ∗ therefore has a ﬁrst-order effect on the number of matches, but only a second-order effect on match surplus, u q − c q . Hence the optimal σ is greater than σ ∗ , which implies that the repayment constraint binds and q < q∗ . 2.5 Further Readings Pairwise credit in a search-theoretic model was ﬁrst introduced by Diamond (1987a, b, 1990). The environment is similar to Diamond (1982), where agents are matched bilaterally and trade indivisible goods. As in our setup, credit is repaid with goods. The punishment for not repaying a loan is permanent autarky. There are several models where agents have some private information about their ability to repay their debt. Aiyagari and Williamson (1999) consider a random-matching model where agents receive random endowments that are private information and exchange is motivated by risk-sharing. The optimal allocations have several features similar to those of real-world credit arrangements, such as credit balances and credit limits. Smith (1989) constructs an overlapping generations model where agents have stochastic endowments and can misrepresent the nature of the liabilities they issue. Jafarey and Rupert (2001) study an economy with alternating endowments, where the set of agents who issue debt is divided into two classes: safer and riskier borrowers. The former have a higher probability of redeeming their debt than the latter. Kocherlakota (1998a, b) describes credit arrangements in different environments, including a search-matching model with a public record of individual transactions. He uses mechanism design to characterize the set of symmetric, stationary, and incentive-feasible allocations. Kocherlakota and Wallace (1998) extend the model to consider the case where the public record of individual transactions is updated after a

Pure Credit Economies

45

probabilistic lag. They establish that society’s welfare increases as the frequency with which the public record is updated increases. As pointed out by Wallace (2000), this is the ﬁrst model that formalizes the idea that technological advances in the payment system improve welfare. The model by Kocherlakota and Wallace has been extended by Shi (2001) to discuss how the degree of advancement of the credit system affects specialization. Kehoe and Levine (1993) develop a theory of endogenous debt limits. If an agent defaults on a contract, he can be excluded from trading on futures markets and can have his assets seized. However, the agent is not forced into autarky as he cannot be excluded from spot market trading and has private endowments that cannot be seized. Kehoe and Levine (1993) ﬁnd that for discount factors close to one, ﬁrst-best allocations may be sustained. Most search-theoretic models of the labor market (e.g., Pissarides 2000) assume long-term partnerships. However, in these economies trades do not involve credit and are free of moral hazard considerations. Corbae and Ritter (2004) consider an economy with pairwise meetings, where agents can form long-term partnerships to sustain credit arrangements. A related model of reciprocal exchange is also presented by Kranton (1996). Appendix The Generalized Nash Bargaining Solution The Nash solution is an axiomatic bargaining solution proposed by Nash (1953). It is based on four axioms—Pareto efﬁciency, scale invariance, independence of irrelevant alternatives, and symmetry—and it predicts a unique outcome to a bargaining problem. Moreover it has solid strategic foundations (e.g., see Osborne and Rubinstein 1990). In our context, the generalized Nash bargaining solution, which generalizes the Nash solution by dropping the axiom of symmetry, is given by the solution to (2.6) i.e., (q, y) = arg max θ ln u q − y + (1 − θ ) ln y − c q . q,y

The ﬁrst-order conditions are θ u q (1 − θ ) c q = 0, − u q −y y−c q

(2.48)

46

Chapter 2

θ (1 − θ ) = 0. − + (2.49) u q −y y−c q It is immediate that u q = c q , or q = q∗ , and y = (1 − θ ) u q∗ +θ c q∗ . Derivation of Equations (2.33) and (2.34) The system (2.31)–(2.32) can be rewritten under the following matrix form:

b

Ve 1 − (1 − λ)β −λβ u(q) − y = . −σ 1 − (1 − σ )β 0 Vub By inverting the ﬁrst matrix, we obtain

Veb 1 − (1 − σ )β λβ u(q) − y −1 = , σ 1 − (1 − λ)β 0 Vub where = [1 − (1 − λ)β] [1 − (1 − σ )β] − σ λβ. The determinant of the matrix can be re-expressed as = (1 − β) [1 − (1 − σ )(1 − λ)β] ∈ (0, 1) . Consequently the closed-form solutions for the value functions of a buyer are [1 − (1 − σ )β] u(q) − y Veb = , (2.50) (1 − β) [1 − (1 − σ )(1 − λ)β] σ u(q) − y b Vu = . (2.51) (1 − β) [1 − (1 − σ )(1 − λ)β] By similar reasoning we can solve for the closed-form solution of a seller: [1 − (1 − σ )β] −c(q) + y s Ve = , (2.52) (1 − β) [1 − (1 − σ )(1 − λ)β] σ −c(q) + y s Vu = . (2.53) (1 − β) [1 − (1 − σ )(1 − λ)β]

3

The Role of Money

In the previous chapter we showed that credit arrangements allow agents to take advantage of intertemporal gains from trade. If, however, creditors do not trust debtors to repay their debts, then trade by credit may not be incentive feasible. If agents are to trade with one another, then some sort of tangible medium of exchange must emerge. According to Kiyotaki and Moore (2002, p. 64), a lack of trust is of primary importance for a theory of money: As they say, “distrust is the root of all money.’’ In this chapter we assume that buyers and sellers never trust one another because they cannot commit to repay their debts, and there is no record-keeping technology or reputational device that can make debt contracts self-enforcing. In the absence of some tangible means of payment, buyers and sellers cannot trade and will live in a world of autarky. In order to illustrate the role of a tangible medium of exchange, we consider the following simple experiment. We endow each buyer with one unit of an indivisible, durable, but intrinsically useless object, called ﬁat money. We then characterize the set of allocations that can be achieved in this economy. If the set of allocations that is feasible with money is larger than the set that is feasible without money, and if it includes some socially preferred allocations, then we will say that money is essential. Money can play an essential role because it allows society to achieve (preferred) outcomes that could not be achieved without it. The main purpose of this chapter is to demonstrate that money is essential in an environment where agents cannot commit and there is no record-keeping technology. Moreover all the allocations that are feasible under a public record-keeping environment will also be feasible

48

Chapter 3

in a monetary economy. Hence the technological role of money can be identiﬁed as one of record keeping. In Kocherlakota’s (1998a, b) words, money is memory. We show that the best allocations (from society’s point of view) can be decentralized by simple trading protocols such that the terms of trade for the search good are determined by a bargaining procedure, where the buyer has all of the bargaining power, and the general good is traded in a competitive market. 3.1 Money Is Memory In this section we compare the set of allocations that are incentive feasible in a monetary economy to that of a credit economy. To facilitate this comparison, we will structure the environment of the monetary economy so that it is similar to a credit economy with public record keeping. In particular, we assume that (1) a buyer and seller are matched with probability σ during the day in the decentralized market, DM; (2) matched agents stay together during the night, and (3) all matches that are formed during the DM are dissolved before the start of the next period, never to be formed again. In addition we assume, for the time being, that there is no centralized market during the CM: only matched agents can interact at this time. In the next section we relax this assumption and show that it is unimportant for our insights. The sequence of events in a typical period is illustrated in ﬁgure 3.1. At the beginning of the DM, a measure σ of buyers and sellers are randomly matched. The buyer has mb ∈ {0, 1, 2, . . .} ≡ N0 units of indivisible money and the seller has ms ∈ N0 . A trading mechanism proposes a contract or allocation to matched buyers and sellers. A trading mechanism can be thought of as a mapping from the money holdings of agents in a match, into the set of of the form s ), am (mb , mpm contracts, or allocations, q, d , y, d . Allocation q, dam , y, dpm has the interpretation that the seller produces q units of the search good for the buyer and the DAY (DM)

NIGHT

σ bilateral matches are formed. Matched sellers produce q. Matched buyers transfer d am units of money.

Matched buyers produce y. Matched sellers transfer d pm units of money. Matches are destroyed.

Figure 3.1 Timing of the representative period

The Role of Money

49

buyer transfers dam ∈ N units of indivisible money to the seller during the DM, and the buyer produces y units of the general good in exchange for dpm units of money in the CM. The proposed allocation must be feasible, which implies that −ms ≤ dam ≤ mb and −mb + dam ≤ dpm ≤ ms + dam , i.e., agents cannot be asked to transfer more money than they hold. After the allocation is proposed, agents simultaneously accept or reject it. If one of the agents rejects the offer, then the outcome will be the notrade allocation [(0, 0) , (0, 0)]. If the allocation is accepted by both the buyer and seller, then agents produce and exchange money balances as speciﬁed in the contract. At the end of the period, all matches are dissolved. We assume that at the start of time each buyer is endowed with one unit of money. We make this assumption for two reasons. First, because we are interested in allocations that maximize social welfare, we want all buyers to be able to trade. This implies that, at least initially, all buyers must be endowed with some money. Second, because we will focus our attention on symmetric allocations, buyers must all have the same quantity of money. We choose this quantity to be equal to one. We examine allocations that require the buyer to transfer one unit of money to the seller in exchange for the search good in the DM, and the seller to transfer one unit of money to the buyer in exchange for the general good in the CM. In particular, inthe DM the trading mecha nism proposes the allocation (q, 1), (y, 1) to a buyer in a match who is holding mb ≥ 1 units of money. This allocation requires that, in the DM, agents should trade q ≥ 0 units of output for one unit of money and, at night, trade y ≥ 0 units of output for one unit of money. If a matched buyer does not have a unit of money, i.e., mb = 0, then the trading mechanism proposes the no-trade allocation [(0, 0) , (0, 0)]. The value function, or the lifetime expected discounted utility, for a buyer holding m ≥ 1 units of money at the beginning of the DM when the mechanism proposes allocation (q, dam ), (y, dpm ) = (q, 1), (y, 1) is V b (m) = σ u(q) − y + βV b (m − dam + dpm ) + (1 − σ ) βV b (m) (3.1) = σ u(q) − y + βV b (m). According to (3.1) the buyer meets a seller with probability σ . When this event occurs, and the buyer and seller agree to the proposal (q, 1), (y, 1) , the buyer will consume q units of the search good during the DM in exchange for his unit of money. At night the buyer gets

50

Chapter 3

his unit of money back from the seller by producing y units of the general good. From (3.1), the closed-form expression for the value of a buyer b is V (m) = σ u(q) − y / (1 − β) for m ≥ 1. If the buyer does not hold money at the beginning of the DM, then the mechanism proposes the no-trade allocation, and hence V b (0) = βV b (0) = 0. This implies that a buyer without money is in autarky forever since it is impossible for him to ever get a unit of money. The value function of a seller holding m ≥ 0 units of money at the of the beginning DM when the mechanism proposes allocation (q, dam ), (y, dpm ) = (q, 1), (y, 1) is V s (m) = σ −c(q) + y + βV b (m + dam − dpm ) + (1 − σ ) βV s (m) (3.2) = σ −c(q) + y + βV s (m). According to (3.2) a seller meets a buyer with probability σ . In the event that a match occurs, the seller produces q units of the search good for the buyer in exchange for one unit of money in the DM. At night the seller exchanges one unit of money for y units of the general good. If the seller chooses not to exchange his unit of money for the general good, then the seller will never be able to spend that unit of money in the future (since the trading mechanism proposes allocation (q, 1), (y, 1) in all future matches). Indeed, by this mechanism, the seller spends the unit of money he acquired in the DM from his matched buyer at night and does not have the opportunity to spend any money acquired from previous trade matches. Hence, from (3.2), the closed form expression s for the value of a seller is V (m) = σ −c(q) + y / (1 − β) for all m ≥ 0. In the next section we allow agents to trade in a competitive market at night, so they are able to spend any money acquired from previous trade if they desire to do so. Since agents cannot commit and there is no public record of the trades, the buyer can always renege on his implicit promise to produce y units of output at night in exchange for dpm units of money. Hence the proposed allocation must be self-enforcing. In order for allocation (q, 1), (y, 1) to be incentive feasible, agents must be willing to participate in trades in both the DM and at night. Incentive feasibility requires that u(q) − y + βV b (1) ≥ βV b (1),

(3.3)

− c(q) + y + βV s (0) ≥ βV s (0),

(3.4)

The Role of Money

51

− y + βV b (1) ≥ βV b (0) = 0,

(3.5)

y + βV s (0) ≥ βV s (1) = βV s (0).

(3.6)

Conditions (3.3) and (3.4) are participation constraints that say a matched buyer and the seller, respectively, will accept the proposed allocation. Note that these conditions imply that c(q) ≤ y ≤ u(q), i.e., the buyer and seller must receive nonnegative surpluses. Conditions (3.5) and (3.6) are incentive-compatibility constraints that say the buyer will produce y units of the general good in the DM and the seller will exchange his unit of money for y units of the general good at night. It should be clear that condition (3.6) will be satisﬁed whenever y ≥ 0, which implies that condition (3.6) will hold whenever (3.4) holds. Condition (3.5), in conjunction with the value function (3.1), can be rewritten as y≤

σ u(q), r+σ

(3.7)

which implies that condition (3.3) will hold when (3.5) holds. Hence the set of incentive-feasible allocations in this monetary economy, AM , is described by conditions (3.4) and (3.5), or by

σ AM = (q, y) ∈ R2+ : c(q) ≤ y ≤ u(q) . (3.8) r+σ Note that this set is larger than the set of incentive-feasible allocations that would prevail without money, which is {(0, 0)}. Agents are forced into autarky in an economy without money because, owing to a lack of commitment and a record-keeping technology, credit arrangements are not incentive feasible. Moreover the set AM includes allocations that are preferred to the autarky allocations by both buyers and sellers. It is in this sense that money plays an essential role in this economy. Money allows for allocations that are not incentive feasible otherwise, and these new allocations increase the welfare of society. What is the exact role that money performs? To answer this question, we compare the set of allocations in the monetary economy with the set obtained in an economy with public record keeping. The set of incentive-feasible allocations in the monetary economy is identical to the set of incentive-feasible allocations in a credit economy with a public record, i.e., AM = APR where the set APR is described by

52

Chapter 3

equation (2.24) of chapter 2. It is in this sense that money is equivalent to a public record-keeping mechanism. In fact the equivalence is complete between the two when the public record-keeping mechanism does not punish agents to autarky for rejecting the proposed allocation in the DM since all of the comparable incentive constraints are identical. Money has the technological role of memory because an agent’s money balances act as a state variable that conveys some information about his past trading behavior. By holding one unit of money at the beginning of a period, a buyer indicates that he has produced in the past whenever it was his turn to produce. If he does not hold one unit of money, it means that in the past he reneged on his promise to produce for a seller. In this situation the buyer is “punished’’ by being unable to consume in all periods that follow this deviation from the required allocation. 3.2 Decentralizing Allocations We now impose explicit trading protocols for both the day and night markets to see if it is possible to generate the socially desirable allocations described in the previous section. In the DM, the terms of trade in bilateral matches are determined by a bargaining process. We assume that the buyer makes a take-it-or-leave-it offer to the seller. At night, we assume that all the matched agents, as well as the unmatched ones, meet together in a competitive market, CM. Here price-taking agents can buy and sell units of money in exchange for general goods at the marketclearing price φt , where φt represents the quantity of general goods that can be purchased in the CM for one unit of money in period t. We restrict our attention to stationary equilibria, where φt = φ is constant over time, and maintain the assumption that money is indivisible. The timing of events in a typical period is similar to that of the previous section; see ﬁgure 3.2. At the beginning of the DM, a measure σ of buyers and sellers are randomly matched, where the buyer has mb ∈ N0 ≡ {0, 1, 2, . . .} units of money, and the seller has ms ∈ N0 . In each match, DAY (DM)

NIGHT (CM)

σ matches are formed. Sellers produce q. Buyers transfer d.

Agents trade money for goods at the competitive price ϕ. Agents split apart.

Figure 3.2 Timing of a representative period

The Role of Money

53

the buyer makes a take-it-or-leave-it offer q, d to the seller, where q represents the amount of the search good to be produced by the seller and d ∈ N0 the amount of money that he receives, where −ms ≤ d ≤ mb . At night, all buyers and sellers participate in the CM. In the stationary equilibrium buyers produce the general good and receive money, while sellers purchase the general goods with money. At the end of the period all the agents split apart. The model can be solved in four steps: 1. Characterize some key properties of the value functions in the CM. 2. Determine the terms of trade in a bilateral match in the DM. 3. Characterize the Bellman equations for the DM. 4. Determine the buyer’s and seller’s choice of money holdings in the CM. The ﬁrst step characterizes the CM value functions. The value function for a buyer at the beginning of the CM satisﬁes W b (m) = max x − y + βV b (m ) (3.9) m ∈N0 ,x,y

subject to

x + φm = y + φm.

(3.10)

The budget constraint (3.10) says that the buyer ﬁnances his endof-period money balances, m , and general good consumption, x, with production of the general good, y, and with money balances brought into the CM, m. Substituting the budget identity (3.10) into the maximand of (3.9), we get W b (m) = φm + max −φm + βV b (m ) . (3.11) m ∈N0

Equation (3.11) tells us that the buyer’s CM value function is linear in the money balances, m, brought into the CM. This is an important result that comes about from the linearity of the CM utility function, x − y. An implication of such preferences is that the buyer’s wealth, which is composed only of real balances, does not affect his choice of money holdings for the future. This result is crucial for the tractability of the model because, otherwise, the idiosyncratic trading shocks in the DM—a buyer is matched with probability σ —would create a nondegenerate distribution of money holdings at the end of the DM. The nondegeneracy would occur because buyers who are matched in the DM hold fewer money

54

Chapter 3

balances in the CM than buyers who are unmatched. In the presence of wealth effects, the heterogeneity in money holdings that results from the trading shocks in the DM would persist in the subsequent CM, as well as in subsequent periods. Generally speaking, it is difﬁcult, if not impossible, to obtain analytical solutions when this sort of heterogeneity is present; one must instead rely on numerical methods. We avoid these issues by restricting preferences to be quasi-linear. The evolution of the distributions of money holdings for buyers and sellers over a period is represented in ﬁgure 3.3. Buyers start the period with mb units of money, where mb will typically be equal to the money supply, M. Sellers start with ms , typically equal to zero. The fraction σ of buyers who are matched end the DM with mb − d units of money, where d is the amount spent in a match. Similarly the fraction σ of sellers who are matched end the DM with ms + d units of money. Buyers and sellers in the CM readjust their money holdings so that all buyers end the period with mb units of money and all sellers end the period with ms units of money. The second key feature of (3.11) is that the value function at the beginning of the CM is linear in money holdings, i.e., W b (m) = φm + W b (0). This property will prove especially usefulwhen solving the bargaining problem and determining terms of trade, q, d , in the bilateral matches in the DM. DM

σ Buyers

mb – d

mb = M

mb = M 1– σ

σ Sellers

CM

mb ms + d

ms = 0

ms = 0 1– σ

ms

Figure 3.3 Evolution of the distributions of money holdings over a period

The Role of Money

55

Following a similar line of reasoning, a seller’s value function in the CM is W s (m) = φm + max −φm + βV s (m ) , m ∈N0

(3.12)

where V s (m) is the value function for the seller at the beginning of the period. The second step determines the terms of trade in a bilateral match in the DM. We assume that the buyer, who holds mb units of money, makes a take-it-or-leave-it offer to the seller, who holds ms units of money. The buyer’s problem is max u(q) + W b (mb − d) (3.13) q,d

subject to and

− c(q) + W s (ms + d) ≥ W s (ms )

− ms ≤ d ≤ mb .

(3.14) (3.15)

The inequality (3.14) is the seller’s participation constraint. An optimal offer is such that this constraint holds at equality. According to (3.15), a transfer of money is feasible if the buyer does not offer to transfer more units of money than he holds, or he does not ask for money units of money that the seller holds. Given the linearity of the value functions W b and W s , we can simplify and rewrite the buyer’s problem as max u(q) − φd (3.16) q,d

subject to and

− c(q) + φd = 0

− ms ≤ d ≤ mb .

(3.17) (3.18)

Given that production of the search good is nonnegative, q ≥ 0, it should be clear from (3.17) that a solution to the buyer’s problem cannot be such that d < 0. As a consequence the terms of trade, q, d , only depend on the buyer’s money holdings. The third step characterizes the DM value functions. The value function for a buyer holding m units of money at the beginning of the period satisﬁes V b (m) = σ {u q(m) + W b m − d(m) } + (1 − σ )W b (m),

(3.19)

56

Chapter 3

According to (3.19) a buyer in the DM is randomly matched with a seller with probability σ . In a match he consumes q units of the search good and delivers d units of money to the seller, where q and d are the solution to (3.16)–(3.18). Using the linearity of W b (m) one more time, and the buyer’s bargaining problem, (3.16)–(3.18), the value function of the buyer in the DM can be simpliﬁed to V b (m) = σ max u ◦ c−1 (φd) − φd + φm + W b (0). (3.20) d∈{0,...,m}

According to (3.20) the buyer is matched with probability σ , in which case he chooses a transfer of money balances that maximizes his surplus, which is equal to the entire surplus of the match. His continuation value is linear in his real balances. The beginning-of-period value function of a seller who holds ms units of money solves, s V (ms ) = σ −c q(mb ) + W s ms + d(mb ) dF(mb ) + (1 − σ )W s (ms ),

(3.21)

where F(mb ) is the distribution of money holdings across buyers. According to (3.21) the seller is in a trade match with probability σ , and receives an offer (q, d) from the buyer, where the offer depends only on the money holdings of the buyer. Assuming the seller’s utility from accepting the offer, −c(q) + W s (ms + d), is at least as large as the utility from rejecting the trade, W s (ms ), then the offer is accepted. In this case the seller suffers the disutility c(q) in order to produce q units of the search good and receives d units of money in exchange. Using the linearity of W s (ms ), i.e., W s (ms ) = φms + W s (0), (3.21) can be re-expressed as V s (ms ) = σ −c q(mb ) + φd(mb ) dF(mb ) + φms + W s (0). (3.22) With probability σ , the seller is matched and enjoys a surplus that only depends on the buyer’s money holdings. Consequently the seller’s value function in the DM is linear in his own money balances. From the seller’s participation constraint, (3.17), −c q(mb ) + φd(mb ) = 0. Hence the Bellman equation (3.22) becomes V s (ms ) = φms + W s (0).

(3.23)

The Role of Money

57

In the ﬁnal step we characterize the buyers’ and sellers’ choices of money holdings in the CM. Substituting φm +W s (0) for V s (m ) in (3.12), we can rewrite the seller’s choice of money balances problem as max −(1 − β)φm + βW s (0) .

m ∈N0

(3.24)

Clearly, the seller’s optimal choice of money holdings is m = 0. Sellers do not want to carry money into the DM since their money holdings do not affect their terms of trade in the DM and since, due to discounting, it is costly to carry money from one period to the next. Because the seller receives no surplus in a trade match, his beginning-of-period value function, (3.23), simpliﬁes to V s (0) = W s (0) = 0. Consider next the buyer’s choice of money holdings. Substituting the expression for V b (m) given by (3.20) into the buyer’s CM value function, (3.11), his money holdings problem can be re-expressed more compactly as

max −rφm + σ max u ◦ c−1 (φd) − φd , (3.25)

m∈N0

d∈{0,...,m}

where r = (1 − β)/β . Equation (3.25) has a simple interpretation. The buyer faces a trade-off when determining his money holdings. There is a cost associated with holding real balances, which is equal to the agents’ rate of time preference, r, per unit of real balances. But there is a beneﬁt associated with holding real balances, which is equal to the expected surplus that can be obtained in the DM, σ u(q) − φd . Since r > 0, it is easy to check from (3.25) that buyers will hold no more money than they expect to spend in the DM if they ﬁnd themselves in a trade match. This implies that d = m and, therefore, c(q) = φm. As a result the buyer’s portfolio problem, (3.25), can be further simpliﬁed to max −rφm + σ u ◦ c−1 (φm) − φm . (3.26)

m∈N0

Since u ◦ c−1 (·) is strictly concave in m, the buyer’s maximization problem (3.26) has a unique solution if money was perfectly divisible. This solution is denoted as m∗ ∈ R+ in ﬁgure 3.4. However, because money is indivisible, this solution may not be feasible. Let [m∗ ] denote the integer part of m∗ . Consequently (3.26) has at most two solutions, which are [m∗ ] and [m∗ ] + 1.

58

Chapter 3

−rϕm + σ [u ° c −1 (ϕm) − ϕm]

m*

m

m

Figure 3.4 Buyer’s net payoff from holding money

At the beginning of time all buyers receive exactly one unit of money, M = 1. The market-clearing condition in the CM requires that buyers prefer holding one unit of money instead of two or zero. These conditions can be written as follows: −rφ + σ u ◦ c−1 (φ) − φ ≥ −r2φ + σ u ◦ c−1 (2φ) − 2φ , (3.27) −rφ + σ u ◦ c−1 (φ) − φ ≥ 0. (3.28) Condition (3.27) is the requirement that a buyer prefers holding one unit of money to two, and (3.28) is the condition that a buyer prefers holding one unit of money to none. A stationary equilibrium is any φ ≥ 0 that satisﬁes (3.27) and (3.28). We can now determine the conditions under which q = q∗ is part of an equilibrium. Since buyers have all of the bargaining power, as seen from (3.17), we have, for d = m = 1, φ = c(q∗ ). Condition (3.27) can be rewritten as −r2φ + σ u ◦ c−1 (2φ) − 2φ ≤ −rφ + σ u(q∗ ) − c(q∗ ) .

The Role of Money

59

Since u ◦ c−1 (2φ) − 2φ < u(q∗ ) − c(q∗ ), it is immediate that (3.27) is satisﬁed. Intuitively, by accumulating one unit of money, the buyer maximizes the match surplus. A second unit of money is clearly not useful since it cannot increase the match surplus, but it is costly to hold. Condition (3.28) can be rewritten as c(q∗ ) ≤

σ u(q∗ ). r+σ

(3.29)

The inequality (3.29) is similar to the condition described in (3.8). Hence, if the efﬁcient level of consumption and production in the DM is incentive feasible; i.e., it is implementable under the trading mechanism described in section 3.1, then it can be implemented by the trading protocols that gives all of the bargaining power to the buyer in the DM and provides a competitive market at night. By giving all of the bargaining power to the buyer, this trading protocol maximizes the buyer’s incentive to hold money. As before, an efﬁcient allocation can be achieved provided that agents are sufﬁciently patient and search frictions are sufﬁciently small. If the efﬁcient ∗level of consumption ∗ and production are not incentive feasible, i.e., c q > [σ/ (r + σ )] u q , then the best incentive-feasible allocation can still be obtained by the trading protocol that gives all of the bargaining power to the buyer. To see this, suppose that q∗ is not incentive feasible; then the highest incentive feasible q < q∗ is such that the right inequality of (3.29) holds at equality, i.e., c q =

σ u(q). r+σ

In that case the buyer is just indifferent between holding one unit of ¯ = 1. So is clear that the buyer has no money or zero. In ﬁgure 3.4, m incentive to accumulate a second unit of money. 3.3 Further Readings The record-keeping role of money is emphasized by Ostroy (1973), Ostroy and Starr (1974, 1990), and Townsend (1987, 1989), among others. Kocherlakota (1998a, b) uses a mechanism design approach to establish that the technological role of money is to act as a societal memory device that gives agents access to certain aspects of the histories of their trading partners. As a corollary, imperfect knowledge of individual histories is

60

Chapter 3

necessary for money to play an essential role in the economy (Wallace 2000). For further discussions on the essential role of money as memory, see Araujo (2004), Aliprantis, Camera, and Puzzello (2007), and Araujo and Camargo (2009). Kiyotaki and Wright (1989, 1991, 1993) introduce into a searchtheoretic environment a double-coincidence-of-wants problem to explain the emergence of a medium of exchange, and the essentiality of ﬁat money. Kiyotaki and Wright (1993), Shi (1997b), and Camera, Reed, and Waller (2003) endogenize the extent of specialization in decentralized markets. In all these models individuals are matched randomly. Corbae, Temzelides, and Wright (2002, 2003) endogenize the matching process so that who meets who is part of the equilibrium. Engineer and Shi (1998, 2001) and Berentsen and Rocheteau (2003) introduce asymmetries in matches and show that money can be useful even when there is double coincidence of wants in all matches.

4

Money in Equilibrium

In the previous chapter we uncovered the technological role of ﬁat money by describing the set of allocations that are incentive feasible with indivisible money. The objective of this chapter is to study money in equilibrium under trading protocols that have either explicit axiomatic or strategic foundations, and to explore the normative and positive implications of such protocols. In contrast to chapter 3, we do not restrict money to be indivisible, nor do we restrict equilibria to be stationary. The model presented in this chapter is the core framework to study issues related to money, payments, and liquidity for the rest of the book. For all of the trading protocols we study, the model has similar positive implications. For example, the value of money depends on the fundamentals of the economy, such as preferences, technologies, and search frictions. In terms of normative implications, we isolate a key inefﬁciency of monetary exchange that is common to all trading protocols: quantities traded in the DM tend to be too low. In terms of policy, money is neutral, in the sense that a one-time change in the money supply does not affect the allocations or welfare for all the trading protocols we consider. Even though we identify a number of features that are common to the various trading protocols, equilibrium allocations, welfare, and the value of ﬁat money are not invariant to the protocol. For example, quantities traded and social welfare tend to be higher under a competitive protocol in the DM compared to bargaining when the buyer does not have all of the bargaining power. We characterize both stationary and nonstationary equilibria. We show that ﬁat-money economies can have a rich set of nonstationary equilibria. Some of these equilibria are characterized by inﬂation, even

62

Chapter 4

though there is a constant money supply. Such equilibria re-enforce the notion that the value of ﬁat money is sustained by self-fulﬁlling beliefs. Other equilibria are characterized by the value of money ﬂuctuating over time and generating output cycles, even though there are no fundamental changes in the environment. 4.1 A Model of Divisible Money The environment is similar to chapter 3.2. Amajor departure, however, is that we now assume that ﬁat money is perfectly divisible. We continue to assume that the aggregate stock of ﬁat money is constant over time, and equal to M. In this section the terms of trade in the DM are determined by a take-it-or-leave-it offer by a buyer in a match. We consider alternative trading protocols in subsequent sections. The general good and money are traded at night in a CM. In contrast to the previous sections, we do not restrict our attention to stationary equilibria; we will allow the price of money, φt , and the quantity of the DM good produced and consumed, qt , to vary over time. The timing of events in a typical period is as follows: At the beginning of the DM, a measure σ of buyers and sellers are randomly matched, where the buyer has m ∈ R+ units of money and the seller has ms ∈ R+ . Unless otherwise speciﬁed, we assume that the measures of buyers and sellers in the economy are bothequal to 1. In each match the buyer makes a take-it-or-leave-it offer q, d to the seller, where q represents the amount of the search good to be produced by the seller and d ∈ R+ the amount of money that he receives. At the end of the day, all matches are broken up. At night, all buyers and sellers participate in the CM, where agents can exchange money for general goods at price φt . As in chapter 3.2, the model is solved in four steps: 1. Characterize some key properties of the value functions in the CM. 2. Determine the terms of trade in a bilateral match in the DM. 3. Characterize the Bellman equations for the DM. 4. Determine the buyer’s and seller’s choice of money holdings in the CM. The value function for a buyer holding m ∈ R+ units of money, evaluated at the beginning of the CM—or simply the buyer’s CM value function—satisﬁes

Money in Equilibrium

Wtb (m) =

max

m ∈R+ ,x,y

63

b x − y + βVt+1 (m )

subject to x + φt m = y + φt m.

(4.1) (4.2)

From (4.2) the buyer ﬁnances his end-of-period money balances, m , and general good consumption, x, with production of the general good, y, and money balances brought into the CM, m. Notice that the value functions in (4.1) and prices in (4.2) are indexed by time since we now allow the value of money and allocations to vary over time. Substituting x − y from (4.2) into the maximand of (4.1), we get an equation similar to (3.11): b Wtb (m) = φt m + max m + βV (m ) . (4.3) −φ t t+1 m ≥0

As in chapter 3.2 the value function of the buyer at the beginning of the CM is linear in real balances, m. This linearity allows us to obtain a degenerate distribution of money holdings at the beginning of each period, despite the idiosyncratic trading shocks in the DM, since the choice of m is independent of m in (4.3). The value function of the buyer in the CM is represented in ﬁgure 4.1. By a similar line of reasoning, the seller’s CM value function is s Wts (m) = φt m + max −φt m + βVt+1 (m ) . m ≥0

(4.4)

The seller’s value function, like the buyer’s, is linear in real balances. The linearity of the buyer’s and seller’s value functions will prove to be convenient when solving the bargaining problem. The terms of trade in the DM of period t are determined in a bilateral match between a buyer holding m units of money and a seller holding ms units. The buyer chooses an offer, (q, d), that maximizes his expected utility subject to satisfying the seller’s participation constraint. The buyer’s offer solves max u(q) + Wtb (m − d) (4.5) q,d

subject to and

− c(q) + Wts (ms + d) ≥ Wts (ms )

− ms ≤ d ≤ m.

(4.6) (4.7)

64

Chapter 4

Wtb(m)

Wtb(0)

ϕt

m Figure 4.1 Buyer’s value function

The seller’s participation constraint is (4.6), and (4.7) is a feasibility constraint that says the buyer cannot offer to transfer more units of money than he holds, or he cannot ask for more units of money than the seller holds. Since the value functions Wtb and Wts are linear, (4.5)–(4.7) can be simpliﬁed to max u(q) − φt d subject to − c(q) + φt d ≥ 0. (4.8) q,d≤m

From (4.8) the terms of trade, q, d , do not depend on the seller’s money holdings. As well, the seller’s participation constraint must hold at equality. If this were not the case, then the buyer could increase his surplus by slightly reducing the amount that he offers to pay the seller so that the seller still ﬁnds the offer acceptable. Therefore the solution to (4.8) is

∗

q ≥ q= if φt m c(q∗ ), (4.9) c−1 (φt m)

β, then money is costly to hold. Buyers do not carry more money balances than they expect to spend in the DM, i.e., d = m. The ﬁrst-order (necessary and sufﬁcient) condition of the buyer’s problem (4.13) is given by u ◦ c−1 (φt+1 m) 1 φt /φt+1 − β . = 1+ σ β c ◦ c−1 (φt+1 m)

(4.14)

By a similar line of reasoning, the seller’s DM value function, which is evaluated at the beginning of the period, is Vts (m) = σ −c q + φt d + Wts (m) = φt m + Wts (0) , where we have used the fact that the seller does not receive any surplus in the DM, i.e., c q = φt d. Hence we can rewrite the seller’s choice of money balances problem in the CM, described in (4.4), as

φt /φt+1 max − (4.15) − 1 φt+1 m . m≥0 β From (4.15), if (φt /φt+1 )/β = 1, then the seller is indifferent between holding money or not, and m ≥ 0. If, however, (φt /φt+1 )/β > 1, then m = 0 since the seller’s money holdings are costly to carry from one period to the next but do not affect the terms of trade in the DM. As above, if (φt /φt+1 )/β < 1, then a solution does not exist. The aggregate money demand correspondence, Md (φt ), is the sum of the individual money demands across buyers and sellers. From the cases above, the aggregate demand correspondence is ∗

c(q ) d M (φt ) = , +∞ if φt = βφt+1 φt+1 = {m}

where m solves (4.14) if φt > βφt+1 .

The aggregate money demand correspondence is represented in ﬁgure 4.3. It is equal to an interval when φt = βφt+1 , and is single-valued otherwise. Moreover it is decreasing in φt . Market clearing requires that

68

Chapter 4

M d (ϕt)

c (q*) ϕt+ 1

M

ϕt

βϕt+1 Figure 4.3 Aggregate money demand

M ∈ Md (φt ). If M ≥ c(q∗ )/φt+1 , then φt = βφt+1 . Otherwise, φt solves (4.14) with m = M. Consequently the sequence of the value of money, {φt }∞ t=0 , is the solution to the difference equation φt = βφt+1

+

u ◦ c−1 (φt+1 M) 1 + σ −1 −1 c ◦ c (φt+1 M)

,

(4.16)

where [x]+ ≡ max(x, 0). According to (4.16) the price of money in period t is equal to its discounted price in period t + 1 plus a liquidity fac+ tor, σβφt+1 u ◦ c−1 (φt+1 M) /c ◦ c−1 (φt+1 M) − 1 , that captures the marginal beneﬁt of holding real balances in the DM. If money is costly to hold, then buyers don’t bring enough real balances in the DM to purchase q∗ if they are matched, φt+1 M < c(q∗ ). As a consequence the liquidity factor is positive since a buyer would value having an additional unit of money to spend in the DM. If money is costless to hold, then in the CM of period t buyers accumulate sufﬁcient balances to purchase q∗ in the DM of period t + 1 if they are matched, which

Money in Equilibrium

69

implies that φt+1 M ≥ c(q∗ ). Here the liquidity factor is zero because + −1 u ◦ c (φt+1 M) /c ◦ c−1 (φt+1 M) − 1 = 0 and a buyer would not value having an additional unit of money to spend in the DM. An equilibrium of an economy with divisible money is a sequence {φt }∞ t=0 solving the ﬁrst-order difference equation (4.16). Note that we do not impose an initial condition because the dynamic equation for the value of money, (4.16), is forward looking. The value of money is not determined by what happened in the past; it depends entirely on expectations about its future value. Feasibility, however, does require that the value of money is bounded from above. To see this, note that the expected lifetime discounted utility for a buyer at the beginning of a period is Vtb (m) = σ u q(φt m) − c q (φt m) + φt m + Wtb (0), and for a seller is Vts (m) = φt m. Hence the sum of the lifetime utilities of buyers and sellers is such that Vtb (m) + Vts (m) ≥ φt M. Given that the production and consumption of CM goods generate no utility for society, and the surplus of a match is bounded above by u(q∗ )− c(q∗ ), we also have that σ u(q∗ ) − c(q∗ ) b s Vt (m) + Vt (m) ≤ . 1−β Consequently feasibility requires that σ u(q∗ ) − c(q∗ ) φt M ≤ . 1−β Aggregate real balances, which are a promise to utility in the future, cannot be greater than the discounted sum of the maximum match surpluses in the DM. 4.1.1 Steady-State Equilibria We ﬁrst examine stationary equilibria, where φt = φt+1 = φ ss . Since ﬁat money has no intrinsic value, there always exists an equilibrium

70

Chapter 4

where money has no exchange value, where φt = φt+1 = 0. Now consider stationary equilibria where the production and consumption of ss ss the search −1 ssgood are strictly positive, qt = qt+1 = q > 0, and q = ∗ min c (φ M) , q . Equation (4.16) can be simpliﬁed to u (qss ) r = 1+ , c (qss ) σ

(4.17)

where r = (1−β)/β. The left side of (4.17), which is decreasing in qss , goes to inﬁnity as qss approaches zero, i.e., u (0)/c (0) = ∞, and is equal to one if qss = q∗ . See ﬁgure 4.4. Consequently there is a unique qss satisfying (4.17). Since r/σ > 0, qss < q∗ , and the unique φ ss is pinned down by c qss φ = . M ss

(4.18)

Hence output will be inefﬁciently low when r > 0. Moreover output increases as trading frictions decrease, ∂qss /∂σ > 0, and decreases as money becomes more costly to hold, ∂qss /∂r < 0. One can interpret the term r/σ in (4.17) as a measure of the cost of holding real balances: it is

u'(0) c'(0) u'(q) c'(q)

1+

r σ 1

q ss

q*

Figure 4.4 Determination of the steady-state equilibrium

q

Money in Equilibrium

71

the product of the rate at which agents depreciate future utility, r, and the average number of periods it takes to get matched, 1/σ . As this cost increases, buyers reduce their real balances, and DM output falls. As the rate of time preference approaches zero, qss tends to q∗ . Finally, notice that a one-time change in the stock of money, M, does not affect the real allocation: Money is neutral. From (4.17) output in the DM, qss , is unaffected by a change in the aggregate stock of money, since a change in the aggregate stock affects neither the frequency of trade, σ , nor the rate of time preference, r. Hence aggregate real balances, φM, are constant—equal to c qss —and the change in the price level, 1/φ, is proportional to the change in M. 4.1.2 Nonstationary Equilibria There exist other equilibria that are not stationary. The exact nature of these equilibria, however, depends on functional forms and parameter values. Diagrammatically speaking, the relationship between φt+1 and φt as deﬁned by (4.16)—the so-called phase line—is continuous, goes through both the originand the steady-state point (φ ss , φ ss ), where φ ss > 0. For all φt+1 M > c q∗ , the phase line is linear, φt = βφt+1 . Consequently, in the (φt , φt+1 ) space, the phase line intersects the 45o line from below. See the appendix for details. Consider the following functional forms: c(q) = q and u(q) = q1−a /(1 − a) with a < 1. For this speciﬁcation, q∗ = 1 and (4.16) becomes if φt+1 M < 1, β (1 − σ ) φt+1 + σ (φt+1 )1−a (M)−a (4.19) φt = if φt+1 M ≥ 1. βφt+1 As shown in ﬁgure 4.5, the phase line, given by (4.19), is monotonically increasing and convex in the (φt , φt+1 ) space when φt+1 M < 1 and linear with slope β −1 = 1 + r otherwise. There are a continuum of equilibria converging to the nonmonetary equilibrium: if φ0 < φ ss , then φt approaches 0 as t goes to inﬁnity. For all these equilibria, φt is decreasing and the price level, 1/φt , is increasing over time. Hence it is possible to have a positive inﬂation, in equilibrium, even though the money supply is constant. In this situation beliefs about a depreciating value of currency can be selffulﬁlling. If φ0 = φ ss , then the equilibrium is stationary. If φ0 > φ ss , then there does not exist an equilibrium, since {φt }∞ t=0 is unbounded and thereby violates feasibility.

72

Chapter 4

ϕt+1 ϕt +1 = ϕt

>

ϕ0 ϕ2

ϕ1

ϕt

Figure 4.6 Phase line: σ = 1, c(q) = q, u(q) = [(q + b)1−a − b1−a ]/(1 − a), b > 0 and a(β 1/a − b) > 2β 1/a .

In the bottom right panel, we plot both the phase line, φt = (φt+1 ), and its mirror image with respect to the 45o line, φt+1 = (φt ). We enlarged the phase diagram in the neighborhood of the steady state. Two-period cycles are obtained at the intersection of the phase line and its mirror image. If the intersection is not on the 45o line, we obtain a proper cycle. In our example, there is a two-period cycle where the value of money alternates between a low value, φL ≈ 0.85, and a high value, φH ≈ 0.95. Hence output in a bilateral match alternates between qL ≈ 0.85 and qH = q∗ = 0.9. Note that in the high state, buyers’ holdings are larger than the level required to buy the efﬁcient quantity. The buyer is willing to hold this additional currency because the rate of return on currency is exactly equal to r. 4.1.3 Sunspot Equilibria To conclude this section, we introduce the notion of extrinsic uncertainty—uncertainty that does not affect fundamentals such as technologies and preferences. The sample space of the extrinsic random variable, called a sunspot, is E = { , h}. The sunspot e ∈ E is observed by

Money in Equilibrium

75

φ t+1 1.0

φ t+1 1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.2

0.4

0.6

0.8

1.0

φt

0.4

0.6

0.8

1.0

φt

φ t+1 1.10 1.05 1.00

φ t+1

0.95

1.2 1.0 0.8 0.6 0.4 0.2

0.90 0.85

0.5

1.0

1.5

2.0

2.5

3.0

φt

0.85 0.90 0.95 1.00 1.05 1.10

φt

Figure 4.7 Phase diagrams. (top left) a = 0.5; (top right) a = 1.5; (bottom left) a = 2.2; (bottom right) a=4

all agents at the beginning of the CM, and follows a two-state Markov chain, with λee = Pr[et+1 = e |et = e ]. That is, there is a (possibly new) sunspot realization at the beginning of each CM, and the probability that the new realization is e , given that the previous realization was e, is λee . We now characterize stationary equilibria where there is extrinsic uncertainty, where by stationarity we mean that the value of money, φe , depends on the realization of the sunspot state but does not depend on time. Following the same reasoning as above, the buyer’s choice of money holdings in state s is given by max −φe m + βσ u q(φ¯ e m) − c q(φ¯ e m) + β φ¯ e m , (4.23) m≥0

76

Chapter 4

where φ¯ e = e ∈E λee φe is the expected price of money in the next CM conditional on the current state e. We have that q(φ¯ e m) = q∗ if φ¯ e m ≥ c(q∗ ) and q(φ¯ e m) = c−1 (φ¯ e m) otherwise. According to (4.23), in the sunspot state e, the buyer purchases m units of money at the price φe . In thesubsequent DM, buyers purchase q(φ¯ e m) units of goods and transfer c q(φ¯ e m) real balances to sellers. In the DM, agents value money according to its future expected price in the CM. The ﬁrst-order condition of the buyer’s problem, (4.23), together with the market-clearing condition m = M, is u q(φ¯ e M) −1 . φe = β φ¯ e 1 + σ (4.24) c q(φ¯ e M) As above, the value of money is equal to its expected discounted value in the next CM plus a liquidity premium factor. The liquidity premium factor is strictly positive if an additional unit of money relaxes the budget constraint of the buyer in a bilateral match in the DM. A stationary sunspot equilibrium is a pair (φ , φh ) that satisﬁes (4.24) for e = , h. There is an equilibrium where agents simply ignore sunspots, φ = φh = φ ss , since sunspot states do not affect fundamentals in any way. There can also be proper sunspot equilibria where the economy jumps from one state to another state, where states are associated with different values for money and different quantities traded in the DM. In general, one can construct sunspot equilibria from the multiplicity of steady-state equilibria. However, this won’t work in our case where one equilibrium is the nonmonetary one because the value of money in the low state is constrained to be nonnegative. This would work in the case where there is a cost to carry money, as in chapter 5.2, and if we use the two monetary equilibria to construct a sunspot equilibrium. One can construct a sunspot equilibrium from a two-period-cycle equilibrium when it exists, as mentioned above. Suppose λ h = λh = 1. Then (φ , φh ) solution to (4.24) corresponds to the values of money in the two-period cycle. By continuity, for λ h and λh close to one, there exists other proper sunspot equilibria where the change in the state is not deterministic. 4.2 Alternative Bargaining Solutions We have considered a trading protocol in the DM where a buyer in a match makes a take-it-or-leave-it offer to the seller. This protocol is interesting because the agent who holds money during the day is

Money in Equilibrium

77

able to extract all the gains from trade. This arrangement, however, is quite special, and one should examine the positive and normative implications associated with alternative trading protocols. We now propose a number of different trading protocols for the DM, which include alternative bargaining solutions, a Walrasian protocol, where agents are price-takers, and a price-posting protocol, where sellers compete to attract buyers. We start by deﬁning the bargaining set in a bilateral match, and then review the solutions to alternative bargaining protocols. 4.2.1 Bargaining Set Consider a match between a buyer holding m units of money and a seller holding none. (It is straightforward to generalize the argument to the case where sellers hold positive money balances.) An agreement is a pair (q, d), where the buyer receives q ≥ 0 units of the search good produced by the seller in exchange for d ∈ [0, m] units of money. If an agreement is reached, then the buyer’s utility is ub = u(q) + W b (m − d), and the seller’s is us = −c(q) + W s (d). If there is no agreement, then the buyer’s utility is ub0 = W b (m), and the seller’s is us0 = W s (0). Because W b and W s are linear in money, we have ub = u(q) − φd + ub0 and us = φd − c(q) + us0 . Hence the buyer’s surplus from an agreement is ub − ub0 = u(q) − φd, the seller’s surplus is us − us0 = φd − c(q), and the total surplus, the sum of the buyer’s and seller’s surpluses, is u q − c q . To illustrate the role that money plays in exchange, suppose that the buyer cannot spend more than τ ≤ m units of money. Let S(τ ) represent the set of feasible utility levels for the buyer and seller, when the buyer can spend at most τ units of money, i.e., S(τ ) = (u(q) − φd + ub0 , φd − c(q) + us0 ) : d ∈ [0, τ ] and q ≥ 0 . The equation frontier of S is derived from the program for the Pareto b b u = maxq,d u(q) − φd + u0 subject to −c(q) + φd ≥ us − us0 and d ≤ τ , for us ≥ us0 . If φτ ≥ c(q∗ ) + us − us0 , then the solution to the Pareto problem is q = q∗ , φd = c(q∗ ) + us − us0 , and if φτ < c(q∗ ) + us − us0 , then the solution is q = c−1 φτ − (us − us0 ) , d = τ.

78

Chapter 4

The equation for the Pareto frontier is u

s

− us0

=

u(q∗ ) − c(q∗ ) − (ub − ub0 ) φτ − c u−1 (ub − ub0 + φτ )

if φτ ≥ c(q∗ ) + us − us0 , otherwise. (4.25)

If τ units of money are sufﬁcient to purchase q∗ and to provide the seller with a surplus of us − us0 , i.e., if φτ ≥ c(q∗ )+ us − us0 , then the buyer and seller will split the total surplus, u(q∗ ) − c(q∗ ), according to ub − ub0 and us − us0 , respectively. If, however, τ units of money are insufﬁcient to purchase q∗ and to provide a surplus of us − us0 to the seller, then the buyer will spend all τ units of his money and q < q∗ . It can be checked from (4.25) that d2 us /(dub )2 = 0 if φτ − c(q∗ ) − us + us0 ≥ 0 and d2 us /(dub )2 < 0 otherwise. That is, when q = q∗ , the Pareto frontier is linear, and when q < q∗ , it is strictly concave. In ﬁgure 4.8 we illustrate the bargaining set S(τ ) for τ3 > τ2 > τ1 . The maximum possible surplus of a match is denoted by ∗ , where ∗ = u(q∗ )−c(q∗ ). Note that the match surplus is always less ∗ for bargaining us

u0s + Δ*

S (τ3 )

S (τ2 )

S (τ1 )

(u0b, u0s ) Figure 4.8 Bargaining set

u0b + Δ*

ub

Money in Equilibrium

79

set S(τ1 ) in ﬁgure 4.8, i.e., by construction, all of the allocations in S(τ1 ) are characterized by q < q∗ since φτ < c q∗ . For the bargaining sets S(τ2 ) and S(τ3 ), the match surplus is equal to ∗ along their linear portions of the respective sets. However, trades are characterized by q < q∗ where the frontiers are strictly concave. Note that the bargaining set is larger when the buyer is able to use more of his money balances, i.e., S(τ1 ) ⊂ S(τ2 ) ⊂ S(τ3 ). This expansion of the bargaining set illustrates the idea that ﬁat money allows traders to achieve utility and output levels that otherwise would not be attainable. 4.2.2 The Nash Solution A solution to the bargaining problem can be interpreted as a function that assigns a pair of utility levels to every bargaining game. The generalized Nash solution to the bargaining problem is based on three axioms: Pareto efﬁciency, invariance to rescaling of agents’ payoffs, and independence to irrelevant alternatives. It can be shown that these three axioms imply that the solution maximizes the weighted geometric average of the buyer’s and seller’s surpluses from trade, where the weights are given by the agents’ bargaining powers. Figure 4.9 provides a graphical illustration of the Nash solution. The gray shaded area, denoted by S, represents the set of feasible utility levels for the buyer and seller—the bargaining set—and the bargaining solution chooses one point from this set. Since the Nash solution here is Pareto efﬁcient, the solution will lie on the Pareto frontier. The downward-sloping, convex curve represents the combinations of the weighted geometric average of the agents’ surpluses that generates the same value. The Nash solution is given by the tangency of this curve with the bargaining set. Since, in ﬁgure 4.9, the tangency occurs on the strictly concave part of the Pareto frontier, the Nash solution here is characterized, in part, by q < q∗ . In the context of our monetary environment, the generalized Nash solution, q(m), d(m) , can be expressed as

θ 1−θ , q(m), d(m) = arg max u(q) − φd −c(q) + φd q,d≤m

(4.26)

where θ ∈ [0, 1] represents the buyer’s bargaining power, 1−θ represents the seller’s, and m is the buyer’s money holdings. If the constraint d ≤ m does not bind, then the solution to (4.26) is q = q∗ ,

80

Chapter 4

us

u0s + Δ*

S

u0b + Δ*

(u0b, u0s )

ub

Figure 4.9 Nash solution

d = m∗ ≡

(1 − θ)u(q∗ ) + θ c(q∗ ) . φ

If, however, m < m∗ , then the constraint d ≤ m binds, meaning d = m. In this case the generalized Nash solution for the level of DM output can be expressed as arg max θ ln u(q) − φm + (1 − θ ) ln −c(q) + φm . (4.27) q

The solution to (4.27) is φm ≡ zθ (q) =

(1 − θ )c (q)u(q) + θu (q)c(q) . θ u (q) + (1 − θ )c (q)

(4.28)

According to (4.28), in order to buy q < q∗ units of output, the buyer spends all of his money, and DM output, q < q∗ , is determined by a weighted mean of the buyer’s utility of consuming q and seller’s disutility of producing q. It should be clear from (4.26) or (4.27) that the

Money in Equilibrium

81

outcome, [q(m), d(m)], is independent of the seller’s money balances due to the linearity of the seller’s value function. Since money is costly to hold and the seller’s money holdings do not inﬂuence the terms of trade, the seller will not accumulate money in the CM to bring into the DM. For the remainder of this chapter, we will focus only on steady-state equilibria. The buyer’s DM value function is, therefore, given by V b (m) = σ u[q(m)] + W b m − d(m) + (1 − σ )W b (m), (4.29) Since the buyer’s CM value function is linear in money, i.e., W b (m) = φm + W b (0), the choice of his money holdings is given by the solution to max −rφm + σ u q(m) − φd(m) . (4.30)

m∈R+

Provided that r > 0, the buyer will never accumulate more balances in the CM than he would spend in the DM, which implies that m ≤ m∗ . From (4.28) and (4.30), the buyer’s choice of consumption in the DM at a steady-state equilibrium is max −rzθ (q) + σ u(q) − zθ (q) . (4.31) ∗ q∈[0,q ] Note that problem (4.31) is a generalization of problem (4.13) when φt = φt+1 = φ, where in the latter problem the buyer has all of the bargaining power. The the expected surplus from a buyer maximizes trade in the DM, σ u(q) − zθ (q) , minus the cost of holding real balances, rzθ (q). While the objective function in (4.31) is not necessarily concave—because zθ q is not convex—it is continuous and the choice of q is in the compact set [0, q∗ ]. Therefore a solution exists. Assuming an interior solution, the ﬁrst-order condition to (4.31) is u (q) r = 1+ . zθ (q) σ

(4.32)

If θ = 1, then zθ (q) = c(q) and the solution to (4.32) coincides with (4.16). In particular, as r tends to zero, the quantities traded in the DM approach q∗ . In contrast, however, if θ < 1, then q < q∗ even in the limit when r → 0. To see this, we can express the relationship between real balances and output in (4.28) as zθ (q) = 1 − (q) u(q) + (q)c(q), (4.33)

82

Chapter 4

where (q) =

θ u (q) . θ u (q) + (1 − θ )c (q)

It is easy to check (q∗ ) = θ , and (q) < 0 for all q. Hence zθ (q∗ ) = that ∗ ∗ ∗ u (q ) − (q ) u(q ) − c(q∗ ) > u (q∗ ). Therefore, as q approaches q∗ , the buyer’s surplus, u(q) − zθ (q), falls. There are two effects with associated an increase in q < q∗ : ﬁrst, total match surplus, u q − c q increases, and second, the buyer’s share of the surplus decreases. For q close to q∗ , the second effect dominates the ﬁrst. Consequently, even if it is not costly to hold real balances, r ≈ 0, the buyer will not bring sufﬁcient real balances into the DM to be able to purchase q∗ . So, in addition to the monetary inefﬁciency created by discounting, there is an inefﬁciency associated with Nash bargaining. This result is a consequence of the fact that the buyer’s surplus is not always increasing in his real balances: the generalized Nash bargaining solution is said to be nonmonotonic. that if the buyer has no bargaining power, θ = 0, then zθ q = Note u q , and the solution to problem (4.31) is q = 0. Since the buyer receives no surplus from purchasing the DM good from the seller, and it is costly to accumulate real balances, the buyer optimally chooses not to trade in the DM. Hence a necessary condition for trade to take place is θ > 0; buyers must have some bargaining power. 4.2.3 The Proportional Solution In contrast to the generalized Nash solution, the proportional bargaining solution requires that agents’ surpluses increase as the bargaining set expands, which implies that the solution is monotonic. The proportional bargaining solution is also Pareto efﬁcient, i.e., (ub , us ) lies in the Pareto frontier of S and has each player receiving a constant share of the match surplus, i.e., u q − φd = θ u q − c q and −c q + φd = (1 − θ) u q − c q or ub − ub0 =

θ s u − us0 , 1−θ

(4.34)

where, as above, θ ∈ (0, 1] is the buyer’s bargaining power. The outcome of the proportional solution is illustrated in ﬁgure 4.10. In the context of our monetary model, (q, d) solves (q, d) = arg max[u(q) − φd] d≤m

(4.35)

Money in Equilibrium

83

us

u0s + Δ*

u s − u 0s 1 − θ = θ u b − u 0b

S

u0b + Δ*

(u0b, u0s )

ub

Figure 4.10 Proportional bargaining solution

u(q) − φd =

subject to and

θ φd − c(q) 1−θ

d ≤ m.

(4.36) (4.37)

Substituting φd by its expression from (4.36), i.e., φd = (1−θ)u(q)+θ c(q), into (4.35) and (4.37), this problem can be simpliﬁed to q = arg max θ u(q) − c(q) (4.38) q

subject to (1 − θ )u(q) + θc(q) ≤ φm.

(4.39)

If (4.39) binds, then q is simply the solution to φm ≡ zθ (q) = θ c(q) + (1 − θ )u(q).

(4.40)

This expression is similar to the Nash solution, (4.33), except that the buyer’s share in the Nash solution, q , is a function of q, whereas for the proportional solution it is a constant.

84

Chapter 4

The buyer’s problem in the DM is given by (4.31), which, thanks to (4.40), can be rewritten as max −rzθ (q) + σ θ u(q) − c(q) (4.41) ∗ q∈[0,q ] or max [σ θ − r (1 − θ )] u q − (r + σ ) θc(q) . ∗ q∈[0,q ]

(4.42)

The analysis assumes that (4.39) binds; it should be pointed out that as long as r > 0, (4.39) will always bind. A necessary condition for the buyer’s CM problem to admit a positive solution is σ θ − r (1 − θ ) > 0 or θ/(1 − θ ) > r/σ . This condition implies that buyers must have enough bargaining power if money is to be valued. If θ/(1 − θ) > r/σ , then the buyer’s objective in (4.42) is strictly concave. The ﬁrst-order condition to (4.42) is given by (4.32). This condition, with the help of (4.40), can be rewritten as u (q) − c (q) r = . zθ (q) θσ

(4.43)

The left side of (4.43) is the marginal increase of the match surplus generated by an increase of the buyer’s real balances, while the right side of (4.43) is a monetary wedge introduced by discounting, r, search frictions, σ , and the buyer’s bargaining power, θ . An increase in the seller’s bargaining power—which reduces θ —raises the monetary wedge through a holdup problem. The buyer will tend to underinvest in real balances since he incurs the proportional cost r/σ from holding real balances but only receives a fraction θ of the match surplus. It can be checked that q increases with θ . As r tends to zero, the cost of holding real balances vanishes, as does the holdup problem. Consequently match output approaches q∗ as r approaches zero, which is in contrast to the generalized Nash solution. With proportional bargaining, the buyer’s surplus is strictly increasing in his real balances, until the match output q∗ is achieved. Hence, if the cost of holding money balances is zero, then the buyer will accumulate sufﬁcient real balances to purchase the efﬁcient level of the search good, q∗ . 4.3 Walrasian Price Taking We have assumed so far that buyers and sellers meet bilaterally in the DM. We favor this sort of arrangement because it provides an explicit

Money in Equilibrium

85

description of how trades take place and prices are formed. We show in subsequent chapters that the assumption of bilateral meetings is crucial for generating certain optimal policy results, and the coexistence of assets with different rates of return. Nevertheless, the notion of competitive markets is pervasive in economics. We can accommodate such a trading protocol in the DM by assuming that buyers and sellers meet in large groups during the day in a competitive market and that they are anonymous. Since agents are anonymous during the day, they are unable to use credit arrangements. We continue to label the day market as decentralized, DM, and reinterpret the idiosyncratic matching shocks, σ , as preference and productivity shocks. In particular, a fraction σ of buyers want to consume during the day, while a fraction 1 − σ do not, and a fraction σ of sellers are able to produce, while a fraction 1−σ cannot. We denote the price of the day good expressed in terms of the night good as p, i.e., if pˆ is the dollar price for a unit of DM output and φ is amount of CM goods that can be purchased with a dollar, then p ≡ pˆ φ. The problem that an active seller faces in the DM, i.e., a seller who can produce, is to choose the quantity to supply, qs . This problem is given by qs = arg max −c(q) + pq . q

(4.44)

The ﬁrst-order condition to this problem is p = c (qs ).

(4.45)

Sellers produce until their marginal disutility is equal to the real price of the DM good, measured in terms of the night good. The problem that the buyer faces in the CM is how much money to bring into the DM or, equivalently, how much of the DM good to consume. The buyer makes this choice before he learns whether he is active in the DM. The buyer’s problem is qb = arg max −rpq + σ u(q) − pq . q

(4.46)

From (4.46), in order to consume q in the DM, the buyer must accumulate pq real balances—measured in terms of the next period’s CM output—in the CM, where the cost of holding real balances is equal to the rate of time preference, r. The ﬁrst-order condition to (4.46) is r u (qb ) = 1 + p. σ

(4.47)

86

Chapter 4

From (4.47) there is a monetary wedge equal to r/σ between the buyer’s marginal utility of consumption and the price of the good in the DM. This wedge arises because the buyer must accumulate real balances the period before entering the DM. As well, there is a risk that the buyer will not need the real balances if he receives a preference shock that implies he does not want to consume. Since the measures of active buyers and sellers are both equal to σ , the clearing condition for the DM goods market requires that qb = qs = q. From (4.45) and (4.47) we have that u (q) r = 1+ . c (q) σ

(4.48)

This equation is identical to the one obtained under the bargaining protocol, where the buyer has all of the bargaining power, (4.17). In both cases q approaches q∗ as r tends to zero. The value of money is given by the solution to pq = φM, which, by (4.45), implies that c q q = φM or c q q φ= . M If c q is strictly convex, then the value of money is larger than in a bargaining environment where the buyer makes a take-it-or-leave-it offer. Intuitively, when the buyer makes a take-it-or-leave-it offer, DM goods are priced according to average cost, and when pricing is Walrasian, DM goods are priced according to marginal cost. Hence the value of money will be greater under the latter than the former. 4.4 Competitive Price Posting In many markets, sellers post prices for their goods. Buyers observe these prices—or contracts—and then decide where to buy. We formalize this notion of trade by appealing to the concept of competitive search. Competitive search has been developed to provide a foundation for competition in environments where agents meet in pairs, and their participation decisions are associated with thick-market and congestion externalities. By having sellers compete before matches are formed, competitive search allows one to price congestion or waiting times in the market, where the surplus that an agent receives reﬂects his social contribution to the matching process.

Money in Equilibrium

Sellers post (q, d ) for the next DM.

87

Period t − 1

Period t

CM

DM

Buyers choose Buyers enter their money a submarket holdings. with posted (q, d ).

Each buyer finds a seller with probability σ min (1, 1/n)

In each match agents trade according to the posted (q, d ).

Figure 4.11 Competitive search: Timing of events

We assume that the economy is composed of different submarkets in the DM, where a submarket is identiﬁed by its terms of trade, q, d . The terms of trade for the DM good in period t are posted by sellers at the beginning of the previous night, in period t − 1. Sellers can commit to their posted prices. Buyers can observe all of the terms of trade in all of the submarkets. Based on the observed terms of trade, buyers decide which particular submarket they will visit in the subsequent DM, and the amount of real balances to accumulate in that CM. The timing of events is illustrated in ﬁgure 4.11. Submarkets are not frictionless. The search frictions that exist in competitive search environments attempt to capture heterogeneity of goods and capacity constraints. For example, in each submarket, buyers and sellers face the risk of being unmatched. So, even though a buyer can direct his search to a location where he knows the terms of trade, he still has to ﬁnd a match with a seller who produces the type of good he wants. In addition, even if the buyer ﬁnds a desirable seller, sellers may face capacity constraints, such as only being able to produce for one buyer. We can describe the matching process more formally. Suppose that there is a measure of B buyers and S sellers in a submarket that has posted terms of trade q, d . Denote the ratio of buyers per seller as n = B/S. A matching technology speciﬁes the measure of matches in a submarket as a function of the matching friction, σ , and the measures of buyers and sellers. We assume that the matching technology is given by σ min (B, S); i.e., the measure of matches is a fraction, σ , of the measure of agents on the short side of the market. If σ = 1, then all agents on the short side of the market are matched. The actual buyers and sellers that are matched are chosen at random in their

88

Chapter 4

respective submarkets. Consequently the matching rate of a buyer is σ min (B, S) /B = σ min (1, 1/n), and the matching rate of a seller is σ min (B, S) /S = σ min (n, 1). When a seller posts his terms of trade at the beginning of the night subperiod, he takes as given the utility that buyers expect to receive when optimally choosing the submarket to search for sellers. If U b represents the expected surplus of a buyer in the DM, net of the cost of holding real balances, then in any active submarket,

1 −rφd + σ min 1, (4.49) u(q) − φd = U b . n A seller’s choice of his terms of trade, (q, d), determines the length of the queue, n, in his submarket, where n is given by the solution to (4.49). The length of queue is such that a buyer is indifferentbetween going to a particular submarket associated with terms of trade q, d and going to his best alternative that guarantees him an expected utility equal to U b . The seller’s posting problem can be represented by max σ min (1, n) −c(q) + φd subject to (4.49). (4.50) q,d,n

The seller chooses the terms of trade to post, (q, d), and, via constraint (4.49), the implied queue length, n, so as to maximize his expected utility in the DM. ¯ b represent the upper bound of the buyer’s expected utility in Let U any equilibrium. This upper bound will be attained if the buyer receives the entire match surplus, u(q) − c(q), and if his matching probability is at its maximum value, σ . In this case the buyer will only bring enough real balances to compensate the seller for his production cost, c(q). More formally, the upper bound of the buyer’s expected utility is given by ¯ b = max −rφd + σ u(q) − φd U q

or

subject to

− c q + φd = 0,

¯ = max −rc(q) + σ u(q) − c(q) . U b

q

Qualitatively speaking, the buyer’s expected utility, U b , can fall into one of four ranges. ¯ b , then sellers have no incentives to make markets, or 1. If U b > U post prices, since they cannot offer buyers their market expected utility

Money in Equilibrium

89

¯b without generating a negative payoff for themselves. Clearly, U b > U is inconsistent with an equilibrium. ¯ b , then the buyer’s surplus is at its maximum value. Any 2. If U b = U solution to (4.50) implies that buyers receive the entire surplus of a match, i.e., φd = c(q), and that they are on the short side of the submarket, n ≤ 1. Hence the seller’s payoff is zero. ¯ b , then u(q) − φd > 0. This implies, however, that n > 1 3. If U b ∈ 0, U

cannot be an equilibrium, i.e., a solution to (4.50). If n > 1 were a solution, then the seller could slightly increase d such that, via (4.49), n decreases but still remains greater than or equal to 1. Hence the seller’s expected utility increases, which is a contradiction. Intuitively, if n > 1, then there is congestion on the buyer’s side. Sellers don’t beneﬁt from this congestion since their matching probability is σ , which is independent of n, whereas buyers must be compensated for the congestion by better terms of trade. Clearly, it is optimal for sellers to eliminate the congestion, since doing so results in better terms of trade for themselves without affecting their matching probability. Therefore, in any equilibrium it must be the case that n ≤ 1. Since n ≤ 1, min (1, 1/n) = 1. If we substitute the expression for φd given by (4.49) with min (1, 1/n) = 1 into the objective function, (4.50), the seller’s problem becomes −rc(q) + σ u(q) − c(q) − U b max σ n . (4.51) q,n≤1 r+σ The ﬁrst-order condition with respect to q is u (q) r = 1+ . c (q) σ

(4.52)

¯ b is the same as that for the Hence the quantity traded when U b ∈ 0, U Walrasian price taking protocol and the bargaining protocol where the ¯ b , the ratio buyer has all of the bargaining power. Finally, if U b ∈ 0, U expression in the inner braces of (4.51) will be strictly positive, which implies that n = 1 is the solution. 4. If U b = 0, then a buyer is indifferent between (actively) participating or not in the DM. If a buyer participates, then solution to (4.50) is such that n = 1 and q solves (4.52). The value of the money transfer, d, will adjust so that the left side of (4.49) is zero. There may also be some buyers who choose not to participate and enter an inactive submarket, i.e., a

90

Chapter 4

submarket that implicitly has d = q = 0 and n = ∞, i.e., σ min(1, 1/n) = 0; see constraint (4.49). The equilibrium value of being a buyer, U b , is determined such that the ratio of buyers per seller in the different submarkets is consistent with the measures of buyers and sellers in the economy. Suppose that the market is composed of a unit measure of sellers and a measure N of buyers, where N > 0. Then we can deﬁne the aggregate demand for active buyers by sellers, N d , and the aggregate supply of active buyers, N s , by d N ≡ n(j)dj = N s ≡ N − n0 , (4.53) where n(j) is the measure of buyers per seller in the submarket of seller j and n0 is the measure of buyers who do not participate, i.e., they enter ¯ b , then from the inactive submarket. From the results above, if U b < U b points 3 and 4, n(j)dj = 1. Moreover, if U > 0, then n0 = 0 and N s = N. ¯ b , then from point 2, n(j) ∈ [0, 1] and n(j)dj ∈ [0, 1]. Finally, if If U b = U b U = 0, then from point 4, buyers are indifferent between participating and not participating, which means that n0 ∈ [0, N]. We illustrate the various equilibrium outcomes in ﬁgure 4.12. The step function in ﬁgure 4.12 labeled N d represents the aggregate number of buyers desired by sellers across all submarkets for various levels of the buyer’s surplus. The step function labeled N s represents the aggregate number of buyers who are willing to participate, i.e., the active buyers. Note that step function N s corresponds to the case where N > 1. The “market-clearing price’’ that equalizes aggregate supply of buyers and aggregate demand of buyers is the buyer’s expected utility, U b . If N > 1, then U b = 0since N s intersects N d at U b = 0 in ﬁgure 4.12. In any equilibrium, n j = 1 for all sellers; a measure N − 1 of buyers go to the inactive market, and a unit measure will allocate themselves one-for-one with sellers. The inactive buyers get zero utility and, since they are inactive in the DM, they do not accumulate real balances in the CM. The unit measure of buyers who are active also receive zero utility, and the seller’s posted price is characterized by φd = [σ/(r + σ )]u(q), which follows from constraint (4.49). ¯ b , since the horizontal dashed line, which If N < 1, then U b = U d represents the supply of active buyers when N < 1, intersects N at b b ¯ in ﬁgure 4.12. In any equilibrium, n j ≤ 1 for all sellers j. The U =U seller’s posted contract, q, d , is the one that maximizes the expected surplus of the buyer, subject to the seller receiving zero surplus, i.e., it

Money in Equilibrium

91

N s, N d

Ns

N

1

N 0. Differentiate (4.16) for φt+1 such that φt+1 < c(q∗ )/M and use (4.17) to get ∂φt u (qss )c (qss ) − u (qss )c (qss ) s < 1, = 1 + σβφ M 3 ∂φt+1 c (qss )

94

Chapter 4

where qss = c−1 (φ ss M). In the space (φt , φt+1 ) the phase line representing RHS intersects the 45o line from below. Shi’s (1997) Large Household Model In section 4.1 we described a simple search-theoretic model with divisible money. Even though there are idiosyncratic trading shocks in the DM, the distribution of money holdings at the beginning of each periods is degenerate, which keeps the model tractable. This result arises thanks to a competitive market in the second subperiod and quasi-linear preferences. The former allows agents to readjust their money holdings and latter eliminates wealth effects so that the choice of money holdings of an agent is independent of his trading history in the previous decentralized markets. The ﬁrst search model with divisible money and a degenerate distribution of money holdings was proposed by Shi (1997, 1999, 2001). This model does not assume competitive markets nor quasi-linear preferences. The trick to keep the model tractable, which is borrowed from Lucas (1990), is to assume that households are composed of a large number of members that can pool their money holdings, thereby providing insurance against the idiosyncratic trading shocks in the DMs. We will describe a slightly modiﬁed version of the large household model that is similar to the model used in this book. Each household consists of a unit measure of buyers and a unit measure of sellers. Buyer and sellers carry out different tasks but regard the household’s utility as the common objective. Buyers attempt to exchange money for consumption goods, and sellers attempt to produce goods for money. When carrying out these tasks, household members follow strategies that have been given to them by their households. In each period, the probability that a seller of a given household meets a buyer from another household is σ , and the probability that a buyer meets a seller from another household is σ . At the end of each period, buyers and sellers of the same household pool their money holdings, which eliminates aggregate uncertainty for households. Finally, the utility of the household is deﬁned as the sum of the utilities of its members. We refer to an arbitrary household as household h. Decision variables for this household are denoted by lowercase letters. Uppercase letters denote other households’ variables, which are taken as given by the representative household h. Because we focus on steady-state equilibria,

Money in Equilibrium

95

we omit the time index t. Nevertheless, variables corresponding to the next period are indexed by +1, and those corresponding to the previous period are indexed by −1. The chronology of events within a period is as follows: At the beginning of each period, household h has m units of money per buyer, which it divides evenly among its buyers. The household speciﬁes the trading strategies for its members. Then agents are matched and carry out their exchanges according to the prescribed strategies. Within a period, a buyer cannot transfer any of his money to another member of the same household. After trading concludes, buyers consume the goods they acquired, and sellers bring the money that they received for producing back to the household. At the end of a period, the household has money holdings m+1 that is carried into period t + 1. The quantity of money in the economy is assumed to be constant and equal to M units per buyer. The (indirect) marginal utility of money of household h is φ = βV (m+1 ), where V(m) is the lifetime discounted utility of a household holding m units of money. We assume that the terms of trade in bilateral matches are determined by a take-it-or-leave-it offer by the buyer. When matched, household members cannot observe the marginal value of money of their trading partners, βV (m+1 ). As a consequence households’ strategies depend on the distribution of their potential bargaining partners’ valuations for money. In a symmetric equilibrium this distribution is degenerate: all households have the same marginal value ofmoney, . A buyer’s take-it-or-leave-it offer is a pair q, d , where q is the quantity of goods produced by the seller for d units of money. If the seller accepts the offer, then the acquired money, d, is added to his household’s money balances at the beginning of the next period. Because each seller is atomistic, the amount of money obtained by a seller is valued at the marginal utility of money, . Since theseller’s cost associated with producing q is c(q), the seller accepts offer q, d if d ≥ c(q). Thus any optimal offer—optimal from the buyer’s household perspective— satisﬁes d = c(q).

(4.54)

Because a buyer cannot offer to exchange more money than he has, offer q, d satisﬁes d ≤ m.

(4.55)

96

Chapter 4

In each period household h chooses m+1 , and the terms of trade q, d to solve the following problem: V(m) = max σ u q − c(Q) + βV(m+1 ) (4.56) q,d,m+1

subject to

(4.54), (4.55)

and m+1 − m = σ (D − d) .

(4.57)

The variables taken as given in the above problem are the state variable m and other households’ choices, Q and D. The ﬁrst term in the maximand of (4.56), σ u q , speciﬁes the consumption utility of the household. This utility is deﬁned as the sum of utilities of all its members, (recall there is no aggregate uncertainty at the household level). The measure of buyers is one, and the probability of meeting an appropriate seller is σ , so that the number of single-coincidence meetings involving a buyer in each period is σ . The second term in the maximand, −σ c (Q), speciﬁes the household’s disutility of production. The law of motion of the household’s money balances is given by (4.57). The ﬁrst term on the right side speciﬁes sellers’ money receipts from producing goods, and the second term speciﬁes buyers’ expenses when exchanging money for goods. If we denote λ as the the multipliers associated with constraints (4.55), recognizing that these constraints are applicable only when buyers are involved in single-coincidence meetings that occur with probability σ and take note that (4.54) can be written as q = c−1 (d), then the household’s problem (4.56)-(4.57) can be expressed as V(m) = max σ u ◦ c−1 (d) − c(Q) + βV m + σ (D − d) + σ λ (m − d) . d

The ﬁrst-order conditions and the envelope condition are u (q) λ + φ = , c (q)

(4.58)

λ (d − m) = 0,

(4.59)

φ−1 = σ λ + φ. β

(4.60)

Equation (4.58) states that, for a buyer in a match, the marginal utility of consumption must equal the opportunity cost of the amount of money that must be paid to acquire additional goods. To buy another

Money in Equilibrium

97

unit of a good, the buyer must give up c (q)/ units of money (see equation 4.54). Increasing the monetary payment has two costs to the buyer. He gives up the future value of money φ, and he faces a tighter constraint (4.55). Together φ and λ measure the marginal cost of obtaining a larger quantity of goods in exchange for money. Equation (4.59) is the Kuhn–Tucker condition associated with the multiplier λ. Finally, equation (4.60) describes the evolution of the marginal value of money. It states that the marginal value of money today, φ−1 /β = V (m), equals the discounted marginal value of money tomorrow, φ = βV (m+ ), plus the marginal beneﬁt of relaxing future cash constraints, σ λ. We focus on symmetric, stationary equilibria, where the value of money is the same across all households, φ = and across time. In addition symmetry implies that the values for the different variables associated with household h equal the values of the same variables of all other households. Consequently upperand lowercase vari ables equal to one another, m = M and d, q = (D, Q), and φ−1 = φ = βV (M). A steady-state, symmetric, monetary equilibrium is a collection q, λ, d, φ , satisfying equations (4.54) and (4.58)–(4.60), and φ > 0. From (4.58), u (q) λ = 1+ , c (q) φ and from (4.60), rφ = σ λ. Consequently u (q) r = 1+ . c (q) σ This equation is identical to the one found in our model of section 4.1; see (4.17). However, instead of a large household, there we considered a competitive market that allows agents to rebalance their money holdings. For all r > 0, the quantities traded in bilateral matches are inefﬁciently low, q < q∗ . Moreover, as r increases or σ decreases, the quantities traded fall. The transfer of money in a match is d = M, and the value of money is φ = c(q)/M. A key difference between the large household model and the model with alternating market structures and quasi-linear preferences is that in the former the value of money, φ, is household speciﬁc, whereas in the latter it is a market price taken as given by all households. This subtle difference can generate intricate technicalities, which are discussed in Rauch (2000), Berentsen and Rocheteau (2003), and Zhu (2008).

5

Properties of Money

In a simple state of industry, money is chieﬂy required to pass about between buyers and sellers. It should, then, be conveniently portable, divisible into pieces of various size, so that any sum may readily be made up, and easily distinguishable by its appearance, or by the design impressed upon it. —William Stanley Jevons, Money and the Mechanism of Exchange (1875, ch. 5)

The role that an asset plays as a medium of exchange depends on the nature of the frictions in the economy and on its physical characteristics. In chapters 3 and 4 the absence of record keeping and commitment implied that a tangible medium of exchange was needed to facilitate trade, and ﬁat money fulﬁlled that role. In those chapters although some physical properties of ﬁat money were made explicit—such as its divisibility or lack thereof—other important, and desirable, properties were left implicit. For example, it was implicitly assumed that ﬁat money did not depreciate or wear out over time, that it could be carried costlessly from one market to another, and that it could not be counterfeited. In this chapter we examine how the physical properties of money can affect its value and ability to perform the role of a medium of exchange. We re-examine the issue of divisibility, and investigate the implications for a medium of exchange that is costly to carry or that can be counterfeited. We are interested in how allocations and equilibria are affected when the physical properties of money depart for their ideal state. Commodity money systems have, at times, been plagued with a scarcity of certain types of coins. Since money cannot be scarce if it is perfectly divisible, we examine an environment where money is indivisible and there are fewer units of money than there are buyers. Obviously in this situation the total number of trades will be too low. In order to illustrate other important inefﬁciencies associated with indivisible and scarce money, we assume that buyers have heterogeneous valuations

100

Chapter 5

for the goods produced by sellers. Because of this, the economy will be characterized by a number of trade inefﬁciencies, where some of these inefﬁciencies would not arise if money was perfectly divisible. The second property of money we investigate is its portability. According to Jevons (1875, ch. 5): Many of the substances used as currency in former times must have been sadly wanting in portability. Oxen and sheep, indeed, would transport themselves on their own legs; but corn, skins, oil, nuts, almonds, etc., though in several respects forming fair currency, would be intolerably bulky and troublesome to transfer.

If we assume that it is costly to carry units of money, then money will not be held nor valued when the cost of carrying money is higher than some threshold. If, however, the carrying cost of money is not too large, then there are multiple stationary equilibria where money has a positive value in exchange. This suggests that fundamentals, such as carrying costs, as well as conventions matter for the use of an object as a means of payment. In the equilibrium where money has its highest value, the value of money decreases as the carrying cost increases. Finally, money is not neutral since, in a monetary equilibrium, an increase in the money supply implies that agents will hold more nominal balances, which increases the total cost of holding money and, hence, reduces welfare. The ﬁnal property of money that we examine is its recognizability or, in Jevons’s (1875, ch. 5) words, its cognizability: By this name we may denote the capability of a substance for being easily recognized and distinguished from all other substances. As a medium of exchange, money has to be continually handed about, and it will occasion great trouble if every person receiving currency has to scrutinize, weigh, and test it. If it requires any skill to discriminate good money from bad, poor ignorant people are sure to be imposed upon. Hence the medium of exchange should have certain distinct marks which nobody can mistake.

The art of counterfeiting has been around for as long as money. In medieval Europe, individuals clipped the edges of silver and gold coins and tried to pass off the depreciated coin as full bodied. During the nineteenth century in the United States, vast quantities of counterfeit banknotes were produced and passed off as the real thing. To address the issue of counterfeiting, we examine an environment where ﬁat money can be counterfeited at a ﬁxed cost, and sellers are unable to distinguish genuine from counterfeit notes. We show that the lack of recognizability results in an upper bound on the quantity of real balances that a buyer

Properties of Money

101

can transfer to the seller in a match. Even though counterfeiting does not occur in equilibrium, our model provides support for policies that make a currency harder to counterfeit: By raising the cost to produce counterfeits, policy makers can increase the velocity of money, output, and welfare. 5.1 Divisibility of Money In this section we investigate the implications of money being indivisible. In chapter 3.2 we considered a model with indivisible money and assumed that the supply of money was such that all buyers could exactly hold one unit of money, i.e., M = 1. We now assume that money is scarce or, equivalently, that there is a currency shortage, i.e., M < 1. In order to identify the inefﬁciencies associated with indivisible and scarce money we introduce buyer heterogeneity in the DM matches. In particular, the utility of a buyer in a bilateral match is εu(q), where ε is the realization of an idiosyncratic preference shock drawn from some cumulative distribution function F(ε) with support in R+ . A high ε means that the buyer’s marginal utility for the seller’s good is high, and a low ε means that it is low. The preference shocks are independent across time and across matches. They capture the idea that even though agents are ex-ante identical, buyers have idiosyncratic preferences over the goods produced by sellers in the DM. The timing of events in a representative period is illustrated in ﬁgure 5.1. A fraction σ of buyers and sellers are matched in the DM. Upon being matched, a buyer draws a preference shock ε for the output produced by the seller. If the buyer has some money, then he can make a take-it-or-leave-it offer to the seller. At night, buyers and sellers trade money and the general good in a centralized competitive market, CM, where the price of a unit of money in terms of the general good is φ. We focus on stationary equilibria where this price is constant over time. DAY (DM)

σ bilateral matches are formed.

Buyers receive a preference shock ε. Buyers make a take-it-or-leave-it offer.

Figure 5.1 Timing of a representative period

NIGHT (CM) Money is traded competitively against the general good at the price ϕ.

102

Chapter 5

5.1.1 Currency Shortage Since there is less than one indivisible unit of money per buyer, clearing of the money market in the CM requires that a fraction M of buyers end up with one unit of money and the remaining 1 − M end up with none. Moreover buyers are indifferent between holding one unit of money and holding zero unit. As in chapter 3.2, the concavity of the buyer’s value function implies that the buyer has no incentive to hold more than one unit of money. See the appendix. Therefore −φ + βV1 = βV0 ,

(5.1)

where V1 is the value of a buyer holding one unit of money in the DM and V0 is the value of a buyer holding no money. The left side of (5.1) is the expected discounted utility of a buyer who obtains one unit of money in the CM: the unit of money costs him φ and his continuation value in the next DM is V1 . The right side of (5.1) is the expected discounted utility of a buyer who exits the CM without money. In the DM, a matched buyer with one unit of money can make a takeit-or-leave-it offer, q, d , to the seller. The offer must satisfy the seller’s participation constraint, −c(q) + φ ≥ 0. Hence the buyer will choose the largest q he can afford with his unit of money, q = c−1 (φ), which is independent of the realization of his preference shock. The value of a buyer without money in the DM solves V0 = max (−φ + βV1 , βV0 ) = βV0 = 0.

(5.2)

The buyer cannot trade in the DM because he has no means of payment. In the CM, equilibrium requires that buyers are indifferent between holding or not holding one unit of money. The value of a buyer with one unit of money at the beginning of the DM is V1 = σ max εu(q) − φ + βV1 , βV1 dF(ε) + (1 − σ )βV1 . (5.3) With probability σ the buyer ﬁnds a seller. He draws a preference shock, ε, for the good produced by the seller. If the buyer chooses to make an offer, then his lifetime utility is εu(q) − φ + βV1 : he enjoys the utility of consumption and his continuation value in the CM is −φ +βV1 = βV0 . If the buyer chooses not to make an offer, his continuation value is simply βV1 . The value function (5.3) can be simpliﬁed to V1 = σ max εu(q) − φ, 0 dF(ε) + βV1 . (5.4)

Properties of Money

103

If the surplus from trading is positive, εu(q) − φ ≥ 0, then the buyer makes an offer. Otherwise, he chooses not to trade. Since q = c−1 (φ), the buyer chooses to trade if εu ◦ c−1 (φ) − φ ≥ 0. Let εR (φ) = φ/u ◦ c−1 (φ) denote the threshold for ε, below which the buyer chooses not to trade. Since u ◦ c−1 (φ) is strictly concave and u ◦ c−1 (0) = 0, it can be shown that εR (φ) is an increasing function of φ. That is, as money becomes more valuable, buyers become more choosy, and are only willing to spend their indivisible unit of money on goods that they highly value. Using (5.1) and (5.2), i.e., βV1 = φ, we can rewrite (5.4) as rφ = σ

∞

εR (φ)

εu ◦ c−1 (φ) − φ dF(ε).

(5.5)

According to (5.5), the value of money in equilibrium is such that the opportunity cost of holding one unit of money, the left side of (5.5), is equal to the expected surplus from a trade in the DM, the right side of (5.5). A steady-state equilibrium of the economy corresponds to a φ solution to (5.5). We ﬁrst examine the special case where ε = 1 in all matches. Then (5.5) becomes rφ = σ {u ◦ c−1 (φ) − φ},

(5.6)

or φ=

σ u ◦ c−1 (φ). r+σ

(5.7)

Given our assumptions about c and u, it is easy to check that there exists a unique φ > 0 that satisﬁes (5.7). In terms of comparative statics associated with the purchasing power of money, note that φ is independent of the quantity of money, M. From (5.7), ∂φ/∂σ > 0 and ∂φ/∂r < 0. Intuitively, as the matching probability σ increases, a buyer has a higher chance of trading in the DM, which makes money more valuable. As a consequence the quantities traded during the DM increase. And as the rate of time preference, r, increases, agents become more impatient, and the cost of holding money increases. As a consequence the value of money falls, and agents trade less in the DM. We now generalize these results to the case where the distribution of preference shocks is nondegenerate. If we divide both sides of (5.5) by

104

Chapter 5

u ◦ c−1 (φ) and use εR = φ/u ◦ c−1 (φ), then we get rεR = σ

∞ εR

(ε − εR ) dF(ε).

(5.8)

Equation (5.8) is a standard optimal stopping rule in sequential search models. It determines the reservation utility, above which it is optimal to accept a trade. For the sake of interpretation it can be rewritten as rεR = σ [1 − F(εR )]E[ε − εR |ε ≥ εR ]. The left side is the ﬂow value from agreeing to trade at the reservation utility, while the right side is the expected return from the search activity. The return from search for a buyer is the probability of meeting a seller, σ , times the probability that the match speciﬁc component is larger than the reservation value, 1 − F(εR ), times the expected difference between ε and εR conditional on ε being larger than εR . Integrating the right side of (5.8) by parts, we get rεR = σ

∞ εR

[1 − F(ε)] dε.

(5.9)

There is a unique εR > 0 that solves (5.9). To see this, notice that the left side is increasing in εR from 0 to ∞ as εR goes from 0 to ∞, while the right side is decreasing from σ εe , where ε e denotes the mean of the distribution F, to 0 as εR goes from 0 to ∞. See ﬁgure 5.2. It is also immediate from (5.9) that ∂εR /∂σ > 0 and ∂εR /∂r < 0. If it is easier to ﬁnd a seller in the DM, then buyers become more demanding and raise their reservation utility. In contrast, if buyers become less patient, then they lower their reservation utility. Since εR = φ/u ◦ c−1 (φ), there is a positive relationship between the value of money and the buyer’s reservation utility. Consequently ∂φ/∂σ > 0 and ∂φ/∂r < 0. We now turn to normative considerations. We measure social welfare by the discounted sum of utilities of buyers and sellers, −1

W = σ (1 − β)

M

∞ εR

[εu(qε ) − c(qε )]dF(ε),

where M ∈ (0, 1) and qε is the output traded in a match with idiosyncratic shock ε. (The net aggregate utility from consuming and producing the general good in the CM is zero.) In this situation, since a change in M affects the extensive margin—the number of trade matches—an increase in M raises welfare. This extensive margin result disappears when money is perfectly divisible. A change in M, however, has no

Properties of Money

105

σε e r εR

∞

σ ∫ 1 − F(ε)dε εR

εR Figure 5.2 Reservation utility in the model with indivisible money

effect on the intensive margin—the quantity produced in a particular trade match. In terms of efﬁcient allocations, a social planner would choose εR∗ and q∗ε such that εR∗ = 0, εu (q∗ε ) = c (q∗ε ). The social planner would like agents to trade in all matches, and the quantities traded should equalize the marginal utility of consumption of the buyer with the marginal disutility of production of the seller. In contrast, in equilibrium, εR > εR∗ = 0. Buyers do not trade in matches when they have a low valuation for the seller’s output. Hence, for low values of ε, there is a no-trade inefﬁciency. When ε = εR , then by definition, εR u(q) − c(q) = 0. However, when the socially efﬁcient level of output is produced, we get εR u(q∗εR ) − c(q∗εR ) > 0. In this situation agents trade too much from a social perspective, i.e., q > q∗εR . Finally, for values of ε sufﬁciently large, agents trade too little from a social perspective, i.e., q < q∗ε . To explain the no-trade and too-much-trade inefﬁciencies, consider a buyer’s consumption decision when his preference shock is in a

106

Chapter 5

qε qε*

q

ε

εR No trade

Too much trade

Too little trade

Figure 5.3 Trade inefﬁciencies with indivisible money

neighborhood of εR ; see ﬁgure 5.3. If ε = εR , the buyer is just indifferent between consuming q units of the good in exchange for his unit of money and not trading. The seller is also indifferent between producing q units for one unit of money and not trading. If ε is slightly below εR , then no trade takes place because the bid price of money—the quantity qb = c−1 (φ) the seller is willing to produce for one unit of money—is smaller than the ask price of money—the quantity of output qa = u−1 (φ/ε) the buyer demands to give his unit of money up. In contrast, if ε is slightly above εR , then the bid price of money is larger than its ask price, and because of the buyer-takes-all bargaining protocol, a trade takes place at the bid price. The consumed quantity, however, is inefﬁciently large because of the buyer’s low valuation for the seller’s output. In summary, the following inefﬁciencies arise when money is indivisible; see ﬁgure 5.3: 1. The number of trade matches will be too low if there is a shortage of currency, i.e., when M < 1. 2. For low values of ε, buyers do not trade even though it would be socially optimal to do so.

Properties of Money

107

3. For intermediate values of ε, agents trade too much. 4. For high values of ε, agents trade too little. 5.1.2 Indivisible Money and Lotteries When agents don’t trade, ε < εR , or when they trade too much, q > q∗ε , they could achieve a pairwise superior outcome in the DM if the buyer could somehow give up only a fraction of his unit of money to the seller. But this is not feasible since each unit of money is indivisible. The buyer could, however, overcome this indivisibility by offering to transfer his unit of money with some probability by using a lottery device. Since output is perfectly divisible, a lottery is only needed for the money balances that are transferred from the buyer to seller in the DM. When lotteries are used, a take-it-or-leave-it offer by the buyer can be compactly described by (qε , ςε ), where qε is the amount of the DM good produced by the seller, and ςε ∈ [0, 1] is the probability that the buyer transfers his unit of money to the seller. Consider a match between a buyer and a seller. The take-it-or-leave-it offer that a buyer with one indivisible unit of money makes to the seller, (qε , ςε ), solves the problem, max εu(q) − ς φ subject to − c(q) + ςφ = 0, 0 ≤ ς ≤ 1. (5.10) q,ς

The buyer maximizes his expected surplus, which is the difference between his utility of consumption in the DM minus the probability that he gives up his unit of money times the value of money in the CM. The offer is such that the seller is indifferent between accepting and rejecting. If c(q∗ε ) ≤ φ, then the solution to (5.10) is qε = q∗ε , ςε =

c(q∗ε ) ; φ

if c(q∗ε ) > φ, then the solution is qε = q = c−1 (φ) and ς = 1. In contrast to an environment without lotteries, buyers trade in all matches, which implies that εR = 0, and they never trade too much, qε ≤ q∗ε . By the same reasoning as in (5.5), the value of money is given by the solution to rφ = σ

0

∞

εu(qε ) − ςε φ dF(ε).

(5.11)

108

Chapter 5

From the seller’s participation constraint, ςε φ = c qε , and (5.11) can be written as rφ = σ

∞

εu qε − c qε dF(ε).

0

(5.12)

The opportunity cost of holding money, the left side of (5.12), is equal to the expected match surplus in the DM, the right side of (5.12). Denote ε˜ as the threshold for the preference shock below which agents trade the socially efﬁcient quantity, i.e., ε˜ is implicitly deﬁned by q∗ε˜ = c−1 (φ). Then (5.12) can be rewritten as rφ = σ

0

ε˜

εu q∗ε − c q∗ε dF(ε) + σ

ε˜

εu ◦ c−1 (φ) − φ dF(ε).

∞

(5.13)

It is easy to check that (5.13) determines a unique φ > 0: the left side is linear in φ, while the right side is strictly increasing and concave in φ. In the absence of lotteries, if a buyer’s valuation for a good is very low, then the ask price of money, qa = u−1 (φ/ε), is larger than the bid price of money, qb = c−1 (φ), and consequently no trade takes place. This no-trade inefﬁciency disappears with lotteries because when a buyer’s valuation for a good is low, he simply delivers the indivisible money with a probability greater than zero, but less than one, in exchange for a small (and efﬁcient) amount of the good. In the absence of lotteries, if the buyer’s valuation for the good is low, but not too low, then the ask price of money is smaller than the bid price, and consequently exchange takes place but at a level of DM output that is larger than the efﬁcient level. Similar to the no-trade inefﬁciency, the too-muchtrade inefﬁciency disappears with lotteries on indivisible money since the buyer can effectively deliver, in expected terms, less than a unit of money for the efﬁcient level of DM output. Finally, note that lotteries do not eliminate the “too-little-trade’’ inefﬁciency, which occurs when ε > ε˜ . If the support of the distribution of the preference shocks is not too large, it is possible to have agents trade the socially efﬁcient quantity in all matches. Consider the case where ε = 1 in all matches. We saw that with divisible money the output is too low provided that r > 0. With indivisible money and lotteries, the value of money is determined by (5.13):

Properties of Money

109

if φ > c(q∗ ), rφ = σ u q∗ − c q∗ = σ u ◦ c−1 (φ) − φ otherwise.

(5.14)

The determination of the equilibrium is illustrated in ﬁgure 5.4. It can easily be checked that q = q∗ if and only if the left side of (5.14) evaluated at φ = c(q∗ ) is less than the right side of (5.14) evaluated at q = q∗ , i.e., c(q∗ ) ≤

σ u(q∗ ). r+σ

If the allocation (q, y) = (q∗ , c(q∗ )) is incentive feasible in the environment with money—see chapter 3.1 and the deﬁnition of AM in (3.8)—or in a credit environment with public record keeping—see chapter 2.3 and the deﬁnition of APR in (2.24)—then it can be implemented as an equilibrium in a monetary economy with indivisible money by a take-it-or-leave-it offer when buyers can use lotteries. Note, however, there is still an inefﬁciency due to the shortage of currency, M < 1, which reduces the number of matches.

rϕ σ [u(q)−c(q)]

σ [u(q*)−c(q*)]

c(q*) Figure 5.4 Equilibrium with indivisible money and lotteries

ϕ

110

Chapter 5

5.1.3 Divisible Money In this section we examine the case of a perfectly divisible money to see how these allocations compare to those with indivisible money. We focus on stationary equilibria. When money is divisible and a buyer’s preference shock is ε, a takeit-or-leave-it offer by the buyer in the DM is now a pair (qε , dε ), where qε is the amount of the search good produced by the seller, dε ∈ [0, m] is the transfer of money from the buyer to the seller, and m is the buyer’s money holdings. The buyer solves the problem max εu(q) − dφ q,d

− c(q) + dφ = 0

subject to and

0 ≤ d ≤ m.

(5.15)

This problem is analogous to (5.10). If c(q∗ε ) ≤ mφ, then the solution is qε = q∗ε , dε =

c(q∗ε ) ; φ

if c(q∗ε ) > mφ, then qε = q = c−1 (mφ) and d = m. The divisibility of money, just like the use of lotteries when money is indivisible, removes the no-trade and too-much-trade inefﬁciencies, i.e., εR = 0 and qε ≤ q∗ε . The expected lifetime utility of a buyer in the CM is

b

W (m) = φm + max −φm + βσ m

b

+β(1 − σ )W (m ) ,

∞ 0

εu(qε ) + W b (m − dε ) dF(ε) (5.16)

where qε and dε are functions of the buyer’s money holdings in the DM, m . According to (5.16) the buyer readjusts his money holdings in the CM by acquiring m − m new units, which costs him φ(m − m) in terms of the general good. In the next DM, if the buyer is in a trade match, which occurs with probability σ , then he consumes qε units of the DM output and delivers dε units of money. We use the linearity of W b , i.e., W b (m) = φm + W b (0), and φdε = c(qε ) to solve for the buyer’s choice of money holdings:

Properties of Money

max −rφm + σ m≥0

111

εu(qε ) − c(qε ) dF(ε)

0

= max −rφm + σ m≥0

+σ

∞

∞

ε˜ (φm)

ε˜ (φm)

εu(q∗ε ) − c(q∗ε ) dF(ε)

0

εu ◦ c

−1

(φm) − φm) dF(ε) ,

(5.17)

where ε˜ solves q∗ε˜ = c−1 (φm). The ﬁrst-order condition with respect to m is ∞ −1 r εu ◦ c (mφ) = − 1 dF(ε). (5.18) σ c ◦ c−1 (mφ) ε˜ (φm) For market clearing, m = M, which implies that (5.18) determines a unique φ > 0 since the right side of (5.18) is decreasing in φ. For any r > 0, the right side of (5.18) must be positive, which implies that qε < q∗ε for some ε even if the support of F(ε) is ﬁnite. The divisibility of money does not remove the too-little trade inefﬁciency. This inefﬁciency arises because there is a cost of holding real balances due to discounting. If this cost is driven to zero, i.e., r → 0, then the right side of (5.18) is zero, meaning that real balances are sufﬁciently large to trade the socially efﬁcient quantities in all matches. Moreover, because money is divisible, it is feasible to endow all buyers with M units money at the beginning of a period, even when M < 1. Therefore, when money is perfectly divisible, currency shortages cannot occur and the number of trade matches is at its maximum. 5.2 Portability of Money We now consider another important physical attribute of a medium of exchange: portability. Portability describes the ease at which an object can be carried to where it is needed, i.e., into bilateral meetings. We equate portability with the cost of bringing money into the DM, and assume that at the beginning of each period, the buyer incurs a real cost κ > 0 for each unit of money he holds. As in chapter 4.1 the buyer’s choice of money holdings in the CM of period t is given by b max −φt m + βVt+1 (m) . (5.19) m≥0

112

Chapter 5

However, the value of being a buyer in the DM is now given by b b (m) = −κm + σ max u ◦ c−1 (φt+1 d) − φt+1 d + φt+1 m + Wt+1 (0), Vt+1 d∈[0,m]

(5.20) b (m) = φ b where we have used Wt+1 t+1 m + Wt+1 (0), and from the buyer−1 takes-all bargaining assumption, qt+1 = c (φt+1 d). The ﬁrst term on the right side of (5.20) is new and represents the proportional cost from b (m) into holding m units of money. Substituting this expression for Vt+1 (5.19), we solve for the choice of money holdings as

φt /φt+1 −1 max − − 1 φt+1 m − κm + σ max u ◦ c (φt+1 d) − φt+1 d . β d∈[0,m] m∈R+

(5.21) The cost of accumulating φt+1 m units of real balances has two components: the part due to inﬂation and discounting, (φt /φt+1 − β)/β, and the part due to the imperfect portability of money, κ/φt+1 . Provided that (φt /φt+1 − β) /β + κ/φt+1 > 0, it is costly to hold money and, hence, d = m. Substituting c(qt ) = φt m into (5.21) and rearranging, we get

φt /φt+1 κ max − c(qt+1 ) + σ u(qt+1 ) − c(qt+1 ) . (5.22) −1+ β φt+1 qt+1 ∈R+ Assuming an interior solution, the ﬁrst-order condition to this problem is given by u (qt+1 ) φt /φt+1 − β κ = 1+ . + c (qt+1 ) σβ σ φt+1

(5.23)

The money market clears if m = M, which implies from (5.23),

u (qt+1 ) φt = βφt+1 σ − 1 + 1 − βκ, (5.24) c (qt+1 ) where qt+1 = min q∗ , c−1 (φt+1 M) . This equation generalizes (4.16) in an obvious way. Even though φt = φt+1 = 0 does not solve (5.24), it should be noticed that for all κ > 0 there is a nonmonetary equilibrium, where the solution to (5.22) is a corner solution, and agents dispose of their money holdings since they have no value and they are costly to hold.

Properties of Money

113

A monetary equilibrium is a sequence {φt }∞ t=0 solving the ﬁrst-order difference equation (5.24), where φt is bounded. Consider ﬁrst stationary equilibria where money is valued, qt = qt+1 = qss > 0. At a steady state, (5.24) can be rewritten as u (qss ) r κM = 1+ + . c (qss ) σ σ c(qss )

(5.25)

In contrast to the previous section, the steady-state monetary equilibrium with positive output, if it exists, is no longer unique. To see this, we assume the following functional forms and parameter values: c(q) = q, u(q) = q1−a /(1 − a), a < 1, and σ = 1. Then (5.25) can be rewritten as (qss )1−a = (1 + r) qss + κM.

(5.26)

The left side is a strictly concave function of qss , while the right side is linear with a positive intercept. Consequently, if κ is below a threshold, then there are two solutions qss > 0 to (5.26); otherwise, there is no monetary equilibrium. Suppose, for example, that a = 1/2. Then the two solutions to (5.26) are √

2 1 − 4(1 + r)κM , 2(1 + r) √

2 1 − 1 − 4(1 + r)κM ss qL = , 2(1 + r)

qss H=

1+

if 4(1 + r)κM < 1. The intuition behind the multiplicity of steady-state equilibria is that the cost of holding one unit of real balances is κ/φ, which depends on the value of money. If the value of money is low, then the cost of holding real balances is high. Buyers do not want to accumulate large real balances, which makes the value of money low. A similar logic applies to the case where the value of money is high. A monetary equilibrium is more likely to exist if the candidate object to be used as money is not too costly to hold. Hence fundamentals matter for the use of an object as a means of payment. But good fundamentals are not sufﬁcient for an object to be used as money since the liquidity property of the object—its acceptability—is endogenous. The following example illustrates this point.

114

Chapter 5

Suppose there are two objects that can serve as a means of payments, called object 1 and object 2. There is a ﬁxed supply of both objects, M1 and M2 . The storage cost of object 1 is κ1 , and the storage cost of object 2 is √ κ2 . For simplicity, assume that u(q) = 2 q, c(q) = q, and σ = 1. A steadystate monetary equilibrium where only object 1 is used as money exists if 4(1 + r)κ1 M1 < 1; a steady-state monetary equilibrium where only object 2 is used exists if 4(1 + r)κ2 M2 < 1. If κ2 M2 > κ1 M1 , then whenever there exists an equilibrium where object 2 is used as money, there is also an equilibrium where object 1 is used as money, but the reverse is not true. In this sense, object 1 is more likely to be used as means of payment than object 2. An object is more likely to be used as means of payment if the aggregate cost from carrying this object is low, i.e., the storage cost per unit must not be too large and the object must not be too abundant. Still the object with a large storage cost can emerge as the medium of exchange because of self-fulﬁlling beliefs. Consider now the effects of an increase in κ on the high steady-state equilibrium. It can easily be checked that there is a negative relationship between qss H and κ. When ﬁat money is more costly to carry, the DM output falls. Moreover, money is no longer neutral. As M increases, qss H decreases since carrying money involves additional real resources. The comparative statics at the low steady-state monetary equilibrium are opposite to those at the high steady-state monetary equilibrium. Finally, let’s consider nonstationary equilibria. If we adopt the same functional form and parameter values as above, then (5.24) becomes 1−a qt = β qt+1 − βκM.

(5.27)

As illustrated in ﬁgure 5.5, there are a continuum of trajectories leading to the low steady-state monetary equilibrium, while there is a unique trajectory—the stationary one—that leads to the high steady-state monetary equilibrium. We have considered the case where κ ≥ 0. If κ < 0, then the medium of exchange can be interpreted as a commodity money, or a real asset since it provides its holder with a real dividend. When κ < 0, the phase line in ﬁgure 5.5 would shift down and intersect the horizontal axis at a positive value of qt . In contrast to the case where κ > 0, the phase line for κ < 0 would have a unique intersection with the 45o line. As well, when κ < 0, a nonmonetary equilibrium no longer exists. This is because the price of money, φ, is bounded below by its fundamental value, which is given by −βκ/(1 − β) > 0. If the price of money were below its fundamental

Properties of Money

115

qt + 1 qt +1 = qt

(κM) /

1 (1−a)

qLss

qHss

qt

Figure 5.5 Dynamic equilibria under imperfect portability

value, say zero, then (5.22) would have no solution, as agents would demand an inﬁnite amount of money in order to enjoy its real dividend. Moreover, for the functional form used as above, there is a unique monetary equilibrium, and it is the stationary monetary equilibrium, qt = qt+1 = qss . In a ﬁat monetary system there is a continuum of equilibria that lead to the autarkic outcome, and in all these equilibria the value of money at any date is lower than what would prevail in a stationary (monetary) equilibrium. As a result the stationary monetary equilibrium dominates, from a social welfare perspective, any of the inﬂationary equilibria. Since the presence of a commodity component eliminates any equilibria where money loses value overtime, there is a welfare gain associated with having a commodity money system. 5.3 Recognizability of Money In this section we analyze the implications of money being imperfectly recognizable. In particular, sellers are unable to distinguish genuine money from counterfeit money in the DM, and buyers can produce

116

Chapter 5

counterfeit bills at night, after the CM has closed. There is a ﬁxed cost k > 0 associated with engaging in counterfeiting activities each night, but the marginal cost of producing a counterfeit note is zero. Moreover the technology to produce counterfeits becomes obsolete after one period, so whenever an agent chooses to produce counterfeits at night he must incur the cost k. Counterfeits produced in period t − 1 are all detected and conﬁscated as agents enter the CM of period t. Hence the only venue to pass a counterfeit bill produced in period t − 1 is the DM of period t. A seller in the DM of period t would never knowingly accept a counterfeit since it is worthless at night. The terms of trade in the DM are determined by take-it-or-leave-it offers by buyers. Sellers do not observe the money holdings of the buyers or their decisions to produce counterfeits. To simplify the presentation, we assume that there are no search frictions in the DM, σ = 1. The strategic interactions between a buyer—who decides the amount of genuine money to hold, whether to produce counterfeits, and the terms of trade—and a seller—who must accept or reject the buyer’s offer—can be represented by a simple game, where the buyer makes the ﬁrst three moves. Things can be simpliﬁed a bit by noting that since the marginal cost of producing counterfeits is zero, a buyer will not need (or want) to accumulate genuine money if he produces counterfeits. Moreover, since money is costly to hold, i.e., r > 0, if a buyer decides to accumulate genuine money balances, then he will never hold more than what he intends to spend, i.e., m = d. Consequently two of the buyer’s moves can be collapsed into one, in which he chooses either to produce d counterfeits or to accumulate d units of genuine money. At the beginning of the CM of period t − 1, the buyer anticipates that he will make the offer q, d in the subsequent DM. If the buyer chooses to accumulate d units of genuine money balances in the CM, given his anticipated offer q, d , then, assuming this offer is accepted, his lifetime utility is −φd + β u q + W b (0) . (5.28) If instead he chooses to counterfeit d units of money, then his lifetime utility is −k + β u q + W b (0) . (5.29)

Properties of Money

117

The buyer is willing to accumulate genuine money if (5.28) exceeds (5.29), or if φd ≤ k.

(5.30)

Suppose that in the DM a seller is in a match with a buyer that offers terms of trade q, d . Recall that the seller is unable to distinguish between genuine and counterfeit money. If (5.30) holds, then the seller concludes that the buyer is holding genuine money. The seller’s reasoning is that (5.30) implies that a buyer’s strategy of offering terms of trade q, d and producing d counterfeits is dominated by the strategy of offering q, d and accumulating genuine money. So, if (5.30) holds, a buyer has no incentive to use counterfeit notes independent of the seller’s decision to accept or reject. If, however, (5.30) does not hold, then we assume that the seller believes that the buyer in the match is holding counterfeit notes, and he rejects the offer. Thereforea necessary, but not sufﬁcient, condition for the seller to accept offer q, d is that (5.30) holds. An equilibrium offer, q, d , by the buyer must satisfy (5.30), as well as the seller’s participation constraint. The buyer’s equilibrium offer satisﬁes (q, d) = arg max − (1 − β) φd + β u(q) − φd (5.31) subject to and

− c(q) + φd ≥ 0

φd ≤ k,

(5.32) (5.33)

where (5.31) is the buyer’s expected utility, net of the continuation value W b (0), (5.32) is the seller’s participation constraint, and (5.33) is the no-counterfeiting constraint. The problem that determines the equilibrium terms of trade q, d , (5.31)–(5.33), is similar to the one of the previous section, except that it incorporates an additional constraint, (5.33). Constraint (5.32) ensures that the seller will accept the offer with probability one, while constraint (5.33) ensures that the buyer has no incentive to produce counterfeit notes. The latter constraint places an upper bound on how many real balances the buyer can transfer. The amount of real balances that the buyer can transfer to the seller is equal to the cost of producing counterfeits. A noteworthy property of this equilibrium is that no counterfeiting ever takes place. The buyer cannot beneﬁt from counterfeiting since the seller understands the buyer’s incentives, and accordingly adjusts his

118

Chapter 5

acceptance rule. We now examine this idea in greater detail. Suppose that constraint (5.33) does not bind, i.e., k islarge. Then the offer q, d is given by the solution to (5.31)–(5.32); i.e., q, d solves u (q) = 1 + r, c (q) φd = c(q).

(5.34) (5.35)

Suppose, on the contrary, that constraint (5.33) binds, i.e, k is small. Then the offer q, d is given by the solution to the constraints (5.32)–(5.33), i.e, q, d solves q = c−1 (k) , φd = k.

(5.36) (5.37)

We can now deﬁne what we mean by k being large or small. There is a ¯ such that the solutions to (5.34) and (5.36) critical value for k, denoted k, coincide; that is, the critical value k¯ is given by the solution to u c−1 k¯ = (1 + r) c c−1 k¯ . (5.38) ¯ then the constraint (5.33) does not bind and the Therefore, if k ≥ k, buyer’s offer q, d is given by the solution to (5.34) and (5.35); if, how ¯ ever, k < k, then constraint (5.33) binds and the buyer’s offer q, d is given by (5.36) and (5.37). In either case, the clearing of the money market implies d = M, which pins down the value of money, φ. The determination of the equilibrium level of the DM good production, q, is illustrated in ﬁgure 5.6. When constraint (5.33) is not binding, ¯ the equilibrium q is given by the intersection of or equivalently if k > k, the horizontal line representing the cost of holding money, 1 + r, and a downward sloping curve representing the function u (q)/c (q). Provided that u (0)/c (0) > 1 + r, which is true since we assume that u (0) = ∞ and c (0) = 0, there exists a monetary equilibrium. This condition is independent of k. The threat of counterfeiting does not make the monetary equilibrium less likely to prevail. In particular, if φM is sufﬁciently small, it would be more costly for a buyer to incur the ﬁxed cost to produce counterfeit money rather than going into the CM to produce φM units of the general good. ¯ the equilibrium level q is When constraint (5.33) binds, i.e., k < k, given by the intersection of the horizontal line representing the cost of

Properties of Money

119

1+ r

u'(q) c'(q) c −1(k)

q

Figure 5.6 Determination of the equilibrium

holding money, 1 + r, and the vertical line emanating from (5.36), q = c−1 (k). In this case, note that ∂q/∂k > 0 and ∂φ/∂k > 0. Diagrammatically speaking, an increase in k shifts the vertical line to the right, resulting in a higher production level of the DM good; as a result money becomes more valuable. An implication of this result is that policies designed to make it harder to counterfeit ﬁat money, e.g., the use of special paper and ink, the frequent redesign of the currency, and so on, can have real effects even when counterfeiting does not take place. 5.4 Further Readings The ﬁrst generations of search-theoretic models of monetary exchange assumed indivisible money and currency shortage. This includes Diamond (1984), Kiyotaki and Wright (1989, 1991, 1993), Shi (1995), Trejos and Wright (1995), and Wallace and Zhou (1997). Rupert, Schindler, and Wright (2001) extend the work of Trejos and Wright (1995) by generalizing agents’ production choices and bargaining power. Berentsen, Molico, and Wright (2002) and Lotz, Schevchenko, and Waller (2007) introduced lotteries into the analysis. Shevshenko and Wright (2004)

120

Chapter 5

show that one can obtain partial acceptability of a means of payment by introducing heterogeneity across agents. Rupert, Schindler, and Wright (2000) provide a survey of search-theoretic models with indivisible money. Camera and Corbae (1999) and Taber and Wallace (1999) relax the unit upper bound on money holdings and study price dispersion and divisibility of money. Molico (2006) has one of the ﬁrst models with perfectly divisible goods and money. Redish and Weber (2008) build a random-matching monetary model with two indivisible coins with different intrinsic values and study small change shortages. The assumption of indivisible money in the presence of match-speciﬁc preference shocks, and its implications for the efﬁciency of monetary exchange, are studied in Berentsen and Rocheteau (2002, 2003). Matchspeciﬁc shocks have also been used in Shi’s (1997) large household model by Shi and Peterson (2004) and in the search labor literature by Marimon and Zilibotti (1997) and Pissarides (2000, ch. 6). Kiyotaki and Wright (1989) and Aiyagari and Wallace (1991) studied how storage costs affect the ability of a commodity to be used as means of payment. See also Kehoe, Kiyotaki, and Wright (1993) and Renero (1998, 1999). The role of money as a recognizable asset has been emphasized in Brunner and Meltzer (1971) and Alchian (1977), and it has been formalized by King and Plosser (1986), Williamson and Wright (1994), and Banerjee and Maskin (1996). Williamson and Wright showed that money could be valued in a double-coincidence-of-wants environment if sellers have private information about the quality of the good they hold. Kim (1996) extended the model to endogenize the fraction of informed agents (who can recognize the quality of goods) in the economy. Trejos (1999) studied a version of the Williamson–Wright model with divisible goods, and Berentsen and Rocheteau (2004) considered the case with both divisible money and divisible goods. In order to establish the robustness of the monetary institution, Cuadras-Morato (1994) and Li (1995) showed that a good can be used as a medium of exchange even if its quality is uncertain. Kultti (1996) and Green and Weber (1996) were the ﬁrst papers to study counterfeiting of currency in a random-matching model with exogenous prices. Williamson (2002) investigated the counterfeiting of banknotes in a random-matching model with indivisible money but divisible output. Nosal and Wallace (2007) and Li and Rocheteau (2008)

Properties of Money

121

introduced lotteries as a proxy for divisible money and showed that it allows buyers to signal the quality of their money holdings. Cavalcanti and Nosal (2007) and Monnet (2005) adopted a mechanism design approach and focused on pooling allocations. A model of counterfeiting with perfectly divisible money, as examined in this chapter, was initially studied in Rocheteau (2008) and Li and Rocheteau (2009). These papers provide a more detailed analysis of the seller’s beliefs. Quercioli and Smith (2009) introduced multiple denominations and a costly decision to verify currency in a nonmonetary counterfeiting model. Appendix Optimal Choice of Money Holdings in the Indivisible Money Model We establish that the buyer has no incentive to accumulate more than one unit of money so that the support for the money distribution across buyers is {0, 1}. Consider a buyer in a match holding m units of money with a preference shock ε. The buyer is willing to spend at least d ∈ {1, . . ., m} units of money if εu ◦ c−1 (φd) − φd ≥ εu ◦ c−1 φ(d − 1) − φ(d − 1). According to the inequality above, the buyer’s surplus from spending d units of money is greater than the surplus from spending d − 1 units of money. Deﬁne εR,d the threshold for ε above which it is optimal to spend the dth units of money. εR,d =

φ u ◦ c−1 (φd) − u ◦ c−1

.

φ(d − 1)

From the concavity of u◦c−1 (φd), it is easy to check that εR,d is increasing in d. Let υ(m) denote the expected surplus of the buyer in the DM from holding m units of money. It is given by υ(m) = σ

m−1 εR,d+1 d=1 εR,d ∞

εu ◦ c−1 (φd) − φd dF(ε)

+σ

εR,m

εu ◦ c−1 (φm) − φm dF(ε).

122

Chapter 5

Consequently the utility gain associated with the mth unit of money is υ(m)−υ(m−1) = σ

∞ εR,m

ε u ◦ c−1 (φm) − u ◦ c−1 (φ(m − 1)) − φ dF(ε).

By the deﬁnition of εR,m , υ(m) − υ(m − 1) = σ u ◦ c−1 (φm) − u ◦ c−1 (φ(m − 1)) ∞ ε − εR,m dF(ε). εR,m

Integration by parts obtains υ(m) − υ(m − 1) = σ u ◦ c−1 (φm) − u ◦ c−1 (φ(m − 1))

∞

εR,m

1 − F(ε)dε.

Due to the concavity of u ◦ c−1 (φm) and the fact that εR,m is increasing in m, υ(m) − υ(m − 1) is decreasing with m. Since the cost of holding an additional unit of money is rφ, it is optimal to hold m units of money if υ(m) − υ(m − 1) ≥ rφ, υ(m + 1) − υ(m) ≤ rφ. By the deﬁnitions of εR,m and εR,m+1 , these inequalities can be rewritten as σ σ

∞

εR,m ∞ εR,m+1

1 − F(ε)dε ≥ rεR,m , 1 − F(ε)dε ≤ rεR,m+1 .

In the case of a currency shortage, M < 1, the ﬁrst inequality holds at equality for m = 1, and the second inequality is satisﬁed from the fact that εR,m is increasing in m, i.e., εR,2 > εR,1 . The Shi–Trejos–Wright Model of Indivisible Money In chapter 5.1.1 we presented a model with a currency shortage and indivisible money. A related model was ﬁrst proposed by Shi (1995) and Trejos and Wright (1995). The environment in those models is similar to

Properties of Money

123

the one we consider, except that there is no centralized market where agents can readjust their money holdings. Moreover individual money holdings are restricted to the set {0, 1}, i.e., an agent cannot accumulate more than one unit of money. An agent holding one unit of money is called a buyer, while the agent without money is called a seller. The model is in continuous time. The Poisson arrival rate of a singlecoincidence meeting—an encounter between an agent and a producer of a good he wishes to consume—is denoted by σ . This means that on a small interval of time of length dt, the probability of a single-coincidence meeting is σ dt. For simplicity, we rule out double-coincidence-of-wants meetings, where two matched agents would like to consume their partner’s output. For example, suppose that there are J ≥ 3 types of goods and J types of agents, where agents are evenly divided across types. An agent to type j produces good j but wishes to consume good j + 1 (modulo J). Then, the probability of a single-coincidence meeting is σ = 1/J and the probability of a double-coincidence-of-wants meeting is zero. Finally, agents are matched at random. So conditional on a meeting, the probability that the partner holds one unit of money is M, while the probability that he doesn’t hold money is 1 − M. Given these assumptions, we can write the ﬂow Bellman equations of a buyer and a seller as follows: rV1 = σ (1 − M) u q + V0 − V1 , (5.39) rV0 = σ M −c(q) + V1 − V0 . (5.40) These ﬂow Bellman equations (5.39) and (5.40) can be interpreted as asset pricing equations where V1 and V0 are the values of an asset in two different states. The left side of the ﬂow Bellman equation represents the opportunity cost of holding the asset, while the right side is the expected return from holding the asset (dividend ﬂows and capital gains or losses). According to (5.39) a buyer meets a seller who produces a good that he wishes to consumes with Poisson arrival rate σ (1 − M). In this event the buyer enjoys the utility from consuming q units of the output produced by the seller, u(q), and transfers his indivisible unit of money to the seller, which generates a capital loss V1 − V0 . According to (5.40) a seller meets a buyer who wishes to consume the good he produces with probability σ M. In this event the seller suffers the disutility of producing q units of output, c(q), but he receives one unit of money, which generates a capital gain of V1 − V0 .

124

Chapter 5

The quantity of output produced in a bilateral match, q, is determined by a take-it-or-leave-it offer by the buyer. The offer makes the seller indifferent between accepting and rejecting a trade, c(q) = V1 − V0 .

(5.41)

It is then immediate from (5.40) that V0 = 0, the seller gets no surplus from a trade. Substituting V1 = c(q) into (5.39), we obtain c(q) =

σ (1 − M) u q . r + σ (1 − M)

(5.42)

Asteady-state equilibrium is a q that solves (5.42). First, q = 0 is a solution to (5.42). There always exists a nonmonetary equilibrium. Second, since the left side of (5.42) is convex and the right side of (5.42) is strictly concave, there is a unique q > 0 that solves (5.42). So there is a unique steady-state monetary equilibrium. It is easy to check that ∂q/∂σ > 0, ∂q/∂M < 0, and ∂q/∂r < 0. So, in contrast to the model presented in chapter 5.1.1, the quantity traded in bilateral matches is affected by the supply of money. This difference can be explained by the fact that in the Shi–Trejos–Wright model, a buyer is matched at random with any agent from the whole population, whereas in chapter 5.1.1 buyers are only matched with sellers. Except for this difference, the two models have the same equilibrium condition. We now describe the dynamics of the Shi–Trejos–Wright model. For simplicity, we adopt the normalization c(q) = q. The ﬂow Bellman equations become ˙ 1, rV1 = σ (1 − M) u q + V0 − V1 + V (5.43) ˙ 0, rV0 = σ M −q + V1 − V0 + V (5.44) where a dot over a value function indicates a time derivative. Equations (5.43)–(5.44) are generalizations of (5.39)–(5.40) where the expected return of the asset also includes the change of the value of the asset over time, the last terms on the right sides of (5.43) and (5.44). From the buyer-take-all assumption, (5.41), q = V1 and V0 = 0. Substituting V1 = q into (5.43), we obtain the following ﬁrst-order differential equation: q˙ = [r + σ (1 − M)] q − σ (1 − M) u q . (5.45) The phase line associated with this differential equation, the right side of (5.45), goes through the origin and is strictly convex. It is represented in

Properties of Money

q⋅

125

[r+σ(1−M)]q−σ(1−M)u(q)

q

Figure 5.7 Dynamics of the Shi–Trejos–Wright model

ﬁgure 5.7. It has a unique intersection with the horizontal axis such that q > 0, which corresponds to the unique steady-state monetary equilibrium. The initial value of money cannot be greater than the positive steady-state value since otherwise the value of money would become unbounded and the match surplus would be negative. If the initial value of money is lower than the positive steady-state value, then the value of money decreases over time. Consequently there are a continuum of nonstationary monetary equilibria converging to the nonmonetary equilibrium.

6

The Optimum Quantity of Money

Milton Friedman’s (1969) doctrine regarding the “optimum quantity of money’’—according to which an optimal monetary policy would involve a steady contraction of the money supply at a rate sufﬁcient to bring the nominal interest rate down to zero—is undoubtedly one of the most celebrated propositions in modern monetary theory, probably the most celebrated proposition in what one might call “pure’’ monetary theory . . . [T]he general equilibrium literature has shown that the question of optimal monetary policy cannot be settled—in the sense of producing explicit quantitative advice for policy makers—without needing to specify in relative detail a model of how money is used in the economy. —Michael Woodford, The Optimum Quantity of Money, in Handbook of Monetary Economics (1990, ch. 20)

By not specifying the frictions that make monetary exchange useful, reduced-form models do not fully articulate how monetary policy affects the economy. Here we adopt instead the strategy of constructing economic environments where the presence of ﬁat money is essential, and the societal beneﬁts of monetary exchange are explicitly spelled out. By following this strategy, we are able to show that the same frictions that support positively valued ﬁat money can also provide new insights for monetary policy. So far we have only considered a one-time change in the money supply. In this chapter we go one step further and assume that monetary policy takes the form of a constant money growth rate. By changing the rate of growth of money supply, the monetary authority is able to affect the rate of return of currency, and hence agents’ incentives to hold real balances. This in turn has implications for equilibrium allocations, and society’s welfare. Under standard trading protocols, such as bargaining, price taking, and price posting, optimal monetary policy is characterized by the

128

Chapter 6

so-called Friedman (1969) rule. According to this policy prescription, the policy maker must engineer a rate of return for money that compensates agents for the cost of holding money balances. This can be accomplished by contracting the money supply at a rate approximately equal to the agent’s rate of time preference. By doing this, the policy maker can drive the cost associated with holding real balances to zero, which in turn implies that agents will hold sufﬁcient money balances to maximize their surpluses from trade. While the Friedman rule is optimal under most trading protocols, it does not necessarily implement socially efﬁcient allocations. For example, under the Nash bargaining solution, the quantities traded are inefﬁciently low even when the cost of holding real balances is driven to zero. The optimality of the Friedman rule is a robust ﬁnding across various kinds of monetary models, but it is rarely observed in practice. In order to reconcile this observation with the predictions of our model, we ﬁrst discuss the incentive feasibility of the Friedman rule when the government’s coercive power (to tax) is limited. Even though the policy maker would like to implement the Friedman rule through a contraction of the money supply, this policy may not be feasible. In particular, agents may choose not to participate in the market in order to avoid incurring the tax that is required to make the money supply contract at the optimal rate. An alternative explanation for the nonobservance of the Friedman rule in practice is that it may not be the optimal monetary policy for some environments. We provide two extensions of the model where running the Friedman rule is feasible but not optimal. In the ﬁrst extension, we suppose that the number of trades in the decentralized market depends on the relative numbers of buyers and sellers in the market, and agents can choose whether to be buyers or sellers. It will turn out that the number of matches is inefﬁcient because agents ignore the effect of their participation decisions on other agents’ matching probabilities. Because inﬂation acts as a tax on participation, a deviation from the Friedman rule may be optimal. In the second extension, we suppose that buyers receive uninsurable idiosyncratic productivity shocks. A growing money supply allows some redistribution of real balances among young buyers, and provides some valuable insurance. Hence in this situation a strictly positive inﬂation rate is socially desirable.

The Optimum Quantity of Money

129

6.1 Optimality of the Friedman Rule In this section we determine the optimal growth rate of the money supply in the context of the divisible monetary economy studied in chapter 4. Let Mt represent the aggregate stock of money at the beginning of period t, and γ ≡ Mt+1 /Mt the gross growth rate of the money supply. Money is injected, or withdrawn, in a lump-sum fashion in the competitive market, CM, at night. If γ > 1, then injections of money occur at the beginning of the CM; if γ < 1, then money is withdrawn at the end of the CM. If γ < 1, we assume that the government has sufﬁcient coercive power to force agents to pay the lump-sum taxes. The government is only able to tax in the CM because agents are anonymous in the decentralized market, DM, during the day, and, hence, cannot be monitored or coerced at that time. Since agents have quasi-linear preferences in the CM—preferences that eliminate wealth effects—we will assume without loss of generality that only buyers receive the monetary transfers. The timing of events is illustrated in ﬁgure 6.1. We focus on steady-state equilibria, where the real value of the money supply is constant over time, i.e., φt Mt = φt+1 Mt+1 . Note that the gross rate of return on money is φt+1 /φt = Mt /Mt+1 = γ −1 . Period t

Period t +1

Transfers

Transfers

NIGHT (CM)

Mt

DAY (DM)

NIGHT (CM)

Mt+1

Mt+2

Agent’s real balances:

z = ϕt m

Figure 6.1 Timing of a representative period

γ −1z = ϕt+1 m

130

Chapter 6

Since the price of money is not constant across time, we will write the value functions, V b and W b , as functions of the buyer’s real balances, z = φt mt , expressed in terms of the general good traded in the current period. The transfer of real balances in a bilateral match from the buyer to the seller in the DM will be denoted d. (We keep the same notation as the one used for the transfer of nominal money balances in the previous chapter.) The value function of the buyer at the beginning of the CM, W b (z), satisﬁes W b (z) = max x − y + βV b z (6.1) x,y,z

subject to x + φt m = y + z + T

(6.2)

and z = φt+1 m ,

(6.3)

where T corresponds to the real value of the lump-sum transfer from the government, i.e., T = φt (Mt+1 − Mt ) = (γ − 1)φt Mt . The ﬁrst constraint, (6.2), represents the buyer’s budget constraint in the CM and (6.3) describes the real value that m units of money will have in the next period, t + 1. Substituting m = z /φt+1 from (6.3) into (6.2), and then into (6.1), and recalling that φt /φt+1 = γ , the buyer’s value function at the beginning of the CM can be expressed as b −γ z W b (z) = z + T + max z . (6.4) + βV z ≥0

According to (6.4) the lifetime expected utility of a buyer in the CM is the sum of his real balances, the lump-sum transfer from the government, and his continuation value at the beginning of the next DM minus the investment in real balances. Recall that in order to hold z real balances in the next DM, the buyer must obtain γ z real balances this CM. The buyer’s value function at the beginning of the DM, V b (z), is given by V b (z) = σ u q (z) + W b z − d (z) + (1 − σ ) W b (z) = σ u q (z) − d(z) + W b (z) , (6.5) where we use the linearity of W b (z) in going from the ﬁrst equality to the second. According to (6.5) the lifetime expected utility of a buyer at the beginning of the DM is the sum of his expected surplus in the

The Optimum Quantity of Money

131

DM plus his continuation value in the subsequent CM. The trade surplus in the DM is the difference between the utility of consumption and the transfer of real balances. We will consider trading protocols where the terms of trade q, d depend only on the real balances of the buyer. The buyer’s problem can be simpliﬁed by substituting V b (z) from (6.5) into (6.4), i.e., max −iz + σ u q(z) − d(z) , (6.6) z≥0

where 1 + i = (1 + r)γ and i can be interpreted as the nominal rate of interest on an illiquid bond, i.e., the bond cannot be used as a medium of exchange in the DM. If a one-period (illiquid) nominal bond issued in period t − 1 pays one dollar in period t, then the dollar price of the newly issued bonds in period t − 1 is ωt−1 , where ωt−1 solves ωt−1 φt−1 = βφt , so agents are indifferent between holding and not holding the bond. Hence ωt−1 = β(φt /φt−1 ) = β/γ . The nominal interest rate is then i = 1/ωt−1 − 1 = (γ − β)/β = (1 + r)γ , as stated above. The buyer chooses his real balances so as to maximize his expected surplus in the DM minus the cost of holding money balances, where the cost of holding money, i, is a function of the rate of time preference and the inﬂation rate. Because it is costly to hold money, buyers will not hold more money than they intend to spend in a bilateral match in the DM; this implies that d = z. As a benchmark we assume that the terms of trade are determined by a take-it-or-leave-it offer by the buyer. Since i > 0, buyers will not hold more real balances than what is necessary to compensate the seller for the efﬁcient level of output, i.e., z ≤ c q∗ . The quantity traded in a match satisﬁes c(q) = z whenever z ≤ c(q∗ ).Since there is a one-to-one relationship between q and z when z ≤ c q∗ , the buyer’s problem (6.6) can be rewritten as a choice of q, i.e., (6.7) max −ic(q) + σ u(q) − c(q) . q∈[0,q∗ ] The ﬁrst-order (necessary and sufﬁcient) condition to the buyer’s problem (6.7) is u (q) i = 1+ . c (q) σ

(6.8)

This equation is similar to (4.14) in chapter 4, except the rate of time preference, r, has been replaced by the nominal interest rate, i. In chapter 4

132

Chapter 6

the money supply was assumed to be constant, which implies that γ = 1 and, hence, i = r. The cost of holding real balances, i, generates a wedge between the marginal utility of consuming and the marginal cost of producing q that is proportional to the average length of time to complete a trade in the DM, 1/σ . The steady-state solution, qss , to (6.8) is depicted in ﬁgure 6.2. From (6.8) it is clear that the optimal monetary policy requires a zero nominal interest rate, i = 0, or, equivalently, γ = 1/(1 + r) < 1. As a consequence prices contract at a rate that is approximately equal to the rate of time preference. This is the so-called Friedman rule. By reducing the cost of holding real balances to zero, buyers will accumulate sufﬁcient real balances in the previous CM so that they can purchase the quantity, q∗ , that maximizes the gains from trade in the DM. It is also clear from (6.8) and ﬁgure 6.2 that an increase in inﬂation and hence, i, decreases output produced in the DM. In summary, when buyers have all of the bargaining power, the allocation of the monetary equilibrium under the Friedman rule coincides with the socially efﬁcient allocation of the DM good, q = q∗ .

u'(0) c'(0) u'(q) c'(q)

i 1+ σ 1

q ss

q*

Figure 6.2 Stationary monetary equilibrium under a constant money growth rate

q

The Optimum Quantity of Money

133

6.2 Interest on Currency We will now show that a policy that generates a rate of return for currency equal to the rate of time preference, i.e., the Friedman rule, does not need to be implemented by a contraction of the money supply. Instead, the policy maker can pay an interest on currency. This is effectively what happens when the central bank pays interest on reserves. Suppose that an agent holding m units of money at the beginning of the CM receives im m units of money, where the interest on currency is equal to im ≥ 0. The budget constraint of the government is T + im φt Mt = φt (Mt+1 − Mt ).

(6.9)

According to (6.9) the government ﬁnances its lump-sum transfer to buyers, T, and the interest payment on currency by the increase in the money supply. The value of a buyer in the CM is W b (z) = max x − y + βV b z (6.10) x,y,z

subject to x + φt m = y + z (1 + im ) + T

(6.11)

and z = φt+1 m .

(6.12)

From (6.11) the buyer receives a lump-sum transfer, T, and an interest payment on his money balances that he holds at the beginning of the CM, im z. The latter implies that the real value of one unit of money in the DM, measured in terms of the general good, is (1 + im )φt . From (6.10)–(6.12), the buyer’s value function at the beginning of the CM can be expressed as b W b (z) = T + (1 + im )z + max −γ z + βV (z ) , (6.13) z ≥0

where V b (z) = σ u(qt ) − c(qt ) + W b (z), (6.14) and c(qt ) = min (1 + im )z, c(q∗ ) from the buyer-takes-all assumption. From (6.14) the buyer enjoys the whole surplus from a match. The value of being seller in the CM is simply W s (z) = (1 + im )z.

134

Chapter 6

From (6.13) and (6.14) the buyer’s choice of money balances solves max −γ z + σβ u(qt ) − c(qt ) + β(1 + im )z . (6.15) z≥0

We must assume that γ ≥ β(1 + im ); otherwise, the buyer’s problem has no solution. Note that the rate of return of currency is φt+1 (1 + im )/φt = (1 + im )/γ . We focus on stationary equilibria where φt Mt = φt+1 Mt+1 or φt /φt+1 = Mt+1 /Mt = γ . The buyer’s problem, (6.15), can be rearranged as

γ − β(1 + im ) max − z + σ u(qt ) − c(qt ) . (6.16) z≥0 β Notice from (6.16) that the interest on currency reduces the cost of holding money, [γ − β(1 + im )]/β. Since we are assuming an interior solution, the ﬁrst-order condition is u (q) γ − β(1 + im ) = 1+σ . c (q) β(1 + im )

(6.17)

It is clear from (6.17) that in order to achieve the socially efﬁcient allocation the policy maker must choose a combination for γ and im that satisﬁes γ = β(1 + im ).

(6.18)

If im = 0, i.e., is no interest on currency, then γ = β and the money supply must contract at a rate that is approximately equal to the rate of time preference. If for some reason the policy maker wants to avoid a deﬂation, it can set γ = 1. From (6.18) β(1 + im ) = 1 or, equivalently, im = r. Hence the quantity traded will be at the efﬁcient level if the policy maker sets the interest on currency equal to the rate of time preference, r, and maintains a constant money supply (through lump-sum taxes). This implies that the Friedman prescription for optimal monetary policy need not be associated with a contracting money supply. Finally, if the policy maker does not make any lump-sum transfers i.e., T = 0, then the interest on currency must be ﬁnanced by an increase in the money supply. From (6.9) we have im = γ − 1, and from (6.17) we get u (q) = 1 + r. c (q)

The Optimum Quantity of Money

135

If the change in the money supply is engineered through proportional transfers to money holders, i.e., by interest on money holdings, then the quantity traded in a bilateral match is independent of the interest on currency and will be inefﬁciently low due to agents’ impatience. 6.3 Friedman Rule and the First Best When buyers receive the entire surplus from trade, the Friedman rule implements the efﬁcient allocation, q∗ . We want to check the robustness of this result by considering alternative trading protocols for the DM. We will see that the Friedman rule need not implement the efﬁcient allocation for some trading protocols. Let’s ﬁrst consider the generalized Nash bargaining solution. The Nash bargaining solution is appealing because it has strategic foundations, i.e., there are explicit alternating-offer bargaining games that generate the same outcome. The terms of trade, (q, d), are determined by the solution to max[u(q) − d]θ [−c(q) + d]1−θ q,d

subject to d ≤ z,

(6.19)

where θ ∈ [0, 1] measures the buyer’s bargaining power. Note that (6.19) is identical to (4.26) in chapter 4.2.2. The constraint d ≤ z binds in any monetary equilibrium because buyers do not hold more money than they intend to spend. So the solution to (6.19) describes a relationship between q and z, and is given by (1 − θ )c (q)u(q) + θu (q)c(q) z = zθ q ≡ = (q)c(q) + 1 − (q) u(q), (1 − θ )c (q) + θ u (q) (6.20)

where (q) =

θu (q) θ u (q) + (1 − θ )c (q)

.

The deﬁnition of the transfer of real balances, to (4.28) ∗ (6.20), is identical ∗ in the chapter 4.2.2, provided that z ≤ θ c q + (1 − θ ) u q . Since there is a one-to-one relationship between q and z, the buyer’s choice of real balances can be rewritten as a choice of q:

136

Chapter 6

max −izθ (q) + σ u(q) − zθ (q) . q∈[0,q∗ ]

(6.21)

Note that (6.21) generalizes (4.31), which assumes a constant money supply. From the solution to the bargaining problem, q is no greater than q∗ . At the Friedman rule, i = 0 and the buyer chooses q to maximize u(q) − zθ (q). We established in chapter 4.2.2 that his surplus, u q∗ − zθ q∗ < 0 whenever θ < 1, which implies that the buyer’s surplus is decreasing in q, when q is close to q∗ . Therefore, to maximize his surplus, a buyer will choose, from a social perspective, an inefﬁciently low value for q when the cost of holding real balances is zero. This inefﬁciency is due to a nonmonotonicity property of the Nash solution, according to which the buyer’s surplus can fall even if the match surplus increases. The buyer receives his maximum surplus at a value of q < q∗ when θ < 1, as illustrated by the middle curve in ﬁgure 6.3, and maximizes his surplus at q = q∗ when θ = 1, as illustrated in the top curve. Despite the real balances being too low when θ < 1, the optimal monetary policy is Buyers-take-all

u(q)−c(q) Buyer’s surplus

Generalized Nash bargaining

Θ(q) [u(q)−c(q)]

Proportional bargaining

θ[u(q)−c(q)]

q* Figure 6.3 Buyer’s surplus under alternative bargaining solutions

q

The Optimum Quantity of Money

137

still the Friedman rule. If, for example, i > 0, then buyers will choose an even lower amount of real balances, which implies an even lower social welfare. The nonmonotonicity property of the Nash solution is crucial for the inability of the Friedman rule to generate the efﬁcient allocation. But it is not a generic property of all bargaining solutions. To see this, consider the proportional solution where the buyer gets a constant share θ of the match surplus. With proportional bargaining we have zθ (q) = θ c(q) + (1 − θ )u(q). The buyer’s choice of real balances under proportional bargaining is given by the solution to max −izθ (q) + σ θ u(q) − c(q) , (6.22) q∈[0,q∗ ] which generalizes (4.41) in chapter 4.2.3. It is obvious that as i tends to 0, q approaches q∗ . So, although buyers do not have all the bargaining power, the fact that the buyer’s surplus is increasing in the total match surplus implies that both of these surpluses are maximized at q = q∗ . See the bottom curve in ﬁgure 6.3. Under proportional bargaining, the Friedman rule is optimal and guarantees that the efﬁcient allocation, q∗ , will prevail. Finally, if the terms of trade are determined by either a Walrasian pricing protocol or a competitive posting protocol in the DM, then, as we demonstrated in chapters 4.3 and 4.4, q is given by the solution to (6.8). Under both of these protocols, the buyer is able to extract the entire marginal contribution of his real balances to the match surplus. As a consequence the Friedman rule implements the efﬁcient allocation, q∗ . To summarize the results so far, while the Friedman rule is the optimal monetary policy under many trading protocols, it does not always achieve the efﬁcient allocation. If the buyer obtains the marginal social return of his real balances, as is the case under buyers-take-all, competitive price posting or Walrasian price taking, then the Friedman rule implements the efﬁcient allocation. And even if this condition does not hold, the Friedman rule can achieve the socially efﬁcient allocation provided that the buyer’s surplus from a trade increases with the total surplus of a match.

138

Chapter 6

6.4 Necessity of the Friedman Rule For each of the trading protocols that we have considered—buyerstake-all, Nash and proportional bargaining, Walrasian price taking, competitive price posting—it is necessary to run the Friedman rule in order to implement the efﬁcient allocation. The intuition that underlies this result is clear: if the rate of return on money is less than the discount rate, then it is costly to hold real balances, and agents economize on real balance holdings. This results in too little trade in the DM. In this section we show that the Friedman rule, although optimal, is, in fact not necessary to achieve the efﬁcient allocation. We do this by constructing a trading protocol that implements the efﬁcient allocation when it is costly to hold money. We would like to emphasize that the trading protocol is in no way “odd’’ and, if anything, possesses a couple of good properties. In particular, the trading protocol is characterized by individual rationality and pairwise Pareto efﬁciency. The former means that buyers and sellers voluntarily subject themselves to the protocol, and the latter that there is no room for a joint defection by a pair of matched agents. Consider a trading protocol that proposes the following terms of trade: for all z ≥ c(q∗ ), q(z) = q∗ , d(z) = c(q∗ ), and for all z < c(q∗ ), q(z) = u−1 (z) < q∗ , d(z) = z. This trading protocol is illustrate in ﬁgure 6.4. The bargaining set and the utility levels assigned to the buyer and the seller by the mechanism, ub = u(q) − d and us = d − c(q), respectively, are functions of the buyer’s real balances, z. If the buyer brings enough real balances to compensate the seller for producing q∗ , then the trading protocol prescribes the efﬁcient output, q∗ , and gives the entire match surplus to the buyer. This case is depicted in the right panel of ﬁgure 6.4, where the Pareto frontier of the bargaining set has a linear portion and the payoffs assigned by the mechanism are at the intersection of the Pareto frontier and the horizontal axis. If, on the contrary, the buyer

z < c(q*)

u b + u s = u(q*) − c(q*)

ub

us

Figure 6.4 Choice of utility levels as a function of the buyer’s real balances

us

z ≥ c(q*)

: Utility levels selected by the mechanism

ub

The Optimum Quantity of Money 139

140

Chapter 6

brings less than c(q∗ ) real balances into a match, then the seller receives the entire match surplus and the buyer receives nothing, u(q) − z = 0. This case is illustrated in the left panel of ﬁgure 6.4, where the Pareto frontier of the bargaining set is strictly concave and the agents’ payoffs are at the intersection of the Pareto frontier and the vertical axis. The statement of the buyer’s problem in the CM is similar to (6.6), except he only receives a positive surplus, equal to u q∗ − ∗ that now c q , if z ≥ c(q∗ ). Hence the buyer’s problem is given by max −iz + σ u q∗ − c q∗ I{z≥c(q∗ )} , z≥0

(6.23)

where I{z≥c(q∗ )} is an indicator function that is equal to one if z ≥ c(q∗ ), and zero otherwise. The buyer’s surplus, as a function of his real balances, is illustrated in the top panel of ﬁgure 6.5. It is a step function. The buyer’s net payoff from holding real balances is illustrated in ∗ the ∗ bottom ∗ panel of ﬁgure 6.5, where it is assumed that −ic q + σ u q − c q > 0. The solution to (6.23) is ⎧ ∗ if − ic(q∗ ) + σ u q∗ − c q∗ > 0, ⎨ = c(q ) z ∈ {0, c(q∗ )} if − ic(q∗ ) + σ u q∗ − c q∗ = 0, ⎩ =0 otherwise. For this trading protocol, when a monetary equilibrium exists, inﬂation is neutral in the sense that a (small) increase or decrease in inﬂation has no effect on the allocation. In this case money is said to be superneutral. A monetary equilibrium, however, need not always exist. Fiat money ∗ will ∗ ∗ only be valued for this trading protocol if −ic(q ) + σ u q − c q ≥ 0. As a result, if a monetary equilibrium exists, then the money growth rate, γ , must be lower than some threshold, γ¯ , where σ u q∗ − c(q∗ ) γ ≤ γ¯ ≡ β + . (1 + r)c(q∗ )

(6.24)

It is, of course, possible to implement the ﬁrst-best allocation for a nonnegative rate of inﬂation. From (6.24) a necessary and sufﬁcient condition for γ¯ ≥ 1 is c(q∗ ) ≤

σ u q∗ . r+σ

(6.25)

If condition (6.25) is satisﬁed, then q∗ can be implemented under a policy of price stability. Note that this condition is the same necessary condition

The Optimum Quantity of Money

141

u[ q ( z)] − d ( z)

u ( q*) − c ( q*)

− iz + σ {u[ q( z)] − d (z)}

c (q *)

z

− ic(q*) + σ [u(q*) − c(q*)]

c (q *)

− ic (q*)

Figure 6.5 Necessity of the Friedman rule: Buyers’ payoffs

z

142

Chapter 6

we obtained for implementing the efﬁcient allocation in the model with indivisible money with no currency shortage; see (3.29) in chapter 3.2. Perhaps this result should not be surprising. Because our trading protocol replicates the same discreteness for the buyer’s payoff as indivisible money, i.e., holding c(q∗ ) of real balances is like holding an indivisible unit of money. If γ > γ¯ , then one could design an alternative mechanism that would maximize social welfare, subject to incentive-feasibility constraints. In this situation incentive-feasible trades could take place but not at the ﬁrst-best level. 6.5 Feasibility of the Friedman Rule We have assumed that the government has enough coercive power in the CM to force buyers to pay the lump-sum tax required to implement a deﬂation consistent with the Friedman rule. In this section we weaken the enforcement power of the government. We assume that the government has the ability to collect taxes from buyers in the form of money balances at the end of the CM but cannot force buyers to produce or accumulate money balances. Consequently a buyer can avoid paying the lump-sum tax by simply not producing in the CM and, hence, not accumulating money balances. If the buyer does not have enough money balances to pay all his taxes, but has some money balances, the government conﬁscates everything the buyer has. In this environment taxes will only be collected from buyers. Since sellers have no incentive to accumulate money, they will never leave the CM with money balances and hence cannot be taxed. If a buyer chooses to hold real balances at the end of the CM, then he will accumulate the optimal money balances deﬁned by problem (6.6) in addition to the lump-sum tax. The reason is straightforward: if the buyer holds some money at the end of the CM but not enough to pay the tax, the government will conﬁscate all his money. So, if the buyer is going to pay the tax, he might as well accumulate the optimal amount of real balances for the subsequent DM. A buyer will be willing to pay the lump-sum tax if W b (z) ≥ z + βV b (0),

(6.26)

where the right side says that the buyer consumes his real balances, z, in the CM and exits with no money balances. We assume that agents

The Optimum Quantity of Money

143

do not accumulate tax liabilities across periods, (e.g., the government has no memory). Nevertheless, if it is optimal for the buyer not to pay his taxes in, say, period t, then it is never optimal for him to pay any (current period) tax liabilities in future periods. From (6.4), (6.26) can be re-expressed as T + max −γ z + βV b (z) ≥ βV b (0) = βW b (0), z≥0

where T < 0 when there is a contraction of the money supply. Using (6.5), we can rewrite this inequality as T + max −γ z + βσ u[q(z)] − z + βz ≥ 0, z≥0

(6.27)

i.e., the expected surplus from trade in the DM, net of the cost of holding money, must be greater than the lump-sum taxes collected by the government. The lump-sum transfer in the CM of period t is T = (γ − 1)φt Mt = (γ − 1)Z, where Z represents aggregate real balances. Let’s assume that buyers make take-it-or-leave-it offers in bilateral matches in the DM. Then z = c(q) and from (6.27), q solves i ≡ (γ − β)/β = σ [u (q)/c (q)] − 1 . In equilibrium z = Z = c(q), and (6.27) can be expressed as −(1 − β)c(q) + βσ u(q) − c(q) ≥ 0. Dividing by β, and rearranging terms, the inequality above holds if and only if c(q) ≤

σ u(q), r+σ

(6.28)

The policy that consists in setting i equal to zero is incentive feasible if c(q∗ ) ≤

σ u(q∗ ). r+σ

(6.29)

Clearly, if r is sufﬁciently small, then the Friedman rule will be incentive feasible; that is, buyers will be willing to pay the tax that is required to generate the optimal deﬂation. It is important to point out that condition (6.29) coincides with the condition under which q∗ can be implemented under a constant money supply if the trading protocol is chosen optimally, (6.25). This ﬁnding suggests that the Friedman rule is not an essential policy in environments with bilateral trades. Whenever the ﬁrst best can be achieved

144

Chapter 6

under the Friedman rule, it can also be achieved by a constant money supply, provided that the trading protocol is designed appropriately. If (6.29) is violated, then the Friedman rule is not incentive feasible, and there is a lower bound γ ∈ (β, 1) for the incentive-feasible money growth rate. Notice that this lower bound is less than one, meaning that the optimal feasible policy is characterized by deﬂation. 6.6 Trading Frictions and the Friedman Rule Although the Friedman rule is the optimal policy in many monetary environments, it is rarely implemented in practice. We have provided a couple of reasons for this. First, the government may lack the enforcement power required to implement the lump-sum tax needed to generate a deﬂation. Second, the Friedman rule may not be needed if the trading protocol that determines the terms of trade in the DM is appropriately designed. In this section we describe an environment where the government has sufﬁcient enforcement power to implement the Friedman rule, and the terms of trade are determined by a standard bargaining solution. However, the government may choose not to implement the Friedman rule—even though it is feasible—because it may be suboptimal. The novelty in this section is that DM search frictions are endogenously determined. We slightly amend our benchmark model to endogenize the composition of buyers and sellers in the DM. We assume that there is a unit measure of ex ante identical agents that can choose to be either buyers or sellers in the DM. The decision to become a buyer or seller in period t is taken at the beginning of the previous CM, in period t − 1. Suppose, for example, that at the beginning of the CM, individuals invest in a (costless) technology that allows them to either produce DM goods or consume them, and it is only possible to invest in one technology. One can think of the DM good as being an intermediate good, where sellers produce the intermediate good and buyers produce a ﬁnal good that requires the intermediate good as an input. The ﬁnal good is produced after the buyer and seller split apart. Therefore the ﬁnal good cannot be consumed by both the buyer and seller. We assume that the government has coercive power in the CM to tax individuals. However, since it cannot observe agents’ histories in the DM, the government cannot tax buyers and sellers at different rates. Figure 6.6 illustrates the timing of events for a typical period.

The Optimum Quantity of Money

DAY (DM)

n buyers and 1 − n sellers are matched bilaterally and at random

145

NIGHT (CM)

Choice of being buyers or sellers in the next day

Choice of real balances

Figure 6.6 Timing of the representative period

Let n denote the fraction of buyers in the DM and 1 − n the fraction of sellers. The technology that matches buyers and sellers is the following: A buyer meets a seller with probability 1 − n, the fraction of sellers in the population, and a seller meets a buyer with probability n, the fraction of buyers in the population. Therefore the number of matches in the DM is n(1 − n), and it is maximized when n = 1/2. As before, W b (W s ) denotes the value function of an agent in the CM who chooses to be a buyer (seller) the next DM, and V b (V s ) denotes the value function for a buyer (seller) in the DM. The value function at the beginning of the CM is analogous to (6.4), and satisﬁes j W j (z) = T + z + max + βV (z ) , (6.30) −γ z z ≥0

where j ∈ {b, s}. Since buyers spend all their money holdings in the DM if they are matched, the value of being a buyer in the DM satisﬁes V b (z) = (1 − n) u q(z) − z + max W b (z), W s (z) . (6.31) We substitute (6.31) into (6.30), and use the linearity of W b (z) and W s (z), to get the value of a buyer with z units of real balances at the beginning of the CM that must satisfy W b (z) = T + z + max β −iz(q) + (1 − n)[u(q) − z(q)] ∗ q∈[0,q ] + β max W b (0), W s (0) .

(6.32)

From (6.32) the buyer chooses the quantity to trade in the next DM, taking as given his matching probability, 1 − n. By similar reasoning, the value of being a seller with z units of real balances satisﬁes

146

Chapter 6

W s (z) = T + z + βn[z(q) − c(q)] + β max W b (0), W s (0) .

(6.33)

Equation (6.33) embodies the result that sellers do not carry money balances into the DM—since they do not need them—and that the quantity traded q , or equivalently the buyers’ real balances, is taken as given. Since both W b (z) and W s (z) are linear in z, the choice of being a buyer or a seller does not depend on z. In a monetary equilibrium, agents must be indifferent between being a seller or a buyer; otherwise, there will be no trade, and ﬁat money will not be valued. Consequently we focus on monetary equilibria where n ∈ (0, 1) and W b (z) = W s (z). From (6.32) and (6.33), n must satisfy n[z(q) − c(q)] = (1 − n) u(q) − z(q) − iz(q). (6.34) The left side of (6.34) is the seller’s expected surplus in the DM, whereas the right side is the buyer’s expected surplus, minus the cost of holding real balances. Hence in any monetary equilibrium n=

u(q) − (1 + i)z(q) . u(q) − c(q)

(6.35)

Note that for given q, an increase in i reduces the measure of buyers. Intuitively, higher inﬂation increases the cost of holding real balances and, hence, reduces the incentives to be a buyer in the DM. From (6.32), q solves (6.36) max −iz(q) + (1 − n)[u(q) − z(q)] . ∗ q∈[0,q ] A steady-state monetary equilibrium is a pair (q, n) such that q > 0 is a solution to (6.36) and n ∈ (0, (6.35). Suppose that z(q) is 1)∗satisﬁes strictly increasing with q for q ∈ 0, q , and the buyer’s objective function in (6.36) is strictly concave and twice continuously differentiable. Then the equilibrium is unique at the Friedman rule: q solves the ﬁrst-order condition from (6.36), u (q) − z (q) = 0, and, given q, the measure of sellers is uniquely determined by (6.35). Assuming the solution for n is interior, the effects of a change in i in the neighborhood of i = 0 are given by

The Optimum Quantity of Money

dq di dn di

= i=0

147

z (q) , (1 − n)[u (q) − z (q)]

= − [u(q) − c(q)]−1

i=0

n 1−n

(6.37)

z (q)[u (q) − c (q)] + z(q) , u (q) − z (q)

(6.38)

where n and q are evaluated at i = 0. Inﬂation has a direct effect on the equilibrium allocation by raising the cost of holding real balances and therefore by reducing q. The effect of inﬂation on the measure of buyers, n, is, in general, ambiguous. If, however, the pricing mechanism delivers q = q∗ under the Friedman rule, then n decreases with inﬂation since dn di

= i=0

−z(q∗ ) < 0. u(q∗ ) − c(q∗ )

The intuition here is straightforward: since inﬂation is a direct tax on agents who hold money, as inﬂation increases agents have less incentives to be buyers. As a result the matching probability of buyers, 1 − n, increases with inﬂation (close to the Friedman rule). Because there are fewer buyers, they spend their money balances in the DM faster. This is the so-called hot potato effect of inﬂation. We measure social welfare by the sum of all trade surpluses in a period, i.e., W = n(1 − n)[u(q) − c(q)]. Equivalently, we could divide by 1 − β to consider the discounted sum of those surpluses. Welfare is maximized when the surplus of each match is maximized—which requires q = q∗ —and when the number of matches in the DM is maximized—which requires n = 1/2. Suppose that the trading protocol in the DM implements q∗ at the Friedman rule. This would be the case, for example, under proportional bargaining. The ﬁrst condition for efﬁciency, q = q∗ , requires that the Friedman rule be implemented. The second condition for efﬁciency, n = 1/2, requires, from (6.35), that u q∗ − z q∗ 1 = . ∗ ∗ 2 u q −c q

(6.39)

Note that the left side of (6.39) represents the buyer’s share of the match surplus.

148

Chapter 6

Equation (6.39) turns out to be a restatement of the so-called Hosios condition for efﬁciency in models with search externalities. Search externalities arise when agents’ decisions to participate in a market affect the trading probabilities of other agents in the market. These search externalities are internalized when the elasticity of the matching function with respect to the measure of buyers is equal to the buyer’s share in the match surplus. In other words, the buyer’s contribution in the creation of matches in the DM must be rewarded by giving buyers a share in the match surplus that is equal to the fraction of matches that buyers are responsible for. The number of matches, i.e., the matching function, is = bs/(b + s), where b is the measure of buyers and s is the measure of sellers. Hence the Hosios condition requires that d/ s u(q∗ ) − z(q∗ ) = = 1−n = . db/b b+s u(q∗ ) − c(q∗ )

(6.40)

But, from (6.35), the right side of (6.40) is equal to n, and n = 1 − n means that n = 1/2. The welfare effect of a change in i in the neighborhood of i = 0 can be evaluated by totally differentiating the social welfare function and using (6.37) and (6.38), i.e., dW di

u q [u q − c (q)] n2 + (2n − 1) z(q). = [u (q) − z (q)] 1 − n i=0

(6.41)

Assuming that q = q∗ at the Friedman rule—which is valid under proportional bargaining—we can evaluate the welfare implications of a deviation from the Friedman rule by evaluating the second term in (6.41), (2n − 1) z(q). Adeviation will be optimal, i.e., dW/di|i=0 > 0, if and only if n > 1/2. From (6.35), this occurs when the buyer’s share of the surplus is greater than one-half. When the buyer’s share of match surplus is greater than one-half, there are too many buyers from a social perspective. Under proportional bargaining a deviation from the Friedman rule will be optimal whenever θ ∈ (0.5, 1). In this case the policy maker is willing to trade off efﬁciency on the intensive margin—the quantities traded in each match—in order to improve the extensive margin—the number of trade matches in the DM—by raising inﬂation. An increase in inﬂation will increase the number of sellers and decrease the number of buyers.

The Optimum Quantity of Money

149

If instead θ < 1/2, then, at the Friedman rule, there are too many sellers in equilibrium. In this situation a small deviation from the Friedman rule reduces welfare, since this would only further increase the number of sellers in the economy. If pricing in the DM is given by proportional bargaining, then, using (6.35), the measure of buyers is θc(q) + (1 − θ )u(q) n = θ −i . u(q) − c(q) This means that for all i > 0, n < θ. Recall that the total number of trades, n (1 − n), is increasing in n for all n < 1/2. Consequently, if θ < 1/2, then n < θ < 1/2, and the total number of trades is less than what it would be at the Friedman rule, θ (1 − θ ). In this case a deviation from the Friedman rule reduces both the number of trades and the quantity traded in each match. So it is unambiguous that the Friedman rule be optimal, even though it fails to achieve a constrained efﬁcient allocation. It should be pointed out that these sorts of welfare results depend critically on the DM trading protocol. For example, it can be shown that under a competitive search pricing protocol, the Hosios condition emerges endogenously, and as a consequence the search externalities are internalized, i.e., the extensive margin is efﬁcient. Therefore, since the competitive search pricing protocol results in an efﬁcient intensive margin under the Friedman rule, a Friedman rule policy can implement an efﬁcient allocation. The envelope-type argument used above is only valid if the Friedman rule achieves an efﬁcient intensive margin outcome, i.e., if q = q∗ . The argument would not be valid for the generalized Nash bargaining protocol since q < q∗ . When q < q∗ , the ﬁrst term on the right side of (6.41) is not equal to zero, and as a result one cannot evaluate the welfare implications of a departure from the Friedman rule by simply examining the value of n. One can, however, use numerical examples to establish that a deviation from the Friedman rule under the generalized Nash bargaining protocol can be optimal when the buyer’s bargaining power is sufﬁciently high. Hence the result that a deviation from the Friedman rule can be optimal is robust across different bargaining solutions. At this point it would be natural to ask if there are other policy instruments that could be used to correct the extensive margin when n = 1/2 without distorting the intensive margin. If the policy maker could tax buyers and sellers differently, it would not need to resort to inﬂation to affect agents’ incentives to participate in the market. However, because

150

Chapter 6

agents’ trading roles in the DM are private information, the inﬂation tax seems to be a natural policy instrument to reduce agents’ incentives to be buyers. 6.7 Distributional Effects of Monetary Policy An inﬂationary monetary policy can be desirable when the distribution of money balances across agents is not degenerate. Indeed a positive inﬂation, engineered by lump-sum money injections, redistributes wealth from the richest to the poorest agents in the economy. If some agents are poor because of uninsurable idiosyncratic shocks, then this redistribution can raise social welfare. Money injections do not have a distributional effect in our benchmark model because, by construction, the distribution of money holdings across buyers at the end of the CM or beginning of the DM is degenerate. The assumption of quasi-linear preferences—which eliminates wealth effects—along with the fact that all buyers have access to the CM implies that all buyers will choose the same level of money holdings in the CM under the standard trading protocols we have examined. One can obtain a nondegenerate distribution of money holdings by introducing some heterogeneity across buyers. For example, buyers could differ in terms of their marginal utility of consumption in the DM. Buyers with high marginal utilities of consumption would want to consume more than buyers with low marginal utilities of consumption and as a result would hold larger real balances. In such an environment, however, the Friedman rule is still optimal since each type of buyer holds an amount of real balances that maximize his expected surplus in the DM. To capture a distributional effect of monetary policy, we modify the benchmark model. We suppose that buyers and sellers live for only three subperiods. They are born at the beginning of the night subperiod and die at the end of the following period. Agents can potentially trade three times: in the CM when they are born, in the DM of the next period, and in the CM just before they die; see ﬁgure 6.7. For simplicity we assume that agents do not discount across periods, i.e., r = 0. This implies that the Friedman rule corresponds to a constant money supply or a zero inﬂation rate. The utility function of a buyer is xy + u(q) + xo , where xy ∈ R is the utility of consumption net of the disutility of production in the CM when young, xo is the net utility of consumption in the CM when old, and u(q)

The Optimum Quantity of Money

151

Generation t DAY

NIGHT

Productivity shocks Transfers Competitive markets

DAY

Bilateral trades

NIGHT

DAY

Generation t+1

Figure 6.7 Overlapping generations

is the utility of consumption in the DM. Similarly the utility function of a seller is xy − c(q) + xo . This overlapping generations structure does not alone alter the allocation relative to the inﬁnitely lived agents model. In order to obtain a nondegenerate distribution of money balances across agents, we assume that newly born buyers differ in terms of their productivity in the ﬁrst period of their lives. A fraction ρ ∈ (0, 1) of newly born buyers are productive, while the remaining fraction is unproductive. As a result newly born productive buyers can participate in the CM to accumulate money balances, while unproductive ones cannot. Under a constant money supply policy, unproductive buyers do not consume in the DM because they have no money and cannot commit to repay their debt. Moreover the productivity shocks to newly born buyers are private information, and as a result the government is unable to make differentiated transfers to productive and unproductive buyers. The problem of a productive newly born buyer, which is similar to (6.6), is max −φt m + σ u[q(φt+1 m)] − c[q(φt+1 m)] + φt+1 m . (6.42) m≥0

The productive buyer produces φt m units of the general good in exchange for m units of money in the CM when he is born. If he doesn’t meet a seller in the subsequent DM, then he spends his money balances in the CM before he dies; if he does meet a seller, we assume that the buyer captures the entire surplus from the match. Denote z = φt+1 m as the choice of real balances for a productive buyer born in period t for the subsequent DM. The productive buyer’s problem (6.42) can be simpliﬁed to read

152

Chapter 6

max −(γ − 1)z + σ u[q(z)] − c[q(z)] . z≥0

(6.43)

The ﬁrst-order condition for this problem is u (q) γ −1 = 1+ . c (q) σ

(6.44)

Therefore, if the money supply is constant, i.e., γ = 1, newly born productive buyers consume q∗ units of the DM good if they are matched. However, unproductive newly born buyers do not consume in the DM since they cannot produce in the CM when they are born. Assume now that there is a constant positive inﬂation, γ > 1, and that money is injected into the economy through lump-sum transfers to all newly born buyers in the CM. Let t denote a transfer at night in period t − 1 that can be used in the DM of period t. We have t = Mt − Mt−1 =

γ −1 Mt . γ

(6.45)

Let mt represent the money balances of a buyer in the DM of period t who had access to the CM when he was young. Equilibrium in the money market requires that ρmt + (1 − ρ)t = Mt .

(6.46)

The fraction ρ of productive buyers hold mt units of money while the 1 − ρ unproductive buyers hold t . The sum of the individual money holdings must add up to the money supply, Mt . Substituting t from (6.45) into (6.46) and rearranging, we get

Mt 1 + ρ(γ − 1) mt = , (6.47) ρ γ and from (6.45) and (6.47) we get t =

ρ(γ − 1) mt . 1 + ρ(γ − 1)

(6.48)

Equation (6.48) implies that t < mt : unproductive buyers are poorer than productive ones. Let q˜ denote the DM consumption of unproductive buyers. Unproductive buyers will spend all of their money balances in the DM because t < mt , and productive buyers spend all of their balances. From the

The Optimum Quantity of Money

153

buyer-takes-all bargaining assumption, c(qt ) = φt mt and c(˜qt ) = φt t . Hence (6.48) implies that c(˜qt ) =

ρ(γ − 1) c(qt ). 1 + ρ(γ − 1)

(6.49)

From (6.49), q˜ t < qt . As γ increases, qt decreases through a standard inﬂation-tax effect; see (6.44). But inﬂation also affects the distribution of real balances across buyers. Indeed the dispersion of real balances, as measured by [c(qt ) − c(˜qt )]/c(qt ) = 1/[1 + ρ(γ − 1)], decreases as γ increases. The policy maker therefore faces a trade-off between smoothing consumption across buyers and preserving the purchasing power of real balances. We treat buyers and sellers from all the different generations symmetrically when we measure social welfare. In this case the allocations of the general good are irrelevant, and welfare can be measured by the sum of all surpluses across matches, W = σρ[u(q) − c(q)] + σ (1 − ρ)[u(˜q) − c(˜q)].

(6.50)

In the neighborhood of price stability, an increase in inﬂation only has a second-order effect on the match surpluses of productive buyers, d[u(q) − c(q)]/dγ γ =1+ = 0. However, it has a ﬁrst-order effect on the match surpluses of unproductive buyers. Differentiating (6.49) with respect to γ , we get d˜qt dγ

γ =1+

=

ρc(q∗ ) . c (0)

The welfare effect of an increase in inﬂation from price stability is, from (6.50), given by dW dγ

γ =1+

u (0) = σ (1 − ρ) − 1 ρc(q∗ ) > 0, c (0)

since u (0) /c (0) = ∞. Hence an increase in inﬂation from γ = 1 is welfare improving because it allows unproductive buyers to consume, while the negative effect on productive buyers’ welfare is only of second-order consequence.

154

Chapter 6

6.8 The Welfare Cost of Inﬂation Under most of the trading protocols examined thus far—bargaining, price taking, and price posting—inﬂation distorts allocations by inducing agents to reduce their real balances, and hence the quantities they trade in the DM. Qualitatively speaking, inﬂation typically reduces social welfare. The next step is to quantify this effect in order to determine whether the costs associated with inﬂation are large or small. If the costs associated with a moderate level of inﬂation are very small, then inﬂation will not be an important policy concern. A typical calibration procedure adopts a representative-agent version of the model studied so far. The CM utility function takes the form B ln x − h, where x is consumption, h is the hours of work, and h hours produces h units of the general good. With the linear speciﬁcation used so far, CM output would be indeterminate. In contrast, with the quasi-linear preferences, production in the CM maximizes B ln x − h, so x = B. One can interpret B as the quantity of goods that do not require money to be traded. The functional forms for utility in the DM are u(q)= q1−η /(1 − η) and c(q) = q. The parameters (η, B) are chosen to ﬁt money demand, as described in the model, to the data. The cost of holding real balances, i, is measured by the commercial paper rate and M is measured by M1, which is cash plus demand deposits. The model has been calibrated over long time periods, such as 1900 to 2000. The typical measure of the cost of inﬂation is the fraction of total consumption—in both the CM and DM—that agents would be willing to give up to have zero inﬂation instead of 10 percent inﬂation. The results of existing studies are summarized in table 6.1. Under the buyer-take-all bargaining solution, the welfare cost of 10 percent inﬂation is typically between 1 percent and 1.5 percent of GDP per year. One ﬁnds a similar magnitude for the welfare cost of inﬂation under Walrasian price taking or competitive posting, i.e., competitive search equilibrium. This is a sizable number. Graphically this number is approximately equal to the area underneath the money demand curve. To see this, integrate the inverse (individual) function, which is given by i(z) = money demand σ u q(z) /c q(z) − 1 ; see (6.8). This will obtain

z1

z1.1

i(z)dz = σ u q(z1 ) − c q(z1 ) − σ u q(z1.1 ) − c q(z1.1 ) ,

The Optimum Quantity of Money

155

Table 6.1 Summary of studies on the cost of inﬂation Trading mechanism

Cost of inﬂation (% of GDP)

Buyers-take-all Nash solution Generalized Nash Egalitarian Price posting (private info) Price taking General Nash with external margin Proportional with external margin Comp. search with external margin

1.2–1.4 3.2–3.3 Up to 5.2 3.2 6.1–7.2 1–1.5 3.2–5.4 0.2–5.5 1.1

where z1 represents real balances when γ = 1 and z1.1 represents real balances when γ = 1.1. The left side of the expression above is the area underneath the individual money demand curve while the right side is the change in society’s welfare. In ﬁgure 6.8 we represent the individual money demand function, i(z). As the nominal interest rate approaches to 0, real balances approach their maximum level, z∗ . Under the buyers-take-all bargaining protocol, z∗ = c(q∗ ). Consider two nominal interest rates, i > 0 and i > i. The welfare cost from raising the nominal interest rate from i to i corresponds to the area, ABDE, underneath money demand curve. The welfare cost from raising the interest rate from the rate associated with the Friedman rule, zero, to i is given by the area ABC. If sellers have some bargaining power, then the welfare cost of inﬂation is larger. Under the (symmetric) Nash solution or the egalitarian solution (i.e., proportional with θ = 0.5), the welfare cost of 10 percent inﬂation is between 3 and 4 percent of GDP. The explanation for this large welfare cost of inﬂation is the following: whenever θ < 1 and i > 0, any bargaining solution generates a holdup problem for money holdings. Buyers incur a cost from investing in real balances in the CM that they cannot fully recover once they are matched in the DM. The severity of this holdup problem depends on the seller’s bargaining power, 1 − θ , and the average cost of holding real balances, i/σ . As inﬂation increases, the holdup problem is more severe, which induces buyers to underinvest in real balances. This argument can be illustrated using the area underneath the money demand function; see ﬁgure 6.9. The inverse (individual) money

156

Chapter 6

i i(z) = σ

{

}

u'[q(z)] −1 z'[q(z)]

B

i'

D

i A

0

z'

E

z

C

z*

z

Figure 6.8 Welfare cost of inﬂation and the area underneath money demand

demand function is i(z) = σ u q(z) /z q(z) − 1 . The area underneath money demand is

z1

z0

i(z)dz = σ u q(z1 ) − z1 − σ u q(z1.1 ) − z1.1 .

Under proportional bargaining, u q(z) − z = θ u q(z) − c q(z) , the area underneath the money demand function is

z1

z0

i(z)dz = θ σ u q(z1 ) − c q(z1 ) − θ σ u q(z1.1 ) − c q(z1.1 ) .

The private loss due to an increase in the inﬂation rate corresponds to left side of the above expression. It is equal to a fraction θ of the welfare loss for society, the right side of the above expression. In ﬁgure 6.9 we represent the individual demand for real balances as well as the social return of those real balances (the dashed curve). The welfare cost from raising the nominal interest rate from 0 to i is given by the area ADC, while the welfare cost to the buyer is the area underneath

The Optimum Quantity of Money

157

i i(z) = σ

{

}

u'[q(z)] −1 z'[q(z)]

σ

D

i

0

{

}

i(z) u'[q(z)] − c'[q(z)] = θ z'[q(z)]

B

A

C

z

z*

z

Figure 6.9 Holdup problem and the cost of inﬂation

money demand, ABC. To see this, notice that the marginal social return of ﬁat money is the increase in society’s welfare arising from a marginal increase in real balances. Since social welfare is σ u q (z) − c q (z) , this is equal to dq σ u q − c q σ u q (z) − c q (z) . = dz z q The private return to the buyer is u q − z q u q − c q i (z) = σ = θσ . z q z q So the individual money demand does not accurately capture the social value of holding money since it ignores the surplus that the seller enjoys when the buyer increases his real balances. If, for example, θ = 1/2, the egalitarian solution, then the social welfare cost of inﬂation is approximately twice the private cost for money holders. This private cost has

158

Chapter 6

been estimated to be about 1.5 percent of GDP, so the total welfare cost of inﬂation for society is then about 3 percent of GDP. The introduction of an endogenous participation decision, as in section 6.6, can either mitigate or exacerbate the cost of inﬂation, depending on agents’ bargaining powers. As we saw earlier, in some instances, the cost of small inﬂation can be negative. 6.9 Further Readings The result that the optimal monetary policy requires the nominal interest rate to be zero or, equivalently, deﬂation equal to the rate of time preference, comes from Friedman (1969). Different deﬁnitions and interpretations of the Friedman rule are discussed in Woodford (1990). The optimal monetary policy in a search model with divisible money was ﬁrst studied by Shi (1997a), who showed that the Friedman rule is optimal when agents’ participation decisions are exogenous. The ability of the Friedman rule to generate an efﬁcient allocation when the terms of trade are determined according to the Nash solution is discussed in Rauch (2000) and Lagos and Wright (2005). Aruoba, Rocheteau, and Waller (2007) prove that an efﬁcient allocation can be obtained even if sellers have some bargaining power, provided that the bargaining solution is monotonic. Lagos (2010) characterizes a large family of monetary policies that are necessary and sufﬁcient to implement zero nominal interest rates. The optimality of the Friedman rule in different monetary models with heterogeneous agents is discussed in Bhattacharya, Haslag, and Martin (2005, 2006) and Haslag and Martin (2007). Berentsen and Monnet (2008) study the conduct of monetary policy through a channel system. Araujo and Camargo (2008) discuss reputational concerns for the monetary authority. The policy of paying interest on reserves has been advocated by Friedman (1960), and studied in overlapping generation economies by Sargent and Wallace (1985), Smith (1991) and Freeman and Haslag (1996). Andolfatto (2010) studies the payment of interest on money in a model similar to the one used in this book. Using a mechanism design approach, Hu, Kennan, and Wallace (2009) show that the Friedman rule is not necessary to obtain good allocations. The incentive feasibility of the Friedman rule when the government has limited coercive power is discussed in Andolfatto (2008), Hu, Kennan, and Wallace (2009), and Sanches and Williamson (2010).

The Optimum Quantity of Money

159

The importance of trading frictions and search externalities for the design of monetary policy was ﬁrst emphasized by Victor Li (1994, 1995, 1997), who established that an inﬂation tax could be welfare enhancing when agents’ search intensities are endogenous. However, his results are subject to the caveat that prices are exogenous. Shi (1997) found a related result in a divisible-money model where prices are endogenous. In Shi’s model each household has a large number of members who can be divided between buyers and sellers. See the appendix of chapter 4 for a presentation of this model. When the composition of buyers and sellers is inefﬁcient, a deviation from the Friedman rule can be welfare improving. Faig (2008), Aruoba, Rocheteau, and Waller (2007), and Rocheteau and Wright (2009) discuss Shi’s ﬁnding under alternative trading mechanisms. Berentsen, Rocheteau, and Shi (2007) establish that the efﬁcient allocation is achieved when both the Hosios rule and the Friedman rule are satisﬁed. A necessary condition for a deviation from the Friedman rule to be optimal is that the Hosios condition be violated. Rocheteau and Wright (2005) study the optimal monetary policy in a model with free entry of sellers under alternative pricing mechanisms. Camera, Reed, and Waller (2003) show that search externalities and hold up problems can arise when specialization is endogenous. Shi (1998) and Shi and Wang (2006) calibrate a model with an endogenous extensive margin to the US time series data to examine the model’s quantitative predictions on aggregate variables and, in particular, on the variability of consumption velocity of money. The ﬁrst attempt to formalize the hot potato effect of inﬂation in a search model of money is in Li (1994, 1995, 1997); in his model, prices are exogenous. Lagos and Rocheteau (2005) show that this effect vanishes in a model with divisible money and endogenous prices. Several attempts to resuscitate this hot potato effect have been provided by Ennis (2009), Nosal (2008), and Liu, Wang, and Wright (2009). The welfare-improving role of a monetary expansion through distributional effects has been studied by Levine (1991), and in a searchtheoretic environment by Deviatov and Wallace (2001), Berentsen, Camera, and Waller (2004, 2005) and Molico (2006). Waller (2009) and Zhu (2008) also show a beneﬁcial role for a positive inﬂation in the context of a search model with overlapping generations and strictly concave preferences. The traditional approach to measuring the cost of inﬂation as the area underneath a money demand curve was developed by Bailey (1956). Lucas (2000) revisited this methodology and provided theoretical

160

Chapter 6

foundations using a general equilibrium model where money is an argument of the utility function. Lagos and Wright (2005) were the ﬁrst to apply this methodology in the context of a model of monetary exchange. Rocheteau and Wright (2009) and Aruoba, Rocheteau, and Waller (2007) evaluate the cost of inﬂation under alternative trading mechanisms and in the presence of an extensive margin. Ennis (2008) considers a model with price posting under private information, Reed and Waller (2006) consider price taking, and Faig and Jerez (2006) study competitive posting. Aruoba, Waller, and Wright (2007) study quantitatively the effects of inﬂation in a search model with capital. Boel and Camera (2010) calibrate a model and estimate the welfare cost of anticipated inﬂation for twenty-three different OECD countries. Gomis-Porqueras and PeraltaAva (2008) and Aruoba and Chugh (2010) study the optimality of the Friedman rule in the presence of distortionary taxes. A review of this literature is provided in Craig and Rocheteau (2008).

7

Information, Monetary Policy, and the Inﬂation–Output Trade-Off

The main ﬁnding that emerged from the research of the 1970s is that anticipated changes in money growth have very different effects from unanticipated changes. Anticipated monetary expansions have inﬂation tax effects and induce an inﬂation premium on nominal interest rates, but they are not associated with the kind of stimulus to employment and production that Hume described. Unanticipated monetary expansions, on the other hand, can stimulate production as, symmetrically, unanticipated contractions can induce depression. —Robert E. Lucas, Monetary Neutrality, Nobel Prize Lecture, 1995.

How does money affect output? This is a classic and largely unresolved question in economics, dating back at least to David Hume. In the monetary economy described in chapter 4, we show that money is neutral: a one-time, anticipated change in the money supply has no real effects, and nominal prices vary proportionally with the stock of money. Money is not, however, superneutral because a change in the rate of growth of money supply, even if anticipated, has real effects by reducing aggregate real balances, real output, and welfare. In this chapter we revisit the relationship between changes in money supply, output, and welfare. In contrast to chapter 6, we assume that changes in the money supply are random and cannot be fully anticipated. Although the stochastic process driving the money supply is known to all, we make different assumptions regarding what agents know about the value of money at the time of trade. We consider the cases where information regarding the value of money is evenly distributed across agents, and where it is not. We show that if all agents are uninformed about the realization of the money supply, then output is constant and uncorrelated with the changes in the money supply. In contrast, if all agents are fully informed,

162

Chapter 7

then output is negatively correlated with inﬂation: agents can trade larger quantities when the value of money is high. We will spend most of this chapter analyzing the case where the buyers—the agents who hold and spend money in the decentralized market—have private information regarding the realization of the current money growth rate. In this case inﬂation and output are positively correlated. A short-run Phillips (1958) curve emerges even though, in equilibrium, information is fully revealed, and prices are fully ﬂexible. Buyers signal the high value of ﬁat money in the low-inﬂation state by hoarding a large fraction of their real balances, thereby reducing their consumption in bilateral trades. As a result the model predicts a positive correlation between the velocity of money and inﬂation. We demonstrate, by means of examples, that if agents assign a positive probability to a high-inﬂation state, then a policy maker can raise aggregate output and welfare by increasing the frequency of this high-inﬂation state. The optimal monetary policy, however, requires the monetary authority to target the money growth rate—to make it deterministic—and to implement a rate of return for ﬁat money that is equal to the rate of time preference, namely the Friedman rule. If the informational asymmetry between buyers and sellers is reversed, meaning sellers have some private information regarding the future value of money, then the positive correlation between inﬂation and output disappears. In this case buyers spend less money in the high-inﬂation state in order to reduce the informational rent captured by informed sellers. Hence the informational structure regarding monetary policy is crucial in order to understand the output effects of unanticipated changes in the money supply. 7.1 Stochastic Money Growth We extend the pure monetary economy described in chapters 4 and 6. Let Mt represent the stock of money at the beginning of period t, and γt ≡ Mt+1 /Mt the gross growth rate of the money supply in period t. Money is injected or withdrawn in a lump-sum fashion in the centralized market, CM. This implies that in period t, the money supply in the decentralized market, DM, is Mt , and in the CM, after the monetary transfers have taken place, is Mt+1 . We assume that agents always know the value of γt at the beginning of the CM and, without loss of generality, that only buyers receive the monetary transfers.

Information, Monetary Policy, Inﬂation–Output Trade-Off

163

The value of ﬁat money in period t, φt , refers to the amount of CM goods that can be purchased by a unit of ﬁat money in period t. The novelty in this chapter is the assumption that the money growth rate, γt , is random. In each period the money growth rate can take one of two values, high, γ¯ , or low, γ < γ¯ , where the probability of a high money growth rate is α, i.e., γ¯ with probability α ∈ (0, 1), γt = with probability 1 − α. γ We focus on stationary equilibria where the real value of the money supply in the CM after the money transfer has taken place is constant over time, φt Mt+1 = φt−1 Mt ≡ Z. Note that if the growth rate of the money supply is constant, then this steady-state condition can be expressed as φt Mt = φt−1 Mt−1 , since Mt+1 = γ Mt for all t, which is the condition we speciﬁed for a constant real money supply in the previous chapter. Conditional on γt = γ , the value of money in the CM is φt = φt−1 /γ , and conditional on γt = γ¯ , the value of money is φt = φt−1 /γ¯ . Hence the gross expected rate of return of money, condi tional on the information available in the CM of t − 1, Et−1 φt /φt−1 , is equal to (1 − α)/γ + α/γ¯ . In the DM of period t, all agents know the current stock of money that is available for trade, Mt , and the value of money that prevailed in the previous period, φt−1 . But in order to determine the terms of trade in the DM of period t, agents need to know the value of money that will prevail in the upcoming CM, φt . All agents will learn the money growth rate, γt , in the CM of period t, but some agents may learn it earlier. In order to formulate the buyer’s problem recursively, we will express the buyer’s money holdings as a fraction of aggregate money balances. The value of a buyer in the CM of period t − 1 after the transfer of money balances has been realized is

m b b m W = max x − y + βV Mt Mt x,y,m subject to x + Z

m m = y+Z , Mt Mt

where φt−1 = Z/Mt . Note that the budget constraint above does not include the lump-sum transfer from the government because the utility of the buyer is measured after the transfer has been realized.

164

Chapter 7

Substituting the budget constraint into the objective function, we obtain

m m m b b m W =Z . (7.1) + max −Z + βV Mt Mt m ≥0 Mt Mt As before, the buyer’s value function is linear, W b (m/Mt ) = Z (m/Mt ) + W b (0), and the buyer’s choice of money balances, m , is independent of the balances he has when he enters the CM, m. The value of the buyer at the beginning of the DM is

m m − dt + (γt − 1)Mt Vb = Et u(qt ) + W b . Mt γt M t Buyers form expectations about the future growth rate—and hence value—of money, the terms of trade in the DM, and about the trading shock, σ , in the DM. Note that for the latter, the buyer must form expectations regarding what agents will know about the future value of money at the beginning of the DM. Note that in the expression for W b , we take into account the lump-sum transfer from the government. Using the linearity of W b , we can rewrite the value function above as m (γt − 1) Z Z b = Et u(qt ) − dt + Et V +m Z + W b (0). Mt γt M t γt M t γt Hence the buyer’s choice of money holdings, given by the second term on the right side of (7.1), can be expressed as

m Z max −Z , + βEt u(qt ) + (m − dt ) m≥0 Mt γt M t or, since φt−1 = Z/Mt ,

φt−1 max −φt−1 m + βEt u(qt ) + (m − dt ) . m≥0 γt

(7.2)

Before we examine the interesting case where different agents know different things about the money growth rate, we ﬁrst consider the case where buyers and sellers are symmetrically informed regarding the money growth rate. 7.1.1 Symmetrically Uninformed Agents Consider the situation where both buyers and sellers learn the realization of the money growth rate, γt , at the beginning of the CM

Information, Monetary Policy, Inﬂation–Output Trade-Off

165

of period t. In this case buyers and sellers will determine terms of trade in the DM based on the expected value of money, φte , where

φte = Et (φt−1 /γt ) = (1 − α)/γ + α/γ¯ φt−1 . If we assume that buyers make take-it-or-leave-it offers to sellers in the DM, which implies that c qt = φte d, then, from (7.2), the buyer’s choice of money holdings in the CM of period t − 1 is the solution to, max −φt−1 m + β σ u(qt ) − c(qt ) + φte m , (7.3) m≥0

where c(qt ) = min c(q∗ ), φte m . In period t − 1, the buyer incurs the cost φt−1 m to accumulate m units of money; in period t, the buyer enjoys the expected surplus from a trade, σ u(qt ) − c(qt ) , and can expect to resell his money holding for φte m units of output in the CM. Problem (7.3) can be rearranged to max −ic(qt ) + σ u(qt ) − c(qt ) , qt ∈[0,q∗ ] where i=

1

β (1 − α)/γ + α/γ¯

−1

is the nominal interest rate. The ﬁrst-order condition for this problem is u (qt ) i = 1+ . c (qt ) σ

(7.4)

There is a unique qt that solves (7.4), andit is independentof time. Given

q, the value of money is determined by (1 − α)/γ + α/γ¯ φt−1 Mt = c(q) and, hence, real balances are constant across time. The level of output traded in the DM may differ from the efﬁcient level because of a wedge between agents’ rate of time reference and ﬁat money’s expected rate of return. If, however, the expected rate of return on money is (1 − α)/γ + α/γ¯ = β −1 , then qt = q∗ . In words, if the rate of return on money is equal to the rate of time preference, then agents will trade the efﬁcient level of output in the DM. This implies that there are many combinations of γ and γ¯ that can implement the Friedman rule. While the DM output depends on the expected rate of return of money, it does not depend on the realization of the money growth rate in the current period. Consequently the model predicts no correlation between inﬂation and output.

166

Chapter 7

7.1.2 Symmetrically Informed Agents Consider now the situation where buyers and sellers learn the period t money growth rate, γt , and hence the value of money, φt , before entering the DM of period t. One can imagine that the monetary authority makes a credible announcement at the beginning of each DM regarding the money growth rate that will prevail in the CM. If the monetary authority announces γt = γ , then a buyer holding mt units of money will ask for qt = qH , where φt−1 ∗ c(qH ) = min mt , c(q ) . γ If, alternatively, it announces γt = γ¯ , then the buyer asks for qt = qL , where φt−1 ∗ c(qL ) = min mt , c(q ) . γ¯ Since γ¯ > γ , it is obvious that qL (mt ) ≤ qH (mt ). Buyers consume more in periods where the money growth rate is low and the value of ﬁat money is high. From (7.2) the buyer’s choice of money holdings in the CM of period t − 1 is given by the the solution to max −φt−1 m + βσ (1 − α) u qH (m) − c qH (m) m≥0

(7.5) + βσ α u qL (m) − c qL (m) + βφte m , where φte = (1 − α)/γ + α/γ¯ φt−1 . This problem differs from (7.3) because now the quantity traded in the DM depends on the information regarding the money growth rate that agents receive before being matched. The ﬁrst-order condition for this problem is

ıˆ u (qL ) u (qH ) 1−α α + = −1 −1 , (7.6) σ c (qH ) γ c (qL ) γ¯ where ıˆ ≡ β −1 − (1 − α)/γ + α/γ¯ . (Notice that i (1 − α)/γ + α/γ¯ = ıˆ.) Equation (7.6) determines a unique value for φt−1 mt = φt−1 Mt (from the clearing of the money market). If (1 − α)/γ + α/γ¯ = β −1 , then i = ıˆ = 0 and qL = qH = q∗ , which means that the Friedman rule achieves the

Information, Monetary Policy, Inﬂation–Output Trade-Off

167

efﬁcient level of DM output in both inﬂation states. The fact that the money growth rate is stochastic does not matter for implementing the efﬁcient level of DM output, provided that the expected rate of return of money is equal to the (gross) discount rate. In summary, when agents are symmetrically informed, there is either no correlation between inﬂation and output, or a negative one, depending on whether agents are imperfectly or perfectly informed, respectively. 7.2 Bargaining under Asymmetric Information We now consider situations where buyers and sellers are asymmetrically informed in the DM regarding the money growth rate that will prevail for that period. At the beginning of the DM of period t, buyers receive a perfectly informative private signal, χ ∈ {L, H}, regarding the value of money or, equivalently, the money growth rate for period t. If χ = L, then buyers learn that the value of money will be low, φt = φL = φt−1 /γ¯ ; if χ = H, then buyers learn that the value of money will be high, φt = φH = φt−1 /γ . A buyer will be called an H-type buyer if he receives the signal H and an L-type buyer if he receives the signal L. Although sellers do not receive any informative signals in the DM, they understand the stochastic process that drives the money supply and will learn the actual money growth rate at the beginning of the CM. The relevant timing of events for a typical period is illustrated in ﬁgure 7.1. Consider a match between a buyer holding m units of money and a seller holding no money in the DM of period t. Assume that the buyer’s money holdings are common knowledge in the match. This simpliﬁes the presentation since agents have no incentives to misrepresent their money holdings. The bargaining game between a buyer and a seller in DAY (DM)

Buyers receive a private signal about γ t

σ bilateral matches are formed

NIGHT (CM)

Competitive markets open. Money growth rate, γ t, is realized

Figure 7.1 Timing of a representative period, t, under asymmetric information

168

Chapter 7

w Lo lat inf ion

Hig hi nfl ati on

N

B

B

Offer

Offer

S Yes

S No

Yes

No

Figure 7.2 Game tree of the bargaining game in the DM

the DM has the structure of a signaling game. This game is illustrated in ﬁgure 7.2, where the label N represents the player Nature who chooses the money growth rate—or, equivalently, the value of money—the label B represents the buyer, and the label S represents the seller. A strategy for the buyer speciﬁes an offer (q, d) ∈ R+ × [0, m], where q is the output produced by the seller in the DM, d is the transfer of money from the buyer to the seller. A strategy for the seller is an acceptance rule that speciﬁes the set A ⊆ R+ × [0, m] of acceptable offers. The buyer’s payoff is u(q) − φd IA (q, d), where IA (q, d) is an indicator function that is equal to one if (q, d) ∈ A and zero otherwise. If an offer is accepted, then the buyer enjoys the utility of consumption, u(q), net of the utility he forgoes by transferring d units of money to the seller, −φd. The seller’s payoff is −c(q) + φd. The seller uses the information conveyed by the buyer’s offer (q, d) to update his prior belief regarding the value of money in the subsequent CM. Let λ(q, d) ∈ [0, 1] represent the updated belief of a seller that the value of money is high, φ = φH . If (q, d) corresponds to an equilibrium offer, then the updated belief λ(q, d) is derived from the seller’s prior belief according to Bayes’s rule. If (q, d) is an out-of-equilibrium offer, then λ(q, d) is, to some extent, arbitrary, as will be discussed below. Given his updated—or posterior—belief, the seller optimally chooses to accept or reject offers. For a given belief system, λ, the set of acceptable

Information, Monetary Policy, Inﬂation–Output Trade-Off

169

offers for a seller, A(λ), is given by A(λ) = (q, d) ∈ R+ × [0, m] : −c(q) + λ(q, d)φH + 1 − λ(q, d) φL d ≥ 0 . (7.7) If offer q, d is acceptable, then the seller’s cost of production, c(q), must be no greater than the expected value of the transfer of money that he receives. The buyer will choose an offer that maximizes his surplus, taking as given the acceptance rule of the seller. The buyer’s bargaining problem is given by max u(q) − φd IA (q, d), (7.8)

q,d≤m

where the value of money is φ ∈ {φL , φH }. A seller’s belief following an out-of-equilibrium offer is somewhat arbitrary. To get sharper predictions, we require that the equilibrium satisﬁes the intuitive criterion. Denote Uχb as the surplus that a χ type buyer, χ ∈ {L, H}, receives in the proposed equilibrium of the bargaining game. A proposed equilibrium fails to satisfy the intuitive criterion—and hence cannot be an equilibrium—if there exists an ˜ such that the following conditions are out-of-equilibrium offer (˜q, d), satisﬁed: b u(˜q) − φH d˜ > UH ,

(7.9)

u(˜q) − φL d˜ < ULb ,

(7.10)

−c(˜q) + φH d˜ ≥ 0.

(7.11)

˜ would make an H-type buyer strictly According to (7.9), the offer (˜q, d) better off if it were accepted but, according to (7.10), would make an L-type buyer strictly worse off. Since the L-type buyer has no incentive to make this offer, the seller should believe that it came from an H-type buyer, and will accept it if condition (7.11) holds. We provide a characterization of an equilibrium offer by ﬁrst demonstrating what cannot be an equilibrium. In particular, a pooling offer— where the H- and L-type buyers make the same offer—cannot be an equilibrium. Figure 7.3 illustrates the argument. Consider a proposed pooling equilibrium, where both types of buyers make the offer ¯ = (0, 0) to the seller in the bargaining game, and the offer is (¯q, d) ¯ generates a surplus U b ≡ u(¯q) − φL d¯ for the accepted. The offer (¯q, d) L

170

Chapter 7

d

ULb b

UH

s

UH d

q

q

Offers violating the intuitive criterion Figure 7.3 Ruling out pooling equilibria b ≡ u(¯ L-type buyer and UH q) − φH d¯ for the H-type buyer. The indifb b ference curves, UL and UH , depicted in ﬁgure 7.3 represent a set of offers (q, d) for each type of buyer that generates a surplus equal to the equilibrium surplus associated with that buyer’s type. Note that ULb is b since φ > φ . The participation constraint of a seller steeper than UH H L who believes he is facing an H-type buyer is represented by the locus s ¯ UH ≡ (q, d) : −c(q) + φH d = 0 . The proposed equilibrium offer (¯q, d) s ¯ lies above UH since it is accepted when λ(¯q, d) < 1. The shaded area in ﬁgure 7.3 identiﬁes the set of offers, when compared to the proposed equilibrium, that (1) increase the surplus of an b ; (2) reduce the surplus of an H-type buyer—offers to the right of UH b L-type buyer—offers to the left of UL ; and (3) are acceptable to the seller s . The offers in the shaded area assuming that λ = 1—offers above UH satisfy conditions (7.9)–(7.11), which implies that the proposed equi¯ violates the intuitive librium where both types of buyers offer (¯q, d) criterion. Indeed the H-type buyer is able to make an offer different ¯ that, if accepted, would make him better off, while makfrom (¯q, d) ing an L-type buyer strictly worse off. Moreover, provided that the seller believes that this offer is coming from an H-type buyer, then it is acceptable.

Information, Monetary Policy, Inﬂation–Output Trade-Off

171

Since pooling offers are not compatible with equilibrium, if an equilibrium exists, it must be characterized by separating offers, i.e., the Land H-type buyers make different offers. But if the offers are separating, then in equilibrium the seller can attribute each offer to a buyer’s type. This means that in the equilibrium the seller knows exactly what type of money he is receiving, either low or high value. If offers are separating, then the L-type buyer can do no worse than to make the offer that he would make under complete information, since this complete information offer is always acceptable to the seller, independent of his beliefs. The L-type buyer cannot do any better than this; otherwise, the offer would have to be pooled with an H-type buyer offer. But such offers have been ruled out as possible equilibrium outcomes. Hence the payoff of an L-type buyer is given by ULb = max u(q) − φL d q,d≤m

subject to

− c(q) + φL d ≥ 0.

The solution to problem (7.12) is qL = min q∗ , c−1 (φL m) , ∗ c(q ) dL = min ,m . φL

(7.12)

(7.13) (7.14)

If the L-type buyer’s money holdings are sufﬁciently large, then the trade in L-type matches is efﬁcient, qL = q∗ . If instead the value of the money holdings is less than the cost of producing q∗ —where a unit of money is valued at φL —then the L-type buyer is unable to purchase the efﬁcient quantity of output and qL < q∗ . In both cases, the buyer appropriates the entire surplus of the match. Consider now the offermadeby an H-type buyer, (qH , dH ), given the offer of the L-type buyer, qL , dL . An H-type buyer’s offer, (qH , dH ), will be part of an equilibrium ifthe L-type buyer does not have a strict pref erence to offer it instead of qL , dL . Hence (qH , dH ) solves the following problem: b UH = max u(q) − φH d q,d≤m

subject to

− c(q) + φH d ≥ 0

and u(q) − φL d ≤ ULb = u qL − c qL .

(7.15) (7.16)

172

Chapter 7

From (7.15) the buyer maximizes his expected surplus subject to the participation constraint of the seller—where the seller has the correct belief that he faces an H-type buyer—and the incentive-compatibility condition, (7.16), that an L-type buyer cannot be made better off by offering (qH , dH ) instead of qL , dL . Note that the solution satisﬁes the intuitive criterion, since there is no other acceptable offer that the H-type buyer could make that would raise his payoff and would not increase the payoff to the L-type buyer. A belief system consistent with the equilibrium offers has the seller attributing all offers that violate (7.16) to L-type buyers, and all other out-of-equilibrium offers to H-type buyers; see ﬁgure 7.4. Notice that the intuitive criterion selects the Pareto-efﬁcient equilibrium among all separating equilibria. The solution to (7.15)–(7.16) has both constraints binding. To see this, consider ﬁrst the incentive-compatibility condition (7.16). Suppose that this condition does not bind; then the solution, qH , dH , to problem (7.15)–(7.16) is the complete information offer—given by problem (7.15)—and u(qH ) − φL dH = u(qH ) − c qH + (φH − φL ) dH > ULb ,

d s

UL

dL

ULb

b

UH s

UH

dH

qH Acceptable offers Figure 7.4 Separating offer

q*

q

Offers attributed to L-type buyers

Information, Monetary Policy, Inﬂation–Output Trade-Off

173

where we have used that c qH = φH dH . The inequality follows from the observation for a given m, the complete-information output in state ∗ that, −1 H is min q , c (φH m) , and hence the complete-information payoff of an H-type buyer exceeds that of an L-type buyer. In words, the condition above states that the L-type buyer can be made better off, to compared offer qL , dL , by mimicking the H-type buyers’ offer, qH , dH . This is not compatible with equilibrium, and hence constraint (7.16) must bind. Consider now the participation constraint, −c(q) + φH d ≥ 0, given in (7.15). Suppose that this constraint does not bind. Then problem (7.15)– (7.16) becomes b UH = max (φL − φH ) d + ULb = ULb . d≤m

b = U b > 0, which implies The solution to this problem is dH = 0 and UH L qH > 0. But this solution violates the seller’s participation constraint, which implies that the seller’s participation constraint must bind. In summary, the solution to problem (7.15)–(7.16) satisﬁes

u(qH ) −

φL c(qH ) = u(qL ) − c(qL ), φH dH =

u(qH ) − ULb . φL

(7.17) (7.18)

Using (7.15) and (7.16) with a strict equality, we ﬁnd the payoff to an H-type buyer to be b UH = u qH − φH d = ULb − (φH − φL ) d.

(7.19)

Substituting (7.18) for d, (7.19) can be written as

φH b φH − φL b UH u(qH ), = UL − φL φL which is decreasing in qH . Consequently the solution (qH , dH ) to problem (7.15)–(7.16) corresponds to the lowest qH that solves equation (7.17). But note that (7.17) determines a unique qH in the interval 0, qL . To see this, notice that if qH = 0, then the left side is less than the right side; if qH = qL , then the opposite is true. Moreover, for all qH ≤ q∗ , the left side is increasing in qH . Hence (7.17) has a unique solution qH ∈ (0, qL ). Given qH , dH is determined by (7.18). The most notable feature of this solution is that qH < qL , which implies c(qH ) = φH dH < c(qL ) = φL dL , and hence

174

Chapter 7

dH < dL ≤ m. The lower velocity of money in the H-state is a consequence of H-type buyers separating themselves from L-type buyers. √ If we adopt the functional forms c(q) = q and u(q) = 2 q, we can obtain closed-form solutions for the expression of the quantities traded in the DM. From (7.13), qL = min [1, φL m], and (7.17) becomes φL √ √ qH − 2 qH + 2 qL − qL = 0. φH The smallest value of qH that solves this equation is qH =

φH φL

!

2

φL √ 2 qL − qL 1− 1− φH

.

(7.20)

It is clear from this expression that the quantities traded in the H-state depend on the discrepancy of the value of money in the different states, φH /φL , and on the quantity traded in the L-state, qL . Note that if φH = φL , then qH = qL . The buyers’ offers are illustrated in ﬁgure 7.4 for the case where the constraint dL ≤ m does not bind. The offer of the L-type buyer is given by the point where the iso-surplus curve of the seller who knows that s he is facing an L-type buyer, UL ≡ (q, d) : −c(q) + φL d = 0 , is tangent to the iso-surplus curve of the L-type buyer, ULb . In order for the H-type buyer to satisfy the seller’s participation constraint, c qH − φH dH = 0, and condition (7.16) with an equality, he must make an offer that is in the region to the left of (and including) curve ULb and above (and including) s . This region is identiﬁed as “Acceptable offers’’ in ﬁgure 7.4. curve UH The utility-maximizing offer in this region is given by the intersection s curves. of the ULb and UH 7.3 Equilibrium under Asymmetric Information The terms of trade in the DM of period t are a function of the buyer’s private signal and the money balances he accumulated in the CM of period t − 1. Using (7.2), we write the buyer’s choice of money holdings in the CM of period t − 1 as

φt−1 max −φt−1 m + βσ α u(qL ) + (m − dL ) m≥0 γ¯ φt−1 + (1 − α) u(qH ) + (m − dH ) . γ

Information, Monetary Policy, Inﬂation–Output Trade-Off

175

Since (φt−1 /γ¯ )dL = c(qL ), (φt−1 /γ )dH = c(qH ), and φte = α(φt−1 /γ¯ ) + (1 − α)(φt−1 /γ ), this problem becomes max −φt−1 m + βσ α u(qL ) − c(qL ) + (1 − α) u(qH ) − c(qH ) + βφte m , m≥0

(7.21) where qL and qH solve

∗ −1 φt−1 m qL = min q , c , γ¯ and u(qH ) −

γ γ¯

c(qH ) = u(qL ) − c(qL ),

(7.22)

where (7.22) is identical to (7.17) since γ φL φt−1 /γ¯ = = . φH φt−1 /γ γ¯ According to (7.21) the buyer accumulates φt−1 m real balances in the CM of period t − 1. With probability α, the value of money in t is low and the buyer consumes qL , and with probability 1 − α, it is high and the buyer consumes qH . In both cases the buyer enjoys the whole surplus of the match in the DM of period t. Finally, the buyer can resell any money he has leftwhen entering the CM of t at the expected price φte = α/γ¯ + (1 − α)/γ φt−1 . By grouping the m terms and then dividing by β, we can rearrange (7.21) as max −ˆı φt−1 m + σ α u(qL ) − c(qL ) + (1 − α) u(qH ) − c(qH ) , (7.23) m≥0

where ıˆ ≡ β −1 − (1 − α)/γ + α/γ¯ . The buyer chooses his money holdings in order to maximize his expected surplus in the DM, net of the cost of holding real balances. The cost of holding real balances, ıˆ, is the difference between the gross rate of time preference and the expected gross rate of return of money, the surplus in the L-state is SL = u(qL ) − c(qL ) and the surplus in the H-state is SH = u(qH ) − c(qH ). Observe that both SL and SH are increasing functions of DM output in the L-state, qL , and are strictly increasing if qL < q∗ . This can be seen

176

Chapter 7

by differentiating the buyer’s surpluses in the low and high states with respect to qL : dSL ≡ SL = u (qL ) − c (qL ) ≥ 0, dqL dqH dSH ≡ SH = u (qH ) − c (qH ) dqH dqL u (qH ) − c (qH ) u (qL ) − c (qL ) ≥ 0, = γ u (qH ) − γ¯ c (qH )

(7.24) (7.25)

where, from (7.22), we used −1 γ dqH u (qL ) − c (qL ) ≥ 0. = u (qH ) − c (qH ) γ¯ dqL

(7.26)

The value of an additional unit of output in the low state, SL , is simply the marginal match surplus, u (qL ) − c (qL ), which gives us (7.24). An additional unit of output in state L relaxes incentive-compatibility constraint (7.16), which allows the buyer to raise his consumption by the amount given in (7.26) in state H. Since the buyer obtains the whole surplus of the match, each additional unit of consumption in the DM raises his surplus by u (qH ) − c (qH ), which gives us (7.25). Since the surpluses in both the H- and L-states are increasing functions of qL , and since the buyer will never bring more money than what is required to buy qL in the L-state because it is costly to hold money and qH < qL , we can re-express the buyer’s problem (7.23) as a choice of qL . Given that the buyer chooses φt−1 m = γ¯ c qL , the buyer’s problem can be expressed as max −ˆı γ¯ c(qL ) + σ α u(qL ) − c(qL ) + (1 − α) u(qH ) − c(qH ) . qL ∈[0,q∗ ] (7.27) From (7.24) and (7.25) the marginal surplus functions dSL /dqL and dSH /dqH are decreasing in qL and qH for all qH , qL ∈ [0, q∗ ]. Since qH is increasing with qL , we can deduce that the buyer’s objective function in problem (7.27) is concave in qL . The ﬁrst-order (necessary and sufﬁcient) condition for the buyer’s choice of output in the L-state is given by

u (qH ) − c (qH ) α u (qL ) ıˆ = σ (1 − α) + −1 . (7.28) γ¯ u (qH ) − γ c (qH ) γ¯ c (qL )

Information, Monetary Policy, Inﬂation–Output Trade-Off

177

The cost of holding money, the left side of (7.28), must be equal to the marginal beneﬁt from holding money in the DM, which is the right side of (7.28). The right side of (7.28) varies from +∞ to 0 as qL varies from 0 to q∗ . Hence there is a unique qL that solves (7.28). Market-clearing requires that m = Mt , so that the value of money in period t − 1 is uniquely determined by c(qL ) = (φt−1 /γ¯ )Mt , i.e., φt−1 = γ¯ c(qL )/Mt . Finally, notice that (7.24) and (7.25) imply that dSH u (qH ) − c (qH ) dSL = . (7.29) γ dqH u (qH ) − c (qH ) dqL γ¯

Since between qL and φt−1 m, i.e., qL = there is aone-to-one relationship −1 c (φt−1 /γ¯ )m , one can interpret SH and SL as the liquidity value of real balances in the H- and L-states, respectively. Hence, since the squared bracketed term on the right side of (7.29) is less than one, the liquidity value of an additional unit of real balances is lower in the H-state than it is in the L-state, i.e., SH ≤ SL (and with a strict inequality when qL < q∗ ). So, paradoxically, the liquidity value of money is lower when its market price is high or, equivalently, when inﬂation is low. 7.4 The Inﬂation and Output Trade-Off We now discuss some basic properties of the equilibrium when there is asymmetric information. The model makes some predictions regarding correlations between inﬂation, output, and the velocity of money. First, the model predicts a positive correlation between output and inﬂation. If γt = γ¯ , then qt = qL ≤ q∗ ; if γt =γ , then qt = qH < qL . Consequently the model generates an upward-sloping Phillips curve, and an apparent trade-off between inﬂation and output. Second, the model predicts a positive correlation between the velocity of money and inﬂation. If γt = γ , then dt = dH < Mt ; if γt = γ¯ , then dt = dL = Mt . Buyers spend all their money holdings in the highinﬂation state, but only a fraction of it in the low-inﬂation state. These correlations are illustrated in ﬁgure 7.5, where we plot the quantities traded and the money transfers in the DM as a function of the inﬂation rate. Ashort-run Phillips curve emerges because of the informational asymmetry that prevails between buyers and sellers regarding the future value of money. When buyers learn that the inﬂation rate is low and the value of ﬁat money is high, they signal this information to sellers

178

Chapter 7

q qL

qH

γ

γ

γ

dH

dL = M

d Figure 7.5 Output, velocity, and inﬂation

by retaining a fraction of their (valuable) money holdings, and reducing their DM consumption. It is because buyers are willing to hold onto their money balances that sellers can be convinced that ﬁat money has a high value. If the inﬂation rate is high, and the value of ﬁat money is low, buyers do not have to signal its value, and hence they spend it all in the DM. The structure of the asymmetric information mechanism in our model suggests a new explanation for the non-neutrality of money and the inﬂation–output trade-off. A related explanation, based on agents’ imperfect information about monetary policy, suggests that output rises when inﬂation is high because agents are unable to disentangle nominal and real shocks. Agents who face this “signal extraction problem’’ attribute a high nominal price for the good they produce to both an increase in the real price of this good and an increase in the stock of

Information, Monetary Policy, Inﬂation–Output Trade-Off

179

money. The precise division between the real and nominal components will depend on how often the monetary authority generates high inﬂation. So the reason why output is high when inﬂation is high is because agents incorrectly attribute an increase in the price of the good they produce to real factors as opposed to monetary ones. In contrast, the positive correlation between output and inﬂation in our model is not due to agents being mistaken, since in equilibrium both buyers and sellers know the true value of ﬁat money. Another popular explanation for changes in the money supply having real effects is the presence of price rigidities. If, for some reason, producers set nominal prices and can only adjust these prices infrequently, then an unanticipated increase in the money supply can lead to a higher demand for those goods whose prices have not been adjusted. In our model the real effects of monetary policy are not based on any notions of nominal rigidities that might arise from the existence of informational asymmetries. To see this, suppose that the seller’s cost function in the DM is linear, c(q) = q. Then, according to (7.12) and (7.15), the price of output in the DM of period t is deﬁned as the monetary payment divided by the output traded, and is given by γ dH 1 = = , qH φH φt−1 dL 1 γ¯ = = . qL φL φt−1 In both the high- and low-inﬂation states, the nominal price is proportional to the money growth rate. We now ask whether the monetary authority can take advantage of the apparent trade-off between inﬂation and output by implementing the high money growth rate more often. Suppose that the monetary authority increases the frequency for the high money state, meaning it increases α. The equilibrium condition (7.28) can be compactly reexpressed as (α, qL ) = 0, where 1−α α −1 (α, qL ) = β − + γ γ¯

u (qH ) − c (qH ) α u (qL ) − σ (1 − α) + −1 , γ¯ u (qH ) − γ c (qH ) γ¯ c (qL )

180

Chapter 7

and from (7.17), qH is an increasing function of qL . By totally differentiating the equilibrium condition, we obtain that dqL /dα = −α / qL , where α and qL are the partial derivatives of with respect to α and qL , respectively. Using the fact that is increasing in qH and qL , it can easily be seen that qL > 0. Differentiating (α, qL ) with respect to α, we obtain

1 1 u (qH ) − c (qH ) 1 u (qL ) α = − + σ − − 1 . γ γ¯ γ¯ u (qH ) − γ c (qH ) γ¯ c (qL ) Consider the case where ıˆ is close to zero so that qL is close to q∗ ; see equation (7.28). This implies that the liquidity premium in the low state, u (qL )/c (qL ) − 1 , is close to zero and hence α ≈ (1/γ ) − (1/γ¯ ) > 0. Consequently for ıˆ close to zero, an increase in α reduces the value of money and the output in all states. If the policy maker attempts to exploit the trade-off between inﬂation and output in a more systematic way, then agents will change their expectations about the occurrence of the different states, which in turn will adversely affect the value of money and output in the different states. The overall effect of increasing the frequency of the high-inﬂation state on expected aggregate output, however, is ambiguous because the high-inﬂation state, which is associated with a higher level of output, occurs more often. To see this, suppose that γ = β < γ¯ and α ≈ 0. Then ıˆ = β −1 − (1 − α)/γ + (α/γ¯ ) ≈ 0 and qH < qL = q∗ . From (7.17), dqH dα

α≈0+

dqL u (qL ) − c (qL ) /dα = 0. = u (qH ) − γ c (qH )/γ¯

A change in α affects qH indirectly through the buyer’s surplus in the L-state. Since qL = q∗ , a change in α only has a second-order effect on the buyer’s surplus in the on the quantities traded L-state and, hence, in the H-state. Let Y = σ αqL + (1 − α)qH . Then dY dα

α≈0+

= σ qL − qH > 0.

If the rate of return on money is close to the rate of time preference, and if the high money growth rate occurs infrequently, then an increase in the

Information, Monetary Policy, Inﬂation–Output Trade-Off

181

frequency of the high money growth rate can lead to higher aggregate output. The existence of this trade-off has also implications for social welfare measured here by the expected surplus in the DM, W = σ α u(qL ) − c(qL ) + (1 − α) u(qH ) − c(qH ) . By the same reasoning as above, dW dα

α≈0+

=σ

u(qL ) − c(qL ) − u(qH ) − c(qH ) > 0.

By increasing the frequency of the high-money-growth-rate state, the policy maker can raise welfare. When γ = γ = β and α ≈ 0, prices (on average) fall over time. In this situation buyers do not want to spend all of their cash (in state H) and prefer to wait until the subsequent CM, where the value of money is realized. This description is loosely related to a common-held view that deﬂation hurts society because agents hoard their money balances when they anticipate the value of money will increase over time, and a small expected inﬂation will increase output and welfare since agents will spend their money holdings faster. While this view is difﬁcult to capture in our environment with symmetric information, it is quite natural when information is asymmetric. It is worth emphasizing that the allocations and output levels are not continuous at α = 0. If α is exactly equal to zero, then the money growth rate is deterministic, and there is no uncertainty about the value of money. There is no informational asymmetry in the DM and buyers do not need to signal the value of money to sellers. In that case, if γ = β, then qH = q∗ . In contrast, if there is a chance that the policy maker chooses a high money growth rate, this possibility affects the quantities traded in the low-money-growth-rate state no matter how small α is. The mere possibility that the policy maker might implement a high money growth rate, even it is a very rare event, has a nonvanishing negative externality on the quantities traded in the low-inﬂation state. This feature of the model is a consequence of a separating equilibrium— selected by the intuitive criterion—where the terms of trade in the low inﬂation state are determined by the incentive-compatibility condition (7.16). If the intuitive criterion for equilibrium selection is dropped, then the discontinuity may no longer exist in a pooling (perfect Bayesian) equilibrium. But in that case, there would be no correlation between inﬂation and output.

182

Chapter 7

Output as a function of α

0.6

0.5

γ– = 1.1

0.4

γ– = 1.5

0.3

γ– = 2

0.2 0.0

0.2

0.4

0.6

0.8

1.0

Welfare as a function of α 0.95

0.90

γ– = 1.1 0.85

γ– = 1.5

0.80

0.75

γ– = 2

0.70 0.0

0.2

0.4

0.6

0.8

1.0

Figure 7.6 Inﬂation-output trade-off

The analytic results obtained so far, regarding the effect that monetary policy has on aggregate output and welfare, are valid for small values of α. We now use a simple numerical example to investigate the case where α is not close to 0. We take the functional forms c(q) = q and u(q) = √ 2 q, and set σ = 1 and β = 0.9. We assume that γ = β, and take three possible values for the high inﬂation state, γ¯ ∈ {1.1, 1.5, 2}. In ﬁgure 7.6 we plot aggregate output and welfare—as measured by the expected

Information, Monetary Policy, Inﬂation–Output Trade-Off

183

match surplus in the DM—as a function of α, the frequency with which the high-money-growth-rate state occurs. Provided that the difference between the money growth rates in the two states is not too large, the model predicts that there is an exploitable trade-off between inﬂation and output. Moreover increasing the frequency at which the monetary authority implements the high money growth rate can raise society’s welfare. The reasoning behind this exploitable trade-off is as follows: In the low-money-growth-rate state, buyers hoard money balances in order to signal the high value of money to sellers. As a result output in the DM is quite low, and this is costly for society. In contrast, in the high-moneygrowth-rate state, since buyers do not hoard any cash, output is higher than it is in the low-money-growth-rate state. If we assume that the value of money is ﬁxed, then by implementing the high-money-growth-rate state more often, the monetary authority can reduce the welfare cost associated with signaling. We will refer to this as the (positive) outputcomposition effect associated with increasing α. Of course, the value of money does not remain ﬁxed if the monetary authority implements the high-money-growth-rate state more often; it falls. As a result the amount of output that is purchased in both the high- and low-money-growth states falls. We will refer to this as the inﬂation tax effect associated with increasing α. If the output-composition effect dominates the inﬂation tax effect, then increasing the frequency of the high-inﬂation state actually increases output and welfare. Our numerical examples indicate that if the difference between money growth rates is not too big, then the output-composition effect can dominate the inﬂation tax effect for all values of α = 0. When the difference between money growth rates is not too big, such as γ¯ = 1.1 in our numerical example, if the monetary authority cannot implement γ = γ with certainty, then in fact it is optimal to choose the high money growth rate with probability one. If, however, the difference between monetary growth rates is not small, such as γ¯ = 1.5 or γ¯ = 2 in our numerical example, then output and welfare are nonmonotonic in α. This means that as the monetary authority increases the frequency of the high-moneygrowth-rate state, at some point the inﬂation tax effect dominates the output-composition effect, which implies that output and welfare will fall. For these cases there is an optimal frequency to implement the high-money-growth-rate state, and it is less than one. Up to this point, the policy takes the form of a choice of α, taking γ and γ¯ as given. Now let’s examine the optimal monetary policy when the

184

Chapter 7

policy maker can also choose γ and γ¯ . One may wonder if the inﬂationoutput trade-off justiﬁes a deviation from the Friedman rule. We saw in chapter 6 that the optimal monetary policy in an environment where the money supply is growing at a constant rate sets the cost of holding real balances to zero. In our model this version of the Friedman rule would require that β −1 = (1 − α)/γ + (α/γ¯ ). Since at the Friedman rule the expected rate of return of ﬁat money must equal the gross rate of time preference, we have γ < β < γ¯ , if α ∈ (0, 1) and γ = γ¯ . Hence, the ex-post rate of return of ﬁat money is larger than the rate of time preference in the low-inﬂation state, but it is smaller in the high-inﬂation state. From (7.28) the quantity traded in the high-inﬂation state approaches the ﬁrst-best level, q∗ , as the expected cost of holding real balances, ˆi, approaches zero. And from (7.17) the quantity traded in the low-inﬂation state, qH , solves u(qH ) −

γ γ¯

c(qH ) = u(q∗ ) − c(q∗ ).

(7.30)

Since γ /γ¯ < 1, the smallest solution to (7.30) has qH < q∗ . So equalizing the expected rate of return of currency to the rate of time preference is not enough to implement the efﬁcient allocation. The informational asymmetry between buyers and sellers causes the quantity traded in the low-inﬂation state to be inefﬁciently low at the Friedman rule. This inefﬁciency can only be removed if the monetary authority eliminates the ﬂuctuations of the money supply, i.e., if γ = β = γ¯ .

(7.31)

Clearly, if (7.31) holds, then from (7.30), qH = q∗ . Hence targeting the nominal interest rate is not sufﬁcient to ensure that the efﬁcient allocation is implementable; the optimal policy consists in targeting the rate of growth of money supply. This is one instance where the distinction between the two policies—targeting nominal interest rates versus money growth rates—really matters. 7.5 An Alternative Information Structure Thus far we have assumed that buyers receive an informative signal about the rate of growth of money supply and, hence, the future value

Information, Monetary Policy, Inﬂation–Output Trade-Off

185

of money, while sellers do not. This assumption is consistent with the view that agents have greater incentives or opportunities to learn about the future value of the assets they hold. It is equally plausible that sellers receive prior information regarding monetary policy. To see how the information structure affects the relationship between inﬂation and output, we will now suppose that sellers are informed about monetary policy while buyers are not. The bargaining game that occurs in the DM, which is illustrated in ﬁgure 7.7, has the structure of a screening game. The assumption that buyers are uninformed about monetary policy is captured by the dotted line in ﬁgure 7.7 that represents an information set. An offer by the buyer consists of a menu of various terms of trade. Since there are two possible signals that the seller can receive, we need only to consider menus with two items, {(qH , dH ), (qL , dL )}, where (qH , dH ) is the terms of trade intended for sellers in the low-inﬂation state—when the value of ﬁat money is high—and (qL , dL ) is the terms of trade for sellers in the high-inﬂation state. A buyer holding m units of money offers a menu {(qH , dH ), (qL , dL )} that solves max (1 − α) u(qH ) − φH dH + α u(qL ) − φL dL (7.32) qH ,qL ,dH ,dL

subject to dL ≤ m, dH ≤ m,

inf lat

lat inf

[α] [1−α]

ion

Hig h

w Lo

ion

N

B

B

Offer

Offer

S Yes

S No

Yes

No

Figure 7.7 Bargaining game when buyers are uninformed

186

Chapter 7

and −c(qH ) + φH dH ≥ 0,

(7.33)

−c(qL ) + φL dL ≥ 0,

(7.34)

−c(qH ) + φH dH ≥ −c(qL ) + φH dL ,

(7.35)

−c(qL ) + φL dL ≥ −c(qH ) + φL dH .

(7.36)

According to (7.32)–(7.36) the buyer maximizes his expected surplus, subject to individual rationality and incentive compatibility constraints. The conditions (7.33) and (7.34) are the individual rationality constraints for sellers in the low-inﬂation and high-inﬂation states, respectively, while conditions (7.35) and (7.36) are the incentive-compatibility constraints. According to (7.35) a seller who knows that the money growth rate will be low in the current period prefers allocation qH , dH to the terms of trade intended for the high-inﬂation-type seller. Inequality (7.36) has a similar interpretation. In the appendix we show that the solution to (7.32)–(7.36) has individual-rationality constraint (7.34) and incentive-compatibility constraint (7.35) binding: −c(qL ) + φL dL = 0, −c(qH ) + φH dH = −c(qL ) + φH dL > 0.

(7.37) (7.38)

From (7.37) buyers leave no surplus to sellers in the high-inﬂation state. In contrast, from (7.38) the seller is able to extract a positive surplus, or informational rent, in the low-inﬂation state that is equal to −c(qH ) + φH dH = −c(qL ) + φL dL + (φH − φL ) dL = (φH − φL ) dL .

(7.39)

Intuitively, a seller in the low-inﬂation state, state H, is the one who has an incentive to misrepresent his private information since, when the value of money is low, he does not have to produce much for the same transfer of money. This incentive to lie by the seller in state H explains why his incentive-compatiblity constraint is binding. And since a seller has no incentive to claim that inﬂation is low, state H, when it is actually high, state L, the buyer is able to extract all of the match surplus in the high-inﬂation state. This is why the individual-rationality constraint for the seller binds in state L. One can check (see the appendix) that dL ≤ dH and qL ≤ qH , and that either dL = dH and qH = qL or dL < dH and qH > qL . So, when

Information, Monetary Policy, Inﬂation–Output Trade-Off

187

the allocation is a separating one, both output and velocity are negatively correlated with inﬂation. Hence the nature of the informational asymmetry between buyers and sellers is crucial for the sign of the correlation between inﬂation and output. If buyers are informed, then there is a positive correlation between inﬂation and output. This trade-off emerges because buyers signal the high value of money by retaining a fraction of their money holdings. If instead sellers are informed, then the correlation between inﬂation and output is negative. In this situation buyers reduce their cost of extracting sellers’ information by spending less money in the high-inﬂation state, which reduces sellers’ rent. Consider a policy where the cost of holding real balances is zero, ˆi = 0. For this policy, the buyer’s problem (7.32)–(7.36) can be greatly simpliﬁed. First, buyers will accumulate sufﬁcient real balances so that they are unconstrained in all states, which implies that the constraints dL ≤ m and dH ≤ m can be ignored. Second, since the seller receives an informational rent equal to (φH − φL ) dL in the low-inﬂation state, the buyer’s objective function, (7.32), thanks to (7.39), can be written as (1 − α) u(qH ) − c qH − (φH − φL ) dL + α u(qL ) − φL dL . And ﬁnally, since the seller does not receive any surplus in the highinﬂation state, (7.37), the buyer’s objective function can be further rewritten as (φH − φL ) (1 − α) u(qH ) − c qH − c qL + α u(qL ) − c qL . (7.40) φL The buyer’s problem therefore is simply to choose qH and qL so as to maximize (7.40). The ﬁrst-order conditions for this problem with respect to qH and qL are u qH = c qH or qH = q∗ and

u (qL ) 1−α γ¯ −1 , (7.41) = 1+ c (qL ) α γ respectively. When the Friedman rule is implemented, the quantity traded in the low-inﬂation state is socially efﬁcient, while the quantity traded in the high-inﬂation state is inefﬁciently low, provided that γ < γ¯ . As in the previous section, a policy that consists in setting the expected cost of holding real balances equal to zero is not sufﬁcient to obtain the efﬁcient allocation when buyers and sellers are asymmetrically informed. In order to implement the efﬁcient allocation, the money growth rate must also be constant, γ¯ = γ .

188

Chapter 7

7.6 Further Readings Lucas (1972, 1973) introduces models with imperfect information to explain how unanticipated monetary shocks affect output. Lucas (1972) adopts an overlapping generations model, in which young producers are divided unevenly across markets and the supply of money is stochastic. The producers observe the price in their market, but they do not know the average price level. Therefore, conditional on the price they observe, producers will have to disentangle real from nominal disturbances. A tractable version of the model with aggregate shocks is provided by Wallace (1992). Benassy (1999) provides analytical solutions to the model. Wallace (1997) considers an unanticipated change of the money supply in a random-matching model, and shows that the short-run effects are predominantly real while the long-run effects are predominantly nominal. Faig and Li (2009) introduce the Lucas signal extraction problem into the Lagos–Wright model and estimate the welfare costs of expected and erratic inﬂations. Araujo and Camargo (2006) consider a search-theoretic model in which information about the value of indivisible ﬁat money is imperfect and learning is decentralized. Araujo and Shevshenko (2006) consider an economy where agents have incomplete information with respect to the value of money, and they learn from private experiences. Models with sticky prices include Taylor (1980), Rotemberg (1982), and Calvo (1983). Benabou (1988) and Diamond (1993) introduce menu costs in search-theoretic models without money and show that inﬂation can be welfare improving. Craig and Rocheteau (2008) develop a continuous-time version of Lagos–Wright with menu costs. The optimal monetary policy corresponds to a deﬂation. Aruoba and Schorfheide (2011) also introduce nominal rigidities into a search model with divisible money. Mankiw and Reis (2002) explain the inﬂation–output trade-off by the assumption that information about macroeconomic conditions diffuses slowly through the population. In the context of a search model with divisible money, Williamson (2006) assumes that agents participate only infrequently in the competitive market where monetary injections take place. Finally, Sanches and Williamson (2010) introduce an asymmetry of information regarding the seller’s value of money in the context of the Lagos–Wright model.

Information, Monetary Policy, Inﬂation–Output Trade-Off

189

Appendix Informed Sellers and Uninformed Buyers Consider a match between a buyer holding a portfolio of m units of money and a seller. The buyer is uninformed about the future value of money, but he knows that with probability α the value of money is φL while with probability 1 − α the value of money is φH . The seller is informed about the value of money. The buyer offers a menu {(qH , dH ), (qL , dL )} where (qH , dH ) are the terms of trade for the seller in the high state and (qL , dL ) are the terms of trade for the seller in the low state. The buyer commits to these terms of trade. The buyer’s problem is max α u(qL ) − φL dL + (1 − α) u(qH ) − φH dH (7.42)

(qH ,dH ),(qL ,dL )

subject to the feasibility constraints dH ∈ [0, m], dL ∈ [0, m], and to the following incentive constraints: −c(qL ) + φL dL ≥ 0,

(7.43)

−c(qH ) + φH dH ≥ 0,

(7.44)

−c(qL ) + φL dL ≥ −c(qH ) + φL dH ,

(7.45)

−c(qH ) + φH dH ≥ −c(qL ) + φH dL .

(7.46)

The conditions (7.43) and (7.44) are individual-rationality constraints for sellers in the low and high states, respectively. Conditions (7.45) and (7.46) are incentive-compatibility constraints. We now establish that for any optimal menu, (7.43) and (7.46) are binding: −c(qL ) + φL dL = 0, −c(qH ) + φH dH = −c(qL ) + φH dL = (φH − φL )dL .

(7.47) (7.48)

First, (7.44) and (7.46) cannot both hold with a strict inequality since, if this were the case, the buyer could raise his expected surplus by increasing qH and keeping (qL , dH , dL ) unchanged without upsetting (7.43)–(7.46). By identical reasoning, (7.43) and (7.45) cannot both hold with strict inequality. Second, (7.46) must bind. To see this, assume the contrary: that (7.46) holds with a strict inequality. Then (7.46) and (7.43) imply that

190

Chapter 7

−c(qH ) + φH dH > −c(qL ) + φH dL ≥ 0. This set of inequalities implies that whenever (7.46) holds with a strict inequality, then so does (7.44). A contradiction with our ﬁrst point. Hence, (7.46) must bind. Third, to show that (7.43) binds assume to the contrary that (7.43) holds with a strict inequality. Then, from the reasoning above, (7.45) must bind. From (7.46) and (7.45) at equality, φL (dH − dL ) = c(qH ) − c(qL ) = φH (dH − dL ) . This implies that the menu offered by the buyer is pooling, dH = dL and qL = qH . But then (7.43) cannot be slack since otherwise the seller would be able to increase his expected payoff by lowering qL and qH without upsetting any constraint (7.43)–(7.46). A contradiction. The reasoning above shows that buyers leave no surplus to sellers in the low state whereas sellers in the high state can extract an informational rent equal to (φH − φL )dL . Moreover sellers in the low state transfer less money than sellers in the high state, dL ≤ dH . To see this, rearrange (7.45) and (7.46) to read φL (dH − dL ) ≤ c(qH ) − c(qL ) ≤ φH (dH − dL ) .

(7.49)

It also implies that qL ≤ qH . We will make use of the previous insights to reduce the buyer’s problem to the maximization of (7.42) subject to the constraints (7.47) and (7.48). From (7.48) it is immediate that (7.44) holds. Moreover from (7.47)–(7.48), (7.45) holds whenever dL ≤ dH , which, as we demonstrated above, is the case. The buyer’s maximization problem can be divided into two steps. First, taking dL as given, the buyer chooses the terms of trade in the H-state subject to the constraint that sellers must receive a surplus equal to (φH −φL )dL . The buyer’s surplus in the H-state solves b (d ) = max SH L (qH ,dH ) u(qH ) − φH dH subject to and

− c(qH ) + φH dH = (φH − φL )dL

dH ∈ [0, m] .

(7.50) (7.51)

After φH dH from (7.50) is substituted into (7.6), the buyer’s surplus in the high state becomes b SH (dL ) = max u(qH ) − c(qH ) − (φH − φL )dL qH

subject to

c(qH ) + (φH − φL )dL ∈ [0, φH m] .

Information, Monetary Policy, Inﬂation–Output Trade-Off

191

If c(q∗ ) + (φH − φL )dL ≤ φH m, then qH = q∗ and φH dH = c(q∗ ) + (φH − φL )dL . If the buyer holds enough money balances, he will compensate the seller for producing q∗ and he will offer him an informational rent to guarantee that he chooses the terms of trade intended for the H-state. If c(q∗ ) + (φH − φL )dL > φH m, then the feasibility constraint dH ≤ m is binding and qH = c−1 (φH m − (φH − φL )dL ). Consequently b SH (dL ) = u(q∗ ) − c(q∗ ) − (φH − φL )dL if c(q∗ )

+ (φH − φL )dL ≤ φH m.

(7.52)

= u ◦ c−1 (φH m − (φH − φL )dL ) − φH m

otherwise.

(7.53)

b (d ) is a decreasing function of It is immediate from (7.52)–(7.53) that SH L dL and it is differentiable: b SH (dL ) = −(φH − φL ) if c(q∗ ) + (φH − φL )dL ≤ φH m

=−

u (qH ) (φH − φL ) c (qH )

otherwise.

Moreover, since qH is a decreasing function of dL , it is easy to check that b (d ) is a concave function of d . SH L L In the second step, the buyer chooses the terms of trade in the Lstate in order to maximize his expected surplus. The buyer’s expected surplus is b S b = max (1 − α)SH (dL ) + α u(qL ) − φL dL (7.54) qL ,dL

subject to

− c(qL ) + φL dL = 0

and dL ∈ [0, m] .

(7.55) (7.56)

According to (7.54) the buyer takes into account that the surplus in the H-state depends on the transfer of money in the L-state through the incentive-compatibility conditions. Substitute φL dL from (7.55) into (7.54) to rewrite this problem as

c(qL ) b b S = max (1 − α)SH + α u(qL ) − c(qL ) qL φL subject to

c(qL ) ∈ [0, φL m] .

If the constraint c(qL ) ≤ φL m does not bind, then the choice of qL is given by the following ﬁrst-order condition:

192

−(1 − α)

Chapter 7

u (qH ) c (qH )

φH − φL c (qL ) + α u (qL ) − c (qL ) = 0. φL

(7.57)

We distinguish three cases. 1. dH ≤ m and dL ≤ m are not binding. From (7.52), qH = q∗ and from (7.57) qL solves

u (qL ) 1−α φH − φL . (7.58) = 1+ c (qL ) α φL Let q˜ L < q∗ denote the solution to (7.58). From (7.55), dL = c(˜qL )/φL , and from (7.50), dH = dL + [c(q∗ ) − c(˜qL )]/φH . This condition dH ≤ m is equivalent to m ≥ c(˜qL )/φL + [c(q∗ ) − c(˜qL )]/φH . 2. dH ≤ m is binding and dL ≤ m is not binding. From (7.53), qH = c−1 φH m − [(φH − φL )/φL ] c(qL ) . From (7.57), qL solves

u (qL ) 1 − α u (qH ) φH − φL . = 1 + c (qL ) α c (qH ) φL

(7.59)

The left side of (7.59) is decreasing from ∞ to 1 as qL increases from 0 to q∗ and the right side of (7.59) is increasing in qL (since qH is decreasing in qL ) and is greater than 1 when qL = 0. Consequently there is a unique qL = qˆ L ∈ [0, q∗ ] that solves (7.59). The condition dL ≤ m is equivalent to c(qL ) ≤ φL m, which can be rearranged as qL ≤ qH . It can be checked that qL and qH are increasing in m. Moreover from (7.59), u (qL )/c (qL ) / u (qH )/c (qH ) is increasing in m. Hence there is a threshold for m below which dL ≤ m binds. This threshold is deﬁned from (7.59) where qL = qH = q = c−1 (φL m), u (q) φL − (1 − α) φH = 1. αφL c (q) This threshold exists if φL > (1 − α) φH . 3. dH ≤ m and dL ≤ m are binding. From (7.53), qH = c−1 {φH m− [(φH − φL )/φL ] c(qL ) and from (7.55), qL = c−1 (φL m). This gives qH = qL = c−1 (φL m). Now that we have determined the terms of trade in a bilateral match, we can solve for the buyer’s choice of money holdings in the CM of period t. In this case, φH = φt /γ and φL = φt /γ¯ . The buyer’s problem is

Information, Monetary Policy, Inﬂation–Output Trade-Off

max −ˆı φt m + σ (1 − α) u(qH ) − c(qH ) − m≥0

+σ α u(qL ) − c(qL ) ,

γ¯ − γ γ

193

c(qL ) (7.60)

where ıˆ = β −1 − (α/γ¯ ) + (1 − α)/γ . In (7.60) we used the fact that the buyer’s surplus in the H-state is the whole match surplus net of the informational rent received by the seller, (φH − φL )dL , and that qH solves as ⎧ ⎨q∗ if φt m ≥ γ c(q∗ ) + γ¯ − γ c(qL ), qH = ⎩c−1 φt m − γ¯ −γ c(qL ) otherwise, γ γ and qL solves as

φt m qL = min c−1 , qˆ L . γ¯ Since it is costly to hold money, it is immediate m that the constraintdH ≤ must be binding, in which case qH = c−1 φt m/γ − (γ¯ − γ )/γ c(qL ) . The ﬁrst-order condition with respect to φt m gives u (qH ) ıˆ 1−α = − 1 σ γ c (qH )

α u (qL ) 1 − α u (qH ) γ¯ − γ + . (7.61) −1 − γ¯ c (qL ) γ¯ c (qH ) γ

If the constraint dL ≤ m does not bind, then the second term on the right side is 0 and qL = qˆ L . As ˆi tends to 0, then qH approaches q∗ and qL approaches q˜ L < q∗ .

8

Money and Credit

The key distinction between monetary and credit trades is that monetary trades are quid pro quo: goods and services are exchanged simultaneously for currency, and do not involve future obligations. Credit trades, on the contrary, are intertemporal and involve a delayed settlement. In reality, some trades are conducted through credit arrangements, other trades are based on monetary exchange. The coexistence of these different forms of payment raises some interesting questions: Are the frictions that make monetary exchange essential, such as lack of commitment and record keeping, compatible with the existence of credit? How does the presence of monetary exchange affect the use and the availability of credit? And, how does the availability of credit affect the value of money? We address these questions in this chapter. Astraightforward way to model the coexistence of monetary exchange and credit arrangements is to introduce some heterogeneity between trading matches. For example, suppose that in some markets there is no record-keeping technology, while in others there is a record-keeping technology and agents’ identities can be costlessly veriﬁed. In the former markets, agents can only trade with money, while in the latter they can resort to credit arrangements. We consider such an environment, where there is a costless technology that enforces debt contracts in some markets but not in others. In this kind of economy we do obtain coexistence of money and credit, but there is a dichotomy between the monetary and credit sectors. The amount of output that is traded with credit is determined independently from the amount of output that is traded with money. Moreover monetary policy has no effect on credit use. Since this dichotomy is an artifact of costless enforcement, we break it by making the use of record keeping both costly and a choice variable of the individual. Assuming that the gains from trade vary across

196

Chapter 8

matches in the decentralized market, DM, the mix between monetary and credit transactions is endogenous, and depends on monetary policy. We show that credit will be used for large transactions and money will be used for smaller ones, and that as inﬂation increases, the fraction of credit transactions will increase. As well, if veriﬁcation requires that sellers undertake an ex ante investment, then multiple equilibria can emerge, where different equilibria are characterized by different payment arrangements. In this situation we show that a transitory change in monetary policy can lead to a permanent change in payment arrangements. We conclude that the emergence of a payment system depends not only on fundamentals and policies but also on histories and social customs. The sort of lending and borrowing that we have considered so far have agents borrowing goods and repaying with goods. That is, a credit transaction does not involve money. We consider an environment with a market for loanable funds, where agents borrow and lend money. This market is useful in the presence of idiosyncratic shocks since it allows liquid assets to be reallocated from agents with low liquidity needs to agents with high liquidity needs. As we mentioned above, a critical distinction between monetary trades and credit transactions is that the former are atemporal, involving a quid pro quo exchange, while the latter are intertemporal, involving some degree of commitment. We capture the notion of commitment using the idea of reputation. We do so by assuming that some decentralized market matches are short-lived, lasting only for that period, while others are longer lived and can be productive for many periods. The use of credit is not incentive feasible in short-lived matches, since, owing to the lack of commitment and record keeping, the buyer will always default on repaying his obligation. In contrast, the buyer’s behavior in a longer lived match is disciplined by reputational considerations that will trigger the dissolution of a valuable relationship following a default. In this environment we show that the availability of credit depends on the value of money and monetary policy, as well as on the extent of the trading frictions. 8.1 Dichotomy between Money and Credit In this section we modify the model with divisible money described in chapter 4.1 by dividing the day market into two subperiods: a morning

Money and Credit

197

(DM1) and an afternoon (DM2). The morning and afternoon subperiods are similar in terms of agents’ preferences and specialization—buyers can consume in both subperiods but cannot produce, whereas sellers can produce but cannot consume—and in terms of the trading process— buyers and sellers trade in bilateral matches. The buyer’s instantaneous utility function is U b (q1 , q2 , x, h) = υ(q1 ) + u(q2 ) + x − h, where q1 is the consumption in the ﬁrst subperiod, q2 is the consumption in the second subperiod, x is the consumption of the general good in the third subperiod, and h is the utility cost of producing h units of the general good. The utility functions υ(q) and u(q) are strictly increasing and concave, with υ(0) = u(0) = 0, υ (0) = u (0) = +∞, and υ (+∞) = u (+∞) = 0. Without loss of generality, we assume that there is no discounting between subperiods. The discount factor across periods is β. The utility function of a seller is U s (q1 , q2 , x, h) = −ψ(q1 ) − c(q2 ) + x − h, where ψ(q) and c(q) are strictly increasing and convex, with ψ(0) = c (0) = 0, ψ (0) = c (0) = 0, and ψ (+∞) = c (+∞) = +∞. We denote q∗1 the solution to υ (q) = ψ (q) and q∗2 the solution to u (q) = c (q). These are the quantities that maximize the match surpluses in the ﬁrst two subperiods. The timing and preferences in a representative period are described in ﬁgure 8.1. Both the morning market, DM1, and the afternoon market, DM2, are characterized by search frictions. A buyer meets a seller in the DM1 with probability σ1 ∈ [0, 1], and in the DM2 with probability MORNING (DM1)

AFTERNOON (DM2)

NIGHT (CM)

Utility of consumption:

υ ( q1 )

u (q2 )

Disutility of production:

− ψ ( q1 )

− c (q2 )

−h

Anonymity

Record-keeping enforcement

Record-keeping enforcement Figure 8.1 Timing of a representative period

198

Chapter 8

σ2 ∈ [0, 1], where buyer–seller matches in the morning and the afternoon are independent events. The DM1 and DM2 differ in the following important dimension: In the former, there is a record-keeping technology and all agents’ identities are known to all other agents, whereas in the latter there is no record keeping and all agents are anonymous. Moreover any contract written in the DM1 will be enforced at night since agents who renege on their obligations can be subject to arbitrarily large ﬁnes in the CM. As a result buyers can get output in the DM1 by using credit—or, equivalently, by issuing an IOU—to be repaid at night. We will assume that all the IOUs are one period in nature in that they are repaid in the subsequent competitive night market, CM. Moreover the authenticity of the IOUs issued in DM1 cannot be veriﬁed in DM2, and hence they cannot be used as medium of exchange in the afternoon (e.g., because fake IOUs can be produced at zero cost). Remember, buyers are anonymous in the DM2, so sellers will not accept IOUs for output produced in this subperiod since buyers would renege on these at night. Because of the anonymity of agents in the DM2, money has an essential role in this environment. We assume that the stock of money grows at a constant rate γ ≡ Mt+1 /Mt , and that this is accomplished by a lump-sum transfers to buyers in the CM. We focus on stationary equilibria where real balances and the quantities traded in the different subperiods are constant over time. The former implies that φt+1 /φt = Mt /Mt+1 = γ −1 . Consider a buyer at the beginning of the CM who holds z = φt m units of real balances and has issued b units of IOUs in the previous DM1, where each unit is normalized to be worth one unit of general good. The value function for this buyer, W b (z, −b), is given by W b (z, −b) = max x − h + βV b (z ) , (8.1) x,h,z

x + b + γ z = z + h + T,

(8.2)

where V b is the value of a buyer at the beginning of the day market. According to (8.2) the buyer ﬁnances his nighttime consumption, x, the repayment of his IOU, b, and his next-period real balances, γ z , with his current real balances, z, his labor income, h, and the lump-sum transfer from the government (expressed in terms of the general good), T = φt (Mt+1 − Mt ). Recall that the rate of return of real balances is φt+1 /φt = γ −1 . Hence, in order to hold z units of real balances in the next period,

Money and Credit

199

the buyer must acquire γ z units of real balances in the current period. Substituting x − h from (8.2) into (8.1), we get b W b (z, −b) = z − b + T + max + βV (z ) . (8.3) −γ z z ≥0

As before, the value function is linear in the buyer’s current portfolio, and the buyer’s choice of real balances is independent of his current portfolio. The value function of a seller who holds z units of real balances and b IOUs at the beginning of the CM period is denoted by W s (z, b). Since sellers have no incentive to accumulate real balances at night, this value function is given by W s (z, b) = z + b + βV s ,

(8.4)

where V s is the value function of a seller at the beginning of the next period. Recall that sellers do not receive transfers in the CM. Consider now a bilateral match in the DM2 between a buyer holding z units of real balances and a seller. The buyer is anonymous and cannot use credit. Hence he can transfer at most z units of real balances to the seller in exchange for afternoon output. We assume that the buyer makes a take-it-or-leave-it offer. Because real balances enter the beginning-ofthe-night value functions of the buyer and seller in a linear fashion, the buyer’s offer to the seller is given by the solution to the following simple problem: max u(q2 ) − d2 q2 ,d2

subject to

− c(q2 ) + d2 ≥ 0

and d2 ≤ z, where the ﬁrst inequality represents the seller’s participation constraint and the second is a feasibility constraint. The solution to this maximization problem is q2 = min q∗2 , c−1 (z) , (8.5) d2 = c(q2 ),

(8.6)

that is, the buyer purchases the efﬁcient level of output if he has sufﬁcient real balances; otherwise he spends all of his balances on output.

200

Chapter 8

The value function of a buyer with z units of real balances and b units of debt at the beginning of DM2 is V2b (z, −b) = σ2 u q2 (z) − c[q2 (z)] + W b (z, −b).

(8.7)

Similarly the value function of a seller is V2s (z, b) = W s (z, b). We can now turn to the buyer’s bargaining problem in DM1. The buyer who holds z units of real balances makes a take-it-or-leave-it offer that solves max υ(q1 ) + V2b (z − d1 , −b1 )

q1 ,d1 ,b1

subject to

− ψ(q1 ) + W s (d1 , b1 ) ≥ W s (0, 0)

and d1 ≤ z. Using the linearity of W s and the seller’s participation constraint at equality, d1 + b1 = ψ q1 , the buyer’s problem can be simpliﬁed to max υ ◦ ψ −1 (d1 + b1 ) + V2b (z − d1 , −b1 ) , d1 ,b1

(8.8)

The ﬁrst-order conditions, ignoring the constraint d1 ≤ z, are υ (q1 ) − 1 ≤ 0, ψ (q1 ) υ (q1 ) u (q2 ) − σ2 − 1 − 1 ≤ 0, ψ (q1 ) c (q2 )

“ = ’’ if b1 > 0,

(8.9)

“ = ’’ if d1 > 0.

(8.10)

If q2 < q∗2 , then it is immediate that d1 = 0. If the buyer is constrained by his real balances in the DM2, he should not spend them in the DM1 and he should trade with credit only. If q2 = q∗2 , then the buyer is indifferent between using credit or cash as long as he keeps enough real balances to purchase q∗2 in DM2. So, with no loss in generality, we can assume that in the DM1 the buyer trades with credit only. From (8.9), it is immediate that q1 = q∗1 . We can now write the value function of a buyer at the beginning of a period: V b (z) = σ1 υ(q∗1 ) + V2b z, −ψ(q∗1 ) + (1 − σ1 )V2b (z, 0) . (8.11)

Money and Credit

201

Using the linearity of V2b with respect to its second argument, and substituting V2b (z, 0) from its expression given by (8.7) into (8.11), we obtain V b (z) = σ1 υ(q∗1 ) − ψ(q∗1 ) + V2b (z, 0) (8.12) ∗ ∗ b = σ1 υ(q1 ) − ψ(q1 ) + σ2 u[q2 (z)] − c q2 (z) + z + W (0, 0) . If we substitute V b (z) from (8.12) into (8.3), then the buyer’s portfolio problem in the CM can be represented by max −iz + σ2 u[q2 (z)] − c q2 (z) , (8.13) z≥0

where i ≡ (γ − β)/β. Note that the buyer’s real balances only affects his surplus in the DM2. The ﬁrst-order (necessary and sufﬁcient) condition for problem (8.13) is u (q2 ) i = 1+ . c (q2 ) σ2

(8.14)

This expression for the output traded in the DM2 is identical to the one we derived in chapter 6.1, equation (6.8). ∗ An equilibrium is a list (q1 , q2 , b1 , d2 , {φt }) that solves q1 = q1 , b1 = ∗ ψ q1 , (8.6), (8.14), and φt = c(q2 )/Mt . The allocation is dichotomic in the sense that the output traded in the DM1, q1 , is independent of both the quantity traded in the DM2, q2 , and the value of money, φt . As well, when inﬂation increases, q1 is unaffected and remains at the efﬁcient level, while q2 decreases; see equation (8.14). So there are no interactions between the DM1 and the DM2. Another noteworthy feature of the model is that in the DM1, a fraction σ1 of the buyers issue debt, while at the same time they hold a positive amount of money. Credit is a preferred means of payment because it involves no opportunity cost. However, credit can only be used in transactions when agents’ identities are known and debt contracts can be enforced. Buyers will hold money, even though it is more costly than credit, because it allows them to consume in the DM2 when they are anonymous. Finally, as the cost of holding money, i, approaches zero, the quantity traded in the DM2 approaches its efﬁcient level, q∗2 . When the cost of holding money is exactly equal to zero, there is no cost associated with

202

Chapter 8

holding real balances, and buyers will be indifferent between trading with money and credit in the DM1. 8.2 Costly Record Keeping We now consider an environment where money and credit coexist, and monetary policy affects the composition of monetary and credit transactions. The model is similar to the one in chapter 5.1.3, where a typical period has a decentralized market, DM, a competitive night market, CM, and the gains from trade in the DM vary across bilateral matches. To this environment we add a costly record-keeping technology. Hence credit transactions are feasible but costly. The instantaneous utility function of a buyer is given by U b = εu(q) + x − h, where ε ∈ R+ is a match-speciﬁc preference shock. The preference shock, ε, is drawn from a cumulative distribution, F(ε), with support [0, εmax ]. Matched agents in the DM have the option to record a credit transaction at a utility cost of ζ > 0. We assume that the buyer incurs this cost. This cost could capture the resources needed to authenticate both the buyer’s identity and his IOU. If a credit transaction is recorded in the DM, we assume that its repayment is enforced at night. The value functions for buyers and sellers at the beginning of the CM, W b (z, −b) and W s (z, b), are given by equations (8.3) and (8.4), respectively. Consider a match in the DM between a buyer with a match-speciﬁc preference shock ε holding z units of real balances, and a seller. We assume that the buyer makes a take-it-or-leave-it offer to the seller. Owing to the linearity of the buyer’s and seller’s CM value functions, the terms of trade, q, b, d , are given by the solution to max εu(q) − d − b − ζ I{b>0} q,d,b

subject to

− c(q) + d + b ≥ 0

and d ≤ z, where I{b>0} = 1 if b > 0 and I{b>0} = 0, otherwise. The buyer chooses his consumption, q, the amount of real balances to transfer to the seller, d, and the size of the loan, b. If the buyer chooses to use credit as a means of payment, he must incur the ﬁxed cost ζ due to record keeping. If the buyer incurs the ﬁxed cost, then the solution is q = q∗ε with

Money and Credit

203

d + b = c(q∗ε ), where q∗ε solves εu q∗ε = c q∗ε . Without loss of generality, we assume that in this case the buyer only uses credit in the transaction. If the buyer does not incur the ﬁxed cost to use credit, then q = qε (z) = min q∗ε , c−1 (z) and d = c(q). In other words, if he has enough real balances, the buyer purchases the efﬁcient level of output for his particular preference shock; otherwise, he spends all of his real balances. Consequently the buyer’s surplus from a trade match in the DM is Sb (z, ε) = max εu(q∗ε ) − c(q∗ε ) − ζ , εu qε (z) − c qε (z) .

(8.15)

In ﬁgure 8.2 we illustrate the utility gain to the buyer from using a credit arrangement. The gray area represents the set of utility levels (us = −c(q) + d for the seller and ub = εu(q) − d for the buyer) that are incentive feasible when the buyer uses money only. The dashed line is the Pareto

us

u b + u s = εu (qε* ) − c(qε* )

Bargaining set with money

z < c (qε*)

ub Utility gain from using credit Figure 8.2 Utility gain from using credit

204

Chapter 8

frontier of the bargaining set if the buyer uses credit, which excludes the ﬁxed cost, ζ , associated with record keeping and enforcement. This Pareto frontier is linear because the match surplus is maximum and equal to εu(q∗ε ) − c(q∗ε ). The gain for a buyer using credit can been seen on the horizontal axis: it is the distance between the intercepts of the two Pareto frontiers, the one with money only and the one with credit. Note that Sb (z, ε) is increasing in ε; that is, both terms in the maximization problem (8.15) increase with ε. We represent each of these terms as a function of ε in ﬁgure 8.3. From an envelope argument, ∗ ). The slope of the second is u q (z) the slope of the ﬁrst term is u(q ε ε since u qε (z) + εu qε (z) − c qε (z) ∂qε (z) /∂ε = 0, i.e., εu qε (z) = c qε (z) if qε (z) = q∗ε , and qε (z) = c−1 (z), and hence ∂qε (z) /∂ε = 0, if qε (z) < q∗ε . Let ε¯ denote the value of ε such that c(q∗ε¯ ) = z or, equivalently, ε¯ u c−1 (z) = c c−1 (z) , i.e., ε¯ is a threshold for the idiosyncratic preference shock for a given z, below which the buyer has enough real balances to purchase the efﬁcient level of DM output. For all ε < ε¯ , u qε (z) = u(q∗ε ), which implies that the slopes of the two terms in the maximization problem (8.15) are equal. For all ε > ε¯ , u qε (z) < u(q∗ε ), and the slope of the second term in the maximization problem (8.15) is independent of ε and lower than the slope of the ﬁrst term. When ε = 0, the ﬁrst term is equal to −ζ , while the second is equal to zero. For ε > ε¯ sufﬁciently large,

εu(q∗ε ) − c(q∗ε ) − ζ − εu qε (z) − c qε (z) > 0,

since for large ε, qε is negligible compared to q∗ε , and hence the left side of the inequality goes to inﬁnity. Consequently there exists a threshold εc > ε¯ above which the buyer uses credit as means of payment—i.e., the ﬁrst term in the maximization problem (8.15) exceeds the second—and below which he uses money. This threshold is given by εc u(q∗εc ) − c(q∗εc ) − ζ = εc u c−1 (z) − z. (8.16) Graphically the ﬁrst term in the maximization problem (8.15) intersects the second term from below at ε = εc ; see ﬁgure 8.3. It should be (re)emphasized that the value of the threshold, εc , is for a given level of real balances, z. From (8.16), εc increases with z, i.e., dεc [εc u c−1 (z) /c c−1 (z) ] − 1 = > 0, dz u q∗εc − u c−1 (z)

Money and Credit

205

Buyer’s surplus

εu(qε*) − c(qε*) − ζ

ε

εu[qε (z)]− c[qε (z)]

ε

εc

−ζ Trades with money

Trades with credit

Figure 8.3 Credit versus monetary trades

since q∗εc > c−1 (z). Graphically, as z increases, ε¯ increases, and for all ε > ε¯ , the second term of the maximization problem (8.15) moves upward. Buyers increase their surplus by holding more real balances in all trades where they don’t trade the efﬁcient quantity. Consequently the two terms intersect at a larger value of ε. As buyers hold more real balances, the fraction of trades conducted with credit decreases: money and credit are substitutes. Using the linearity of W b , the value of being a buyer at the beginning of the period, V b (z), is V b (z) = σ

0

εmax

Sb (z, ε)dF(ε) + W b (z, 0).

(8.17)

206

Chapter 8

With probability σ the buyer meets a seller, and he draws a realization for the preference shock from the distribution F(ε). The buyer enjoys a surplus of Sb (z, ε), given by (8.15), which depends on both the buyer’s real balances and the match speciﬁc component. Substituting V b (z) from (8.17) into (8.3), and simplifying, we get max −iz + σ z≥0

0

εmax

Sb (z, ε)dF(ε) .

(8.18)

The buyer chooses his real balances in order to maximize his expected surplus in the DM, where the expectation is with respect to the random preference shock, minus the cost of holding real balances. The objective function in (8.18) is continuous and for all i > 0 the solution to (8.18) must lie in the interval [0, c(q∗εmax )]. If z > c(q∗εmax ), then the surplus is maximum in all matches and independent of z. But by reducing z, the buyer can reduce his cost of holding real balances without affecting his expected surplus in the DM. Since a continuous function is being maximized over a compact set, there exists a solution to (8.18). An equilibrium corresponds to a pair (εc , z) that solves (8.16) and (8.18) and can be determined recursively: a value for z is determined independently by (8.18), and given this value for z, (8.16) determines a unique εc . We now investigate the effects that monetary policy has on the use of ﬁat money and credit as means of payment. The ﬁrst-order (necessary but not sufﬁcient) condition associated with (8.18) is εc (z) −1 εu c (z) − 1 dF(ε). i=σ (8.19) c c−1 (z) ε¯ (z) From (8.19) real balances have a liquidity return when the realization of the preference shock is not too low—so that the buyer’s budget constraint in the match is binding—and when the preference shock is not too high—so that it is not proﬁtable for buyers to use credit—i.e., when ε¯ (z) < ε < εc (z). Suppose that inﬂation and, hence, the cost of holding money, i, increases. This implies that the right side of (8.19) must also increase. One would conjecture that an increase in inﬂation decreases real money balances z. In order to check this conjecture, consider two monetary policies with resulting nominal interest rates i and i such that i < i . (Recall that i is referred to as a nominal interest rate because it is the interest rate paid by an illiquid nominal bond that can only be traded in the CM.)

Money and Credit

207

Let z and z denote the solutions of (8.18) for i and i , respectively. From (8.18) we have εmax εmax −iz + σ Sb (z, ε)dF(ε) ≥ −iz + σ Sb (z , ε)dF(ε), (8.20) −i z + σ

0 εmax

0

0

Sb (z , ε)dF(ε) ≥ −i z + σ

0

εmax

Sb (z, ε)dF(ε).

(8.21)

These inequalities imply that i z − z ≤ σ

Sb (z, ε) − Sb (z , ε) dF(ε) ≤ i z − z ,

εmax 0

which in turn imply z ≥ z since, by assumption, i < i and i z − z ≤ i z − z , from the inequality above. Moreover it can be checked that z = z when i < i is inconsistent with (8.19). Hence z > z . An increase in inﬂation reduces buyers’ real balances and increases the use of costly credit. As the cost of holding real balances approaches zero, it is immediate from (8.18) that real balances approach c(q∗εmax ) and buyers ﬁnd it proﬁtable to trade with money only. 8.3 Strategic Complementarities and Payments So far we have described environments where buyers make offers to sellers and choose the means of payment that will be used in bilateral meetings. Typically, however, in order to be able to accept credit, sellers must invest ex ante—i.e., before trades take place—in a technology that authenticate buyers’ IOUs. Buyers will form rational expectations about sellers’ investment decisions and choose the amount of means of payment(s) to carry into meetings. As we will see, these decisions made by buyers and sellers create strategic complementarities for payment choices and network-like externalities. The model with network externalities is similar to that of the previous section, but modiﬁed in the following ways: First, for simplicity, assume all matches are identical, i.e., ε = 1. Second, and more substantially, assume that it is the seller who invests in the record-keeping technology and that this investment is undertaken at the beginning of the DM before matches are formed. The utility cost to invest in this technology is ζ > 0. The pricing mechanism must be changed from the previous section to one that permits sellers to extract a fraction of the match surplus; otherwise, sellers could not recover their ex ante investment

208

Chapter 8

costs and would have no incentive to invest in the record-keeping technology. We will adopt the proportional bargaining solution described in chapter 4.2.3, where the buyer receives a constant share θ ∈ [0, 1) of the match surplus, while the seller gets the remaining share, 1 − θ > 0. We start by describing the determination of the terms of trade in a bilateral match in the DM, depending on whether sellers have invested or not in the record-keeping technology. Consider ﬁrst a match between a buyer holding z units of real balances and a seller who has invested in the technology. The terms of trade are given by the solution to the following problem: max u(q) − d − b (8.22) q,d,b

subject to

− c(q) + d + b =

1−θ u(q) − d − b θ

and d ≤ z,

(8.23) (8.24)

where we have used the linearity of the buyer’s and seller’s value functions with respect to their wealths. According to problem (8.22)– (8.24), the buyer maximizes his utility of consuming the DM good net of the transfer of real balances, d, and IOUs, b, subject to the constraints that (1) the seller’s payoff is equal to (1 − θ )/θ times the buyer’s payoff and (2) the buyer cannot transfer more money than he has. Since b is unconstrained—buyers can borrow as much as they want in the DM—it should be obvious that d ≤ z never constrains the purchase of q. When sellers have invested in the record-keeping technology, buyers can ﬁnance all of their daytime purchases with credit alone. The output produced in the DM will be at the efﬁcient level, q = q∗ , and d + b = (1 − θ) u(q∗ ) + θ c(q∗ ), i.e., the seller gets the fraction 1 − θ of the match surplus. Without loss of generality, assume that d = 0, so that the trade is conducted with credit only. Consider next the case where the seller has not invested in the recordkeeping technology. The terms of trade are still determined by the problem (8.22)–(8.24), but with the added constraint that b = 0. If z ≥ (1 − θ ) u(q∗ ) + θ c(q∗ ), then the buyer will have sufﬁcient money balances to purchase the efﬁcient level of output and q = q∗ ; otherwise, the level of DM output, q(z), will satisfy z = z(q) ≡ (1 − θ) u(q) + θ c(q), where q (z) < q∗ .

(8.25)

Money and Credit

209

We now turn to the seller’s decision to invest in the record-keeping technology. We consider situations where all buyers hold the same real balances, z. It is optimal for a seller to invest in the technology if σ (1 − θ ) u(q(z)) − c(q(z)) ≤ σ (1 − θ ) u(q∗ ) − c(q∗ ) − ζ ,

(8.26)

where we have used the linearity of the value function of the seller in the CM. The left side is the seller’s expected payoff if he does not invest in the record-keeping technology. In this case the seller can only accept the buyer’s real balances and will not provide credit. The right side is the seller’s expected payoff if he invests in the technology to accept IOUs. From (8.26) the ﬂow cost to invest in the record-keeping technology must be less than the increase in the seller’s expected surplus associated with accepting credit instead of money. left side in z: The of (8.26) is increasing ∗ ∗ ∗ it equals 0 if z = 0 and σ (1 − θ) u(q ) − c(q ) if z ≥ (1 − θ ) u(q ) + θc(q∗ ). Consequently, if ζ < σ (1−θ ) u(q∗ ) − c(q∗ ) , then there exists a threshold zc > 0 for the buyer’s real balances below which sellers invest in the record-keeping technology. This threshold is given by the solution to u q(zc ) − c q(zc ) = u(q∗ ) − c(q∗ ) − Let be the measure technology. Then ⎧ ⎧ ⎨ =1 ⎨ < ∈ [0, 1] if z = ⎩ ⎩ =0 >

ζ . σ (1 − θ )

(8.27)

of sellers who invest in the record-keeping ⎫ ⎬ ⎭

zc .

(8.28)

The seller’s reaction function is depicted in ﬁgure 8.4. It is a step function that is decreasing with the buyer’s real balances. As buyers hold more money, sellers have less incentives to invest in the costly record-keeping technology. In ﬁgure 8.5 we illustrate the gains from using credit for the buyer and the seller. The gray area represents the set of surpluses that are incentive feasible when the buyer uses money only, while the dashed line is the Pareto frontier of the bargaining set if the buyer uses credit. The outcome to the proportional bargaining problem is given by the intersection of the line us /ub = (1 − θ )/θ with the relevant Pareto frontier. The seller’s gain is the vertical distance between the intersections the Pareto frontiers with the line (1 − θ )/θ, and the gain for the buyer is given by the horizontal distance. It follows immediately that the buyer’s gain from

210

Chapter 8

z

z0 Sellers’ reaction function

zc

Buyers’ reaction function

Λc =

σθ − (1 − θ )i σθ

1

Figure 8.4 Buyers’ and sellers’ reaction functions

using credit is θ/(1−θ ) times the seller’s gain. The fact that the seller cannot appropriate the entire gain from using the credit technology, which requires an ex ante investment, creates a standard holdup problem. Given the seller’s decision to invest in the record-keeping technology, (8.28), we now consider the buyer’s decision to hold real balances. Following a similar line of reasoning as in chapter 6.3, the buyer’s decision problem is given by max −iz + σ (1 − )θ u[q(z)] − c q(z) + σ θ u(q∗ ) − c(q∗ ) . (8.29) z≥0

The buyer chooses his real balances in order to maximize his expected surplus in the DM, net of the cost of holding real balances. The buyer obtains a fraction θ of the entire match surplus in all meetings. From (8.29) the buyer’s surplus depends on his real balances only if the seller does not have the recording-keeping technology, an event that occurs with probability 1 − . If the seller has the technology to accept credit, an event that occurs with probability , the match surplus is at its

Money and Credit

211

us

us =

Utility gain to the seller from using credit

1−θ b u θ

u b + u s = u (q * ) − c (q * )

ub Utility gain to the buyer from using credit Bargaining set with money (z< c(q * )) Figure 8.5 Gains from using costly credit

maximum and the quantity traded is q∗ . The ﬁrst-order condition for problem (8.29) is [σ (1 − )θ − i (1 − θ)] u (q) − [i + σ (1 − )] θc (q) ≤ 0, (1 − θ ) u (q) + θ c (q)

(8.30)

and holds with an equality if z > 0. If z > 0, then the numerator of (8.30) will equal to zero, and u q [i + σ (1 − )] θ = . [i + σ (1 − )] θ − i c q

(8.31)

The right side of (8.31) is increasing with , which implies that an increase in will decrease q, and hence z. Therefore, as illustrated in the ﬁgure 8.4, the buyer’s choice of real balances is decreasing in . Intuitively, if it is more likely to ﬁnd a seller who accepts credit, then

212

Chapter 8

money is needed in a smaller fraction of matches, and since it is costly to hold money, buyers will ﬁnd it optimal to hold fewer real balances. Moreover there is a critical value for above which buyers hold no real balances, and this happens when the denominator of equation (8.31) is equal to zero, or when c = [σ θ − (1 − θ)i] /σ θ , where c > 0 if i < σ θ/(1 − θ ). A stationary symmetric equilibrium is a pair (z, ) that solves (8.28) and (8.29). If ζ > σ (1 − θ ) u(q∗ ) − c(q∗ ) , then it is a strictly dominant strategy for sellers not to invest in the record-keeping technology. In this case there is a unique equilibrium ∗ where = 0. Let’s now ∗ consider the case where ζ < σ (1 − θ ) u(q ) − c(q ) . From (8.27), zc ∈ 0, (1 − θ ) u(q∗ ) + θ c(q∗ ) . Let z0 be the solution to (8.29) when = 0, i.e., z0 is the buyer’s money holdings if no seller invests in the recordkeeping technology. If z0 > zc , which happens if i is sufﬁciently low, then there are multiple equilibria. This can be seen in ﬁgure 8.4, where the buyers’ and sellers’ reaction functions intersect three times. There exists a pure monetary equilibrium with = 0 and z > 0; a pure credit equilibrium, with = 1 and z = 0; and a “mixed’’ monetary equilibrium, where buyers use both credit and money, accumulating zc > 0 real balances. A fraction 1 − ∈ (0, 1) of sellers accept only money, while other sellers, ∈ (0, 1) of them, are willing to accept both money and credit. The multiplicity of equilibria arises from the strategic complementarities between the buyers’ decisions to hold real balances and the sellers’ decisions to invest in the record-keeping technology. To understand this, suppose, for example, that buyers believe that all sellers have invested in the record-keeping technology. Then they have no need to hold real balances. But, if sellers think that buyers are not holding any money, then they have an incentive to invest in the record-keeping technology, provided, of course, that the cost of the technology is not too high. And, for exactly the same fundamentals, buyers may anticipate that sellers choose not to invest in the record-keeping technology. In this situation, buyers will hold a large quantity of real balances. But if sellers believe that buyers hold enough real balances, then they do not have an incentive to invest in the record-keeping technology. Given the existence of multiple equilibria, history is able to explain why seemingly identical economies can end up with different payment systems. Consider, for example, an economy with a low inﬂation where agents play the pure monetary equilibrium. Suppose that this economy subsequently experiences a period of high inﬂation. In terms of ﬁgure 8.4, the buyer’s reaction function will shift downward and,

Money and Credit

213

provided that the increase in the inﬂation rate is sufﬁciently large, z0 < zc . With this higher level of inﬂation, the equilibrium is unique, and all sellers invest in the record-keeping technology, = 1. Suppose that the high-inﬂation episode is temporary, and inﬂation reverts back to its initial low level; will agents go back to playing the pure monetary equilibrium? Since the pure credit equilibrium is still an equilibrium, one can imagine that agents will continue to coordinate on this equilibrium after the inﬂation rate reverts back to its initial level. Interestingly, even though the change in inﬂation was temporary, the change in the payment system has become permanent: the payment system exhibits hysteresis. We conclude this section by turning to some normative considerations. When there are multiple equilibria, which one is preferred from the society’s viewpoint? Society’s welfare is measured by the surpluses of all matches in the DM minus the real resource cost incurred by sellers to accept credit, W = σ u(q∗ ) − c(q∗ ) + σ (1 − ) u q(z) − c q(z) − ζ . Consider a case where z0 is greater than but close to zc . There is a pure monetary equilibrium with z = z0 , = 0, and social welfare is W0 = σ u q(z0 ) − c q(z0 ) . There is also a pure credit equilibrium with = 1, and social welfare is W1 = σ u(q∗ ) − c(q∗ ) − ζ . Then, given the deﬁnition of zc in (8.27), ζ ≈ σ (1 − θ ) u(q∗ ) − c(q∗ ) − u(q(z0 )) − c(q(z0 )) < σ u(q∗ ) − c(q∗ ) − u(q(z0 )) − c(q(z0 )) , where we get the strict inequality because θ > 0. In this case the difference in the surpluses associated with credit and monetary transactions strictly exceeds the cost of investment in the record-keeping technology. Hence, W1 > W0 , which means that the pure monetary equilibrium is dominated, from a social welfare perspective, by the pure credit equilibrium. However, the socially inefﬁcient monetary equilibrium can prevail because of a holdup externality. If a seller decides to adopt the technology to accept credit, he incurs the full cost of the technology adoption, but he only receives a fraction 1 − θ < 1 of the increase in the match surplus. Hence sellers fail to internalize the effect of the credit technology on buyers’ surpluses, which can lead to excess inertia in the decision to adopt the record-keeping technology.

214

Chapter 8

Consider next the case where the cost of holding money, i, is close ∗ ∗ to zero. In this situation z0 will be close to θ c(q ) + (1 − θ)u(q ) and ∗ ∗ ∗ q(z0 ) ≈ q . Hence W0 ≈ σ u(q ) − c(q ) . Provided that ζ > 0, the pure monetary equilibrium dominates the pure credit equilibrium from a social welfare perspective. The resources allocated to the record-keeping technology are “wasted’’ in the sense that a monetary equilibrium avoids costs associated with record-keeping and provides an allocation that is almost as good as the credit allocation. Still, if ζ < σ (1−θ) u(q∗ ) − c(q∗ ) , agents can end up coordinating on the inferior (credit) equilibrium because of the strategic complementarities between the buyers’ and sellers’ choices. 8.4 Credit and Reallocation of Liquidity In this section we describe an economy where credit is used to reallocate liquidity from agents with an excess supply of money to agents with an excess demand for money. To do this, we introduce some heterogeneity in terms of agents’ liquidity needs: some buyers need more money than others to trade in the DM. We reinterpret the matching shocks in the DM as preference shocks. With probability σ , a buyer has a positive marginal utility of consumption in the DM, while with the complement probability, 1 − σ , his marginal utility of consumption is zero. These shocks are realized at the beginning of a period before agents are matched and are independent across buyers and time. In the DM, after the preference shocks are realized, each buyer gets matched with a seller with probability one. It should be clear that this model is isomorphic to the one we have been studying so far. If buyers are unable to borrow or lend before being matched, then when the money supply is constant, the quantity traded in the DM in a stationary monetary equilibrium solves the familiar equation u (q) r = 1+ . c (q) σ

(8.32)

An important feature of (8.32) is that the quantities traded decrease if buyers face a higher risk of a negative preference shock—i.e., if σ is lower—because a buyer’s money holdings are unproductive more often. We now modify the environment by allowing a loan market to operate at the beginning of each period, after preference shocks are realized

Money and Credit

215

MORNING

DAY (DM)

NIGHT (CM)

Loan market

Pairwise meetings

Competitive market for money and general goods

Preference shocks Figure 8.6 Timing

but before bilateral matches are formed. The sequence of events is represented in ﬁgure 8.6. In the loan market, agents cannot produce, but they can buy and sell loans, i.e., they can borrow or lend money for a promise to repay or receive money in the subsequent CM. The nominal interest rate on a loan is i : a loan of one dollar is repaid in the subsequent CM for 1 + i dollars. Finally, there is a technology to enforce the repayment of loans contracted at the beginning of a period. However, the IOU that represents a loan does not circulate in the DM because it cannot be authenticated in that market. We denote as the size of a loan. If > 0, then the buyer is a creditor, and if < 0, then the buyer is a debtor. Deﬁne m as the amount of money held after the loan market closes. The expected lifetime utility of a buyer who has positive marginal utility of consumption in the DM who holds m units of money and dollars in loans is ˆ b (m , ) = u(q) − c(q) + W b (m , ), V

(8.33)

where c(q) = min c(q∗ ), φm since we assume that buyers make a takeit-or-leave-it offer to the seller. The value function of the buyer in the CM, W b (m , ), is given by b W b (m , ) = φm + (1 + i )φ + max + βV (m ) , (8.34) −φm m ≥0

where V b (m) is the expected utility of the buyer at the beginning of a period before his preference shock is realized. According to (8.34), a buyer in the CM can sell each unit of money at the competitive price φ, and he receives 1 + i dollars for each unit of loan he owns. The choice of money holdings for the next period, m , is independent of both the size

216

Chapter 8

of the loan, , and the amount of money held by the buyer, m . Hence W b (m , ) = φm + (1 + i )φ + W b (0, 0). The expected utility of the buyer who holds m units of money at the beginning of a period before his preference shock is realized, V b (m), satisﬁes ˆ b (m + d , − d ) + (1 − σ ) max W b (m − s , s ), V b (m) = σ max V s ≤m

d ≥0

(8.35)

where we interpret d ≥ 0 as the demand of loans and s ≥ 0 as the supply of loans. With probability σ the buyer receives a positive preference shock and wants to consume in the DM. In this case he demands a loan of size d . With probability 1 − σ the buyer does not want to consume, but he is willing to lend part or all of his money holdings. Hence, if a buyer is a borrower, m = m + d , and if he is a lender, m = m − s ≥ 0. ˆb −V ˆ b ≤ 0, From (8.35) the optimal demand for loans satisﬁes V m ˆ b (V ˆ b ) represents the derivative of with a strict equality if d > 0. (V m ˆ b with respect to its ﬁrst (second) argument.) From (8.33) the beneﬁt V ˆ b = [u (q)/c (q)]φ, while the cost from borrowing one unit of money is V m ˆ b = (1 + i ) φ. Hence is V u (q) − 1 − i ≤ 0, c (q)

“ = ’’

if d > 0,

(8.36)

where c(q) = min c(q∗ ), φ(m + d ) . Notice that if the solution to (8.36) is interior, the quantity of money held by the buyer before entering the DM is independent from his money holdings at the beginning of the period. If the solution to (8.36) is interior, then ˆ b (m + d , − d ) = max u ◦ c−1 (φm ) − (1 + i )φm max V d ≥0

m

+ (1 + i )φm + W b (0, 0). From (8.35) the individual supply of loans satisﬁes s = m whenever i > 0 and s ≥ 0 if i = 0. Consequently max W b (m − s , s ) = (1 + i )φm + W b (0, 0). s ≤m

We will check later that sellers have no strict incentives to borrow or lend.

Money and Credit

217

i Ld = σ

L = (1 − σ)

d

s

s

u ' c −1 ϕM −1 c ' c −1 (ϕM)

r

(1 −σ )M

σ c(q *) − ϕM ϕ

Ld, Ls

Figure 8.7 Equilibrium of the loan market

The equilibrium of the loan market is represented graphically in ﬁgure 8.7. The aggregate demand for loans, Ld = σ d , is downward-sloping because as the interest rate on loans increases, the individual demand for loans decreases. If u ◦ c−1 (φM) / c ◦ c−1 (φM) ≤ 1 + i , then the beneﬁt of borrowing is less than its cost, and buyers do not ﬁnd it proﬁtable to borrow funds, i.e., Ld = d = 0. If i = 0, then buyers will borrow enough money to trade q∗ in the DM. The size of an individual loan is greater than or equal to [c(q∗ ) − φM]/φ. The aggregate supply of loans, Ls = (1 − σ ) s , is vertical at Ls = (1 − σ )M. Let us turn to the demand for money in the CM. From (8.34) and (8.35) the optimal choice of money holdings satisﬁes

u (q) φ = β σ φ + (1 − σ )(1 + i )φ . (8.37) c (q) As usual, the left side of (8.37) represents the cost of accumulating an additional unit of money, while the right side of (8.37) is the beneﬁt from

218

Chapter 8

holding an additional unit of money. The beneﬁt has two components. With probability σ , the buyer has positive marginal utility of consumption, in which case he can use his marginal unit of money to buy φ/c (q) units of output in the DM. With probability 1 − σ , the buyer does not want to consume, in which case he can lend his unit of money for 1 + i units of money in the CM. So, compared to the environment where there is no borrowing or lending, the buyer can obtain an additional return on his money holdings if he does not have an opportunity to consume. This additional return tends to make money more valuable. We now show that the loan market will be active. To see this, suppose that instead s = d = 0. Since d = 0, (8.37) represents the demand for money, as does (8.32), and buyers with a low marginal utility of consumption will not lend their money balances, i.e., s = 0, only if i = 0. This implies that u (q) r u (q) − 1 − i = − 1 = > 0. c (q) c (q) σ But this inequality violates (8.36). Clearly, at i = 0, buyers that have a positive marginal utility of consumption have an incentive to borrow some money in order to relax their budget constraint in a bilateral match. As a result the loan market is active: i > 0 and s = m. From (8.36) and (8.37), we can solve for the quantities traded in the DM and the interest rate on loans, u (q) = 1 + r, c (q)

(8.38)

i = r.

(8.39)

Acomparison between (8.32) and (8.38) reveals that the quantities traded when the loan market is active are greater than the quantities traded when there is no loan market to reallocate the liquid assets. This implies that the existence of a loan market after preference shocks are realized but before agents are matched in the DM is welfare improving. So the use of credit is essential to reallocating the liquidity that is needed to trade in the DM. Notice also that the allocation obtained with an active loan market is the one that would prevail if buyers learn their preference shocks in the CM, at the time when they choose their money holdings. In this situation there would be no precautionary demand for money holdings.

Money and Credit

219

The market clearing for loans requires that σ d = (1 − σ ) s . Since, in equilibrium, s = M, the size of the buyer’s loan is

1−σ M. d = σ Since we assume that buyers make a take-it-or-leave-it offer to sellers, the quantity traded in the DM solve c(q) = φ(M + d ). Consequently the value of money in equilibrium is φ=

σ c(q) . M

(8.40)

The stock of money per active buyer is M/σ . As σ increases, the quantity of money per active buyer decreases and hence the value of money increases. According to (8.39) the interest rate on a loan is exactly equal to the rate of time preference, r. Buyers with a high marginal utility of consumption are willing to pay up to the rate of time preference to borrow an additional unit of money, which is the marginal beneﬁt of money holdings in the DM. We can now check that sellers have no strict incentives to participate in the loan market. It is clear that sellers do not want to borrow money at a positive interest rate since they don’t need it in the DM. And they are indifferent in terms of accumulating money or not in the CM and lending it in the next DM at the interest r. In the presence of a growing money supply, it can be checked that the nominal interest on the loans is i = i ≡ (γ − β)/β ≈ (γ − 1) + r. According to the Fischer effect, an increase in the inﬂation rate, γ − 1, has a one-to-one effect on the nominal interest rate. 8.5 Short-Term and Long-Term Partnerships So far we have assumed that credit transactions that are recorded during the day (or DM) can be enforced at night (or CM). In this section we assume that there is no enforcement technology and buyers cannot commit to repay their debt. Therefore debt contracts must be self-enforcing. If there can be repeated interactions with a seller, a buyer will want to generate a reputation for paying his debts, and the buyer’s desire for this reputation results in contracts being self enforced.

220

Chapter 8

We allow for the possibility of both short-term and long-term partnerships, and model this by combining the pure monetary environment with short-term partnerships in chapter 4.1 with the long-termpartnership environment described in chapter 2.4. At the beginning of a period, unmatched agents can enter into a long-term trade match with probability σL or a short-term trade match with probability σS . A shortterm match corresponds to a situation where the buyer and the seller know they will not meet again in the future. In contrast, in a long-term match the buyer and the seller have a chance to stay together for more than one period. We assume that 0 < σL + σS < 1. A short-term match is destroyed with probability one at the beginning of the CM, while a long-term match will be exogenously destroyed with probability λ < 1 at the beginning of the CM. In addition either party to a long-term match that is not exogenously destroyed can always choose to terminate the relationship at the beginning of the DM. The timing of the relevant events are described in ﬁgure 8.8. Buyers enter the day market, DM, either attached, i.e., in a long-term trade match, or unattached. At this time matched buyers and sellers in a longterm partnership simultaneously decide whether to continue or split apart. Unattached buyers and sellers participate in a random-matching process. Since the measures of buyers and sellers are equal, there are also equal measures of unattached buyers and unattached sellers. After the matching process is completed, all matched sellers—those in either a long-term or short-term relationship—produce the DM good for buyers. The night period begins with buyers who are in a long-term partnership, producing the general good for sellers if trade was mediated by credit in the previous DM. A fraction λ of buyers in the long-term partnership then realize a shock that dissolves the relationship they have with their currently matched seller, and all of the short-term partnerships are destroyed. This is followed by the opening of the CM, where the general good and money are traded. In terms of pricing mechanisms, DAY

A fraction σ (σs) Matched sellers Matched buyers of unmatched agents produce q (qs) in long-term find a long-term in long-term matches (short-term) match. (short-term) matches. produce y .

Figure 8.8 Timing of a representative period

NIGHT

A fraction λ Agents can of long-term readjust their matches are money holdings. destroyed.

Money and Credit

221

we assume that buyers make take-it-or-leave-it offers to sellers in the DM, and that the night market is competitive, where one unit of money trades for φt units of the general good. We will restrict our attention to a particular class of equilibria that exhibit two features. First, money is valued, but is only used in short-term trade matches. Second, the buyer’s incentive-compatibility constraint in long-term matches—that the buyer is willing to produce the general good for the seller to extinguish his debt obligation—is not binding. This latter assumption implies that a buyer in a long-term partnership will be able to purchase the efﬁcient quantity of the DM good, q∗ , with credit alone. So these equilibria will be such that money and credit coexist but are used in different types of meetings, as in the previous sections, but we do not need to impose enforcement or commitment. The value of being an unmatched buyer in the CM, Wub (z), is given by Wub (z) = z + T + max {−γ z + βVub (z )}, z ≥0

(8.41)

where Vub (z ) is the value of being an unmatched buyer holding z units of real balances at the beginning of a period. The buyer can consume z units of general good from his z units of real balances, he receives a lump-sum transfer (tax) of real balances if γ > 1 (γ < 1), and he accumulates γ z units of real balances in the current period in order to start the next period with z real balances, where γ −1 = φt+1 /φt is the rate of return on money in a steady-state equilibrium. The value function of an unmatched buyer in the DM who holds z units of real balances, Vub (z), is given by Vub (z) = σL VLb (z) + σS VSb (z) + (1 − σL − σS )Wub (z).

(8.42)

With probability σL , the buyer ﬁnds a long-term partnership with value VLb (z), and with probability σS , he ﬁnds a short-term match whose value is VSb (z). With probability 1 − σL − σS , the buyer remains unattached and enters the night market with his z units of real balances that provide value Wub (z). Following a similar reasoning, the expected lifetime utility of an unmatched seller at night is Wus (z) = z + βVus ,

(8.43)

222

Chapter 8

where we take into account that sellers have no incentives to hold real balances in the DM. So an unmatched seller with z units of real balances at night consumes z units of general goods and starts the next period unmatched and with no money. In the DM, the value of an unmatched seller is Vus = σL VLs + σS VSs + (1 − σL − σS )Wus (0),

(8.44)

where VLs (VSs ) is the value of a seller in a long-term (short-term) match in the DM. The interpretation of (8.44) is similar to (8.42), except that sellers at the beginning of the DM do not hold real balances. The buyer in a short-term trade match makes a take-it-or-leave-it offer, (qS , dS ), to the seller, where qS is the amount of the DM good that the seller produces and dS is the amount of real balances transferred from the buyer to the seller. The value function of a buyer holding z units of real balances in a short-term trade match, VSb (z), is given by VSb (z) = u qS (z) + Wub z − dS (z) = u qS (z) − dS (z) + z + Wub (0), (8.45) where the second equality is obtained from the linearity of Wub . The buyer consumes qS units of the search good in the day and enters the competitive general goods market with z − dS units of real balances. Similarly the value function of a seller (with no real balances) in a shortterm trade match is VSs = −c qS (z) + dS (z) + Wus (0),

(8.46)

where z represents the buyer’s real balances. Thetake-it-or-leave-it offer by the buyer maximizes the buyer’s surplus, u q , subject to the − d S S seller’s participation constraint, −c qS + dS ≥ 0, and the feasibility constraint, dS ≤ z. It is characterized either by qS (z) = q∗ and dS (z) = c(q∗ ) if z ≥ c(q∗ ), or by qS = c−1 (z) if z < c(q∗ ). Hence (8.45) becomes VSb (z) = u qS (z) − c qS (z) + z + Wub (0) ,

(8.47)

and from (8.46), VSs = Wus (0). The value function for a buyer in a long-term relationship holding z units of real balances at the beginning of the period is VLb (z) = u qL (z) + WLb z − dL (z), −yL (z) ,

(8.48)

Money and Credit

223

where WLb (z−dL , −yL ) is the value of the matched buyer at night holding z − dL units of real balances, with a promise to produce yL units of the general good for his trade-match partner. So a buyer in a long-term partnership consumes qL units of search goods in exchange for dL units of real balances and a promise to repay yL units of general goods. Even though we allow the terms of trade (qL , dL , yL ) to depend on the buyer’s real balances, z, we consider equilibria where buyers don’t use money in long-term partnerships, dL = 0 and (qL , yL ) independent of z. The value function of a buyer in a long-term partnership at the beginning of the night, WLb (z, −yL ), satisﬁes WLb (z, −yL ) = z − yL + T + λ max {−γ z + βVub (z )} z ≥0

(8.49)

b −γ z + βV (z ) . + (1 − λ) max L z ≥0

At the beginning of the night, the buyer fulﬁlls his promise and produces yL units of the general good for the seller. If the trade match is not exogenously destroyed, then the buyer produces in order to hold z real balances in the CM. If the partnership breaks up at night—an event that occurs with probability λ—then the buyer produces to hold z real balances in the CM before he proceeds to the next period in search of a new trading partner. By a similar reasoning, the value function for a seller in a long-term relationship at the beginning of the period is VLs = −c qL (z) + WLs dL (z), yL (z) .

(8.50)

The seller produces qL for the buyer in the DM in exchange for a promise to receive yL units of general good at night and dL units of real balances (where z represents the buyer’s real balances). The value function of the seller at night is WLs (z, yL ) = z + yL + (1 − λ)βVLs + λβVus .

(8.51)

The seller receives yL units of general goods from the buyer he is matched with and he spends his z real balances in the CM. With probability λ the long-term partnership is destroyed, in which case the seller starts the next period unmatched. We now turn to the formation of the terms of trade in long-term partnerships. We will assume that the buyer makes a take-it-or-leave-it

224

Chapter 8

offer, (qL , yL , dL ). If the offer is rejected, no trade takes place in that period, but the buyer and the seller remain matched in the subsequent period unless an exogenous destruction shock occurs with probability λ. (In equilibrium, sellers are indifferent between being matched and unmatched.) Moreover the offer must satisfy the incentive-compatibility constraint according to which the buyer is willing to repay his debt b at night. So the buyer chooses (qL , yL , dL ) in order tomaximize VL(z) s subject to the seller’s participation constraint, −c qL + WL dL , yL ≥ WLs (0, 0), and the incentive compatibility constraint, WLb z − dL , −yL ≥ Wub (z − dL ). The incentive-compatibility constraint states that the buyer is better off paying his debt than walking away from his partnership. From (8.48) and using the linearity of WLs , WLb , and Wub , the buyer’s problem can be expressed as max u q − y − d (8.52) q,y,d

subject to and

− c q + y + d ≥ 0,

y ≤ WLb (0, 0) − Wub (0).

d ≤ z, (8.53)

We will focus on equilibria where the incentive-compatibility constraint (8.53) does not bind for all values of z. As a result qL = q∗ and yL + dL = c(q∗ ). This ﬁnding conﬁrms that the terms of trade in long-term partnerships are independent of the buyer’s real balances. Without loss of generality, we can assume that buyers pay with credit only, dL = 0. It is also immediate from (8.48) and (8.49) that a buyer in a longterm partnership at night will not accumulate real balances (in z = 0, equation 8.49). Let us consider the choice of real balances by unmatched buyers. From (8.41)–(8.49) the optimal choice of real balances at night, z, for a buyer who is not in a long-term relationship satisﬁes max{−iz + σS u qS (z) − c qS (z) }. (8.54) z≥0

Since real balances are not needed in long-term partnerships, the buyer only takes into account his expected surplus in a short-term match when choosing his money holdings. This leads to the familiar ﬁrst-order condition, u (qS ) i = 1+ . c (qS ) σS

(8.55)

Money and Credit

225

The last thing we need to check is that the incentive-compatibility condition, (8.53), is not binding. Using that yL = c(q∗ ), (8.53) becomes c(q∗ ) ≤ WLb (0, 0) − Wub (0).

(8.56)

With the help of equations (8.42)–(8.49), and after some rearranging (see the appendix), inequality (8.56) can be rewritten as c(q∗ ) ≤ (1 − λ)β (1 − σL )u(q∗ ) + ic(qS ) − σS u(qS ) − c(qS ) ,

(8.57)

where qS satisﬁes (8.55). If inequality (8.57) holds, then there exists an equilibrium where buyers and sellers in long-term relationships consume and produce qL = q∗ units of the search good during the day and yL = c(q∗ ) units of the general good at night, using credit arrangements to implement these trades. Buyers and sellers in short-term partnerships trade qS units of the search good for yS = c(qS ) units of real balances during the day. Perhaps not surprisingly, if σS = 0, then from (8.55), qS = 0 and the incentive condition (8.57) is identical to the one obtained in a model where money was absent and trade in long-term relationships was supported by reputation; see the deﬁnition of AR given by (2.46) in chapter 2.4. If the frequency of short-term matches, σS , increases, then from (8.55) agents will increase their real balance holdings; as a result the incentive constraint (8.57) becomes more difﬁcult to satisfy. Hence the availability of monetary exchange in the presence of a long-term partnership increases the attractiveness of defaulting on promised performance. However, if inﬂation increases, then, from the envelope theorem, the term −ic(qS ) + σS u(qS ) − c(qS ) decreases, which relaxes the incentive constraint (8.57). Hence a higher inﬂation rate reduces the buyer’s incentive to default on this long-term partnership obligations. 8.6 Further Readings Shi (1996) considers a search-theoretic environment where ﬁat money and credit can coexist, even though money is dominated by credit in rate of return. A credit trade occurs when two agents are matched and the buyer in the match does not have money. Collateral is used to make the repayment incentive compatible, and debt is repaid with money. In this approach, monetary exchange is superior to credit in the sense that monetary exchange allows agents to trade faster.

226

Chapter 8

Li (2001) extends Shi’s model to allow private debt to circulate and she investigates various government policies, including open-market operations. Telyukova and Wright (2008) develop a model similar to that in chapter 8.1 where IOUs are issued in a competitive market. They show that such a model can explain the credit card debt puzzle, the observation that a large fraction of US households owe a sizable amount of credit card debt and hold liquid assets at the same time. In Camera and Li (2008), agents are anonymous, and they choose between using money and credit to facilitate trade. There exists a costly technology that allows limited record keeping and enforcement. Money and credit can coexist if the cost of using the technology is sufﬁciently small. Lucas and Stokey (1987) propose a model where the distinction between goods purchased with cash and goods purchased with credit is exogenous. Schreft (1992) and Dotsey and Ireland (1996) endogenize the composition of trades involving cash or credit. They assume that agents trade in different markets where they can hire the services of a ﬁnancial intermediary who can verify the buyer’s identity. The cost paid to the intermediary is higher the greater the distance between the borrower’s and the lender’s home locations. This formalization is also related to the model by Prescott (1987) and Freeman and Kydland (2000), where some goods are bought with cash and others with demand deposits. The second means of payment involves a ﬁxed cost of record-keeping associated with bank drafts. Townsend (1989) investigates the optimal trading mechanism in an economy with different locations, where some agents stay in the same location and other agents move from one location to another. The optimal arrangement implies the coexistence of currency and credit: Currency is used between strangers, namely by agents whose histories are not known to one another, and credit is used among agents who know their histories. Kocherlakota and Wallace (1998) consider a random-matching economy with a public record of all past transactions that is updated only infrequently. They show that in this economy there are roles for both monetary transactions and some form of credit. Jin and Temzelides (2004) consider a search-theoretic model with local and faraway trades. There is record keeping at the local level so that agents in local meetings can trade with credit. In contrast, agents from different neighborhoods need to trade with money. Li (2007) considers an environment with a random-matching sector and organized

Money and Credit

227

markets in which bills of exchange circulate as a general medium of exchange. The model on credit and the reallocation of liquidity has been inspired from the work by Berentsen, Camera, and Waller (2007) on banking. Instead of considering a loan market, they introduce banks that make loans and accept deposits. Another interpretation is the one from Kocherlakota (2003) on the societal beneﬁts of illiquid bonds. In Kocherlakota’s model, agents trade their excess liquidity for interest-bearing illiquid government bonds. Kahn (2009) uses a similar model to study round-the-clock private payments arrangements. Williamson (1999) constructs a model where banks intermediate a mismatch between the timing of investment payoffs and when agents wish to consume; claims on banks may serve as media of exchange, namely private money. Cavalcanti, Erosa, and Temzelides (1999) develop a model of money and reserve-holding banks where private liabilities can circulate as media of exchange. Li (2006) studies competition between inside and outside money in economies with trading frictions and ﬁnancial intermediation. Corbae and Ritter (2004) consider a model of long-term and shortterm partnerships similar to the one presented in chapter 8.5. Williamson (1998) constructs a dynamic risk-sharing model where there is private information about agents’ endowments. Risk-sharing is accomplished though dynamic contracts involving credit transactions and monetary exchange. Aiyagari and Williamson (2000) construct a dynamic risksharing model where agents can enter into a long-term relationship with a ﬁnancial intermediary. They introduce a transaction role for money, by assuming random limited participation in the ﬁnancial market. In each period agents can defect from their long-term contracts and trade in a competitive money market in thereafter. Aiyagari and Williamson show that the value of this outside option depends on monetary policy. Appendix Derivation of (8.57) From (8.48) and (8.49), WLb (0, 0) = T + (1 − λ)β u(q∗ ) − c(q∗ ) + WLb (0, 0) + λ −γ z + βVub (z) ,

(8.58)

228

Chapter 8

where z is the optimal choice of real balances of an unmatched buyer, and buyers in long-term partnerships do not accumulate real balances. From (8.41). Wub (0) = T − γ z + βVub (z).

(8.59)

From (8.58) and (8.59), WLb (0, 0) − Wub (0) = (1 − λ)β

γ u(q∗ ) − c(q∗ ) + WLb (0, 0) − − z + Vub (z) β

(8.60)

From (8.42), (8.47), and (8.48), Vub (z) = σL u q∗ − c(q∗ ) + WLb (0, 0) − Wub (0) + σS u qS − c qS + z + Wub (0). Substituting Vub (z) by its expression into (8.60) obtains [1 − (1 − λ) (1 − σL ) β] WLb (0, 0) − Wub (0) = (1 − λ)β (1 − σL ) u(q∗ ) − c(q∗ ) − −iz + σS u qS − c qS , where we have used that i ≡ (γ − β)/β. Condition (8.56), c(q∗ ) ≤ WLb (0, 0) − Wub (0), can then be expressed as [1 − (1 − λ) (1 − σL ) β] c(q∗ ) ≤ (1 − λ)β (1 − σL ) u(q∗ ) − c(q∗ ) − −iz + σS u qS − c qS , and simpliﬁed to c(q∗ ) ≤ (1 − λ)β (1 − σL )u(q∗ ) + ic(qS ) − σS u qS − c qS , where we use z = c(qS ).

9

Money, Negotiable Debt, and Settlement

In large value payment systems, such as the Federal Reserve’s Fedwire, participants make and receive payments throughout the day. In an ideal world the payments process would be seamless in the sense that agents receive payments at, or just before, the time they have to make them. In such a world agents will always have sufﬁcient liquidity on hand to make their required payments. In practice, however, the payments process is not so perfectly synchronized; agents may have insufﬁcient liquidity on hand when they wish to, or have to, make a payment. In such circumstances the agent can always wait for an incoming payment. But waiting may be costly. Alternatively, the payments network may provide the agent with liquidity, say, via a daytime loan, so that the agent can make time critical payments without delay, and loans can be paid back when the agent receives (the delayed) payments. The importance of the timing of payments is not conﬁned to large value payment systems. This issue also arises in short-term money markets, such as the tri-party repos, in the clearing and settlement of ﬁnancial securities, and so on. In this chapter we examine the implications associated with settlement frictions and possible policy responses if the frictions have adverse implications for the economy. To do so, we modify the economic environment so that ﬁat money plays a dual role: it serves both as a medium of exchange to facilitate trade and as an instrument to settle debt, that is, to make a payment on a prior obligation. We introduce frictions in the settlement of private debt that give rise to negotiable debt. Negotiable debt is debt that can be sold to a third party, and is honored by the issuing debtor when presented for redemption. The settlement friction is that an agent having an immediate need for liquidity or money presently has no liquidity and is awaiting a payment. The agent can always sell the obligation that represents this incoming payment for the liquidity that he currently desires. Depending on the extent of the frictions, the

230

Chapter 9

market for negotiable debt may be sufﬁciently liquid so that the seller of negotiable debt receives the full value of his claim. In this situation the market for negotiable debt overcomes the settlement frictions. But this need not always be the case, and liquidity problems associated with settlement frictions can arise. When the market for negotiable debt fails to overcome the settlement frictions, the liquidity problems that arise in settlement will spill over into credit and product markets, and will have negative implications for the real economy. In this case there is a welfare-enhancing role for central bank intervention. A central bank can pursue either open-market or discount window operations to provide additional liquidity in the settlement phase of the economy. A properly designed policy provides liquidity during the settlement phase but has no long-run effects on the supply of money: Any injection of money for liquidity purposes is immediately undone when the private debt, held by the central bank, is redeemed. If the central bank follows a policy along these lines, then an efﬁcient allocation will be restored. This line of reasoning provides support to the notion of an elastic supply of currency, which is one of the founding principles of the establishment of the Federal Reserve System. We ﬁnd that our basic insights are not altered when there is an exogenous risk of default on behalf of the debtors. 9.1 The Environment We consider an environment where credit and money coexist, and money is used to settle debt obligations. In order to present the ideas in the most economical way, we modify the benchmark model. A period is now divided into four subperiods: morning, day, night, and late night. As in the benchmark model, the day subperiod is a decentralized market, DM, characterized by bilateral matching and exchange of the search good, and the night subperiod is a competitive market, CM2, where the general good is produced and traded. In terms of the two new subperiods, the morning subperiod, like the night subperiod, is a competitive market, CM1, where the general good is produced and traded. In the late-night subperiod, production and consumption are not feasible. In this subperiod, agents have the opportunity to settle any debts that were incurred in previous subperiods. If agents choose to settle their debts in the late-night subperiod, the debts must be settled with money since production is not possible.

Money, Negotiable Debt, and Settlement

231

In order to capture the coexistence of money and credit, and settlement of debt obligations with money, we make the following assumptions: 1. Agents live for only four subperiods. Buyers are born at the beginning of a period, in the morning, and die after the settlement phase in the late night of the same period. Sellers are born at the beginning of the day subperiod and die at the end of the morning subperiod, CM1, in the subsequent period. 2. Buyers are heterogeneous in terms of when they can produce. Half of the buyers can only produce in the CM1, and the other half can only produce in the CM2. We call the former early producers and the latter late producers. 3. In the DM bilateral match, the seller has a technology to verify the identity of the issuer. In the late-night subperiod, there is a technology that authenticates IOUs issued in the DM and enforces the repayment of the IOUs. 4. In the CMs, IOUs cannot be authenticated and can be costlessly counterfeited. Assumption 1 implies that in any particular CM1, the economy is populated with young buyers and old sellers; in all other subperiods, the economy is populated with buyers and sellers who are born in the same period. The assumption of ﬁnitely lived buyers is convenient since all buyers start the period with no money balances. Otherwise, buyers who anticipate they cannot produce in the CM1 may want to accumulate money balances in previous periods. Assumptions 1 and 2 imply that if late producers trade in the DM, they can only do so by issuing IOUs since it is not possible for them to accumulate money balances. Assumption 3 implies that in a DM match, an IOU can be issued. Assumption 4 implies that IOUs issued and authenticated in the DM will not circulate as a means of payment in the CMs, and new IOUs will not be issued, because of the recognizability problem that exists in those subperiods. Collectively, the assumptions above imply that early producers can use money or debt in the DM, late producers only use debt in the DM, and all debts will be settled with money in the late-night settlement period. (Note that this structure is similar in spirit to that of chapter 6.7, where the important link between that structure and the structure in this chapter is that buyers are heterogeneous in terms of their ability to produce.)

232

Chapter 9

Buyers are able to produce the general good in either the CM1 or CM2, depending on their type, but have no desire to consume the general good. They are unable to produce the search good but want to consume it. The preferences for the buyer are described by the instantaneous utility function U b (q, y) = u(q) − y, where y is the buyer’s production of the general good—either in the CM1 or in the CM2, depending on the buyer’s type—and q is the consumption of the search good. Sellers are able to produce the search good in the DM but have no desire to consume it. They are unable to produce the general good but want to consume it. The preferences for the seller are given by U s (q, x) = −c(q) + x, where x is the seller’s consumption of the general good—in the CM1 and the CM2—and q is the amount of the search good that is produced. Note that agents do not discount utility across subperiods over their lifetime. During the DM, buyers and sellers are matched, where buyers consume the search good and sellers produce it. For simplicity, we eliminate any search-matching frictions by setting the matching probability σ to one. The timing of events and the pattern of trade in a representative period are summarized in ﬁgure 9.1. At the beginning of a period, a measure one of buyers are born. Half of them, the early producers, can produce in the CM1. In the CM1 these young buyers produce general goods in exchange for money, and old sellers exchange money for the general good. Old sellers die at the end of the morning, and are replaced by a measure one of newborn sellers at the beginning of the DM. In the DM, each buyer is matched with a seller. Half of the buyers, the early producers, trade with money and the other half, the late producers, trade with credit. (Although it is not indicated in ﬁgure 9.1, in order to simplify the exposition, early producers also have the option to trade with credit in the DM). To settle their debts in the late-night subperiod, buyers who traded with credit produce general goods in exchange for money in the CM2; sellers exchange money for the general good. In the late night settlement period buyers and sellers arrive at a meeting place for the purpose of settling debts. Sellers who receive money in the

Money, Negotiable Debt, and Settlement

233

MORNING (CM1)

DAY (DM)

NIGHT (CM2)

LATE-NIGHT

Competitive market

Bilateral trades

Competitive market

Settlement

Early producers

Early producers

Debtors

Debtors

$

$

Sellers

Old sellers

$ Sellers

$ IOU Creditors

Late-producers (debtors)

IOU Sellers (creditors) Figure 9.1 Timing and pattern of trade

late-night settlement subperiod will spend it in CM1 of the next period, before they die. We focus on stationary equilibria. Since money is traded for general goods in competitive markets in the two different subperiods, we distinguish between two prices for money. We let φ1 be the price of money in terms of general goods in the CM1 and φ2 the price of money in the CM2. 9.2 Frictionless Settlement We ﬁrst examine an economy where there are no frictions in the settlement phase. In particular, all debtors and creditors arrive simultaneously at a central meeting place in the late-night subperiod, and all debts are settled instantaneously. Consider ﬁrst a match in the DM, between a buyer who is an early producer and a seller. This buyer produced general goods in the morning to get m units of money. Suppose that the buyer spends his m units of money in a bilateral match in the day for qm units of the search good. The quantity qm is determined by a take-it-or-leave-it offer by the buyer. The seller’s participation constraint is −c(qm ) + max(φ1 , φ2 )m ≥ 0.

(9.1)

234

Chapter 9

A seller values a unit of money at max(φ1 , φ2 ) because he has the option to spend his money either in the CM2, at the price φ2 , or in the following CM1, at the price φ1 . We can use a simple equilibrium argument to show that max(φ1 , φ2 ) = φ2 . If φ2 < φ1 , then sellers will spend their money in the following CM1. But this outcome is inconsistent with the clearing of the CM2 since late producers need to acquire money at night to settle their debts. Therefore the seller’s participation constraint (9.1) simpliﬁes to −c(qm ) + φ2 m ≥ 0. Note also that an early-producing buyer has no incentive to accumulate money in the CM1 and issue debt in the DM because sellers prefer (weakly) to receive money that they can spend in CM2. Since buyers do not value consumption of the general good, an early buyer’s offer to the seller in the DM is given by the solution to, max u qm m

(9.2)

subject to c qm = φ2 m.

(9.3)

q

The solution to this problem is qm (m) = c−1 (φ2 m), i.e., the buyer spends all his money, subject to satisfying seller participation. In the CM1 the early-producing buyer’s problem of choosing his money holdings, m, is given by the solution to max −φ1 m + u qm (m) . m

(9.4)

The solution to (9.4) is u (qm ) φ1 = c (qm ) φ2

(9.5)

since, from (9.2)–(9.3), dqm /dm = φ2 /c qm . From (9.5), qm = q∗ if and only if φ1 = φ2 ; if φ2 > φ1 , then qm > q∗ . The demand for money from early-producers in the CM1 is then m=

c(qm ) . φ2

(9.6)

The supply of money in the CM1 comes from old sellers who hold the entire stock of money, M. Since there is a measure 1/2 of early-producing buyers, equilibrium in the CM1 money market implies that M = m/2 and, from (9.6), qm satisﬁes

Money, Negotiable Debt, and Settlement

c(qm ) = 2Mφ2 .

235

(9.7)

Now let’s turn to the problem of a late-producing buyer in a bilateral match in the DM. In his bilateral match a late-producing buyer must issue an IOU to pay for the search good, which will be repaid in the late-night settlement subperiod. Recall that a buyer is able to issue an IOU in a bilateral match because the IOU and his identity can be authenticated, and the only other place where the authenticity of the IOU can be established is in the settlement subperiod. The buyer repays the debt by producing output for money in the CM2. The terms of trade in the match are determined by a take-it-or-leave-it offer (qb , b) by the buyer, where qb is the amount of search good produced by the seller and b is the amount of dollars that the buyer commits to repay in the late-night settlement subperiod. (It might be convenient to think of the “m’’ in qm as referring to a buyer who uses money to purchase search goods and the “b’’ in qb as referring to a buyer who issues a bond or an IOU.) The buyer’s offer is given by the solution to max u(qb ) − φ2 b (9.8) qb ,b

subject to

− c(qb ) + φ1 b = 0.

(9.9)

The seller values the buyer’s debt at the price φ1 since the money he receives in the late-night settlement subperiod can only be spent in the next morning. The solution to the buyer’s problem (9.8)–(9.9) is u (qb ) φ2 = . φ1 c (qb )

(9.10)

From (9.10), qb = q∗ if and only if φ1 = φ2 . If φ1 < φ2 , then qb < q∗ . From (9.9) the amount of nominal debt issued by the buyer in the match is b=

c(qb ) . φ1

(9.11)

Consider the equilibrium in the CM2. If φ2 > φ1 , then sellers holding money at the beginning of the CM2 will spend all of it so that at the end of the night all of the money is held by the late-producing buyers, that is, b/2 = M. If φ2 = φ1 , then sellers holding money are indifferent

236

Chapter 9

between spending it in the CM2 or in the following CM1. In this case b/2 ≤ M. In summary,

= > b 2M if φ2 φ1 . (9.12) ≤ = A steady-state equilibrium is a list (qm , qb , φ1 , φ2 , b) that satisﬁes (9.5), (9.7), (9.10), (9.11), and (9.12). It can be easily demonstrated that qm = qb = q∗ , b = 2M, and that φ1 = φ2 = c(q∗ )/2M is an equilibrium. If φ1 = φ2 , then from (9.5) and (9.10), qm = qb = q∗ . And from (9.7), φ1 = φ2 = c(q∗ )/2M. From (9.11), b = 2M, which is consistent with (9.12). We show in the appendix that this is the unique equilibrium for some speciﬁcations u √ and c, such as u(q) = 2 q and c(q) = q. In what follows, we will focus on speciﬁcations for which the equilibrium under frictionless settlement is unique. In this equilibrium the price of money is the same in the CM1 and CM2, and the efﬁcient quantity of the search good q∗ is traded in all matches. 9.3 Settlement and Liquidity We now introduce settlement frictions. Settlement frictions are captured by having debtors and creditors arrive and leave the late-night settlement period at different times. To be more speciﬁc, the timing during the late-night settlement period is as follows: All of the creditors— who are sellers—and a fraction α of debtors—who are late-producing buyers—arrive at a central meeting place at the beginning of the settlement period. Then a fraction δ of the creditors depart, after which the remaining 1 − α debtors arrive. Finally, the remaining 1 − δ creditors and all of the debtors leave the settlement period. At this point all the buyers die, and all the sellers move into the morning of the next period. The timing of arrivals and departures is illustrated in ﬁgure 9.2. We will sometimes refer to creditors (debtors) as being early-leaving (earlyarriving) and late-leaving (late-arriving), where the meaning is obvious. These arrival and departure frictions will create a need for a resale market for debt during the late-night settlement period. We assume that this resale market for debt is competitive, where ρ is the price of one-dollar of debt in terms of money. Sellers who produce the DM good for money are neither creditors nor debtors. These sellers may have an incentive to forgo (some) consumption in the CM2, and instead provide liquidity in the settlement

Money, Negotiable Debt, and Settlement

Sellers with money

Creditors

Early-arriving debtors (α)

237

Late-arriving debtors (1−α)

Early-leaving creditors (δ) Figure 9.2 Frictions in the settlement phase

period. They can do so by buying the IOUs of early-leaving creditors that will be repaid by late-arriving debtors. For simplicity, we assume that sellers with money who do not spend all of it in the CM2 always arrive at the beginning of the settlement period, and always stay until the end; see ﬁgure 9.2. The logic of our arguments would go through if a fraction δ of the sellers holding money had to leave the settlement stage early: in that case a seller with money would be able to buy a second-hand debt with probability 1 − δ. The DM bargaining problem of the buyer must now take into account the possibility that a seller who receives money for producing the DM goods may want to use some of it to purchase debt in the settlement period. In particular, a seller who receives one unit of money in a bilateral match during the DM can spend it in the CM2 for φ2 units of the general good, or he can buy 1/ρ IOUs in the settlement period and then purchase φ1 /ρ units of the general good in the following CM1. In equilibrium sellers must be willing to spend some of their money in the CM2 in order to allow late-producing buyers to acquire money to settle their debt in the late-night subperiod. Since φ2 ≥ φ1 /ρ is required for equilibrium in the CM2, the seller’s participation constraint is still given by c(qm ) = φ2 m. Hence the early-producing buyer’s bargaining problem is the same as in the frictionless settlement environment, where the solution to this

238

Chapter 9

problem is characterized by (9.5), and the quantity produced in this match, qm , satisﬁes (9.7). Consider now the late-producing buyer’s bargaining problem. The participation constraint of a seller who trades output for debt will be affected by the frictions in the settlement phase. More speciﬁcally, creditor sellers may have to sell their IOUs at a discount if they need to leave the settlement phase before their debtors arrive. Let denote the expected value to the seller of a one-dollar IOU expressed in dollars. The buyer’s bargaining problem can be represented by max u(qb ) − φ2 b (9.13) qb ,b

subject to

− c(qb ) + φ1 b = 0,

(9.14)

where satisﬁes

α = δ [α + (1 − α)ρ] + (1 − δ) + (1 − α) . ρ

(9.15)

From (9.13) the buyer maximizes his utility of consumption net of the cost of producing φ2 b units of general good in the CM2 in order to repay his debt in the settlement period. The seller’s participation constraint, (9.14), speciﬁes that the expected value of the IOUs in terms of the general good traded in the next CM1 must cover the disutility of production of the seller in the DM. Equation (9.15) has the following interpretation. With probability δ, a seller holding a one-dollar IOU must leave the settlement place early. If his debtor has already arrived, an event that occurs with probability α, the IOU is redeemed for one dollar. Otherwise, the IOU is sold at the price ρ. With probability 1 − δ, the seller with a one-dollar IOU does not need to leave early. Therefore the IOU that he holds is redeemed for one dollar, independent of the arrival time of his debtor. However, if the debtor of a seller arrives early, an event that occurs with probability α, the creditor can use the dollar he receives to buy 1/ρ IOUs that will be redeemed for 1/ρ dollars at the end of the settlement phase. The expected values for an IOU, conditional on arrival and departure outcomes, and the probabilities associated with the outcomes, are presented in table 9.1. The solution to the late-producing buyer’s bargaining problem (9.13)– (9.14) is given by

Money, Negotiable Debt, and Settlement

239

Table 9.1 Value of $1 IOU in the settlement period (no default) Debtor arrives . . .

Early (α)

Late (1 − α)

1 1/ρ

ρ 1

Creditor leaves . . . Early (δ) Late (1 − δ)

u (qb ) φ2 . = φ1 c (qb )

(9.16)

The quantities traded in the DM in exchange for IOUs are efﬁcient if φ2 = φ1 . From (9.14) the quantity of debt issued by buyers in the DM is b=

c(qb ) . φ1

(9.17)

Consider the equilibrium of the CM2. Denote as the funds that each seller with money—and there is a measure 1/2 of such sellers—retains at night so that he can purchase second-hand IOUs in the late-night settlement period. The total amount of money supplied in the CM2 is equal to the total stock, M, minus money held by sellers to purchase existing IOUs in the settlement period, /2. The demand for money comes from buyers who need to settle their debt, equal to b/2. Hence equilibrium in the CM2 requires that b + = M. 2 2

(9.18)

If φ2 > φ1 /ρ, then sellers who hold money at the beginning of the night prefer to spend it in the CM2 rather than the following CM1. If, however, φ2 = φ1 /ρ, then sellers are indifferent between spending money in the CM2 or in the next CM1. To summarize,

=0 ≥0

if φ2 > if φ2 =

φ1 ρ , φ1 ρ .

(9.19)

Let’s now turn to the equilibrium for the existing-debt market in the settlement period. Note that ρ, the price of IOUs in the settlement period, cannot be greater than one; ρ > 1 implies that anyone who purchases the

240

Chapter 9

IOU will get a strictly negative net payoff. Therefore ρ ≤ 1. There are two possible sources for the supply of funds to purchase existing IOUs in the settlement period. First, there are the creditors who are repaid early and leave late, who hold in total (1 − δ)αb/2 units of money. (Recall that half of the sellers in the DM are paid with IOUs.) Second, there are sellers who received money during the DM and supply /2 units of money in the settlement period. The demand for funds from earlyleaving creditors is ρδ(1 − α)b/2. If the supply of funds, (1 − δ)αb/2 + /2, is greater than the volume of second-hand IOUs to be purchased, δ(1 − α)b/2, then buyers of those IOUs will bid up the price until it reaches ρ = 1. Otherwise, the price of second-hand IOUs will adjust so that the supply of funds, (1 − δ)αb/2 + /2, is equal to the demand, δ(1 − α)bρ/2. To summarize, the market-clearing price of second-hand debt, ρ, satisﬁes ⎧ ⎨ 1 if (1 − δ)α 2b + 2 ≥ δ(1 − α) 2b , ρ= (9.20) ⎩ (1−δ)αb+ otherwise. δ(1−α)b If the supply of funds is large enough to redeem the IOUs of earlyleaving creditors at face value, then the price of existing debt is one. If there is a shortage of funds, then existing debt will be sold at a discount. A steady-state equilibrium is a list (φ1 , φ2 , ρ, qm , qb , b, ) that satisﬁes (9.5)–(9.7) and (9.16)–(9.20). We distinguish between two types of equilibria: one where ρ = 1 and one where ρ < 1. If ρ = 1, then there is no liquidity shortage in the settlement period: Existing IOUs are sold at par, ρ = 1 and, from (9.15), the expected value of a one dollar IOU in the DM is one, i.e., = 1. As a result the equilibrium conditions are identical to those of the economy without any frictions in the settlement period, i.e., qm = qb = q∗ , φ1 = φ2 = c(q∗ )/2M, b = 2M, and = 0. Note that from (9.20), ρ = 1 requires that (1 − δ)α/δ(1 − α) ≥ 1 or, equivalently, α ≥ δ. Intuitively there is no liquidity shortage if the measure of debtors who arrive early in the settlement place, α, is larger than the measure of creditors who leave early, δ. Creditors who are repaid by early-arriving debtors can use this money to purchase the IOUs of creditors who need to sell them, the earlier-leaving creditors. Consider now equilibria where existing debt is sold at a discount in the settlement period, i.e., where ρ < 1. From (9.20) we have

Money, Negotiable Debt, and Settlement

ρ=

241

(1 − δ)αb + . δ(1 − α)b

The equilibrium is liquidity constrained in the sense that the amount of money available at the settlement period just prior to the departure of the early-leaving creditors is insufﬁcient to clear debts at their par value. An important result here is that if ρ < 1, then > 0, which implies that sellers with money provide additional liquidity in the settlement period by only spending a fraction of their money balances in the CM2. To see this, suppose, to the contrary, that = 0. Then from (9.20), ρ = (1 − δ)α/δ(1 − α) < 1, and from (9.15), = 1. But this implies that the equations that determine (qm , qb , φ1 , φ2 ) are identical to those derived for the model that had no settlement frictions, and as a result that φ2 = φ1 . (Recall that we are focusing on speciﬁcations for which the equilibrium under frictionless settlement is unique). But φ2 = φ1 contradicts the no-arbitrage condition, φ2 ≥ φ1 /ρ since ρ < 1. Therefore it must be that > 0 whenever ρ < 1. When ρ < 1 and > 0, condition (9.19) implies that φ2 = φ1 /ρ, which means that sellers with money are indifferent between spending money in the CM2 or the following CM1. Let’s turn to the effect that the liquidity shortage has on the equilibrium allocation. From (9.15), < 1/ρ, and hence φ2 / φ1 > ρφ2 /φ1 = 1. Together with the fact that φ2 > φ1 , (9.5) and (9.16) imply that u (qm ) φ1 u (qb ) φ2 < 1 < . = = m b c (q ) φ2 φ1 c (q ) The quantities traded in the DM must satisfy qb < q∗ < qm : Buyers who trade with money in the DM receive more output than those who trade with credit. Intuitively, a seller who is paid with money can use it to buy interest-bearing debt in the settlement period; in contrast, a seller paid with debt is facing the risk of having to sell his IOUs at a discount in the settlement period. The liquidity shortage during the settlement period affects the allocation of resources by making money more valuable in the CM2 than in the CM1. Indeed, since unsettled debts are sold at a discount during the settlement period, there is an additional demand for liquidity at night. The fact that money is more valuable in the CM2 allows early-producing buyers to consume more, whereas the consumption of late-producing buyers falls.

242

Chapter 9

9.4 Settlement and Default Risk We now introduce an idiosyncratic risk of late-producing buyers defaulting on their debt. We formalize the default risk by assuming that a debtor is able to produce at night with probability and, with probability 1 − , is unable to produce and hence defaults on his debt obligation. Assume that a debtor does not know if he will default before the night period. This assumption implies that during the DM, buyers and sellers have symmetric information in their bilateral matches. We assume that debtors who are unable to produce, and therefore default on their debt, do not show up at the settlement period. This implies that early-leaving creditors who sell their IOUs do not know whether or not these IOUs will be repaid. The bargaining and choice of money holdings problems for an earlyproducing buyer are still given by (9.2)–(9.3) and (9.4), respectively, since the risk of default is irrelevant for transactions conducted with money. The bargaining problem for a late-producing buyer, however, is now given by max u(qb ) − φ2 b (9.21) qb ,b

subject to

− c(qb ) + φ1 b = 0.

(9.22)

According to (9.21) the buyer receives qb from the seller and is able to produce at night with probability , in which case he can repay his debt. According to (9.22) the seller who receives a promise of b dollars can expect to get b dollars at the end of the period, which can be spent the following morning, where , the expected value of a one-dollar IOU, now reﬂects not only any settlement frictions but also the possibility of default. The solution to problem (9.21)–(9.22) implies that u (qb ) φ2 . = φ1 c (qb )

(9.23)

In the absence of any settlement frictions, it will be the case that = . Then (9.23) is identical to (9.10), and the outcome is similar to the one of the economy without default risk. The default risk is simply reﬂected in the (higher) amount of money that the buyer commits to repay, and the quantity of output traded in bilateral matches remains efﬁcient. This result is reminiscent of the exogenous default result in chapter 2.2.

Money, Negotiable Debt, and Settlement

243

Consider now a seller who has money at the beginning of the settlement period, and who contemplates buying existing IOUs from early-leaving creditors when there is a possibility of settlement frictions. The seller must assess the probability that an existing IOU will be repaid, conditional on the fact that the debtor did not arrive early. This probability is Pr no default ∩ no early arrival Pr no default no early arrival = Pr no early arrival =

(1 − α) 1 − + (1 − α)

=

(1 − α) . 1 − α

We have used the fact that there are three possible events for an IOU in deriving the conditional probability above: an IOU is not repaid, which occurs with probability 1 − ; it is repaid early, which occurs with probability α; or it is repaid late, which occurs with probability (1 − α). The maximum price an agent is willing to pay for a unit face value of existing IOU in the settlement period is the actuarial price, ρ ∗ , that is equal to the conditional probability of repayment, i.e., ρ∗ =

(1 − α) . 1 − α

(9.24)

The expected value of a one-dollar IOU in the DM, when the possibility of settlement frictions exists, is

ρ∗ = α δ + (1 − δ) + (1 − α)(1 − δ) + δ(1 − α)ρ, ρ

(9.25)

or, equivalently from (9.24),

ρ∗ ρ = δα + (1 − δ)α + (1 − δ)(1 − α) + δ(1 − α) ∗ . ρ ρ

(9.26)

Equation (9.25) has the following interpretation: The debtor arrives early with probability α. With probability δ, the creditor leaves early, in which case he gets the par value of the IOU. With probability 1 − δ, he can stay late and use his dollar to buy a second-hand IOU at the price ρ, i.e., he buys 1/ρ IOUs. The probability that the second-hand IOU

244

Chapter 9

Table 9.2 Expected value of $1 IOU in the settlement period (default) Debtor arrives . . .

Early (α)

Late ((1 − α))

Never (1 − )

1 ρ ∗ /ρ

ρ 1

ρ 0

Creditor leaves . . . Early (δ) Late (1 − δ)

is repaid is ρ ∗ . The debtor arrives late with probability (1 − α). If the creditor can wait, with probability 1 − δ, he receives one dollar at the end of the settlement phase. Finally, if the debtor does not arrive early, because he either defaults or because he arrives late, an event that occurs with probability 1 − α, and if the creditor leaves early, an event that occurs with probability δ, then the creditor can sell his IOU at the price ρ. The expected value for an IOU for different events are presented in table 9.2. Following the reasoning in chapter 9.3, the clearing of the CM2 requires b + = M, 2 2

(9.27)

where, as above, represents the funds that a seller with money (there is a measure 1/2 of such sellers) retains at night so that he can purchase existing IOUs in the settlement period. The demand of money in the CM2 comes from the late-producing buyers who need to acquire b units of money in order to redeem their IOUs in the settlement period. The only difference with respect to the market clearing condition (9.18) is that only a fraction of the late-producing buyers are able to produce in the CM2 in order to repay their debt. If φ2 > (ρ ∗ /ρ)φ1 , the sellers with money at the end of the DM strictly prefer buying in the upcoming CM2. As a result the supply of funds from a seller who holds money at the beginning of the CM2, , satisﬁes

=0

if φ2 >

∈ [0, 2M]

if φ2 =

ρ∗ ρ φ1 , ρ∗ ρ φ1 .

(9.28)

The only difference associated with this expression, (9.28), compared to that given by (9.19), is that in the former an existing IOU is redeemed with probability ρ ∗ , while in the latter it is with probability one.

Money, Negotiable Debt, and Settlement

245

Finally, we consider the clearing of the market for existing debt. The market-clearing price, ρ, satisﬁes ρ=

⎧ ⎨ ⎩

ρ∗ (1−δ)αb+ δ(1−α)b

if

(1−δ)αb 2

+ 2 ≥

otherwise.

δ(1−α)bρ ∗ , 2

(9.29)

If the supply of funds is large enough—the left side of the inequality on the top line—to redeem the IOUs of early-leaving creditors at their actuarial price—the right side of the inequality—then the price of second-hand debt is ρ ∗ . If there is a shortage of funds, then existing debt will have to be sold at a discount for the market to clear. Substituting 1−α by its expression given by (9.24) into (9.29) and rearranging, we get ⎧ ⎨ 1 if (1−δ)αb + 2 ≥ δ(1−α)b , ρ 2 2 (9.30) = ρ ∗ ⎩ (1−δ)αb+ otherwise. δ(1−α)b

An equilibrium of the model with default risk is a list (φ1 , φ2 , ρ, qm , qb , b, ) that satisﬁes (9.2)–(9.3), (9.4), (9.22), (9.23), (9.27), (9.28), and (9.30). It can be checked that the probability of nodefault, , affects the equilibrium conditions only through the variables b, ρ/ρ ∗ , and /. For example, ρ/ρ ∗ given by (9.30) coincides with ρ given by (9.20) when b is replaced by b. Hence (φ1 , φ2 , qm , qb , ) coincide with their values in the no-default economy. The value of money and the quantities traded in the DM are not affected by the probability of default, which is taken into account in the price of bonds and the transfer of bonds in the DM. See the appendix for further details. The equilibrium is not liquidity constrained whenever the supply of liquidity from the late-leaving creditors who had their IOUs redeemed by early-arriving debtors, α(1 − δ)b/2, is greater than the demand of liquidity from early-leaving creditors, δ(1 − α)ρ ∗ b/2. From (9.24) the condition α(1 − δ) ≥ δ(1 − α)ρ ∗ is equivalent to α ≥ δ. This is precisely the condition we had in the absence of default risk. The fact that the rate of repayment does not inﬂuence the condition for a liquidity shortage can be explained as follows: Consider an increase in the repayment rate. On the one hand, the number of creditors who are repaid early, α(1 − δ)/2, increases, so there is more liquidity in the late-night settlement period. On the other hand, the demand for liquidity, δρ ∗ (1 − α)b/2 = δ(1 − α)b/2, increases with as well. When α = δ these two effects just cancel each other.

246

Chapter 9

In summary, the presence of an idiosyncratic default risk does not make it more likely that the settlement frictions will generate a shortage of liquidity and, hence, a misallocation of resources. 9.5 Settlement and Monetary Policy When liquidity is “plentiful’’ in the settlement subperiod, the efﬁcient allocation, i.e., the allocation that maximizes the surpluses in the DM, can be implemented as an equilibrium, and this is independent of default probabilities. If, however, there is a liquidity shortage, then the allocation is no longer efﬁcient, i.e., qb < q∗ < qm . Is it possible for monetary policy to improve matters in this situation? In addressing this question, we assume that there is no default risk, i.e., = 1, because, as we have seen, the default risk is simply internalized in the pricing mechanism. When there is a liquidity shortage—which occurs when the fraction of creditors who depart early, δ, is greater than the fraction of debtors who arrive early, α—the market clearing price for debt in the settlement period, ρ, will be less than one, and this ultimately leads to inefﬁcient levels of production in the DM. Suppose now that there exists a monetary authority, or central bank, that can provide “liquidity’’ to the settlement period. More speciﬁcally, the central bank purchases cb ≤ δ (1 − α) b/2 amount of IOUs from early-leaving creditors in exchange for ﬁat money. When the latearriving debtors come to the settlement period, the central bank will exchange the IOUs for ﬁat money. Provided that the IOUs are sold at the price ρ = 1, this operation is neutral for the stock of ﬁat money. Recall that the supply of funds by creditors who are paid early and stay late is (1 − δ)αb/2 and that the face value of bonds of the creditors who leave early and whose issuers arrive late is δ(1 − α)b/2. If b b (1 − δ)α + cb ≥ δ (1 − α) , 2 2 then the liquidity problem is solved: the supply of funds by late-leaving creditors and the central bank is enough to satisfy the demand of funds by early-leaving creditors. In this case IOUs are traded at their face value, ρ = 1, and sellers spend all their money in the CM2 so that b/2 = M. Consequently, in order to implement an efﬁcient outcome as an equilibrium when there is a liquidity shortage in the absence of a central bank, the supply of liquidity by the central bank must satisfy

Money, Negotiable Debt, and Settlement

247

(δ − α) M ≤ cb ≤ δ (1 − α) M. The supply of funds by the central bank is large enough to cover the difference between the IOUs supplied by early-leaving creditors and the demand of IOUs that comes from late-leaving creditors, (δ − α) M, but it is not larger than the liquidity needs of early-leaving creditors, δ (1 − α) M. This temporary supply of liquidity by the monetary authority resembles either a discount window policy or an open-market operation. As an open-market operation, the central bank purchases (δ − α) M units of bonds before the early-leaving creditors depart and sells the bonds back after the late-arriving debtors arrive. As a discount window policy, the central bank stands ready to purchase existing IOUs at their par value, with the understanding that the IOUs have to be repurchased at their par value by the late-arriving debtors before the settlement period ends. The increase in the money supply that results from the openmarket operation or discount window policy is not inﬂationary, since the IOUs purchased by the monetary authority are all redeemed within the period so that the stock of money remains constant across periods. This policy is consistent with the real bills doctrine, which says that the stock of money should be allowed to ﬂuctuate to meet the needs of trade by means of self-liquidating loans. A central bank is not necessarily needed to overcome the liquidity problem. Suppose that a late-leaving creditor, say a clearinghouse, purchases the debt of early-leaving creditors with his own IOUs, with the understanding that the IOUs of the clearinghouse can be exchanged for money in the next morning. (This assumes that in the next period repayment by the clearing house can be enforced.) When the late-arriving debtors arrive, the clearinghouse will exchange the debt that it holds for money. The next morning the clearinghouse can repurchase its debt with money. Hence, as long as the clearinghouse is able to repurchase the debt it has issued, the liquidity problem that arises due to the settlement frictions can be overcome by private agents. 9.6 Further Readings The model of settlement presented in this section is closely related to Freeman (1996a,b). Freeman considers an overlapping-generations economy with heterogeneous agents. Some agents trade with debt while others trade with money. Freeman (1999) extends the model to allow for

248

Chapter 9

aggregate default risk. Green (1999) shows that the role of the central bank as a clearinghouse can be undertaken by ordinary private agents. Zhou (2000) discusses this literature. Temzelides and Williamson (2001) consider two related models: a model with spatial separation and a random-matching model. They investigate different types of payment arrangements, such as monetary exchange, banking with settlement, and banking with interbank lending. They show that payment systems with net settlement generate efﬁciency gains, and interbank lending can support the Pareto-optimal allocation in the absence of idiosyncratic shocks. Koeppl, Monnet, and Temzelides (2008) develop a dynamic general equilibrium model of payments that incorporates private information frictions and that uses mechanism design. As in Lagos and Wright (2005), there is a periodic round of centralized trading, the settlement stage, where agents have linear preferences and can trade a general good. There is no currency, but there is a payments system that can record individual transactions and assign balances to its participants. Because some bilateral meetings are not monitored, the payments system relies on individuals reporting their trades truthfully. This type of model can be used to determine the optimal settlement frequency, and the trade-off between trade sizes and settlement frequency. Kahn and Roberds (2009) provide a survey on the payments literature. They identify payments problems as being associated with temporal mismatches in trading demands and limited enforcement of promises. They focus on mixtures of two kinds of payment systems: store of value systems, which include money, and account-based systems, which include credit. With regard to the latter, they point out that collateral can be useful in facilitating payments. They also discuss issues associated with net and gross settlement, and provide a brief overview to the industrial organization of retail payments. Appendix Equilibrium of the Economy with Frictionless Settlement When √ c(q) = q and u(q) = 2 q From (9.5) and (9.10), m

q =

φ2 φ1

2 ,

(9.31)

Money, Negotiable Debt, and Settlement

b

q =

φ1 φ2

249

2 .

(9.32)

From (9.7) and (9.11), qm = 2Mφ2 ,

(9.33)

qb = bφ1 .

(9.34)

Substitute qm by its expression given by (9.33) into (9.31) to get φ2 = 2M (φ1 )2 .

(9.35)

Similarly substitute qb by its expression given by (9.34) into (9.32) to obtain φ1 = b (φ2 )2 .

(9.36)

There is a unique positive solution to (9.35)–(9.36), and it is φ2 = φ1 =

1 (2M)1/3 b2/3 1 (2M)2/3 b1/3

, .

Hence φ2 = φ1

2M b

1/3 .

From (9.12),

= b 2M if ≤

(9.37)

φ2 φ1

> =

1.

(9.38)

From (9.37) and (9.38) the unique solution is such that b = 2M and φ2 /φ1 = 1. Consequently φ2 = φ1 = 1/2M and qb = qm = 1. Equivalence between the Equilibrium Conditions of the Models with and without Default Redeﬁne the endogenous variables as b˜ = b, ρ˜ = ρ/ρ ∗ , and ˜ = /. The equilibrium conditions (9.22), (9.23), (9.26), (9.27), (9.28), and (9.30) can be re-expressed as

250

Chapter 9

−c(qb ) + ˜ φ1 b˜ = 0, u (qb ) φ2 , = b ˜ φ1 c (q ) ˜ = δα +

(1 − δ)α + (1 − δ)(1 − α) + δ(1 − α)ρ, ˜ ρ˜

b˜ + = M, 2 2

=0 if φ2 > ρφ ˜ 1, ∈ [0, 2M] if φ2 = ρφ ˜ 1, ⎧ b˜ ⎪ ⎨ 1 if (1−δ)α + 2 ≥ 2 ρ˜ = ˜ ⎪ ⎩ (1−δ)α b+ otherwise. ˜

δ(1−α)b˜ , 2

δ(1−α)b

It can be checked that these equilibrium conditions are identical to (9.14), ˜ ρ, (9.16), (9.15), (9.18), (9.19), and (9.20), respectively, where b, ˜ ˜ is replaced by (b, ρ, ).

10

Competing Media of Exchange

Even though ﬁat money plays a major role in facilitating exchange in practice, there exists a large variety of assets and commodities that can be, and are, used as means of payment. For example, commodities, such as gold and silver, and ﬁnancial assets, such as demand deposits, checkable mutual funds, and, to some extent, government securities are used for transaction purposes. There is also a plethora of assets (e.g., capital, claims on capital, and stocks) that could be used as means of payment but are not, or only to a limited extent. The presence of competing media of exchange raises the “central issue in the pure theory of money,’’ which is to explain why ﬁat money is valued in the presence of interest-bearing assets. In John Hicks’s (1935, p. 5) own words: The critical question arises when we look for an explanation of the preference for holding money rather than capital goods. For capital goods will ordinarily yield a positive rate of return, which money does not. What has to be explained is the decision to hold assets in the form of barren money, rather than of interestor proﬁt-yielding securities.

This famous quote is a statement of the so-called rate-of-return dominance puzzle. Most macroeconomic models that incorporate multiple assets—such as money, bonds, and capital—evade this question. Typically money is introduced in these models either via a cash-in-advance constraint—the requirement that a subset of consumption goods must be purchased with money—or as an argument of the utility function. The role of assets as means of payment is then assumed rather than explained. According to Hicks (1935, p. 6): [T]he great evaders would not have denied that there must be some explanation of the fact. But they would have put it down to “frictions,’’ and since there was no adequate place for frictions in the rest of their economic theory, a theory of

252

Chapter 10

money based on frictions did not seem to them a promising ﬁeld for economic analysis.

Following Hicks’s advice, our approach is to look the “frictions’’ in the face. We explain why ﬁat money can be useful even when capital goods can be used as media of exchange. When there is no ﬁat money, agents will, from a social perspective, overaccumulate capital if the stock of capital is insufﬁciently large to satisfy their liquidity needs. When ﬁat money is introduced into the economy and valued, the capital stock decreases since less capital is now required for transactions purposes. Moreover, there is a positive relationship between capital and inﬂation: this is the so-called Tobin (1965) effect. In an economy with two currencies, we show that for standard pricing mechanisms, the exchange rate between the currencies is indeterminate. This result should not be surprising since the exchange rate between two intrinsically useless objects can be whatever agents believe it will be. To remedy this indeterminacy, we propose a Pareto-efﬁcient trading mechanism that is biased in favor of the domestic currency. In this case agents will, in equilibrium, hold only the domestic currency and the exchange rate is determined by fundamentals and monetary factors. In an economy with ﬁat money and nominal government bonds, the rate of return on money will equal that of bonds when money and bonds are perfect substitutes in the decentralized market. Hence liquid bonds will not pay any interest in the sense that the purchase price of a bond will equal its face value. If, however, bonds are less liquid than money—for example, one cannot use all of their bonds for transactional purposes—then bonds will dominate money in terms of their rates of return. We rationalize the illiquidity of bonds in two different ways. First, if bonds can be counterfeited at a cost and money is noncounterfeitable, then sellers will only accept bonds for payment up to some limit. As agents must be compensated for holding bonds they cannot use for transactions purposes, the rate of return on bonds will exceed that of money. Second, if agents use a Pareto-efﬁcient trading mechanism that provides the buyer will a greater surplus for transacting in money instead of bonds, then bonds can dominate money in rate of return. 10.1 Money and Capital The most direct way for a buyer to purchase the search good from a seller in a bilateral match in the decentralized market, DM, would be

Competing Media of Exchange

253

to give the seller what he values: the general good, which is produced in the centralized market, CM. In the benchmark model a barter trade is technologically infeasible since it is assumed that goods are perishable: they fully depreciate at the end of the subperiod in which they are produced. The good that is produced in the CM cannot be carried into the next day DM to pay for the DM good. In what follows, we modify the economic environment of the benchmark model by assuming that agents have access to a storage technology that enables them to carry the CM good from one period into the next. The storage technology is represented by a function f . An agent who stores k units of the CM good obtains f (k) units in the subsequent period. The DM good is still assumed to be perishable, and fully depreciates at the end of the DM. 10.1.1 Linear Storage Technology We ﬁrst consider the case where the storage technology, f , is linear, meaning that one unit of the stored CM good at night generates R ≥ 0 units of general good in the following period. We will refer to a CM good that is stored as capital. The gross rate of return from storage is R: k units of capital at night will turn into f (k) = Rk units of CM goods the following period. The Rk units of the CM good can be used as a medium of exchange in the DM, and can be either consumed and/or used as capital at night. The technology f corresponds to pure storage if R = 1, a productive technology if R > 1, and one that is characterized by depreciation if R < 1. In addition to capital, an agent can also use ﬁat money as a store of value. One unit of money balances at date t has real value φt in the CM and has a real gross rate of return from period t to period t + 1 equal to φt+1 /φt . The evolution of the real value of a portfolio (m, k), consisting of m units of ﬁat money and k units of capital, between the night of period t and the day of period t + 1 is described in ﬁgure 10.1. NIGHT (CM)

DAY (DM)

Agent’s portfolio:

ϕt m + k

ϕt+1 m + f(k)

Assets’ returns Figure 10.1 Timing and assets’ returns

254

Chapter 10

Assume, for the time being, that the money supply is constant, and focus on steady-state equilibria where the value of ﬁat money is also constant. Consider a buyer with portfolio (m, k) at the beginning of the DM. Denote (q, dm , dk ) as the terms of trade in a bilateral match in the DM, where q is the amount of the DM good that the buyer receives from the seller, dm is the transfer of (nominal) money balances from the buyer to the seller, and dk is the transfer of capital. We consider a pricing mechanism where the terms of trade, (q, dm , dk ), depend only on the buyer’s portfolio. A buyer with portfolio (m, k) at the beginning of the period has a lifetime expected utility, V b (m, k), given by V b (m, k) = σ u q(m, k) + W b m − dm (m, k), k − dk (m, k) + (1 − σ )W b (m, k).

(10.1)

According to (10.1) a buyer who meets a seller consumes q units of the DM good, and transfers dm units of money balances and dk units of capital to the seller. The value function for a buyer holding a portfolio (m, k) at the beginning of the CM, W b (m, k), obeys (10.2) W b (m, k) = φm + k + max −φm − k + βV b (m , Rk ) . m ,k

The cost of adjusting the buyer’s portfolio in the CM is φ m − m +k −k, where the k units of general good that are stored at the end of the CM will generate Rk units of general goods in the next DM. As it is by now standard, the value function W b (m, k) is linear in the buyer’s wealth. The terms of trades in a bilateral match are determined by a take-itor-leave-it offer by the buyer. If the buyer holds a portfolio (m, k) in the DM, his optimal offer to the seller is given by the solution to max u(q) − dk − φdm q,dm ,dk

subject to

− c(q) + dk + φdm ≥ 0,

(10.3)

dm ≤ m, dk ≤ k. In other words, the buyer will maximize his surplus, subject to covering the seller’s cost. The solution to problem (10.3) is

q(m, k) =

q∗ −1 c (φm + k)

if

φm + k

≥

0, (10.5) c (q) 1 − βR u (q) − +σ − 1 ≤ 0, “ = ’’ if k > 0. (10.6) βR c (q) According to (10.5) and (10.6), a buyer equalizes the cost of having an additional unit of the asset in the DM with its expected liquidity return in the DM. The liquidity return of an asset corresponds to the increase in the buyer’s surplus if he had an additional unit of the asset in the DM. This liquidity return is u (q)/c (q) − 1 for both capital and real balances. To see this, note that the increase in the buyer’s surplus accumulates unit of real asset in if he an additional q ∂q/∂(φm). When q < q∗ , the DM is u q − c q ∂q/∂k = u q − c ∂q/∂k = ∂q/∂(φm) = 1/c q since φm+k = c(q); when q = q∗ , the liquidity return for both assets is zero. Up to this point we have only considered the buyer’s portfolio problem. The seller’s choice of asset holdings is given by the solution to

1−β 1 − βR max − φm − Rk , β βR m≥0,k≥0

256

Chapter 10

since, by our choice of trading mechanism, the seller’s asset holdings do not affect the terms of trade in bilateral matches. A seller will never accumulate money in the CM since β < 1. If βR = 1, then sellers are indifferent between accumulating capital and not. It should be obvious from conditions (10.5) and (10.6) that buyers are willing to hold both money and capital if and only if R = 1, since this implies that both assets offer the same real return. If R > 1, then capital dominates money in its rate of return, and buyers will hold only capital goods to make transactions. In this case ﬁat money will not be valued, and the quantity traded in the DM satisﬁes u (q) 1 − βR = 1+ . c (q) σβR

(10.7)

The quantity of DM goods traded in bilateral matches increases with the rate of return on capital. And, as the rate of return of the storage technology, R − 1, approaches the discount rate, r (equivalently Rβ approaches one), the quantity traded, q, approaches its efﬁcient value, q∗ . (Note that Rβ cannot be greater than 1, otherwise the buyer’s problem does not have a solution.) If R < 1, then the rate of return on capital is lower than that of ﬁat money. In a steady-state equilibrium buyers will only use money for transaction purposes; buyers will not store any of the general good. It should be pointed out that there exist nonstationary equilibria where output is constant and where money and capital coexist. In such nonstationary equilibria the rate of return on ﬁat money is constant and equal to R < 1. This implies that aggregate real balances shrink over time. But the quantity traded in the DM, q, is determined by (10.7) and, hence, is constant. Since c q = Rkt + φt+1 M is also constant, capital kt must be growing over time. In a monetary equilibrium, where money and capital coexist as means of payments, the quantities traded in the DM correspond to those traded in the monetary economy examined in the previous section, namely, q solves u (q)/c (q) = 1 + (r/σ ). Since money and capital are perfect substitutes, which implies R = 1, the composition of a buyer’s portfolio, in terms of money and capital holdings, will be indeterminate. The total value of the portfolio, however, is pinned down by φM + k = c(q). the value of money can be anywhere in Consequently the interval 0, c(q)/M . This indeterminacy, however, is not neutral in the following sense: if ﬁat money completely replaces capital as a means of payment, then society’s welfare improves because there will

Competing Media of Exchange

257

be a one-time gain from consuming all of the capital that was accumulated to be used as a medium of exchange. This is the kind of argument that has been put forth to favor a ﬁat money regime over a commodity standard. 10.1.2 Concave Storage Technology When the storage technology is linear and the money supply is constant over time, money and capital coexist only in the knife-edge case where R = 1 in a steady-state equilibrium where the value of money is constant. Coexistence can be made more robust if the storage technology is strictly concave. Consider now a storage technology that converts k units of general good in the CM into f (k) units of general good at the start of the subsequent period, where f (0) = 0, f > 0, and f < 0. For simplicity, we impose the Inada conditions f (0) = +∞ and f (+∞) = 0. The buyer’s portfolio choice problem, (10.4), is now given by the solution to

1−β k − βf (k) max − φm − +σ u q −c q , (10.8) β β m≥0,k≥0 where, ∗ from the buyer-take-all bargaining assumption, c(q) = min c(q ), f (k) + φm . The middle term of (10.8) represents the cost of having f (k) units of capital in the DM; to get f (k) units in the DM, k units must be stored in the CM, the (net) cost being kβ −1 − f (k). The ﬁrst-order conditions associated with (10.8) are

u q −r + σ − 1 ≤ 0, c (q)

“ = ’’

if m > 0,

(10.9)

u q 1 − βf (k) − +σ − 1 ≤ 0, βf (k) c (q)

“ = ’’

if k > 0.

(10.10)

If q = q∗ , then, from (10.10) with an equality, k = k ∗ where k ∗ solves βf (k) = 1. The quantity k ∗ also corresponds to the quantity that would be chosen by a social planner who can dictate the allocations in both the CM and the DM. But from (10.9) it is immediate that this is inconsistent with a monetary equilibrium. Nonmonetary Equilibria Consider ﬁrst a nonmonetary equilibrium. From the Inada conditions on f , the solution to (10.10) is interior and the buyer’s capital stock satisﬁes

258

Chapter 10

+ 1 u ◦ c−1 [f (k)] − 1 = σ −1 −1 , βf (k) c ◦ c [f (k)]

(10.11)

where [x]+ ≡ max(x, 0). The left side of (10.11) is increasing in k from −1, when k = 0, to inﬁnity, when k = ∞ and is equal to zero when k = k ∗ ; the right side is decreasing in k from inﬁnity, when k = 0, to 0, when f (k) ≥ c(q∗ ). Consequently, as illustrated in ﬁgure 10.2, there is a unique k n ≥ k ∗ that solves (10.11). It can easily be seen that if f (k ∗ ) ≥ c(q∗ ), then the right side of (10.11) intersects the horizontal axis in ﬁgure 10.2 at a lower value than the left side, and hence k n = k ∗ . A buyer who holds f (k ∗ ) units of general goods in the DM has sufﬁcient resources to purchase the efﬁcient level of the DM good, q∗ , if he is matched. This implies that the right side of (10.11) is zero. And the left side of (10.11) also equals zero, since βf (k ∗ ) = 1. If instead f (k ∗ ) < c q∗ , then the socially efﬁcient stock of capital, k ∗ , is not large enough to allow buyers to purchase q∗ in the DM. In this

σ

u ' ( q) −1 c' ( q )

+

1 −1 β f ' ( k)

k*

k

n

−1 Figure 10.2 Nonmonetary equilibrium

f −1[c (q*)]

k

Competing Media of Exchange

259

situation buyers will overaccumulate capital, i.e., k n > k ∗ as depicted in ﬁgure 10.2. Here buyers are willing to accept a lower rate of return because the capital they hold generates a positive liquidity return by serving as means of payment in bilateral matches in the DM. Now we turn to the seller’s choice of capital. Sellers do not need to accumulate a means of payment. They will therefore choose a level of capital that is independent of any liquidity considerations, which is the same choice that an agent would make in a frictionless economy. The seller maximizes −k + βf (k), and his capital choice is k ∗ . Monetary Equilibria Consider next equilibria where ﬁat money is valued. Condition (10.9) holds with an equality, which, from (10.10), implies that (1 − β)/β = [1 − βf (k)]/βf (k) and hence f (k) = 1. Deﬁne k m > k ∗ as the solution to f (k) = 1, i.e., k m = f −1 (1). In a monetary equilibrium, buyers are willing to hold both capital and real balances since both assets have the same expected liquidity return at the margin in the DM, σ u (q)/c (q) − 1 , and they have the same rate of return across CMs, f (k)−1. Consequently our model is able to explain the coexistence of ﬁat money and capital as means of payment, but does not explain the rate-of-return dominance puzzle. The output in the DM, q, is given by the solution to (10.9) at equality, i.e., u (q) r = 1+ , c (q) σ

and of money is given by f (k m ) + φM = c q , i.e., φ = the value c q − f (k m ) /M. Since a necessary condition for a monetary equilibrium is φ > 0, which in turn implies that c(q) > f (k m ), a monetary equilibrium can exist if u ◦ c−1 f (k m ) r > 1+ , σ c ◦ c−1 f (k m ) or, equivalently, u ◦ c−1 f (k m ) 1 σ −1 m − 1 > m − 1, βf (k ) f (k ) c ◦c

(10.12)

(10.13)

since f (k m ) = 1. A comparison of (10.13) with (10.11) reveals that the capital stock in a monetary equilibrium, k m , is less than that in a

260

Chapter 10

nonmonetary equilibrium, k n . Hence, if condition (10.12) (or condition 10.13) holds, then the (gross) rate of return of capital at the nonmonetary equilibrium is less than one, since f (k m ) = 1. In this situation the introduction of a valued ﬁat money allows buyers to reduce their inefﬁciently high capital stock and to raise their consumption of the DM good. We conclude by considering the limiting case where agents’ discount rate approaches zero. When r approaches zero or, equivalently, when β approaches one, k ∗ tends to k m since βf (k ∗ ) = 1. A necessary condition for a monetary equilibrium is that the stock of capital in the nonmonetary equilibrium, k n , is greater than the socially efﬁcient stock of capital, k ∗ . When k n > k ∗ , buyers in the nonmonetary equilibrium accumulate more capital than what is socially efﬁcient. One can interpret this situation as a shortage of capital in the following sense: the efﬁcient level of capital does not allow buyers to purchase the efﬁcient quantity of output in the DM. It is this shortage of capital that generates a role for ﬁat money. 10.1.3 Capital and Inﬂation In order to study the effect that inﬂation has on capital accumulation and output, we let the money supply grow or shrink at a constant rate. As in chapter 6, money is injected (withdrawn) through lump-sum transfers (taxes) to buyers in the CM. The money growth rate is γ ≡ Mt+1 /Mt > β. We will focus on stationary equilibria, where the rate of return of money, φt+1 /φt , is constant and equal to γ −1 . Taking the same approach as in previous sections, the buyer’s portfolio problem in the CM of period t, assuming the buyer makes a take-it-or-leave-it offer in the DM, is given by max −φt m − k + β σ u q − c q + φt+1 m + f (k) , m≥0,k≥0

where c(q) = min c(q∗ ), f (k) + φt+1 m . This problem can be rearranged to

k − βf (k) max −iφt+1 m − . (10.14) +σ u q −c q β m≥0,k≥0 The buyer’s portfolio problem here is identical to problem (10.8), but now prices must be indexed by time, and the opportunity cost of holding money is i = (γ − β)/β. (As shown in chapter 6.1, the opportunity cost of money, i, can be interpreted as a nominal interest rate since this is the interest that would be paid on an illiquid nominal bond that cannot

Competing Media of Exchange

261

be used as means of payment in the DM.) The ﬁrst-order conditions to problem (10.14) are u q −i + σ − 1 ≤ 0, “ = ’’ if m > 0, (10.15) c (q) u q 1 − βf (k) − +σ − 1 ≤ 0, “ = ’’ if k > 0. (10.16) βf (k) c (q) Note that conditions (10.15) and (10.16) generalize (10.9) and (10.10), since i = r when the money stock is constant, γ = 1. If a monetary equilibrium exists, then both (10.15) and (10.16) hold at equality, which gives 1 − βf (k) =i βf (k) or f (k) = γ −1 ;

(10.17)

that is, the rate of return of capital is equal to the rate of return of ﬁat money. Once again, the rate-of-return-equality principle holds. In a monetary equilibrium, the capital stock is k m = f −1 (γ −1 ). Note from (10.17) that as the inﬂation rate, γ − 1, increases, the rate of return on ﬁat money falls, and buyers accumulate more capital to serve as means of payment. Hence monetary policy can affect capital accumulation when capital is used as a means of payment. Monetary policy can also affect the level of output in the DM, which is given by the solution to u (q) i = 1+ . c (q) σ The determination of a monetary equilibrium is illustrated in ﬁgure 10.3. The top left panel depicts the relationship between the capital stock, k, and the return to capital, f (k). The top right panel represents the relationship between the rate of return on ﬁat money, γ −1 , and the cost of holding real balances, i = (γ − β) /β. Finally, the bottom panel plots the expected liquidity return of assets in the DM, σ u q /c q − 1 , as a function of the output traded in that market, q. For a given return on money, γ −1 , the equilibrium capital stock, k m , is determined in the top

262

Chapter 10

f ' ( k), γ −1

β −1

k

i k

m

k*

σ

u ' ( q) −1 c' ( q )

qm

q* q Figure 10.3 Determination of the monetary equilibrium

left panel. The associated cost of holding real balances can be read on the horizontal axis in the top right panel. Given the cost of holding real balances, the equilibrium output in the DM, qm , is determined in the bottom right panel. A monetary equilibrium exists if φt Mt = c(q) − f (k m ) > 0 or, equivalently, if equation (10.12) holds where r is replaced by i. As in the previous section, this requires that k n be greater than k m . As γ approaches β, k m tends to k ∗ , and the condition for a monetary equilibrium becomes k n > k ∗ . Fiat money has a welfare-improving role whenever buyers in the nonmonetary equilibrium accumulate more capital than the socially efﬁcient level. In that case society’s welfare is at a maximum if ﬁat money is valued and the Friedman rule, γ = β, is implemented: at the Friedman rule, qm = q∗ and k m = k ∗ . 10.2 Dual Currency Payment Systems The previous section examined the coexistence of a ﬁat money and a durable, real commodity. In this section we examine the coexistence

Competing Media of Exchange

263

of two intrinsically worthless objects as means of payment. This is a relevant exercise because actual economies have many different currencies. This raises the questions of whether multiple currencies can be valued and used in payments, whether there is a role for multiple currencies, and how the exchange rate is determined. Models in international macroeconomics typically explain the determination of the value of a currency by using exogenous restrictions on payments, such as cash-inadvance constraints. Although we remove these exogenous restrictions, we show how they may endogenously arise. 10.2.1 Indeterminacy of the Exchange Rate Consider an economy where two ﬁat monies—called money 1 and money 2—can be used as media of exchange. For convenience, one can think of money 1 as dollars and money 2 as euros; this will be useful when we deﬁne exchange rates. The stocks of both monies, M1 and M2 , are ﬁxed, and agents are free to use either currency. One unit of money 1 buys φ1 units of the CM good, and one unit of money 2 buys φ2 units of the CM good. We will focus on stationary equilibria where φ1 and φ2 are constant over time. Consider a buyer holding m1 units of money 1 and m2 units of money 2 in the DM. His beginning-of-period value function, V b (m1 , m2 ), satisﬁes V b (m1 , m2 ) = σ u q(m1 , m2 ) + W b m1 − d1 (m1 , m2 ), m2 − d2 (m1 , m2 ) +(1 − σ )W b (m1 , m2 ).

(10.18)

The interpretation of the value function (10.18) is similar to that of value function V b (m, k) given in (10.1). The value function of the buyer at the beginning of the CM is given by ˆ 1 − φ2 m ˆ 2 + βV b (m ˆ 1, m ˆ 2) , −φ1 m W b (m1 , m2 ) = φ1 m1 + φ2 m2 + max ˆ 1 ≥0,m ˆ 2 ≥0 m

(10.19) where the interpretation is similar to that of W b (m, k) given in (10.2). The terms of trade in the DM are determined by a take-it-or-leave-it offer by the buyer, i.e.,

q(m1 , m2 ) =

q∗ if φ1 m1 + φ2 m2 c−1 (φ1 m1 + φ2 m2 )

≥

c q . If, alternatively, it is assumed that φ1 m1 + φ2 m2 < c q∗ , then the entire intermediate frontier would be curved and would lie below the upper frontier. The lower dashed line frontier represents the pairs of utility levels that can be achieved when the buyer is restricted to use only the domestic currency as means of payment. For this frontier it is assumed that φ1 m1 < c q∗ . In terms of ﬁgure 10.4, our pricing mechanism speciﬁes that the buyer’s surplus is given by the intersection of the dashed lower frontier and the horizontal axis: it is the maximum surplus that the buyer can extract if he can only use the domestic currency in trade. Given the buyer’s s surplus, U b , the seller’s on surplus, U , lies the Pareto frontier directly b above the point U , 0 . Note that U b , U s is pairwise Pareto efﬁcient, given the buyer’s portfolio (m1 , m2 ). We now turn to the buyer’s portfolio choice problem in the CM. Using the same kind of reasoning that led to (10.20), the portfolio choice problem for a buyer who resides in country 1 is given by max −r (φ1 m1 + φ2 m2 ) + σ1 U1b (m1 , m2 ) . m1 ≥0,m2 ≥0

Since, from (10.23), m2 has no affect on the buyer’s surplus, it is immediate that the buyer will choose m2 = 0. As a result our model rationalizes a cash-in-advance constraint, one where buyers hold only the domestic currency. From (10.23), U1b (m1 , m2 ) = u(q1 ) − c(q1 ), where c(q1 ) = min c(q∗1 ), φ1 m1 . The ﬁrst-order condition with respect to m1 for the buyer’s portfolio problem is u1 (q1 ) r = 1+ , c1 (q1 ) σ1

(10.27)

with c1 (q1 ) = φ1 M1 . By analogy, a buyer’s choice of money holdings in country 2 is u2 (q2 ) r = 1+ , c2 (q2 ) σ2

(10.28)

with c2 (q2 ) = φ2 M2 . The intuition for the result that buyers hold only their domestic currency is straightforward. If a buyer purchases goods with foreign

Competing Media of Exchange

269

currency, he will obtain terms of trade that are worst than those associated with holding the domestic currency. More speciﬁcally, from (10.23), with an additional unit of real domestic currency, i.e., 1/φ1 units of money 1, the buyer in country 1 can obtain 1/c1 (q1 ) units of output. According to (10.25), with an additional unit of real foreign currency, i.e., 1/φ2 units of money 2, the buyer obtains 1/u1 (q1 ) < 1/c1 (q1 ) units of output. This implies that the marginal surplus from using the foreign currency, u1 (q1 )[∂q1 /∂ (φ2 m2 )] − 1, is zero, while it is strictly positive from using the domestic currency. As a result agents in each country will only hold the domestic currency, even though there are no restriction on which currencies can be used as means of payment, and there is no cost associated with trading in the foreign exchange market. The nominal exchange rate, ε ≡ φ1 /φ2 , is equal to ε=

c1 (q1 ) M2 . c2 (q2 ) M1

(10.29)

This exchange rate depends on technologies and preferences through the ﬁrst term, and on monetary factors in the two countries through the second term. In order to obtain an expression for the exchange rate that is easier to interpret, we adopt the following functional forms: Agents in both economies have the same utility function for DM goods, u1 (q) = u2 (q) = q1−a /(1 − a) with a ∈ (0, 1). The disutility of production is cj (q) = Aj q, which implies that a productive country has a low A. From (10.27) −1/a −1/a and (10.28), qj = Aj 1 + (r/σj ) . From (10.29) the expression for the exchange rate is then ε=

A2 A1

(1−a)/a

1 + (r/σ2 ) 1 + (r/σ1 )

1/a

M2 . M1

(10.30)

If country 1 becomes more productive, or if its supply of money shrinks, then its currency appreciates vis-à-vis the currency of country 2. The exchange rate depends also on trading frictions. If it becomes easier to ﬁnd trading partners in country 1, then the exchange rate increases. The model can be readily extended to account for the effects that monetary policies in each country have on the exchange rate. Suppose that the gross growth rate in money for country j = 1, 2 is γj ≡ Mj,t+1 /Mj , t > β. (If agents from the two countries have different discount factors, then

270

Chapter 10

the money growth rate in each country must be greater than the discount factor of the most patient agents.) The cost of holding real balances in country j is ij , where 1 + ij = (1 + r)γj . Since a buyer gets zero surplus in the DM from holding the foreign money, he will only accumulate the domestic money, even if its inﬂation rate is higher than that of the foreign money. Hence the model can explain a version of the rate-of-return dominance puzzle, where agents trade with their domestic money even if it is dominated in its rate of return by foreign money. The DM output in each country is given by an equation analogous to (10.27) and (10.28), i.e., uj (qj ) cj (qj )

= 1+

ij , σj

j = 1, 2.

Using the same functional forms as described above, we obtain the expression for the exchange rate in period t as εt =

A2 A1

(1−a)/a

1 + (i2 /σ2 ) 1 + (i1 /σ1 )

1/a

M2,t . M1,t

The exchange rate appreciates at a rate equal to the difference between the inﬂation rates in countries 2 and 1. Note that the (gross) growth rate of the exchange rate, εt+1 /εt , is equal to γ2 /γ1 . 10.3 Money and Nominal Bonds In the previous section we looked at economies with multiple intrinsically useless objects—ﬁat currencies—that serve as means of payment. We now consider economies with ﬁat money and nominal bonds, which are claims on ﬁat money. The presence of nominal bonds allows us to determine a key policy variable, the nominal interest rate. We ﬁrst show that under the standard assumptions used so far, the model predicts that the nominal interest rate is zero. This result constitutes a puzzle—the so-called rate-of-return dominance puzzle—since, in reality, bonds dominate money in terms of rate of return. We then provide conditions under which the rate-of-return dominance puzzle can be resolved, and discuss the implications of the resolution of the puzzle for the determinants of the nominal interest rate.

Competing Media of Exchange

271

10.3.1 The Rate-of-Return Dominance Puzzle Consider an economy where agents can use both money and government bonds as media of exchange. A one-period government bond is issued in the CM and is redeemed for one unit of money in the subsequent period CM. The ﬂow of bonds sold by the government each period is constant and equal to B. We will also assume that the aggregate money supply is constant, i.e., Mt+1 = Mt , or equivalently, γ = 1. Government bonds are of the pure discount variety, perfectly divisible, payable to the bearer, and default-free. These assumptions make money and bonds close substitutes. Since matured bonds are exchanged for money one for one, the price of matured bonds, in terms of CM goods, is φ. Let ω be the price of newly issued bonds in terms of CM goods. If ω < φ, then newly issued bonds are sold at a discount for money. The one-period real rate of return on newly issued bonds is rb = φ/ω − 1. If rb > 0, then the government ﬁnances the interest payments on bonds by lump-sum taxation in the CM. The tax per buyer is (φ − ω) B = rb ωB. We assume that the terms of trade in bilateral matches in the DM are determined by a take-it-or-leave-it offer by the buyer. By the same reasoning as in the previous sections, the expected lifetime utility of a buyer holding a portfolio (m, b), composed of m units of money and b units of bonds at the beginning of the period, is V b (m, b) = σ u q(m, b) − c q(m, b) + W b (m, b),

(10.31)

where q(m, b) = q∗ if φ(m + b) ≥ c(q∗ ) and q(m, b) = c−1 φ(m + b) , otherwise. The expected lifetime utility of a buyer entering the CM with portfolio (m, b) is W b (m, b) = φ(m + b) + T + max −φm − ωb + βV b (m , b ) , (10.32) m ≥0,b ≥0

where T represents a lump-sum transfer in terms of general goods by the government in the CM. If the government needs to ﬁnance the interest payment on bonds, then T = −rb ωB < 0. Note that this equation is similar to (10.19) in the context of two currencies. If we substitute V b from (10.31) into (10.32), then the buyer’s portfolio problem becomes

r − rb max −rφm − φb + σ u q(m, b) − c q(m, b) , (10.33) 1 + rb m≥0,b≥0

272

Chapter 10

where the cost of holding nominal bonds, (ω − βφ)/βφ, can be rearranged to (r −rb )/(1+rb ), which is approximately equal to the difference between the rate of time preference and the real interest rate of the bonds. The ﬁrst-order (necessary and sufﬁcient) conditions for problem (10.33) are

u (q) − 1 ≤ 0, c (q)

“ = ’’

if m > 0,

(10.34)

r − rb u (q) − +σ − 1 ≤ 0, 1 + rb c (q)

“ = ’’

if b > 0.

(10.35)

−r + σ

If bonds are sold at a discount, i.e., if rb > 0, then the cost of holding bonds is lower than that of money, since (r − rb )/(1 + rb ) < r. But then, from (10.34) and (10.35), buyers would only hold bonds, and ﬁat money would not be valued. This, however, cannot be an equilibrium outcome since a nominal bond is a claim to ﬁat money. Consequently, in equilibrium, ﬁat money and newly issued bonds must be perfect substitutes, i.e., ω = φ and rb = 0. Therefore, if there are no restrictions on the use of bonds as means of payment, then interest-bearing government bonds cannot coexist with ﬁat money. This is the rate-of-return dominance puzzle. The output in the DM is given by the solution to (10.34) or (10.35) at equality, i.e., u (q) r = 1+ c (q) σ

(10.36)

and from the seller’s participation constraint, the value of money satisﬁes φ=

c(q) . M+B

(10.37)

The value of money decreases with the stock of money and bonds. The allocations and prices are identical to the ones in a pure monetary economy, where the stock of money in the pure monetary economy is equal to M + B. This implies that the composition of money and bonds, B/M, has no effect on output, prices, and the interest rate. In other words, openmarket operations that consist in substituting money for bonds, or vice versa, are irrelevant because money and bonds are perfect substitutes.

Competing Media of Exchange

273

10.3.2 Money and Illiquid Bonds In an attempt to explain the rate-of-return dominance of bonds over money, we now introduce an admittedly arbitrary restriction on the use of bonds in bilateral meetings in the DM. We assume that a buyer holding a portfolio of b units of bonds can use only a fraction g ∈ [0, 1] of these bonds as a means of payment if he ﬁnds himself in a bilateral match. If g = 0, then bonds are completely illiquid, and if g = 1, then they are perfectly liquid. In practice, the illiquidity of bonds can result from legal restrictions, from the indivisibility of bonds, or from the presence of costs incurred to recognize bonds. While we provide foundations for this restriction in the next sections, for the time being we simply take it as given. The value functions for buyers, V b (m, b) and W b (m, b), are given by (10.31) and (10.32), respectively, where q(m, b) is now deﬁned as follows: ∗ ∗ −1 q(m, b) = q if φ(m + gb) ≥ c(q ), and q(m, b) = c φ(m + gb) , otherwise. The buyer’s portfolio problem is given by the solution to

r − rb max −rφm − φb + σ u q (m, b) − c q (m, b) . (10.38) 1 + rb m≥0,b≥0 The illiquidity of bonds affects the terms of trade in the DM by restricting the amount of wealth that buyers can transfer to sellers. The ﬁrst-order conditions for problem (10.38), assuming an interior solution, are u q r − rb = 1+ , σ g(1 + rb ) c q u q r = 1+ . σ c q

(10.39)

(10.40)

Equating the right sides of (10.39) and (10.40), which implies that buyers are indifferent between holding money and bonds, we obtain rb =

r(1 − g) . 1 + gr

(10.41)

The rate of return on bonds depends on the degree of their liquidity: If bonds are perfectly liquid, i.e., g = 1, then rb = 0 and ω = φ. If bonds are partially illiquid, then the model is able to generate the rate of return dominance of bonds over money; i.e., if g < 1, then rb > 0. In particular, if bonds are illiquid, i.e., if g = 0, then rb = r. While the composition of

274

Chapter 10

money and bonds does not affect the interest rate or output (see equation 10.41), it does affect the value of money, since φ = c(q)/(M + gB). 10.4 Recognizability and Rate-of-Return Dominance Thus far we have shown that interest-bearing bonds and ﬁat money can coexist if there are restrictions on the use of bonds as means of payment. We have not, however, explained the origin of such restrictions. In earlier literatures, physical properties have been used to motivate why bonds are not as liquid as money. Aclassic motivation is that bonds are available only in large denominations and therefore are not useful as a means of payment for typical (small) transactions. We too will appeal to a physical property and that is the recognizability or counterfeitability of a bond. The notion of imperfect recognizability of assets seems plausible at times when bonds are produced on paper, just like banknotes. We suppose that ﬁat money cannot be counterfeited, or only at a very high cost, while bonds can. In particular, agents can produce any amount of counterfeit government bonds in the CM by incurring a ﬁxed real disutility cost of κ > 0. The technology to produce counterfeits in period t becomes obsolete in period t + 1, so paying the cost only allows buyers to produce counterfeit assets for one period. In the DM a seller is unable to recognize the authenticity of bonds. The government has a technology to detect and conﬁscate counterfeits: Any counterfeit bonds produced in period t are detected and conﬁscated before agents enter the CM of period t + 1. Consequently the only outlet for a counterfeit bond produced in period t is in the DM of period t + 1. To simplify the exposition, we assume that there are no search frictions in the DM, i.e., σ = 1, and the terms of trade, (q, dm , db ), are determined by a take-it-or-leave-it offer by the buyer, where q represents the output produced by the seller, dm is the transfer of money, and db is the transfer of bonds—genuine or counterfeit—from the buyer to the seller. The counterfeiting game is similar to the one analyzed for the recognizability of money in chapter 5.3, except that it is bonds, and not money, that can be counterfeited. Following the same reasoning as in chapter 5.3, the buyer’s offer in the DM, (q, dm , db ), must satisfy a no-counterfeiting constraint, −ωdb − φdm + βu(q) ≥ −κ − φdm + βu(q).

(10.42)

Competing Media of Exchange

275

The left side of (10.42) is the buyer’s payoff if he does not produce counterfeits. The buyer accumulates db units of genuine bonds at the price ω, and dm units of money at the price φ, and enjoys the utility of consuming q units of DM output. The right side of (10.42) is the payoff to a buyer who chooses to produce counterfeit bonds. By producing counterfeits, the buyer saves the cost of investing into bonds, ωdb , but he incurs the ﬁxed cost, κ, of producing counterfeits. From (10.42) a buyer in the CM at date t − 1 who anticipates he will be making the offer (q, dm , db ) in the DM at date t will accumulate genuine bonds instead of counterfeits if ωdb ≤ κ.

(10.43)

Inequality (10.43) is an endogenous liquidity constraint that speciﬁes an upper bound on the quantity of bonds that buyers can transfer in the DM. The real value of the newly issued bonds cannot be greater than the ﬁxed cost of producing counterfeits. If it is more costly to produce counterfeits, then the liquidity constraint (10.43) is relaxed. Buyers in the CM will choose a portfolio of money and genuine bonds in order to maximize their expected surplus in the subsequent DM, net of the cost of holding the assets. They will anticipate that the offer they make in the DM must satisfy both the seller’s participation constraint and the no-counterfeiting constraint, (10.43). Because of the opportunity cost of holding money, buyers will not hold more money than they intend to spend in the DM; hence dm = m. Moreover, if ω > βφ, then holding genuine bonds is costly, and buyers will choose to hold the exact amount they spend in the DM; hence db = b. If ω = βφ, then buyers can hold more bonds than they spend in the DM, i.e., b ≥ db . Using these observations, we can write the buyer’s portfolio problem as

(ω/φ) − β max −rφdm − (10.44) φb + u(q) − φ (dm + db ) β q,dm ,db ,b subject to

− c(q) + φ(dm + db ) ≥ 0

and ωdb ≤ κ, db ≤ b,

(10.45) (10.46)

where b is the buyer’s bond holdings. According to (10.44), the cost of holding money is [(1 − β)/β = r, the rate of time preference, while the cost of holding bonds is [(ω/φ) − β]/β. According to inequality (10.45) the offer must be acceptable to sellers, given that sellers interpret all offers satisfying (10.46) as coming from noncounterfeiting buyers.

276

Chapter 10

The Lagrangian associated with this problem is

(ω/φ) − β max −rφdm − φb + u ◦ c−1 φ(dm + db ) − φ (dm + db ) β dm ,db ,b

φκ +λ − φdb + µφ(b − db ) , ω where λ is the Lagrange multiplier associated with the liquidity constraint and µ is the Lagrange multiplier associated with the feasibility constraint on the transfer of bonds. The ﬁrst-order (necessary and sufﬁcient) condition with respect to dm determines the output traded in the DM, q: u (q) = 1 + r. c (q)

(10.47)

The ﬁrst-order condition with respect to db is u (q) − 1 − λ − µ ≤ 0. c (q)

(10.48)

From (10.47) and (10.48), r − λ − µ ≤ 0. So λ = µ = 0 cannot occur in equilibrium, i.e., not all constraints are relaxed. If the solution is interior, then the above inequality, with a strict equality, and (10.47), imply that r = λ + µ.

(10.49)

It can be checked that db = 0 only if ω = φ, i.e., bonds don’t pay interest, in which case buyers are indifferent between holding money and bonds so that (10.49) still holds. Finally, the ﬁrst-order condition with respect to b, assuming an interior solution (since in equilibrium the bonds market must clear) is µ=

(ω/φ) − β . β

(10.50)

Together with (10.49) this gives λ=

1 − (ω/φ) . β

(10.51)

The prices of money and bonds, φ and ω, are determined so as to clear the markets in the CM. Since the portfolio choice of the buyer

Competing Media of Exchange

277

need not necessarily be unique, we will focus on symmetric equilibria. The demand for money is equal to dm , and hence the market-clearing condition for the money market is dm = M.

(10.52)

The market-clearing for the bond market requires that b = B.

(10.53)

We consider the following three cases: 1. The no-counterfeiting constraint is not binding, λ = 0. The buyer’s problem is then identical to problem (10.33), where there is no restriction on the use of bonds as means of payment. From (10.51), bonds and money are perfect substitutes, i.e., λ = 0 implies that ω = φ. From (10.50), µ = r > 0, and hence db = b = B. From (10.45) at equality, φ(M + B) = c(q), meaning that the value of money decreases if the total stock of liquid assets, M + B, increases. The no-counterfeiting constraint (10.43) is not binding if B c(q) ≤ κ. M+B

(10.54)

If the cost of counterfeiting bonds is sufﬁciently high, then bonds are perfect substitutes for ﬁat money, and they do not pay interest. The condition (10.54) also depends on the relative supplies of money and bonds; the no-counterfeiting constraint will not bind if bonds are not too abundant in supply, relative to ﬁat money. Figure 10.5 illustrates the relationship between the relative supply of bonds, B/(M + B), and the relative price of bonds, ω/φ. When the relative supply of bonds is less than κ/c(q), bonds and money trade at the same price, implying that the relative price is unity, i.e., ω/φ = 1. 2. The no-counterfeiting constraint binds, λ > 0, but buyers are not constrained by their bonds holdings, µ = 0. Then, ωdb = κ. The output produced in the DM solves (10.47) and it is independent of the quantity of bonds that buyers can use as means of payment. From (10.50), ω = βφ. Buyers must be compensated for their rate of time preference, and the interest rate paid by bonds is rb = (φ/ω) − 1 = β −1 − 1 = r. In this case bonds dominate money in their rates of return. The no-counterfeiting constraint (10.43), along with (10.45), both at equality, implies that φ=

c(q) − (κ/β) . M

(10.55)

278

Chapter 10

ω ϕ

1

β

κ c (q)

κ β c (q )

B M +B

Figure 10.5 Price of bonds

The value of money decreases with the cost of producing counterfeits. This implies that the value of ﬁat money depends not only on its own characteristics but also on the physical properties of the competing asset. As the cost of producing counterfeit bonds increases, buyers can use a larger fraction of their bond holdings as means of payment, which reduces the value of ﬁat money. If counterfeited bonds can be produced at no cost, κ = 0, then, from (10.43), bonds cannot be used as means of payment and the value of money is the one that prevails in a pure monetary economy. When the no-counterfeiting constraint binds, the condition db ≤ B requires that κ B ≤ c(q). β M+B

(10.56)

If the cost to produce counterfeits is sufﬁciently low and if bonds are abundant relative to ﬁat money, then bonds are fully illiquid at the margin and they offer an interest rate equal to the rate of time preference.

Competing Media of Exchange

279

This result can be seen in ﬁgure 10.5,where ω/φ = β if the relative supply of bonds, B/ (B + M), exceeds κ/βc q . 3. The no-counterfeit constraint binds, λ > 0, and buyers are constrained by their bonds holdings, µ > 0. Conditions (10.43) and (10.45) at equality give ω κ M+B , = c(q) B φ

(10.57)

c(q) . M+B

(10.58)

φ=

From (10.57) the relative price of bonds is a function of the relative supply of bonds, B/(M + B). As the relative supply of bonds increases, the relative price of newly issued bonds decreases. As ﬁgure 10.5 shows, this result implies that the interest rate on bonds increases. Notice that when the buyer receives an additional bond, under the previously prevailing market price of bonds, he cannot spend it in the DM. The price of bonds thus must decrease to reﬂect this illiquidity. The price of bonds will decrease to the point where the no-counterfeiting constraint binds again. Intuitively, the stock of bonds is sufﬁciently large so that the no-counterfeit constraint binds at ω = φ. In order to get buyers to hold all the bonds, the price of bonds must fall to where ω < φ. Although it is less costly to use bonds for transactions in the DM than money, buyers do not demand any additional bonds since they cannot use them in the DM, as that would violate the nocounterfeiting constraint and they do not get compensated for their rate of time preference. From (10.50) the condition db ≤ B binds if µ = [(ω/φ) − β] /β > 0, i.e., ω > βφ. From (10.51) λ = [1 − (ω/φ)] /β > 0, i.e., φ > ω. From (10.57) these conditions can be re-expressed as κ B κ < < . c(q) M + B βc(q)

(10.59)

We can summarize our results using ﬁgure 10.5. We can see that bonds will pay interest provided that the supply of bonds is sufﬁciently large, relative to the cost of producing counterfeits, i.e., when B/(M + B) > κ/c(q). Although the conduct of monetary policy affects the interest rate, it has no effect on the real allocation and welfare. When bonds are relatively scarce, money and bonds are perfect substitutes and ω/φ = 1.

280

Chapter 10

Obviously in that case a change in the composition of money and bonds is irrelevant for the allocation. When bonds are more abundant, the constraint on the transfer of bonds is binding. An open-market operation affects the price of bonds, but the output is still determined so that the marginal beneﬁt of an additional unit of real balances is equal to its cost. 10.5 Pairwise Trade and Rate-of-Return Dominance In the previous section, we used the recognizability property of money and bonds to provide an explanation for the rate-of-return dominance puzzle. In this section, we argue that even if ﬁat money and bonds have the same physical properties—both are divisible and recognizable— the model is still able to generate equilibria with outcomes that are consistent with the rate-of-return dominance puzzle. This explanation is based on the idea that social conventions can play a role in explaining the superior liquidity properties of some assets. For example, buyers may prefer to trade with money instead of bonds because the social convention dictates that they receive better terms of trade in the DM when using money as a means of payment. As in chapter 10.2.2 we exploit the fact that the set of pairwise Pareto-efﬁcient allocations in bilateral matches is large, and we construct a trading mechanism that generates asset prices that are consistent with those in chapter 10.3.2, where there we simply restricted the use of bonds as means of payment. We will construct a mechanism where buyers get the same payoff they would in the economy with exogenous liquidity constraints described in chapter 10.3.2. As in chapter 10.2.2 the mechanism can be thought of as a two-step procedure. The ﬁrst step determines the buyer’s surplus in the DM, U b (m, b). It corresponds to what the buyer would obtain if he was making a take-or-leave-it offer but was able to transfer at most a fraction g of his bond holdings to the seller, i.e., U b (m, b) = max u(q) − φ(dm + db ) (10.60) q,dm ,db

subject to and

− c(q) + φ (dm + db ) ≥ 0

dm ∈ [0, m], db ∈ [0, gb].

(10.61) (10.62)

Competing Media of Exchange

281

The buyer’s payoff is uniquely determined, and satisﬁes, b

U (m, b) =

u(q∗ ) − c(q∗ )

if φ(m + gb) ≥ c(q∗ ),

u ◦ c−1 [φ(m + gb)] − φ(m + gb)

otherwise. (10.63)

Once again, it is important to emphasize that this ﬁrst step determines the surplus that the buyer will receive, and not the terms of trade that will be implemented. The latter is determined in the second step. The second step of the pricing procedure determines the seller’s surplus, U s (m, b), and the actual terms of trade, (q, dm , db ), as functions of the buyer’s portfolio in the match, (m, b), and the ﬁrst stage surplus, U b (m, b). By construction, the terms of trade are chosen so that the allocation is pairwise Pareto efﬁcient. The allocation solves the following problem: U s (m, b) = max −c(q) + φ (dm + db )

(10.64)

subject to u(q) − φ (dm + db ) ≥ U b (m, b)

(10.65)

q,dm ,db

and

0 ≤ dm ≤ m, 0 ≤ db ≤ b.

(10.66)

It is important to emphasize that the use of bonds as means of payment is unrestricted, see condition (10.66). Moreover U s (m, b) ≥ 0 since the allocation determined in the ﬁrst step ofthe pricing procedure is still feasible in the second step. If φ (m + b) ≥ u q∗ − U b (m, b), then the terms of trade in a bilateral meeting in the DM satisfy q = q∗

(10.67)

φ (dm + db ) = u(q∗ ) − U b (m, b);

(10.68)

otherwise, the terms of trade are given by q =u−1 φ(m + b) + U b (m, b)

(10.69)

(dm , db ) =(m, b).

(10.70)

The seller’s payoff and output in the DM are uniquely determined. The composition of the payment between money and bonds will be unique if ∗ the output produced in the DM is bstrictly less than the efﬁcient level, q . ∗ If, however, φ (m + b) > u q − U (m, b), then there are a continuum of transfers (dm , db ) that can achieve (10.68). As before, the determination of

282

Chapter 10

the terms of trade is illustrated in ﬁgure 10.4. The lower (dashed) frontier corresponds to the pair of surplus utility levels in the ﬁrst step of the pricing protocol, where the buyer cannot spend more than a fraction g of his bond holdings. The upper frontier corresponds to the pair of utility levels in the second step of the procedure, where payments are unconstrained. Given this pricing mechanism, the expected lifetime utility of the buyer holding portfolio (m, b) in the DM is given by V b (m, b) = σ U b (m, b) + W b (m, b).

(10.71)

With probability σ the buyer is matched, in which case he enjoys the surplus U b (m, b). If we substitute V b (m, b) from (10.71) into the buyer’s portfolio problem, (10.32), and rearrange, the buyer’s choice of portfolio is given by the solution to

r − rb max −rφm − φb + σ U b (m, b) , 1 + rb m≥0,b≥0 where, as above, [(r − rb )/(1 + rb )] represents the cost of holding bonds. Note that this portfolio problem is identical to (10.38). Consequently the buyer’s demands for money and bonds are identical to the ones in the liquidity-constrained economy described in chapter 10.3.2, and the rate of return of bonds is given by (10.41). Our model with bilateral trades is able to generate a rate-of-return differential between money and risk-free bonds, even though there are no restriction on the use of bonds as means of payment. Fiat money and bonds share the same physical properties in terms of divisibility and recognizability, and the allocations in bilateral matches are pairwise Pareto efﬁcient. The explanation for the rate-of-return dominance is that different assets are traded at different prices. Indeed from (10.63), if the buyer an additional unit of money, then his surplus increases by holds φ u (q)/c (q) − 1 , whereas if he holds an additional unit of bonds, then his surplus increases by φg u (q)/c (q) − 1 . Hence the marginal unit of bond commands a surplus that is g times the surplus that the marginal unit of money generates. 10.6 Further Readings Kiyotaki and Wright (1989) construct an environment where commodities are storable and can serve as means of payment, but they differ in

Competing Media of Exchange

283

terms of their storage costs. They show that the goods that emerge as media of exchange depend on the storage costs, as well as preferences and technologies through the pattern of specialization. Models of commodity monies include Sargent and Wallace (1983), Li (1995), Burdett, Trejos, and Wright (2001), and Velde, Weber, and Wright (1999). The existence of a monetary equilibrium when agents have access to a linear storage technology is studied by Wallace (1980) in the context of an overlapping-generations model. Lagos and Rocheteau (2008) study how money and capital can compete as means of payment in a search environment. Shi (1999a, b), Aruoba and Wright (2003), Molico and Zhang (2006), and Aruoba, Waller, and Wright (2007) describe search economies where agents can accumulate capital, but capital is illiquid in the sense that it cannot be used as a means of payment in bilateral matches. The effect of inﬂation on capital accumulation was studied in reduced-form monetary models by Tobin (1965) and Stockman (1981). Two-country cash-in-advance models are described in Obstfeld and Rogoff (1996, app. 8A). The ﬁrst search-theoretic environment with two currencies was proposed by Kiyotaki, Matsui, and Matsuyama (1993) and extended by Zhou (1997) to allow for currency exchange. These authors consider two-country economies and establish conditions on parameters for which one currency is used as an international currency. They also show that a uniform currency dominates in terms of welfare. Other models with multiple currencies include Head and Shi (2003), Camera and Winkler (2003), Craig and Waller (2004), Camera, Craig, and Waller (2004), Liu, and Shi (2010), and Ales et al. (2008). The proposition about the indeterminacy of the exchange rate is established by Kareken and Wallace (1981) in the context of an overlapping-generations economy. Our method to determine the exchange rate adopts the trading mechanism proposed in Zhu and Wallace (2007). Another method uses legal restrictions as in Li and Wright (1998), Curtis and Waller (2000, 2003), Li (2002), Lotz and Rocheteau (2002), and Lotz (2004). Kocherlakota and Kruger (1999) and Kocherlakota (2002) discuss the usefulness of two currencies. Trejos and Wright (1996) and Craig and Waller (2000) survey the search literature on dual-currency payment systems. The coexistence of money and bonds is discussed in Bryant and Wallace (1979), Wallace (1980), Aiyagari, Wallace, and Wright (1996), Kocherlakota (2003), Shi (2004, 2005), and Zhu and Wallace (2007). According to Bryant and Wallace (1979), interest-bearing government bonds are socially inefﬁcient because of intermediation costs to

284

Chapter 10

transform large-denomination bonds into perfectly divisible intermediary liabilities. Aiyagari, Wallace, and Wright (1996) introduce government agents to explain why government bonds are sold at a discount. Kocherlakota (2003), Boel and Camera (2006), and Shi (2008) show that illiquid bonds can raise society’s welfare when agents are subject to idiosyncratic shocks. The approach in this chapter to explain the coexistence of money and interest-bearing bonds due to the counterfeitability of bonds is taken from Li and Rocheteau (2009).

11

Liquidity, Monetary Policy, and Asset Prices

In frictionless economies the price of an asset depends only on the discounted value of its expected future income ﬂows. In monetary economies assets can in addition be valued for their liquidity properties. These assets are helpful in overcoming the frictions that plague monetary economies, which prevent the use of credit arrangements. An obvious case in point is ﬁat money. As shown in earlier chapters, if there is limited enforcement and lack of record keeping, ﬁat money can have a positive value even though its fundamental value, i.e., the discounted sum of its dividends, is zero. Other assets can play a similar role in facilitating exchange by, for instance, being used as collateral to secure loans. This dual role of assets can help account for various asset pricing anomalies that would be difﬁcult to explain in a frictionless economy. In this chapter we investigate how asset prices are determined in monetary economies. We ﬁrst examine an environment where there is a ﬁxed supply of real assets and no money. The real asset, or claims to it, can serve as a medium of exchange in decentralized trades, just as ﬁat money did in earlier chapters. When the amount of real assets is relatively low, then there is a shortage of liquidity and the asset price is higher than its fundamental value. The difference between the asset price and the fundamental value represents the liquidity value (or premium) of the asset. The price of the asset also depends on agents’ liquidity needs and the extent of trading frictions in asset markets. A prediction of the model is that asset prices tend to increase in markets where it is easier to ﬁnd a counterparty for a trade. In order to study the effects of inﬂation on asset prices, we introduce ﬁat money into the environment that has a ﬁxed supply of real assets. Fiat money can be valued if the supply of real assets is low relative to agents’ liquidity needs and if the inﬂation rate is not too large. In

286

Chapter 11

a monetary equilibrium the rate of return of the real asset is equal to the rate of return of ﬁat money. This rate-of-return equality implies a positive relationship between asset prices and inﬂation. The rate-of-return equality breaks down if the real asset pays a risky dividend despite agents being risk neutral in terms of the dividend good. The asset pays a risk premium because its riskiness affects its role as a medium of exchange. More precisely the risky asset pays a high dividend when liquidity needs are low, and a low dividend when liquidity needs are high. This feature makes the risky asset less attractive as a medium of exchange, compared to risk-free assets such as ﬁat money or government bonds. Qualitatively speaking, this suggests that liquidity considerations can provide an explanation for abnormally high risk premia. Finally, we explain rate-of-return differences across assets as stemming from liquidity differences. These liquidity differences can arise from bargaining conventions or social norms, which affect the terms at which these assets are traded. The liquidity differences can also reﬂect informational asymmetries regarding the value of assets. 11.1 A Monetary Approach to Asset Prices In this section we provide a simple model, where monetary considerations matter for asset prices. Consider an economy that is identical to the one studied in previous chapters, where agents trade alternatively in centralized, CM, and decentralized markets, DM. See ﬁgure 11.1. The economy is endowed with a single real asset, such as, a tree, that is in ﬁxed supply, A > 0, and can be traded in both markets. One can think of the bilateral matches in the DM as an over-the-counter asset market. We will provide a description of such a market in chapter 12. At the beginning of each night period, before the CM opens, each unit of the real asset generates a dividend payoff equal to κ > 0 units of the general, or CM, good, for example, the fruits of the tree. Consequently the asset in the CM is traded ex-dividend: the dividend belongs to the agent who holds the asset at the beginning of the CM. Note that if κ approaches 0, then the asset becomes intrinsically useless, and is similar to ﬁat money. The price of the asset, measured in terms of the CM good in period t, is denoted by pt . We consider stationary equilibria, where pt is constant over time.

Liquidity, Monetary Policy, and Asset Prices

NIGHT (CM)

DAY (DM)

287

NIGHT (CM)

Agent’s portfolio:

(p+κ)a

pa

Assets’ returns Figure 11.1 Timing and asset’s returns

The value function of a buyer entering the CM holding a portfolio of a units of the real asset is W b (a) = max x − y + βV b a (11.1) x,y,a

subject to pa + x = y + a(p + κ).

(11.2)

According to (11.1), in the CM, the buyer chooses his net consumption of the CM good, x − y, and the quantity of assets, a , that he will bring into the subsequent DM. Equation (11.2) is the buyer’s budget constraint expressed in terms of the CM good. In the CM, one unit of the real asset generates κ units of the CM good and can be sold at the competitive price, p; see ﬁgure 11.1. Substituting x − y from the budget constraint into (11.1) and rearranging, we get b W b (a) = a(p + κ) + max −pa a . (11.3) + βV a ≥0

The CM value function is linear in the buyer’s wealth, a(p + κ), and his choice of asset holdings, a , is independent of the assets, a, he brought into the CM. If a buyer is matched with a seller in the DM, he makes a take-itor-leave-it offer q, da , where da represents the assets that the buyer transfers to the seller in exchange for q units of the DM good. An alternative interpretation is that the asset is used as collateral for a secured loan, and it is only transferred to the seller if the buyer defaults in the CM. Suppose that the buyer brings a units of the asset to the DM. The value of these assets in the subsequent CM is a(p + κ). If a(p + κ) ≥ c(q∗ ), then the buyer’s offer is characterized by q = q∗ and da = c(q∗ )/(p + κ),

288

Chapter 11

where da is sufﬁcient to compensate the seller for producing q∗. If, how- ever, a(p + κ) < c(q∗ ), then the buyer’s offer is given by q = c−1 a(p + κ) and da = a, i.e., the buyer spends all his asset holdings to get q. Consequently the value function of a buyer holding a units of asset at the beginning of the DM is V b (a) = σ u(q) + W b (a − da ) + (1 − σ )W b (a) = σ u(q) − c(q) + a(p + κ) + W b (0), (11.4) ∗ where (p + κ)da = c(q) = min c(q ), a(p + κ) . In going from the ﬁrst to the second equality above, we used the fact that W b is linear and that the buyer receives all of the surplus from exchange. According to (11.4) the buyer is in a DM match with probability σ , in which case he extracts the entire match surplus, u(q) − c(q). Substituting V b (a) from (11.4) into (11.3), we ﬁnd that the buyer’s choice of asset holdings solves max −ar p − p∗ + σ u(q) − c(q) , (11.5) a≥0

where p∗ ≡ κ/r is the discounted sum of dividends, i.e., the price of the asset in a frictionless economy. The price p∗ will be referred to as the fundamental value of the asset. The buyer maximizes his expected surplus in the DM, net of the cost of holding the real asset. The cost of holding the asset is the difference between the price of the asset and its fundamental value, times the discount rate, r. The ﬁrst-order condition from the buyer’s problem (11.5), assuming an interior solution, is u (q) −r p − p∗ + σ − 1 (p + κ) = 0. (11.6) c (q) If p < p∗ , then (11.5) has no solution; in this situation there would be an inﬁnite demand for the asset. If p = p∗ , then, u q = c q , i.e., q = q∗ . In this situation any a ≥ c(q∗ )/(p∗ + κ) is a solution to the buyer’s problem; the buyer has sufﬁcient wealth to purchase the efﬁcient level of the DM good. Finally, if p > p∗ , then there is a unique a that solves (11.6), and it is ∗ / p + κ is increasing in decreasing with p. To see this note that r p − p p, and that u q /c q is decreasing in p and a. Moreover p > p∗ implies that σ u (q)/c (q) − 1 > 0, where q = c−1 a p + κ , and hence q < q∗ . When the price of the asset is above its fundamental value, it is costly

Liquidity, Monetary Policy, and Asset Prices

289

to accumulate the asset, and buyers will not hold enough of the asset to purchase the efﬁcient level of output in the DM, q∗ . Since sellers do not obtain any surplus in the DM, choice their of asset holdings in the CM is simply given by maxa≥0 −ar p − p∗ . Since, in any equilibrium, p ≥ p∗ , sellers will be willing to hold the asset only if its price is equal to its fundamental value and, at that price, they are indifferent between holding and not holding the asset. So, without loss in generality, we assume that in equilibrium sellers do not hold assets. Let the set of all buyers be the interval [0, 1], and let a j be buyer j’s, j ∈ [0, 1], demand for the asset. The aggregate demand correspondence for the asset is

d A (p) = a(j)dj : a(j) is a solution to (11.5) . [0,1]

The clearing of the asset market requires that A ∈ Ad (p),

(11.7)

where A is the ﬁxed supply of the real asset. The market-clearing price, denoted pe , is illustrated in ﬁgure 11.2. The aggregate demand correspondence, Ad (p), is single-valued for all p > p∗ —see equation (11.6)—and is equal to c(q∗ )/(p∗ + κ), + ∞ for p = p∗ . Consequently there is a unique p ≥ p∗ that solves (11.7). Graphically the solution is given by the intersection of the aggregate demand correspondence, Ad (p), and the ﬁxed supply of the real asset, A. If A ≥ c(q∗ )/(κ + p∗ ), then there is enough wealth in the economy, ∗ (p + κ)A, to purchase the efﬁcient level of the DM good, q∗ . In this case the asset is priced at its fundamental value, p = p∗ , because the expected increase in the buyer’s surplus associated unit of the with an additional real asset at the beginning of the DM, σ u q /c q − 1 (p + κ), is equal to zero. In other words, the asset has no liquidity value at the margin. In contrast, if A < c(q∗ )/(κ + p∗ ), then there is insufﬁcient wealth in the economy to purchase the efﬁcient quantity of the DM good. Here the expected increase in the buyer’s surplus associated with an additional unit of the real asset is strictly positive, which implies that the price of the real asset, p, is above its fundamental value, p∗ ; see equation (11.6). Buyers are now willing to pay more than the fundamental value for the asset since, in addition to the dividend ﬂow, an additional unit of the asset provides some liquidity—or additional surplus—in the DM. This difference between p and p∗ would be viewed as an anomaly in a frictionless

290

Chapter 11

Ad (p)

c (q*) p* +κ

A

p*

pe

p

Figure 11.2 Equilibrium of the asset market

economy since in that economy an additional unit of the asset would not provide any additional surplus in the DM. This simple model has predictions regarding the effects that trading frictions and the supply of the asset have on the asset price. The expression for the asset price, (11.6), can be rewritten as

u (q) p+κ ∗ p = p +σ , (11.8) −1 c (q) r where c(q) = min c(q∗ ), A(p + κ) . The ﬁrst term on the right side of (11.8) represents the fundamental value of the asset and the second term is the liquidity value of the asset, i.e., the increase in the expected surplus of the buyer in the DM from holding an additional unit of asset. Assuming that q < q∗ , as the trading friction σ is reduced, the asset price increases, i.e., ∂p/∂σ > 0, since the asset can be used more often as means of payment and, as a consequence, its liquidity value goes up. As well, as agents become more impatient, the asset price falls, i.e., ∂p/∂r < 0. In this case agents discount both the dividend of the asset and

Liquidity, Monetary Policy, and Asset Prices

291

its future liquidity returns more heavily, which results in lower asset values. Finally, as κ tends to zero, the asset becomes like ﬁat money since p∗ → 0, and from (11.6) its price is given by the solution to u (q) −r + σ − 1 = 0. c (q) Not surprisingly, as the value of the dividend approaches zero, the price of the asset approaches the value of ﬁat money that was derived in chapter 4.1, equation (4.14), when φt+1 = φt . 11.2 Monetary Policy and Asset Prices What is the relationship between monetary policy and asset prices? Does monetary policy affect asset prices, and what is the optimal monetary policy when asset prices respond to a change in the money growth rate? We use the model developed in the previous section to answer these questions. In order to talk about monetary policy, we must reintroduce ﬁat money into our economy. We assume that the stock of money grows at the constant rate, γ = Mt+1 /Mt , and is injected or withdrawn via lump-sum transfers to buyers in the CM. We will focus on stationary monetary equilibria, where real balances are constant over time, i.e., φt+1 Mt+1 = φt Mt , and φt is the amount of the CM good that one unit of ﬁat money can buy in period t. The value function of a buyer holding portfolio (a, z) at the beginning of the CM, where a represents the buyer’s holdings of the real asset and z = φt m represents his holding of real money balances, generalizes (11.3) in the obvious way. This value function, W b (a, z), is given by b − γ z + βV , z W b (a, z) = a(p + κ) + z + T + max −pa a , (11.9) a ≥0,z ≥0

where the lump-sum transfer or tax received by buyers, T ≡ φt (Mt+1 − Mt ), is expressed in terms of the CM good. As above, the buyer’s CM value function is linear in his wealth, which now includes his real balance holdings. Note that if the buyer wishes to hold z = φt+1 m units of real balances in period t + 1, he must produce φt m = (φt /φt+1 )z = γ z in period t, see ﬁgure 11.3. Since the terms of trade in the DM are determined by the buyer making a take-it-or-leave-it offer to the seller, the value function for a buyer holding portfolio (a, z) at the beginning of the period, which generalizes

292

Chapter 11

NIGHT (CM)

DAY (DM)

NIGHT (CM)

Agent’s portfolio:

(p+κ)a + γ −1 z

pa + z

Assets’ returns Figure 11.3 Timing and assets’ returns

(11.4), is given by V b (a, z) = σ u(q) − c(q) + a(p + κ) + z + W b (0, 0), (11.10) ∗ where c(q) = min c(q ), a(p + κ) + z . The buyer’s portfolio problem—which is described by the last term in (11.9)—can be re-expressed by substituting the expression for V b (a, z) given by (11.10) into (11.9), and simplifying, i.e., max −iz − ar p − p∗ + σ u(q) − c(q) , (11.11) a≥0,z≥0

where i = (γ − β)/β is the cost of holding real balances. Note that this problem generalizes (11.5) in the previous section. The buyer chooses his portfolio, composed of money and the real asset, in order to maximize his expected surplus in a bilateral match, net of the cost of holding the real asset and money. In order to characterize the buyer’s asset demand correspondence, it will be convenient to deﬁne ≡ z + a(p + κ) as the buyer’s liquid wealth that is available to purchase the DM good in a bilateral match. The buyer’s portfolio problem, (11.11), can be, equivalently, written as max −i − (r − i)p − (1 + i)κ a + σ u(q) − c(q) (11.12) a,

subject to a(p + κ) ≤ , (11.13) ∗ where c(q) = min c(q ), . The squared bracketed term that premultiplies a in problem (11.12) has an interesting and intuitive interpretation. ∗ − i κ + p . The This term can be rearranged to read as − r p − p ﬁrst term in this difference, r p − p∗ , is the cost of holding a unit of the real asset between one CM and the next, and the second term is the cost of holding the equivalent amount of real balances. Hence

Liquidity, Monetary Policy, and Asset Prices

293

r p − p∗ − i κ + p represents the relative cost of holding wealth in the real asset compared to holding it in ﬁat money. There are three cases to consider: 1. r p − p∗ < i κ + p : Money is more costly to hold than the real asset. The constraint a(p + κ) ≤ will bind, which implies that z = 0. If we substitute z = 0 into (11.11), then the buyer’s problem is exactly the same as problem (11.5), and therefore his choice of asset holdings, a, is given by (11.6). 2. r p − p∗ > i κ + p : The real asset more costly to hold than money. Buyers will demand only real balances and a = 0. 3. r p − p∗ = i κ + p : Money and the real asset are equally costly to hold, which implies that the buyer is indifferent between holding the real asset and ﬁat money. In this case, the value of the portfolio, , solves the ﬁrst-order condition, u ◦ c−1 ( ) i = σ −1 −1 , (11.14) c ◦ c ( ) and the asset price is p=

(1 + i)κ . r−i

(11.15)

We denote (i) as the solution to (11.14); (i) is the demand for liquid assets, as a function of the cost of holding real balances. The aggregate asset demand correspondence, Ad (p), is illustrated in ﬁgure 11.4. The correspondence is constructed assuming that i > 0. If p = p∗ , then necessarily z = 0—since the real asset is costless to hold but money is not—and (11.11) or, equivalently, (11.12) simpliﬁes problem to maxa σ u q − c q , which implies that any a ≥ c(q∗ )/(p∗ + κ) is a solution and that q = q∗ . If p ∈ p∗ , (1 + i)κ/(r − i) , then z = 0, and a is the unique solution to (11.6), and is decreasing in p. In this situation, although the real asset is costly to hold, money is even more costly. This part of the asset demand correspondenceis identical to the one in ﬁgure 11.2. If p = [(1 + i)κ]/(r − i), then any a ∈ 0, (i)/(p + κ) is a solution to (11.12) since the buyer is indifferent between holding the real asset and money, i.e., the real asset and money are equally costly to hold. In this case the real value of the buyer’s portfolio, (i), is given by solution to (11.14). Finally, if p > [(1 + i)κ]/(r − i), then it is cheaper to hold money than the real asset and, as a result, a = 0.

294

Chapter 11

Ad (p)

Nonmonetary equilibrium

c (q*) p* +κ

A' Monetary equilibrium

r −i (i) κ (1 + r) A

p*

κ (1 + i) r−i

p

Figure 11.4 Fiat money and the demand for assets

Market clearing in the asset market requires that A ∈ Ad (p). The asset price is uniquely determined by the intersection of the aggregate demand correspondence, Ad p , and the horizontal supply, A; see ﬁgure 11.4. A monetary equilibrium exists if A < [(r − i)/κ(1 + r)] (i). A necessary, but not sufﬁcient, condition for money to be valued is that the stock of real assets is less than c(q∗ )/(p∗ + κ) or, in other words, the supply of the real asset must not be large enough to allow agents to trade the efﬁcient quantity in the DM. In a monetary equilibrium, r − i > 0, which implies that the inﬂation rate must be negative or that, equivalently, the money supply must contract, i.e., γ < 1. Note that in ﬁgure 11.4, if the supply of assets is A , then the inﬂation rate is too large for ﬁat money to be valued. In a monetary equilibrium the price of the real asset is increasing with the rate of inﬂation; see (11.15), where ∂p/∂i > 0. Graphically, as

Liquidity, Monetary Policy, and Asset Prices

295

i increases, the vertical portion of the aggregate demand curve Ad (p) moves to the right. As inﬂation increases, it becomes more costly to hold real balances, and buyers demand a higher quantity of real assets to be used as means of payment, which in turn drives asset prices up. Notice the difference here—where the asset supply is ﬁxed—compared to the analysis in chapter 10.1—where capital goods could be produced one for one from the CM good. In that case an increase in inﬂation did not affect the price of capital—which was always equal to one—but instead resulted in buyers over accumulating capital. The gross rate of return of the real asset is Ra = (p + κ)/p = 1 + (κ/p). In a monetary equilibrium, the rate of return of the real asset can, from (11.15), be expressed as Ra =

1+r = γ −1 , 1+i

(11.16)

i.e., the rate of return of the real asset equals the rate of return of ﬁat money. We have seen this principle of the equality of rates of return on assets earlier in chapter 10.1. Since Ra = (p + κ)/p > 1, the gross growth rate of money must be less than one, γ < 1. This is an alternative way to see that in order for money to be valued, the money supply must contract, i.e., there must be a deﬂation. In a monetary equilibrium the optimal monetary policy will drive the cost of holding real balances, i, to zero. From (11.14), as i tends to zero, the buyer’s liquid wealth, , tends to c(q∗ ), and the output traded in bilateral matches approaches its efﬁcient level, q∗ . In this situation the asset price converges to its fundamental value (see equation 11.15), since real balances are costless to hold and, as a result, at the margin, the real asset does not provide any additional liquidity. When the asset price converges to its fundamental value, the gross rate of return on all assets will converge to one plus the rate of time preference, 1 + r. 11.3 Risk and Liquidity So far we have assumed that the real asset is risk-free in the sense that it provides a constant ﬂow of dividend in every period. Given agents’ quasi-linear preferences, the riskiness of the asset is irrelevant for asset pricing provided that one of the following two conditions is valid: (1) the real asset plays no role as a means of payment, or (2) the value of the dividend is not realized until after the DM closes. In this section we assume that neither condition 1 nor 2 holds, i.e., the real asset is useful for

296

Chapter 11

NIGHT (CM) Agent’s portfolio:

pa + z

DAY (DM)

NIGHT (CM)

Dividend shock:

πH

πL

κ = κH

(p+κ) a + γ −1 z

κ = κL

Assets’ expected returns Figure 11.5 Timing and assets’ returns

facilitating exchange and the dividend realization is known at the time of bilateral exchange. These assumptions allow us to uncover a new channel through which the riskiness of an asset affects its liquidity and price. We assume that the dividend of the real asset follows a simple stochastic process: with probability πH , the dividend payment is high, κH , and with complementary probability, πL ≡ 1 − πH , it is low, κL , where κL < κH . The dividend shocks are independent across time. We denote the expected dividend by κ¯ = πH κH + πL κL and assume that buyers and sellers learn the dividend realization at the beginning of the period, before they are matched in the DM. The timing and information structure are illustrated in ﬁgure 11.5. At the beginning of the CM, the value function of a buyer is similar to (11.9), i.e., W b (a, z, κ) = a(p + κ) + z + W b (0, 0, κ). We introduce κ as an explicit argument since it is no longer constant over time. The terms of trade in a bilateral match in the DM are determined by a take-it-or-leave-itoffer by the buyer to the seller. The output traded solves c(qH ) = min c(q∗ ), a(p +κH ) + z in the high-dividend state and c(qL ) = min c(q∗ ), a(p + κL ) + z in the low-dividend state. The value function for a buyer holding portfolio (a, z) at the beginning of the DM, before the dividend realization is known, V b (a, z), is given by V b (a, z) = σ πH u(qH ) + W b (a − da,H , z − dz,H , κH ) + σ πL u(qL ) + W b (a − da,L , z − dz,L , κL ) + (1 − σ ) πH W b (a, z, κH ) + πL W b (a, z, κL )

Liquidity, Monetary Policy, and Asset Prices

297

= σ πH u(qH ) − c(qH ) + πL u(qL ) − c(qL ) + a(p + κ) ¯ + z + πH W b (0, 0, κH ) + πL W b (0, 0, κL ),

(11.17)

where (da,H , da,L , dz,H , dz,L ) is a vector of asset transfers in the two dividend states. Going from the ﬁrst equality to the second equality in (11.17), we have used the linearity of W b . According to (11.17), independent of the realization of the dividend, the buyer always extracts the entire surplus of the match. With probability πH , the realization of the dividend is high and agents trade qH in the DM, and with probability πL , the dividend is low and agents trade qL . If a(p + κL ) + z < c(q∗ ), then the quantity traded in the low dividend state is less than that traded in the high-dividend state, i.e., qL < qH . If we substitute V b (a, z) from (11.17) into (11.9), then the buyer’s portfolio problem in the CM can be expressed as max −iz − ar p − p∗ + σ πH u(qH ) − c(qH ) + πL u(qL ) − c(qL ) ,

a≥0,z≥0

where now p∗ = κ/r. ¯ The ﬁrst-order (necessary and sufﬁcient) conditions for this problem are

u (qH ) u (qL ) −i + σ πH − 1 + πL − 1 ≤ 0, (11.18) c (qH ) c (qL )

u (qH ) −r p − p∗ + σ πH p + κH − 1 c (qH ) u (qL ) +πL p + κL − 1 ≤ 0, (11.19) c (qL ) where (11.18)holds at equality if z > 0, and at equality if a > (11.19) holds 0. The terms p + κH u (qH )/c (qH ) − 1 and p + κL u (qL )/c (qL ) − 1 represent the liquidity values of having an additional unit of the real asset in the high and low dividend states, respectively, for a buyer in a trade match. According to (11.18) and (11.19) the buyer chooses his portfolio so as to equalize the cost of holding an asset with its expected liquidity return in the DM. In any equilibrium the ﬁxed stock of real assets must be held, and therefore (11.19) must hold at equality. The asset price, p, satisﬁes

σ u (qH ) u (qL ) p = p∗ + πH (p + κH ) − 1 + πL (p + κL ) − 1 . (11.20) r c (qH ) c (qL )

298

Chapter 11

The ﬁrst component on the right side of (11.20) is the fundamental value of the asset, while the second component is the expected discounted liquidity value of the asset in the DM. We ﬁrst consider the case where the efﬁcient allocation, qH = qL = q∗ , can be achieved. From the pricing equation (11.20) this implies that p = p∗ , and from (11.18) ﬁat money will not be valued for any i > 0. A sufﬁcient condition for q∗ to be implementable in all states is that the real value of the stock of assets in the low dividend state is large enough to compensate sellers for their costs associated with producing q∗ , i.e., A(p∗ + κL ) ≥ c(q∗ ).

(11.21)

If (11.21) holds, then the efﬁcient allocation can be implemented as an equilibrium without ﬁat money. If condition (11.21) fails to hold and i > 0, then qL < q∗ , and the price of the asset will rise above its fundamental value. Provided that i is sufﬁciently small, ﬁat money can have a strictly positive value. equilibrium, (11.18) and (11.19) imply that In any monetary i ≤ (p − p∗ )/(p + κL ) r < r. To see this, divide (11.19) by (p +κL ) to obtain −r

p − p∗ p + κL

p + κH u (qH ) u (qL ) + σ πH − 1 + π − 1 ≤ 0. L p + κL c (qH ) c (qL )

Since (p + κH )/(p + κL ) > 1 and u (qH )/c (qH ) − 1≥ 0, (11.18) and (11.19) will hold at equality only if r (p − p∗ )/(p + κL ) ≥ i. Hence, as in the previous section, any monetary equilibrium will be characterized by a negative inﬂation rate. To better illustrate the pricing relationship between ﬁat money and the real asset, we introduce the covariance of the value of the real asset at the beginning of the period, p + κ, and the marginal return of wealth in the DM, u (q)/c (q)−1. Denote this covariance as ρ, where, by deﬁnition, u (qH ) u u (qL ) u + πL (κL − κ) , (11.22) ρ = πH (κH − κ) − − ¯ ¯ c (qH ) c c (qL ) c and u /c = πH u (qH )/c (qH ) + πL u (qL )/c (qL ) . Using (11.18) and (11.22), we can express the price for the real asset, p, given by (11.20) simply as p=

(1 + i)κ¯ + σρ . r−i

(11.23)

Liquidity, Monetary Policy, and Asset Prices

299

For the derivation of this asset price, see the appendix. Comparing this expression for the price of the real asset with the expression (11.15)— where there was no information revealed regarding the dividend payoff before the opening of the CM—we see that the former has an additional component, σρ/(r − i), which is proportional to the covariance between the risky dividend and the marginal utility of wealth in the DM. To determine the sign of the covariance term, note that πH (κH − κ) ¯ + πL (κL − κ) ¯ = 0 and, since qH > qL , u (qH )/c (qH ) < u (qL )/c (qL ). These two observations imply that u (qH ) u u (qL ) u ρ = πH (κH − κ) + πL (κL − κ) ¯ ¯ − − c (qH ) c c (qL ) c

u (qH ) u (qL ) = πH (κH − κ) − < 0. ¯ c (qH ) c (qL ) We now discuss the effect that this new component has on the asset pricing. Let’s ﬁrst compare the (gross) rates of return on money, γ −1 , with the return of the real asset, Ra = (p + κ)/p. ¯ The rate of return on the real asset, using equation (11.23), can be expressed as

(1 + r)κ¯ + σρ (γ − 1) Ra = = γ −1 1 + σρ , (11.24) (1 + i)κ¯ + σρ κ(1 ¯ + i) + σρ since (1 + r) = (1 + i) γ −1 . From (11.23) in any monetary equilibrium r > i, which implies that γ < 1. Since (γ − 1) σ/ [κ(1 ¯ + i) + σρ] < 0, the rate−1 of-return differential Ra − γ has the opposite sign of the covariance between the risky dividend and the marginal utility of wealth in the DM, ρ. Since the covariance, ρ, is negative, from (11.24), Ra > γ −1 .

(11.25)

Therefore the risk-free real asset with a dividend payment equal to κ, ¯ will be more expensive than a risky real asset that delivers an expected dividend of κ; ¯ see equation (11.23). When real assets can be used for transactions purposes and agents know the dividend realization in the DM matches, the rate-of-returnequality principle no longer holds. A rate-of-return differential arises because the real asset is used as a means of payment in the DM and individuals are risk-averse. The real asset yields a high dividend in matches where the marginal value of wealth is low, and a low dividend in matches where the marginal value of wealth is high. In contrast, the

300

Chapter 11

rate of return of money is constant and uncorrelated with the marginal utility of wealth in the DM. Consequently money has a higher liquidity return than the real asset, and hence a lower rate of return than that of the real asset. Finally, as i → 0, qH → q∗ and qL → q∗ , which implies that ρ → 0 and Ra = γ −1 = β −1 . In words, at the Friedman rule, ﬁat money and the real asset will have the same rate of return equal to the (gross) rate of time preference, and the ﬁrst-best allocation is obtained. 11.4 The Liquidity Structure of Assets’ Yields In this section we examine the structure of assets’ yields and how it is affected by monetary policy. We extend the model in chapter 11.2 to allow for a ﬁnite number K ≥ 1 of inﬁnitely lived real assets indexed by k ∈ {1, . . ., K}. Denote Ak > 0 as the ﬁxed stock of asset k ∈ {1, . . ., K}, κk as its expected dividend in terms of the CM good in the CM, and pk as its price in the CM in terms of the CM good. In contrast to chapter 11.3, we assume that agents do not learn the dividend realization (if the dividend is risky) until the beginning of the CM. Consequently the terms at which the asset is traded in the DM will only depend on its expected dividend, κk . In order to generate rate-of-return differentials, we now assume that a buyer in a bilateral match can only transfer a fraction νk ∈ [0, 1] of his holdings of asset k to the seller. Asset k is said to be partially illiquid if 0 < νk < 1, and is more liquid than asset k if νk > νk . The parameters νk can be interpreted as capturing either institutional constraints or informational frictions that make some assets harder to liquidate than others. In the subsequent section we will deal with these liquidity constraints more formally, (see also chapters 10.4 and 10.5). Consider a buyer in a bilateral match in the DM with a portfolio ({ak }K k=1 , z), where ak is the quantity of the kth real asset and z is real balances. We assume that the terms of trade are determined by the buyer making a take-it-or-leave-it offer, (q, dz , {dk }K k=1 ), to the seller, where q is the buyer’s consumption of the DM good, dz is the transfer of real balances, and dk is the transfer of the asset k. The buyer’s surplus from a match in the DM is given by U b = max

q,dz ,{dk }

u(q) − dz −

K k=1

dk (pk + κk )

(11.26)

Liquidity, Monetary Policy, and Asset Prices

subject to

− c(q) + dz +

K

dk (pk + κk ) ≥ 0

301

(11.27)

k=1

and

dz ≤ z, dk ≤ νk ak .

(11.28)

According to (11.26) the buyer maximizes his utility of consumption net of the transfer of assets. The transfer of one unit of real balances is worth one unit of the CM good, while the transfer of one unit of asset k is worth pk + κk units of the CM good. Condition (11.27) is the seller’s participation constraint. The ﬁnal constraint, (11.28), is a feasibility condition that says the buyer cannot transfer more than his real balances and a fraction νk of asset k. The solution to (11.26)–(11.28) is

if ≥ c(q∗ ), u(q∗ ) − c(q∗ ) U b ( ) = (11.29) −1 u ◦ c ( ) − otherwise, where = z + K k=1 νk ak (pk + κk ) is the value of the assets that the buyer can transfer to the seller in exchange for the DM good. We will refer to as the buyer’s liquid portfolio. If the value of this liquid wealth is greater than c(q∗ ), then the buyer can ask for the efﬁcient quantity q∗ ; otherwise, he will transfer all his liquid wealth in exchange for a quantity q of output less than q∗ . Suppose that the buyer’s liquidity constraint dz + K k=1 dk (pk + κk ) ≤ is binding, so that c q = . Then u q ∂U b = νk (pk + κk ) − 1 , ∂ak c q ∂U b u q = − 1, ∂z c q which implies that (pk + κk )−1

∂U b ∂U b = νk . ∂ak ∂z

In words, 1/(pk + κk ) units of the kth asset, which is a claim to one unit of CM good, allows the buyer to raise his surplus in a bilateral match in the DM by a fraction νk of what he would obtain if he would accumulate one additional unit of real balances. This way the parameter νk is a measure of the liquidity of the asset k, and the extent to which it allows buyers to capture a fraction of the gains from trade in the DM market. If we

302

Chapter 11

assume that the liquidity coefﬁcients are ranked as ν1 ≥ ν2 ≥ . . . ≥ νK , then ﬁat money is the most liquid asset and the asset K is the least liquid. The buyer’s portfolio problem in the CM is a straightforward generalization of the problem with two assets, (11.11), and is given by the solution to K ∗ b max −iz − r (11.30) ak pk − pk + σ U ( ) , {ak },z

k=1

where pk∗ = κk /r represents the fundamental price of asset k. According to (11.30) the buyer maximizes the expected surplus in the DM, net of the cost of holding the different assets in his portfolio. The cost of holding asset k is the difference between the price of the asset and its fundamental value times the discount rate, r, while the cost of holding real balances is i = (γ − β)/β. Since z = − K k=1 νk ak (pk + κk ), the buyer’s portfolio choice problem, (11.30), can be rewritten as max −i + {ak },

K

ak iνk (pk + κk ) − r

pk − pk∗

b

+ σ U ( )

(11.31)

k=1

subject to

K

νk ak (pk + κk ) ≤ .

(11.32)

k=1

In a monetary equilibrium, constraint (11.32) does not bind since z > 0, and the ﬁrst-order condition with respective to is u ◦ c−1 ( ) i = σ −1 −1 . c ◦ c ( ) Let (i) denote the solution to this equation. The demand for liquid assets decreases with i, i.e., (i) < 0. In a monetary equilibrium, buyers must be indifferent between holding asset k and ﬁat money; hence iνk (pk + κk ) − r pk − pk∗ = 0, or pk =

1 + iνk κk , r − iνk

(11.33)

for all k ∈ {1, . . ., K}. Notice the similarities between (11.31), (11.32), and (11.33)—where the real asset is not “fully liquid’’—and (11.12), (11.13) and (11.15), respectively, where it is. From (11.33) it is obvious that r > iνk for the asset price to be nonnegative. In contrast to the previous section,

Liquidity, Monetary Policy, and Asset Prices

303

where the real asset is assumed to be fully liquid, it is now possible to have a monetary equilibrium with a strictly positive inﬂation rate. From (11.32) and (11.33) ﬁat money is valued if K

νk A k

k=1

1+r κk < (i). r − iνk

(11.34)

For money to be valued, the total liquid stock of real assets, the left side of (11.34), must be less than the quantity of real balances that a buyer would accumulate in a pure monetary economy, (i). Otherwise, the buyer would have no incentive to complement his portfolio of real assets with real balances given the cost of holding money. Now let’s examine the effect that monetary policy has on asset prices. From (11.33), ∂ ln pk νk (1 + r) . = ∂i (1 + iνk ) (r − iνk ) The price of real asset k increases with inﬂation, provided that νk > 0, as buyers try to substitute the real asset for real balances when inﬂation is higher, and money is more costly to hold. If νk = 0, then the asset is completely illiquid—in the sense that it cannot be used as means of payment in the DM—and monetary policy has no affect on its price. In this situation it should be obvious that the asset will be priced at its fundamental value, κk /r. Note that ∂ ln pk /∂i is increasing with νk , which means that inﬂation has a bigger effect on the price of assets that are more liquid. The gross rate of return of asset k ∈ {1, . . . , K} is Rk =

κk + pk 1+r = . pk 1 + iνk

(11.35)

If the nominal interest rate, i, is strictly positive, then the model predicts a nondegenerate distribution of rates of return, where the ordering depends on the liquidity coefﬁcients {νk }. In any monetary equilibrium, RK ≥ RK−1 ≥ . . . ≥ R1 ≥ γ −1 , where Rk 1 + iνk = >0 Rk 1 + iνk

for νk > νk .

(11.36)

It is both interesting and important to point out that these rate-of-return differentials emerge in an environment where agents are essentially risk

304

Chapter 11

neutral in that they have linear preferences over the CM good. The nondegenerate structure of asset yields arises because the different assets are used in different degrees as means of payments. We now examine the effect that monetary policy has on the structure of asset yields. From (11.35) we have ∂ ln Rk νk =− . ∂i 1 + iνk Provided that νk > 0, as inﬂation increases, the rates of return of the real assets decrease, and real asset prices are bid up. As a consequence in any monetary equilibrium the structure of asset yields {Rki }K k=1 associated > i in a ﬁrstwith a cost of money i dominates {Rk }K associated with i k=1 order stochastic sense. Moreover |∂ ln Rk /∂i| increases with νk , so that the effect of inﬂation on an asset’s rate of return is larger if the asset is more liquid. Note that ln Rk − ln Rk ≈ i (νk − νk ) , which means that the rate of return differences across assets reﬂect the liquidity differences in the assets, and inﬂation acts as a scaling factor that ampliﬁes these liquidity differences. Finally, consider two assets k and k such that νk > νk . If i > 0, then Rk − Rk > 0. Using (11.36), we have ∂ ln (Rk − Rk ) 1 − i 2 νk νk . = ∂i i (1 + iνk ) (1 + iνk ) Hence ∂ ln (Rk − Rk ) /∂i > 0 if and only if 1 − i2 νk νk > 0. That is, the premia paid to the less liquid asset, Rk − Rk , increases with inﬂation, provided that i is not too large. In the case where νk = 0, i.e., the least liquid asset is illiquid, then ∂ ln (Rk − Rk ) /∂i > 0 always holds. So far we have taken the liquidity coefﬁcients {νk } as exogenous. Although it has been useful to describe how liquidity differences across assets can generate differences in asset returns and different responses to changes in monetary policy, taking {νk } as exogenous is not satisfactory. One would like to understand what frictions in the economy would generate such restrictions on the use of assets as means of payment, and how these frictions might interact with monetary policy. To conclude this section, we provide a number of explanations that may underlie

Liquidity, Monetary Policy, and Asset Prices

305

the liquidity coefﬁcients {νk }. (The reader can also go back to chapters 10.3.2 and 10.5 for some explanations for the rate-of-return dominance puzzle as a result of liquidity differences across assets.) First, the partial illiquidity of some assets may be the result of moral hazard frictions. Suppose, for example, that the transfer of goods from a seller to a buyer in the DM is immediate, but the transfer of assets from a buyer to a seller, although initiated immediately, is ﬁnalized with a slight lag. If the buyer has the ability to divert a fraction 1−νk of his asset holdings after the transfer is initiated, but before it is ﬁnalized, then he won’t be able to credibly promise more than νk of his asset holdings. Similarly agents might be able to produce fraudulent assets or claims on assets, such as counterfeits, at different costs. As we demonstrated in chapters 5.3 and 10.4, the possibility of counterfeiting generates an upper bound on the real value of the asset that will be transferred to the seller in a bilateral match. Second, the illiquidity of some assets can reﬂect an adverse selection problem. Suppose that asset k is risky in the sense that in some states the dividend is higher than in other states. If the holder of the asset has some private information about the future dividend of the asset in the DM, then a buyer holding a high-dividend asset may signal its quality by retaining a fraction of his asset holdings. This will make the asset partially illiquid, i.e., νk < 1, in the high dividend states. We used a related explanation in chapter 7.4 to explain how stochastic inﬂation affects output. Third, the differences in assets’ liquidity may be the result of the pricing mechanism in the DM. Indeed, as shown in chapter 10.5, one can construct a pricing mechanism that generates the same payoff for the buyer as the one in problem (11.26)–(11.28), but the constructed pricing mechanism is pairwise Pareto efﬁcient, and does not restrict the transfer of assets in a bilateral match, as does problem (11.26)–(11.28). For this kind of pricing mechanism, one could interpret the differences in liquidity among assets as coming from a convention that allows some assets to be traded at better terms of trades for the buyer than others. 11.5 Endogenous Recognizability, Information, and Liquidity In this section we endogenize the recognizability and, hence, the liquidity of an asset. We adopt the economic environment of chapter 11.2 where ﬁat money and a single risk-free real asset coexist. We assume that the real asset is not portable but that agents can trade claims on it. In

306

Chapter 11

the DM agents have the technology to counterfeit those claims instantly and at zero cost. In contrast, ﬁat money cannot be counterfeited. If the seller is unable to distinguish genuine claims from counterfeits, then claims on the real asset will not be traded since sellers understand that it is a dominant strategy for buyers to try to pass counterfeits once an offer has been accepted. (See chapter 10.4 for a more formal argument.) In contrast to chapters 5.3 and 10.4, sellers can choose to be informed or not. At the beginning of each period, a seller can invest in a costly technology that allows him to recognize genuine claims from counterfeited ones. The cost of this technology is ψ > 0, measured in terms of utility. We denote ν ∈ [0, 1] as the fraction of informed sellers, and it is common knowledge in the match whether the seller invested in the technology. The parameter ν, which is related to the parameter νk of the previous section, will also indicate the probability that a claim on the real asset is accepted in payment by a random seller in the DM. 11.5.1 Equilibrium If buyers make take-it-or-leave-it offers to sellers in the DM, then sellers have no incentive to invest in the costly technology that allows them to recognize counterfeits since they do not receive any surplus from their DM trades. We will therefore adopt the proportional bargaining solution; see chapter 4.2.3. In this case sellers receive the share 1 − θ > 0 of the total match surplus. Consider a buyer in the DM holding z units of real balances and a units of the real asset. Denote the maximum wealth that the buyer can transfer to the seller in a match. If the seller is informed, then = z+(p+κ)a; if he is not, then = z since uninformed sellers will not accept claims on the real asset. Under the proportional bargaining scheme the quantity traded in informed matches, q, solves ω(q) = min ω(q∗ ), z + (p + κ)a , (11.37) where ω(q) ≡ θ c(q) + (1 − θ )u(q) is the transfer of wealth from the buyer to the seller. In uninformed matches the quantity traded, qu , solves ω(qu ) = min ω(q∗ ), z .

(11.38)

The right sides of (11.37) and (11.38) differ because the liquid wealth of a buyer in an informed match is composed of money and the real asset, while it is only money in an uninformed match.

Liquidity, Monetary Policy, and Asset Prices

307

The value of being a buyer in the DM with portfolio (a, z) is given by V b (a, z) = σ νθ u(q) − c(q) + σ (1 − ν)θ u(qu ) − c(qu ) + z + a(p + κ) + W b (0, 0),

(11.39)

where, to simplify matters, we use the linearity of the value function b W and the fact that u q − ω q = θ u q − c q . According to (11.39) the buyer meets a seller with probability σ . The seller is informed with probability ν. In an informed match, the seller produces q, and in an uninformed match, he produces qu . The buyer receives a fraction θ of the total trade surplus in all trade matches. If we substitute V b from (11.39) into (11.9), the buyer’s value function at the beginning of the CM, then the buyer’s portfolio problem is given by max

a≥0,z≥0

−iz − ar p − p∗ + σ νθ u(q) − c(q) + σ (1 − ν)θ u(qu ) − c(qu ) , (11.40)

which is a straightforward generalization of (11.11). In the appendix we show that problem (11.40) is concave. The ﬁrst-order (necessary and sufﬁcient) conditions are i u (q) − c (q) u (qu ) − c (qu ) − +ν + (1 − ν) ≤ 0, σθ θ c (q) + (1 − θ )u (q) θ c (qu ) + (1 − θ)u (qu ) (11.41) ∗ r p−p u (q) − c (q) − +ν ≤ 0, (11.42) σ θ(p + κ) θ c (q) + (1 − θ )u (q) where we have used that dq 1 1 dq 1 , = = = dz p + κ da ω q θ c q + (1 − θ ) u q dqu 1 1 , = u = u dz ω q θ c q + (1 − θ ) u qu and dqu = 0. da

308

Chapter 11

Condition (11.41) is satisﬁed with an equality if z > 0, as is condition (11.42) if a > 0. An important difference between (11.41) and (11.42) is that a buyer can spend his marginal unit of real balances in both informed and uninformed matches, but a claim on the marginal unit of the real asset can only be transferred in informed matches. We focus on symmetric equilibria where all buyers make the same portfolio choice. We now turn to the seller’s problem. Without loss of generality, we assume that sellers do not hold assets, since they have no strict incentive to do so. At the beginning of each period, a seller must choose whether or not to invest in the technology to recognize claims on the real asset. The seller’s makes this choice by comparing his lifetime expected utility if he does invest in the technology with that if he does not invest. So the seller’s problem is max −ψ + σ (1 − θ ) u(q) − c(q) , σ (1 − θ ) u(qu ) − c(qu ) .

(11.43)

Note that we omit the continuation value of the seller in the CM, W s (0, 0), from both expressions in the above maximization problem. According to (11.43), if the seller chooses to be informed, then he incurs the disutility cost ψ, which allows him to accept claims on the real asset. In this case the quantity traded is q, and the seller extracts a fraction 1 − θ of the match surplus. If the seller chooses to be uninformed, then he only accepts money, and the quantity traded is qu . From (11.43) the measure of informed sellers will satisfy ⎧ ⎧ ⎫ ⎨ =1 ⎨ > ⎬ ν ∈ [0, 1] if − ψ + σ (1 − θ ) u(q) − c(q) = ⎩ ⎩ ⎭ =0 < σ (1 − θ ) u(qu ) − c(qu ) . (11.44) A stationary symmetric equilibrium is a list (q, qu , z, p, ν) that satisﬁes conditions (11.37) with a = A, (11.38), (11.41), (11.42), and (11.44). 11.5.2 Equilibria with Recognizable Assets Consider, ﬁrst, equilibria where all sellers get informed, i.e., ν = 1. Except for the pricing mechanism in the DM, these equilibria are essentially the same as those in chapter 11.2, where ﬁat money and claims on the real asset are equally liquid. From (11.41) the output traded in the DM in a monetary equilibrium, q1 , is solution to

Liquidity, Monetary Policy, and Asset Prices

i u (q1 ) − c (q1 ) = , σθ θ c (q1 ) + (1 − θ )u (q1 )

309

(11.45)

which is identical to equation (4.43) in chapter 4.2.3. The subscript “1’’ refers to an equilibrium with ν = 1. When ν = 1, the price of the real asset is p1 =

(1 + i)κ , r−i

since the cost of investing in one unit of the asset, r p1 − p∗ , must equal the cost of obtaining an equivalent payoff by holding money, i p1 + κ . In a monetary equilibrium, the buyer’s real balances are z1 = ω(q1 ) − (p1 + κ)A > 0.

(11.46)

The right side of (11.46) is decreasing in i. Note that if i = 0, then z1 = ω(q∗ ) − (p∗ + κ)A, and as i approaches r, z1 approaches minus inﬁnity. Consequently, if (p∗ +κ)A < θ c(q∗ )+(1−θ )u(q∗ ), then there is a ı¯ ∈ (0, r), such that for all i < ı¯ there is an equilibrium with informed sellers and valued ﬁat money. If i > ı¯, then the equilibrium will be a nonmonetary one, and the asset price will be given by the solution to (11.42) at equality, with ω(q) = (p + κ)A. If (p∗ + κ)A ≥ θ c(q∗ ) + (1 − θ )u(q∗ ), then ﬁat money is not valued and q1 = q∗ in all matches. In this equilibrium the stock of real asset is sufﬁciently large to satiate the economy’s need for a medium of exchange. It should be emphasized that even if q1 = q∗ , the equilibrium is not socially efﬁcient since sellers incur a real cost associated with being informed. We now need to verify that it is optimal for sellers to get informed. From (11.44), ν = 1 requires that ψ ≤ ψ1 ≡ σ (1 − θ ) u(q1 ) − c(q1 ) − u(qu1 ) − c(qu1 ) ,

(11.47)

where qu1 represents output in the DM if a seller chooses not to get informed when all other sellers are informed, and is given by the solution to ω(qu1 ) = z1 if the equilibrium when all sellers are informed is monetary, and qu1 = 0 if it is not. Hence there exists an equilibrium where all sellers are informed, provided that the cost to be informed is sufﬁciently low, i.e., lower than ψ1 > 0. 11.5.3 Equilibria with Unrecognizable Assets Now let’s consider equilibria where all sellers are uninformed, i.e., ν = 0. In this case genuine and counterfeit claims on the real asset cannot be

310

Chapter 11

distinguished by sellers in the DM, and hence they will not be accepted as means of payment. The only medium of exchange is ﬁat money, i.e., the model generates an endogenous cash-in-advance constraint. The equilibrium outcome will be similar to the pure monetary economy described in chapter 4.2.3. From (11.41) the output traded in the DM is qu0 solution to u (qu ) − c (qu0 ) i = u 0 , σθ θ c (q0 ) + (1 − θ )u (qu0 )

(11.48)

and from (11.42) with ν = 0, the price of the real asset is p0 = p∗ . The subscript “0’’ refers to an equilibrium with ν = 0. The asset is priced at its fundamental value since it cannot be used as medium of exchange owing to its lack of recognizability. The buyer’s real balances are z0 = ω(qu0 ). Condition (11.44) implies that it is optimal for sellers to remain uninformed with regard to claims on the real asset if ψ ≥ ψ0 ≡ σ (1 − θ ) u(q0 ) − c(q0 ) − u(qu0 ) − c(qu0 ) ,

(11.49)

where q0 represents output in the DM if a seller chooses to get informed when all other sellers are not informed and is given implicitly by (11.37), with p = p0 = p∗ , a = A and z = z0 . From (11.48), if i tends to 0, then qu0 approaches q∗ , and z0 approaches θc(q∗ ) + (1 − θ)u(q∗ ). Consequently q0 = q∗ and ψ0 = 0. Hence, if the monetary authority implements the Friedman rule, then there exists an equilibrium where agents trade the ﬁrst-best level of output in all matches, and ﬁat money is the only means of payment. Note that this equilibrium is socially efﬁcient because sellers do not need to invest in a costly recognition technology; ﬁat money, in conjunction with the Friedman rule, allows society to save on information costs. 11.5.4 Multiple Monetary Equilibria If ψ0 < ψ1 , then there will exist multiple equilibria—an equilibrium where sellers get informed and one where they do not—for any ψ ∈ [ψ0 , ψ1 ] since conditions (11.47) and (11.49) can be simultaneously satisﬁed. We now demonstrate that ψ0 < ψ1 . First, notice that the asset price is higher in an equilibrium where sellers are informed compared to one where they are uninformed, i.e.,

Liquidity, Monetary Policy, and Asset Prices

311

p1 ≥ p0 = p∗ . This is because the price of the real asset can rise above its fundamental value only if it is recognizable and is used as medium of exchange. Hence (p0 + κ)A ≤ (p1 + κ)A. Moreover, if we assume that the conditions for a monetary equilibrium are satisﬁed when ν = 1, then from (11.45) and (11.48), the quantities traded in the DM in a monetary equilibrium with informed sellers and in a monetary equilibrium with uninformed sellers are the same, qu0 = q1 . This implies that z1 + (p1 + κ)A =z0 from (11.37) and (11.38). In addition the surplus S( ) ≡ u q( ) − c q( ) as a function of the buyer’s liquid wealth, , is concave, and strictly concave if < θc(q∗ ) + (1 − θ )u(q∗ ). Therefore σ (1 − θ )S (z0 )(p1 + κ)A < σ (1 − θ ) u(q1 ) − c(q1 ) − u(qu1 ) − c(qu1 ) ≡ ψ1

(11.50)

and ψ0 ≡ σ (1 − θ ) u(q0 ) − c(q0 ) − u(qu0 ) − c(qu0 ) < σ (1 − θ )S (z0 )(p∗ + κ)A.

(11.51)

See ﬁgure 11.6. Since (p1 + κ)A ≥ (p∗ + κ)A, conditions (11.50) and (11.51) imply that ψ0 < ψ1 . Consequently, if a monetary equilibrium exists with ν = 1, and ψ ∈ [ψ0 , ψ1 ], then there is also a monetary equilibrium with ν = 0 and p = p∗ . There are two other interesting cases to consider. The ﬁrst case assumes that i is close to 0 and (p∗ + κ)A < ω(q∗ ). In an equilibrium with uninformed sellers, z0 approaches θ c(q∗ ) + (1 − θ)u(q∗ ), and the asset price is its fundamental p0 = p∗ . Consequently ∗ −1 value, u −1 ∗ ∗ q0 = ω (z0 ) = q , q0 = min q , ω (z0 + (p + κ)A) = q∗ , and ψ0 = 0. In a monetary equilibrium with informed sellers, z1 = ω(q∗ ) − ∗ u ∗ ∗ −1 ∗ (p + κ)A q1 = q while and uq1 = ωu (p + κ)A < q . Hence ψ1 = ∗ ∗ σ (1 − θ ) u(q ) − c(q ) − u(q1 ) − c(q1 ) > 0 = ψ0 . If the cost of acquiring information is sufﬁciently small, then there are multiple equilibria. In this case an equilibrium where sellers are uninformed, and therefore money is the only means of payment, dominates from a social welfare view point an equilibrium where sellers are informed, since information acquisition is costly. The second case illustrates the existence of multiple equilibria when (p∗ + κ)A ≥ ω(q∗ ). The equilibrium with informed sellers is such that q1 = q∗ and money is not valued, so that qu1 = 0. If i is sufﬁciently small to allow for the existence of a monetary equilibrium in the case where

312

Chapter 11

σ (1 −θ ) u(q( )) − c(q( ))

ψ0

ψ1

( p1 + κ ) A

z1

( p *+ κ ) A z1 + ( p1 + κ ) A = z0

z0 + ( p* + κ ) A

Figure 11.6 Information acquisition and multiple equilibria

sellers are uninformed, then it is immediate that ψ1 = σ (1 − θ ) u(q∗ ) − c(q∗ ) > ψ0 = σ (1 − θ ) u(q0 ) − c(q0 ) − u(qu0 ) − c(qu0 ) , since q0 = min q∗ , ω−1 z0 + (p∗ + κ)A = q∗ and qu0 > 0. The intuition that underlies the multiplicity of equilibria is as follows. Suppose that sellers believe that buyers hold few real balances. Then sellers have incentives to be informed because, otherwise, they will only be able to accept the reduced real balances of the buyers. But, if buyers believe that all sellers are informed, it is optimal for them to reduce their real balances and bid up for the real asset until the rates of return of money and the real asset are equalized. If, on the contrary, sellers believe that buyers hold a large amount of real balances, then they do not need to acquire costly information to recognize claims on the real asset because the use of those assets would not increase the match surplus by

Liquidity, Monetary Policy, and Asset Prices

313

more than the cost of information. And if all sellers are uninformed, it becomes optimal for buyers to accumulate large amount of real balances provided that inﬂation is not too high. This multiplicity of equilibria captures the strategic complementarities that make the liquidity of an asset a self-fulﬁlling phenomenon and is similar to the viability of credit being a self-fulﬁlling phenomenon in chapter 8.3. 11.6 Further Readings The canonical macroeconomic model of asset pricing is due to Lucas (1978) in the context of a frictionless, exchange economy. Risk-free assets in ﬁxed supply have been introduced into the Lagos–Wright model of monetary exchange by Geromichalos, Licari, and Suarez-Lledo (2007). The case where the asset is risky and agents are symmetrically informed about the dividend of the asset in a bilateral match has been studied by Lagos (2006). In his model ﬁat money is replaced by risk-free bonds. He adds an exogenous constraint on the use of the risky asset as means of payment, and he calibrates the model to explain the risk-free rate and equity premium puzzles following the methodology of Mehra and Prescott (1985). He shows that a slight restriction on the use of the risky asset is necessary to allow the model to match the risk-free rate and the size of the equity premium in the data for plausible degrees of risk aversion. Our model with multiple assets is related to the ones of Wallace (1996, 2000) and Cone (2005) who, in contrast to us, emphasize asset divisibility, or lack of divisibility, to explain the coexistence of money and interest-bearing assets, and the liquidity structure of asset yields. The analysis is similar to the one in Nosal and Rocheteau (2009). Several explanations for the liquidity differences across assets can be found in the literature. Kiyotaki and Moore (2005) assume that the transfer of ownership of capital is not instantaneous so that an agent can steal a fraction of his capital before the transfer is effective. Similarly Holmström and Tirole (1998, 2001) develop a corporate ﬁnance approach to liquidity, where a moral hazard problem prevents claims on corporate assets from being written. Freeman (1985), Lester, Postlewaite, and Wright (2008), and Kim and Lee (2008) explain the illiquidity of capital goods by the assumption that claims on capital can be costlessly counterfeited and can only be authenticated in a fraction of meetings. Following Kim (1996) and Berentsen and Rocheteau (2004), Lester, Postlewaite, and Wright (2008) endogenize this fraction

314

Chapter 11

of meetings by assuming that agents can invest in a costly technology to recognize claims on capital. Chapter 11.5 is similar to Lester, Postlewaite, and Wright except that we use proportional bargaining instead of Nash, and we do not assume that sellers are heterogeneous in terms of their costs of getting informed. The idea that ﬁat money is a substitute for information acquisition can be found in Brunner and Meltzer (1971) and King and Plosser (1986). Li and Rocheteau (2009) and Rocheteau (2009b) investigate the case where counterfeits are produced at a positive cost and show that the lack of recognizability manifests by an endogenous upper bound on the transfer of assets in uninformed matches. Asymmetries of information are used to endogenize transaction costs in ﬁnancial markets (e.g., Kyle 1985; Glosten and Milgrom 1985), security design (e.g., DeMarzo and Dufﬁe 1999), and capital structure choices (e.g., Myers and Majluf 1984). Hopenhayn and Werner (1996) develop a model with multiple indivisible assets traded in bilateral meetings under private information. Rocheteau (2009a) proposes a search-theoretic monetary model in which buyers have some private information about the future value of their risky assets. He shows that buyers in the high-dividend states retain a fraction of their asset holdings in order to signal their quality. Finally, Nosal and Rocheteau (2008) extend the trading mechanism in Wallace and Zhu (2007) and show that a search-theoretic monetary model can generate rate-of-return differences among seemingly identical assets without imposing trading restrictions and without violating Pareto efﬁciency in bilateral trades. Appendix Derivation of (11.23) From (11.18),

u (qH ) u (qL ) i = σ πH − 1 + πL −1 . c (qH ) c (qL )

(11.52)

From (11.19) at equality,

u (qH ) u (qL ) ∗ r p − p = (p + κ)σ ¯ πH − 1 + πL −1 c (qH ) c (qL )

u (qH ) u (qL ) + σ πH (κH − κ) ¯ (κ − κ) ¯ − 1 + π − 1 . L L c (qH ) c (qL ) (11.53)

Liquidity, Monetary Policy, and Asset Prices

315

From (11.52) the ﬁrst term on the right side of (11.53) is equal to i(p + κ), ¯ and hence the expression for p can be rearranged as

u (qH ) ∗ r p − p = i(p + κ) ¯ + σ πH (κH − κ) ¯ −1 c (qH ) u (qL ) +πL (κL − κ) −1 . (11.54) ¯ c (qL ) It is straightforward to demonstrate that the ρ given by (11.22) is identical to the above bracketed term that is premultiplied by σ . Replacing this bracketed term with ρ, and rearranging, we get equation (11.23). Concavity of the Problem (11.40) The buyer’s objective function is (a, z) = −iz − ar p − p∗ + σ νθ u(q) − c(q) + σ (1 − ν)θ u(qu ) − c(qu ) where q and qu are given by (11.37) and (11.38). The partial derivatives of the buyer’s objective function are

+ u (q) − c (q) z (a, z) = − i + σ νθ θ c (q) + (1 − θ )u (q) + u (qu ) − c (qu ) + σ (1 − ν)θ θ c (qu ) + (1 − θ )u (qu ) + u (q) − c (q) ∗ (p + κ), a (a, z) = − r p − p + σ νθ θ c (q) + (1 − θ )u (q) ∗ where [x]+ = max(x, 0). For all (a, z) such z + (p + κ)a ≥ θ c(q ) + that ∗ ∗ ∗ (1 − θ )u(q ), q = q , and a (a, z) = −r p − p . The objective function (a, z) is concave, but not strictly jointly concave. If i > 0, then q = q∗ if and only if p = p ∗ . In this case the of real balances choice is uniquely u u u u determined by u (q ) − c (q ) / θ c (q ) + (1 − θ)u (q ) = i/σ (1 − ν)θ and the choice of asset holdings is z ∈ [θc(q∗ ) + (1 − θ)u(q∗ ) − z, +∞). Let us turn to the case where p > p∗ . Then we can restrict our attention to portfolios such that z + (p + κ)a < θc(q∗ ) + (1 − θ)u(q∗ ), which implies that q < q∗ . The second and cross partial derivatives are then u zz (a, z) = σ νθ + σ (1 − ν)θ za (a, z) = σ νθ(p + κ) < 0, aa (a, z) = σ νθ(p + κ)

2

< 0,

< 0,

316

Chapter 11

where u (q)c (q) − u (q)c (q) = 3 , θc (q) + (1 − θ )u (q) u (qu )c (qu ) − u (qu )c (qu ) u = 3 , θc (qu ) + (1 − θ )u (qu ) The determinant of the Hessian matrix is then det H = (σ θ )2 ν(1 − ν)(p + κ)2 u > 0. Hence, for all (a, z) such that z + (p + κ)a < θ c(q∗ ) + (1 − θ)u(q∗ ), the objective function (a, z) is strictly jointly concave. Consequently, if p > p∗ , then the buyer’s problem has a unique solution.

12

Liquidity and Trading Frictions

Liquidity is the ability to trade large size quickly, at low cost, when you want to trade. It is the most important characteristic of well-functioning markets. . . . Liquidity—the ability to trade—is the object of a bilateral search in which buyers look for sellers and sellers look for buyers. The various liquidity dimensions are related to each other through the mechanics of this bilateral search. Traders must understand these relations in order to trade effectively. —Larry Harris, Trading and Exchanges: Market Microstructure for Practitioners (2003, ch. 19)

In previous chapters we deﬁned the liquidity of an asset in terms of its ability to function as a medium of exchange in goods markets that are characterized by trading frictions. In this chapter we revisit the notion of liquidity. In contrast to previous chapters, there are no trading frictions associated with the purchase and consumption of goods, and as a result the asset does not play any role as a means of payment. Instead, trading frictions are introduced directly into an asset market best described as an over-the-counter market, with bilateral matches between investors and dealers. Our simple model will be able to capture different dimensions of liquidity that have been identiﬁed in the ﬁnance literature, such as the volume of trade, bid–ask spreads, and trading delays. We consider an economy where investors accumulate capital goods to produce a general consumption good, as in chapter 10.1. But idiosyncratic productivity shocks give investors a reason to wanting to reallocate their asset holdings. In particular, investors with low productivity want to sell their capital holdings to agents with high productivity. Investors, however, do not have direct access to a centralized market where they can readjust their asset holdings instantly. Instead, they adjust their asset holdings via a network of dealers.

318

Chapter 12

An investor’s asset demand depends not only on his productivity at the time he is able to access the market but also on his expected productivity over the period of time that he does not have the opportunity to adjust his asset holdings. When asset markets are illiquid, investors put more weight on their future expected productivity, and as a result investors will adjust their asset positions in a way that reduces their need to trade. Conversely, a reduction in trading frictions makes the investor less likely to remain locked into an undesirable asset position and therefore induces him to put more weight on his current productivity when determining his asset position. As a result a reduction in trading frictions induces an investor to demand a larger asset position if his current productivity is relatively high, and a smaller position if it is relatively low. This effect on the dispersion of the distribution of asset holdings is a key channel through which trading frictions determine trade volume, bid–ask spreads, and trading delays. If it is easier to trade the asset, or if dealers have less bargaining power, investors take more extreme asset positions, which leads to a higher volume of trade. As well, bid–ask spreads tend to be lower, and trading delays shorter. We also examine how asset market frictions affect asset prices. Finally, we endogenize trading frictions by allowing free entry of dealers in the market-making sector. As the number of dealers increases, trading delays fall. We show that the presence of complementarities between investors’ asset-holding decisions and dealers’ entry decisions can lead to multiple equilibria, so that liquidity in the market can dry up because of self-fulﬁlling beliefs. 12.1 The Environment We depart from the standard environment along a number of dimensions. We now assume that time is continuous. This assumption simpliﬁes the analysis; for example, on a small time interval we can rule out the possibility of multiple events occurring. Even though we must drop the assumption that periods are divided into day and night subperiods, as this distinction is meaningless in continuous time, there still exist centralized and decentralized markets. There is one type of consumption good, the general good. There are two types of inﬁnitely lived agents, called investors and dealers, with a unit measure of each type. Both agents consume the general good,

Liquidity and Trading Frictions

319

Figure 12.1 Trading arrangement

Investors

Competitive interdealer market

Dealers

Dealers

Investors

where the utility of consuming x units of the general good is x. Agents discount future utility at rate r. The general good can be produced with two different technologies. One technology has h units of the general good being produced from h units of labor (and h units of labor generates h units of disutility). The general good can also be produced by a technology that uses capital as an input, and depends on the investor’s productivity. This technology is described by fi (k), where k ∈ R+ represents capital invested, i ∈ {1, . . . , I} indexes the productivity of the investor who operates the capital, and fi (k) is twice continuously differentiable, strictly increasing, and strictly concave. Capital is a durable, perfectly divisible asset that is in ﬁxed supply, K ∈ R+ . With instantaneous probability equal to δ, each investor receives a productivity shock. This means that productivity shocks occur according to a Poisson process with arrival rate δ, i.e., the inter-arrival time between two shocks is exponentially distributed with mean 1/δ. Conditional on receiving this shock, the investor draws productivity type j ∈ {1, . . . , I} with probability πj > 0, where Ii=1 πi = 1. These δ shocks capture the idea that investors’ productivities vary over time, which result in investors wanting to rebalance their asset positions. Dealers do not have access to the capital technology to produce the general good, and do not hold positions in capital. Dealers can, however, trade capital assets continuously in a competitive market. Investors do not have direct access to the competitive asset market, but they do have periodic contact with dealers who can trade in this market on their behalf. The arrival rate with a dealer for the investor is σ > 0. The bilateral matching process between investors and dealers plays the part of the decentralized market in earlier chapters. The trading process for the capital asset is depicted in ﬁgure 12.1. Once a dealer and an investor have contacted one another, they negotiate over the quantity of assets that the

320

Chapter 12

dealer will acquire in competitive markets on behalf of the investor, and the intermediation fee that the dealer charges for his services. 12.2 Equilibrium Let Vi (k) denote the maximum expected discounted utility attainable by an investor who is of productivity type i and is holding k units of the asset. The ﬂow Bellman equation that determines Vi (k) is rVi (k) = fi (k) + σ Vi (ki ) − Vi (k) − p(ki − k) − φi (k) +δ

I

πj Vj (k) − Vi (k) .

(12.1)

j=1

The ﬂow Bellman equation can be interpreted as an asset-pricing equation, where the asset to be priced is an investor in state (i, k). The left side is the opportunity cost from holding this asset, while the right side is the dividends and capital gains or losses from holding the asset. According to (12.1) an investor with productivity type i and asset holdings k produces fi (k) of the general good, which can be interpreted as a ﬂow dividend. With instantaneous probability σ , the investor contacts a dealer, and readjusts his asset holdings from k to ki . This readjustment raises his lifetime expected utility by Vi (ki ) − Vi (k), which can be interpreted as a capital gain, net of the fee, φi (k), he pays to the dealer and the value of the assets he purchases, p(ki − k). We will show that the intermediation fee, φi , depends on the capital stock held by the investor, but the desired capital stock, ki , does not. With instantaneous probability δ he receives a productivity shock: conditional on receiving this shock, his productivity type becomes j ∈ {1, . . . , I} with probability πj > 0. The maximum expected discounted utility attained by a dealer is denoted by V d and solves d rV = σ φi (k)dH(k, i), (12.2) where H represents the distribution of investors across asset holdings and preference types states. With instantaneous probability σ , the dealer meets an investor who is drawn at random from the population of investors. The dealer trades in the competitive asset market on behalf of the investor and receives an intermediation fee, φi (k), for his services. The size of the intermediation fee depends on the productivity type of the investor, i, and his asset holdings at the time he contacts the dealer, k.

Liquidity and Trading Frictions

321

We now examine the determination of the terms of trade in a bilateral meeting between a dealer and an investor. Suppose that the investor’s productivity type is i, and he holds k units of capital. The terms of trade will specify a new asset position for the investor, k , and the intermediation fee, φ, paid to the dealer. If agreement (k , φ) is reached, then the payoff to the investor is Vi (k ) − p(k − k) − φ.

(12.3)

The investor enjoys the expected lifetime utility associated with his new stock of capital, Vi (k ), minus the cost of his investment in capital, p(k − k), and the intermediation fee paid to the dealer, φ. The payoff of the dealer is simply V d + φ.

(12.4)

If no agreement is reached, the payoff of the investor is Vi (k), and the payoff of the dealer is V d . We assume that the agreement (k , φ) is given by the solution to a generalized Nash bargaining problem, where the dealer’s bargaining power is θ ∈ [0, 1]. This agreement is given by ki , φi (k) = arg max[Vi (k ) − Vi (k) − p(k − k) − φ]1−θ φ θ . (k ,φ)

(12.5)

The solution to (12.5) is ki = arg max[Vi (k ) − pk ], k

φi (k) = θ [Vi (ki ) − Vi (k) − p(ki − k)].

(12.6) (12.7)

According to (12.6) the choice of capital is the one that an investor would make if he could trade directly in the competitive asset market at the price p: it maximizes the value of the investor, net of the cost of acquiring the capital. According to (12.7) the intermediation fee is chosen so that the dealer gets a fraction θ of total match surplus. If we substitute φi (k), given by its expression in (12.7), into (12.1), we get rVi (k) = fi (k) + σ (1 − θ ) Vi (ki ) − Vi (k) − p(ki − k) +δ

I j=1

πj Vj (k) − Vi (k) .

(12.8)

322

Chapter 12

The investor’s ﬂow payoff, given by (12.8), is equivalent to what he would receive in an economy where he is able to extract the entire surplus from his match with a dealer, but meets a dealer with an instantaneous probability equal to only σ (1 − θ ). Thus, from the investor’s point of view, the stochastic trading process and the bargaining solution are payoff-equivalent to an alternative trading arrangement, in which he has all the bargaining power in bilateral negotiations with dealers, but only gets to meet dealers with instantaneous probability σ (1 − θ). We now proceed to provide a closed-form solution for the investor’s value function. We can rearrange (12.8) to read as [r + δ + σ (1 − θ )] Vi (k) = fi (k) + σ (1 − θ )pk + δ

I

πj Vj (k) + !i ,

(12.9)

j=1

where !i ≡ σ (1 − θ ) maxk Vi (k ) − pk . If we multiply both sides of (12.9) by πi , sum across i’s, and rearrange, we get I

πi Vi (k) =

i=1

I

¯

i=1 πi fi (k) + σ (1 − θ )pk + !

r + σ (1 − θ )

,

(12.10)

¯ ≡ Ii=1 πi !i . By substituting (12.10) into (12.9), we are able to where ! get a closed-form solution for the value function of investors, Vi (k) =

f¯i (k) + σ (1 − θ )pk + i , r + σ (1 − θ )

(12.11)

where i ≡

¯ !i δ! + r + δ + σ (1 − θ ) [r + δ + σ (1 − θ )] [r + σ (1 − θ)]

and f¯i (k) =

(r + σ (1 − θ )) fi (k) + δ

r + σ (1 − θ ) + δ

j πj fj (k)

.

(12.12)

From (12.12) we see that f¯i (k) is a weighted average of the productivities in the different states. The weights on the current productivity, fi (k), and future ones, fj (k), are functions of the transition rates σ and δ, the discount rate r, and the dealer’s bargaining power, θ . As the trading frictions vanish, i.e., as σ goes to inﬁnity, f¯i (k) approaches the current productivity, fi (k). It can be shown, from (12.11), that Vi (k) is continuous

Liquidity and Trading Frictions

323

and strictly concave in k. From (12.6) and (12.11), the optimal choice of capital is given by ki = arg max[f¯i (ki ) − rpki ]. ki ≥0

(12.13)

From the strict concavity of f¯i (k), ki is uniquely determined. Moreover from (12.7) and (12.11) the expression for the intermediation fee is φi (k) =

θ f¯i (ki ) − f¯i (k) − rp (ki − k) . r + σ (1 − θ )

(12.14)

The intermediation fee depends on the dealer’s bargaining power, θ , the discount factor, r, and the transition rates, σ and δ. It also varies with the change in the investor’s asset position. Intuitively the intermediation fee is proportional to the gain that the investor enjoys from readjusting his asset holdings. We now characterize the steady-state distribution of investors’ types, H(k, j). The individual state of an investor is the pair (k, j) ∈ R+ × {0, . . . , I}, where k is his current asset holdings and j his productivity type. Note that any state (k, j) such that k ∈ / {ki }Ii=1 is transient since, whenever an investor adjusts his asset holdings in a steady state, he chooses k ∈ {ki }Ii=1 . Thus the set of ergodic states is {ki }Ii=1 × {1, . . . , I}. This allows us to simplify the exposition by denoting state (ki , j) by ij ∈ {1, . . . , I}2 . Hence, for state ij, i represents the quantity of capital the investor currently has, i.e., the one corresponding to the productivity shock he had at the time he last rebalanced his asset holdings, and j represents his current productivity shock. The measure of investors in state ij is denoted nij . In a steady state the ﬂow of investors entering state ij must equal the ﬂow of investors leaving state ij: δπj nik − δ(1 − πj )nij − σ nij = 0 if j = i, (12.15) σ

k=i

k=j

nki + δπi

nik − δ(1 − πi )nii = 0.

(12.16)

k=i

According to (12.15) the measure of investors in state ij, j = i, increases whenever an investor in some state ik, k = j, i, receives a productivity shock j, which occurs with instantaneous probability δπj . The measure of investors decreases whenever an investor in state ij receives a new productivity shock different from j, which occurs with instantaneous

324

Chapter 12

δπ 3 δπ 2

n 11

δπ 1

n 12

δπ 3

n13

δπ 2

δπ 1 δπ 3 δπ 2

n 21

δπ 1

n22

δπ 3

n 23

δπ 2

δπ 1 δπ 3 δπ 2

n31

δπ 1

n32

δπ 3 δπ 2

n 33

δπ 1 Figure 12.2 Flows across states

probability δ(1 − πj ), or whenever such an investor is able to readjust his asset holdings, which occurs with instantaneous probability σ . Equation (12.16) has a similar interpretation for agents in state ii. The ﬂows between states is depicted in ﬁgure 12.2 for I = 3. Each circle represents a state. The horizontal arrows represent ﬂows due to productivity shocks, whereas the vertical arrows indicate ﬂows due to asset holdings readjustments. The individual states shaded in gray, which lie on the diagonal, are those states for which there is no mismatch between the investor’s current productivity type and his capital holdings. It can be shown that the steady-state distribution (nij )Ii,j=1 satisﬁes nij =

δπi πj σ +δ

nii =

δπi2 + σ πi . σ +δ

for j = i,

(12.17) (12.18)

Liquidity and Trading Frictions

325

The marginal distributions, deﬁned by ni· = j nij and n·j = i nij , have the property that ni· = n·i = πi . So the measure of investors with productivity type i is equal to πi , the probability of drawing productivity shock i, conditional on getting a productivity shock. Note that the distribution of probabilities across states is symmetric, i.e., nij = nji . Note also that ∂nij /∂σ < 0 if i = j and ∂nii /∂σ > 0, which means that the measure of investors who are matched to their desired capital increases as the rate at which investors get to meet dealers increases. The only remaining equilibrium variable to be determined is the price of capital in the competitive market, p. This price equates the demand and supply of assets, i.e., i,j nij ki = K. Using that j nij = πi , this market-clearing condition can be expressed as πi ki = K. (12.19) i

The intuition behind equation (12.19) is the following: In a steady state the measure of investors that have productivity type i is πi . Each investor of type i demands ki independent of his stock of capital at the time he meets a dealer. Hence the aggregate demand of capital, in ﬂow terms, is σ i πi ki . By the Law of Large Numbers, the ﬂow supply of capital is the average capital stock held by the investors who contact dealers, σ K. There exists a unique steady-state equilibrium. The distribution of investors across asset holdings and productivity types is given by (12.17) and (12.18). The individual choices of asset holdings, the ki ’s in (12.13), depend negatively on p, the equilibrium price in the interdealer market. Assuming an interior solution, we have ki = f¯i−1 rp . Given these individual demands, the market-clearing condition (12.19) determines a unique price, the solution to

πi f¯i−1 rp = K.

i

To illustrate how a reduction in trading frictions affects the equilibrium, consider the limiting case where search frictions vanish, i.e., where σ → ∞. Investors can trade in the asset market continuously. In the limit, from (12.12) and (12.13), we get

326

Chapter 12

fi (ki ) =p r

(12.20)

for i = 1, . . . , I. From (12.14) we see that φi (k) → 0 for all k and i when σ → ∞. Combining (12.19) and (12.20), we see that the price of the asset converges to the solution to i πi fi−1 (rp) = K. The limiting distribution of investors across asset holdings and productivity types is nii = πi for each i, and nij = 0 for j = i. In this case investors with productivity i choose ki continuously by equating the marginal return from the asset, fi (ki ), to its ﬂow price, rp. When search frictions vanish, the equilibrium fee, asset price, and distribution of asset positions are the ones that would prevail in a Walrasian economy. 12.3 Trading Frictions and Asset Prices In chapter 11 we looked at asset prices in economies with trading frictions. We established that the price of an asset can depart from its “fundamental’’ value if the asset has a role in facilitating trade in the DM, that is, the asset can be used as a medium of exchange. In this case the asset price will decrease if trading frictions in the DM increase. The approach we take in this chapter is different. The asset is not used to facilitate trade, but trading frictions plague the asset market itself. In this section we will revisit the effects that trading frictions have on asset prices. We assume the following speciﬁcation for the technology that investors possess: fi (k) = Ai k α ,

0 < α < 1,

¯= and A1 < A2 < . . . < AI . Let A j πj Aj denote the average productivity. From (12.13) the demand for capital by an investor with productivity type j is kj =

¯ α (r + σ (1 − θ )) Aj + δ A rp r + σ (1 − θ ) + δ

1/(1−α) .

(12.21)

It is easy to see from (12.21) that, for a given price of capital, p, kj increases ¯ That is, investors with a productivity shock with σ as long as Aj > A. above average increase their demand for capital when σ increases.

Liquidity and Trading Frictions

327

¯ have a current marginal productivity that is higher Agents with Aj > A than what they expect it to be in the future. Because of search fric 1/(1−α) , tions, their choice of capital, kj , will be lower than kj∞ = αAj /rp which is what they would choose in a world with no trading frictions. If ¯ then the investor’s choice of capital investors’ productivity is equal to A, 1/(1−α) ¯ the investor anticipates ¯ . Since Aj is higher than A, is k¯ = α A/rp ¯ in the future; when that his productivity is likely to revert toward A this happens, he may be unable to rebalance his asset holdings for some time. As a result the investor’s optimal choice of capital holdings is a weighted average of the optimal holdings for productivities Aj and ¯ i.e., A, kj =

1−α r + σ (1 − θ ) ∞ 1−α δ kj k¯ + r + σ (1 − θ ) + δ r + σ (1 − θ ) + δ

1/(1−α) .

An increase in σ means that it will be easier for the investor to ﬁnd a dealer in the future, and this makes him put more weight on his current marginal productivity from holding the asset relative to its expected value. Hence, as σ increases so does kj . Conversely, investors with a ¯ reduce their demand for productivity shock below the average, Aj < A, capital when σ increases. From all of this, we can conclude that, for given p, as σ increases, the dispersion of asset holdings will also increase. Figure 12.3 illustrates the effect that a reduction on trading frictions has on the distribution of asset holdings. The black bars represent the distribution of asset holdings when the frequency of meetings with dealers is σ , and the gray bars represent the distribution when σ > σ . From the market-clearing condition (12.19) the asset price is given by ⎡ ⎤1−α I ¯ 1/(1−α) + δ A + σ (1 − θ )) A (r α j ⎦ πj . p = K −(1−α) ⎣ r r + σ (1 − θ ) + δ

(12.22)

j=1

Note that the expected value of the terms in round brackets, I j=1

πj

¯ α (r + σ (1 − θ )) Aj + δ A , r r + σ (1 − θ ) + δ

¯ is constant and equal to α A/r, and that the function x1/(1−α) is convex in x when 0 < α < 1. If σ increases, the dispersion of the

328

Chapter 12

πi

π πj

πI π1

k1

kj

K

k

kI

k

Figure 12.3 Trading frictions and distribution of asset holdings

¯ / [r + σ (1 − θ ) + δ] terms in (12.22) increases but (r + σ (1 − θ )) Aj + δ A their mean remains constant. From the convexity of the function x1/(1−α) the asset price will increase. Therefore, when fi (k) = Ai k α , 0 < α < 1, our model predicts a negative relationship between asset prices and trading frictions, just as in chapter 11. The reasoning behind these negative relationships is, however, different. In chapter 11 asset prices decrease as trading frictions become more severe because the asset is used less frequently as a means of payment. In this section trading frictions generate a mismatch between investors’ productivities and their capital holdings, so when frictions increase, asset prices will fall because mismatches increase. It should be emphasized that the negative relationship between trading frictions and asset prices derived above is not a general proposition; the relationship depends on the speciﬁcation of the production function, fi (k). To see this, suppose that the production function is logarithmic, fi (k) = Ai ln(1 + k). Then the demand for capital goods, assuming an interior solution, is given by kj =

¯ (r + σ (1 − θ )) Aj + δ A − 1. [r + σ (1 − θ ) + δ] rp

(12.23)

Liquidity and Trading Frictions

329

In this case the demand for the asset is linear in the productivity. As a consequence the market-clearing price is p=

¯ A . r (1 + K)

(12.24)

The asset price is now independent of the speed at which investors can access the market and dealers’ bargaining power. The price given by (12.24) is in fact the Walrasian price that would prevail in an economy without trading frictions. This suggests that the price of an asset is a poor indicator of the trading frictions that prevail in the market for the asset. The reason why σ does not affect the asset price is quite simple. As one aggregates the individual changes in demands induced by an increase in σ , the increases in kj for investors with values of Aj larger ¯ cancel out the decreases in kj for investors with values of Aj lower than A ¯ As a result σ has no effect on the aggregate demand for assets than A. and, hence, on the equilibrium price, even though the quality of the match between the stock of capital and investors is affected. We will close this section with the special case where the investors’ technologies are linear, i.e., fi (k) = Ai k. From (12.13), ¯ (r + σ (1 − θ )) Ai + δ A − rp ≤ 0, r + σ (1 − θ ) + δ with an equality if ki > 0. Market clearing implies kj = 0 for all j < I so that only investors with the highest productivity demand the asset. In this case the asset price is given by p=

¯ [r + σ (1 − θ )] AI + δ A , r [r + σ (1 − θ ) + δ]

(12.25)

The price is a weighted average of the marginal productivity of the highest investor type and the average marginal productivity in the market. The weight on the marginal productivity of the highest productivity investor—and hence the asset price—is increasing in σ , and decreasing in θ and δ. 12.4 Intermediation Fees and Bid–Ask Spreads An asset is said to be liquid if it can be readily bought or sold at a low transaction cost. We can measure this notion of liquidity by the intermediation fee that investors pay to dealers or, equivalently, by bid–ask

330

Chapter 12

spreads. In this section we study how changes in trading frictions affect intermediation fees. In the subsequent section we will examine alternative measures of liquidity, such as trading delays. We specialize the analysis to the production function, fi (k) = Ai k α for α ∈ (0, 1). From (12.14) the equilibrium fee that a dealer charges an investor who holds a capital stock ki and wishes to hold kj is ¯ [r + σ (1 − θ )] Aj + δ A θ φj (ki ) = kjα − kiα − rp kj − ki , r + σ (1 − θ ) r + σ (1 − θ ) + δ (12.26) where kj and p are given by (12.21) and (12.22), respectively. We see that an increase in σ has opposing effects on the intermediation fee. On the one hand, a higher σ implies more competition among dealers, which tends to reduce the fees they charge for any given trade size. This effect is captured by the ﬁrst term on the right side of (12.26). On the other hand, a higher σ also induces investors to conduct larger asset holding reallocations every time they trade, and this translates into larger fees for dealers, on average. To show that the intermediation fees can vary in a nonmonotonic fashion with the trading frictions, consider the case where r is small, i.e., agents are inﬁnitely patient. From (12.21), kj ≈

¯ α σ (1 − θ )Aj + δ A rp σ (1 − θ ) + δ

1/(1−α) .

If σ tends to inﬁnity, i.e., the asset market is very liquid, it is clear from (12.26) that φj (ki ) approaches 0. If σ tends to zero, i.e., the asset 1/(1−α) ¯ market is very illiquid, then kj ≈ α/rp A which is independent of j. So when it takes a very long time to readjust one’s asset position, investors choose asset holdings that reﬂect their average productivity and not their current one. As a consequence they don’t need to readjust their asset holdings as their idiosyncratic productivities change, and the intermediation fee, φj (ki ), goes to 0. Finally, when σ is neither too small nor too large, then ki = kj for all i = j so that intermediation fees are positive. This demonstrates that intermediation fees will be maximum for an intermediate level of the trading frictions. So far we have interpreted transaction costs in terms of intermediation fees, i.e., the total amount paid by the investor to the dealer to readjust his asset holdings. Alternatively, one can interpret the results in terms of

Liquidity and Trading Frictions

331

bid–ask spreads, which provides a measure of transaction costs per unit of asset traded. Consider the limiting case where α → 1, i.e., technologies are linear. In this case we showed above that kj → 0 for all j = I and ¯ / [r + σ (1 − θ ) + δ]. This implies that (12.26) rp → (r + σ (1 − θ )) AI + δ A yields φj (ki ) → 0 for all (i, j) ∈ / {I}×{1, . . . , I −1}. Obviously dealers do not obtain any fee when investors do not want to readjust their portfolios. Perhaps more surprisingly, when investors are buying the asset (i = I and j = I), dealers do not charge a fee either. The reason is that when buying capital, the investor pays his marginal product for the asset, and since the technology is linear, this means that he is indifferent between holding or not holding the asset. Finally, investors in state ij, where i = I and j = I, are holding kI units of capital but wish to hold kj → 0. From (12.26) we ﬁnd that φj (kI ) =

θ (AI − Aj ) kI , r + σ (1 − θ ) + δ

(12.27)

i.e., the fee is proportional to the quantity traded. Since the intermediation fee (12.27) is linear in the quantity traded, the previous results can be readily interpreted in terms of bid–ask spreads. The fact that an investor pays no fee when buying from the dealer is equivalent to a transaction in which the dealer charges an ask price pa equal to the price of the asset in the competitive market, i.e., pa = p. When an investor of type j < I sells his capital holdings kI through a dealer, the investor receives pkI − φj (kI ). Using (12.27), we see that this transaction is equivalent to one in which the dealer pays investors of type j a bid price pjb = p − θ (AI − Aj )/ [r + σ (1 − θ ) + δ] < p. The difference between the effective price at which the dealer sells, pa , and buys, pjb , is akin to a bid–ask spread of pa − pjb =

θ (AI − Aj ) . r + σ (1 − θ ) + δ

This spread is decreasing in the rate at which investors can rebalance their asset holdings, σ . As σ increases, it is quicker for an investor to ﬁnd a dealer, which tends to raise the investor’s disagreement point in the bargaining. This competition effect reduces the per unit fees that dealers can ask for. The bid–ask spread also decreases with δ, since the value of rebalancing one’s asset holdings is lower when productivity shocks are more frequent. The spread increases with the dealer’s bargaining power, θ , and with the difference between the marginal productivity

332

Chapter 12

of the most productive investor and that of the investor involved in the trade. Dealers buy assets at a lower effective price from investors with low marginal productivity because these investors incur a larger opportunity cost from holding on to their capital. 12.5 Trading Delays In this section we endogenize the speed at which investors can rebalance their asset holdings by extending the model to allow for free entry of dealers. The Poisson rate at which an investor contacts a dealer is σ and, since all matches are bilateral, the Poisson rate at which a dealer serves an investor is σ/υ, where υ is the measure of dealers in the market. Suppose that σ is a continuously differentiable function of υ, where σ (υ) a strictly increasing function and σ (υ)/υ a strictly decreasing function of υ. As well, assume that σ (0) = 0, σ (∞) = ∞ and σ (∞)/∞ = 0. As υ increases, investors’ orders are executed faster, but the ﬂow of orders per dealer decreases due to a congestion effect. There is a large measure of potential dealers who can choose to participate in the market. Dealers who choose to operate incur a ﬂow cost of κ > 0 that represents the ongoing costs of running the dealership, including the cost of searching for investors, advertising their services, and so on. The free entry of dealers implies that, in equilibrium, σ (υ) φj (ki )dH(ki , j) = κ, (12.28) υ i,j i.e., the expected instantaneous proﬁt of a dealer equals his ﬂow operation cost. Using (12.14), we can rewrite this condition as σ (υ) θ nij f¯j (kj ) − f¯j (ki ) = κ, υ r + σ (1 − θ )

(12.29)

i,j

since i,j nij kj − ki = 0. It can be shown that there exists a steady-state equilibrium with free entry provided that dealers have some bargaining power, θ > 0. If dealers have no bargaining power, then intermediation fees would equal zero in every trade, and dealers would be unable to cover their operating costs, κ. In this case υ = 0. Suppose instead that θ > 0. As the measure of dealers becomes large, the instantaneous probability that a dealer meets an investor is driven

Liquidity and Trading Frictions

333

to zero, which implies that the expected proﬁt for a dealer becomes negative, since the cost to participate in the market is strictly positive. Conversely, if the measure of dealers approaches zero, then the rate at which a dealer meets an investor grows without bound, and the expected proﬁt of dealership becomes arbitrarily large. Expected proﬁts are positive because investors with different productivities choose different capital stocks even when σ = 0, provided that r > 0; see, for example, equation (12.21). Consequently, since a dealer’s expected proﬁt is continuous in the contact rate, there exists an intermediate value of υ such that the expected proﬁt of a dealer equals zero. Before we proceed, consider the level of dealer entry for the limiting case where the dealer’s operating cost, κ, tends to zero. Since the average fee is positive and bounded away from zero for any σ < ∞, the free-entry condition (12.29) implies that υ → ∞. This in turn implies that σ → ∞, so the equilibrium converges to the frictionless competitive equilibrium. Although the equilibrium does not need to be unique when κ > 0, we now analyze two cases where the equilibrium with entry is, in fact, unique. Suppose ﬁrst that θ = 1, i.e., dealers receive the entire surplus from trade. From (12.29) the free-entry condition becomes σ (υ) δπi πj f¯j (kj ) − f¯j (ki ) = κ. υ σ (υ) + δ r

(12.30)

i=j

Since θ = 1 implies that f¯j and kj are independent of σ , from (12.12) and (12.13), the average fee depends on σ (υ) through the distribution of investors and the dealer’s contact rate. As the number of dealers increases, a larger measure of investors hold their desired portfolios, which reduces dealers’ opportunities to intermediate trades, i.e., an increase in υ increases σ (υ), which in turn decreases nij for i = j. Clearly, the left side of (12.30) is a strictly decreasing function of υ, which implies uniqueness of the steady-state equilibrium with entry. We obtain the comparative static result that higher operation costs, by reducing expected proﬁts, reduce the measure of active dealers, i.e., dυ/dκ < 0. Suppose now that 0 < θ < 1 but that, in the limit, investors’ technologies are linear, i.e., fi (k) → Ai k. Let A1 < A2 < . . . < AI , and recall that in this case only investors with the highest marginal productivity, AI , want to hold assets. From (12.29), δπi πj δπi πj σ (υ) θ (−f¯j (kI )) + f¯I (kI ) = κ. υ r + σ (1 − θ ) σ (υ) + δ σ (υ) + δ i=I,j 0, and dυ/dδ ≷ 0. Lower operation costs naturally imply more entry of dealers. Higher bargaining power for dealers means that they can extract a larger share from the gains from trade in a meeting with an investor, so the measure of dealers increases. Similarly, if the stock of assets increases, the size of each trade is larger and dealers make more proﬁt. Finally, an increase in the frequency of productivity shocks has an ambiguous effect on the equilibrium measure of dealers. On the one hand, a higher δ generates more mismatch, which raises the return to intermediation. On the other hand, since with larger δ the investor’s current productivity reverts back to the mean produc¯ faster, an increase in δ lowers the expected utility of the highest tivity, A, productivity investor relative to the lower productivity investors, which implies smaller gains from trade and consequently lower intermediation fees. We have examined two special cases for which the equilibrium with entry is unique. In general, however, the steady-state equilibrium with free entry need not be unique. The basic reason behind multiple steadystate equilibria is as follows: an increase in the number of dealers leads to an increase in σ (υ). Faster trade means more competition among dealers, which tends to reduce intermediation fees. But as we have pointed out earlier, an increase in σ (υ) also induces investors to take on more extreme asset positions, i.e., more in line with their current as opposed to the mean productivity shock. This means that dealers will, on average, intermediate larger asset-holding reallocations, which implies larger fees, as fees are increasing in the volume traded. If this second effect is sufﬁciently strong, then the model will exhibit multiple steady states. It should now be clear what drives the uniqueness result in the two examples provided above: in both cases this second effect is absent.

Liquidity and Trading Frictions

335

σ (υ) n ϕ −κ υ ∑ ji ji

υ

−κ Figure 12.4 Multiple steady states

In ﬁgure 12.4 we provide a typical representation of a dealer’s expected proﬁt, [σ (υ)/υ] i,j nji φji − κ, where φji = φi (kj ), as a function of the measure of dealers, υ. As υ approaches zero, the contact rate for dealers goes to inﬁnity, while i,j nii φij stays bounded away from zero. Therefore dealers’ expected proﬁts are strictly positive for small υ. As υ goes to inﬁnity, the dealers’ expected proﬁts approach −κ. Thus generically there will be an odd number of steady-state equilibria. In our numerical examples we typically ﬁnd either one or three equilibria. In case of multiple equilibria, the market can be stuck in a low-liquidity equilibrium—an equilibrium where few dealers enter and investors engage in relatively small transactions. The low-liquidity equilibrium exhibits large bid–ask spreads, small trade volume and long trade-execution delays. The high and low equilibria share the following comparative static: a decrease in the participation cost of dealers increases the measure of dealers in the market. And, if the decrease in the participation cost is large enough, the multiplicity of equilibria can be removed. To see this, note that the expected proﬁts curve in ﬁgure 12.4 shifts upward when κ decreases. We conclude this section by considering a linear matching function, σ (υ) = σ0 υ, with σ0 > 0. For this speciﬁcation there is no congestion

336

Chapter 12

effect associated with the entry of dealers: the rate at which dealers ﬁnd orders to execute, σ (υ)/υ = σ0 , is independent of the measure of dealers in the market. From the free-entry condition, υ = 0 if σ0 φ¯ < κ, υ = ∞ if σ0 φ¯ > κ, and υ ∈ [0, ∞] if σ0 φ¯ = κ, where φ¯ represents the average fee ¯ of the dealer. If the average fee as a function of υ, φ(υ) = i,j nij φij , is hump shaped, then the number of equilibria will be either one or three. ¯ To see that φ(υ) can be hump-shaped, recall that when r is close to 0 individual fees, φij , vary in a nonmonotonic fashion with the trading frictions. If the market is either very liquid or very illiquid, then fees are close to 0; for an intermediate level of the trading frictions, fees are ¯ If the strictly positive. The same property holds for the average fee, φ. measure of dealers is very large, the competition effect drives the average fee to zero. If the measure of dealers is very small and investors are very patient, then they choose asset positions that reﬂect their average productivity so that trade sizes are close to zero. In this case the average fee is also close to zero. The average fee is highest for intermediate levels of the trading frictions. If there are multiple equilibria, then one of the equilibria is υ = 0, as illustrated in ﬁgure 12.5. Note that by reducing the cost of dealership κ, or by improving the efﬁciency of the matching technology σ0 , it is possible to eliminate the multiplicity of equilibria.

κ σ0

nji ϕji

υ Figure 12.5 Linear matching and multiple steady states

Liquidity and Trading Frictions

337

12.6 Further Readings Dufﬁe, Gârleanu, and Pedersen (2005, 2007) are the ﬁrst to propose a description of over-the-counter markets based on a search-theoretic model, and to use this approach to explain bid–ask spreads. The model is extended by Weill (2007) to allow for dealers’ inventories. The version in this chapter is based on Lagos and Rocheteau (2007, 2009). In contrast to earlier models, this version relaxes the asset-holding restrictions imposed by Dufﬁe, Gârleanu, and Pedersen, whereby investors can have general asset holdings, not just 0 or 1 units of the asset. Moreover it incorporates more general forms of investor heterogeneity, and it endogenizes the measure of dealers. Gârleanu (2009) also has a version of the model with endogenous asset holdings, and he shows that trading frictions have a second-order effect on asset prices. Lagos, Rocheteau, and Weill (2009) consider a model with both endogenous asset holdings and dealers’ inventories. They investigate how dealers respond to a crash and a stochastic recovery. Vayanos and Weill (2008) use a similar approach to explain the onthe-run phenomenon according to which government securities with identical cash ﬂows can trade at different prices. Weill (2008) develops a search-theoretic model of the cross-sectional distribution of asset returns, abstracting from risk premia and focusing exclusively on liquidity. Ashcraft and Dufﬁe (2007) use a search-theoretic approach to study the market for federal funds, while Gavazza (2009) studies the effects of trading frictions in the commercial aircraft markets. Other papers in this search-theoretic approach to liquidity and ﬁnance include Miao (2004), Rust and Hall (2003), Vayanos and Wang (2002), Kim (2008), and Afonso (2010). Also related is the work of Spulber (1996), who considers a search environment where middlemen intermediate trade between heterogeneous buyers and sellers. There is also a large non–search-based, related literature that studies how exogenously speciﬁed transaction costs affect the functioning of asset markets. This literature includes Amihud and Mendelson (1986), Constantinides (1986), Aiyagari and Gertler (1991), Heaton and Lucas (1996), Vayanos (1998), Vayanos and Vila (1999), Huang (2003), and Lo, Mamaysky, and Wang (2004). See Heaton and Lucas (1995) for a survey of this body of work. There is a collection of papers on search environments with intermediaries, following the pioneering work by Rubinstein and Wolinsky (1987). Rubinstein and Wolinsky consider a market with search frictions

338

Chapter 12

in which a class of agents, called middlemen, have a higher probability of getting matched than nonmiddlemen; see also Yavas (1994). Shevshenko (2004) considers a related environment with a more general inventory problem where middlemen emerge to overcome a double-coincidenceof-wants problems; see also Camera (2001). In Li (1998, 1999) middlemen do not have an advantage in terms of their matching probability, but they invest in a technology to recognize the quality of goods in the presence of private information; see also Biglaiser (1993). Finally, there is a large literature in market microstructure theory that seeks to explain liquidity and trading costs. Broadly speaking, there are two approaches: one based on inventory models, and one based on information asymmetries. Inventory-theoretic models include Amihud and Mendelson (1980), Stoll (1978), and Ho and Stoll (1983). The private information approach of trading costs was pioneered by Kyle (1985) and Glosten and Milgrom (1985).

Bibliography

Chapter numbers following references refer to book chapters where these references are cited. Afonso, Gara. 2010. Liquidity and congestion. Staff report 349. FRB of New York. Chapter 12 Aiyagari, S. Rao, and Mark Gertler. 1991. Asset returns with transactions costs and uninsured individual risk. Journal of Monetary Economics 27: 311–31. Chapter 12 Aiyagari, S. Rao, and Neil Wallace. 1991. Existence of steady states with positive consumption in the Kiyotaki-Wright model. Working paper 428. Federal Reserve Bank of Minneapolis. Chapter 5 Aiyagari, S. Rao, Neil Wallace, and Randall Wright. 1996. Coexistence of money and interest-bearing securities. Journal of Monetary Economics 37: 397–419. Chapter 10 Aiyagari, S. Rao, and Stephen Williamson. 1999. Credit in a random matching model with private information. Review of Economic Dynamics 2: 36–64. Chapter 2 Aiyagari, S. Rao, and Stephen Williamson. 2000. Money and dynamic credit arrangements with private information. Journal of Economic Theory 91: 248–79. Chapter 8 Alchian, Armen. 1977. Economic Forces at Work. Indianapolis: Liberty Press. Chapter 5 Ales, Laurence, Carapella, Francesca, Maziero, Pricila and Weber, Warren. 2008. A model of banknote discounts. Journal of Economic Theory 142: 5–27. Chapter 10 Aliprantis, Charalambos D., Gabriele Camera, and Daniela Puzzello. 2006. Matching and anonymity. Economic Theory 29: 415-32. Chapter 4 Aliprantis, Charalambos D., Gabriele Camera, and Daniela Puzzello. 2007. A random matching theory. Games and Economic Behavior 59: 1–16. Chapter 4 Aliprantis, Charalambos D., Gabriele Camera, and Daniela Puzzello. 2007. Contagion equilibria in a monetary model. Econometrica 75: 277–82. Chapter 3 Amihud, Yakov, and Haim Mendelson. 1980. Dealership markets: Market-making with inventory. Journal of Financial Economics 8: 31–53. Chapter 12 Amihud, Yakov, and Haim Mendelson. 1986. Asset pricing and the bid-ask spread. Journal of Financial Economics 17: 223–49. Chapter 12

340

Bibliography

Andolfatto, David. 2008. The simple analytics of money and credit in a quasi-linear environment. Working paper. Chapter 6 Andolfatto, David. 2010. Essential interest-bearing money. Journal of Economic Theory 145: 1495–1507. Chapter 6 Araujo, Luis. 2004. Social norms and money. Journal of Monetary Economics 51: 241–56. Chapter 3 Araujo, Luis, and Braz Camargo. 2006. Information, learning, and the stability of ﬁat money. Journal of Monetary Economics 53: 1571–91. Chapter 7 Araujo, Luis, and Braz Camargo. 2008. Endogenous supply of ﬁat money. Journal of Economic Theory 142: 48–72. Chapter 6 Araujo, Luis, and Braz Camargo. 2009. Money vs. memory. Working paper. Michigan State University. Chapter 3 Araujo, Luis, and Andrei Shevshenko. 2006. Price dispersion, information and learning. Journal of Monetary Economics 53: 1197–1223. Chapter 7 Arrow, Kenneth J., and Gerard Debreu. 1954. Existence of an equilibrium for a competitive economy. Econometrica 22: 265–90. Introduction Aruoba, S. Boragan, and Sanjay Chugh. 2010. Optimal ﬁscal and monetary policy when money is essential. Journal of Economic Theory 145: 1618–47. Chapter 6 Aruoba, S. Boragan, Guillaume Rocheteau, and Christopher Waller. 2007. Bargaining and the value of money. Journal of Monetary Economics 54: 2636–55. Chapters 4, 6 Aruoba, S. Boragan, and Frank Schorfheide. 2011. Sticky prices versus monetary frictions: An estimation of policy trade-offs. American Economic Journals: Macroeconomics 3: 60–90. Chapter 7 Aruoba, S. Boragan, Christopher Waller, and Randall Wright. 2007. Money and capital: A quantitative analysis. Working paper 0714. Federal Reserve Bank of Cleveland. Chapters 6, 10 Aruoba, S. Boragan, and Randall Wright. 2003. Search, money, and capital: A neoclassical dichotomy. Journal of Money, Credit, and Banking 35: 1085–1106. Chapter 10 Ashcraft, Adam, and Darrell Dufﬁe. 2007. Systemic illiquidity in the federal funds market. American Economic Review 97: 221–25. Chapter 12 Azariadis, Costas. 1993. Intertemporal Macroeconomics. Malden, MA: Blackwell. Chapter 4 Bailey, Martin. 1956. The welfare costs of inﬂationary ﬁnance. Journal of Political Economy 64: 93–110. Chapter 6 Banerjee, Abhijit, and Eric Maskin. 1996. AWalrasian theory of money and barter. Quarterly Journal of Economics 111: 955–1005. Chapter 5 Benabou, Roland. 1988. Search, price setting and inﬂation. Review of Economic Studies 55: 353–76. Chapter 7 Bénassy, Jean-Pascal. 1999. Analytical solutions to a structural signal extraction model: Lucas 1972 revisited. Journal of Monetary Economics 44: 509–21. Chapter 7

Bibliography

341

Berentsen, Aleksander, Gabriele Camera, and Christopher Waller. 2004. The distribution of money and prices in an equilibrium with lotteries. Economic Theory 24: 887–906. Chapter 6 Berentsen, Aleksander, Gabriele Camera, and Christopher Waller. 2005. The distribution of money balances and the nonneutrality of money. International Economic Review 46: 465–87. Chapter 6 Berentsen, Aleksander, Camera, Gabriele, and Christopher Waller. 2007. Money, credit and banking. Journal of Economic Theory 135: 171–95. Chapter 8 Berentsen, Aleksander, Miguel Molico, and Randall Wright. 2002. Indivisibilities, lotteries, and monetary exchange. Journal of Economic Theory 107: 70–94. Chapter 5 Berentsen, Aleksander, and Cyril Monnet. 2008. Monetary policy in a channel system. Journal of Monetary Economics 55: 1067–80. Chapter 6 Berentsen, Aleksander, and Guillaume Rocheteau. 2002. On the efﬁciency of monetary exchange: How divisibility of money matters. Journal of Monetary Economics 49: 1621–49. Chapter 5 Berentsen, Aleksander, and Guillaume Rocheteau. 2003. On the Friedman rule in search models with divisible money. Contributions to Macroeconomics 3: arti. 11. Chapter 5 Berentsen, Aleksander, and Guillaume Rocheteau. 2003. Money and the gains from trade. International Economic Review 44: 263–97. Chapters 3, 4, 5 Berentsen, Aleksander, and Guillaume Rocheteau. 2004. Money and information. Review of Economic Studies 71: 915–44. Chapters 5, 11 Berentsen, Aleksander, Guillaume Rocheteau, and Shouyong Shi. 2007. Friedman meets Hosios: Efﬁciency in search models of money. Economic Journal 117: 174–95. Chapter 6 Bhattacharya, Joydeep, Joseph Haslag, and Antoine Martin. 2005. Heterogeneity, redistribution, and the Friedman rule. International Economic Review 46: 437–54. Chapter 6 Bhattacharya, Joydeep, Joseph Haslag, and Antoine Martin. 2006. Sub-optimality of the Friedman rule in Townsend’s turnpike and stochastic relocation models of money: do ﬁnite lives and initial dates matter? Journal of Economic Dynamics and Control 30: 879–97. Chapter 6 Biglaiser, Gary. 1993. Middlemen as experts. RAND Journal of Economics 24: 212–23. Chapter 12 Boel, Paola, and Gabriele Camera. 2006. Efﬁcient monetary allocations and the illiquidity of bonds. Journal of Monetary Economics 53: 1693–1715. Chapter 10 Boel, Paola, and Gabriele Camera. 2010. The welfare cost of inﬂation in OECD countries. Macroeconomic Dynamics (forthcoming). Chapter 6 Brunner, Karl, and Allan Meltzer. 1971. The uses of money: Money in the theory of an exchange economy. American Economic Review 61: 784–805. Chapters 5, 11 Bryant, John, and Neil Wallace. 1979. The inefﬁciency of interest-bearing national debt. Journal of Political Economy 87: 365–81. Chapter 10 Burdett, Kenneth, Alberto Trejos, and Randall Wright. 2001. Cigarette money. Journal of Economic Theory 99: 117–42. Chapter 10

342

Bibliography

Calvo, Guillermo. 1983. Staggered prices in a utility-maximizing framework. Journal of Monetary Economics 12: 383–98. Chapter 7 Camera, Gabriele. 2001. Search, dealers and the terms of trade. Review of Economic Dynamics 4: 680–94. Chapter 12 Camera, Gabriele, and Dean Corbae. 1999. Money and price dispersion. International Economic Review 40: 985-1008. Chapter 5 Camera, Gabriele, Ben Craig, and Christopher Waller. 2004. Currency competition in a fundamental model of money. Journal of International Economics 64: 521–44. Chapter 10 Camera, Gabriele, and Yiting Li. 2008. Another example of a credit system that coexists with money. Journal of Money, Credit and Banking 40: 1295-1308. Chapter 8 Camera, Gabriele, Rob Reed, and Christopher Waller. 2003. Jack of all trades or master of one? Specialization, trade and money. International Economic Review 44: 1275–94. Chapters 3, 6 Camera, Gabriele, and Johannes Winkler. 2003. International monetary exchange and the law of one price. Journal of Monetary Economics 50: 1531–53. Chapter 10 Cass, David. 1965. Optimum growth in an aggregative model of capital accumulation. Review of Economic Studies 37: 233–40. Introduction Cavalcanti, Ricardo, Andres Erosa, and Ted Temzelides. 1999. Private money and reserve management in a random matching model. Journal of Political Economy 107: 929–45. Chapter 8 Cavalcanti, Ricardo, and Ed Nosal. 2007. Counterfeiting as private money in mechanism design. Working paper. 0716. Federal Reserve Bank of Cleveland. Chapter 5 Chiu, Jonathan, and Miguel Molico. 2010. Liquidity, redistribution, and the welfare cost of inﬂation. Journal of Monetary Economics 57: 428–38. Chapter 1 Clower, Robert. 1967. A reconsideration of the microfoundations of monetary theory. Western Economic Journal 6: 1–8. Introduction Coles, Melvyn, and Randall Wright. 1998. A dynamic equilibrium model of search, bargaining, and money. Journal of Economic Theory 78: 32–54. Chapter 4 Cone, Thomas. 2005. Anticipated inﬂation and rates of return in a search-theoretic monetary model. Journal of Monetary Economics 52: 567–674. Chapter 11 Constantinides, George. 1986. Capital market equilibrium with transaction costs. Journal of Political Economy 94: 842–62. Chapter 12 Corbae, Dean, and Joseph Ritter. 2004. Decentralized credit and monetary exchange without public record keeping. Economic Theory 24: 933–51. Chapters 2, 8 Corbae, Dean, Ted Temdelizes, and Randall Wright. 2002. Money and matching. American Economic Review: Papers and Proceedings 92: 67–71. Chapter 3 Corbae, Dean, Ted Temdelizes, and Randall Wright. 2003. Directed matching and monetary exchange. Econometrica 71: 731–56. Chapter 3 Craig, Ben, and Guillaume Rocheteau. 2008a. Inﬂation and welfare: A search approach. Journal of Money, Credit and Banking 40: 89–119. Chapter 6

Bibliography

343

Craig, Ben, and Guillaume Rocheteau. 2008b. State-dependent pricing, inﬂation and welfare. European Economic Review 52: 441–68. Chapter 7 Craig, Ben, and Christopher Waller. 2000. Dual-currency economies as multiple-payment systems. Economic Review 36: 2–13. Chapter 10 Craig, Ben, and Craig Waller. 2004. Dollarization and currency exchange. Journal of Monetary Economics 51: 671–89. Chapter 10 Cuadras-Morato, Xavier. 1994. Commodity money in the presence of goods of heterogeneous quality. Economic Theory 4: 579–91. Chapter 5 Curtis, Elisabeth S., and Christopher Waller. 2000. A search-theoretic model of legal and illegal currency. Journal of Monetary Economics 45: 155–84. Chapter 10 Curtis, Elisabeth S., and Christopher Waller. 2003. Currency restrictions, government transaction policies and currency exchange. Economic Theory 21: 19–42. Chapter 10 Debreu, Gerard. 1959. Theory of Value: An Axiomatic Analysis of Economic Equilibrium. New York: Wiley. Reprint. New Haven: Yale University Press, 1971. Introduction DeMarzo, Peter, and Darrell Dufﬁe. 1999. A liquidity-based model of security design. Econometrica 67: 65–99. Chapter 11 Deviatov, Alexei, and Neil Wallace. 2001. Another example in which lump-sum money creation is beneﬁcial. Advances in Macroeconomics 1: arti. 1. Chapter 6 Diamond, Peter. 1982. Aggregate demand management in search equilibrium. Journal of Political Economy 90: 881–94. Chapters 1, 2, 12 Diamond, Peter. 1984. Money in search equilibrium. Econometrica 52: 1–20. Chapters 1, 3, 5 Diamond, Peter. 1987a. Credit in search equilibrium. In Meir Kohn and Sho-Chieh Tsiang, eds., Financial Constraints, Expectations, and Macroeconomics. Oxford: Oxford University Press. Chapter 2 Diamond, Peter. 1987b. Multiple equilibria in models of credit. American Economic Review 77: 82–86. Chapter 2 Diamond, Peter. 1990. Pairwise credit in search equilibrium. Quarterly Journal of Economics 105: 285–319. Chapter 2 Diamond, Peter. 1993. Search, sticky prices, and inﬂation. Review of Economic Studies 60: 53–68. Chapter 7 Dotsey, Michael, and Peter Ireland. 1996. The welfare cost of inﬂation in general equilibrium. Journal of Monetary Economics 37: 29–47. Chapter 8 Dufﬁe, Darrell, Nicolae Garleanu, and Lasse H. Pedersen. 2005. Over-the-counter markets. Econometrica 73: 1815–47. Chapter 12 Dufﬁe, Darrell, Nicolae Garleanu, and Lasse H. Pedersen. 2007. Valuation in over-thecounter markets. Review of Financial Studies 20: 1865–1900. Chapter 12 Engineer, Merwan, and Shouyong Shi. 1998. Asymmetry, imperfectly transferable utility, and ﬁat money in improving the terms of trade. Journal of Monetary Economics 41: 153–83. Chapter 3

344

Bibliography

Engineer, Merwan, and Shouyong Shi. 2001. Bargains, barter, and money. Review of Economic Dynamics 4: 188–209. Chapter 3 Ennis, Huberto. 2001. On random matching, monetary equilibria, and sunspots. Macroeconomic Dynamics 5: 132–42. Chapter 4 Ennis, Huberto. 2004. Macroeconomic ﬂuctuations and bargaining. Journal of Economic Theory 115: 322–40. Chapter 4 Ennis, Huberto. 2008. Search, money, and inﬂation under private information. Journal of Economic Theory 138: 101–31. Chapter 6 Ennis, Huberto. 2009. Avoiding the inﬂation tax. International Economic Review 50: 607–25. Chapter 6 Faig, Miquel. 2008. Endogenous buyer-seller choice and divisible money in search equilibrium. Journal of Economic Theory 141: 184–99. Chapters 4, 6 Faig, Miquel, and Belen Jerez. 2006. Inﬂation, prices, and information in competitive search. Advances in Macroeconomics 6, 1313. Chapters 4, 6 Faig, Miquel, and Zhe Li. 2009. The welfare cost of expected and unexpected inﬂation. Journal of Monetary Economics 56: 1004–13. Chapter 7 Fischer, Stanley. 1974. Money and the production function. Economic Inquiry 12: 517–33. Introduction Freeman, Scott. 1985. Transaction costs and the optimal quantity of money. Journal of Political Economy 93: 145–57. Chapter 11 Freeman, Scott. 1996a. Clearinghouse banks and banknote over-issue. Journal of Monetary Economics 38: 101–15. Chapter 9 Freeman, Scott. 1996b. The payments system, liquidity, and rediscounting. American Economic Review 86: 1126–38. Chapter 9 Freeman, Scott. 1999. Rediscounting under aggregate risk. Journal of Monetary Economics 43: 197–216. Chapter 9 Freeman, Scott, and Joseph Haslag. 1996. On the optimality of interest-bearing reserves in economies of overlapping generations. Economic Theory 7: 557–65. Chapter 6 Freeman, Scott, and Finn Kydland. 2000. Monetary aggregates and output. American Economic Review 90: 1125–35. Chapter 8 Friedman, Milton. 1969. The optimum quantity of money. In The Optimum Quantity of Money and Other Essays. 1–50. Chapter 6 Friedman, Milton. 1960. A Program for Monetary Stability. New York: Fordham University Press. Chapter 6 Galenianos, Manolis, and Philipp Kircher. 2008. A model of money with multilateral matching. Journal of Monetary Economics 55: 1054-66. Chapter 4 Garleanu, Nicolae. 2009. Pricing and portfolio choice in illiquid markets. Journal of Economic Theory 144: 532–64. Chapter 12 Gavazza, Alessandro. 2009. Leasing and secondary markets: Theory and evidence from commercial aircraft. Working paper. New York University. Chapter 12

Bibliography

345

Geromichalos, Athanasios, Juan Licari, and Jose Suarez-Lledo. 2007. Monetary policy and asset prices. Review of Economic Dynamics 10: 761–79. Chapter 11 Glosten, Lawrence, and Paul Milgrom. 1985. Bid, ask and transaction prices in a specialist market with heterogeneously informed traders. Journal of Financial Economics 14: 71–100. Chapters 11, 12 Gomis-Porqueras, Pere, and Adrian Peralta-Ava. 2008. Optimal monetary and ﬁscal policies in a search theoretic model of monetary exchange. Working paper 2008–015. Federal Reserve Bank of St. Louis. Chapter 6 Grandmont, Jean-Michel. 1985. On endogenous competitive business cycles. Econometrica 53: 995–1046. Chapter 4 Green, Edward. 1999. Money and debt in the structure of payments. Federal Reserve Bank of Minneapolis, Quarterly Review 23: 13–29. Chapter 9 Green, Edward, and Warren Weber. 1996. Will the new $100 bill decrease counterfeiting? Working paper 571. Federal Reserve Bank of Minneapolis. Chapter 5 Green, Edward, and Ruilin Zhou. 1998. A rudimentary random-matching model with divisible money and prices. Journal of Economic Theory 81: 252–71. Chapter 4 Green, Edward, and Ruilin Zhou. 2002. Dynamic monetary equilibrium in a randommatching economy. Econometrica 70: 929–69. Chapter 4 Harris, Larry. 2003. Trading & Exchanges, Market Microstructure for Practitioners. New York: Oxford University Press. Chapter 12 Haslag, Joseph, and Antoine Martin. 2007. Optimality of the Friedman rule in an overlapping generations model with spatial separation. Journal of Money, Credit and Banking 39: 1741–58. Chapter 6 Head, Allen, and Shouyong Shi. 2003. A fundamental theory of exchange rates and direct currency trades. Journal of Monetary Economics 50: 1555–91. Chapter 10 Heaton, John, and Deborah Lucas. 1995. The importance of investor heterogeneity and ﬁnancial market imperfections for the behavior of asset prices. Carnegie-Rochester Conference Series on Public Policy 42: 1–32. Chapter 12 Heaton, John, and Deborah Lucas. 1996. Evaluating the effects of incomplete markets on risk sharing and asset pricing. Journal of Political Economy 104: 443–87. Chapter 12 Hicks, J. R. 1935. A suggestion for simplifying the theory of money. Economica 2: 1–19. Introduction Ho, Thomas, and Hans Stoll. 1983. The dynamics of dealer markets under competition. Journal of Finance 38: 1053–74. Chapter 12 Holström, Bengt, and Jean Tirole. 1998. Private and public supply of liquidity. Journal of Political Economy 106: 1–40. Chapter 11 Holström, Bengt, and Jean Tirole. 2001. LAPM—A liquidity-based asset pricing model. Journal of Finance 56: 1837–67. Chapter 11 Hopenhayn, Hugo, and Ingrid Werner. 1996. Information, liquidity, and asset trading in a random matching game. Journal of Economic Theory 68: 349–79. Chapter 11

346

Bibliography

Howitt, Peter. 2005 Beyond search: Fiat money in organized exchange. International Economic Review 46: 405–29. Chapter 4 Hu, Tai-wei, John Kennan, and Neil Wallace. 2009. Coalition-proof trade and the Friedman rule in the Lagos–Wright model. Journal of Political Economy 117: 116–37. Chapter 6 Huang, Ming. 2003. Liquidity shocks and equilibrium liquidity premia. Journal of Economic Theory 109: 104–29. Chapter 12 Iwai, Katsuhito. 1996. The bootstrap theory of money: A search-theoretic foundation of monetary economics. Structural Change and Economic Dynamics 7: 451–77. Chapters 1, 3 Jafarey, Saqib, and Peter Rupert. 2001. Limited commitment, money, and credit. Journal of Economic Theory 99: 22–58. Chapter 2 Jevons, William S. 1875. Money and the Mechanism of Exchange. New York: D. Appleton. Chapter 5 Jin, Yi, and Ted Temzelides. 2004. On the local interaction of money and credit. Review of Economic Dynamics 7: 143–56. Chapter 8 Jones, Robert A. 1976. The origin and development of media of exchange. Journal of Political Economy 84: 757–75. Introduction, Chapter 1 Julien, Benoit, John Kennes, and Ian King. 2008. Bidding for money. Journal of Economic Theory 142: 196–217. Chapter 4 Kahn, Charles. 2009. Collateral policy in a world of round-the-clock payment. Working paper. University of Illinois. Chapter 8 Kahn, Charles, and William Roberds. 2009. Why pay? An introduction to payments economics. Journal of Financial Intermediation 18: 1–23. Chapter 9 Kamiya, Kazuya, and Takashi Shimizu. 2006. Real indeterminacy of stationary equilibria in matching models with divisible money. Journal of Mathematical Economics 42: 594–617. Chapter 4 Kamiya, Kazuya, Morishita, Noritsugu, and Takashi Shimizu. 2005. On the existence of single-price equilibria in a matching model with divisible money and production cost. International Journal of Economic Theory 1: 219–31. Chapter 4 Kamiya, Kazuya, and Takashi Sato. 2004. Equilibrium price dispersion in a matching model with divisible money. International Economic Review 45: 413–30. Chapter 4 Kamiya, Kazuya, and Takashi Shimizu. 2007a. Existence of equilibria in matching models of money: A new technique. Economic Theory 32: 447–60. Chapter 4 Kamiya, Kazuya, and Takashi Shimizu. 2007b. On the role of tax-subsidy scheme in money search models. International Economic Review 48: 575–606. Chapter 4 Karaken, John, and Neil Wallace. 1981. On the indeterminacy of equilibrium exchange rates. Quarterly Journal of Economics 96: 207–22. Chapter 10 Kehoe, Timothy, and David Levine. 1993. Debt-constrained asset markets. Review of Economic Studies 60: 865–88. Chapters 2, 8 Kehoe, Timothy, Nobuhiro Kiyotaki, and Randall Wright. 1993. More on money as a medium of exchange. Economic Theory 3: 297–314. Chapter 5

Bibliography

347

Kim, Yong. 2008. Liquidity and selection in asset markets with search frictions. Mimeo, University of Southern California. Chapter 12 Kim, Young Sik. 1996. Money, barter and costly information acquisition. Journal of Monetary Economics 37: 119–42. Chapters 5, 11 Kim, Young Sik, and Manjong Lee. 2008. Recognizability and liquidity. Mimeo. Korea University. Chapter 11 King, Robert, and Charles Plosser. 1986. Money as the mechanism of exchange. Journal of Monetary Economics 17: 93–115. Chapters 5, 11. Kiyotaki, Nobuhiro, Akihiko Matsui, and Kiminori Matsuyama. 1993. Toward a theory of international currency. Review of Economic Studies 60: 283–307. Chapter 10 Kiyotaki, Nobuhiro, and John Moore. 2002. Evil is the root of all money. American Economic Review, Papers and Proceedings 92: 62–66. Chapter 3 Kiyotaki, Nobuhiro, and John Moore. 2005. 2002 Lawrence R. Klein lecture: Liquidity and asset prices. International Economic Review 46: 317–49. Chapter 11 Kiyotaki, Nobuhiro, and Randall Wright. 1989. On money as a medium of exchange. Journal of Political Economy 97: 927–54. Chapters 1, 3, 4, 5, 10 Kiyotaki, Nobuhiro, and Randall Wright. 1991. A contribution to the pure theory of money. Journal of Economic Theory 53: 215–35. Chapters 1, 3, 5 Kiyotaki, Nobuhiro, and Randall Wright. 1993. A search-theoretic approach to monetary economics. American Economic Review 83: 63–77. Chapters 1, 3, 4, 5 Kocherlakota, Narayana. 1998a. Money is memory. Journal of Economic Theory 81: 232–51. Chapters 2, 3 Kocherlakota, Narayana. 1998b. The technological role of ﬁat money. Federal Reserve Bank of Minneapolis Quarterly Review 22: 2–10. Chapters 2, 3 Kocherlakota, Narayana. 2002. The two-money theorem. International Economic Review 43: 333–46. Chapter 10 Kocherlakota, Narayana. 2003. Social beneﬁts of illiquid bonds. Journal of Economic Theory 108: 179–93. Chapters 8, 10 Kocherlakota, Narayana, and Thomas Krueger. 1999. A signaling model of multiple currencies. Review of Economic Dynamics 2: 231–44. Chapter 10 Kocherlakota, Narayana, and Neil Wallace. 1998. Incomplete record-keeping and optimal payment arrangements. Journal of Economic Theory 81: 272–89. Chapters 2, 8 Koeppl, Thorsten, Cyril Monnet, and Ted Temzelides. 2008. A dynamic model of settlement. Journal of Economic Theory 142: 233–46. Chapter 9 Koopmans, Tjalling. 1965. On the concept of optimal economic growth. In The Econometric Approach to Development Planning. Amsterdam: North Holland, 225–87. Introduction Kranton, Rachel. 1996. Reciprocal exchange: A self-sustaining system. American Economic Review 86: 830–51. Chapter 2 Kultti, Klaus. 1996. A monetary economy with counterfeiting. Journal of Economics 63: 175–86. Chapter 5

348

Bibliography

Kydland, Finn E., and Edward Prescott. 1982. Time to build and aggregate ﬂuctuations. Econometrica 50: 1345–70. Introduction Kyle, Albert. 1985. Continuous auctions and insider trading. Econometrica 53: 1315–35. Chapters 11, 12 Lagos, Ricardo. 2006. Asset prices and liquidity in an exchange economy. Staff report 373. Federal Reserve Bank of Minneapolis. Introduction, Chapter 11 Lagos, Ricardo. 2010. Some results on the optimality and implementation of the Friedman rule in the search theory of money. Journal of Economic Theory 145: 1508–24. Chapter 6 Lagos, Richard, and Guillaume Rocheteau. 2005. Inﬂation, output, and welfare. International Economic Review 46: 495–522. Chapters 1, 6 Lagos, Richard, and Guillaume Rocheteau. 2007. Search in asset markets: market structure, liquidity, and welfare. American Economic Review 97: 198–202. Chapter 12 Lagos, Ricardo, and Guillaume Rocheteau. 2008. Money and capital as competing media of exchange. Journal of Economic Theory 142: 247–58. Chapter 10 Lagos, Richard, and Guillaume Rocheteau. 2009. Liquidity in asset markets with search frictions. Econometrica 77: 403–26. Chapter 12 Lagos, Ricardo, Guillaume Rocheteau, and Pierre-Olivier Weill. 2009. Crises and liquidity in over-the-counter markets. Working paper. New York University. Chapter 12 Lagos, Ricardo, and Randall Wright. 2003. Dynamics, cycles, and sunspot equilibria in “genuinely dynamic, fundamentally disaggregative’’ models of money. Journal of Economic Theory 109: 156–71. Chapter 4 Lagos, Ricardo, and Randall Wright. 2005. A uniﬁed framework for monetary theory and policy analysis. Journal of Political Economy 113: 463–84. Chapters 1, 4, 6, 9 Laing, Derek, Victor Li, and Ping Wang. 2007. Inﬂation and productive activity in a multiple-matching model of money. Journal of Monetary Economics 54: 1949–61. Chapter 4 Lester, Benjamin, Andrew Postlewaite, and Randall Wright. 2008. Information, liquidity and asset prices. PIER Working paper archive 08-039. Penn Institute for Economic Research. Chapter 11 Levine, David. 1991. Asset trading mechanisms and expansionary policy. Journal of Economic Theory 54: 148–64. Chapter 6 Li, Victor. 1994. Inventory accumulation in a search-based monetary economy. Journal of Monetary Economics 34: 511–36. Chapter 6 Li, Victor. 1995. The optimal taxation of ﬁat money in search equilibrium. International Economic Review 36: 927–42. Chapter 6 Li, Victor. 1997. The efﬁciency of monetary exchange in search equilibrium. Journal of Money, Credit, and Banking 28: 61–72. Chapter 6 Li, Yiting. 1995. Commodity money under private information. Journal of Monetary Economics 36: 573–92. Chapters 5, 10 Li, Yiting. 1998. Middlemen and private information. Journal of Monetary Economics 42: 131– 59. Chapter 12

Bibliography

349

Li, Yiting. 1999. Money and middlemen in an economy with private information. Economic Inquiry 37: 1–12. Chapter 12 Li, Yiting. 2001. A search model of money and circulating private debt with applications to monetary policy. International Economic Review 42: 925–46. Chapter 8 Li, Yiting. 2002. Government transaction policy and Gresham’s law. Journal of Monetary Economics 49: 435–53. Chapter 10 Li, Yiting. 2006. Banks, private money, and government regulation. Journal of Monetary Economics 53: 2067–83. Chapter 8 Li, Yiting. 2007. Inside money, organized markets, and specialization. Macroeconomic Dynamics 11: 388–404. Chapter 8 Li, Yiting, and Guillaume Rocheteau. 2008. On the threat of counterfeiting, Working paper 0809. Federal Reserve Bank of Cleveland. Chapter 5 Li, Yiting, and Guillaume Rocheteau. 2009. Liquidity constraints. Working paper. University of California at Irvine. Chapters 5, 10 Li, Yiting, and Randall Wright. 1998. Government transaction policy, media of exchange, and prices. Journal of Economic Theory 81: 290–313. Chapter 10 Liu, Lucy Qian, Liang Wang, and Randall Wright. 2009. The “hot potato’’ effect of inﬂation. Working paper. University of Pennsylvania. Chapter 6 Liu, Qing, and Shouyong Shi. 2010. Currency areas and monetary coordination. International Economic Review 51: 813–836. Chapter 10 Lo, Andrew, Harry Mamaysky, and Jiang Wang. 2004. Asset prices and trading volume under ﬁxed transactions costs. Journal of Political Economy 112: 1054–90. Chapter 12 Lomeli, Hector, and Ted Temzelides. 2002. Discrete time dynamics in a random matching monetary model. Economic Theory 20: 259–69. Chapter 4 Lotz, Sebastien. 2004. Introducing a new currency: Government policy and prices. European Economic Review 48: 959–82. Chapter 10 Lotz, Sebastien, and Guillaume Rocheteau. 2002. On the launching of a new currency. Journal of Money, Credit and Banking 34: 563–88. Chapter 10 Lotz, Sebastien, Andrei Shevshenko, and Christopher Waller. 2007. Heterogeneity and lotteries in monetary search models. Journal of Money, Credit and Banking 39: 703-12. Chapter 5 Lucas, Robert. 1972. Expectations and the neutrality of money. Journal of Economic Theory 4: 103–24. Chapter 7 Lucas, Robert. 1973. Some international evidence on output-inﬂation tradeoffs. American Economic Review 63: 326–34. Chapter 7 Lucas, Robert. 1978. Asset prices in an exchange economy. Econometrica 46: 1429–45. Chapter 11 Lucas, Robert. 1980. Two illustrations of the quantity theory of money. American Economic Review 70: 1005–14. Introduction

350

Bibliography

Lucas, Robert. 1990. Liquidity and interest rates. Journal of Economic Theory 50: 237–64. Chapter 4 Lucas, Robert. 2000. Inﬂation and welfare. Econometrica 68: 247–74. Chapter 6 Lucas, Robert, and Nancy Stokey. 1987. Money and interest in a cash-in-advance economy. Econometrica 55: 491–513. Chapter 8 Mankiw, N. Gregory, and Ricardo Reis. 2002. What measure of inﬂation should a central bank target? Working paper 9375. National Bureau of Economic Research, Cambridge, MA. Chapter 7 Marimon, Ramon, and Fabrizio Zilibotti. 1997. Unemployment vs. mismatch of talents: Reconsidering unemployment beneﬁts. Economic Journal 109: 266–29. Chapter 5 Mehra, Rajnish, and Edward Prescott. 1985. The equity premium: A puzzle. Journal of Monetary Economics 15: 145–61. Chapter 11 Miao, Jianjun. 2006. A search model of centralized and decentralized trade. Review of Economic Dynamics 9: 68–92. Chapter 12 Moen, Espen. 1997. Competitive search equilibrium. Journal of Political Economy 105: 385– 411. Chapter 4 Molico, Miguel. 2006. The distribution of money and prices in search equilibrium. International Economic Review 47: 701–22. Chapters 5, 6 Molico, Miguel, and Yahong Zhang. 2006. Monetary policy and the distribution of money and capital. Computing in Economics and Finance: 136. Chapter 10 Monnet, Cyril. 2005. Counterfeiting and inﬂation. Working paper 512. European Central Bank. Chapter 5 Mortensen, Dale, and Randall Wright. 2002. Competitive pricing and efﬁciency in search equilibrium. International Economic Review 43: 1–20. Chapter 4 Myers, Stewart, and Nicholal Majluf. 1984. Corporate ﬁnancing and investment decisions when ﬁrms have information that investors do not have. Journal of Financial Economics 13: 187–221. Chapter 11 Nash, John. 1953. Two person cooperative games. Econometrica 21: 128–40. Chapter 2 Niehans, Jurg. 1971. Money and barter in general equilibrium with transactions costs. American Economic Review 61: 773–83. Introduction Niehans, Jurg. 1978. The Theory of Money, Baltimore: Johns Hopkins University Press. Introduction Nosal, Ed. 2008. Search, welfare and the hot potato effect of inﬂation. Working paper. Federal Reserve Bank of Chicago. Chapter 6 Nosal, Ed, and Guillaume Rocheteau. 2008. The economics of payments. Policy discussion paper. Federal Reserve Bank of Cleveland. Chapter 9 Nosal, Ed, and Guillaume Rocheteau. 2009. Pairwise trades, asset prices, and monetary policy. Working paper. Federal Reserve Bank of Chicago. Chapter 11 Nosal, Ed, and Neil Wallace. 2007. A model of (the threat of) counterfeiting. Journal of Monetary Economics 54: 994–1001. Chapter 5

Bibliography

351

Obstfeld, Maurice, and Kenneth Rogoff. 1996. Foundations of International Macroeconomics. Cambridge: MIT Press. Chapter 10 Oh, Seonghwan. 1989. A theory of a generally acceptable medium of exchange and barter. Journal of Monetary Economics 23: 101–19. Chapter 3 Osborne, Martin, and Ariel Rubinstein. 1990. Bargaining and markets. San Diego: Academic Press. Chapters 2, 4 Ostroy, Joseph. 1973. The informational efﬁciency of monetary exchange. American Economic Review 63: 597–610. Chapter 3 Ostroy, Joseph, and Ross Starr. 1974. Money and the decentralization of exchange. Econometrica 42: 1093–1113. Chapter 3 Ostroy, Joseph, and Ross Starr. 1990. The transactions role of money. In Benjamin Friedman and Frank Hahn, eds., Handbook of Monetary Economics. Chapter 3 Patinkin, Don. 1965. Money, Interest and Prices, 2nd ed. New York: Harper Row. Introduction Peterson, Brian, and Shouyong Shi. 2004. Money, price dispersion and welfare. Economic Theory 24: 907–32. Chapter 5 Phillips, A. W. 1958. The relationship between unemployment and the rate of change of money wages in the United Kingdom 1861–1957. Economica 25: 283–299. Chapter 7 Pissarides, Christopher. 2000. Equilibrium Unemployment, 2nd ed. Cambridge: MIT Press. Chapters 2, 5 Prescott, Edward. 1987. A multiple means-of-payment model. In W. Barnett and K. Singleton, eds., New Approaches to Monetary Economics. Cambridge: Cambridge University Press. Chapter 8 Quercioli, Elena, and Lones Smith. 2009. The economics of counterfeiting. Working paper 1325892. Social Science Research Network. Chapter 5 Rauch, Bernhard. 2000. A divisible search model of ﬁat money: a comment. Econometrica 68: 149–56. Chapters 4, 6 Redish, Angela, and Warren E. Weber. 2008. Coin sizes and payments in commodity money systems. Staff report 416. Federal Reserve Bank of Minneapolis. Chapter 5 Reed, Robert, and Christopher Waller. 2006. Money and risk sharing. Journal of Money, Credit, and Banking 38: 1599–618. Chapter 6 Renero, Juan M. 1998. Unstable and stable steady-states in the Kiyotaki–Wright model. Economic Theory 11: 275–94. Chapter 5 Renero, Juan M. 1999. Does and should a commodity medium of exchange have relatively low storage costs? International Economic Review 40: 251–64. Chapter 5 Rocheteau, Guillaume. 2008. Money and competing assets under private information. Working paper 0802. Federal Reserve Bank of Cleveland. Chapter 5 Rocheteau, Guillaume. 2009a. A monetary approach to asset liquidity. Working paper 09-01. Federal Reserve Bank of Cleveland. Chapter 11

352

Bibliography

Rocheteau, Guillaume. 2009b. Information and liquidity: A discussion. Working paper 09-02. Federal Reserve Bank of Cleveland. Chapter 11 Rocheteau, Guillaume, Peter Rupert, and Randall Wright. 2007. Inﬂation and unemployment in general equilibrium. Scandavian Journal of Economics 109: 837–55. Chapter 1 Rocheteau, Guillaume, Peter Rupert, Karl Shell, and Randall Wright. 2008. General equilibrium with nonconvexities and money. Journal of Economic Theory 142: 294–317. Introduction Rocheteau, Guillaume, and Randall Wright. 2005. Money in search equilibrium, in competitive equilibrium, and in competitive search equilibrium. Econometrica 73: 175–202. Chapters 1, 4, 6 Rocheteau, Guillaume, and Randall Wright. 2009. Inﬂation and welfare in models with trading frictions. In Ed Nosal and Dave Altig, eds., Monetary Policy in Low Inﬂation Economies. Cambridge: Cambridge University Press. Chapter 6 Rotemberg, Julio. 1982. Sticky prices in the United States. Journal of Political Economy 90: 1187–211. Chapter 7 Rubinstein, Ariel, and Asher Wolinsky. 1987. Middlemen. Quarterly Journal of Economics 102: 581–94. Chapter 12 Rupert, Peter, Martin Schindler, and Randall Wright. 2000. The search-theoretic approach to monetary economics: A primer. Federal Reserve Bank of Cleveland Economic Review 36: 10–28. Chapter 5 Rust, John, and George Hall. 2003. Middlemen versus market makers: A theory of competitive exchange. Journal of Political Economy 111: 353–403. Chapter 12 Samuelson, Paul. 1958. An exact consumption-loan model of interest with or without the social contrivance of money. Journal of Political Economy 66: 467. Introduction Sanches, Daniel, and Stephen Williamson. 2010. Money and credit with limited commitment and theft. Journal of Economic Theory 145: 1525–49. Chapters 6, 7 Sargent, Thomas, and Neil Wallace. 1983. A model of commodity money. Journal of Monetary Economics 12: 163–87. Chapter 10 Sargent, Thomas, and Neil Wallace. 1985. Interest on reserves. Journal of Monetary Economics 15: 279–90. Chapter 6 Schreft, Stacey. 1992. Transaction costs and the use of cash and credit. Economic Theory 2: 283–96. Chapter 8 Shevshenko, Andrei. 2004. Middlemen. International Economic Review 45: 1–24. Chapter 12 Shevshenko, Andrei, and Randall Wright. 2004. A simple search model of money with heterogeneous agents and partial acceptability. Economic Theory 24: 877–85. Chapter 5 Shi, Shouyong. 1995. Money and prices: A model of search and bargaining. Journal of Economic Theory 67: 467–96. Chapters 4, 5 Shi, Shouyong. 1996. Credit and money in a search model with divisible commodities. Review of Economic Studies 63: 627–52. Chapter 8 Shi, Shouyong. 1997a. A divisible search model of ﬁat money. Econometrica 65: 75–102. Chapters 4, 6

Bibliography

353

Shi, Shouyong. 1997b. Money and specialization. Economic Theory 10: 99–113. Chapter 3 Shi, Shouyong. 1998. Search for a monetary propagation mechanism. Journal of Economic Theory 81: 314–52. Chapter 6 Shi, Shouyong. 1999a. Search, inﬂation, and capital accumulation. Journal of Monetary Economics 44: 81–103. Chapters 4, 10 Shi, Shouyong. 1999b. Money, capital and redistributive effects of monetary policies. Journal of Economic Dynamics and Control 23: 565–90. Chapters 4, 10 Shi, Shouyong. 2001. The extent of the market and the optimal degree of specialization. Mimeo. Chapters 2, 4 Shi, Shouyong. 2004. Liquidity, interest rates, and output. Mimeo. Chapter 10 Shi, Shouyong. 2005. Nominal bonds and interest rates. International Economic Review 45: 579–612. Chapter 10 Shi, Shouyong. 2006. Viewpoint: A microfoundation of monetary economics. Canadian Journal of Economics 39: 643–88. Chapter 1 Shi, Shouyong. 2008. Efﬁciency improvement from restricting the liquidity of nominal bonds. Journal of Monetary Economics 55: 1025–37. Chapter 10 Shi, Shouyong, and Weimin Wang. 2006. The variability of velocity of money in a search model. Journal of Monetary Economics 53: 537–71. Chapter 6 Smith, Bruce. 1989. Limited information, money, and competitive equilibrium. Canadian Journal of Economics 19: 780–97. Chapter 2 Smith, Bruce. 1991. Interest on reserves and sunspot equilibria: Friedman’s proposal reconsidered. Review of Economic Studies 58: 93–105. Chapter 6 Spulber, Daniel. 1996. Market microstructure and intermediation. Journal of Economic Perspectives 10: 135–52. Chapter 12 Starr, Ross. 1972. The structure of exchange in barter and monetary economies. Quarterly Journal of Economics 86: 290–302. Introduction Stockman, Alan. 1981. Anticipated inﬂation and the capital stock in a cash-in-advance economy. Journal of Monetary Economics 8: 387–93. Chapter 10 Stoll, Hans R. 1978. The supply of dealer services in securities markets. Journal of Finance 33: 1133–51. Chapter 12 Taber, Alexander, and Neil Wallace. 1999. A matching model with bounded holdings of indivisible money. International Economic Review 40: 961–84. Chapter 5 Taylor, John. 1980. Aggregate dynamics and staggered contracts. Journal of Political Economy 88: 1–23. Chapter 7 Telyukova, Irina, and Randall Wright. 2008. A model of money and credit, with application to the credit card debt puzzle. Review of Economic Studies 75: 629–47. Chapter 8 Temzelides, Ted, and Stephen Williamson. 2001. Payments systems design in deterministic and private information environments. Journal of Economic Theory 99: 297–326. Chapter 9

354

Bibliography

Tirole, Jean. 1985. Asset bubbles and overlapping generations. Econometrica 53: 1499–1528. Chapter 4 Tobin, James. 1965. The monetary interpretation of history. American Economic Review 55: 464–85. Chapter 10 Townsend, Robert. 1980. Models of money with spatially separated agents. In John Kareken and Neil Wallace, eds., Models of Monetary Economies. Federal Reserve Bank of Minneapolis, 265–303. Chapter 4 Townsend, Robert. 1987. Economic organization with limited communication. American Economic Review 77: 954–71. Chapter 3 Townsend, Robert. 1989. Currency and credit in a private information economy. Journal of Political Economy 97: 1323–44. Chapters 3, 8 Trejos, Alberto. 1999. Search, bargaining, money and prices under private information. International Economic Review 40: 679–95. Chapter 5 Trejos, Alberto, and Randall Wright. 1993. Search, bargaining, money and prices: Recent results and policy implications. Journal of Money, Credit, and Banking 25: 558–76. Chapter 4 Trejos, Alberto, and Randall Wright. 1995. Search, bargaining, money, and prices. Journal of Political Economy 103: 118–41. Chapters 4, 5 Trejos, Alberto, and Randall Wright. 1996. Search-theoretic models of international currency. Federal Reserve Bank of St Louis Review 78: 117–32. Chapter 10 Vayanos, Dimitri. 1998. Transaction costs and asset prices: A dynamic equilibrium model. Review of Financial Studies 11: 1–58. Chapter 12 Vayanos, Dimitri, and Jean-Lu Vila. 1999. Interest rate and liquidity premium with transaction costs. Economic Theory 13: 509–39. Chapter 12 Vayanos, Dmitri, and Tan Wang. 2002. Search and endogenous concentration of liquidity in asset markets. Working paper. Massachusetts Institute of Technology. Chapter 12 Vayanos, Dmitri, and Pierre-Olivier Weill. 2008. A search-based theory of the on-the-run phenomenon. Journal of Finance 63: 1361–98. Chapter 12 Velde, François, Warren Weber, and Randall Wright. 1999. A model of commodity money, with applications to Gresham’s law and the debasement puzzle. Review of Economic Dynamics 2: 291–323. Chapter 10 Wallace, Neil. 1980. The overlapping generations model of ﬁat money. In John Kareken and Neil Wallace, eds., Models of Monetary Economies. Federal Reserve Bank of Minneapolis, 49–82. Introduction, Chapters 4, 10 Wallace, Neil. 1992. Lucas’s signal extraction model: A ﬁnite-state exposition with aggregate real shocks. Journal of Monetary Economics 30: 433–47. Chapter 7 Wallace, Neil. 1996. Questions concerning rate-of-return dominance and indeterminacy in absence-of-double-coincidence models of money. Mimeo. Chapter 11 Wallace, Neil. 1997. Short-run and long-run effects of changes in money in a randommatching model. Journal of Political Economy 105: 1293–1307. Chapter 7

Bibliography

355

Wallace, Neil. 1998. Introduction to modeling money and studying monetary policy. Journal of Economic Theory 81: 223–31. Chapter 1 Wallace, Neil. 2000. Knowledge of individual histories and optimal payment arrangements. Federal Reserve Bank of Minneapolis Quarterly Review 24: 11–21. Chapters 1 2, 3, 11 Wallace, Neil. 2001. Whither monetary economics? International Economic Review 42: 847–69. Introduction, Chapter 6 Wallace, Neil. 2010. The mechanism-design approach to monetary theory. In Benjamin Friedman and Michael Woodford, eds., Handbook of Monetary Economics, 2nd ed. Amsterdam: North Holland. Introduction, Chapter 1 Wallace, Neil, and Ruilin Zhou. 1997. A model of a currency shortage. Journal of Monetary Economics 40: 555–72. Chapter 5 Waller, Christopher. 2009. Random matching and money in the neoclassical growth model: Some analytical results. Macroeconomic Dynamics, forthcoming. Chapter 6 Weill, Pierre-Olivier. 2007. Leaning against the wind. Review of Economic Studies 74: 1329–54. Chapter 12 Weill, Pierre-Olivier. 2008. Liquidity premia in dynamic bargaining markets. Journal of Economic Theory 140: 66–96. Chapter 12 Williamson, Stephen. 1998. Payment systems with random matching and private information. Journal of Money, Credit and Banking 30: 551–69. Chapters 8, 9. Williamson, Stephen. 1999. Private money, Journal of Money, Credit and Banking 31: 469–491. Chapter 8 Williamson, Stephen. 2002. Private money and counterfeiting. Federal Reserve Bank of Richmond, Economic Quarterly 88: 37–57. Chapter 5 Williamson, Stephen. 2006. Search, limited participation, and monetary policy. International Economic Review 47: 107–28. Chapter 7 Williamson, Stephen, and Randall Wright. 1994. Barter and monetary exchange under private information. American Economic Review 84: 104–23. Chapter 5 Williamson, Stephen, and Randall Wright. 2010a. New monetarist economics: Methods. Federal Reserve Bank of St. Louis Review 92: 265–302. Introduction Williamson, Stephen, and Randall Wright. 2010b. New monetarist economics: models. In Benjamin Friedman and Michael Woodford, eds., Handbook of Monetary Economics, 2nd ed. Amsterdam: North Holland. Introduction Woodford, Michael. 1990. The optimum quantity of money. Handbook of Monetary Economics 2: 1067–1152. Chapter 6 Woodford, Micheal. 2003. Interest and Prices: Foundations of a Theory of Monetary Policy. Princeton: Princeton University Press. Introduction Wright, Randall. 1994. A note on sunspot equilibria in search models of ﬁat money. Journal of Economic Theory 64: 234–41. Chapter 4 Wright, Randall. 1995. Search, evolution, and money. Journal of Economic Dynamics and Control 190: 181–206. Chapter 4

356

Bibliography

Yavas, Abdullah. 1994. Middlemen in bilateral search markets. Journal of Labor Economics 12: 406–29. Chapter 12 Zhou, Ruilin. 1997. Currency exchange in a random search model. Review of Economic Studies 64: 289–310. Chapter 10 Zhou, Ruilin. 1999. Individual and aggregate real balances in a random-matching model. International Economic Review 40: 1009–38. Chapter 4 Zhou, Ruilin. 2000. Understanding intraday credit in large-value payment systems. Federal Reserve Bank of Chicago, Economic Perspectives, 3rd quarter, 29–44. Chapter 9 Zhu, Tao. 2003. Existence of a monetary steady state in a matching model: indivisible money. Journal of Economic Theory 112: 307–24. Introduction Zhu, Tao. 2005. Existence of a monetary steady state in a matching model: Divisible Money. Journal of Economic Theory 123: 130–60. Introduction Zhu, Tao. 2008. An overlapping-generations model with search. Journal of Economic Theory 142: 318–31. Chapter 6 Zhu, Tao, and Neil Wallace. 2007. Pairwise trade and coexistence of money and higherreturn assets. Journal of Economic Theory 133: 524–35. Chapter 10

Index

Adverse selection, 305 Afonso, Gara, 337 Aggregate money demand asset prices and, 289, 294–95 credit and, 217 divisible money model and, 62, 67–69, 71 trading frictions and, 325, 329 Aiyagari, S. Rao, 44, 120, 227, 283–84, 337 Alchian, Armen, 120 Ales, Laurence, 283 Aliprantis, Charalambos D., 60, 92 Amihud, Yakov, 337–38 Andolfatto, David, 158 Anonymity, 92, 198 Araujo, Luis, 60, 158 Arrow–Debreu model, 3, 9, 11, 13, 21 Aruoba, S. Boragan, 92, 158–60, 188, 283 Ashcraft, Adam, 337 Asset prices, 12 aggregate money demand and, 289, 294–95 bargaining and, 286 bid–ask spreads and, 11, 317–18, 329–31, 335, 337 bilateral matches and, 286, 292, 295 collateral and, 285, 287 consumption and, 287 equilibrium and, 286, 289, 294–95 ﬁat money and, 291–95 fundamental value and, 288–89 inﬂation and, 285–86, 294–95 interest rates and, 203 liquidity and, 285, 305 (see also Liquidity) monetary policy and, 285–95 output and, 289 payments and, 290 trading frictions and, 289–91, 326–29

Assets’ yields, 300–305 Asymmetric information, 5 alternative information structures and, 184, 187 asset prices and, 286 bargaining under, 167–74 bilateral matches and, 167 equilibrium and, 174–81 Autarky concept of, 2–3 credit and, 31–38, 44–45 punishments and, 31, 33, 44, 52 role of money and, 47, 50–52 Axiomatic approach, 36, 45 Azariadis, Costas, 92 Bailey, Martin, 159 Banerjee, Abhijit, 3, 120 Banking central bank and, 133, 230, 246–48 credit and, 226–27 (see also Credit) interest on currency and, 133 record-keeping costs and, 226 settlement and, 230, 246–48 Bargaining, 8–9 alternative solutions to, 76–84 asymmetric information and, 167–74 Coles–Wright solution and, 92 Nash, 45–46 (see also Nash bargaining) Pareto frontier and, 26–27, 78–79, 82, 138, 140, 204, 209, 267–68 proportional solution and, 82–84 Benabou, Roland, 188 Bénassy, Jean-Pascal, 188 Berentsen, Aleksander, 60, 97, 119–20, 158–59, 227, 313 Bhattacharya, Joydeep, 158

358

Bid–ask spreads, 11, 317–18, 329–31, 335, 337 Bid price, 106, 108, 331 Biglaiser, Gary, 338 Bilateral matches, 9, 192 asset prices and, 286, 292, 295 asymmetric information and, 167 competing media of exchange and, 252–56, 259, 267, 271, 273, 280, 282–83 credit and, 25, 197–99, 202, 208, 215, 218 liquidity and, 296, 300–301, 305, 313, 317, 319 optimum quantity of money and, 130–31, 135, 143 role of money and, 48, 52–55 settlement and, 230–37, 242–43 trading frictions and, 317, 319 Bilateral trade, 6, 8, 11 competing media of exchange and, 282 monetary policy and, 162, 314 optimum quantity of money and, 143 Boel, Paola, 160, 284 Bonds, 10–12 competing media of exchange and, 251–52, 260, 270–74, 283–84 counterfeit, 252, 274–79, 284 credit and, 206, 227 illiquid, 131, 273–74 interest-bearing, 227, 251, 272, 274, 283–84, 313 monetary policy and, 286, 313 nominal, 131, 206, 260, 270–74 optimum quantity of money and and, 131 rate-of-return dominance puzzle and, 270–74 settlement and, 235, 245–47 Borrowing. See Credit Brunner, Karl, 120, 314 Bryant, John, 283–84 Burdett, Kenneth, 283 Calibration, 154, 159–60, 313 Calvo, Guillermo, 188 Camargo, Braz, 158, 188 Camera, Gabriele, 9, 60, 92, 120, 159–60, 226–27, 283–84, 338 Capital, 10, 12 cash-in-advance and, 265–70 competing media of exchange and, 251–62 credit and, 23

Index

dual currency payment systems and, 262–70 illiquid bonds and, 273–74 inﬂation and, 260–62 monetary policy and, 295, 313–14 money and, 252–62 nominal bonds and, 270–74 optimum quantity of money and, 160, 251–62, 283 properties of money and, 123 rate-of-return dominance and, 271–82 Tobin effect and, 252 trading frictions and, 317–33 Cash-in-advance, 265–70 Cass, David, 4 Cavalcanti, Ricardo, 121, 227 Central bank, 133, 230, 246–48 Chiu, Jonathan, 19 Chugh, Sanjay, 160 Clower, Robert, 4 Coins, 99–100, 120 Coles, Melvyn, 92 Coles–Wright bargaining solution, 92 Collateral, 10, 225, 248, 285, 287 Commitment, 2, 13 credit and, 21–27, 31–34, 195–96, 219–21 trust and, 21–22, 47 Commodities, 6–9, 103 competing media of exchange and, 251, 257, 262–63, 282–83 properties of money and, 99, 114–15, 120 Competitive search equilibrium, 93, 154. See also Price posting Complementarities (strategic), 318 credit and, 207–14 multiple equilibria and, 313 Concave storage technology, 257–60 Cone, Thomas, 313 Constantinides, George, 337 Corbae, Dean, 9, 45, 60, 120, 227 Cost of inﬂation alternative trading mechanisms and, 160 traditional measurement of, 159 welfare and, 154–58 Counterfeits banknotes and, 100 bonds and, 252, 274–79, 284 clipped coins and, 100 cost of producing, 100–101, 115–17 ﬁat money and, 100, 119 IOUs and, 231

Index

liquidity and, 305–306, 309, 313–14 monetary policy and, 101 properties of money and, 99–101, 115–21 recognizability and, 115–19 Craig, Ben, 160, 188, 283 Credit, 11–12 Arrow–Debreu model and, 21 asset prices and, 285, 287 bargaining and, 23–27, 36, 45, 200, 203–204, 208–11 bilateral matches and, 25, 197–99, 202, 208, 215, 218 bonds and, 206, 227 capital and, 23 collateral and, 225 commitment and, 21–27, 31–34, 195–96, 219–21 competing media of exchange and, 248 costless enforcement and, 195–96 debt and, 21–23, 26, 28, 31–35, 44–45, 195, 200–201, 215, 219–21, 224–26 default and, 22, 26–31, 33–34, 36, 45, 196, 225 delayed settlement and, 195 divisible money and, 196 dynamic models and, 227 ﬁat money and, 206, 225 gains from trade and, 21–25, 195, 202 incentive feasible allocations and, 24–25, 28–35, 42–44, 196, 203, 209 inﬂation and, 196, 201, 206–207, 212–13, 219, 225 interest rates and, 206–207, 215, 217–19 IOUs and, 10, 198–99, 202, 207–209, 215, 226, 231–47 liquidity and, 196, 206, 214–19, 227 long-term partnerships and, 219–25 Nash bargaining and, 25–26, 36, 45–46 output and, 25, 28, 33, 36, 195, 198–204, 208, 218 overlapping generations (OLG) model and, 5, 44 Pareto optimality and, 203–204, 209 partnerships and, 22, 38–45, 219–25, 227–28 payments and, 21, 28, 31, 38, 44–45, 195–96, 198, 201–15, 225–27 production and, 22–25, 28–35, 42–43, 197 pure credit economies and, 21–46 random matching and, 39–41, 44 real balances and, 198–212, 221–25, 228

359

reallocation of liquidity and, 214–19 record keeping and, 22, 31–38, 195, 198, 202, 207–14, 226 reputation and, 38–44 risk and, 26, 28, 31, 44, 214, 227 role of money and, 47–51 search-theoretic model and, 44–45, 225–26 settlement and, 230, 242–46, 248–49 short-term partnerships and, 219–25 specialization and, 45, 197 strategic complementarities and, 207–14 terms of trade and, 22, 28–29, 44, 202, 208, 223–24 trading frictions and, 35, 42–43, 195–97, 227 welfare and, 43–45, 49, 213–14, 218 Cuadras-Morato, Xavier, 120 Currency. See also Money counterfeit, 100–101, 120–21 (see also Counterfeits) credit and, 226 depreciation and, 71 domestic, 251–52 dual payment systems and, 262–70 elastic supply of, 230 emergence of uniform, 14 interest on, 133–35 international, 7, 283 monetary trades and, 195 portability and, 12, 99–100, 111–15 rate of return and, 66–67, 71, 74, 127, 184 redesign of, 119 settlement and, 248 shortage of, 102–107, 109, 122, 142 two-country models and, 7, 283 Curtis, Elisabeth S., 283 Cycles divisible money and, 92 output, 62 two-period, 74, 76 Dealers, 337 access issues and, 319 bargaining power and, 318 delays and, 332–36 intermediation fees and, 320, 329–31 networks and, 317 trading frictions and, 317–36

360

Debreu, Gerard, 3, 9, 11, 13, 21 Debt commitment and, 13 (see also Commitment) credit and, 21–23, 26, 28, 31–35, 44–45, 195, 200–201, 215, 219–21, 224–26 dynamic models and, 14 government, 10 negotiable, 229–49 optimum quantity of money and, 151 role of money and, 47 settlement and, 22–23, 229–49 Default asset prices and, 287 competing media exchange and, 271 credit and, 22, 26–31, 33–34, 36, 45, 196, 225 punishments and, 21–22, 31–38 strategic, 31 DeMarzo, Peter, 314 Deviatov, Alexei, 159 Diamond, Peter, 6–7, 10, 19, 44, 119, 188 Dichotomy, 195–202 Difference equation, 93–94 Distribution of money holdings, 9, 53–56, 63, 92–94, 150 Divisible money, 8–9, 91 aggregate money demand and, 62, 67–69, 71 credit and, 196 currency shortage and, 102–107 Friedman rule and, 129 indivisible money and, 107–109 large household model and, 94–96 lotteries for, 121 monetary policy and, 158 pairwise trade and, 280 scarcity and, 99 Dotsey, Michael, 226 Double coincidence of wants exchange and, 1–2, 6–7, 13–14, 19, 60, 120, 123, 337 lack of, 13 role of money and, 60 Dual currency payment systems capital and, 262–70 cash-in-advance and, 265–70 indeterminacy of the exchange rate and, 263–64 Dufﬁe, Darrell, 314, 336–37

Index

Enforcement, 22, 248, 285 credit issues and, 195–106, 204, 219–21, 226 optimum quantity of money and, 142, 144 Ennis, Huberto, 92, 159, 160 Equity premium, 10–11, 313 Erosa, Andres, 227 Essentiality of money, 8, 60 Exchange, 313 asset prices and, 285 autarky and, 2–3, 31–38, 44–47, 50–52 bilateral, 19, 296 commitment and, 13 (see also Commitment) competing media of, 251–83 credit and, 10–11, 21, 44–45, 195–99, 223, 225, 227 divisible money model and, 62–76 double coincidence of wants and, 1–2, 6–7, 13–14, 19, 60, 120, 123, 337 dual currency payment systems and, 262–70 ease of, 1–3 ﬁat money and, 251–64 (see also Fiat money) gains from trade and, 1 (see also Gains from trade) information and, 22 (see also Information) liquidity and, 296, 301, 309–11 (see also Liquidity) matching probability and, 33 (see also Matching probability) medium of, 6–7, 12–14, 251–83 nominal bonds and, 270–74 OLG model and, 283 (see also Overlapping generations (OLG) model) pairwise trade and, 280–82 Pareto optimality and, 252, 265–68, 280–82 payments and, 3–4 (see also Payments) role of money and, 47–52, 56, 60 trading frictions and, 317, 326 Exchange rate, 252, 263–65, 269–70, 283 Extensive margin, 104, 148–49, 159–60 Externalities, 86, 148–49, 159, 181, 207, 213 Faig, Miquel, 92, 159, 188 Fedwire, 229

Index

Fiat money asset prices and, 291–95 competing media of exchange and, 251–64, 270–74, 277–82 counterfeits and, 100, 119 credit and, 206, 225 dynamic models and, 14 equilibrium and, 61–62, 69, 79, 92 exchange and, 3–7, 10, 12, 14, 19 Friedman rule and, 262, 300 indivisible, 188 liquidity and, 298–310 monetary policy and, 162–66, 177–79, 184–85, 188, 285–86, 291–95, 298–310, 313–14 optimum quantity of money and, 127, 140, 146, 157 properties of, 99–100, 114–15, 119 record keeping and, 12 role of, 47, 60 search-theoretic model and, 188 settlement and, 229, 256 Fischer, Stanley, 4 Fischer effect, 219 Fluctuations, 62, 73, 184, 47 Freeman, Scott, 226, 247–48, 313 Frictionless markets Arrow–Debreu model and, 13 competing media of exchange and, 259 liquidity and, 285, 288–89, 313, 333 settlement and, 236–41 Friedman, Milton, 127, 158 Friedman rule alternative information structures and, 184, 187 bargaining and, 127–28, 132, 135–40, 144, 147–49, 158 competing media of exchange and, 262 extensive margin and, 148–49, 159–60 feasibility of, 142–44 ﬁat money and, 262, 300 ﬁrst-best allocation and, 12, 135–37, 140–44, 184, 300, 310 gains from trade and, 132 inﬂation and, 184 intensive margin and, 148–49 interest on currency and, 133–35 liquidity and, 300, 310 necessity of, 138–42 optimality of, 128–32

361

optimum quantity of money and, 127–50, 155, 158–60 Pareto optimality and, 138, 140 rate of time preference and, 162 stochastic money growth and, 165, 167 taxes and, 128, 142–44, 147, 149–50, 160 trading frictions and, 144–50 Walrasian price taking and, 137 welfare and, 149 Gains from trade, 1 costless enforcement and, 195–96 credit and, 21–25, 195, 202 dual currency payment systems and, 265 exchange and, 1 exploitation of intertemporal, 21 Friedman rule and, 132 liquidity and, 301 pure credit economies and, 21–25 role of money and, 47 trading delays and, 301 Garleanu, Nicolae, 336–37 Gavazza, Alessandro, 337 Geromichalos, Athanasios, 313 Gertler, Mark, 337 Glosten, Lawrence, 314, 338, 341 Gomis-Porqueras, Pere, 160 Grandmont, Jean-Michel, 92 Green, Edward, 9, 120, 248 Growth of money supply, 4, 8, 12 competing media of exchange and, 260, 269–70 interest rates and, 165 monetary policy and, 161–68, 179–87, 291, 295 optimum quantity of money and, 127, 129, 140, 144 stochastic, 162–67 Hall, George, 337 Harris, Larry, 317 Haslag, Joseph, 158 Head, Allen, 283 Heaton, John, 337 Hicks, John, 7, 251–52 Ho, Thomas, 338 Holdup problem, 84, 155, 210, 213 Holmström, Bengt, 313 Hopenhayn, Hugo, 314 Hosios condition, 148–49, 159 Hot potato effect, 147, 159

362

Hu, Tai-wei, 158 Huang, Ming, 337 Illiquidity, 11, 313 assets’ yields and, 300, 303–305 bonds and, 273–74 competing media of exchange and, 252, 260, 273, 278–79, 283–84 credit and, 206, 227 optimum quantity of money and, 131 trading frictions and, 318, 330, 336 Incentive feasible allocations, 5, 8, 12 credit and, 24–25, 28–35, 42–44, 196, 203, 209 optimum quantity of money and, 142–44 properties of money and, 109 role of money and, 47–51, 59, 61 Indeterminacy, 154 of equilibrium, 91–92 of exchange rate, 252, 256, 263–66, 283 Indivisible money competing media of exchange and, 273 ﬁat money and, 47, 188 Kiyotaki–Wright model and, 8 liquidity and, 314 lotteries and, 107–109 optimum quantity of money and, 142, 188 role of, 47–49, 52, 57 search-theoretic model and, 119–20 Shi–Trejos–Wright model and, 9, 92, 122–25 Wallace dictum and, 5 Inﬂation, 8, 12 asset prices and, 285–86, 294–95 capital and, 260–62 competing media of exchange and, 252, 260–62, 265, 270, 283 consumption and, 178 credit and, 196, 201, 206–207, 212–19, 225, 227 extensive margin and, 104, 148–49, 159–60 Friedman rule and, 184 intensive margin and, 105, 148–49 liquidity and, 177, 180, 298, 303–305, 313 monetary policy and, 285–86, 294–95, 298, 303–305, 313 optimum quantity of money and, 128, 131–32, 140, 146–60 output and, 161–62, 165–68, 171, 175–87

Index

properties of money and, 112, 115 settlement and, 247 short-run Phillips curve and, 177 signal extraction problem and, 178–79 stochastic, 305 superneutral money and, 140, 161 taxes and, 153, 159, 161, 183 Tobin effect and, 252 trade-off between output and, 161–62, 165–67, 177–88 welfare and, 154–60, 184, 188 Information, 2, 12 asset prices and, 286 asymmetric, 5, 162, 167–81, 184, 187, 286 credit and, 22, 28–29, 44 endogeneous recognizability and, 305–13 Friedman rule and, 184, 187 inﬂation–output trade-off and, 177–81 interest rates and, 184 liquidity and, 296, 299–300, 305–14, 338 optimum quantity of money and, 150–51, 160 private, 7, 28, 44, 120, 150–51, 160, 162, 186, 227, 248, 305, 314, 338 settlement and, 227, 242, 248 signal extraction problem and, 178–79 stochastic money growth and, 163, 166 trading frictions and, 338 Wallace dictum and, 5 Interest rates alternative information structures and, 184 asset prices and, 203 competing media of exchange and, 260, 270–74, 277–79 credit and, 206–207, 215, 217–19 inﬂation and, 161 liquidity and, 303 optimum quantity of money and, 127, 131–32, 155–58 stochastic money growth and, 165 Intermediation, 227, 283–84 fees for, 329–31 trading frictions and, 320, 323, 329–34 IOUs credit and, 10, 198–99, 202, 207–209, 215, 226, 231–47 default risk and, 242–46 monetary policy and, 246–47 settlement and, 231–47 Ireland, Peter, 226

Index

Jafarey, Saqib, 44 Jevons, William Stanley, 7, 99–100 Jin, Yi, 226 Jones, Robert A., 6, 19 Kahn, Charles, 227, 248 Kamiya, Kazuya, 92 Kareken, John, 283 Kehoe, Timothy, 45, 120 Kennan, John, 158 Kim, Yong, 120, 313, 337 King, Robert, 120, 314 Kiyotaki, Nobuhiro, 6–8, 19, 47, 60, 92, 119–20, 282–83, 313 Kiyotaki–Wright model, 6–7, 19, 60, 92, 119–20 Kocherlakota, Narayana, 8, 44–45, 59, 226–27, 283–84 Koeppl, Thorsten, 248 Koopmans, Tjalling, 4 Kranton, Rachel, 45 Krueger, Thomas, 283 Kultti, Klaus, 120 Kydland, Finn E., 4, 226 Kyle, Albert, 314, 338 Lagos, Ricardo, 9, 11, 91, 158–60, 248, 283, 313, 337 Lagos–Wright model, 9, 188, 313 Laing, Derek, 91 Large household model, 9, 92, 94–97, 120 Lee, Manjong, 313 Legal restrictions, 273, 283 Lending. See Credit Leontief matching function, 93 Lester, Benjamin, 313–14 Levine, David, 45, 159 Li, Victor, 91, 159 Li, Yiting, 120–21, 226–27, 283–84, 314, 338 Li, Zhe, 188 Licari, Juan, 313 Linear storage technology, 253–57 Liquidity, 12 assets’ yields and, 300–305 bid–ask spreads and, 11, 317–18, 329–31, 335, 337 bilateral matches and, 296, 300–301, 305, 313, 317, 319 competing media of exchange and, 252, 255, 259, 261, 273, 275–76, 279–82 constraints, 241, 245, 275–76, 280, 300–301

363

consumption and, 300–301, 317–18 counterfeits and, 306, 309, 313–14 delays and, 332–36 endogenous recognizability and, 305–13 exchange and, 296, 301, 309–11 ﬁat money and, 298–310 frictionless markets and, 285, 288–89, 313, 333 Friedman rule and, 300, 310 gains from trade and, 301 indivisible money and, 314 inﬂation and, 177, 180, 298, 303–305, 313 information and, 296, 299–300, 305–14, 338 interest rates and, 303 intermediation fees and, 320, 329–31 matching probability and, 232 output and, 295–96, 301, 305, 308–10 payments and, 295–296, 299, 303–306, 310–13, 317, 328 properties of money and, 113 real assets and, 295–312 real balances and, 295, 300–303, 306–15 reallocation of, 214–19 risk and, 295–300, 303, 305–306, 313–14, 337 search-theoretic model and, 314, 336–37 settlement and, 229–30, 236–41, 245–47 shocks and, 296, 317–20, 323–27, 331, 334 specialization and, 329 trading frictions and, 10–11, 285, 300, 304–305, 313, 317–38 welfare and, 311 Liquidity premium, 76, 180 Liu, Lucy Qian, 159 Liu, Qing, 283 Lo, Andrew, 337 Loan market, 214–19, 227 Lomeli, Hector, 92 Lotteries divisible money and, 121 indivisible money and, 107–109 properties of money and, 107–10, 119, 121 Lotz, Sebastien, 119, 283 Lucas, Robert E., 4–5, 9, 94, 159–61, 188, 226, 313, 337 Majluf, Nicholai, 314 Mamaysky, Harry, 337

364

Mankiw, N. Gregory, 188 Marimon, Ramon, 120 Martin, Antoine, 158 Maskin, Eric, 3, 120 Matching frictions, 42–43, 87, 232 Matching probability credit and, 33, 40 liquidity and, 232 optimum quantity of money and, 128, 145, 147 trading frictions and, 232 Matsui, Akihiko, 7, 283 Matsuyama, Kiminori, 7, 283 Mechanism design, 8, 44, 59, 121, 158, 248 Mehra, Rajnish, 313 Meltzer, Allan, 120, 314 Mendelson, Haim, 337–38 Miao, Jianjun, 337 Microfoundations of monetary exchange, 19 Milgrom, Paul, 314, 338 Mismatch, 227, 248, 324, 328, 334 Molico, Miguel, 19, 119–20, 159, 283 Monetary policy, 10–11 asset prices and, 285–95 belief and, 168–72 bilateral trade and, 162, 314 bonds and, 286, 313 capital and, 295, 313–14 commitment and, 189 counterfeits and, 101 distributional effects of, 150–53 divisible money and, 158 ﬁat money and, 162–66, 177–79, 184–85, 188, 285–86, 291–95, 298–310, 313–14 Friedman rule and, 12, 127–50, 155, 158–60, 162, 165, 167, 184, 187, 262, 300, 310 inﬂation and, 177–84, 285–86, 294–95, 298, 303–305, 313 IOUs and, 246–47 liquidity and, 295–313 (see also Liquidity) optimum quantity of money and, 127–160 overlapping generations (OLG) model and, 188 Pareto optimality and, 172, 305, 314 real balances and, 161–62, 165, 175–77, 184, 187 settlement and, 246–47 signal extraction problem and, 178–79

Index

Money autarky and, 47, 50–52 capital and, 252–62 cash-in-advance and, 265–70 coins and, 99–100, 120 counterfeit, 99–101, 115–21, 231, 252, 274–79, 284, 305–306, 313–14 credit and, 195–228 (see also Credit) currency shortage and, 102–107 dichotomy between credit and, 195–202 divisible, 99 (see also Divisible money) dual currency payment systems and, 262–70 in equilibrium, 61–97 (see also Equilibrium) as essential, 4–9 illiquid bonds and, 273–74 indeterminacy of the exchange rate and, 263–64 (see also Exchange) indivisible, 5–9, 19, 44, 47–49, 52, 57, 61, 92–93, 99, 101–103, 106–10, 119–23, 142, 188, 314 inﬂation and, 260–62 (see also Inﬂation) interest on currency, 133–35 Kiyotaki–Wright model and, 6–7, 19, 60, 92, 119–20 as memory, 48–52 nominal bonds and, 270–74 optimum quantity of, 127–60 overlapping generations model (OLG) and, 5, 44, 92, 151, 158–59, 188, 247, 283 pairwise trade and, 280–82 portability and, 12, 99–100, 111–15 properties of, 99–125 rate-of-return dominance and, 271–82 recognizability and, 7, 10, 12, 14, 100, 115–19, 231, 274–80, 282, 305–14 role of, 47–60 settlement and, 229–50 superneutral, 140, 161 turnpike model and, 5 two-country model and, 265–70 Wallace dictum and, 4–5 Monnet, Cyril, 121, 158, 248 Moore, John, 47, 313 Moral hazard, 45, 305, 313 Morishita, Noritsugu, 92 Mortensen, Dale, 93

Index

Multiple equilibria, 5, 196, 212–13, 310–12, 318, 335–36 Myers, Stewart, 314 Nash bargaining axiomatic approach and, 36, 45 credit and, 25–26, 36, 45–46 generalized solution for, 45–46 optimum quantity of money and, 128, 135–38, 149, 155, 158 Pareto optimality and, 79 trading frictions and, 321 Niehans, Jurg, 4 Nominal bonds, 131, 206, 260, 270–74 Nominal rigidities, 179, 188 Nonmonetary equilibria, 257–59 Nonstationary equilibrium, 61, 71–74, 114, 125, 256 Nosal, Ed, 120–21, 159, 313–14 Obstfeld, Maurice, 283 Oh, Seonghwan, 19 Optimum quantity of money bilateral matches and, 130–31, 135, 143 distributional effects of monetary policy and, 150–53 extensive margin and, 148–49, 159–60 feasibility and, 142–44 Friedman rule and, 127–50, 155, 158–60 intensive margin and, 148–49 interest on currency and, 133–35 interest rates and, 127, 131–32, 155–58 output and, 131–32, 138, 154 payments and, 133, 158 real balances and, 127–32, 135–47, 150–57 taxes and, 142–44, 147, 149–50, 153, 159–60 trading frictions and, 144–50, 159 welfare cost of inﬂation and, 154–60 Osborne, Martin, 8, 45, 92 Ostroy, Joseph, 5–6, 59 Output, 7 asset prices and, 289 competing media of exchange and, 255–56, 259–62, 265–77, 280–81 credit and, 25, 28, 33, 36, 195, 198–204, 208, 218 inﬂation and, 161–62, 165–68, 171, 175–87 information and, 12 liquidity and, 295–96, 301, 305, 308–10 monetary shocks and, 188

365

neutral money and, 161 optimum quantity of money and, 131–32, 138, 154 properties of money and, 101, 104–10, 113–14, 120, 123–24 role of money and, 49–50 settlement and, 235, 238, 241–42 short-run Phillips curve and, 177 signal extraction problem and, 178–79 Overlapping generations (OLG) model competing media of exchange and, 283 credit and, 5, 44 monetary policy and, 188 optimum quantity of money and, 151, 158–59 settlement and, 247 Over-the-counter markets, 11–12, 286, 317, 336–37 Pairwise trade, 280–82 Pareto optimality Arrow–Debreu model and, 3, 13 bargaining frontier and, 26–27, 78–79, 82, 138, 140, 204, 209, 267–68 competing media of exchange and, 252, 265–68, 280–82 credit and, 203–204, 209 Friedman rule and, 138, 140 monetary policy and, 172, 305, 314 Nash solution and, 79 optimum quantity of money and, 138, 140 overlapping generations (OLG) model and, 5 pure credit economies and, 26, 45 settlement and, 248 Partnerships credit and, 22, 38–45, 219–25, 227–28 long-term, 38, 43, 45, 219–25, 228 relocation shock and, 38, 41 short-term, 219–25 Patinkin, Don, 4 Payments, 3–4, 7, 10–11, 14 asset prices and, 290 cash-in-advance and, 265–70 competing media of exchange and, 251–52, 256, 259–74, 277–83 credit and, 21, 28, 31, 38, 44–45, 195–96, 198, 201–15, 225–27 dual currency systems and, 262–70 Fedwire and, 229

366

Payments (cont.) indeterminacy of the exchange rate and, 263–64 inﬂation–output trade-off and, 179 liquidity and, 295–96, 299, 303–306, 310–13, 317, 328 optimum quantity of money and, 133, 158 properties of money and, 100, 102, 113–14, 120 role of money and, 61 settlement and, 229–50 Pedersen, Lasse H., 336–37 Peralta-Ava, Adrian, 160 Peterson, Brian, 120 Phillips curve, 177 Physical properties (of money), 10, 99, 274, 278, 280, 282 Pissarides, Christopher, 45, 120 Plosser, Charles, 120, 314 Pooling equilibrium, 169–70 Portability of money, 12, 99–100, 111–15 Postlewaite, Andrew, 313–14 Prescott, Edward, 4, 226, 313 Price posting, 9 competitive, 86–91, 92 equilibrium and, 77, 86–91, 92–93 optimum quantity of money and, 127, 137–38, 154, 160 Price taking equilibrium and, 84–86, 89, 92 optimum quantity of money and, 127, 137–38, 154, 160 role of money and, 52 Walrasian, 84–86, 89 Production, 4, 6 asymmetric information and, 169 competing media exchange and, 269 credit and, 22–25, 28–35, 42–43, 197 liquidity and, 298, 328–29 optimum quantity of money and, 150, 154 output and, 16 (see also Output) properties of money and, 105, 118–19 role of money and, 53, 55, 59 settlement and, 230, 232, 238, 246 shocks and, 27, 85, 128, 151, 317, 319–20, 323–27, 331, 334 trading frictions and, 328–29 Proportional solution, 82–84 Punishments

Index

autarky and, 31, 33, 44, 52 default and, 21–22, 31–38 record keeping and, 21–22, 31–38 Pure credit economies Arrow–Debreu model and, 21 commitment and, 21–27, 31–34 default and, 26–31 exchange and, 21, 44–45 gains from trade and, 21–25 Pareto optimality and, 26, 45 Record keeping and, 31–38 reputation and, 38–44 Puzzello, Daniela, 60, 92 Quasi-linear preferences, 9, 11, 54, 94, 97, 129, 150, 154, 295 Quericioli, Elena, 121 Quid quo pro, 195–96 Random matching, 9 counterfeits and, 120 credit and, 39, 39–41, 44 equilibrium and, 92 monetary policy and, 188 properties of money and, 120 settlement and, 220, 226, 248 Rate-of-return dominance puzzle, 7, 10, 12, 305 bonds and, 270–74 competing media of exchange and, 251, 259, 270–82 illiquid bonds and, 273–74 pairwise trade and, 280–82 recognizability and, 274–80 Rauch, Bernhard, 97, 158 Real assets exchange and, 3, 114, 255, 285–312 liquidity and, 295–312 monetary policy and, 285–312 properties of money and, 114 Real balances asset prices and, 291–93 competing media of exchange and, 255–56, 259–64, 270, 280 credit and, 198–212, 221–25, 228 liquidity and, 295, 300–303, 306–15 monetary policy and, 161–62, 165, 175–77, 184, 187 optimum quantity of money and, 127–32, 135–47, 150–57

Index

properties of money and, 100, 111–13, 117 role of money and, 53, 56–57 Recognizability, 7, 10, 12, 14 cognizability and, 100 endogenous, 305–13 equilibrium and, 306–13 Jevons on, 99–100 liquidity and, 305–14 properties of money and, 100, 115–19 rate-of-return dominance and, 274–80, 282 settlement and, 231 unrecognizable assets and, 309–10 Record keeping costless enforcement and, 195–96 costs of, 201–207 credit and, 22, 31, 31–38, 195, 198, 202, 207–14, 226 ﬁat money and, 12 private memory and, 38–44 punishments and, 21–22, 31–38 pure credit economies and, 31–38 role of money and, 47–52, 59 transparency and, 22, 31–32 Redish, Angela, 120 Reed, Robert, 60, 159–60 Reis, Ricardo, 188 Renero, Juan M., 120 Reputation, 38–44 Reserves, 133, 158, 227 Risk assets’ yields and, 300, 303, 305–306 competing media of exchange and, 282, 286 credit and, 26, 28, 31, 44, 214, 227 default and, 22, 26–31, 33–34, 36, 45, 196, 225, 271, 287 dividend payments and, 286, 296, 299 idiosyncratic, 242, 246, 248 liquidity and, 295–300, 303, 305–306, 313–14, 337 random matching process and, 9 risk-free assets and, 11, 282, 286, 295, 299, 305, 313 settlement and, 230, 241–42, 245–48, 249–50 sharing, 44, 227 trading frictions and, 337 Risk-free rate puzzle, 11, 313 Ritter, Joseph, 45, 227

367

Roberds, William, 248 Rocheteau, Guillaume, 9, 19, 60, 92, 97, 120–21, 158–60, 188, 283–84, 313–14, 337 Rogoff, Kenneth, 283 Rotemberg, Julio, 188 Round-the-clock payment arrangement, 227 Rubinstein, Ariel, 8, 45, 92, 337 Rupert, Peter, 19, 44, 119–20 Rust, John, 337 Samuelson, Paul, 5 Sanches, Daniel, 158, 188 Sargent, Thomas, 158, 283 Sato, Takashi, 92 Schevshenko, Andrei, 119, 337–38 Schindler, Martin, 119–20 Schorfheide, Frank, 188 Schreft, Stacey, 226 Screening, 185 Search frictions, 18 competing media of exchange and, 274 credit and, 25, 35, 54–55, 197 liquidity and, 325–27, 337 optimum quantity of money and, 144 properties of money and, 116 role of money and, 59 Search-theoretic model credit and, 44–45, 225–26 double-coincidence-of-wants problem and, 60 ﬁat money and, 188, 225 indivisible money and, 119–20 large household model and, 94 liquidity and, 314, 336–37 properties of money and, 119–20 search-theoretic model and, 188, 225 two-currency, 283 Securitization, 10 Self-fulﬁlling, 7, 10, 62, 114, 313, 318 Separating equilibrium, 172, 181 Settlement allocations and, 230, 241, 246, 248 bargaining and, 237–38, 242 bilateral matches and, 230–37, 242–43 bonds and, 235, 245–47 choice of money holdings and, 242 commitment and, 235, 242 consumption and, 230–38, 241 costs and, 229 credit and, 195

368

Settlement (cont.) debt and, 22–23, 229–49 default and, 230, 242–46, 248–50 exchange and, 229–232, 239, 246–248 Fedwire and, 229 ﬁat money and, 229, 256 frictionless, 233–41, 248–49 gross, 248 inﬂation and, 247 information and, 227, 242, 248 IOUs and, 231–47 liquidity and, 229–30, 236–41, 245–47 monetary policy and, 246–247 net, 248 output and, 235, 238, 241–42 overlapping generations (OLG) model and, 247 Pareto optimality and, 248 production and, 230, 232, 238, 246 risk and, 230, 241–42, 245–48 shocks and, 248 trading frictions and, 229–33, 236–43, 246–48 welfare and, 230 Shi, Shouyong, 8–10, 19, 45, 60, 92, 94–97, 119–25, 158–59, 225–26, 283–84 Shimizu, Takashi, 92 Shi–Trejos–Wright model, 8–9, 92, 122–25 Signal extraction problem, 178–79 Signaling game, 168 Smith, Bruce, 44, 158 Smith, Lones, 121 Specialization, 6 competing media of exchange and, 283 credit and, 45, 197 liquidity and, 329 optimum quantity of money and, 159 role of money and, 60 trading frictions and, 329 Spulber, Daniel, 337 Starr, Todd, 5–6, 59 Stationary allocations, 24, 28, 32, 44 Sticky prices, 188 Stockman, Alan, 283 Stokey, Nancy, 226 Stoll, Hans R., 338 Suarez-Lledo, Jose, 313 Sunspot equilibrium, 74–76, 92 Superneutral money, 140, 161 Surplus from trade, 26, 33, 79, 128, 135, 143, 333

Index

Taber, Alexander, 120 Taxes Friedman rule and, 128, 142–44, 147, 149–50, 160 inﬂation and, 153, 159, 161, 183 lump-sum, 129–30, 133–34, 142–44, 150, 221, 260, 271, 291 optimum quantity of money and, 128–29, 134, 142–44, 147, 149–50, 153, 159–60 Taylor, John, 188 Technology concave storage, 257–60 counterfeiting and, 116, 306 linear storage, 253–57 record keeping and, 38–44 (see also Record keeping) Telyukova, Irina, 226 Temzelides, Ted, 60, 92, 226–27, 248 Terms of trade asset prices and, 291 asymmetric information and, 174 bilateral matches and, 9 competing media of exchange and, 254, 256, 263–74, 280–82 credit and, 22, 202, 208, 223–24 inﬂation and, 181 liquidity and, 296, 300, 305, 321 optimum quantity of money and, 131, 135–38, 144, 158 properties of money and, 116–17 role of money and, 48, 52–57 settlement and, 235 stochastic money growth and, 163–66 trading frictions and, 321 Tirole, Jean, 92, 313 Tobin, James, 283 Tobin effect, 252 Townsend, Robert, 5, 59, 92, 226 Trading frictions, 3, 12 allocations and, 330, 334 asset prices and, 289–91, 326–29 bargaining and, 318, 321–23, 329–34 bid–ask spreads and, 11, 317–18, 329–31, 335, 337 bilateral matches and, 317, 319 capital and, 317–33 competing media of exchange and, 251–52, 259, 269, 274 consumption and, 317–18 continuous time and, 318 credit and, 22, 25, 35, 42–44, 195–97, 227

Index

delays and, 332–36 double coincidence of wants and, 337 exchange and, 317, 326 explicit protocols and, 52, 61, 84–85, 92 Friedman rule and, 144–50 gains from trade and, 301 information and, 338 intermediation fees and, 320, 329–31 liquidity and, 10–11, 285, 300, 304–305, 313 market clearing and, 325–29, 333 matching, 42, 87, 220, 226, 232, 248 Nash bargaining and, 321 optimum quantity of money and, 127, 144–50, 159 pairwise meeting, 15 properties of money and, 99, 116 risk and, 337 role of money and, 59 search, 18, 25, 35, 43–44, 59, 61, 84, 87, 116, 144, 197, 274, 325–27, 337 settlement and, 229–33, 236–43, 246–48 shocks and, 317–20, 323–27, 331, 334 specialization and, 329 terms of trade and, 321 Walrasian price and, 326, 329 Transaction costs and, 4, 11, 314, 329–30, 337 Trejos, Alberto, 8–9, 92, 119–25, 283 Trust, 21–22, 47 Turnpike model, 5 Two-country model, 265–70 U.S. Federal Reserve, 229 Vayanos, Dimitri, 337 Velde, François, 283 Velocity, 101, 159, 162, 174, 177, 187 Vila, Jean-Lu, 337 Wallace, Neil, 4–5, 8, 19, 44–45, 59–60, 92, 119–20, 148–59, 188, 226, 283–84, 313–14 Waller, Christopher, 60, 92, 119, 158–60, 227, 283 Walrasian price, 9 equilibrium and, 77, 84–86, 89 optimum quantity of money and, 137–38, 154 trading frictions and, 326, 329 Wang, Jiang, 337 Wang, Liang, 159 Wang, Ping, 91

369

Wang, Tan, 337 Wang, Weimin, 159 Wealth effect (lack of), 9, 16–17, 54, 94, 129, 150 Weber, Warren E., 120, 283 Weill, Pierre-Olivier, 337 Welfare, 4 asymmetric information and, 167 changes in money supply and, 161–63, 167, 179, 184, 188 competing media of exchange and, 256, 262, 264, 280, 283–84 credit and, 43–45, 49, 213–14, 218 extensive margin and, 104 Friedman rule and, 149 inﬂation and, 154–60, 181–83, 188 liquidity and, 311 optimum quantity of money and, 127, 137, 142, 147–50, 153–60 properties of money and, 100–101, 104, 115 role of money and, 51 settlement and, 230 stochastic money growth and, 162–63 Werner, Ingrid, 314 Williamson, Stephen, 7, 9, 19, 44, 120, 158, 188, 227, 248 Williamson–Wright model, 120 Winkler, Johannes, 283 Wolinsky, Asher, 337 Woodford, Michael, 127, 158 Wright, Randall, 6–11, 19, 60, 91–93, 119–25, 158–60, 188, 226, 248, 282–84, 313–14 Yavas, Abdullah, 337 Zhang, Yahong, 283 Zhou, Ruilin, 9, 91, 119, 248, 283 Zhu, Tao, 9, 97, 159, 283, 314 Zilibotti, Fabrizio, 120